Xz PROCEEDINGS CONFERENCE ON J 1 I [1 APPROACHES TO POPULATION STUDIES TITRE OCTOBER 1-2, 1964 NATIONAL INSTITUTES OF HEALTH BETHESDA, MARYLAND hy be CAT. FOR PUBLIC HEALTH p= “NTS DEPARTMENT oct 21 1886 {BRARY : UNIVER OF CA FORKIA (15 Public. Beal Some. PIS. put 1486 yssn PROCEEDINGS CONFERENCE ON METHODOLOGICAL APPROACHES TO POPULATION STUDIES IN DIABETES October 1-2, 1964 National Institutes of Health Bethesda, Maryland Editor: James W. Pratt, Ph.D. Sponsored By National Institute of Arthritis and Metabolic Diseases National Institutes of Health and Diabetes and Arthritis Program Division of Chronic Diseases Bureau of State Services, Community Health U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C., 20402 - Price $1.25 Public Health Service Publication No. 1486 PC460 AICé CONFERENCE ON METHODOLOGICAL APPROACHES TO RN HEALTH POPULATION STUDIES IN DIABETES by OBJECTIVES To discuss and evaluate specific methodological approaches to population studies in diabetes. To concentrate principally on those methodological problems, techniques, procedures, and approaches which are unique to studies in diabetes. To discuss basic techniques required to assure comparability of data. To discuss aspects of the disease and of objectives in population studies which may require further development and refinement of methods. CONFERENCE PARTICIPANTS Dr. Roy M,. Acheson Department ‘of Epidemiology and Public Health Yale University School of Medicine New Haven, Connecticut 06520 Dr. Ruben Andres Baltimore City Hospital Baltimore, Maryland 21224 Dr. Thomas A. Burch National Institute of Arthritis and Metabolic Diseases National Institutes of Health Bethesda, Maryland 20014 Dr. Benjamin T., Burton National Institute of Arthritis and Metabolic Diseases National Institutes of Health Bethesda, Maryland 20014 Dr. Joseph H. Crampton The Virginia Mason Foundation Seattle, Washington 98101 Dr. Stefan S. Fajans University Hospital Ann Arbor, Michigan 48104 Dr. Robert Feldman Department of Medicine The Permanente Medical Group Oakland, California Mrs. Gail F. Fisher Division of Chronic Diseases Bureau of State Services, USPHS Washington, D. C. Dr. Norman S. Hayner School of Public Health University of Michigan Ann Arbor, Michigan 48104 iv Dr. John M. Kalbfleisch University of Oklahoma Medical Center Oklahoma City, Oklahoma 73104 Dr. Gerald T. Kent School of Medicine Western Reserve University Cleveland, Ohio 44106 Dr. Christian R. Klimt Division of Epidemiology and Biostatistics School of Medicine University of Maryland Baltimore, Maryland Dr. Keatha K. Krueger National Institute of Arthritis and Metabolic Diseases National Institutes of Health Bethesda, Maryland 20014 Dr. Arnold Lazarow Chairman, Department of Anatomy University of Minnesota Minneapolis, Minnesota Dr. Jack R. Leonards School of Medicine Western Reserve University Cleveland, Ohio 44106 Dr. Francis D. W. Lukens Hospital of University of Pennsylvania Cox Institute Philadelphia, Pennsylvania 19104 Dr. Glen W. McDonald Division of Chronic Diseases Bureau of State Services, USPHS Washington, D. C. r. Max Miller School of Medicine Western Reserve University Cleveland, Ohio 44106 Dr. Benno K. Milmore State Department of Public Health 2151 Berkeley Way Berkeley, California 94704 Dr. George J. Mouratoff Foundation for Clinical Pharmacology, Inc. Berkeley, California Dr. Vaun A. Newill School of Medicine Western Reserve University Cleveland, Ohio 44106 Dr. James W. Pratt Division of Chronic Diseases Bureau of State Services, USPHS Washington, D. C. Dr. John O'Sullivan Chief, Research Field Section Diabetes and Arthritis Program, USPHS 77 Warren Street Brighton 35, Massachusetts Dr. George G. Reader Department of Medicine The New York Hospital Cornell Medical Center New York, New York 10021 Mr. Quentin R. Remein Division of Chronic Diseases Bureau of State Services, USPHS Washington, D. C. Dr. Henry T. Ricketts Department of Medicine University of Chicago Chicago, Illinois 60637 Dr. Thomas N. Roberts Clinical Asst. Professor of Medicine Cornell Medical Center New York, New York 10021 Dr. Howard F. Root President, Diabetes Foundation, Inc. Boston, Massachusetts Dr. Thomas Palmer Sharkey 60 Wyoming Street Dayton, Ohio 45409 Dr. Felix Silverstone Maimonides Hospital Brooklyn, New York 11219 Dr. Elena Villavicencio-Suarez School of Medicine University of Puerto Rico San Juan, Puerto Rico 00905 Dr. Robert L. Vought National Institute of Arthritis and Metabolic Diseases National Institutes of Health Bethesda, Maryland 20014 Dr. Masahisa Wada First Department of Internal Medicine Osaka University Medical School 33 Joan-cho, Kita-ku Osaka, Japan Dr. Kelly M. West Department of Medicine University of Oklahoma Medical Center Oklahoma City, Oklahoma 73104 Dr. G. Donald Whedon Director, National Institute of Arthritis and Metabolic Diseases National Institutes of Health Bethesda, Maryland 20014 Mr. Theodore D. Woolsey Deputy Director, National Center for Health Statistics U.S. Public Health Service Washington, D. C. Dr. Louis K. Alpert School of Medicine George Washington University Washington, D. C. Mr. Tavia Gordon Division of Health Examination Statistics, USPHS Washington, D. C. Dr. Curtis L. Meinert School of Medicine University of Maryland Baltimore, Maryland Dr. Charles A. Rosenberg Chief of Research in Metabolism Endocrinology Veterans Administration Washington, D. C. vi II. ITI. IV. CONTENTS THE CONFIRMATION OF DIABETES: REQUIRED CRITERIA FOR DIAGNOSIS#* Stefan S. Fajans, M.D. Discussion STUDIES OF DIABETES WITHIN COMMUNITIES#* Gerald T. Kent, M.D. Discussion PROBLEMS IN OBTAINING OBJECTIVES AND COMPARABLE DATA FROM DIFFERENT INVESTIGATORS AND FACILITIES Christian R. Klimt, M.D. Discussion CONCLUSION A. SYNOPSIS AND DISCUSSION Max Miller, M.D. B. DISCUSSION AND RECOMMENDATIONS Vaun A. Newill, M.D. APPENDIX TABLES AND FIGURES *All tables and figures are in the Appendix. vii PAGE 21 50 57 87 102 154 162 219 3 = hohe un ET = Ye ia g i # . = dma i THE CONFIRMATION OF DIABETES: REQUIRED CRITERIA FOR DIAGNOSIS STAFAN S. FAJANS, M.D.! The subject assigned to me this afternoon is the confirmation of diabetes, or in other words, a discussion of the diagnostic tests and diagnostic criteria employed for the early recognition of asymptomatic diabetes mellitus. Before entering on a discussion of tests, techniques, and criteria, I would like to spend a few minutes in reviewing current concepts of the natural history of diabetes. It is only against such a background that we can appreciate some of the difficulties which we have had and will continue to have with diagnostic tests and standards in use today. There is general agreement that diabetes mellitus is a genetically transmitted disease although there is no similar agreement concerning the nature of this hereditary component. Thus, in recent years, it has been variously postulated that the primary defect affects (1) some aspects of insulin polypeptide synthesis with elaboration of an imperfect insulin; or (2) a primary excess of an insulin antagonist, as postulated by Valance-Owen; or (3) an initial overproduction of insulin followed by excessive antagonist activity, as postulated by Neel, a geneticist at the University of Michigan; or (4) an abnormality 1 The University of Michigan Medical School, Ann Arbor, Michigan -d a of triglyceride metabolism with a higher rate of release of fatty acids and ketone bodies for oxidation with as one consequence an impaired sensitivity to insulin -- you all recognize this as being Randle's hypothesis -- or (5) an enzyme defect common to peripheral tissues, liver and islet cells. When this hypothetical enzyme defect involves primarily peripheral tissues and liver it may lead to decreased insulin sensitivity followed by compensatory increase in insulin release. When and if the enzyme defect involves the islet cells, it may lead to insulin insufficiency; and as a final hypothesis (6) that the primary defect results from an abnormality in the basement membrane of the small blood vessels. Whatever the nature of this defect may be, the presence of a hereditary component suggests its existence from conception and birth. The concomitant metabolic abnormality may remain undetectable by present methods or undetected for many years because of the presence of compen- satory factors. Eventually, however, the primary defect may lead to insufficient insulin activity and a change from apparently normal to abnormal metabolism of carbohydrate and lipids. The duration of time from conception until this occurs, the so-called prediabetic period, is extremely variable as is the severity of insulin insufficiency which develops. Both these factors may depend on the severity of the genetic etiological defect as well as on superimposed exogenous factors such as pregnancy, obesity, infection, nutrition, et cetera. “3 When viewed in this way, it is evident that the earliest recogni- zable loss of carbohydrate tolerance would represent not the earliest stage in the evolution of diabetes mellitus but a more advanced stage in the natural history of the disease. Before proceeding, we need to establish some definitions for purposes of identification. Table I presents a scheme depicting the natural history of diabetes as divided into four stages. Frank, or overt, diabetes is the most advanced of these stages. Classical symptoms may be present; there is fasting hyperglycemia; the glucose tolerance is not necessary for diagnosis. In latent diabetes there are no symptoms referable to the disease. However, a definite diagnosis of diabetes can be established by presently accepted laboratory procedures such as by an elevated fasting blood sugar level or by an abnormal glucose tolerance test, whatever our definition of abnormality shall be. An earlibe stage is subclinical diabetes. Here not only the fasting blood sugar level but also the glucose tolerance test is normal under usual circumstances. However, diabetes may be suspected because of evidence of insufficient functional reserve of the islet cells. An example would be a woman with a normal glucose tolerance test in the non-pregnant state but with a history of transient abnormality of standard glucose tolerance during pregnancy. This has been termed gestational or . pregnancy diabetes. A high proportion of such women develop latent or overt diabetes in the years that follow. Another example of subclinical diabetes may be a non-pregnant individual with a normal standard glucose tolerance but an abnormal cortisone-glucose tolerance test. “4 = The earliest stage is prediabetes. Presently the prediabetic period can be defined as the time from conception to the earliest recognizable abnormality of carbohydrate tolerance diagnostic of diabetes. Until a definite criterion has been found which accurately identifies those non- diabetic individuals who will become diabetic in the future, the term ""prediabetes' cannot be applied to an individual. Prediabetes can, however, be suspected, as in (1) nondiabetic individuals strongly pre- disposed to diabetes on genetic grounds, such as an identical twin of a diabetic or an offspring of a diabetic father and a diabetic mother, or (2) a mother who gives birth to a large baby. During the prediabetic period glucose tolerance and cortisone-glucose tolerance tests are normal. A number of findings indicate that this prediabetic period is a dynamically active one. Steinke, Renold, and collaborators have found elevated levels of serum insulin-like activity as determined by the epididymal fat pad technique, in some nondiabetic children of two diabetic parents. Elevated levels of ILA were also found in latent and overtly diabetic individuals. Valance-Owen has reported finding increased levels of synalbumin insulin antagonist in nondiabetic relatives of diabetics and in nondiabetic subjects with myocardial infraction suggesting that these are 'prediabetic individuals'. He has also found this antagonist in subjects with latent and overt diabetes. Finally, vascular changes suggestive of the diabetic state have been found in genetically predia- betic individuals. Undoubtedly these definitions of the stages of diabetes will have to be revised when various chemical or morphologic -5 - diagnostic procedures can identify the prediabetic state with a high degree of accuracy. Also note that progression or regression from the earliest stage or stages to the next may never occur, may progress very slowly over many years, or may be rapid or even explosive. Elucidation of the factors which exist in the prediabetic period and which lead to eventual development of clinical diabetes constitutes a major goal of current and future research. It is entirely possible that abnormalities existing during the prediabetic period may be sus- ceptible to exogenous influences which may retard the evolution or even prevent the clinical manifestations of the later stages of the disease. The same may be true when prophylactic procedures are applied to early latent diabetes -- and I think this is the reason for the current interest in the early recognition of diabetes. This is one of the reasons, I believe, for this conference being held today. Let us now turn to a discussion of minimal diagnostic criteria for the early detection of latent diabetes. The standard oral glucose tolerance test remains the most sensitive current test for the detection of latent asymptomatic diabetes mellitus. The following procedures are employed by us for the performance of the test. All subjects are instructed in a diet containing 300 grams of carbohydrate to be ingested for three days before performance of the test. I think that 250 grams of carbohydrate would also be adequate. The test meal consists of 1.75 grams of glucose per kilogram of ideal body weight given as a 25% solution flavored with lemon extract. Ideal fw body weight is calculated from the subject's height and age. All ages beyond 25 are equated at 25 since, according to Life Insurance Tables, ideal body weight stays constant for a given height after this age. Venous blood is obtained in the sitting position and is analyzed by the Somogyi-Nelson technique or by the Technicon Auto-Analyzer standardized to give the same readings as the former method. Samples are obtained in the fasting state and every half-hour for three hours. I think the critical levels are those obtained during the first two hours. The timing starts at the beginning of glucose administration, since absorption starts then and continues for several hours. Some of the data which I will show you are data taken from a prospective long-term study which was initiated by us 14 years ago, with the hope of learning something about the natural history of diabetes. Obviously, I cannot show you the results of many repeat tests -- this is out of the context of this conference today. I should also like to mention that Dr. Neel, our geneticist at the University of Michigan was in on the study from the early 1950's on, and helped us plan the type of individual to be employed for the testing. Figure 1 shows the mean curve for 111 control subjects of both sexes, 20 to 39 years of age, whose family histories were negative for diabetes mellitus. It should be emphasized that these were not the usual cursory medical family histories but that detailed pedigrees were obtained on each individual. All of the mean values for the males were slightly higher than those for females, the difference at two hours being the only significant one, namely 5.8 mg. per 100 ml. Since the sex difference . Jim was small in relation to the standard deviation, we have felt justified in deriving one composite curve. Analysis of the data showed that there was no clear age trend in this age interval, the curves for age 30 to 39 being the same as those for age 20 to 29. Also indicated are two standard deviations above and below the mean values. In an otherwise healthy and ambulatory individual we have previously defined a diabetic glucose tolerance curve as one exhibiting a one-hour value of 160 mg. or more, a 1% hour value of 140 mg. or more, and a two-hour value of 120 mg. per 100 ml. or more. Although it is obvious that any criteria which provide a dividing line between normal and abnormal must be somewhat arbitrary, it will be seen that these values fall two standard deviations above the control curve. This definition, therefore, is suitably con- servative for this age group. A small group of control subjects age 40 to 49 had curves similar to those of the younger groups. It should also be noted that some of the comtrol subjects may later develop diabetes despite the negative family histories. Indeed, in a group of 12 control subjects who have been followed for 10 years or more, two now have a definite family history of diabetes. Fasting hyperglycemia has been detected in the parent of one and in the grandparent of the other of these two ''control" subjects. I should also mention that another subject who is not included among these 111 subjects and who was a 'control” subject in 1950 now has a diabetic father and a diabetic mother, both having fasting hyperglycemia. Thus a "negative family history' may be of limited value in the selection of control subjects. Individuals whose glucose tolerance test curves exhibited a one-hour value of 160 mg. or more, a 1% hour value of 135 mg. or more, and a - 8 - two-hour value between 110 and 120 mg. per 100 ml. were placed into a borderline group designated as ''probable diabetes'' or ''diabetes suspect". Justification for setting apart the group of "probable diabetes’ from the "normal group" will be given later. Since it has been reported repeatedly that tolerance to glucose diminishes with advancing age -- Dr. Silverstone, who is here, reported such a study in 1957; Dr, West reported one in 1964 -- the question of what constitutes diabetes in older people has become a very significant one. Unfortunately no good data exist regarding reliable minimal diagnostic criteria for individuals over the age of 50. This problem is fraught with complexities. Figure 2 presents two-hour postglucose blood sugar levels which were obtained on University of Michigan faculty members with a negative family history of diabetes. There is a clear trend toward higher values with age in both the means and standard deviations, particularly in subjects over the age of 50. These over-all figures undoubtedly include some values from individual subjects with mild diabetes mellitus, since their two-hour blood sugar levels were well above the means plus two standard deviations for their age group. Subjects with fasting hyper- glycemia over 120 mg. per 100 ml. have been excluded from this group. Nevertheless, the magnitude of the increases indicates that the diagnostic criteria employed for the interpretation of the glucose tolerance tests in the younger age group have to be modified for those with advancing age. For clinical purposes it needs to be emphasized that although a single two-hour postglucose blood sugar level is an excellent screaning procedure, it gives only limited information in an evaluation of carbo- hydrate tolerance. To be diagnostic for diabetes mellitus a slightly elevated two-hour blood sugar level must be preceded by a considerably greater elevation of the blood sugar level at one and one-and-a-half hours as has been previously defined for the younger age groups. What these criteria should be remains open at present. In this connection, it should be noted that diagnostic criteria for normal and abnormal glucose tolerance in older people cannot be established by studying an unselected population group. It is well known that (1) the incidence of diabetes mellitus increases with age; (2) even in advanced age the disease need not be overt; (3) diabetes occurs more frequently in relatives of diabetics; and (4) the true prevalence of the diabetic diathesis is presently unknown but appears to be very high. In addition, absence of family histories of diabetes mellitus in preceding generations is even less meaningful in older people than in younger individuals. Since there is no evidence for a bimodal distribution of glucose tolerance tests even in younger relatives of diabetics, it will prove difficult to establish diagnostic criteria on a statistical basis from unselected population studies. Thus it appears that data other than those derived from a study of carbohydrate tolerance alone will have to be used eventually to provide valid conclusions in regard to the interpretation of glucose tolerance in the aged. Having defined criteria for the interpretation of the glucose tolerance test in a younger age group, let us now apply them to subjects under the age of 50. In 124 control subjects we found only one glucose - 10 - tolerance test indicating diabetes mellitus and one indicating "probable diabetes. On the other hand, among 609 healthy close relatives of known diabetic patients, consisting of parents, siblings, or children of a known diabetic, 99 or 16% were found to be diabetic. Another 17, or 3%, had curves indicating ''probable diabetes". Another comparison between glucose tolerance tests of normal subjects and those of relatives of diabetic subjects is shown in Figure 3, which gives the distribution of the summed blood glucose values of the glucose tolerance tests for the two groups. In the asymptomatic relatives of the diabetics which include 99 clearly diabetic subjects, the range of blood sugar values indicates a continuous unimodal distribution and does not show evidence of two distinct population groups, although there is definite evidence of skewing toward the higher values. Thus, although one can establish standards of normal versus abnormal for certain age groups on a statistical basis, there is no sharp cut-off point between normal and abnormal with respect to carbohydrate tolerance in relatives of diabetic patients. When there is sufficient abnormality, the disease can be recognized as a discrete entity. The blood sugar aspect of diabetes is not characterized by a sudden onset of deranged carbohydrate metabolism but by subtle and progressive changes in glucose tolerance for years prior to the onset of recognizable chemical disease. This can be illustrated further by an analysis of glucose tolerance tests derived from offspring of marriages in which both partners had clinical diabetes. Figure 4 contrasts the mean curve of non-diabetic offspring of conjugal diabetics with that obtained from control subjects of a similar age + 11 = group. The offspring of conjugal diabetics show consistently higher mean values than do the control subjects. Since subjects with latent diabetes have been eliminated, the higher levels in the offspring of conjugal diabetics are not due to the contribution of a relatively few highly aberrant individuals. This points out that minor abnormalities in glucose tolerance exist for this group even though one could not label a single one of these individuals as being abnormal. Further evidence to point to the importance of minor abnormalities of glucose tolerance which provide added security for our diagnostic criteria come from careful, long-term re-evaluation of patients with mild abnormalities of glucose tolerance. Table II shows the progression of "probable diabetes to diabetes in 15 subjects, age 12 to 47, who have been retested after intervals ranging from one month to eleven years, with a mean of three years, It has been appreciated for some time that mild abnormality in glucose tolerance may be found in asymptomatic middle-aged diabetics. Because in children the first manifestation of symptomatic diabetes is frequently of sudden or explosive onset, many have believed -- and I may say, still believe -- that diabetes in the young can rarely be recognized at an early stage. However, if a search is made for diabetes, the disease can be recognized even in the young many years before symptoms occur. In this stage it has the characteristics of "maturity- onset type" of diabetes. Table III shows an interesting example. A diagnosis of diabetes mellitus was made in this ten year old child of conjugal diabetics on -12 = the basis of a routine glucose tolerance test. The test was normal ten months later, after the patient had lost weight on a 1400 calorie diet. Four years later the test was mildly abnormal but not diagnostic in spite of a weight gain of 67 pounds. A year later the test was again diagnostic in spite of maintenance of weight. After a weight loss of 18 pounds, the test was again near normal. Three years later, at the age of 18, she gave birth to an infant weighing 11 pounds 8 ounces. At the age of 19, having gained weight to 205 pounds, the glucose tolerance test was again diagnostic of diabetes as was an oral tolbutamide test. Thus, over a period of nine years, this child of conjugal diabetics manifested all the characteristics of the "adult type' of diabetes mellitus. Table IV illustrates what may be found when a search is made for diabetes in individuals with a close family history of diabetes. Routine glucose tolerance tests were performed in ten asymptomatic children of a diabetic father. All were of normal weight. Seven of these offspring were shown to have latent diabetes. The abnormality ranged in severity from a fasting blood sugar level of 322 mg. per 100 ml. with a four-plus glycosuria and acetonuria, found in the oldest daughter, age 30, to the mildest, a peak blood sugar level of 178 mg. and a two-hour level of 140 mg. per 100 ml. found in a 21 year old sibling. The 16 and 11 year old children manifested peak glucose levels of 210 and 219 mg. and two-hour levels of 160 and 161 mg. per 100 ml. This family, as well as the case study of the ten year old child previously presented, illustrate that diabetes mellitus may progress slowly in the young as well as in the old and that ample “13 = time may be available for institution of prophylactic procedures. Although the disease may have a very slow evolution in childhood, it may also progress rapidly. This is illustrated by the findings in a 16 year old boy. Absolutely normal glucose tolerance and cortisone glucose tolerance tests were followed by symptomatic diabetes tean months later. It is well recognized that variations in glucose tolerance may be encountered if the test is repeated at intervals on the same individual. When the position of individual points on the glucose tolerance curve is subjected to statistical analysis, some variability may be observed in the results of repeated tests. For diagnostic purposes it must be emphasized that in the great majority of instances both members of a pair of glucose tolerance tests either fall within or outside the limits defined as normal. However, changes from an abnormal glucose tolerance to a normal and vice-versa do occur when the test is performed on repeated occasions in subjects with borderline tests. As was emphasized and illustrated earlier in this discussion, diabetes is not necessarily a disease with steady progression, and spontaneous variations in carbo- hydrate tolerance do occur. Extreme examples have been documented in individuals who had even been in diabetic coma and subsequently exhibited normal glucose tolerance. Thus, repeated fluctuation between abnormal and normal glucose tolerance is part of the early natural history of the disease. Only long-term studies will determine how frequently this occurs in "non-diabetic" individuals. Reactive hypoglycemia three to five hours after ingestion of glucose occurs in a large number of individuals whose glucose tolerance tests = 14 = disclose mild diabetes mellitus. Reactive hypoglycemia does not rule out the presence of diabetes mellitus. Symptoms of spontaneous hypo- glycemia may indeed be one of the earliest clinical manifestations of the disease. An explanation of this phenomenon, confirmed by insulin assays, is that the early hyperglycemia is due to a delay in insulin release after administration of glucose but that the hyperglycemia which results causes delayed but supernormal release of insulin with the resulting hypoglycemia. I bring this out for an important reason. When one considers diagnostic criteria for the glucose tolerance test one should not consider the three-hour point as having to be elevated because, if so, one would eliminate from consideration many glucose tolerance tests, characteristic of diabetes mellitus. A diagnosis of diabetes mellitus cannot be made with confidence on the basis of an abnormal elevation of the blood glucose level at one-half or one hour if it is accompanied by normal two-hour blood glucose levels during the glucose tolerance test. However, careful follow-uprof otherwise healthy individuals with abnormal high peak blood glucose levels during the glucose tolerance test is indicated. Progression of a curve with a high peak blood glucose level to one diagnostic of diabetes mellitus may occur within a period of a few months. Thus, when one considers screening procedures, one may increase sensitivity by selecting an earlier time for detection of a possible abnormality than the usual two-hour glucose blood glucose level. Renal or non-diabetic glycosuria has been thought to be a benign condition. However, we have found unsuspected diabetes in 62% of - 15 - patients with non-diabetic glycosuria in whom follow-up glucose tolerance tests were performed up to 30 years later. Approximately one-third of the group had initially exhibited fasting glycosuria. This indicates that non-diabetic or renal glycosuria may be indicative of potential diabetes. I would like to stress once more that the diagnostic criteria given apply only to otherwise healthy, ambulatory subjects under the age of 50 and do not apply to individuals with acute or chronic disease. In particular, of course, they do not apply to patients with pancreatic disease; certain patients with endocrine diseases such as acromegaly, Cushing's syndrome, pheochromocytoma; patients with liver and brain disease; patients with poor nutritional intake; patients under great emotional stress; and individuals who are seriously immobilized, such as patients with orthopedic handicaps or severe arthritis. It is recognized that judgment has to be exercised in the interpre- tation of the glucose tolerance test when blood sugar values deviate by only a few milligrams per 100 ml. above or below the critical figures given and that the results of repeated tests have to be evaluated under such circumstances. Variability of results is greater in individuals with borderline curves than in subjects in whom the results fall clearly into the normal or abnormal range. I have mentioned that we have used 1.75 g. of glucose per kilogram of ideal body weight. In many women this will come very close to 100 g. of glucose. In most men this dose will exceed 100 g. of glucose, and will go up to approximately 150 g. of glucose in a 90-kilogram man. For ordinary clinical purposes there is probably no great difference in the - 16 - results obtained with the 100 g. dose as compared to 1.75 g. per kilogram, although I know of no exact comparison. The British have been using 50 g. of glucose as the standard dose for the oral glucose tolerance test. With this dose the peak levels reached may not be quite as high, although the greatest difference probably is that the two-hour value will be lower than with the larger dose of glucose. The main principle of the larger dose is to insure absorption of glucose over the first two-hour period of the test. Maximal absorption of glucose is probably close to 0.8 g. per kilogram body weight so that with the 50 g. glucose dose there may not be much absorption of glucose after the first hour. This puts less stress on the insulinogenic mechanisms. In diabetics a 50 g. carbohydrate breakfast may be followed by a normal 2-hour postprandial blood sugar level while with a 100 g. glucose load this is not so. Also, increases in plasma insulin are smaller and less prolonged after 50 g. than after 100 g. of glucose in normal subjects. The advantage of the smaller dose may be that it produces nausea less frequently, particularly in pregnant individuals. Undoubtedly Drs. Kent and Leonards will speak about the use of 75-gram dose, and I shall not say anything about this. As far as variation in blood sugar methods, this used to be a problem. Now with automation there is probably less variability in methods than there has been in the past. Of course, we still haven't reached the ideal, namely automation with the glucose oxidase method. Such a method has been proposed but it is not working well on a large-scale to my knowledge. As far as venous versus capillary blood is concerned, in the United | States most people use venous blood, so I don't think there is much of a “ 1] = variability here. The British, of course, use capillary blood primarily. One word about the intravenous glucose tolerance test employing the rapid injection of glucose. By plotting the rate of fall of blood glucose per unit time on semilog paper, a single K value representing glucose utilization can be calculated. The advantages of the intravenous glucose tolerance test are: (1) avoidance of the variables of gastrointestinal absorption; (2) evaluation of the test by a single figure; (3) less time necessary for performance of the test; and (4) avoidance of nausea. There are also disadvantages. First of all, with the usual load used, namely 50 cc of 50% glucose or 25 g. of glucose, this test is less sensitive in patients with mild diabetes -- and we have many examples of that. Dr. Morehouse published some information along these lines in the November 1963 issue of "Journal of Clinical Endocrinology and Metabolism'. One can see definite abnormality of the oral glucose tolerance test at a time when the intravenous glucose tolerance test gives a normal result. This may be due to two reasons. In states of only mildly diminished carbohydrate utilization, a small load given may be removed in a normal fashion. On the other hand, with the oral test a larger load is given with persistent absorption over two hours necessitating not only removal of the amount absorbed initially but removal of the continuing influx of glucose. Secondly, with the intravenous test the blood sugar is raised suddenly to high levels. The earliest abnormality in mild diabetes is not an inability to release insulin but an inability to release insulin promptly in response to a normal elevation of the blood glucose of only 20 or 30 milligram per cent after administration of glucose. However, in mildly diabetic subjects a large elevation of the blood sugar may « 18 = release insulin quite normally or as Yalow and Berson have shown, when there is a supernormal elevation of the blood sugar there is also super- normal release of insulin. Thus, with the oral glucose tolerance test there is a delay in insulin secretion after administration of glucose and therefore the ensuing hyperglycemia. However, when glucose is introduced in the circulation suddenly to a much higher level by the intravenous route, this may be followed by normal insulin release and therefore normal glucose utilization, hence the decreased sensitivity of the intravenous test. The second disadvantage of the intravenous glucose tolerance test is a theoretical one. The intravenous glucose tolerance test is a less physiological test than the oral glucose tolerance test. It has now been demonstrated that physiologic insulin release depends not only on the height of the blood sugar in the pancreatic artery but also on a stimulus arriving in the portal circulation. When the same amount of glucose is given to normal subjects at the same rate intrajejunally and intravenously, intrajejunal administration causes greater insulin release and less hyperglycemia than when the same amount of glucose is given intravenously. The magnitude of the hypoglycemic response to sodium tolbutamide, the tolbutamide response test, has been proposed as another means for the early detection of latent diabetes. The criteria for the interpre- tation of the intravenous tolbutamide response tests have been discussed previously. For a normal oral test, the cut-off point is a decrease in blood sugar to 78% or lower from fasting levels at 30 minutes. We have made w 19 = a comparison of the glucose tolerance test and the tolbutamide response tests in the same mildly diabetic patients who manifested various degrees of abnormality of carbohydrate tolerance. We have previously reported that the glucose tolerance test is a more sensitive test for the early detection of latent diabetes than is the intravenous tolbutamide test. Table V shows similar data when the oral tolbutamide test is compared with the oral glucose tolerance test. An abnormal response to the tolbutamide test in the absence of liver. disease, starvation, or pregnancy may be interpreted as evidence for the presence of diabetes mellitus. However, a normal response to the tolbu- tamide test should not rule out the presence of mild diabetes since the glucose tolerance test is a more sensitive indicator of the presence of mild diabetes than is the tolbutamide response test. I had planned to continue the discussion by taking up possible detection of subclinical diabetes in non-pregnant individuals since this will address itself to the problem of further refinements of diagnostic techniques as listed in your outlines. However, time does not permit. I was going to show you our most recent evaluation of the cortisone glucose tolerance test for the detection of subclinical diabetes, and that there is certainly progression of the incidence of positive results with advancing age. I was going to pose the question as to the cause of this progression with age =-- what is this due to? It may indicate that the diabetic diathesis is detected with increasing frequency as age progresses, just as the incidence of clinical diabetes mellitus progresses over the same period of time, or that different criteria for the inter- pretation of the cortisone-glocuse tolerance test may have to be «30 w established during different periods of life, especially for the very young and the older groups. I think both factors apply. I had hoped to show you some follow-up tests in individuals with positive cortisone glucose tolerance tests between the ages of 20 and 40. In this group the later incidence of diabetes was 26% and of probable diabetes 6%. Thus, we feel that this test does detect diabetes at an early stage. But I should like to emphasize that the cortisone-glucose tolerance test should not be considered as a test for detecting genetic susceptibility to diabetes, but as a more sensitive test than the standard glucose tolerance test. Below age 40 and perhaps below age 50, conversion from a negative to a positive cortisone-glucose tolerance test, which persists on repeated occasions, can be interpreted tentatively as indicating the progression from prediabetes to subclinical diabetes. Not every individual with a positive cortisone-glucose tolerance test should be expected to progress to latent diabetes, just as a latent diabetic will not necessarily progress to overt diabetes. Although the cortisone-glucose tolerance test may prove to be of value for the possible detection of subclinical diabetes, it is still premature to allow it to be used for routine clinical testing and interpretation. DISCUSSION STEFAN S, FAJANS, M.D. DR. McDONALD: TI would like to ask if there are differences due to dilution of a loading agent? Does it make a difference how much is in a solution? DR. FAJANS: TI think it does make some difference. The less the dilution, the greater the incidence of nausea. Nausea per se can cause variations in gastrointestinal absorption. And while I am speaking of gastrointestinal absorption, a great deal has been said in the literature about the variables of gastrointestinal absorption with the oral glucose tolerance test and there is no doubt there are some variables. However, when there is a decrease in gastrointestinal absorption this certainly could not contribute to a falsely abnormal glucose tolerance test. With increased gastrointestinal absorption, such as in the patient with a partial gastrectomy, one may get an elevated peak blood sugar level, but the two-hour blood sugar level will be normal and this will not con- stitute a diabetic glucose tolerance curve. DR, LEONARDS: I would like to recall for the group a statement Dr. Lazarow made last night, namely when we are talking about glucose tolerance tests - times, variability, nausea - we ought to talk about the experimental data to tell us about the incidence of nausea. I don't know of experimental data to tell us about the differences of different concentrations of glucose on nausea, and I would appreciate hearing from anybody who has such information. DR, ROOT: I am not rising to answer that question but to ask whether Dr. Fajans' experience has been somewhat like ours. It has been so w 29 commonly said that diabetes increases with age. I think that statement should be corrected at least in our experience. Following the period of 50 to 60 years, the frequency of diabetes mellitus as a clinical state sharply falls. When we talk about diabetes and age, what we really mean is diabetes increasing in a peak up to 55 or 60. After the age of 56 or 60, to our knowledge and that of others, too, the frequency of the development of clinical diabetes has a very sharp fall. I don't know what this means. Perhaps Dr. Fajans has some different experience or some comments to make on that, but in talking about methods and population studies, we ought to take this into con- sideration. DR, FAJANS: Actually there were two questions here. First, as to Dr. Leonards' question, I also do not know of any strictly controlled studies. Nevertheless, the question has come up, an opinion has been asked for, and an opinion has been given. But I cannot back it up with figures. Nevertheless, people who do a lot of glucose tolerance tests have found universally that the greater the load and the greater the concentration the greater the incidence of nausea. As far as Dr. Root's question is concerned, I was under the impression that the rising incidence of diabetes goes actually beyond the age of 50. In our studies, as you have seen from the figures, we are dealing primarily with younger individuals, since we are interested in the natural history of diabetes. We actually have very few individuals beyond the age of 50 among the group of relatives of diabetics. Among the University of Michigan employees, the ages went from the early 20's to the 80's. But individuals with symptomatic diabetes or with “«q3 definite fasting hyperglycemia had been excluded. The point of the slide (Figure 2) was to show, as others have shown, that there is a continuous increase in both the mean two-hour blood sugar level and the standard deviation as age increases. DR. ROOT: Our peak has been not at 50. The peak is somewhere between 55 and 60. I am not talking about latent but clinical diabetes. It is striking to me that there is something about the age period following 60 years which seems to protect people against diabetes. DR. ACHESON: They all die off. DR. ROOT: Oh, no, these were corrected for age. With the increasing age of people in the country, the onset of clinical diabetes =-- not prediabetes or chemical diabetes, but clinical diabetes =-- seems to decrease after that period. DR. KLIMT: I think there is a little confusion about terms here. I believe I know what Dr. Root is referring to, and this phenomenon is shown as a cumulative prevalence curve, which indeed does show what he says, namely that it rises in middle age, reaches a plateau, and then declines -- a cumulative prevalence. Now, that is not surprising because after certain ages the age-specific prevalence tends to decline again, as a greater proportion of diabetics has died than deaths have occurred in the total population. I don't know of any incidence data at all, and I think these terms should not be used in connection with diabetes. We are speaking of cumula- tive prevalence, and in cumulative prevalence this phenomenon is quite observable and clear. - 3 = DR. WEST: I may be wrong but I think Dr. Pyke, in England, did accumulate data with respect to the frequency with which diabetes is discovered in the various age groups. I think he showed what Dr. Root has mentioned, that when all the adjustments are made -- and they are indeed difficult to make -- incidence appears to be highest in the fifth and sixth decades and begins to drop off in the seventh and eighth decades. Someone may remember this more specifically. DR, NEWILL: TI would be highly surprised if this were true, Dr. West, because I don't think the incidence of diabetes is something that has been observed in population groups. DR. WEST: This is the incidence of clinical diabetes, not the incidence of diabetes. DR. NEWILL: That has no different meaning to me. Some diseases require prevalence measures and some incidence measures and in long-duration diseases there is a tremendous amount of difference between incidence and prevalence. In any actual age group the incidence of diabetes is so small I don't believe there has ever been a population study that has been done where observed incidence rate has been noted. All of the information that we have on incidence is indirect and has been computed from existing prevalence rates where it will give us this same kind of information because it is based on the prevalence rate which does drop off as Dr. Root has mentioned. If Pyke has done this, I have not seen it, and I specifically looked. DR. WEST: I will say that this, if it exists, is apparent incidence data. As Dr. Fajans brought out, an individual may have had diabetes at age 30 and - 95 reports to the physician and exhibits symptoms at age 60. This is the sort of data that is being dealt with. Actually, there would be a possible explanation for this phenomenon. It may be that if you have the genetic defect, you eventually will express it, usually in your fourth, fifth, or sixth decade. By the seventh decade, almost all of those who have the trait may have already developed the clinical or laboratory expressions of the genetic defect. DR. LAZAROW: I don't know if rats have any bearing on this problem but I can say some of our own data at least indicate that this kind of phenomenon can indeed occur in the rat. We have a series of studies where we induce what we call the sub-diabetic state in the rat by giving a dose of alloxan, which was enough to alter the glucose tolerance but not enough to affect fasting blood sugar levels so the individuals never had hyperglycemia, or any significant glycosuria, but you could differentiate them from the non- injected rats on the basis of glucose tolerance. Then we did quite a large series of glucose tolerances on these animals at monthly intervals. We were interested in seeing whether you could start with abnormal subdiabetic animals of this type and push them into manifest diabetes. And this sometimes happened. But in a very large group of animals we found during the course of the first year the glucose tolerance became progressively more abnormal as the year progressed. Few of the animals went into manifest diabetes; most of them did not. But then between the first and second year, there was a progressive improvement in the glucose tolerance in these same animals, So something is happening in these animals, in terms of their response to the same dose per kilo, which “ Of w would be consistent with what Dr. Root had said in terms of the manifesta- tion. In other words, this may not mean that the diabetic state is any different, except that it might not show the abnormal response to glucose with that given dose. You might have to have a still larger dose in order to demonstrate it. DR. MILLER: This is a single insult, is it not? DR. LAZAROW: We gave one dose of alloxan at age three months. Nothing else was done to the animal except glucose tolerance and we have as many as 20 on one animal. DR. MILLER: Comparisons may not be valid in this case because perhaps in humans there is a continuing influence on the state we are discussing. DR. LAZAROW: But I would suggest that even if there isn't a progressive input there may be other factors which influence the tolerance to glucose, which would result in values indicating glucose intolerance and maybe manifest diabetes as well. It is conceivable that the same individual, if Beta cell damage is the underlying problem, might show glycosuria and release mechanism aberrations at age 50 and not at age 70. DR. ACHESON: TI am neither a diabetologist nor a metabolist, but there are certain aspects which worry me. Dr. Fajans referred to the diabetic diathesis. Now, I expect he means by this the inborn liability to diabetes. But, I wonder if it wouldn't be more useful if he talked about diabetic diatheses. I think there are many things that bear on this, the pituitary, the pancreas, the liver cells and fat metabolism =-- these are all variables w- DF associated with diabetes, all presumably controlled by different enzymatic systems and therefore different genes. So I wonder if it isn't useful to think about various gene weaknesses which may or may not make a person liable to diabetes. If this is the case, one has diatheses and having diatheses they become graded. Surely there cannot be just one diathesis. And then I wonder if we are looking at the glucose tolerance curve the best way we can. The conventional way of doing it is to have glucose on one axis and time on the other =-- one hour, two hours, and perhaps half hours in between and then to plot a line. Now with one person this is splendid but once you start looking at populations, there are various problems that creep in. One has confidence limits about each point not only for glucose, but also for time, so in fact each point begins to have an elliptical shape. The next thing we have to do is try to devise some method of isolating different shapes of curve. There may be the individual who peaks up during part of the test and goes down to normal levels during the test course, but taking blood from people at a few arbitrary points on the curve, as an epidemiologist must, one may miss them. Diabetes, like coronary heart disease, may not be a single simple disease. I think we can say with some confidence that in coronary heart disease there is a variety of lesions with different epidemiologies. Thrombosis has a different epidemiology from various kinds of arterial disease and so on. So I wonder if it would be possible in population studies to try to isolate various kinds of glucose tolerance curves and - 28 = then look at each in populations. DR, KLIMI: We indeed now have jumped into the middle of the problem, if I may say so. If we look at our assignment, so well expressed by Dr. McDonald, how do we express the results of the glucose tolerance test in an optimal fashion? And to realize that each point has sampling variation around time and variation of error due to chemical determinations and other factors is one thing. The next thing is to recognize that these points are not independent of each other, and in fact, as we were able to show, every one point on this line is significantly correlated to the next one. Therefore, the totality of the curve needs really to be expressed in such a manner that we discriminate between diabetics and non-diabetics in an optimal fashion. I mentioned before, Dr. Fajans, that we do not have, as you also observe, an optimal way out of a population study to see where the curve becomes a bimodal. Another approach, less perfect but still an approach, is perhaps the one on which the University Group Diabetes Program at twelve universities based their diagnostic standards. The basis is the comparison of a known group of diabetics, symptomatic diabetics, I mean now, where the diagnosis does not come from the glucose tolerances, with a general population sample, in this case prison volunteers. We took four values, fasting one, two, and three-hour levels, and determined the frequency distribution of the normals and abnormals for each of these and determined the degree of overlap or, in other words, the power of each 0G test at each level to exclude people from further suspicion of diabetes. As has been recognized by many predecessors, this has a certain order of sequence. The fasting =-- the frequency distribution of the normals and the symptomatic abnormals has a great degree of overlap, in fact so great that if we draw a line where they no longer overlap we can only exclude 127% of a general population sample from further suspicion of diabetes. DR. KLIMT: If the same is done for one hour post-glucose ingestion, we find quite a different picture. The general population will distribute one way, and the symptomatic diabetics will distribute another way, so we can now exclude some 827% of a general population sample from further suspicion of diabetes. If this is done for two hours it excludes approximately 64%, and for three hours, 35%. There is not perfect correlation of these values, but there is correlation. In other words, if there were perfect correlation, all we need to do is to take the one hour and we would have excluded all the non-diabetics. However, if we cumulate these curves we can exclude as much as 947%, leaving 67 of a general population sample still suspect of diabetes, which is approximately the maximum prevalence found in the highest age groups in the Oxford survey. If you now combine these values with their various power of exclusion -- 12%, 82%, 64, and 35 -- weighing each of the blood sugar values obtained in the course of a glucose tolerance test and then add, you get a summed value which you can also graph as has been done for each single glucose value. And that pulls apart in a maximum manner the general population from the symptomatic diabetics. -w 30 = This approach is, by coincidence, rather similar to the simple summation of the glucose values which, as I saw, Dr. Fajans is also using in the demonstration of his data, because we are all in need of one single parameter to express the glucose tolerance test. We sum over four values because it is easier than to weigh each value with 12%, 82%, et cetera. We use as diagnostic criterion a summed value of 500 or more, which includes the fasting, one, two, and three hour value after oral challenge with 30 grams of glucose per square meter of body surface. DR. LAZAROW: Before we proceed, are these not diagnostic criteria for manifest diabetes? They would tell you nothing about a lesser degree which may be as important in this study. DR. KLIMT: That is perfectly correct but I maintain we don't have any intrinsic information in the glucose tolerance test for the gray zone. We must classify the glucose tolerance test by further information because intrinsically it doesn't contain the separation of which we are talking, between normal and abnormal. It isn't in the test. It is only by virtue of the fact that we associate the signs and symptoms with the glucose tolerance that we begin to make a distinction. DR, WEST: Dr. Klimt, I am trying to imagine what symptoms of diabetes you might have with a normal two-hour value. You are not talking about polyuria, are you? What symptoms are we talking about? DR. KLIMT: I am not aware that I mentioned anything in that respect. DR. WEST: Didn't you say you had a group of symptomatic diabetics? DR, KLIMT: Yes, but not at two hours or anything of the kind. It is a group of symptomatic diabetics who were given the glucose tolerance test and the frequency distribution of their summed values determined. ow 30 = DR. WEST: Did you not mention that there was a considerable overlap between the normals and the symptomatic diabetics? DR, KLIMT: There is, because we didn't take normals. We took just a general population sample which should include diabetics. In other words, we have plotted a distribution of the general popula- tion and failing to have discovered bimodality we use a substitute, less good technically but the only thing we have at the moment. We use symptomatic diabetics diagnosed on some grounds other than the glucose tolerance test. DR. LAZAROW: But how clearly defined is that? Couldn't you get a half-dozen different varieties of symptomatic diabetics that you could use that would change this completely? DR. LUKENS: I wonder if some of the confusion about what disturbs Dr. Lazarow and Dr. West might be resolved. Are you going to give us an example distinguishing between one normal and one diabetic person as a result of this calculation, or are you talking only about groups? DR. KLIMI: I am trying to explain the rationale behind the diagnostic criteria of the University Group Diabetes Program, which is made up of individuals so diagnosed. DR. LUKENS: I want to learn your rationale, but I want to know whether you are going to apply it to individuals or to a group. Then can you give us a curve between one normal person and one diabetic and say, "With this calculation I make this difference'? DR, KLIMI: That is what I am about to do. I think the answer to Dr. West is that this was a group of out-patients at Johns Hopkins who were diabetics of long standing, with all the classical “ 3 wm signs of diabetes. If you take the distribution of the summed values of the four glucose tolerance test blood levels in these symptomatic diabetics, you get a certain distribution. And we have taken a region where there is a minimum of overlap with corresponding values similarly plotted for a '"mormal" population. And it so happens that this value is 500 or more as a separating point. Now, I fully agree with all the faults you can find clinically on the definition of this group, because it is only a substitute for the popula- tion-based study. However, as a suggestion, if this were done in several different environments on several different populations, both normal, and diagnosed on the basis of criteria independent of the glucose tolerance test, we might then come to a consensus of the optimal separating point between these two populations. DR. VOUGHT: How about the sundering and dispersion concepts? DR. KLIMT: Well, they are skewed so it wouldn't be possible to express them adequately in -- DR. VOUGHT: If you plotted them -- DR. KLIMT: If you take the mean of this group, it is about 900. You may transform it into logs. DR. VOUGHT: Is it normal? DR. KLIMT: On a log transformation? That is a good question. It will come closer to normality but whether a log-normal curve will result I couldn't tell you. DR. ANDRES: What is the age range of your two groups? DR. KLIMT: These were done separately and matched in three age groups: under 35; 35 to 49; and 50 - plus. - 33 - DR. ANDRES: What does the 500 figure refer to then? DR, KLIMT: To the sum of four glucose values. DR. ANDRES: Which age group? DR. KLIMT: This refers to a group which is equally represented by persons under 35, between 35 and 49, and 50 and over, in other words, an artifi- cially standardized population. DR. ANDRES: So if you are recommending this, you would recommend no adjust- ment for age in criteria for the diagnosis of diabetes; is that correct? DR. KLIMT: I haven't been recommending here. I have been explaining one method in use. I think it is a group like this which might recommend that we change criteria and that, if we have this by age group, that indeed we may have different cut-off points as we go along age. That may well be. DR. McDONALD: TI appreciate Dr. Klimt moving this in the direction we hoped it might go. The reason I brought up the first question on the dilution effect, I presumed there would be less divergency on that and perhaps that might be an area of agreement. I have heard nobody contest that the timing should be measured from the beginning of the injection. So, hopefully, we could start with those areas of agreement and search for others. DR. KLIMT: In this context may I ask why the ideal body weight is used as the basis for dosage? DR. FAJANS: Dr. Acheson asked whether there would be any objection to speak- ing in terms of diabetic diatheses instead of diathesis. I can't say I have any objection to it because we don't know. We don't have any informa- tion to say that '"-es" is more correct than "-is''. - 34 = As far as inheritance of diabetes is concerned, Dr. Neel, our geneticist, on the basis of these data, limited as they are, feels he coanot support the recessive hypothesis which has been proposed since 1935 and is still being championed by most as the most likely one. He feels that a more likely explanation is a multi-factorial type of inheritance. This, however, does not mean this makes diathesis diatheses. Obviously, we are talking about secondary diabetes. DR. ACHESON: If inheritance is multifactorial it must mean there are degrees of diathesis. DR. FAJANS: Definitely. As an example, there are certainly many enzymes involved. We know, putting it the other way around, that insulin affects a variety of enzymes. I think people have talked about multiple etiologies for diabetes for a long time in the literature. Actually, I am not aware of any support for more than one variety or type of diabetes. Admittedly, the defense for this is not very strong, either. One can simply enumerate a long list of clinical observations: adult type and juvenile types of diabetes are seen in the same family; adult type may progress to juvenile type; they may have the same type of vascular complications -- these are all very rough clinical correlations, and I think the defense has to rest with that. Then, as far as the points being not independent from each other, I think this is a very important point, and Dr. Klimt mentioned this, too. Certainly the one-half hour point depends on the one-hour point, and so forth. I think I made myself clear, but just to stress it again, we do not feel we can show a bimodal distribution of blood sugar values in relatives of diabetics. w 35 we As far as diagnostic standards are concerned, I used the sum of the seven values of the GIT only for demonstration of absence of bimodality as a simple, convenient test. As far as diagnosis is concerned, we have to think not only along statistical principles, but also along physiologic principles. I think, on the basis of physiologic principles, we now recognize the fact that the mild diabetic has delayed insulin release but then actually may have super-normal insulin release, that the super-normal insulin release may precede the reactive hypoglycemia. And therefore, to me, it doesn't make sense to use a three-four or four-hour blood sugar level in the criteria for diagnosis of diabetes, because, when you use a group of diabetics where we need our diagnostic criteria with greatest urgency, the mild ones, it would detract. We don't need diagnostic criteria for individuals who have symptomatic diabetes. We need them in the asympto- matic group who are very close to normal. So this is why I do not favor the mean of all points. If you want to use the half-hour, one, one-and-a-half, and two hour values -- those four together, this might make physiological sense. DR. KLIMT: The two at the end would get the least weight, either by summa- tion or by weighing in the sense I have indicated. However, the three-hour is still more powerful than fasting. DR. FAJANS: It may be more powerful in those who have the greatest elevation of the glucose tolerance curve, but the point I make is that using the three-hour value may even detract. You may recall in Figure 1 an individual who had a one hour value of 225 and a two hour value of 170 mg. per 100 ml. Well, the three hour value was 46 mg. per 100 ml. - 36 = DR. KLIMT: TI added those. They still would fall under our diagnostic criteria. DR. FAJANS: But the three hour point would detract from the sum? DR, KLIMT: Yes, and maybe at times rightfully detract because they are correlated. But in this particular instance, I used that one and it still comes to 560 in summation. DR. FAJANS: Finally, I would agree completely that the sum total of information is the important thing. As far as determining what should be diagnostic criteria, it is a little difficult for me to see, as it was for Dr. West, how, by comparing symptomatic diabetics with a normal population group, we are going to get the most discriminating information. DR. KLIMT: TI didn't say normal; I said the general population group. Do you know a better definition? We need an outside definition of diabetes to follow this technique and it may not be connected to the glucose tolerance test. And then one sees what is the best way to use the glucose tolerance test. DR. FAJANS: If you talk about symptomatology =-- in other words, weight loss and so on, you are starting with a fasting blood sugar of 250 mg. per 100 ml. or above in 997% of the cases. I think we could all agree that when an individual has a fasting blood sugar of 130 mg. per 100 ml. by a true blood sugar method and in a non-stressed individual a two hour level of 220 mg. per 100 ml. we could all agree that this is diabetes. And if you start with this kind of a population group, you are in the same ball park. But starting with a group of symptomatic diabetics, you are not even close. - 37 = DR. KLIMT: All I want to emphasize is that we need something else to diagnose this group, something other then the glucose tolerance test. DR. FAJANS: Yes; this is the emphasis I made when I said: What are going to be the criteria for the aged population? For the reasons that I listed, I don't see how we will be able to attack this on a statistical basis. We need an outside criterion, be it ILA, or anything else that is going to come along -- I hope will come along in the very near future. But we need, I agree completely, an outside parameter. The last question I was asked was about ideal body weight, and I think this goes back many, many decades, long before I came into the field. The idea at that time was that one should relate the glucose load to tissue capable of metabolic activity. At that time, of course, it wasn't known that fat was a very active tissue as far as glucose metabolism is concerned. When one talked about ideal body weight this really excluded the variations in fat. So I hold no brief for ideal body weight. DR. KLIMT: We use, in fact, a scale based on body surface. But any other scale, as long as we don't give the same challenge to people of different sizes and different mass, should be acceptable. DR. LAZAROW: TI would like to bring this discussion back to the general area that Dr. McDonald raised in terms of getting agreement on the conditions under which you carry out the experiment. I am less concerned with the interpretation of the data, because I hope that in all the studies all of you are doing, that you put in every single blood sugar value that you measure; that you don't mask your data by just giving us the summed number; that you have every bit of broad data, and then when you have a population study over a period of ten or twenty years, you will be able to have the « 38 criteria in terms of what happens to these patients, and go back and you may have it all on the computer and analyze what would have been the best way to do it. So this part of it in your interpretation we can leave for twenty years from now when you have the data on which to base it. It is not based on symptoms but on what actually happens in these patients. But in order to make that kind of data more meaningful, we ought, if possible, to come to some common agreement in terms of dose, times, and so on; when you are drawing your blood, when you are giving your sugar, what concen- tration. And let's look at what information we have, if any, on how important are these factors. Then, if we don't have any finite informa- tion with all the money that is being spent and all the studies that are going on represented in this room, maybe we should address ourselves to the problem of getting precise information on the same patients with varying doses, with varying times, with varying methods. Now, I did a study of this sort in our normal and sub-diabetic rats a couple of years ago where I took one series of animals and did 20 glucose tolerances on each one where, in a symptomatic way, we varied from one gram to two grams to three grams per kilo, oral, intravenous, and peritoneal, and we compared the one-hour, two-hour, three-hour, four-hour blood sugar and all the other parameters in this same group of animals. Maybe we should get the same information on the same group of human subjects, maybe on a cooperative basis and then have a basis for saying how important is 75 grams, or what-have-you. So it may be that with this careful analysis this may not make enough difference that even results that are carried out - 39 - quite differently by different methods may be compared meaningfully. But if we can focus our attention on getting those data, I think we will have gotten on a common tract that will be very useful. DR. SILVERSTONE: I wanted to emphasize the point I thought Dr. Lazarow was going to lead to, and that is not only should we keep records of our raw data -- and this may be pessimistic -- but probably the only point we may be able to agree upon is each of us should be able to have access to data in the literature, so should our feelings about this after or change, the data are not lost. I would at this point suggest that we consider that some means be set up so that on all tolerance tests, if the editor does not accept the raw data to occupy space in his journal, at least let there be a file on record so all the raw data can be preserved, so should re-analysis and comparison come up at a later time, these files are not lost. That is the only thing I can feel strongly on that we may be able to agree upon. I feel very much up in the air about the dose. I am really disturbed that we haven't at this point realized that we have flat glucose tolerance tests where, to all intents and purposes, no glucose has been absorbed. Actually, I am sure this is not what is happening. But what do we think is taking place when we give a 100 gram dose of glucose and find a highest curve on a point being 90 or 85 and the lowest point practically the fasting blood sugar level? I have seen curves of this type. They don't occur often. I wish I knew what was going on. I would like to ask the question in a different sense. How much of a blood sugar rise should we expect in order to feel that the individual - 40 = has been stressed? Basically, we give glucose, not to give it, but to displace the blood sugar. If we give glucose by mouth, we are giving glucose in an odd way. We are giving glucose in a manner that permits us, if we could, to measure two rates going on simultaneously, the rate of absorption of glucose and its rate of utilization or dispensation. Even its rate of utilization is contaminated because, if it rises high, we may get overflow, so we get loss of glucose in some individuals, particularly with intravenous loading. So we haven't escaped that way. But let's get back to the point: How much glucose has to be given to be sure we have stressed the individual? We have some approximations. We have the health examination survey in Michigan where a survey of individuals carefully studied were given 100 gram loads, 50 gram loads, or meals. And actually a difference was discernible only between the 50 and 100 gram loads, as I recall it. And a strange thing is that it didn't make very much difference whether that individual was fasting or whether it was afternoon or morning. I feel there is a difference between 50 and 100. I don't feel we can compare tests done on a 50 gram load versus 100 if we are measuring absolute levels. The question of the 75 versus the straight 100 gram dose, I think we have all agreed, is not critical. I would go along with that. DR. LAZAROW: But do we know? What is the evidence for that? DR. SILVERSTONE: My thought on this is based on the fact that I don't feel that we could discriminate small doses of glucose. I think the difference between 100 grams, or 120, or 80 would be very difficult for us to measure even if there were a difference. I think if we look at the variability about each point on a glucose tolerance test and repeat this and attempt “YL - to assess it against a different dose of glucose -- I think it is obvious we will never discern small differences in doses. Even in the intravenous tests where one avoids the variability of gastrointestinal absorption, which you have already described to us, here again very small differences in doses are not important as far as the slope of the curve is concerned. Amatuzio gave doses of 25 grams and doses of 35 grams with very little difference in the slope of the curve except that the curve, itself, was displaced. Others, I think, have felt that these differences are not quite so important. We have tried various doses of glucose in terms of body weight versus the standard dose and have been satisfied that we couldn't come up with any distinguishable differences based on small dose differences and hence we have used a fixed dose of glucose as our load. Perhaps another factor is we don't know on what basis, really, to vary the load. If we were to have done this on the basis of ideal body weight, this might have been an error in that fat tissue was actively metabolizing glucose. With regard to the configuration of the curve =-- aside from the questions I can't find an answer to -- I feel at present we have done about all we can in terms of trying to assess the variability at the half-hour, one-hour, one-and-a-half hour, and two-hour levels, and perhaps all we can do is try to see whether figures obtained by summating curves, or with a planimeter, as has been done, will provide any important differ- ences in the basic interpretations which you have already established on these points. However, I do feel that there is a difference between intravenous loading and the oral test. Our original interest in this was - G2 stimulated because we were looking for age differences and couldn't distinguish them with the oral test. We felt this was not because they weren't there. This was because we had enough variability to identify them in the normal population group. We had less variability with the others and that is only because we eliminated the increment of intestinal absorption. This again is only a guess. You have pointed out that while the intravenous test avoids GI absorption and comes up with a single figure representative of the individual score, takes a little less time and avoids nausea, the test was less sensitive. I don't know why this opinion exists. I would be interested in your enlarging on that point. The only large figures I have seen where the intravenous test has been applied to diabetic populations, are those which Drs. Conard and Bastini in Belgium have accumulated over the years, among diabetics diagnosed, I assume, by the usual oral test or fasting blood sugar. All their fasting blood sugars have poor K values. We have assessed their values, and their slope in diabetics is neatly distinguished from their control groups, their pregnant groups, their normal controls. Their figures are very close to our figures for normal controls. And I would guess that there should be a sharp difference between the two populations, the diabetic and the controls. And I would know no reason why the test should be regarded as insensitive. The test is sensitive enough to distinquish minute changes, actually, in normal individuals, on the basis of age alone. The magnitude of these changes on the basis of age is less than the magnitude of change between a normal and a diabetic. “ 43 = Now, whether there is some reason to figure that the rise in blood sugar, perhaps, is responsible for this -- I am not clear as to what you meant. You stated that the reason for the insensitivity of the glucose test was that the dose of glucose was small, and yet I had the impression that you stated that the rise in blood sugar was high. Somehow or other, it seems to me that the effective thing is the rise in blood sugar. If the rise in blood sugar is high and this is a disadvantage, then why is the dose being low a disadvantage? The two are somehow or other contradictory. Finally, 1 just want to indicate that with regard to an analysis such as Dr. Acheson has referred to, in terms of the oral test, what is re- sponsible for changes about the first point, changes about the second point, and changes about the third, I think is perhaps better approached not so much from population studies as such, but rather from physiologic studies of the model involved. Changes around the first point are perhaps more influenced by absorp- tion per se, and changes around the last point are perhaps more influenced by forces tending to restore a fasting blood sugar and these reflect other physiologic mechanisms than the processes involved in diabetes itself, so there are other things which influence the points on a curve. DR. FAJANS: There are quite a few points to answer. As far as the flat glucose tolerance test is concerned, it is absolutely normal. Plasma insulin rises as high in the flat glucose tolerance test as in what is thought of as a '"mormal" glucose tolerance test. There is no abnormality to it. It simply means that with the initial rise in blood = dh = sugar, insulin comes out and keeps the blood sugar down in a normal range. If you take such a glucose tolerance curve where the half-hour blood sugar value is the same as the fasting one and if you had obtained a sample at five or ten minutes, you might have found it was 10 or 20 milligrams above the fasting. So there is no diagnostic difficulty about it, and the explanation is not that glucose wasn't absorbed, but that the usual normal insulin output kept it down; that this is an individual who is quite sensitive to his own insulin. As to there being no differences in glucose tolerance tests with variations in dosages, this interpretation is based entirely on the blood sugar levels obtained after the administration of glucose to normal subjects. And that is the important thing. True, in normal subjects you don't see any difference in the blood sugar response -- let's say a very slight difference, with 50, 75, and 100 grams of glucose. However, if you do plasma insulins you find a considerable difference in the plasma insulin response. As a group the ones who are given 50 grams have lower insulin levels. Samols has reported this, Randle has reported this, that the 50 gram test has a much lower insulin response than the 100 gram test. What does this mean? This means that the larger dose is a greater stimulus to insulin release. Now, we are not interested in doing glucose tolerance tests just in normal individuals. We are interested in applying this to potentially diabetic individuals. We would like to know the findings when different doses are given to mildly diabetic subjects. From all we know the larger dose may be more “ 45 « discriminating than the smaller dose =-- and I am speaking of the oral glucose tolerance test now. As far as the intravenous glucose tolerance test is concerned, it is true what you say about Amatuzio's findings. However, Morehouse has extended this range from one half gram to three grams per kg body weight and finds that it makes a considerable difference in the sensitivity of the test. Incidentally, you mentioned there was no overlap between the diabetics and normals. This depends entirely on how you pick your diabetics. Of course, if you pick diabetics who are obviously diabetic by some other criterion and then apply the intravenous glucose tolerance test to those individuals, you will find you have a very sharp delineation. However, if you use mildly diabetic subjects, you will get overlap. Morehouse published overlap between normals and individuals he called diabetic on the basis of the oral glucose tolerance test but when he used a larger intravenous dose the overlap disappeared. So I did not mean to imply that the intravenous test is necessarily less sensitive than the oral one. I merely stated that the test with the usual dose used, 25 grams, is less sensitive than the 100 gram oral glucose tolerance curve. I think what I just said about a good dividing line probably held true in Conard's studies, that he used clearly established diabetic patients. Now I anticipated somebody might say, "There is a paradox in what you said". Actually, I don't think there is, because I think you have to think in terms of two factors: number one, even when giving a small dose of glucose intravenously, you raise the blood sugar very rapidly to a - 40 =~ much higher point than if you had given the same dose orally. We know that in a mildly diabetic patient a high blood sugar level can cause insulin release. That 1s one reason I think this test is less sensitive. Another reason is that with a smaller dose you have a smaller amount to dispose of. If you had a larger amount to dispose of, in spite of the fact you bring your blood sugar to an abnormally high level you may bring out abnormal glucose tolerance. DR. BURCH: I quite agree it would be wonderful to keep a record of every single one of these points on the score for all umpteen thousand individuals who are surveyed, studied in population surveys over the next twenty years, but I cannot but wonder whether it wouldn't need something of the size of the Library of Medicine to keep these records. I would like to toss out another aspect concerning diagnostic testing for diabetes, and that is: In a way there are two rather fundamentally different types of problems in population studies, two major types of population studies which are conducted, not only in this country but elsewhere in the world. One includes those population studies which are conducted at clinics, conducted with a complete laboratory at their beck and call, those studies in which whatever they want to do they can do. But there are those studies done not only in this country, but to a great extent elsewhere, where there are not all these diagnostic facilities available, and what is more, on populations in which they do not have the same incentive to cooperate and participate that many of the population groups which have been discussed here have had. « 47 = For instance, in the program that was suggested over here a while ago, it made my arm ache just to think about all those sticks, all those needles, all those blood specimens. And this is, of course, what you get into in field studies, not only in this country but elsewhere any time you go starting to plan taking seven blood samples, or four, or even three. In fact, oftentimes on population studies, you are lucky to get one. And this is the situation, I think, more so in the, shall we say, hinterlands, not only in this country but elsewhere. This being the case and if we are going to be trying to arrive at standardized procedures which could have wide application, it seems to me that the problem of diagnosing diabetes on population surveys where you are lucky to get one stick at them should be considered. DR, KLIMT: Do you now consider multi-stage screening? Do you think that is best in situations like this? DR, BURCH: Then you are thereby suggesting that if you find a positive urine, that then this individual would have the incentive to allow himself to be stuck. This is, of course, frequently the case, and certainly is the case, I am sure, in a high proportion of the individuals in Washington, D.C., but I don't really know because I haven't actually conducted such a survey. But those of you who have: What proportion of people who are informed that their urine test has been positive will actually come back? DR. ACHESON: I can answer that question as far as the reaction of the British is concerned. In the Bedford study in England an adult population of about 35,000 people was alerted by a good deal of propaganda over a - 48 - period of weeks that a two-week diabetes drive was to be instituted. The way the drive was done was to deliver at the door of each house small plastic pots similar to the one you got your marmalade in this morning. On the 1id of each was a label on which the person wrote his name, age, sex, and whether or not he knew that he had diabetes. The people were told to pass their urine -- I don't remember whether one hour or two hours after a meal and put a sample in their pot. Then the next day the pots were collected. Similar drives were repeated intermittently over a ten-day period to obtain specimens from absentees and laggards. Sixty per cent of the population returned full pots. The urines were all tested by the research team, not the respondent, and urines were then divided into those that were positive and negative. All persons passing positive urine were asked to come for a one-hour glucose tolerance test; also invited were a random sample of the non- cooperators. In both cases we got about 60% of the cooperators, and epidemiologically that doesn't mean much. The blood was taken from the ear lobe, a small prick, done in such a way as to cause the minimum incon- venience and pain. In general terms, even in getting urine, to obtain anything over 60% you need to beat the population around very hard indeed. DR, SHARKEY: I can give some information on this. We tested at one time 69,159 persons postprandially. We had glycosuria on 2,991 and were able to get blood sugars on 2,211. That took us about six months and we had to go after these people to get this done. We wouldn't have had these figures at all if we hadn't pursued it. MR. GORDON: What kind of group was this? DR. SHARKEY: A population study in Dayton, Ohio. - 49 = DR. BURCH: Dr. Sharkey mentioned another point there which I think is extremely important. It took them six months of very hard work to get that, and oftentimes, particularly on field surveys, in contra- distinction to the clinic-based studies, you aren't going to get your staff to stick out there in the field any six months to do a study like that. So, in order to be able to complete some of these population studies out in the field -- this is away from the centers of civiliza- tion, which is somewhat of a different situation -- you need to have your surveys which last a very much shorter interval of time. You cannot afford, then, to do anything which will drive your patients away. Hence, I would feel that on a population survey of that nature, it would be very important to try to devise a meaningful test which could be based on only one blood sample. STUDIES OF DIABETES WITHIN COMMUNITIES GERALD T. KENT, M.D.> Some of this material will be repetitious for those of you who attended the International Congress of Diabetes in Toronto. Diabetes mellitus, a hereditary defect in carbohydrate metabolism is a widespread symptom complex which disables a large segment of the population with illness and ranks as the seventh leading cause of death. Death certificates confuse the statisitcs because this method of reporting emphasizes the end result of diabetes, such as coronary thrombosis and other vascular diseases rather than diabetes per se. Methods for mass detection of this disease have been inadequate. The standard glucose tolerance test has been used to obtain the classical diagnosis. However, this method is too cumbersome for large-scale detection. A technical problem of corraling and feeding to large numbers of people a glucose solution which provokes nausea is alone enough to discourage most detection efforts, When the venipuncture problem is added, this has usually proved too formidable for most mass detection efforts. Ideally, diabetes should be detected in the prediabetic stage. This would require genetic detection methods, at present not available. Next best would be detection in the latent diabetic stage, that is, before sugar spill occurs but the glucose tolerance test is abnormal. Once early diagnosis has been accomplished, it is hoped observation, study, research, and early treatment would give some of the answers so zealously sought. Siestern Reserve University School of Medicine, Cleveland, Ohio « BY = Many questions have been raised and some are as follows: (1) Does early treatment prevent or delay the onset of secondary complications of the disease, or lessen the intensity of the disease process itself? (2) Does early identification help the physician in planning the health program of his patients to ward off other associated diseases? It is apparent that once severe vascular disease has developed, little can be done to mitigate it. However, no one has yet shown what benefits treat- ment started at a much earlier stage may yield. Even weight reduction alone can account for a marked increase in carbohydrate tolerance. (3) Does the use of oral hypoglycemic agents and insulin in addition to dietary management in the early stage of the disease alter the course of the disease? (4) Can the propagation of diabetes be curtailed or prevented by early tdentification and by strong recommendation against marriage between diabetics and/or carriers? If the latent diabetic patient is discovered early certain diabetogenic factors may be modified, as follows: (1) Obesity: This can be modified or corrected. (2) Incorrect diet: Heavy carbohydrate intake can be modified and dispersed throughout the twenty-four hour period. (3) Pregnancy: Birth control measures are available. (4) Hyperthyroidism: This can be corrected. (5) Diabetogenic Drugs: For example, the thiazides and steroids should be used sparingly, if at all. Other diabetogenic factors as yet cannot be modified: 1. Genetic factors; 2. Insulinase; 3. Insulin antagonists, such as antibody develop- ment against insulin protein. The purpose of this paper is to present a method for detection of latent or early diabetes which is relatively inexpensive, rapid, and - 82 applicable to mass screening, and to report the actual application in a metropolitan area. The method consists of: (1) Filling out a card with the pertinent vital statistics. (2) The oral administration of a new carbohydrate loading solution. (3) Analysis of the glucose level of finger blood two hours after the ingestion of the carbohydrate load for preliminary screening purposes. (4) The performance of a confirma- tory, two-hour, carbohydrate tolerance test on all subjects with a screening blood glucose level above 140 mg. per 100 ml. (5) Referral to his personal physician for follow-up and treatment. I would like to take you step by step through the procedure. The screenee reports to the testing station and fills out a data punch card with the pertinent vital statistics. At the same time he is given the carbohydrate loading solution, which has the following formula: glucose, 30%; maltose, 18%; maltotriose, 13%; higher saccharides, 39%, with kola or cherry flavor. It is equivalent to 75 grams of glucose in a volume of 7 ounces, which is usually carbonated. It can be dispensed in one of two forms. First, if the screenee is to drink it on the spot, the solution is dispensed with an ordinary dispenser used in any soda fountain. If the screenee is to take the solution home to consume, or consume it in his work area, it is dispensed in a bottle. This loading solution is unique in that it is refreshing, palatable, carbonated, served cold, and can be drunk by anyone in as little as 10 seconds. It has been administered without nausea to children and pregnant women. It has been given to over 70,000 people without a single incident of vomiting. w 53 w This solution is rapidly hydrolyzed to glucose in the gastrointestinal tract. This new 75 gram preparation of carbohydrate yields blood glucose curves which for practical purposes are identical to those obtained after 100 grams of glucose. Two hours after drinking the "CL" solution, the patient presents himself for finger blood testing. In cold weather it is sometimes difficult to obtain finger blood. In this event a simple, aluminum, heated block finger warmer is used which aids greatly in obtaining blood. In actual practice, 0.075 ml. (75a) of finger blood is drawn into a disposable but accurate capillary micro pipette and diluted with 1.2 ml. of a 1% solution of sodium fluoride in a plastic vial. This vial fits into the sample plate of a modified Autobnalyzers. This machine has been modified by making all the parts smaller so that it is relatively portable. A ferricyanide reduction method is used and results are available seven minutes later. Each machine analyzes 40 samples per hour, and several machines may be used, depending upon the number to be tested. A recent innovation in this detection program has been the use of an impregnated paper test strip (Dextrostix B 4, which can be used to test blood for glucose. The merit of this test is that it can be used as an exclusion procedure. By setting the screening level low, it is possible to identify the obvious negatives. They can be informed and eliminated immediately. Two thousand serial comparisons using the paper test strip and the glucose analyzer have been made. Only one individual screened below the Syechnicon Instruments Corporation, Chauncey, New York 4AMES Company, Inc., Elkhart, Indiana - 54 exclusion level by the paper test strip and above the glucose analyzer screening level of 140 mg. per 100 ml. By using the paper test strip, it is possible to eliminate 75 to 80% of the people to be processed by the glucose analyzer, thus increasing the capacity of this apparatus. In our experience, it is not possible to use the paper test strips alone as a detection program. Too many errors would be made if exact screening were attempted by this method. No attempt is made to keep the patients inactive during the two-hour interval, between the ingestion of the carbohydrate load and the capillary blood test. Notification of a positive test is made by phone or mail and is confidential. Screenees with an abnormal glucose level are given a confirmatory glucose tolerance test in the Cleveland Diabetes Headquarters on Saturday, per appointment. The criteria for an abnormal test are any one of the following: (1) Above 120 mg per 100 ml. fasting; (2) Above 190 mg. per 100 ml. one hour after the carbohydrate load; (3) Above 140 mg. per 100 ml. two hours after the carbohydrate load. The level at one hour may seem very high, but remember these levels are after a carbohydrate load and are on capillary blood. The confirmatory test is important because it identifies the person with an abnormal tolerance curve with very little chance of error. In the original screening the person may have eaten between the 'CL" load and the finger blood test, which would give a false positive. With the report of a positive confirmatory test, the personal physician is sent a brief questionnaire which requests confirmation of the diagnosis, - 55 - report of physical signs of diabetes, treatment instituted, and any other pertinent information. The success of a screening program depends upon the cooperation of the community. Two examples of the testing program are as follows: Industrial program: Cooperation of industry is achieved by acquaint- ing a top official in a company with the program. Business leaders are very conscious of the benefits to their employees and, therefore, are most enthusiastic in supporting this project. Usually the personnel officer, plant physician and our team arrange the exact scheduling for each plant. Distribution of the "CL'" varies from plant to plant, but finger blood is usually obtained in the most central location in the factory. It was found that time could be saved by collecting the blood samples in vials and running the determinations in the central laboratory. This markedly enhanced both the accuracy and the efficiency of this operation. Residential program: This is carried out in a variety of ways, but the most productive is to station the mobile laboratory in a shopping center, or transportation concentration point. When utilization of the local communication media has been directed by professional advertising agencies, the screenees' attendance has been excellent. The mobile laboratories are excellent for processing persons at the rate of 1,000 per week, but when the number reaches 800 to 1,000 per day, they are inefficient, because of inadequate space. We have screened over 70,000 people in the past nine months. The results of the completed pilot study involving 8,790 people are summarized in Table VI. - 56 - The figures presented are not necessarily representative of the total population, because they reflect a motivated group, in that it contains many obese persons and many with a family history of diabetes. In Table VI, the highest number of positives are in the third and fourth decades of age. This corresponds to the industrial population in which the majority of workers are in these age groups. Obviously, all comparisons between studies must take into account the age of the screenees. The prevalence of abnormal glucose tolerance tests increase with each suc- cessive decade. Of the 709 positive screenees, 560 had confirmatory glucose tolerance tests done and 407 of the 560 had abnormal glucose tolerance tests by the above-mentioned standards, for a total over-all prevalence of 4.5%. In effect, these persons have had two abnormal glucose tolerance tests, separated in time, and thus they are highly suspect of having diabetes. The heading "Glucose Tolerance Tests Not Done" indicates that a sub- stantial number of screenees did not return for a confirmatory test, for one reason or another. If the rate of positive glucose tolerance tests - is projected to include these figures, it markedly increases the positive glucose tolerance test figures. In 1961, a careful statistical study was carried out for the Diabetes Association of Greater Cleveland by the Statistical Department of The Welfare Federation of Cleveland, to determine the utilization of the Clinistix’ method of diabetes detection in South Euclid, a city of 30,000 people. No carbohydrate loading was done. SAMES Company, Inc., Elkhart, Indiana - 57 w Results are compared in a crude way with our present results, in the same city, in Table VII. These figures show that twelve times as many people were identified by the finger blood survey, and since each study attempted to exclude known diabetes, this comparison suggests that the finger blood loading technique is more productive in finding new cases. DISCUSSION DR. ALPERT: Does the final positive percent in your table represent the result with the glucose tolerance test? DR. KENT: Yes. We never identify anybody on our finger blood test without two positive tolerance tests. DR. ALPERT: The statistical estimate by the urine test is closer to what we know or think we know than by the finger blood test. How do you explain that? DR. KENT: I indicated that the pilot study was probably a motivated group and thus contained more people with diabetes than a random sample. There were many obese people and those with a family history of diabetes. However, in the industrial testing where we have a captive group we still have a very high prevalence, so I think we are finding more diabetes with this method. DR. ALPERT: If you are going to answer it that way, then you mean the incidence of diabetes is higher than we think it is. DR, KENT: This study gives indication that the prevalence of diabetes is far greater than any previous estimate. I would point out that we have chosen a screening level which in years to come may be thought to be too high. But, we have purposely done this to avoid identifying borderline * 58 w cases that less sophisticated scientists would question. It is most important for us to retain the goodwill of our cooperative groups. DR, WEST: Does the 6% assume that the incidence in the non-reporters is zero? I take it that the people that screened positive and never came in, were not included -- DR. KENT: They are not included in the figures. DR, WEST: So if incidence in this subgroup was the same as in those who came back the incidence might be 7% or 8%? DR, KENT: Exactly right. Allow me to do a little extrapolating. As of September 1, 1964, 69,514 people were tested and 3,094 screened positive. Glucose tolerance tests were done on 1,686, a little better than one-half. We will probably get another 25% of these eventually. Of these 1,686 second glucose tolerance tests, approximately 1,200 were positive. If we assume this same prevalence in the other 50% that have not reported for a glucose tolerance test, there would be 2,400 positive glucose tolerance tests, and we have tried our best to exclude patients with known diabetes. In past studies, the final diagnosis has been left to the judgment of the family doctor. I doubt that this is the ideal way to do this, On occasion the confirmatory tests of the patients are performed by having the patient eat a large meal (maybe all protein) and a two-hour, postprandial blood sugar shows no abnormality, so the detection glucose tolerance tests are sald to be erroneous. Gradually this problem is being eliminated by having the patients loaded with a standard amount of glucose. - 50 « I think the use of two glucose tolerance tests, separated by an interval of time, is much more accurate in determining whether a patient is a diabetic than a loosely controlled confirmatory effort by large numbers of individual physicians. Criticism has been directed at us for not using Public Health nurses for follow-up. Actually we have guaranteed the confidentiality of our detection efforts. If we send a Public Health nurse out to the patient's home, we are to some degree negating our pledge that no one except the patient and his doctor will be told the results of his test. Since diabetes is not infectious nor a contagious public health menace, we question the advisability of this approach at the present time. We are trying to promote a detection program for diabetes, educate the public and the physician population in our community. I think if we go gently along, we can identify many cases of diabetes and maintain our happy relationship with the public and our colleagues in the field of medicine. DR. BURCH: Those two glucose tolerance tests you do are exactly the same? DR. KENT: Not exactly, Dr. Burch. The first glucose tolerance test is not necessarily fasting, it is done by administering 75 grams of glucose, followed by a finger blood analysis two hours later. The second glucose tolerance test is performed on a fasting individual, a fasting blood is determined, a 75 gram glucose load is applied, and a one-hour specimen of finger blood is tested and a two-hour specimen of finger blood is tested. We intend to eliminate the fasting blood sugar, as it has not proved of significance. - BO = DR. BURCH: With the same drink, the same finger prick? DR. KENT: Yes, it is identical. DR. ROOT: All these are capillary bloods? DR. KENT: All of them. Not a single venous blood. DR. BURCH: Other than detecting cases, what uses are contemplated for all this information? DR. KENT: This all goes on sfonde and I have sheets of summary data we have taken off which are cursory but interesting. The question has been raised as to whether there is a higher incidence of diabetes in Jews than in other religions. One figure showed the total Catholic population was 44.67% of the total tested and of those with diabetes 32.27% were Catholic. The Protestants were 41.2% of the total and they had 41.2% of the diabetes prevalence; the Jews had only 14% of the population but 26% of the diabetics. DR, KLIMT: Are these age standardized? DR. KENT: No. DR. KLIMI: But it may be that the Jewish population has a different age than the Protestant. DR. KENT: It can be determined but we haven't done it. DR. KLIMT: Why did you choose the two-hour screening? DR. KENT: Partly on convention. The people with known diabetes all had blood levels above 140 mg. at two hours. Our medical student controls at two hours were well below 140 mg. At one hour, two of our medical students had levels the same as some of our diabetic patients. They may be diabetic. btenationsl Business Machine Corporation, Armonk, New York w OL = However, we were afraid to identify people on this basis. We felt 140 mg. was a pretty safe area to work in. DR, KLIMI: But by all standards I see, the one hour is the more potent screening. DR. KENT: I think this might be so, but again I come back to what we are trying to do, screen the population, not scare them out of their wits. If we scare them, we will cease to test them, So we should not identify too many that we can't prove scientifically. You have to take these things into consideration. DR. MILLER: Let me give you my own personal views. I think the merit of what Dr. Leonards and Dr. Kent have done is clear. They have been able to use the best methods for determining glucose tolerance that are used in hospital practice and applied to large numbers of patients in the communities, I think the fact they have done this large number indicates their success. I think Dr. Kent was wise in asking not to be looked at critically in so far as epidemiologic and statistical questions. That was not the purpose of this study. They have studied large numbers of patients not set up with any experimental designs or to answer the questions which have been bandied about today. I think, as Dr. Lazarow sort of hinted at, with this large amount of information as you look over it, you may find several nuggets which will give interesting ideas and will perhaps suggest or stimulate this group or other groups to use this method which has been so beautifully worked out, and answer these questions specifically. So I think this is a value of it. And also the questions that have been asked bring out the important problem today that Dr. Lazarow has touched upon. Actually, we cannot - 62 - expect Dr. Kent and Dr. Leonards to come up with an answer as to which blood sugar level is the best or at what time you take it or at what dose. This is the thimg Dr. Lazarow touched upon. We do not have the data today to tell us specifically which is the optimum or the right method that we should use. I think this is another question entirely. DR, MILMORE: TI would like to make a comment on the Public Health nurse situation. I was a little surprised to find that in Cleveland a Public Health nurse visit is regarded like a police department visit. I don't think it is universal throughout the country. DR. KENT: I do not wish to impugn the reputation of the Cleveland Public Health Nurse. We can use them successfully, but one of the biggest things we have encountered is the attitude of the labor union. Their first question is: '"Is management going to get this man's test results? If so, we are not interested.'" Therefore, we must convince the labor leader that the results of the tests are absolutely confidentigl, The first time we break this confidence by sending someone out to his house, the worker's interpretation is that this is not confidential. It isn't because the public health nurse does not handle this well, it is that her presence may be interpreted as a breach of confidence. So far, we have been completely able to gain the confidence of labor, and this has been a great help in forwarding our program. We have had over 80% voluntary participation in the major industrial operations in the Cleveland area. There has been excellent cooperation between labor and management, and we would like to maintain this esprit de corps. The screenee is told that if he has a positive screen, he will be called at home, not at work, or the letter will be sent home, - 63 - not to his working address. Our follow-up glucose tolerance test, the second one, is done, not at the business headquarters, but in the diabetes headquarters, which is geographically remote from his place of work. DR, BURCH: It seems to me you are treating this like Bubonic Plague or something. DR, KENT: No, we haven't. None of the people are very frightened after we have talked to them. DR. LAZAROW: Dr. Kent mentioned there were quite a few retests on the part of the doctor that didn't coincide with your tests. How do the tests that Max Miller may have done in his unit agree with yours? Do you have pretty good coincidence there? DR, MILMORE: I think he is asking about the difference in the way this test is done in the business and the way the doctor does it in his office. DR. LAZAROW: What I am wondering is: I think you have suggested, and I think rightly so, that the negative ones the doctors get are probably not as well done as yours. My question is: Do you have data from a university hospital? DR. KENT: We have no comparative tests done in university hospitals. We have a few negatives after our two positive glucose tolerance tests, and I think this supports the work of Dr. McDonald and others who have done serial glucose tolerance tests on a stable population and have found intraindividual differences, probably because some people have just begun to show abnormalities in their glucose tolerance tests - some weeks they will be normal, some weeks they will be abnormal, Eventually, I - Gb w= believe these tests will all become abnormal in these individuals. DR. LAZAROW: Why is there the difference? DR, LEONARDS: I don't think we have any way of knowing why, when we get two poeitive tests, a subsequent one comes out negative. DR, ROOT: Isn't it true that some of the patients went to other doctors, not to Dr. Leonards, and had regular glucose tolerance tests done by the venous method and these were negative? DR. KENT: Yes, that is what I am talking about. DR. LAZAROW: What I am really trying to assess is: Can you at this point sort out how much is due just to the fact that you have cyclical varia- tions in the individual, and how much to the difference between the finger blood method and the venipuncture? DR, KENT: I don't think we can tell you the answer to that. I will give you one example. We had a symposium and one of the doctors in town came to me and said, "Your test is no good." I said, "What happened?" He said, "You sent me six patients and they are all negative." I said, "What did you do?" He said, "I ran a blood sugar." I said, "Did you go to the lab?" He said, "I run my own and take the blood down to the lab." And I said, "How did you load them?" He loaded them with a big meal, I said, "Will you do me one favor. Will you let me supply you with these bottles of glucose." So I got Dr. Leonards to send him the bottles and five out of six were positive. I present that as an example of how we get crucified without justi- fication. DR, MILLER: Let me emphasize that the pragmatic approach has been used here. It is to how that you can do it with the equipment we use in all university - 65 - hospitals, using an Sutchnulyzer,” and that is essentially what you get in the research laboratories. You are giving a standard dose of glucose and getting it at times which are accepted in practices. You are doing this in large numbers of cases. The questions you are asking are terribly pertinent, which we have so carefully avoided in this discussion: What is the variability in subjects in whom glucose tolerance tests are done time after time? And of course, which ones are diabetic? I don't think we are going to answer that today because it is going to take long-term studies to see which ones develop clinical diabetes and, I think we will have to say symptomatic diabetes or diabetes with vascular complications which we might agree are unique for this disease. As for the others, the people with abnormal glucose tolerance tests, is this a disease or is it not? I personally am not sure that it is normal for glucose tolerance tests to get worse with age. It reminds me of the error that arose when we said that obesity was a sequela of age. It was only when the insurance companies looked at the figures and found it wasn't good to weigh more at 60 than you did at 20 that a question was raised. I would raise the question as to whether it is perfectly normal to have this increase in blood glucose. And I would like to know, if we have 100 subjects 20 years of age, whose average blood sugar rises with age, do these all rise the same or are there some of that cohort that are becoming abnormal, which changes the average? Because I know there are some people who are 80 years old who have normal glucose tolerance tests. 7 Technicon Instruments Corporation, Chauncey, New York «- 66 ~ DR. ANDRES: I wish to address myself to the problem raised by the change in glucose tolerance with increasing age. We must face the fact that if we apply the generally accepted standards of normality for performance or. glucose tolerance tests to older subjects, then we will be forced to call close to 50% "diabetic". (Figure 5). This Figure illustrates the results which are rep: esentative of those obtained when any of a variety of glucose tolerance tests are performed on a population group covering the entire adult age span. This Figure happens to be results of cortisone glucose tolerance tests done essentially by the Fajans and Conn technique. The relation between age of subject and blood glucose concentration two hours after glucose ingestion is shown. You will note the tendency for glucose concentration to increase with age. If you will imagine a horizontal line running across the figure at the 140 mg per 100 ml level, you will see that by the later years of life actually more than half of the subjects exceed that concentration. I emphasize again that this type of result applies to intravenous and oral glucose tolerance tests as well as to the cortisone test. Now one has a choice: Either one says that the performance of an individual, regardless of his age, on glucose tolerance tests should be judged against that of a group of young adults, or else one says that an individual should be measured against his age cohorts. Current practice unquestionably favors the first choice. One will look in vain for age-adjusted standards for any of the diagnostic tests for diabetes in text books of medicine, endocrinology, diabetes, clinical laboratory Bet aiis are presented in paper by Pozefsky, Langs, Colker, and Andres; Annals of Internal Medicine, in press. - BF = diagnosis, or in house-staff handbooks. What do we suggest? ‘Figure 6 shows one way of putting the type of data shown previously into what we believe to be an interesting and practical form. A nomogram has been constructed to permit rapid deter- mination of an individual's performance on the cortisone glucose tolerance test in comparison to his age cohorts. Knowing the age of the subject and his glucose concentration at two hours, a straight edge connecting these two points on the appropriate scales will intersect the slanting third scale at the percentile rank of that individual. The rank line slants because of the increasing variarce of test results with increasing age as shown in the first slide. Now, where is the "upper limit or normal on this nomogram? Well obviously one can put a cut-off point wherever one wishes on the Percentile Rank line. Does one wish to have 6% of the population diabetic? Very well, the point goes there. Is the normal range to be the mean plus two standard deviations? Then the cut-off point is approximately at the 2.5% rank. Perhpas this nomogram won't help too much, but I believe it will help some. It cannot make the decision as to what percent of the popula- tion at different ages is abnormal. But it does enable one to get an age-adjusted rank for a subject in a very simple way in a few seconds. It is obvious, I believe, to all of us that longitudinal studies are essential, With follow-up data we will be able to look back and find out which percentile rank carries with it implications concerning the well-being of the subject. Suppose one looks at the development of some of the clinical mani- festations of the overt diabetic state. Indeed, one could examine the - BE = relation between percentile rank at some time in the past and mortality rate. Consider a plot of percentile rank (abscissa) against mortality rate (ordinate). Perhaps we should imagine a family of lines, one line for each decade of life, We might expect to see lines which are hori- zontal for part of their course, but which, at certain critical percen- tile rankings, begin to course upwards. I suggest that you may find the sort of nomogram that I showed useful. If technicians in clinical chemistry laboratories had at their sides a nomogram, and if it were required that the age of the subject be given with each glucose tolerance test, then not only could the actual blood glucose concentrations be reported to the physician but also an "interpretation", if you will, in the form of a percentile ranking. The epidemiologist might find the device useful as well. MR, WOOLSEY: Is it implicit in the use of this nomogram that it is normal for the parameter to increase with age? DR, ANDRES: You can place your cut-off point anywhere you wish. If you wish to place it at 2%% for all subjects regardless of age, that can be done; or if you wish to allow the percentage of abnormal results to increase with age, that can also be done. MR. WOOLSEY: What you are doing in effect is measuring deviations from a trend line with age? DR. ANDRES: That is right. MR, WOOLSEY: So you are holding age fixed in this thing. And it seems to me that it is implicit in this that you are saying that it is normal for it to increase in age, because if it went up with age but stayed on that curve you would still consider -- - 69 - DR, ANDRES: No. I am really not saying that. MR. WOOLSEY: I don't see why you are not. It seems to me you are. DR. ANDRES: You, I think, are inferring this. You see, if I say that blood glucose concentrations fall in a band when plotted against age, I am simply presenting the data. There are no necessary implications con- cerning "normal limits". The same is true of the nomogram. Individuals can in fact be ranked according to their age and their glucose concen- trations. If you wish you can say that only one-tenth -- MR, WOOLSEY: You are measuring deviations away from the trend line. DR. LAZAROW: Could you give the corresponding percentile ranks so we can get a better picture of how it works? DR. ANDRES: Let us consider the plot of glucose concentration versus age again. I will remind you again that this is a cortisone GIT. In that figure you will notice two subjects in their late 30's with glucose concentrations of 175-180 mg per 100 ml. You will also notice two subjects about 70 years of age with glucose concentrations of 200 mg. per 100 ml. Now in a sense the 70 year old subjects have performed more poorly than the two subjects in their 30's - after all their glucose concentrations at two hours are higher. It is equally clear, however, that the younger two subjects are relatively poor performers in comparison to their age cohorts while the two older subjects are simply average performers for their age. DR. LAZAROW: Can you take a couple of points on that graph and say, "This is number two because of why?" Or "this is number 60 because of why?" DR. ANDRES: Every point that falls on the heavy unbroken line has a per- centile rank of 50 on the nomogram. Every point that falls one standard J deviation above the mean line would have a rank of about 18 (regardless of age) and those falling one standard deviation below the mean line would rank at about 82. DR, LAZAROW: Okay. DR. ANDRES: Every point falling two standard deviations above the mean has a rank of about 2.5, and so forth. DR, NEWILL: Is the variation constant with age? DR, ANDRES: It is in my small group of about 100 subjects in the intra- venous glucose tolerance test, but variance increased with age in the cortisone glucose tolerance test series. MRS, FISHER: We did this on a population of 15,000 people and did get an increase in age. In the older age groups, people may still have, in some cases, very low level of glucose but many show levels much higher. The mean increases quite dramatically with age. DR. ANDRES: The band of results plotted against age would then be diverging lines rather than parallel. Is that what you are saying? MRS. FISHER: Yes. DR. ANDRES: The construction of the nomogram would be modified depending upon whether variance increases with age or is constant. In the case of the intravenous glucose tolerance test where variance does not increase with age, the Percentile Rank line is a vertical line parallel to the Age and Glucose Concentration lines, in contrast to the sloping line shown in the nomogram for the cortisone glucose tolerance test. MRS, FISHER: What would that mean in terms of a disease which becomes more apparent with increasing age and stresses and strains? You could assume that a young person, if he is abnormal, may just be showing very low-grade - TL abnormality in his group, but through the years this abnormality will become more apparent. Wouldn't this happen to your regression because the disease itself only becomes apparent with age? In other words, can you assume that mean levels are normal? DR. ANDRES: You have put your finger on the tough question. MR. WOOLSEY: This is exactly the question that worries me because {it doesn't seem to me it fits with the model Dr. Fajans was describing, that it is a genetically determined disease and as you go through life a certain proportion of people, larger and larger as you get older, exhibit signs of it. To measure the deviation away from the age trend line doesn't seem to me to be consistent with that model. At least, conceptually -- DR. ANDRES: You prefer a 60% incidence of diabetes? MR. WOOLSEY: No. We might need to have more sophisticated methods than just a flat level across there. I am oversimplifying this because I don't fully understand it -- I am not sure anybody else does, either -- but if you are born with a certain genetic predisposition to the disease, and in your lifetime a certain fraction of people will show up with signs of the disease, then it seems to me it is quite possible as you go through age you might find a larger and larger proportion who exhibited signs by the time they reached that age. In other words, it would be in accordance with the model to have an increasing prevalence as you grow older. DR. ANDRES: I couldn't agree more. And I come back to what Dr. Fajans and others have stressed today - ultimately you need some technique for diagnos- ing diabetes other than the glucose tolerance test. But we have to diagnose diabetes in 1964 and we are going to do epidemiology studies now. Current “| 72 = textbooks tell us that diabetes reaches a maximum prevalence of 6% in the older age group. If you are happy with a 6% prevalence -- MR. WOOLSEY: Whether one is happy with it or not, I don't think particularly matters, DR. ANDRES: But let me show you what is commonly done in practice. For the older age group, instead of accepting a glucose concentration of 120 mg. per 100 ml, at two hours for the oral glucose tolerance test as the upper limit of normal, we raise the limit to 140 or 160. And the only reason we do it is for the comfort of the people who are handling the diabetic patients. You can put that level anywhere you want it and raise or lower the prevalence of diabetes quite arbitrarily. MR. WOOLSEY: It is like my son when he comes home from school and says he got a 40 in a Math test but thinks the teacher is going to "scale it", which means that everybody else got a 30, and therefore he is doing all right. DR. ANDRES: Let me just say what is done. The usual 120 mg. per 100 ml. level is raised to 160 or so for old subjects and the prevalence of abnormal curves is thus lowered from an uncomfortable 50-60% to a more tolerable 5-6%. DR, MILLER: That is an arbitrary assumption. DR. ANDRES: I am not recommending that procedure. I am just describing the situation as it exists today. DR. BURCH: Then you are indicating ten different normal values, one for each decade. DR, ANDRES: I hope I didn't do that. I don't want to be forced into making use of the nomogram in ways which are at present premature. «73 = DR. FAJANS: I think the criteria which Dr. Kent used are very strict ones and are similar to the ones we use for venous blood. He used 140 mg. per 100 ml. at two hours as the cut-off point, which corresponds roughly to 120 mg per 100 ml. on venous blood. I think we all have realized in recent years that there is an increase in blood sugar levels with advancing age. However, I do not think we should have the pendulum swing all the way to the other side. I think we shouldn't forget, as was just pointed out, that the clinical manifestations of diabetes occur with increasing frequency in an older age group and therefore we should expect an increase in incidence of latent asymptomatic diabetes in this age group also. Let's look at Dr. Kent's statistics. There is no evidence of a 60% incidence of tests above his dividing line in an older age group. At the age of 40 to 49 there was a 3% incidence. At the age of 50 to 59 there was a 6.5% incidence. I think these are very conservative figures. Only with age 60 to 69 do we get a 16% incidence, and with age 70 to 79 we get a 22% incidence. This isn't so terribly high - I am not saying we are ready to accept this as the prevalence of diabetes, but let's not exaggerate and call it a 60% prevalence of "abnormal" glucose tolerance tests. And then the question comes up: What is diabetes? I think it would be wonderful if we would have an independent criterion other than blood sugar. But to require, as Dr. Miller has suggested, the presence of vascular disease before one can make a diagnosis of diabetes or to require symptomatic diabetes to make a diagnosis of diabetes, I don't think will get us any closer to the problem. I think this is moving too - 74 = far into the advanced stages of the disease to require these end stages for the definite diagnosis of diabetes. And lastly, I was very much interested in the discussion of dosage of administered glucose. Dr. Kent showed a comparison in ten normal young subjects of glucose tolerance tests performed with 100 grams of glucose and glucola. With 100 grams of glucose the peak blood level was very close to 120 mg per 100 ml, This is capillary blood. With the glucola it was about 10 milligrams below that, or something like this. DR. KENT: Jack, was that venous or capillary blood? DR. LEONARDS: This is on venous blood. DR. FAJANS: Even better. It is about 120 mg. per 100 ml. for the 100 grams of glucose and 110 or 115 for the glucola. Well, using our load, 1.75 grams per kilogram body weight which as I said before will come to 130 grams of glucose in the male, average peak blood sugar level will come to 130 mg. per 100 ml. So here is an illustration of differences when different loads are administered. In a blood sugar range of 100 to 125 mg. per 100 ml., this difference may be 5 to 15 mg. per 100 ml. Whether this is significant or not from a practical point of view, I don't know. The only point I am trying to make is that is does make a difference what load you give. DR. KLIMI: TI think we are falling into a common trap. I think Dr. Kent has specifically disclaimed the use of these data as a demographic statis- tical basis. We are discussing prevalence figures which we shouldn't. These are volunteers who have come in. It has been specifically stated "4 - 75 - that there was a concentration of people who suspected themselves that they had diabetes. So it is really out of order to discuss these as prevalence figures. They are not prevalence figures of a natural popu- lation. DR, ACHESON: Dr. Kent is inclined to interpret them a little bit that way. I wondered whether he was falling into his own trap. You remember he made interracial comparisons and commented on them. I do think he is doing a magnificent screening job but I do hope he resists the temptation to fall into that trap. DR, KENT: Our population now is heavily weighted in the group from 20 to 50, because we are testing so many workers in industry, and we still have a very high prevalence of diabetes. This population should have a much lower prevalence by previous estimates. DR, O'SULLIVAN: Before we discount the 6% figure, we should look at some valid epidemiologic studies. The National Health Survey sample used one-hour blood sugars following 50 grams of glucose, It suggested a very high prevalence rate. I realize that we.could juggle this figure by changing the criteria, but it was still very high in the National Health Survey as compared with other population studies. Now, all the studies that have been done on postprandial blood sugars, including Oxford, have given the rates about 2%. What we are beginning to witness is that when we do studies involving blood sugars following a glucose challenge we are getting a much higher prevalence rate. There is a lot of supportive evidence for this. Look at the current studies being carried out in England, for example, the Birmingham study, where a screening urine test preceded the Stegrosiic test. The surprise - 70 = was when they took those screening negative, they found a very high pro- portion had abnormal glucose tolerance tests. You can look back through the literature -- the famous food handlers who screeded negative by blood sugar; when Dr. Unger retested them he found a surprising number with abnormal glucose tolerance tests. So we are left with the situation that if you do a population study and use postprandial blood sugars you will get one kind of answer and if you use post-glucose blood sugars you will get another kind of answer. I think if we keep varying the load we will keep getting different answers. What we are trying to learn is: Which of the blood sugars are synonymous with diabetes? We have no trouble clinically when a person has symptomatic diabetes but what about all the people with unsuspected abnormal glucose tolerance curves? Are they diabetic or high-risk persons? There is a difference. I would submit they are in a high-risk category for diabetes. I would guess that if Dr. Kent rechecked his own "diabetics" with his own test that he would find a big variation in his results. Many will become negative. He doesn't have to send them to the local physician to find this out. If he retests them next year, he will find a high percentage will be negative, just as a proportion of the people negative this year will be positive next year. DR, KENT: We do two tests. I don't think anybody should be identified on the basis of a single positive test. But, if they have two positive tests, separated by an interval of a month or more, the odds are high that they will have relatively few negative glucose tolerance tests subsequently. If only one test is positive, then the intraindividual variation will be operative to a devastating degree. - FY iw DR, ACHESON: It is just a question of interpretation. I agree with Dr. O'Sullivan. I think the prevalence is higher than we think. DR. SILVERSTONE: I would like to ask Dr. Kent if he has information on family histories. Is it possible you are attracting people who fear diabetes because it is in their families? DR, KENT: In that one study, yes. In the overall group, no. DR. ALPERT: I hope I am not being impertinent but I would like to raise a practical question. That is, how much does this cost? Can you give us some idea -- in comparison with a type of screening test that you have done in the past, or the type that the American Diabetes Association promulgates, or those, with chest X-rays done by the Health Department? There is some range of cost here, you see. DR. KENT: I think we can give you some idea. We have experimental costs mixed into our overall cost so that it is difficult to give an exact figure. However, if we include our personnel, loading solution expense, apparatus, rent, and overall administration, I think the magnitude of this cost is approximately 50¢ a test. Jack, is this far off? DR. LEONARDS: No, it isn't far off at all. DR. ROOT: Somebody mentioned the mortality rate as being a figure which could possibly be used in some calculation. I think the mortality rate for diabetes is perhaps the most inaccurate figure known to men. I just wonder how you could use that in anything relating to an X-Y-Z new method of detecting diabetes. DR. ANDRES: TI really thought mortality was something we could diagnose with great accuracy. -~ 78 = DR. ROOT: I just say that is the most inaccurate figure known with respect to diabetes. DR. ANDRES: I was just thinking today of what variables were available to us that we could correlate with the results of a glucose tolerance test - that is, what variables that show a high correlation with diabetes. Mortality rate is one such variable. That is, if you could show -- DR. ACHESON: From all causes or diabetes? DR. ANDRES: Let's say all causes to start with. Suppose you found that the overall mortality rate in 50 year old men who had two-hour blood glucose concentrations of 160 mg. per 100 ml. was no higher than in 50 year old men with glucose concentrations of 120. Then we would be justified, on the basis of these mortality data, in saying that the upper limit of normal would not be 120, since a level of 160 carries with it no shortening of life span. I brought up mortality rate then simply as one of the variables that should be examined in a prospective study. DR, ROOT: 25 to 35% of our known diabetic patients never have the word "diabetic" on the death certificate. DR. ANDRES: I really meant this was a study to be carried out by some of the people in the room, perhaps. DR. WADA: TI would like to ask Dr. Kent whether they have measured the concentration of urinary sugar. DR. KENT: No, we have not. We started out with the idea we would try to do urine testing and it didn't take us long to realize this was impossible. If I could take one second, Dr. Leonards and I worked in a partnership in this thing and I want you all to understand very clearly that his con- tribution has been terrific. He developed this new glucose solution which - 79 - I think is the greatest benefit that may come out of this whole program. He also developed little ingenious things like this finger warmer. He also developed this modified glucose analyzer which made it possible for this whole thing to start. When we first started we were going to use candy bars and other obtuse things and it wouldn't have been possible without his genius. It has been through his genius this whole thing has been possible technically. DR. WADA: The reason I asked is that we have been concerned with the con- centration of blood sugar. But when you speak of negative sugar then you should have some measure at which point it becomes negative. DR. KENT: In relation to the kidney filtration level of glucose? DR. WADA: No, the concentration level. At which level does it become negative? What do you mean by that? DR. WEST: Do you define "positive" as one-tenth of one per cent or one hundredth of one per cent? That is what he is talking about. DR. KENT: As far as urinary glucose is concerned? We didn't do any urinary tests in this program. We intended to do this and intended to use a Clintstick’ to ¢orrelate the two, but found it was impractical. DR, KLIMT: The urine test, too, has a different specificity attached to it; in other words, when is the urine spilling not specific as far as diabetes is concerned, or when is it non-glucose? DR, KENT: I think renal diabetes is one condition which invalidates urine testing as a screening test. DR. MILLER: The question of variability of glucose tolerance tests with time in the same individual has been raised. Dr. McDonald, I believe, — Company, Inc., Elkhart, Indiana = 80 = has done such a study and I am surprised we have such silence. DR. McDONALD: I started to mention it but Dr. Lazarow spoke about individual variations so I thought that answered it. But I would have been surprised if among some of those Dr. Kent referred, and a glucose tolerances test done even under indentical circumstances two months later, there were not some that changed to negative and some that changed from negative to positive in the same interval. Because in the prison study about 9% of the presumably normal, healthy males, non-diabetic, varied from a non-diagnostic level into a diagnostic level by several criteria during the course of the study when they had glucose tolerance tests every two months for a year. None of the thirty individuals had all six positive during the course of the year, but 30 of them had at least one that went into diagnostic ranges. So I think we should presume that there is not always going to be agreement between the referral glucose tolerance test and one confirming the diagnosis at a later period of time. But there presumably would be as many who screen negative that would have become positive in the same length of time. DR, KLIMI: If you looked at their ranking, do you think you would get better agreement over time? If you took thirty people and ranked them each time, would you then get good agreement in the order with respect to the glucose tolerance test? DR. McDONALD: No, I think not. DR. LAZAROW: I would like to come back to that blood sugar curve with age. Let's make a couple of hypothetical assumptions and then I want to ask my question. ~ Bl = Let's assume that blood sugar does not change with age. Let's assume that in terms of the syndrome of diabetes, some will go on and develop a progressively increased abnormal glucose tolerance but will never really show manifest diabetes, whereas other groups will, and that the per cent of the population that develop an abnormal glucose tolerance diagnostic for diabetes increases with age, as does the other. Will you get this kind of curve, assuming the blood sugar, itself, is normal, but you are now adding with age an increasing number of people who have diabetes only manifest by abnormal glucose tolerance? Will that give you that curve or can you differentiate between the two questions? With whatever curve you are showing here, you are showing a slope going up. And the question I am raising is: Can this slope, a portion of it or any of it, be attributed to the fact that with age you are having an increasing percentage of your population that has diabetes which is manifested only by abnormal glucose tolerance, and never go on to anything more? This would tend to weight at least your average but could you differentiate this by distribution or what-have-you? DR. ANDRES: No. DR. O'SULLIVAN: It would take a long-term follow-up to answer this. DR. ACHESON: There is one variable in this I think is important. The arterial bed changes with aging and this is an important aspect, so it seems to me, that in older people abnormal glucose tolerance tests are not necessarily diabetic at all. They may indicate many other things. We don't know for sure what the endocrine glands do with increasing age. We know the sex hormones change. - 82 - MRS. FISHER: The problem requires long-term study. We have noted that these regression lines are not the same for different groups of people. If you look at people who are not relatives of diabetics, that is, those who say they don't recall any history in their family, the increase of glucose level with age is less. If you look at people who are not relatives of diabetics and who are of normal weight, you see a smaller progression with age. DR, FAJANS: I think to the best of our ability to answer the question we would have to say that both factors exist at the same time. Age per se is a factor, and secondly, diabetes becomes more prevalent as age progresses, How can we dissect these factors? If we take the philosophical point of view that all diabetes must progress beyond what can be accounted for by age, then we can differentiate these factors with time. Those who have diabetes should get worse along a certain progression line, while those who don't will show only the age factor alone. On the other hand, if we take the other philosophical point of view, that all diabetes does not necessarily have to progress, that the abnor- mality may stay at the same level - then this kind of approach isn't going to answer the question for at least one segment of the "diabetic population. A bad analogy would be the individual who is allergically predisposed. He doesn't have to progress from mild hay fever to severe asthma. He may have mild hay fever all his life. Certainly mortality wouldn't disclose anything, just to use this analogy. In other words, the way I see diabetes it doesn't necessarily have to progress. I don't know whether this is correct, but I think we have «83 « a lot of information along these lines. Mild fasting hyperglycemia does not progress necessarily. We find very mild vascular disease which we think is typical of diabetes and it does not progress. Nothing is 100%, but if this point of view is correct, the approach mentioned just won't give us the answer. DR. McDONALD: Based on these two presentations today, I would like to ask Dr. Fajans if he had to solve Dr. Kent's problem and reduce his seven values to only two, which would be the two that you would choose to keep in for a presumptive diagnosis? DR. FAJANS: Dr. Kent used three. I would use, if I still had three, the one, one-and-a-half, and two hour specimens. If there were two, I would do what he said he was going to do in the future, to use the one and two hour samples. DR, McDONALD: I agree. DR, FAJANS: We all agree that we learn very little from the fasting blood sugar level. DR. McDONALD: That is the kind of discussion I hoped might come out of this assemblage of experts. That is: Which are the most important to preserve when you come to the practical problems of doing mass population sampling? What are the minimals that you would have to have in order to give any value to it at all? So I thank you very much for that answer. DR, KLIMT: If you only had one, which would you use? DR. FAJANS: If you used only one, you are not doing a diagnostic procedure but a screening procedure. And I think we should keep this in mind, that one is a screening procedure and the other a diagnostic procedure. I don't even like a two-point assay as a diagnostic procedure. I have more confidence if we have at least three values. As I pointed out earlier, - 84 - as far as making a diagnosis is concerned, even if you would do all five or all seven, if it is borderline you can't make a diagnosis. DR. McDONALD: Which one would you use for screening? DR. FAJANS: For a screening procedure I would use the two-hour level. DR. KLIMT: Why, if the one-hour is the most potent? DR. FAJANS: I think Dr. Kent answered this before, and I would give the same answer, that although the one-hour is more sensitive than the two-hour level, the two-hour level is a little more specific. DR, KLIMT: What evidence do we have for that? DR. WEST: You can change the sensitivity and specificity at will for either one. DR. KLIMI: That is right. DR. WEST: A disadvantage of the one-hour value is as Dr. Silverstone, I think, mentioned that if you have a varying magnitude of extracellular spaces and you are putting 75 grams into everybody, there will be a variation relating to the relationship between the size of the dose and the extracellular or glucose space you are putting it in. And I think if you will figure this out on a theoretical basis there will be a little greater effect of this variable on the one-hour than on the two-hour value, and furthermore it is probable -- this has not been subjected to test -- that the one-hour value would be a little more subject to the vagaries of glucose absorption than the two-hour. Again, direct evidence on this is meager. While I am talking, may I just mention that I think Dr. Andres' diagram has much to recommend it as far as sort of clarifying a way of thinking about what value lies in what percentile per age group, and if one can - 85 - accumulate a population like the National Health Survey population which was, I think, fairly satisfactorily representative, then one, as Dr. Andres showed, could pilot the degree of risk, or determine the degree of risk, for the eventual development of diabetes for any given percentile. He mentioned mortality as a possible thing to put on the one axis there. Also one could determine the degree of association of any given result with the later risk of each of several symptoms, signs, and complications of diabetes. However, I think the easiest "follow-up index of 'diabetes'", and the one that would be easiest to agree on would be just elevation of fasting blood glucose. We get into the problem of defining what is an abnormal fasting blood glucose, but I think most of us would agree that a fasting level of 130 mg. per 100 ml. was evidence of the disease we call diabetes. We could then follow a large group of individuals for several years and determine, I think, fairly promptly what the degree of risk of developing fasting hyperglycemia was for any percentile of results obtained with a glucose tolerance test in any decade of life. Another thing you could do at once, if you had a large enough number, would be to determine the findings which are associated with the results in any percentile figure, say for two-hour of 160-169 mg. per 100 ml. of a man in his 6th decade of life. You could determine the correlation of results at this level with the stigmata known to be associated with diabetes such as family history of diabetes, or a history of having large babies. Or even -- there are a lot of hookers here -- perhaps the incidence of obesity. This then would give you some indication of the extent to which any given result is associated with an eventual risk - 86 - of diabetes. But I think in the end what we have to do has been brought out by several, including Dr. Andres, and that is to plot the frequency distribution for each decade and then follow these people out and see what the risk is for any given range in the group. And I don't think this will take a generation. I think in probably five years we could add considerably to what we know now provided a well designed study is constructed involving large numbers. PROCEEDINGS OCTOBER 2, 1964 Problems In Obtaining Objective and Comparable Data from Different Investigators and Facilities CHRISTIAN R, KLIMI, M, D, 10 My purpose is to point out the statistical and epidemiological techniques employed in insuring comparability of data in a cooperative clinical trial, the University Group Diabetes Program, and, on the other hand, as a physician, to indicate the selections which have been made for end-point determinations in such a study as this. These are not as easy as the selection of mortality as a differential end point in many studies, because fortunately that is not large enough in diabetics, in contrast to coronary artery disease, to give us a reason- able expectation of determining differences. Now, why do we have to have cooperative clinical trials? The primary reason stems from the requirement of numbers, and adequate numbers, in order to get statistically significant answers. Further- more, it stems from the need to get as broad a representation of types of patients as one can get in order to make the answer more valid for a broader audience, Also, in order to insure quality of data, it is reasonable not to overload a single clinic and participant with too great a patient load. So we consequently found it necessary to establish a minimum require- ment as well as a ceiling of numbers of patients a clinic has to contribute, They do not differ so very much. In our instance, the 10, versity of Maryland School of Medicine, Baltimore, Maryland - 88 = limits are 75 as a minimum, and 100 as a maximum. We feel that beyond 100 when long-term observations are concerned, and to hold down the drop-out rate to a minimum, this is as much as a clinic can reasonably be expected to handle. The number of clinics is determined by the total number of patients, which in turn is derived from the minimum number of observations required and the minimum time expected for observations and for end points to occur, so that we may get an answer to the question posed by the clinician. To get that question properly formulated is one of the difficult tasks. You cannot have him say that any difference would be of interest, because obviously the reply would be, "That requires an infinite number of patients." He must be satisfied with a certain specified difference between treatment groups, usually one which would make him satisfied to institute a life-long treatment if it were to be found superior to any other treatment. One other reason, more of an administrative nature, why we must have studies in human beings and why cooperative trials, is the more recent emphasis by the Food and Drug Administration not to license any new products unless their potency has been established in human beings. Now, when we come to the comparability of data, the first require- ment is common standards for the selection of patients. Whatever variables we have we like to reduce. For example, the duration of the disease -- many things may have happened already to a patient, parti- cularly as we are going to discuss vascular complications in a little while, so that we have a lessened chance of developing an end point if - 89 - we take a patient whose illness is of some duration. On the other hand, the degree of insecurity as to the time of onset, the diagnostic criteria used for a first diagnosis, make it desirable to limit the period retro- actively at which a patient becomes eligible to enter a study of this kind. After many discussions in this regard, which many of the present audience will recall, we decided to go back no further than one year, and to accept as diagnosis only a glucose tolerance test or the fact that a patient was presecibed to take insulin, This did not mean that at the time of the consideration of the patient for the study he did not have to pass standardized diagnostic criteria. I will beg your pardon if I am somewhat repetitive as to what was brought out so well by Dr. Fajans and our discussion, but I would like to reiterate the diagnostic criteria used by the university group diabetes program, because it is one program where a group of 12 university clinics agreed on one single diagnostic criterion, And while it is valid, it was brought out yesterday that in forming this basis, we had to fall back on the rather vague criteria of a symptomatic diabetic in establishing the frequency distribution of summed glucose tolerance test values. You will agree with me that this is a conservative approach, in other words, that we then can be rather sure that those whom we select do fall under a category which could uni- formly be accepted as being diabetic. Before going into the interpretation of the test, it was mentioned yesterday that it is more important to agree on standard procedures for the test. And again this has had long discussions within our group, and - 90 = we came to the following definitions: We did not feel that preceding diet was of exorbitant importance, except that a person not be on a carbohydrate starvation diet. He must be fasting overnight, and he must not receive any hypoglycemic treatment at the time; he would have to suspend it for at least three days prior to taking the test, We then measure height and weight, and from a Dubois nomogram we determine his body surface, and on a scale of 30 grams of glucose per square meter of body surface, we challenge him. This is essentially lower on the average than the standard tolerance test with 100 grams. Usually an adult male gets in the neighborhood of 60 to 65 grams and a woman correspondingly less. We do know by an accidental experience that the amount of challenge is not insignificant to the outcome of the GIT. By some mistake, one clinic's technician equated a glucose solution to 100% glucose for some time, without this being known internally, and we found upon statistical analysis that the mean value of that clinic was around 560 for the summed four glucose values, fasting 1, 2, and 3 hours, while the mean value for all other clinics was in the 900's. We were surprised that this one clinic was getting totally different diabetics from all the other clinics. When, this was discovered, and the error corrected, the mean value jumped to 900 for this clinic, too. - 9]~ So in actuality the criteria we are using are not normally posing any problems. There are very few borderline cases, as we would hope, because in a general population distribution, and in these symptomatic diabetics, the overlap is a minimum at that particular point of 500 milligram per cent which we chose as our arbitrary cut-off point, I don't need to go into some of the other criteria. They have been mentioned. We take venous blood and we try to avoid prolonged compression by tourniquet, and we have finally -- and I may say finally, after years =-- agreed on one method on how to determine glucose. We are using the AutoAnalyzer!! which uses a modified Hoffmann. To standardize a chemical procedure in a number of clinics requires not only that all participants agree on a common protocol and on a common piece of equipment, but it needs checking over time, so that the consistency remains throughout the study. We have a procedure whereby periodically we submit to the clinics, in random order -- and here I mean random rather than haphazard -- a series of versatol specimens of three kinds: versatol, versatol A, and versatol A alternate, whose known values escalate between the three kinds and change from lot to lot. They have the advantage of being in a dry condition so we can mail them without deterioration. Furthermore, by a process of dialysis the contents of glucose and of creatinine, have been taken out and replaced by adding a measured quantity. 1), chien Instruments Corporation, Chauncey, New York “92 = So we not only have a dried serum, but we also have a known amount of the substance in question in addition to the assayed amount which the company and a central laboratory determine. In this manner, we can find out whether the clinic's result corresponds to the true value, and whether its variation -- and there is variation ~~ is greater than the aggregate variation for all clinics, Surprising as it may be to some, this is not a superfluous method, even in glucose determination. We do find that at times there are systematic differences, and at times an unacceptable degree of random variation of the test, So close quality control must be maintained, and 1s maintained in our study. Yesterday, we discussed at length whether diagnostic criteria ought to be uniform for all human beings, or whether we ought to adjust them by age group, by sex group, possibly by other attributes, such as degree of obesity =~ maybe the cohort in which they are born, This is a problem which we recognize, but in a study of this kind where there is a limited period and limiting definitions for types of patients admissable, we have used one single diagnostic standard, although we realize that may- be we are favoring females because in our experience ~-- and this may be a question for discussion ~- contrary to observations mentioned pre- viously in this meeting, we have found that at least non-diabetic females on the average give higher glucose tolerance values than corresponding males of the same age group. And we have found, if we standardize for obesity, by using Metropolitan Life Insurance standards and per cent deviation thereof, we lose the difference in the glucose tolerance test. “53, we This difference is apparent in every single one of the four values, fasting, 1, 2, and 3 hours, averaging about 10 mg. per cent in each one of these four tests, So you can, in a borderline case, see that we are more likely to include females under the same diagnostic standards, than males, This may be due to the average greater obesity of women compared to men, Once we have agreed on a uniform procedure of admitting patients to the study, and agreeing on the procedures of regular checks for continuity among clinics, we use, as all clinical trials do, the technique of random allocation of patients to treatment groups, And again I would like to emphasize that random means anything but unselected. In fact, they are predetermined and preselected according to a computer program of random numbers which has been checked as only a computer can, by means of repeated samplings and Chi square tests, as to whether they are indeed random ~-- and professional statisticians will not be surprised that even tables of random numbers are not random in all combinations. We don't hold any brief that our particular computer program is random in all aspects, It is random for the series we are using, Thus, only a coordinating center, which we represent, holds the central key for treatment allocation, and treatment allocation is done only after a patient has been admitted to the study. In other words, the clinic decides on the selection and we give the predetermined treat- ment allocation, - 094 - This should, and does, insure that variables on which we do not block on the average are equally distributed, such as age and other variables. We do separately allocate at random by clinic and we use a method whereby periodically the equality of numbers by treatment group is assured within each clinic. Thus we limit the random process, so that we can be sure in the end we can only have a predetermined maximal difference of numbers in each treatment group. Having gone that far, our next and most difficult problem has been to agree on the basis of what end points, what criteria we are going to measure the preferability of one treatment over another? In diabetes, over a limited period of years -- and we are speaking of 7 to 10 years in our instance -- it is not too likely thdt mortality will be a suitable one, although it should have little ascertainment error. What clinical measurements ought to be chosen? While I am not competent to go into all the clinical reasoning, I will present the choices made. I may be able to present and defend certain principles for the choices, namely, these choices should be limited to one, or at most twes, end-point determinations per organ system we are concerned with, should reflect an optimum of sensitivity and of specificity and should if possible leave a permanent record. For vascular complications in diabetes we choose five groups, namely for the eye, fundus photographs of the retinal background; for the heart, resting and post-exercise EKG'sj; for the peripheral vascular system, soft tissue X-rays and von Recklinghausen oscillometry; for the kidney, qualitative and quantitative determination of albumin and a ~-95 = creatinine clearance test, And finally, for diabetic neuropathy, where we can only say this is experimental, we use an apparatus called the biothesiometer, which quantitates the patient's reaction to the vibratory impulse, In each of these systems we have different problems of quality control, and the principles are well known to our colleagues who are involved in similar studies and have a similar background. For purposes of this discussion, I would like to differentiate between measurements such as the creatinine clearance, a continuous measurement on one scale, cholesterols, which do not form an end point but nevertheless are being studied and will be measured in their predictive values for vascular complications, and the ascertainment of attributes such as microaneurysms in the retinal picture and the presence or absence of calcification in the arterial tree as deter- mined from soft tissue X-rays. These require different techniques. With regard to measurements, we have two types of procedures: One which requires that every clinic performs the test in a uniform manner. In this case, I will speak of the creatinine clearance test. That test, as all tests, requires a Consensus on uniform procedures of how to perform it, That is to say, the time of day -- which is not unimportant ~-- must be uniform, It is not without importance whether we do it mornings or afternoons. The degree of hydration, the schedule of hydration. -96 ~- The duration of the test; the accuracy of the collection of urine and its measurement; the variability of the time over which it is collected. We aim at a 3-hour collection but, of course, {ft does not matter as long as time measurement is accurate. Having done a study in a prison, where this test was standardized and calibrated, even there you cannot get it done during exactly 180 minutes each time, There are certain physiological barriers, The time of the taking of the blood sample and how it is taken, may be of less importance in that particular test, but we agree, that it is being taken in the middle of the urine collection period, We found that there is some interference of high glucose values with our ability to determine true creatinine, Therefore, it is not desirable, we believe, to do simultaneously a glucose tolerance test, It might give some interference if we happen to have a few patients who have very high blood glucose levels at the time, This is a condensation of the process which took a long time for a large group to agree on. One important question is how to minimize the importance of urine retention, We have experimented with techniques whereby we split the urine component into two ~- collect two amounts, let us say the first half of the collection period and the second half =-- and we see whether we get differences in urine concentration for creatinine, If we do, we can surmise that during the test differences in output and perhaps differences in retention may have taken place. It has been proven in a small number of cases by catheterization, that the urine retention is -97 - not a constant, even for a given patient. He may at one time pretty nearly completely empty his bladder and at another time have a sizeable retention. Of course, this is the reason for hydration, so that this error be minimized, because the more urine output, the less the importance of urine retention. Speaking of quality control, there are many ways to do that. The chemical type I have already referred to. Versatol samples are being tested with each run to check the consistency with known creatinine values, and its variation among duplicate determinations. Another difficulty one has to struggle with, if you change units, or if the units are different from what the laboratory is accustomed to, it is very difficult to get them to change. Now, for some reason or other we have agreed on reporting by milligrams per milliliter. The clinics are largely used to reporting milligrams per hundred milliliters. That looks to you like a silly digit error. However, the coordinating center cannot make the correction or the assumption that it is a digit error, and the computer certainly will not, and the computer will auto- matically reject such a value. This brings me to the fact that in the course of the experience of this study, we have found it absolutely essential to go into computeriza- tion of editing of records, of monitoring and for sequential analysis. The editing consists of two principal types, namely, cross-sectional editing--that is, when each record comes in, it is edited for complete- ness -- and for editing internal consistency. It has been mentioned before, that questionnaires need to be professionally designed. Indeed they do. And we have had such advice and we have included internal checks, “DG repetitive as they may seem to the clinician, to determine that consistency is being maintained. And little items such as height measurements, for example -- you would be surprised what the random variation is per patient, and this needs to be held down. So the computer goes to the record which is immediately punched up as is, and checks for completeness, consistency, and reasonableness of data. The assumption of reasonableness 1s, of course, arbitrary. But if you take certain tests, which I will use as an example, you will agree with me that certain values are simply physiologically unacceptable, Data which led to a crude clearance of less than one ml. per minute are undoubtedly incompatible with life and must be rejected. On the other side of the spectrum, extremely high values have never been observed and ought to be rejected. Nevertheless, be it by digit error or other neglect, they do occur, and the computer will print out a message, and it is being so formatted that it will be doing so on pages which pertain to a clinic only, so that all we need do is take a copy, and we print on the computer 12 copies of the output simultane- ously, and we mail a number of copies to the clinic, so that these can be corrected, At this stage I would like to hand out samples of computer outputs =-- there are two types =-- which show at the top a set of limits for checking reasonableness of the output for the creatinine clearance test, These require serum creatinine aliquot difference, We require duplicate deter- minations, and we expect those to be neither too large nor too small, Clinics which give us zero difference in aliquots are highly suspect in our eyes. That also is impossible. And indeed, it is very difficult to fudge a normal distribution of errors if you want to. So while this -99 looks rather primitive, indeed it is not, and I would challenge anyone who would try to make up a normal distribution of errors between aliquots by virtue of the fact that they are not blind. We ask for urine creatinine aliquot differences, average of serum creatinine aliquots, and average of urine creatinine aliquots, In other words, the samples are split in two and each one is determined twice, so we have actually four different determinations. Then we require that the crude creatinine clearance be checked, and then this be checked on the basis of averages, on the basis of clearances which can be computed for ea:h aliquot determination. These have different limits which we consider acceptable. And then you will find a printout which, in addition to the fact whether it passes the edit test or whether the test needs to be re- peated and why it needs to be repeated, gives the values on the basis of which the test has been rejected. Surprisingly enough, I haven't really touched on a good deal of the spectrum of specifics I wanted to mention, and maybe they ought to be brought out in the course of the discussion. One of the assignments for the discussion is to bring out what kinds of criteria have been used, and how to go about to quantitate and to measure subjective recognition of abnormalities in such records like fundus photographs. That is one I believe which will be of some interest. The general philosophy that you start out with is a very complex scoring system and then you test for reproducibility within and between reader variation, I may add that this program has two expert readers ~100~ for every kind of subjective end point there is, and that each reader, blindly, in random order, reads each record twice, We have a schedule of monitoring whereby all incoming records are monthly sent to both readers and then switched around, random order being changed, and the records are being blinded to avoid specific bias by clinic, and the like. There is a whole lot of methodology to be mentioned in this context, but I think the stage has been set for further discussion, which will take a good part of the remainder of the morning, at which time I would also like to report to you on the possibilities of direct computer analysis of some of these records. I happen to have been with a group last week dealing with the computerization of EKG diagnosis, and perhaps I am in a position in about ten minutes, at a later time, to define for you what the present status of this field is, of interest to diabetologists because the EKG and cardiovascular disease as a predominant complication of diabetes will require our attention. An automated diagnosis is here, and in line with what we heard yesterday from Dr, Lazarow, you will be surprised to what degrees of sophistication present methods already go. The future is really unlimited, and it is undoubtedly possible already to get better extraction of the value of an EKG by computer than it would be by a number of highly regarded experts, The future will open totally different vistas of analysis which the human mind cannot do. Vector analysis, integration of planes and 3~dimensional and multi-dimensional space comes into it, and it has become a highly involved business, ~101~ I think with these remarks I will stop here now and wait for specific questions and discussion. Thank you, Mr. Chairman, DISCUSS ION DR, McDONALD: I would like to ask Dr, Klimt: When you had this error in how much glucose was administered, did you have to throw out those data or was there a tool you could use to say what the values would have been had they been properly loaded? Or if that tool is missing, can you conceive of one that would allow you to bring those values back into comparability, even though they were loaded with less or more than the agent? DR, KLIMI: Dr, McDonald, fortunately for us this was not necessary, The glucose tolerance test does not form an end point on which our treatments are being evaluated. They are criteria for entrance into the study. Now, had we used double the glucose loading dose, we might have soeluled some which would have fallen underneath the diagnostic thres- hold with the prescribed loading dose, and we would have had some inconsistency in the study. As it happened, we had too low a loading dose and all that has happened as a result is that we may have missed a few cases prior to random allocation, which might otherwise have been eligible for admission to the study. I would warn anybody, however, in answer to the second part of your question, against trying to extrapolate what the test result might have been had it been under a different glucose loading scheme. Similarly, I would not suggest that one extrapolate from capillary to venous. That 1s not a standard factor, It is not a standard in terms =103- of percentage; it is not a standard in terms of absolute values, And it obviously differs for each part of the glucose tolerance test. I do not see anyway these could be made comparable or readjusted to each other, DR, CRAMPTON: I would like to make one specific point about this, I think it is important to note that the criteria on the glucose tol- erance test, and more specifically the method here, are set up for this particular study and do not necessarily apply to early diag- nosis of diabetes in the population or any such thing, They are set up to pick up certain diabetics, and not necessarily to apply to population studies to pick up early diabetes, DR, KLIMI: I do want to say that the way the standards were set -- and this was discussed yesterday ~- completely bear out what Dr, Crampton said. In this case we want to be conservative. We want to be sure that in retrospect we are not told, "Oh, you have included a good deal of gray zone diabetics which may indeed not have been diabetics" and considering the investment over ten years in any single patient, this is only good economy. However, the same principle can be so adjusted to include as much of the gray zone as you care to do, DR. BURCH: In that regard, my own feeling is that it would be very advantageous to include some of the gray zone. DR, ACHESON: I know this is repetitious but could you tell us first exactly what the purposes are and could you tell us which the university centers are, -104- DR. KLIMI: The purpose of the study is to evaluate different hypoglycemic treatments, specifically DBI-TD in a standard dose, tolbutamide in a standard dose, insulin in a standard dose, on the basis of body surface, rather than the chemical need to maintain normoglycemia. The three groups are compared with what we term a negative and a positive control. The negative control gets only the uniformly prescribed diet and a placebo =-- these placebos we can only give for the oral drugs, in other words, we do not have a placebo injection, The positive control is insulin ad 1ib, that is to say, as much insulin as is needed and the variety of insulins that are needed and a difference in schedules that are re- quired to maintain normoglycemia as far as that can be determined. The purpose of the study is to evaluate the difference which might occur in the onset rates and in the development of so-called late vascular complications in each one of the organs I have mentioned, the eye, the heart, peripheral vascular system, the kidney, and the periph- eral nervous system, The participating clinics are 12, Our chairman is Dr, Max Miller, at present here from Western Reserve University, and then we have clinics at Harvard, State University of New York in Brooklyn, Johns Hopkins, the University of Alabama, the University of Puerto Rico, the University of West Virginia, the University of Cincinnati,-Washington University, Presbyterian-St, Lukes in Chicago, the University of Minnesota and the Virginia Mason Foundation in Seattle, “Sv. S. Vitamin & Pharmaceutical Corp., New York, New York ~105~ DR, O'SULLIVAN: With regard to the selection of the patients I would like to pursue one of the questions Dr, McDonald has already asked, The error in dosage in one clinic resulted in an average aggregate difference of somewhere in the neighborhood of 300 milligrams. Have you any idea of what the average error was in the dose loading? DR. KLIMI': The loading was half by assumption that the saturated solution was 100%, In other words, they were challenged with 15 grams per square meter of body surface rather than the prescribed 30 grams, so the average per adult male would have been reduced to 30 grams rather than 60 grams and it is in this range where it makes a substantial difference, DR; O'SULLIVAN: I had two other points about the selection by glucose tolerance test, You could have been more conservative if you had used the cut-off point at the upper end of the distribution of the normals, rather than the lower end of the distribution of the diabetics, DR. KLIMI: Excuse me, This would not have been the case because the upper end of the non-normal distribution, positively skewed and asymetrical, is one for a general population including unknown dia- betics. Therefore, we couldn't have used this for a criterion, It goes up all the way and covers the entire 'iabetic' range. This would not have been a practical solution, DR, O'SULLIVAN: One final question, if I may. How many of the people that are selected by clinics in this way have elevated fasting blood sugars, This is of some pertinence with regard to a study on the complications of diabetes. ~106~ DR. KLIMI':t Dr, O'Sullivan, the latter is a question we can answer at any time specifically by writing an appropriate computer program and asking it the question, Offhand, I wouldn't want to make a guess, I don't know how many would have so-called normal fasting sugars, how many would not have, The aggregate of the sum of 500, however, insures that in most instances there would also have to be at least an upper level nermal fasting in order to get to the sum of 500. In order to exceed 500 you see very well that the fasting and the 3~hour values have to be in excess of 80 in order to reach the minimum diagnostic criterias, And doen't for- get the average is 900 for all clinics, So you can well imagine that probably for the majority of patients we also have elevated fasting glucose values, but not for all, DR, MILLER: In general these are mild diabetics and fasting sugars on the average will not be very much elevated, If they were very much elevated they would become symptomatic diabetics and therefore we could not have used the placebo control, I suppose you are hinting at the possibility that people with high blood sugars will have more complications, This was in the back of your mind? DR, O'SULLIVAN: There are two aspects, That is one of them, DR, MILLER: Let me answer that one first, Perhaps that was a lucky shot. This is a conclusion that you are making? DR, O'SULLIVAN: Yes. ~-107- DR, MILLER: We are testing the effect of levels of blood sugar on the development of complications because in this group you will have a certain range. The criticism has already been made by some groups that we are really not testing the group that will get vascular complications, I think this is a matter for discussion at this point because, as far as I can tell, complications occur in all diabetics, It has been said that the older mild diabetic, whose blood sugar on the average is less high than in the younger diabetic, has complications less frequently. From the Joslin clinic, the older diabetic is reported to have retinitis more severely, It develops more rapidly and with more complications than in the younger diabetics, Also, we get the coronary artery diseases, the renal disease. DR, KLIMI: It may be stated that at baseline, which is at most one year after diagnosis, between 15 and 24 per cent, depending on the end point, already show vascular complications, have abnormal EKG's, or have calci~- fication of the vascular tree of the legs, In other words, this group does have vascular complications, DR, MILLER: There is one way of answering: The complications are exactly the same in both groups, old and young; and second, even if they are different, all we can study in this experiment is the older mild diabetics so we can describe the onset of vascular complications in this group. We will be satisfied if we ¢an do this accurately and scientifically. DR. O'SULLIVAN: The other portion of this question, if you select people by glucose tolerance tests, the number of spontaneous remissions will be -108- greater for those with normal fasting blood sugars than for those with elevated fasting blood sugars, So if this 1s of pertinence to the study, the ways they happen to fall in the various study groups would be important, DR, MILLER: Actually, for the first time there will be objective data on this, You have some, too, but in our study since we do glucose tolerance under the same conditions each year, we will be able to give a figure on how much it improves with placebo, with insulin, and pBI-TD and, as you have surmised, there are some very interesting results, DR, KLIMI't While the glucose tolerance test is not the recognized end point by treatment group in this study, it is being repeated annually, having suspended hypoglycemic treatment for the preceding three days. Thus, there will be a sampling annually which will tell us which pro- portion did show a remission, if any, at that time, or what progression is found. MR, WOOLSEY: I wasn't clear from your description how lon® the patients are observed, I am interested in knowing how long they are observed and what are the figures on the numbers lost to observation, and whether you have any knowledge about the conscientousness of the treatment, DR, KLIMI': With regard to the length of observation, we started out our study requiring a minimum of three years of observation. The maximum will be ten years, Of course, they do not enter all at the same time, but in the courge of three years, which is the permitted period for a clinic to reach its complement in the study, So it is from three to ten years =-- the range of observation, 13g, S. Vitamin & Pharmaceutical Corp., New York, New York -109- Surprisingly, and fortunately, drop-out rates and loss to observation are very much less than we had calculated in our expected numbers in the basic study design, It is less than 4% after four years, Obviously, these are all diabetics, They get free treatment; they get free drugs; they get special attention; and we have a limited number by clinic, and a lot of loving, tender care, I think that is one of the most encouraging features, the drop-out rate, We had counted on approximately five times as much in our basic assumptions, and for once we are pleasantly surprised. We do get reports on consclentousness of treatment, It is not chemically monitored, and I would say that the results of a study would reflect a combination of the potency of the drug and of the acceptability of the drug. And as such, I think this is the practical combination as we have it in nature, This is what the patient will do. And I don't need to go into the difficulties of monitoring adherence to a prescribed treatment, It is being periodically checked and we hand out drugs for a three-month supply, but fudging can obviously occur, so we don't have proof of how well a patient really adheres, We make various checks during replenishment of the drug supply but this is not beyond fudging if the patient wants to, DR, FAJANS: I have two or three questions, You mention quite a bit about the drugs and about positive controls and about negative controls, I can't recall having heard anything about the diet, if any, that these patients are on, Can you tell us something about the dietary, or lack of dietary therapy? =110~ DR. KLIMI: I think I did refer to it, but not with enough emphasis. They all get the same standardized dietary advice, and it follows the ADA prescription, But there is no attempt at special enforcement or special emphasis on the diet, except that all are being glven the same advice and the same prescription in its carbohydrate, fat, and protein ratios, We make an attempt to reduce a patient's weight if he is more than 15% over-weight, and to increase his weight if he is more than 15% underweight. Otherwise, we give him a normal caloric diet. DR, FAJANS: 1 think it has been the pretty general experience of dietitians that unless diet instruction is given around the patient's usual food habits and trying to stimulate his previous dietary habits, diet instruc- tion is pretty ineffective, Could you tell us actually how effective this type is? DR. KLIMI: This we don't mind, We have a random allocation system of patients and we are not testing diet as one of the treatments, We want to keep the treatment groups comparable, in every respect except for the specific treatment we are giving, This is not a study to determine the efficiency of diet prescription, DR, FAJANS: This is not my question, My question is: In practice does it come out that the diet is a free diet rather than a standard clinic, DR, KLIML: I think it will vary by clinic, DR, MILLER: The diets have been set up for various caloric levels, 1800, 1500, 1200, and so forth, and percentage of fat to the caloric total has been fixed between 30 and 35 per cent, depending upon the level, Each patient receives dietary instruction in the usual manner at the clinic, and all the clinics have trained dietitians, “111 =~ The question you are asking is a very subtle and difficult one: How well do patients follow diets? I think most physicians who have treated diabetics who are obese -- and most of these patients are obese, I might say -~- are usually quite discouraged about the results, As I look about the room here, I know that people who would like to reduce their weight have great difficulties, So, although we instruct and reinstruct, it is my impression that dietary adherence is not very good even if you persist in it, Now, I think the important thing is that the diet is not one of the variables being tested. Actually, as we look at weight changes, which is one index, the number of tons that our patients weigh in our own clinic remains the same after X years, Some of them go down a little bit, some go up a bit, but the average tonnage is the same, DR, KLIMI: Max, I don't think that the data at the moment bear out that we don't have any weight development, There is an initial loss upon entry in the study, subsequent gain and continuous increase -- as I think is human nature, diabetics or not diabetics. However, we have been concerned differentially by treatment group with weight movements, DR. ANDRES: In trying to think about whether the study included only "black'' diabetics or whether one is getting into the gray range, we need information concerning one {impor tant technical factor, Was glucose analysis done on plasma (or serum) or done on whole blood? You indicated that you sent out serum controls, the versatol controls, and I presume that means you did the glucose tolerance test on serum, -112- DR. KLIMI: Yes. DR, ANDRES: Well, the glucose concentration in serum is roughly 15% higher than it is in blood, a not inconsiderable factor. Dr. Fajans, are your standards set up for whole blood? DR, FAJANS: Whole blood. DR. ANDRES: So if you make a calculation of what your normal standards would be, Dr. Fajans, if serum were used, 500 becomes not only the limit to the black range, but an almost white range, since fasting level would be very close to 100 normally. The 160 level, Dr. Fajans, for you at one hour would become roughly 184 for plasma or serum; the two-hour level instead of 120 would 138, and in three hours you might expect it to be 100 again, which puts you up to about 522. DR, KLIMI: Aren't you thinking that these differences are rather uniform? DR, ANDRES: The difference between glucose concentration in serum and in whole blood is very uniform for subjects with normal hematocrits. DR, KLIMT: And also of a rather larger magnitude than I have been led to believe they are. DR. LEONARDS: They are using whole blood. Are you doing the glucose tolerance test on serum? DR, KLIMI: Yes. The Sutobnslyzer’ method -- DR, McDONALD: We just published some comparisons on whole blood and plasma values in sugar because we needed it for translation because we are running most of ours on plasma. Roughly it is 25 mgs. difference at dtechnicon Instruments Corporation, Chauncey, New York -113- fasting, 35 at one hour, 25, and 25, These are rounded off, but that would be approximately 100 or a little more in the total value, the total sum, The plasma values are always higher than blood, and it varies according to the time after loading. DR, BURCH: Then it is plasma the the AutoAnalyzerl? does this on? DR, O'SULLIVAN: The AutoAnalyzerl® can handle almost any specimen you want to give to it -- whole blood, hemolyzed blood, serum, or plasma, It doesn't matter whether you use a glucose oxidase or the ferricyanide method, you will still show the difference between whole blood and plasma. Plasma and serum give approximately the same values, But if your clinics are using whole blood and some serum, you have the differ- ences that Dr, McDonald has indicated. DR. CRAMPTON: I can answer specifically what we are doing. We are running all these in parallel, but for the criteria of diagnosis, in the shift from the Somogyi-Nelson, which was a whole blood determination =-- we shifted to the whole blood and used the same criteria for diagnosis on 17 with whole blood. That is precisely what we are using. the AutoAnalyzer DR, O'SULLIVAN: Fresh whole blood, or frozen, DR, CRAMPTON: Fresh whole blood, yes. DR, McDONALD: I might add that the correlation line was so good that it allowed individual translations from serum to whole blood values. So if this needs to be done, at least there is a tool so it can allow transla- tion which may not be possible with other values, 15pechnicon Instruments Corporation, Chauncey, New York 161h44, 17 Op. cit, 114 DR. CRAMPTON: That is precisely why we are running them all in parallel in our lab to see if that is true, DR, BURCH: How many here do determinations on whole blood? Well, it looks like the whole-blooders really are in the majority. DR, MILLER: Let me emphasize, The implication I have from Dr, Andres’ discussion is that we are picking people whose sum of glucose is 500, Let me reiterate that the average is 900. We have deliberately tried to get people who are diabetic by any standards, Of course, there is one point that has been raised that perhaps it would be nice to study the people in the gray zone, the people who are Just sometimes diabetics and sometimes not. This is perhaps another study, and we probably will have some who are very close to that 500, But we have tried to get patients that anybody will accept as a diabetic, DR, BURCH: I think this is an important thing that has come up here. What Dr, McDonald just said, is encouraging that even if we all don't use the same blood fraction, the results are adjustable, DR. RICKETTS: If you hemolyze blood, Dr, Crampton, you are letting loose a reducing substance and no longer have true glucose, I wonder to what extent the group believes this is true. It was demonstrated years and years ago, IL suppose it is possible to set the AutoAnalyzer!® to compen=- sate for this but I daresay it would be hazardous because not all glutathione values are the same in red blood cells. DR, ANDRES: I would like to get into the record that my criticism about black, gray, and white earlier was based upon the assumption that glucoge analyses were made on serum since versatol had been sent around for testing. 18 Technicon Instruments Corporation, Chauncey, New York -115- But if whole blood is used, that changes the minimum number in my mind from a very light gray to a very dark gray, DR. MILLER: Versatol is sent around to compare the accuracy of the method. DR. ANDRES: Actually it would be more appropriate to send around the same substance that is being tested but I guess it is messy to send around whole blood. DR. KLIMI: It would be, I would like to restate for the record and correct myself, that the study is using the whole blood method and we are using the AutoAnalyzer uniformly since last July. DR. BURCH: I waild like to have an agreement, if possible, concerning the comparability of these two rather basic approaches to testing the glucose, Dr, Acheson, for instance, uses serum. Some others use plasma, and almost everybody else here used whole blood, Dr. Acheson tells me he uses the serum because this works in better with the other tests that he does, Dr. McDonald, you mentioned that in your study, there was a direct relationship between these, Has anybody else done any comparative evaluation of these tests? DR, KALBFLEISCH: Dr, West and I were involved in diabetes studies in various freign countries. There we were faced with a transportation problem because we had to set up a central laboratory. So rather than being able to determine blood glucose in the field, we had the additional problem of storage. We used frozen plasma that had been collected with oxalate-fluoride mixtures and then transported the specimens back to the United States in the frozen state, 19 Technicon Instruments Corporation, Chauncey, New York =~116~ We also ran a comparability study between plasma and whole blood and found virtually the same thing that Dr, McDonald did, In addition we repeatedly thawed and refroze the plasma and found no change over a period of several months in the glucose content, DR. BURCH: I think this 1s extremely important, Then, if the tests are done, if everybody specifies exactly what he has done, which I assume everybody in this room would, presumably appropriate comparisons could be made, DR. McDONALD: I might say that when we started using the AutoAnalyzer , 29 we were testing it to see whether it could accommodate large numbers for screening purposes, and we were advised at that point that it would be more likely not to have trouble with the membrane if we evaluated plasma instead of whole blood. So that is the reason we chose that to begin with, But then, to make the data acceptable for referral purposes in terms of whole blood, we had to set up a study to find out what the differences were and whether it would allow individual translation of one to the other, and fortunately we found that it would, DR, BURCH: There is another aspect on this same matter that came up in the earlier discussions, and that is that apparently different investi- gators are using at least two different tests for glucose, How com=- parable are the results, for instance, on the same serum with the different tests? Is there any information on this? Becmase this would have an important bearing on the same thing, “Ypechnicon Instruments Corporation, Chauncey, New York -117- DR, O'SULLIVAN: We compared Somogyi-Nelson, svkoAnalyeert and glucose oxidase methods and they all gave fairly comparable results. We didn't do very much with the glucose oxidase method -- the method described by Hill and Kessler -- but what we did approximated the other two methods closely. The Autodnalyzer® over a period of time will give results about 1 to 2 mgs. higher than the Somogyi-Nelson determinations done on the same specimens. DR. BURCH: There is another problem that I can visualize. Of course, we are interested in population studies in which the Sutenaiyzens is almost essential if you are going to have much of a population. But there are those who will be attempting to correlate data from these population studies with their own studies or their own clinic, and many clinics, and particularly doctors, for instance, who do their own glucose tolerance tests aren't going to have an Antohnalyse™® and they are going to use some of these other methods. How do these compare? DR. O'SULLIVAN: Well, the technical variability is fairly great -- great enough to create quite a problem. When we are doing population studies we are looking at a group of people and often drawing conclusions from the group rather than the individual. It is when you come to looking at the individual and the classification of the individual that the problems of both technical and biological variability are very great. lye chnicon Instruments Corporation, Chauncey, New York. 22514. 2300. cit, -118~ DR, RICKETTS: I think our ultimate objective is to determine how mass studies shall be conducted, and if this is true ~- and I hope it is == it seems to me the prior topic, before we talk about serum or plasma or other methods versus Somogyi-Nelson, should be whether in the opinion of this group capillary bloods are to be employed rather than venous, I think you should start from there because the technical problems are rather different in the two situations. For my part I would like to hear a consensus, if possible, as to whether capillary bloods are preferable in mass procedures over venous samples, DR, ACHESON: I don't think we can hope to come to black and white answers on that, It depends entirely, I think, on what you are trying to do. Dr, Ricketts talked about screening. This is what we are doing at the moment in New Haven but we are screening for more than one thing at a time and this is why we are interested in serum, We are interested in total protein and other things that require serum, so we have done the glucose in serum, too, When one is screening or doing epidemiological field work one is often doing it in confined quarters, as you very well know, and one simply cannot use the lab techniques one can in the clinic, One has to keep things as simple as possible, In fact, we have a small quantity of whole blood we use for the hemoglobin and the cells for Rh typing. Another aspect of this is that they are producing multiple-channel sutobnelyzerd > that will do more than one thing at one time in the same specimen and to my knowledge those have been developed only for serum and plasma, 43 Technicon Instruments Corporation, Chauncey, New York -1ll% A further point. If one wants to do what Butterfield has done in Bedford, and do fairly full glucose tolerance tests with four points, one just doesn't get cooperation if one tries to put a needle for every specimen, One must use capillary methods, He uses the ear lobe and he has been able to train girls who aren't even nurses how to prick an ear, So I am a little skeptical as to whether we can ever get a straight answer to Dr, Rickett's question, DR, BURCH: Well, this does bring up a matter such as I brought up yesterday, and that is the fact that when you are working in a clinic or when you are working in the field, you do have an entirely different set of problems, and the procedure which is applicable to your clinic may very well not be applicable if you are having to do your study in a mobile van or trailer. The only ones who have been mentioned who have used capillary blood, it seems to me, have been in England, Is anybody in this country using capillary blood? DR, KENT: All ours was capillary blood. DR, KLIMI': Our population based genetic study used capillary blood exclusively, DR, RICKETTS: May I ask Dr, Klimt: You are not talking about the present university study group, This is a previous study. DR, KLIMI': This is a study of my own, unconnected to the university group, which used capillary blood and the glucose oxidase method. DR, RICKETTS: Unless you have the data, and you may well have, then aside from your study it is necessary to point out that if a decision is reached to use capillary blood in screening, it would be essential to - 120 determine glucose tolerance from a large sample of persons by this method, and to my knowledge, with the possible exception of yours, Dr. Klimt, there is no large series of glucose tolerance tests on capillary blood for a large part of the population, so you will have to establish new standards. This may be worthwhile if the capillary method is so much more desirable. DR. ACHESON: This is what has been done in Bedford. DR. RICKETTS: Standard dose of glucose. DR. ACHESON: Yes, 50 grams, fasting. MR. WOOLSEY: I was very impressed by what Dr. Acheson said about the difficulties of applying one or the other of these methods in different circumstances. It seems to suggest the need, rather, for more calibra- tion studies of one sort or another to show the inter-relationships of these different types of blood and methods of analyzing them, and so on. DR. BURCH: Well,Dr. Klimt, you have now worked with both systems. You did one study with the capillary system and now you are working on another system with the AutoAnalyzeRt using venous blood. What is your view on this? Which do you think would be preferable in population studies? DR, KLIMI: Mr. Chairman, there are populations and populations and we haven't really come around to very closely defining them here. The UGDP is a clinic population, our out-patient clinic population in univer- sity clinics, perfectly suitable for the techniques appropriate there. The other study I was referring to was done in the field, in Anne Arundel 26 Technicon Instruments Corporation, Chauncey, New York a County, Maryland, by travelling hundreds of miles to get into the homes of particular people with a history of diabetes, and controls, There we found the capillary methods to be far superior, We didn't use thé -ear lobe, They didn't like that very much and they rather liked the finger-tip method, We used a heparin sponge to keep it going over the three-hour period so we didn't have to stick again, For different kinds of populations, different surveys, different methods are apparently the most suitable. So we have three methods: venous serum, venous whole blood, and capillary blood. DR, KENT: Yesterday I presented the fact that we had done 70,000 glucose tolerance tests using capillary blood plus some more in the confirma~- tory tests, I think Dr, Leonards could speak from a technical bio- chemical point of view to this very well, and I think he should at this time, DR. LEONARDS: I think before IL can say very much about it, that it is important to define, as Dr, Klimt and Dr, Acheson have stated, just what we are trying to doj that the methods that we have available and are possible for studies such as Dr, Fajans, or of the university group, are entirely different and require different methods than those that are used for multiple screening or those that are used if you want to do 100,000 people a year, Now, once we define which of these three we are talking about, I think I will be in a better position to discuss what the merits and disadvantages of the various types of methodologies are, because the three are going to differ, -122- DR. BURCH: Of course, there is another related problem on that, and that is that, hopefully at least, if you say an individual in any of these types of population is diabetic it should mean the same thing in each population. DR. LEONARDS: That is correct, but the methodology need not be and should not be the same. DR. BURCH: This is what I tried to point out yesterday and today, too. I think that there really is quite universal agreement that in a clinic where you have everything at your beck and call, you would use probably the most elaborate. You said you would need to have the population defined. DR. LEONARDS: No, what I am trying to state is: Why doesn't this group decide what they want to determine, namely, do we want to talk about a study like the university group? Is this what we are trying to settle? Do we want to talk about studies like Dr. Fajans' or Dr. O'Sullivan's on a few hundred or few thousand people? Or do we want to talk about studies that involve hundred of thousands of people a year? Once this is defined then we can talk about that particular methodology. DR. BURCH: My own feeling would be that there would be fewest calls for the surveys involving hundreds of thousands of people in a year. Possibly next in frequency utilization, those involving up to 1,000 or so. And probably the most common would be the clinic-type programs. -123- DR, McDONALD: I would like to bring the discussion back to what Mr, Woolsey mentioned a while ago, and that is, we don't have to have agreement, We can still have flexibility as long as we have correlation data between capillary and venous blood values, I would like to ask if the group is aware of any correlation data comparing the two methods and does it allow translation? DR, RICKETTS: Dr, Mosenthal 20 years ago, maybe 30, did such comparisons and found enormous differences between capillary and venous blood in some parts of the glucose tolerance curve and in some patients, but not anywhere near enough to establish a correlative value, This is my reason for stating if you are going to use capillary blood you have to establish capillary standards, MR, WOOLSEY: Well, if this is the case, then we may be faced with the situation that you are not measuring the same thing with these different tests, and if that is true, then you have really got a problem, it seems to me, I assumed in a naive way, being a kind of layman in this business, that all of these were intended to measure blood sugar and if they don't have some quite strong relationship to each other, right through the whole curve, then I think the whole field is in real trouble, It seems to me the calibration studies are needed in any case. DR, KLIMI: I would have to agree entirely with Dr, Ricketts, I think we need separate standards, and to translate the findings to each other in a few instances may be possible, like whole blood and serum when you say you have such good correlation that even individual predictions can be made, Between capillary and venous blood you definitely cannot do that. -124- DR. NEWILL: I think we ought to distinguish between a screening test and a diagnostic test. There is no reason why the capillary blood samples can- not be used for a screening test and a different blood sample, e.g. , venous blood samples, for the diagnostic test of the disease. DR. BURCH: Would you be implying, then, that one might sort of concentrate on the capillary system for a screening test, and venous for the other? DR. NEWILL: This is what I am implying. DR. O'SULLIVAN: If I may make one comment on that, again it depends on the objective. If you are going into a total population and want to come up with uniform data, it then matters if you use a screening test, because you give the whole population a screening test and, as a result, that portion of the population will have the opportunity of having a second test. You end up using blood sugar determinations with a segment of the population having had the opportunity of being tested twice. I realize people will say this is confirmatory testing but I can give evidence from the Oxford study that you produce data that are not uniform when you do this. This results from intraindividual blood sugar variability. If you give a person a chance of testing twice, you will in many circumstances be more likely to produce a positive result. So if you test your negative population, as has been shown in some of the studies in England, you will find a lot of positives that were initially missed. DR. KLIMT: Dr. O'Sullivan, I think one way to take care of this would be in this two-stage screening procedure which we have in mind now, to test a representative sample of the negative screenees with the more wl2 5 specific test you use the second time, and thus get an estimate of the number of persons missed, and adjust your totals accordingly. DR. ANDRES: I think it is important that we understand the physiological basis of the difference between capillary and venous blood, because this will help us decide which should be used. Capillary blood, especially as collected in Dr. Leonards' technique, is essentially the same as arterial blood and also in all respects (except for blood gases) as mixed venous blood, It represents the average of what all the tissues of the body are pouring into the venous system. It is in essence a weighted mean sample from all of the tissues. Venous Blood as collected antecubitally represents the impact on that "mixed venous sample", if you will, of the local metabolism of the forearm tissues, Furthermore, the concentration of various metabolites depend upon the vein sampled, If you collect blood from a superficial vein, the impact of the metabolism of skin and subcutaneous adipose tissue is seen} if you collect blood from the antecubital vein which drains the deep tissues, you see the impact of the metabolism of skeletal muscle primarily, So it may be important to know whether you wish to have the impact of skeletal muscle metabolism on your glucose determination or the impact of superficial or adipose tissue on the determination; indeed one could build a strong case for preferring the ''weighted mean'' sample, that is, capil- lary or arterial blood, ~ 126 If you are interested in the serum lipids or free fatty acids especially, then it should be realized that the difference between free fatty acid concentration in a superficial forearm vein and a deep forearm vein can be as much as 100%. So if one does a glucose tolerance test and also wants to know what happens to free fatty acid, and during the course of the test sticks sometimes one vein and sometimes another, then you can find enormous fluctuations which are really artifactual. Furthermore, I think it is very difficult to get large capillary samples, although I don't know just how much you can get, But if you are interested in doing anything other than glucose analyses on the blood sample you would be better off to get a venous sample, However, if you wish to do free fatty acids in a delicate study you really should get arterial blood. So the objective of the investigation is all important in making this decision, DR, BURCH: Do you know of any data, Dr, Andres, on comparison of blood glucose values on these two types of veins? DR. ANDRES: Yes, This is all discussed in a paper by Dr, Baltzan, et al,, called "Heterogeneity of Forearm Tissues," and so forth, published in the Journal of Clinical Investigation several years ago. Surprisingly, glucose concentration in superficial veins is signifi- cantly lower than in deep veins. Under basal conditions this difference is very small, but after glucose loading may be very large, DR. ACHESON: And Butterfield has been doing work on that, He looks at the blood sugar gradients between arteries, capillaries and veins. He has been doing this for some years now. =-127- DR, BURCH: I wonder if maybe we could get to a topic which caused a considerable amount of discussion and disagreement yesterday, It seems to me that it is one of the things which actually needs still more airing at this meeting. And that is in regard ‘to variation of the loading dose of glucose which is given, Dr, Klimt determines how much glucose 1s given by the body surface, Dr, Fajans does it by ideal weight, Others give the same dose to everybody, 50, 75, or 100 grams, depending on the investigator. Since there has been some evidence presented showing that with different amounts of load you might well get different results in the glucose tolerance tests, it seems to me it would be desirable to have a little bit better agreement than we have, I think it would be optimistic to expect too many changes in procedures here from this, but I think it would be desirable to have some more discussion on it, DR, CRAMPTON: Well, I would just like to second what Dr, Fajans had to say yesterday about this, that in the normal person, the size of this loading dose probably makes very little relative difference, But it is in the early or in the borderline diabetic, in this gray zone, that the size of the loading dose becomes very important, and it does require a maximal stress dose, What this maximal stress dose is undoubtedly deserves some discussion. There is another factor in this, and that is that adjusting this dose to weight does not lend itself well to mass procedures, It becomes very costly and impossible, That 1s, of course, why the Cleveland group went to 75 grams. And whether this is adequate is, of course, a subject for considerable discussion. DR, KENT: I would like to make a plea for Dr. Hayner to show a slide here on this general subject. I think he has some good data which I think everybody should see, loading times, loading doses, postprandial and non-postprandial blood sugar levels. DR. CRAMPTON: While he is getting his slide ready, I would like to say with regard to the University Group Study, I don't think the dose of 30 grams per square meter of surface area is adequate for what we are talking about here. DR. MILLER: That is a different purpose. DR. CRAMPTON: Yes, but it doesn't apply to this particular purpose we are talking about here. DR. LEONARDS: And may I repeat before you can talk about any of these factors, loading or no loading, time of loading, venous or capillary -- before you can discuss any of this methodology you have to decide what kind of study you are going to do because it will be different for the University Group, for Fajans' group or for mass screening. Let's not get this confused, please. DR. BURCH: I think it is salubrious that we have repeatedly brought out here the fact that we do have these three different types of population studies to contend with, and that we do have to think of them somewhat separately. However, I still think we need to be able to at least have some idea as to what one means in relationship to the other. -129- DR. KLIMT: It is not only the difference of size but the difference of site and the difference of scientific purpose. If you do a field study versus a hospital-based study, it is different. DR, HAYNER: I think Dr. Klimt is reemphasizing an important point, that the research objective is the critical factor on which to decide what type of test procedure may be necessary for a given situation. In the Tecumseh study we are using a test procedure which is essentially the same as what you described, and for other purposes have a screening procedure. But we don't think of this as a screening procedure. In- stead, we speak of this as obtaining a measure of a physiologic variable, We use this not for screening and for then determining whether a person has diabetes which may need clinical attention -- that is left to his own physician -- but rather we use this for classifying individuals, even though there is considerable variability. The procedure we use is to accept people in our clinic in Tecumseh for multi-purpose examination at any time of day that they can come. Participants arrive, then, fasting or not, and the majority are not. They are given a glucose challenge shortly after arrival, and one hour later the blood specimen, which is obtained for various purposes, is collected. The portion which is fluoridated is analyzed in our first round of examinations by the Somogyi-Nelson method. The analyses we will show you now are based upon the first round data. These are frequency disteibutioas® of the one-hour blood sugar level according to sex, broad age ranges, and prior food history. The solid lines represent persons who had taken food in the last four hours; 276». Hayner, et al,, Diabetes 14: 413-23, 1965, fig. 1 -130- the dashed lines represent frequency dis:i. butions for those who had not eaten in this period. You will notice quite consistently that the latter distributions have central values about 25 mg. percent higher than those for the former. DR, KLIMI: May I point out here that not only do you have an age shift, but you have the difference between the sexes, which was referred to before. Each time you can see that the mean value must be greater on the female side. DR, HAYNER: We do not observe a significant difference between the sexes from any of the ways we have done our statistical tests on these data, Dr. Klimt, It may appear so, but if you will try you will, I think, agree that there is very close correspondence between the response of males and of females in this community who are tested under like circumstances. But there is an age shift, as you point out. The next sue? shows the change of the mean values with age. We found this, computing it by decades of age, to be quite linear from the late teens through at least ages 60 to 69. I have shown here the values for those with and without recent food separately. In the eighth decade there were very few persons who had not eaten recently, so those values are not very firm. The next slide shows the median and the 20th and 80th percentiles. You will see that the median increases linearly with age. 28; 1ids Idem, figures 2 and 3 295s cit: Idem, figures 2 and 3 -131- Here you can see that there is some fanning out of the values toward the higher range with increasing age. The standard deviation does have a significant increase with age, but its increase is rather moderate and is not sufficient to prevent us from using a regression approach to the data. Now let's examine the effect of the interval since last food in detail on the next slide. DR. BURCH: With the known diabetics taken out? DR. HAYNER: Yes. The persons who had previously known diabetes were not challenged. There were 68 during the period of time after introduction of this procedure and these slides are based on 2,983 observations in which the test was properly performed, which are about 76% of a repre- sentative sample of all examined adults. DR, MILLER: It is quite evident, though, that in that group you had definite diabetics because some of those blood sugars were as high as 300. DR, HAYNER: It depends on your approach to what you want to call diabetes. Note that the distribution curves shown before do not reveal bimodality. We prefer to deal simultaneously with as much of the distribution as possible. First we did this covariance analysis comparing linear regressions for each one-hour-interval group of persons, timing from the last food to the beginning of the test, You notice that the males that fell within the four-hour period coincide very nicely, and those who were beyond this period seemed to cluster together. 30rb14d: Idem, figures 4 and 5 -132- The females do show a significant variability even within the four-hour period, but this analysis reveals only small differences at the approximate mid-point in the distributions, at age 40, which are minor in comparison with the difference between those who had eaten within four hours and those who had not. The next slide shows a similar analysis for the effect of the amount of carbohydrate in the last four hours, according to classes estimated from the food history. There is here a significant variation in the direction of a fall in blood glucose values with an increase in the amount of carbohydrate, but here again the degree of variation is not very great. It is just perceptible in the males, somewhat more apparent in the females. It is possible women know better what they eat and when. DR. BURCH: Did you use standard loading amounts of glucose? DR. HAYNER: We used 100 grams as a standard loading dose. We did not vary amounts in adult members of this population. That was examined in the prisoner study. I do want to make a point showing an application of this procedure, if T may. We sorted on variables by percentile classes, this being one approach that can be used to take care of the age trend on these data, and we did this for each variable and here we do it against a disease. On the next slide, from a paper presented in Kansas City, at the APHA Myvi Idem, figures 4 and 5 - 133 - meeting, by Dr. Epstein, you will see that one-hour glucose tolerance was associated with coronary heart disease in males at age 40 to 59, a P value of less than .05, and at age 60-plus a suggestively more signi- ficant correlation. In the females this did not show up. There was an association with hypertensive heart disease in the females. There was, interestingly enough, no association between relative weight and glucose tolerance. DR, BURCH: I think, though, this is extremely interesting, but we are getting far afield from the topic. If you have something showing the variations from different loading doses, this would be pertinent, DR, HAYNER: Are you familiar with the procedure used in the prisoner study in Milan? I can say briefly each subject was given 16 tests in a Latin Square design, randomly assigned. There were 24 subjects. There were two tests with each of eight procedures. Concentrating on the right half of the slide, 33 which contains the same data as the left, you will see that the 50 gram challenge, whether given one hour, two hours, or three hours after a meal, produces a one-hour value just slightly above 100 mg. percent. And if given after an overnight fast, the value is slightly higher. Peculiarly, the distribution in response to the 100-gram challenge shows a moderate step-wise increase between one, two, and three hours. DR. BURCH: When you speak of "meal" here, what is this? 32 stein, et al., AM, J. Epidemiology 81:307-322, 1965. 3 vayner, et al.,, Vital and Health Statistics, Series Z, No. 3, Washington, U. S. Government Printing Office, 1963. -134~ DR, HAYNER: The meal was either breakfast or lunch and one-half of the subjects had all of their non-fasting tests performed after either one or the other, not after both, We couldn't include a compar ison between meal times, DR, KLIMI: Was the baseline the same, the fasting? DR. HAYNER: The fasting procedures were under the same conditions, DR, KLIMI: Average levels are not recorded at fasting for the tests. DR, HAYNER: These compare the values at one hour, If you want the numbers =- MR. GORDON: The answer is there was no difference, DR, HAYNER: You realize this is a collaborative study between ours and the National Health Survey. DR, MILLER: This is different from the results you got in the field study? DR. HAYNER: Yes, it is different and I would suspect a large field study would be a little better place to determine things of this sort than a group of prisoners who are under a peculiarly different diet from that in the field, and who have average values that are surprisingly low. These are men 40 to 54 years of age, DR, ACHESON: Is this whole venous blood? DR, HAYNER: This is whole venous blood and it was shipped on ice, but not frozen, to Dr, O'Sullivan who did the determinations of the values. DR, KLIMI: Isn't it peculiar that the average value for the third hour is the highest of the lot in the 100-gram challenge? -135- DR, HAYNER: Yes, this is peculiar since we didn't find this in the behavior of our community. This was an observation in persons who had been subjected to challenges only after lunch, whereas the behavior of the 100-gram challenge after breakfast did not have a _ step-wise increase, DR, KLIMI': But these two are not comparable. DR, FAJANS: TI think the question before the house is: Is one loading dose more effective than another one in disclosing mildly diabetic patients, whatever our definition of "mildly diabetic" will be? I think the answer cannot come from merely testing normal sub- jects or even looking at a population. I think the same challenge has to be given to normal subjects as well as to mildly diabetic patients to ascertain whether there is better discrimination between the normal and abnormal group with one dose versus another dose of glucose. So far we have data only on what differences are in the results when various doses are given to a so-called healthy group. I think it is in the borderline group where we need data. DR. BURCH: I rather surmise from yesterday and today's conversations that these data have not really been collected adequately to show how many individuals, for instance, may be misdiagnosed, or may be con- sidered normal if too small a loading dose is given. Is this true? DR. LEONARDS: We have made some comparisons between 100 grams of glucose and 75 grams of the loading solution we are using, on a group of individuals that did not have diabetes on their records and who said they were not diabetic. Those whom we found to be diabetic are - 136 = just the type of individual that Dr, Fajans is referring to, Among these subjects there is virtually no difference between 75 grams and 100 grams, There obviously is going to be a difference between using 150 grams as compared to 30, and I think we ought to be careful when we are talking about different loading doses to speak of what range we are talking about, But I think this answers your question, DR, ACHESON: These are averages of a series of people? DR. LEONARDS: Yes. DR, BURCH: Seventeen individuals; is that right? DR, LEONARDS: That is right, Dr, Kent showed yesterday we had some~- thing like 20 or 30 normal individuals in which the two average curves were virtually identical for practical purposes, The P value is .05, just at the borderline of statistical significance, DR, FAJANS: I agree that the difference is small between the 75 and 100 gram doses of glucose, but the point is that there is a difference of 15 mg, at two hours which is one of the critical points, The difference you say 1s at the limit of significance but even with the small numbers - 17 = you had a P value of .05. So I could use your own results to say that there is a difference between the 75 and 100 gram loads, particu- larly in individuals who have such borderline curves as this one. The peak blood sugar level isn't over 160 mg, per 100 ml., so that even by our own criteria we could not make a diagnosis of diabetes although I agree it is an abnormal curve, I think one can interpret the results that there is a difference, - 13 7- DR. LEONARDS: There is a difference, but is it of practical importance for screening purposes, and this 1s what I thought this meeting was about, DR, O'SULLIVAN: I would like to make one comment, if I may. We have been talking about the effect of the size of the challenge, I just want to remind people that we have here a single glucose solution containing 100 grams and that it is being compared with a carbonated liquid con~- taining the equivalent of 75 grams. If we are considering the size of dose, it is important to recognize the additional variation introduced by the effect of the carbonation on gastrointestinal absorption, MR, GORDON: With respect to the Milan study, I thought the most interesting conclusion from my point of view was that while there is a difference in the level of response, depending on the size of dose =~ and I don't think there 1s any dispute about that ~~ when you ranked individuals, if you set up a scale for each dose, that individual would be ranked in the same position on the different scale, Now, this may not be true of certain classes of individuals, but I think that would have to be demonstrated, because there was certainly no evidence within our group. DR, ACHESON: Can I make more acutely a point I made too obtusely yester- day, namely this, in my opinion, 1s not the way of handling serial data, These are serial data. These are sets of information =-- pieces of infor- mation about 17 individuals, and each of whom has his own individual curve, If you average each of those points, you are losing all the information you have about the serial component, You might as well get different people for each of these points and connect the points, or in other words your data might as well be cross sectional, 138 = What one must do to get full information on this kind of thing is to look at each curve individually, to include them all in toto. It may be they will superimpose but in fact we know they don't, We know the correlation co-efficient is in the order of .6 or .7, so it means there is a large variance between one point and the next which isn't shown here. DR. BURCH: This would be bad enough with 17 cases, but think if you had 17,000. DR, ACHESON: We all shudder at the magnitude of the task, but unless we do something about it we will not find the answers we are seeking, The best illustration I can give is to look at longitudinal growth data of children, We all know that boys come to puberty after girls, but within each sex puberty growth spurts may range over a period of about six years, But you must look at the growth of each individual child over a period of some years before you can determine this, To return to glucose tolerance curvesj if we look at these as a set of half-hour, hour, and hour-and-a-half averages, I am sure we lose a lot of information. The alternative -- considering individual curves =-=~ gets us into a complicated field, but I think this is what we have to do if we want to find the answers we are looking for. MRS, FISHER: This was reflected in our prison study quite well, We found if we gave these men six 100-gram tests over a period of a year, and after that two 50-gram tests -- if you look at group means they are very stable, In other words, you can use group means to describe the total group. -139- But we did not find what Mr, Gordon said they found in the Milan study, that the rank correlation was good. Actually, our rank correla~- tion 1s very poor, This means that while your group curve is staying the same your individual man is often jumping from one end of the curve to the other end of the curve, and this is the type of thing I think Dr, Acheson is speaking about, You conceal this from yourself when you combine such values, Now, we haven't done a full analysis of the difference between the 50 and 100, but there is a difference which I think will eventually have to be considered in evaluating it, We did not have abnormal people ~~ supposedly they were not abnormal in terms of any presence of diabetes, But we found that there was not much difference in your mean value at fasting, one and three hours between the 50 and the 100-gram tests, but there was a difference at two hours, about 10 mgs,, with the 50-gram test lower than the 100-gram test, Another interesting point is that the variation around the mean tends to be lower at 50-grams and at three hours it is almost back to what it was at fasting with the 50-gram test, This is not true of the 100-gram test, Your mean value is back to fasting but your group varies considerably around that mean, DR, KLIMI: I would like to recall the second sable presented by Dr. Hayner just to make one point, namely, we are talking here about the amount of the load but we probably are not considering the effect of the same load on different people of different body size, and while it might not be statistically significant you do have the difference I was pointing out here, 34 C, P. Hayner, et, al,: Diabetes 14: 413-23, 1965. fig. 2 Look at people aged 20, male and female, where you can surmise the females have less body surface and are smaller and lighter and you have quite a sizable difference between the averages, One is 100 and the other about 115, And this is consistent with several studies, this may be due to the difference in obesity. In this case I would assume it would be due to the fact that the females and males were both challenged with the same amount of glucose which is a harder challenge on the smaller female. DR. HAYNER: We don't find the difference you are trying to point out. You are taking the left-hand point, the group representing ages 16 to 19, I believe; is that right? DR. KLIMI: You have a point at 20 here for males and females and you will find that the females are higher than the males, The same is true for the dotted lines, DR, HAYNER: If you superimpose the whole curve they correspond quite closely, DR, KLIMI': They may have the same slope but they can't correspond, DR, HAYNER: I will show you the tables, DR, KLIMI': This is quite consistent, You are giving females the same challenge as you do males, and on the average the females will have less body mass and will be more severely challenged with the same amount than the males, MR, GORDON: So far as the data from the Health Examination Survey, where we gave a challenge of 50 grams uniformly to all persons, we found the kind of sex difference you speak of. I don't know why Dr, Hayner didn't “141 - find it, but apparently his group responded differently or there may be some peculiarity implicit in one or the other study which either reveals or doesn't reveal this kind of difference. But we found a difference between men and women at the same age of about the same order of magnitude as you record, DR. KLIMI: The same challenge produces automatically a sex difference, because you use the same cut-off point for both sexes, but you are challenging the female more than the male, MR, GORDON: I just want to comment on this, I think it is unlikely that the order of magnitude of the difference that you are describing here could be accounted for by the difference in the body mass, because the average load which is given to women by your formula wouldn't differ that much from the average load given to men, What I am saying is that if you were to give women in your study the same average load that you gave men, you would not raise their level to the degree that we are talking about, I don't think there is that much difference, And so far as Dr, Haymer's study is concerned, what I think would be implied by your remarks is that in fact, if you did give women the same challenge in terms of body mass as you gave men, that women would in fact have lower blood glucose levels because giving them a higher average challenge apparently gives them the same blood glucose levels, DR, BURCH: This would sound as if the field study might even be compound- ed still more by havingto give different amounts of glucose, and I rather agree, myself, with the Cincinnati group, that it is better on -142- a large-scale project like that, and also like the National Health Examination Survey, to keep it uniform, This, again, may be one of these aspects where, when you are in a clinic with scales and heights and all that readily available, it might be feasible to adjust the dosage to a greater extent than in a field survey. However, I think that I would like to get back again to getting some views as to whether or not a 100-gram loading dose would cause enough difference in the response in this gray area from the 75 or even the 50, as used in the National Health Exam Survey so that the recommendation should be made in favor of one or the other of these loading doses, particularly in a field type of study. DR. RICKETTS: It is evident that our discussion about loading dose is purely a matter of opinion. I think this has to be recognized -- a matter of intuition, I don't see why we fail to tie it to the only hook we have got in the way of measuring general metabolic function, namely, the surface area in relation to oxygen consumption. This has been well documented over many years, It is the only general metabolic measurement I know of which has been shown to correlate one function with another -- body surface area with oxygen consumption, basal metabolic rate. I would be strongly in favor of adopting Dr. Klimt's practice of using surface area, perhaps enlarging his dose a little bit, but applying this to a field study. -143- I can't see any mechanical objection to doing it. You can carry little scales with you that are not too big, and carry a tape measure and put the man against the wall and measure his height. With the nomogram this is a very small business. MR. GORDON: I just want to comment on this issue of practicality. First of all, if all you are doing is studying diabetes, I think even in a field study you can do more than if you are studying a variety of things. We chose a 50-gram load not because we felt it was an optimum procedure -- and we are not sure what an optimum procedure is -- but we felt we would have less difficulty using a 50-gram load than 100-gram load with the risk of nausea complicating the examination. The survey was principally for diseases other than diabetes. In varying the load, I think you underestimate the logistical difficulties of this because, even with the fixed 50-gram load given in the circumstances of our examination, we had some difficulty main- taining the 50-gram load. If in fact you are going to have to vary the challenge and be sure that a person is given the appropriate challenge according to your nomogram, I think you are going to have, in a field study, a great deal of difficulty logistically in making arrangements to do this. If all you are doing is making a diabetes survey and you are willing to devote enough personnel and enough time and attention to just this kind of question, the varying dose, and assigning it correctly to each individual, then I agree it could be done. But I think there are other circumstances where it simply is impractical. -144- DR. LEONARDS: I would like to speak to Dr. Ricketts' proposal. I think it is an excellent one if we again address ourselves to the purpose of the field study. If the purpose of the field study is a University Group Diabetes Program, excellent. If it is Dr. Fajans' study, excellent, This is the way to do the thing. But if you try to do this on a mobile unit parked downtown with a thousand people walking through the thing in a day, or in the middle of an industrial plant where you have to catch people coming off a shift, or under circumstances of this nature, it is just absolutely impossible. So again it depends upon the type of study you are doing whether or not you can follow your suggestion or not. DR. KLIMI: This is certainly the case but let me point out that every study will want to look at certain differences. Are males different from females? The results seem to show twice as many female than male diabetics. Is this an artifact or is it true? We have looked at ethnic groups and age groups. All these comparisons depend on the comparability of the test. And if we do a survey which has this in mind, I am afraid we will have to pay attention to it. If our idea is to skim off the top the majority of cases in a case-finding drive, obviously we will have other objectives. DR, KENT: Dr. Ricketts, we do, I would say, 75% of our studies so far in industry, and when I meet with the industrial heads, they ask one question, "How many minutes is this going to take off my line, and how much is this going to cost?" - 145- You can't stand people up along a wall or do other things except move along. We try not to have a delay of two people queued up in a lines; We try to get it really expedited and we have done it and that is why we have been successful. He says, "I pay this man $3.50 an hour and he is away 20 minutes," and he multiplies this against 1,000 employees and you have had it, This is what we are faced with, I am not saying you have to be inaccurate but we have been trying to screen those with diabetes or suspected diabetes to the best of our ability and we have tried to do it with a confirmatory test. We have tried to say this is better than throwing this man out to a physician and saying, "I think this man has diabetes, Prove it.'" We either identify it accurately or not at all because the minute we make a false identification we have trouble because they say, "They cost me time and worried the devil out of me." They call us up at night and say, "You identified me. You said I had diabetes," We say, '"No, we said you had an abnormal glucose tolerance curve," That doesn't mean anything to them, We have a problem of trying to pick out of the population all those, as far as we can identify them, so they can be studied in depth. The minute I hear anybody say ''venin- puncture," I run and hide under the covers because people don't like to be stuck in the arm, They would much rather be stuck in the finger than stuck in the arm, The second thing is we have to train technicians and we have a big turnover. If you don't think it is difficult to train somebody to find a vein in a fat woman, you haven't been around doing these things, because = 146° it is tough and it takes a year or two to train them, If you have someone plunge a needle into the arm two or three times and the woman screaming, that is the end of her and the end of all her neighbors, DR, MILMORE: I can't agree more with Dr, Kent's comments on practicality, but at the same time I don't think we should be slaves to practicality, I do agree fully with Dr, Ricketts that we are floundering in the dark, We don't have the data necessary to make an intelligent choice as to the glucose load for screening purposes, I think all these considera- tions have to be taken together, along with Dr, Fajans' very important point that the variation with reference to glucose load is one thing in nondiabetics and another thing in the very diabetic we are trying to detect in a mass survey, There 1s a great need for specific studies to determine this point, Without the necessary data we are merely guessing, It may be true that our efforts in detection of diabetes will be most fruitful if we use an arbitrary glucose load such as 50 or 100 grams, But we do not actually know that this is so, If we use the body surface or some other approach to glucose load, we may be coming up with a much better test. And only if we have this information can we make an intelligent choice and see to what extent we are gaining something by sacrificing practicality. DR, WADA: I can realize that Dr, Kent and Dr, Leonards' method is simpler, So far we have weighed the sugar according to Dr, Klimt's method in the community. And we have been, at least to some extent, successful, but after trying that method we have found that using a solution prepared =-= =147- and measuring the amount -- is more simple than our previous method. And this will be a better method if we use glucose dosage according to body surface, because weighing the sugar is much more complex, And {f we use a solution and a measuring cylinder, it is much more simple. But I quite agree that Dr. Kent's method is more simple. DR, BURCH: As Dr, Leonards has pointed out repeatedly, it depends on what we are trying to do, and it does seem to me that the type of surveys that they are doing are really quite distinct from general population surveys and the clinic type of surveys, and it might very well be appropriate to decide to use one method on the screening detection surveys and another on the more complete population surveys. MR. WOOLSEY: I would support strongly what Dr. Milmore was saying, IL think this goes along exactly with what I was saying when we were talking about where you take the blood, Let me make one additional point. Through such study, if carefully conducted, you can also determine whether there are different standards that you want 2 sek. In other words, you can look at this thing from two standpoints, One is whether there is any optimum loading dose. Also, you can look at it from the standpoint of: Should we be using different standards for different sexes and different ages, and so on? And the kind of study that he is referring to could throw light on both of these questions, Because, if you found that different standards were suitable for males and females, for example, then it might be possible from the standpoint of practicality to have a uniform loading dose, and yet use different standards for the different sexes. -148- DR. KLIMI': Surely. DR. BURCH: Actually, of course, this would not be the first disease in which this system was used. With hyperurecemia, you have different standards for males and different standards for females, and for various other diseases, too, DR. CRAMPTON: I was just going to try to point this up, In reference to the Cleveland study, you are talking there about screening and I thoroughly agree with what they have said from my own experience, It seems to me the answer is the standard dose in the screening procedure, but when the patient is brought back to the laboratory for the specific definitive study, then is when he should receive the dose related to his surface area. That is when the diagnosis is made, the second step, DR, SHARKEY: I have been listening to this for a long time and it seems as though we may have some crystalization of thought on glucose load for screening purposes, namely probably not less than 75 grams and not more than 100. That seems to be the general trend of the conversation as far as screening levels are concerned, DR, ANDRES: I have a compromise suggestion that I think is very simple, but would meet some of the objections of giving a single dose, and that is to use three different bottles of soda pop, say, and in accordance with Madison Avenue custom, the small one would be called large, and then "middle" and '"jumbo' or something, But if you consider how people vary in terms of body weight or surface area, there is about a two-fold variation, We won't modify the dose if you get up above 220 pounds or something like that, So if you ~149~ give a middling dose to everybody you could be off by 50% for the little guys and big guys. But if you divide everybody up into three groups in a very simple way you would reduce that variability from 50% to about 15% and it would still mean that you would have to store only three different sizes of glucose or three different sizes of soda pop bottles, DR, BURCH: This certainly sounds like a practical way out. DR, VOUGHT: As a complete tenderfoot and non-diabetician, I would like to make an observation here after two days. Number 1t It seems to me I will go with Dr, Leonards, namely that the test you use depends on what you are trying to find out, what are your objectives, Number 2: Then it is up to the investigator to decide what test he is going to use which will best meet his objectives. Now, if he is any kind of an investigator -- and I am sure you all are =-- he will know how accurate his test is, And if he has done this and it will satisfy the requirements of his study, he had done it, I don't see how you can generalize from population to screening test. It all depends on what you are trying to find out. DR. BURCH: Yes, this is of course true, However, it would be nice if different studies which are attempting to find out comparable things used comparable procedures, DR. VOUGHT: How comparable are they? DR, BURCH: I am just newly into the diabetes racket so I won't comment on it, but in most of the diseases in which I have worked, the reported “1507 differences in prevalence and incidence of a disease, I have always felt, are based primarily on differences in the investigators rather than the disease, That may be the case in diabetes, also. DR, WADA: TI have data which was done along with Osaka and supported by The National Institute of Arthritis and Metabolic Diseases. In screening, we have used urine obtained at the site of the examination two hours after the last dinner, And we have found a very high per- centage of urine sugar, The percentage of urine positives in Yao concerning people from age 40 to 64. We have obtained urine at site of examination, And the percentage was 3,9 in the first line. And you can see that for people over the age of 40 in Tajiri the percentage was as high as 12.5. If we use a urine specimen collected two hours after the last dinner, the percentage of finding new, unknown diabetics comes high. Of course this should be proven in the future, but my opinion is if you challenge with glucose or some other drink after a heavy meal, and collect urine, then the percentage will be quite similar to that obtained by Dr. Kent by the blood technique. DR, McDONALD: There is a little-known study that was done in Sullivan County, Missouri and hasn't been published, about urine testing as a screening device, They took histories in addition on these people, so if they either had positive urine -- I don't know what was the level of significance -- or a positive history related to a diabetes or having had a heavy baby, they found as many diabetics by glucose w]5]~ tolerance tests later from a positive history as from those who had a positive urine, So it might arrive at the same percentage, but are they the same people? Because a publication of some years ago indicated that at all levels of specificity, blood is 50% more sensitive, DR, WADA: I would like to show the last figure on my first page, If you take the data obtained by other doctors and me previously, it will show that blood sugar is better as a method of screening. Taking people of 60 to 69, it is 14.7, and the urine was 5.7. But this was taken at the site of the examination, And if you collect urine after a heavy meal, the results will be different. We have studied this point but at present I have no answer, We must be aware whether we take a morning specimen or we take a specimen after a heavy meal -- probably Dr, Kent and Dr. Leonards have clearly shown that if you give glucose after a heavy meal, the results will be quite different. And by doing this, the necessity to take blood will be reduced, DR. BURCH: Actually, there is a matter of comparability which certainly is not limited to diabetes which I would like to comment on. I mentioned that I had worked on various other diseases, several dif- ferent tropical and parasitic diseases, and arthritis and rheumatism, and in all the diseases with which I have worked I have been plagued with one matter of lack of comparability which makes it difficult to compare one individual's data with another, even when they use com=~ parable test procedures, and that is the way in which the results are reported, -152~ This is possibly more in the province of Dr. Woolsey and Mr. Gordon to mention, but it makes it extremely difficult to evaluate and compare different surveys when the population is divided into age groups which are not comparable. There have been studies which show that with people's age, just the same as with blood pressure, there is a concentration of the blood pressure readings around the even numbers; the age there is a concentration of ages at the ten-year age-group intervals. And because of this, the World Health Organization and our own Census and the National Office of Vital Statistics -- if they use ten-year age groups -- straddle the ages ending in zero and go like 25 to 34, 35 to 44, and so forth. If everyone used the same groupings, it would facilitate things in comparing one survey with another. For instance, the Tecumseh survey, I noticed up here, was like from 40 to 49. This is extremely common. I notice that Dr. Wada has done the same thing on one page and on the other page he uses a com- bination. He uses from 40 to 54 and 55 to 64. Actually, in my own work I have done the same thing, too. I go from 30 to 34 and then I go on to the ten-year age groups. DR, MILLER: Is this a specific recommendation? DR, BURCH: This is a specific recommendation of the World Health Organization, the United Nations -- I forget who-all else -- and our own National Office of Vital Statistics. I believe that is right, MR WOOLSEY: We usually use age groups of an overlapping ten-year span. DR. BURCH: But this is a definite recommendation of the World Health Organization for epidemiological studies. ~153~ DR. MILLER: Have all the editors of the journals been alerted to this? DR. BURCH: This is not a new recommendation, This recommendation is 20 years old or almost as old as the World Health Organization, DR, KLIMI: This has application mostly to the countries where you estimate age instead of knowing it, and the human mind centers estimates around zeros and fives, more on zeros than fives, but both have always peaks. You will find the same tendancy in the recording of the tuberculin test, peaks on zeros and fives. Where you have measured ages, birth records and so on, this need not apply. DR, ACHESON: I think it is of even older origin than that, There was some work done in Britain before the World Health Organization was in existence which showed a preference in the female sex for the nine at the end of each decade, r MR. WOOLSEY: With computer methods now it is getting to be much more common to record the date of birth and date of examination and compute the age in the computer, If you have two single years of age in your data, it is possible to retabulate the data, you know, to different age groupings, if you are interested in comparing results from one study with another. So I don't think this 1s quite as serious a problem as some of the others we have been talking about today, There is usually something you can do about it to make the comparisons, Furthermore, you can plot them, SYNOPSIS AND DISCUSSION MAX MILLER mp, At noon today I had the impression I was going to have to tell you exactly what we know about this whole field of diabetes and to clarify the answers to the many questions raised, Obviously, this is an impossible task for it should be clear by now that much more work needs to be done before clear-cut decisions can be made, Let me discuss very briefly with you my impression of yesterday's presentation. One impression I got was that we must think in terms of epidemi- ological principles, and that we cannot solely use the clinical approach, "We saw five cases and made some interesting observations, which we be-~ lieve are worth reporting in order to give us certain leads,'" But we have to go one step beyond that in our present scientific age and come to grips with the problems about which we have talked so long and so wordily, We will never get answers without obtaining adequate data, without obtaining all the facts, Once these are obtained, it should be easier to draw the proper conclusions. Dr, Fajans' most elegant presentation was most helpful, He simplified the definition of diabetes by describing it as a genetic disorder, and I think none of us can oppose this or find any fault with it, He also gave us an excellent outline of the natural history of diabetes, which would be worthwhile to take home for teaching purposes, The four groups he presented, 35 School of Medicine, Western Reserve University, Cleveland, Ohio “155~ the pre-diabetic, the subclinical, the latent, and overt, were logical divisions, and he gave us some idea as to how these might be defined by certain tests, Dr, Fajans made the statement, with which I am in agreement, that at the present time the glucose tolerance test is the most sensitive test available for diagnosing diabetes, A plea for some other way of diagnosing diabetes could be made. If, for example, a mother has delivered a ten-pound infant, there are no data to indicate that she will inevitably develop diabetes, Statistically, of course, she has a good chance of becoming a diabetic, but at the moment in time when she delivers that ten-pound infant we cannot say she is diabetic. She can only be a suspect, It has been emphasized that the diagnosis of pre-diabetes can be made at the present time only retrospectively. After the diagnosis of diabetes has been made, you can go back and say, "Twenty years ago she was pre-diabetic because she delivered a large baby." He indicated, also, that the cortisone-glucose tolerance test perhaps may anticipate diabetes, He said "may anticipate" I put the perhaps’ in because this would require the careful serial studies which he and Jerry Conn have already initiated to determine the pre=- dictability of this test, He then very systematically reviewed his data on glucose tolerance tests in various groups. He demonstrated, as others of you have, the effect of age on the average blood sugar levels of glucose tolerance tests, Part of our problems relate to the methodologic area, and we must -156- separate these in our thinking from the clinical and physiologic problems. For example, we must define somehow which glucose determination is the best. We may have to define according to its specificity; we ought to define it according to its accuracy, reproducibility, etc, We ought to know what it would be in expert hands, such as Jack Leonards, who has done countless blood sugars, carefully and precisely as would be expected from a biochemist, On the other hand, most of the studies that concern us here are not carried out by biochemists, It is necessary, therefore, to know the accuracy of the determinations of the laboratory doing the study. This kind of information should be described in each publication. I see too much of it in the articles I review, "The method of 'X' has been modified" without any explanation why the modification is an improvement or without giving the details of the changes. We all recognize there is probably some effect of the load on a glucose tolerance test, But this wasn't defined in quantitative terms. Some of the experts that I consulted, like Mrs, Fisher, assured me this was not available in the literature, and I think she is probably right, I think there have been attempts in the literature, half-hearted attempts, spotty attempts, in which it has been indicated that there is an effective load. But where is the paper which was done as follows: "We have X individuals, We gave 1,75 gram per kilo, 1,0 gram per kilo, 0.5 gram per kilo to the same individual at different times, and all these tests were repeated to determine the variation in reproducibility"? I am sure we want to do this in "normal individuals," and people who are borderline diabetics or mild diabetics to see if the particular state -157~ of metabolism with regard to carbohydrates has any effect on the reaction to a different load, because we may be interested in this borderline group and perhaps the effect of those might be more apparent in this gray zone group. The effect of load is important, because perhaps ideally, if we gave five grams per kilo in a half hour we would really find out which patients had three islet cells functioning less well than they ought to. But we know there are certain practical aspects we have to consider. Particularly important is the question of how much we can give to a patient. How, we have done glucose tolerance tests on our medical students at Western Reserve and we know when you use 1,75 grams per kilo in the concentration that Steve has told us about, a certain number of them get sick, And they don't like us as teachers because they think we are making them unhappy, and that isn't the spirit in which the new Reserve curriculum has been created, The other factor is concentration, We all know if we put a hypertonic solution of glucose in the stomach it will not be absorbed until it becomes isotonic and if a lot of fluid gets into the stomach and distends it, it perhaps becomes a factor in determining the rate of absorption and the blood glucose loads obtained. What about capillary versus venous blood? Should this be reviewed again a little more systematically? What about this interesting point that Dr. Andres has made, that even the venous blood now is suspect; we have to be careful which vein we get. -158~ That immediately creates all sorts of horrible thoughts, and I think of those fat people we examine in the clinic and how difficult it is to get a vein, Actually, we are satisfied to get any vein, even if it is one on the wrist, Regarding the effect of previous diet: We have the work of the Michigan group, which emphasizes the importance of previous high carbohydrate intake, and yet Wilkerson has come out and casts some doubt on this, Do we already know the answer? Or do we need to do more studies to settle this point? On age, I think we now probably have sufficient data but here again I don't know what happens if you would examine the same group of individuals with time. This is the point that Mrs. Fisher emphasized quite elegantly, that there may be some individuals whose glucose tolerance test remains normal right up to age 92, and that others as they become older, as the islet cells become more tired or that more antagonists are present or what~have~-you -- that these are the ones who are veering toward the diabetic state. And when you average this whole group naturally you will have a higher '"normal' level. But I refuse to accept the statement that it is more normal to have a higher blood sugar as you get older. This doesn't make physiologic sense to me, And then, of course, there is the important problem of reproducibility. We do a single test, or as Dr, Kent does in Cleveland, we do two tests. Obviously, two tests are better than one, but how much better? What is the biological variation? I was very happy to hear that the Bureau of State Services had examined this in at least the prison population =-- that is a real captive group. =159~ And the effect of environment =~ I know it does have some effect and these things should be controlled. And then there is the problem of presentation of data, Should we present all of the data? Should we make the journals publish it? Should we bury it in the desk of some investigator who might move to some place else and when you write him ten years later and say, 'Can I have the original data?" he will ask, "What data?" Or should we start in Washing- ton? How are we going to retrieve it? I think Dr, Lazarow amply demon- strated that we can now store these data and retrieve them easily. If we are not going to present all the data, what point should we use? Dr. Fajans discussed the problem of using three points or two points or one point =~ if you had a choice, which points would you pick? And this certainly warrants further scientific discussion, Or if we wish to make matters "gimpler' should we have a single parameter? Should we add up all the values or should we determine the area under the curve, either with a planimeter or employing the methods of calculus? There is the problem of deciding which test should be used for the detection of diabetes. At the present time, we need not be content with urine testing alone, In Cleveland urine testing was used originally because it could be done on a large scale and because it was impossible to do blood sugars at that time in such numbers, But Leonards and Kent have shown that this is indeed feasible and I believe the answer to Dr. Alpert's question regarding cost is that it can be done quite reasonably. There is the epidemiologic type of study which involves going into a community and asking, "How many of these people have undiagnosed diabetes?" The logical question then arises: 'For what purpose?" -160~ Perhaps the onset of symptomatic diabetes can be delayed by early treatment but I am not sure whether control of carbohydrate metabolism alone will prevent vascular complications, I am hopeful that the UGDP study will provide some of the answers because it is set up specifically to determine the rate of onset and rate of development of vascular com=~ plications in newly-diagnosed stable diabetics treated with various hypoglycemic agents and including a control placebo group. A long-term study should be initiated to determine the best criteria for diagnosis. I emphasized this yesterday and I will say it again, There is no point in anybody saying that one glucose tolerance is better than another for diagnosing diabetes because we really don't know which is better at this time, A population group must be followed serially with various glucose tolerance tests, The development of clinically evident diabetes and of vascular complications could then be correlated with the results of the initial tolerance tests to assess specificity and sensitivity. And this brings up the last point and I am not sure that all of the speakers have clear concepts of what is meant by 'controls," I was interested to hear Dr, Fajans say yesterday that some of his good control group, with a negative family history of diabetes, had eventually developed a family history of diabetes, and therefore they should be dropped from the control values, And of course then you answer, "Suppose you wait another ten years, Which others of this control group should be dropped?" So it seems to me we are in a real dilemma. We will have to follow X number of patients and at the end of the experiment we redefine who the control subjects are. The implications of this type of design for an 161 experiment should be discussed critically by our epidemiologists, I think it is something we should consider very seriously, since the only way we can diagnose diabetes now is by what happens. If a patient develops clinical diabetes with symptoms, (i.e. polydipsia, polyuria, etc.) or the specific vascular complications, only then can we be certain of the diagnosis, Now, I think we might ask the audience whether they know of any populations which are susceptible to such studies, DISCUSSION AND RECOMMENDATIONS VAUN A. NEWILL M,D>° T don't know exactly where we stand in terms of getting recommendations from the group or where to start, to be honest about it, because I think that all of us are afraid of standardization and things that will keep us from doing what we believe we ought to be doing in terms of our own specific work, I think it was expressed by a couple of you yesterday who said that the bio- statistician in the planning session might actually stultify the research because he would put on too much pressure to see that things were done in a fashion he thought should be followed and the clinical investigator would just lose interest in the whole affair, The other side of the coin is, in order to get useful data concerning this particular disease, there has to be some method by which we can actually compare one study to another study to end up with some useful information, Here recommendations should be made (1) that permit flexibility for the investigator to perform the kinds of studies he desires to do and (2) that suggest and specify certain areas of standardization that can be incorporated into his investigations that will permit his work to be comparable with the work of other investigators in the field. DR, ACHESON: We are talking about population studies and we must have just a few words about populations and why we study them, It is just as impor- tant that the population is precisely defined for the purpose at hand, for the hypothesis to be tested, as it is to determine the kind of test you are going to do, 30 ghuol of Medicine, Western Reserve University, Cleveland, Ohio -163 There is one other reason one might want to study populations and I can illustrate this by referring back to yesterday when we were talking about the points on the glucose tolerance curve, Let's consider the first point on this curve and consider the variability of fasting sugars in terms of the people who vary with respect to that single parameter, In fact, one could find out why fasting sugars vary by studying a large population and classifying the individuals in it according to their glucose levels; thus one could learn whether fatness matters, whether social class matters, whether race matters, and many other things, You can get informa- tion about etiology, in other words, And for this purpose the population one defines is different from the one which may be suitable for a prevalence study or retrospective study. I don't want to enlarge on this too much except to say that the defi- nition of a sample and the way it is drawn, must always be precise to the study. DR, NEWILL: Obviously, one of the recommendations that this group can come up with is the kinds of studies they think need to be done, and I think there is little doubt after the discussion that has taken place in the last day or so, that a great deal of work still needs to be done with the glucose tolerance test, We need to know the various sources of variation in the results of the glucose tolerance test: those that are biological and those that are laboratory; the variation introduced by various sized glucose loads, the time of blood sample collection after load or food ingestion, whether venous or capillary blood specimens were collected, whether the blood came -164- from a superficial or deep vein, the amount of exercise permitted before or during the test, etc. I am not aware of any systematic study of the glucose tolerance test specifically aimed at measuring variability introduced into the test by such factors as those mentioned. Useful recommendations that this group could make would be (1) that a group, adequately financed, be set up to pull together and critically review the information available concerning the glucose tolerance test and (2) that specific studies to gather information not already available be set up and given a high priority for financial support. DR. RICKETTS: It seems to me one of the population groups that must be studied first is what we can call, as nearly as possible, a normal population, And this is very different, it is perfectly evident, from a frequency study, prevalence and incidence, and so on, You ask what kind of population we could define as a normal population. I would like to suggest that individuals between 20 and 70 years of age be studied by whatever technique we decide is most proper; that they be required to be in generally good health -- you will jump at this, all of you, because you will say, "How are you going to find them?" I think they can be found if you work through local communities, Individuals in generally good health, with no family history of diabetes =-- and I recognize the pitfalls in this, but what else can you do? People not on any medication. People of normal weight, and people gainfully employed, if they are males. =165- These are some of the criteria that I would suggest ought to define a normal population, I think they can be found and I think until you find them and test them by approved, generally agreed-upon methods, that everything you do will have very little meaning, DR, NEWILL: Can this be interpreted in terms of a recommendation that this kind of study should be done, Dr. Ricketts? DR. RICKETTS: Yes, this would be my suggestion, There may be some other normal qualities you would like to add to it, DR, NEWILL: I do not believe we should go so far as to specify what details should be included. If the consensus is that more systematic studies re- lating to the glucose tolerance tests should be performed then we should make a recommendation to this effect, This type of recommendation would suggest to the group that convened this meeting that they are to set up a local committee to examine what is known and unknown and to stimulate interest among investigators to gather the additional needed data. DR. ALPERT: I agree with Dr. Ricketts, although this is a difficult problem. It has been mentioned here before, and I think we would like to call your attention to this again, that when we are dealing with diabetics we are not dealing with normal people, Criteria which are established by whatever methodology we use by the use of, quote, healthy individuals, may not apply at all to the individual who is not healthy, who is a presumed diabetic. I would like to suggest, then =- I can't think of how this can be done right at the moment in detail -- that the same critical approach be made to groups of known diabetics, - 166 - We have always been careful in our studies to exclude the diabetics. It may be just as important to examine the criteria in the known diabetics at difference levels of severity of diabetes approaching what we are trying to find, that is, the mildest form, sort of working backwards. DR, NEWILL: Yes, but you are still recommending that studies relative to the glucose tolerance test be performed, except you would like the group enlarged. MR. WOOLSEY: I would like to recommend not only that we study the different groups that were spoken of, but that within these groups, either by the use of replication techniques or by the use of inter-penetrating samples, we study the effect of the different types of loading and concentrations and sources of blood, and so forth. I am not sure whether that was in Dr. Ricketts' suggestion or not. DR, RICKETTS: No. it was not. MR. WOOLSEY: But rather than normal variation, the variation due to different techniques. I think Dr. Milmore might support me on this. DR. NEWILL: I think what we are doing here is writing the charge to a committee who should find out what is known about these various things, and I think the consensus of the group is behind this kind of endeavor. DR. KLIMI: I would just like to remind Dr. Alpert that we have learned in this group that diabetes is a continuum, and that normal and abnormal is an arbitrary decision. We are speaking of one distribution. So all studies contemplated should give its distribution rather than a classi- Eication. -167- DR, MILMORE: And I would emphasize here, Dr. Newill, that the best way to do this would be to get a population group that we have every reason to believe could be followed over a long period of time, and that normalcy be defined after the event and not before, because I do remember Eisenhower had a coronary after he was supposedly normal the week before. DR, ACHESON: This is really the only practical solution, I think, to Dr. Ricketts' suggestion. You must take a total population, do your blood glucose examinations, find out who has diabetic uncles and what else. Then study the distributions of the relevant variables in that population and only then make your classification. I think the other approach begs so many questions that it is really never going to get off the ground. DR. WEST: There are problems, of course, in identifying a sample which is representative of a larger universe. In the National Health Survey that Mr. Gordon reported it appears to me that a rather good, that is representative sample was obtained. But this sample of the United States may not be representative of Casper, Wyoming, and that is some- thing you have to give consideration to. But the main point I wanted to make is that it worries me when we begin to talk about excluding people on arbitrary grounds such as whether they are healthy or not and whether they have a family history of diabetes. I think preferable to that would be to select them either at random, or systematically, and simply record whether or not, a family history is present or whether they are healthy, and how many 9-pound babies they have had, and so forth. So you would have the advantage of having a representative ~168~ population, but you could also examine any portion of the sample you wish excluding or including subjects according to any desired criteria, DR, LEONARDS: One of the things that I think this conference was called for was to attempt to decide what we are going to do in the meantime until all these studies get done, because this looks like a ten-year pro- gram and we should at least make some recommendations about what variables ought to be controlled until these studies get done. DR, O'SULLIVAN: Epidemiologic studies of diabetes generally consist of applying blood sugar tests to a whole population often followed by a repeat or other confirmatory test for diabetes, Studies utilizing one test only and performed consistently, for example, at one hour following a glucose load, or between one and two hours postprandially, can give satisfactory results on the epidemiology of blood glucose under such defined conditions, On the other hand, utilizing the initial blood sugar as a screening test introduces often unrecognized lack of uniformity caused by the large biologic variability of blood glucose, A third test, such as a glucose tolerance test, given to a "screened" population in fact gives no reliable information on the epidemiology of abnormal glucose tolerance tests, Obviously, two studies reporting results obtained by glucose tolerance test~- ing on screened populations cannot, therefore, be compared. I would like to propose one way around this problem, Bersons above Persons again above Random Ist Test mcmmmeenny 204 Test 5 3rd Test & 95th Percentile 95th Percentile 7 Sample As indicated in the diagram, retestdng using the population's 95th percentile immediately confers some uniformity with respect to other studies -169 - using different methods but the same model. When using two percentile levels that differ it is important that the lower of the two be used as the screening test and both together as the reported end point, Pro- gression to a third test~-invariably a glucose tolerance test =~ will necessitate an additional step. That is the performance of the third test on a random sample of the whole population, The necessity for this step was also emphasized by Dr, Klimt, In this way, results reporting diabetes in terms of glucose tolerance test results can be adjusted by the finding on the random sample, Whatever end point 1s used, it 1s important to present the data giving the full distribution of values, In this way, studies utilizing differing critical percentile levels may be able to undertake meaningful compar isons, Technical problems should also be mentioned. I think we can go a long way today by the judicious use of quality control procedures, The introduction of duplicates both blind and known, and of samples with known glucose content, as is being done in the University Group Diabetes Program and our studies, should be recommended, Furthermore, the same samples can be sent to other on-going studies, For example, we send specimens to both Tecumseh and Yale and feed them blind to our own laboratory, Where different methods are being used, we can determine the effect of level differences allowing more direct intercomparison of the results, One final word, When urine glucose tests are also introduced, similar sources of variation are operating, They must be given the same detailed attention in order to maintain the required uniformity of test application within the study. =170- I should like to be challenged on some of this, I think this kind of design, while allowing maximum flexibility, ensures results that will facilitate inter=-study comparisons, DR, NEWILL: I think this is carrying us into the second area, Let's come back first to the variation of glucose tolerance tests and then to the interim procedures, Dr. Klimt, you were going to try to summarize some of the things that were said about the variation of glucose tolerance tests in terms of a written recommendation, DR, KLIMI't This is drafted on very short notice but I gather two types of studies will be required, and these need to be defined: I thought that population studies on the glucose tolerance test are required, where classificattons should be made on the basis of the study itself, in retrospect, rather than having a priori classifications Under '"Population'' here is meant a representative sample of a naturally-grown population, in other words, not se- lecting to exclude or define a priori what is normal. A second type of study or studies would be needed to calibrate the sources of variation of the glucose tolerance test, And I have an incom- plete list here: Glucose loading dose, type of sample required, the type of chemical test, individual short-term variation, that is, reproducibility of the test at short intervals, and variation of the test by age, sex, and perhaps obesity. DR, NEWILL: Do you feel there is a reason that we ought to have a group get together to consider what is known as the jumping-off place for making recommendations about the studies that should be done and that there should be a committee set up to review this information? From the report of such a committee more specific recommendations could be made. DR, KLIMI: Undoubtedly, As I undergtood your assignment it was to delineate that two types of studies are required, One is population based -- and this time let's define what we mean by the population, =~ from which we will form our hypothesis on the basis of the results of such studies rather than having it in advance, “171~ The second one would be a designed study of the various sources of variation inherent in the glucose tolerance test, And that involves statistical tasks to come up with the best design to measure the sources of variation some of which I have listed, There may be others I have not thought of, MR, GORDON: It seems to me that there maybe ought to be some sequence in which these things are done, If we are talking about a fairly long- term prospective study to, in effect, retrospectively evaluate different criteria, before you did that you would want to do a fair amount of work on the methodological aspects, the questions of technical and biological variability, DR, NEWILL: Yes, This is one of the reasons I feel there should be a sort of expert committee pulled together to consider what is known and set up the direction which these studies should take, DR, KLIMI: Obviously the calibration study, being a more short-term study, should come first, It can be designed while the population study will have to take what is naturally grown, It will be less expensive and shorter to accomplish, DR, MILLER: May I ask at this time what population groups, that can be looked at critically, might be available for long-term study? I put down "stability essential’. I hope, Dr, O'Sullivan, this expresses what you just said, DR, KLIMI': There are at least two studies in the making, One is already under way, There may be others, One is in Japan where we specifically test the hypothesis of differences in prevalence which has been posed, because of differences in mortality from diabetes in Japan and in the *172= United States, The same suspicion of differences exists with respect to the natural history of vascular complications, Dr, Wada is the responsible representative of that study in Osaka. I think the second one is not that far advanced, but certainly involves a good population laboratory, We are currently thinking of a population-based prevalence study in the Puerto Rico area with Dr, Villavicencio, MR. WOOLSEY: I want to point out these calibration studies which might be short-term might nevertheless require fairly large numbers of people, because the fact is there is a limit to the amount that you can do on the same patient, You can't draw blood from innumerable veins and capillaries as well in the same patient. Therefore, you may have to use inter-penetrating samples and this requires larger numbers, I just want to point out it may be short-term but it has to be a fairly large number of people, DR, KLIMI': This is the second type of study, Mr. Woolsey, and I would have a totally different kind of population in mind, Such studies have been done in the past and can perhaps be done on a so-called captive population where you can design your study and where, with adequate inducements, you can get the possibility of doing repeated tests, and have cross-over and all the refinements of design you may wish in order to balance bias as it may arise, Prison studies or other institutional studies might give you a good deal of the calibration requirements for the glucose tolerance test. -173- DR, BURCH: TI rather disagree with that view in that it seems to me one could wonder as to really whether the prison population is representative of those who haven't gotten there yet, In regard to other populations which might be studied, such as Dr. 0'Sullivan was speaking of, I am going to be doing a study the first part of this next year on a tribe of American Indians, and definitely will be planning to incorporate some of the methodological procedures and objectives that we have discussed, DR, KLIMI: I doubt whether the Indians or any natural=-grown population could be induced to serve as the ''volunteers' to determine repeatability of test, glucose loading dose in various scales on the same persons, the type of samples to be preferred, simultaneous sampling of venous and capillary blood, etc, But that sort of thing can be done best in a captive population, DR, FELDMAN: I am a member of the Permanente Medical Group in California. As most of you know, we have a health plan in Northern California, serving 500,000 people, I can't speak about the stability in California; maybe Dr, Milmore can. But I really think it is an ideal group in this sense. They are not in prison completely and I think we do have the opportunities of long-term follow-up of these people with all age groups represented, and sexes, and so forth, I think it would take a lot of planning and what- not, but it might be a possible solution to a large population study. DR, MILMORE: The Hutterites are quite stable, aren't they? DR, NEWILL: Except they have been studied so much they might not want to submit to anything more, - 174. DR, LEONARDS: Maybe I didn't make myself exactly clear as to the need for this. I don't know how many of you are familiar with the fact that the American Diabetes Association is very strongly considering sending out a recommendation to its local sections that they seriously consider the possibility of experimenting with using blood as a means of dia betes detection, rather than urine, If this were the case, it would be nice for a group like this to make some specific recommendations to them as to what we think is the most practical and yet as close to ideal means of doing this as possible, I don't know that very many of you realize the scope of the State Health Department surveys of diabetes and literally coming up to the million-a-year blood samples that are now being drawn in a rather random fashion with a variety of methods for the purpose of providing health service to the state population, this part of the Health Service being the detection of diabetes, It would be very nice if a group like this could supply to them, who are now in the process of changing their methodology over to using blood specimens in the field, just what we recommend as the most practical way for them to do this. And then I don't know if many of you know that a meeting of the World Health Organization is going to be held in a few months in which they are going to want recommendations from the United States based on our experience of diabetes detection as to what they might incorporate into their recommendations to the various nations. This is why I say there is a real need for our coming up with something definite in the interim before the suggestions are made, ~175~ DR, NEWILL: This separates yours and O'Sullivan's, actually, a little bit, These are two separate recommendations, because you are asking for recom= mendations specifically for detection programs, DR. LEONARDS: I just wanted the group to realize the worldwide interest in diabetes detections. DR, NEWILL: We have Dr, Alpert and Dr, Crampton of the American Diabetes Association with us each of whem has been or is the chairman of the Committee on Public Education and Detection. Can you tell us about what is going to happen? Will there be this recommendation to switch over to blood testing programs, Dr. Alpert? DR. ALPERT: Yes, It will be a recommendation recognizing, of course, at least initially, the obvious difficulties, But it will endorse at least the principle of adding, as far as possible, blood testing to the urine testing which is the standard procedure of the American Diabetes Association. That is not going out this year, incidentally, because it is near November, but it will go out for the 1965 drive, DR, NEWILL: In terms of the World Health Organization meeting that Dr, Leonards mentioned, has there been any request of the committee for representation at this? DR, McDONALD: The composition of the committee has already been set, actually, It is just eight people, But there will be an avenue to put recommendations to that group from this group to bring to their attention what recommendations might come out from here, But it is a rather small committee that 1s meeting, DR, NEWILL: Can we have a specific recommendation in terms of detection? - 176 - DR, LEONARDS: Let me take two minutes to explain why I am going to make this specific recommendation. Obviously, the easiest way of doing mass screening is by a urine specimen. At the other end of the scale, I don't think many people will argue against what I am going to define as the ideal method of detecting diabetes at its earliest stage: to put people on diet of carbohydrate; to induce a stress in those individuals, be it pregnancy or cortisone; to inject intravenously into that individual over a constant rate a large dose of glucose; to have in the other arm a double lumen catheter attached to an AvioSngliyzerS] which will give you a continuous recording of the blood glucose curve; and calculate the ability of that individual to remove carbohydrate on a numerical basis. This sounds funny but it can be done and we have actually done such studies on a few people for a few specific purposes. Now, let's ask ourselves: How far can we back up from this ideal procedure and still retain all of these ideal things that we want, and yet still have the thing practical? And I will discuss each one of these at hand. One is, I think, that everybody will agree that on a large-scale survey you cannot control the diet of the individual. So we will cross that out. I think everybody will agree that you cannot induce a stress, either cortisone or pregnancy, so we will cross that one out. $M pectnicon Instruments Corporation, Chauncey, New York -177- I think everyone will agree we can't give the glucose intravenously, that we can't measure the glucose continuously, and therefore we can't calculate the removal rate, But what can we do? We can load the individual. I think we have shown that this 1s possible, I am not saying it is desirable. This is what I hope the group will make some recommendations on, but it is possible to load the individual, What are you going to load him with? I don't think anybody is going to attempt to do mass screening by loading an individual with food so we will cross that out, I don't think very many people are going to recommend loading an individual with glucose == and I say this because, although there are no data in the literature as far as the incidence of nausea after glucose, I think anybody who has given large numbers of glucose tolerance tests to a variety of people will say, as Dr. Fajans said yesterday, it is the opinion that there is a considerable incidence of nausea, And I think the individual that probably has more information about this than anybody else is Dr, Feldman, and I would like him to correct me if I am wrong. I understand in their multi-screening program in California, one out of three or one out of four individuals that got their 100 grams of glucose loading got sick in one way or another and enough of them vomited that it became a problem, DR, FELDMAN: I can't confirm that, I would say that is an overestimate. But as you know, in terms of cost, I think Dr. Morris S. Collen has an interesting idea for screening, He has taken an ordinary soda vending machine, carbonated, and put 100 grams of glucose in it, and with this method in the last couple of months there has been practically no reaction =~ occasional nausea, - 178 - DR. LEONARDS: So I think there is agreement we must do something to the glucose solution to make it more acceptable to the population, and whether we lemon-flavor it in a vending machine or use corn syrup is a detail. But we have eliminated food and 100 grams of glucose as it is generally used, but some recommendation should be made about glucose modification. As far as the amount of load, I don't think we have any data to make a definite recommendation except on bias, and let me give you the bias I have for this. The British have done very well on 50 grams of glucose. We have done very well on 100 grams and 75 is in between and is more likely to be acceptable by large groups. I think it is of interest that some of you might want to know why we chose 75 grams. This happened to be the largest amount of the corn syrup preparation that we could conveniently put up in a bottle and still have it palatable for individuals to drink. And I think the same thing is going to be true on a large scale basis for whatever type of glucose load. As far as the number of blood samples is concerned, I think there is only one answer to that. If you are doing large-scale screening you can only take one blood sample. This, of course, will be the initial test and it is so recommended, but it is also recommended that all prople screening positive have some form of a second test, as Dr. O'Sullivan has suggested. As far as the time, I think that is one thing you said there has been the greatest agreement on, and that is if you had to make a choice, the two hours is probably the best one for a variety of reasons. -179- As far as venous versus capillary, I think there are four reasons why, on a large-scale survey, one is going to be forced to recommend capillary methods and that is, if you are doing large numbers of people, most of them are going to object to the venous procedure and this will lower your yield. Secondly, there is a cost consideration in that the materials required to draw venous samples cost in excess of those required for capillary. Third is the ability of the technicians to obtain blood, and the training of the individuals. And fourth, if you are doing a large number of individuals, speed is very important. A good technician can take 50, sometimes 60, capillary bloods an hour for two or three hours straight, and you can't come any- where near that using venous blood, With regard to the chemical tests, there really are only two possi- bilities open at the moment about which one can recommend for large numbers of individuals. One is the use of the Aueohnalyzer. 38 The second is the possibility of using the paper strips that have recently been introduced. And I think I would like to make some comments about this as part of a recommendation. I would not recommend that anybody attempt to do screening for diabetes on a mass basis using the paper strips alone. They are subject to too much error and variability from technician to technician, so that you cannot use them alone for mass screening. This does not mean that the 3Byechnicon Instruments Corporation, Chauncey, New York - 180 - paper strips have no value. They are of extreme value in a wide variety of instances in the clinic, in the doctor's office, treating diabetics, et cetera, but they are no good to screen diabetes in the sense that we are interested in doing to find early cases. You can, however, reduce the load on your Aucohnalyzers? by using the paper strips set at a low level, eliminating all of the obvious negatives, and then retesting the positive ones by taking a sample immediately and sending it in to the laboratory. The other recommendation that I would make is if anybody considers “the use of the strips, it is hoped that mass production would considerably lower the price. DR, ACHESON: I should like to comment on the question of the two hours. As I said before, I don't know a good deal about the pros and cons of one hour against two from the clinical point of view but from the point of view of practical epidemiology this is twice the amount of time. Dr. Leonards and Dr. Kent have already stated this and we already know the longer you keep the people the less likely they are to cooperate. DR, KENT: We don't keep them. They stop in and drink their fluid, come in and get their finger pricked and go shopping or back to their place. We don't keep them two hours. It is really as easy to get them back in two hours. DR, ACHESON: First of all, it depends on who they are, what they are doing, what the time of day is. And the other thing is you must have a wide margin in your interpretation of two hours. 3yechnicon Instruments Corporation, Chauncey, New York <181~ DR, KENT: No, we are pretty accurate, actually, DR. WADA: We have screened a sizable amount of people and we have found that keeping people for two hours is very difficult, We are also con- cerned with the possibility of people coming back later than two hours and we will lose some of them, And therefore, I would like to take a one-hour specimen, if possible, MRS, FISHER: Data we have looked at indicate that there is not a great deal of difference in results of one or two-hour samplings. It is generally agreed that these are the two best if you are trying to pre- dict the full results of an oral GIT, But there is not a tremendous difference in the two in their sensitivity and specificity for this purpose, Since some of the data indicate that differences using different loading mechanisms are seen in the two-hour bloods more fre- quently than in the one-hour bloods, it would seem that these two pleces of information might make it important to use the one-hour specimen rather than the two-hour, DR. O'SULLIVAN: I would like to make a few points about the capillary blood specimens, The first is to repeat what Dr, Ricketts said this morning -- not repeat it, but just to remind the meeting about it, Secondly, I would like to say that the blood sugar results with capillary specimens are subject to more variability and this is some- thing you have to keep in mind. The reasons are obvious, Thirdly, when you take a capillary specimen you don't have the apportunity of repeating lab errors. It is usually one shot affair unless, you draw two 0,1 ml, or two 0.2 ml. specimens, if you can get them, This is a real drawback, -182-~ DR, LEONARDS: Those are good points, DR. HAYNER: . I think Drs. Kent and Leonards have proved the feasibility of their method quite well with the two-hour interval. And I suppose there are some instances where this may be required in an industry where you want to use the coffee-break and lunch-time, let's say. But we also made a computation of the sensitivity -- in this case we computed the ranking ability for the one-hour and two-hour point, and the values here were almost identical for these two points from the Milan study -- a smaller number of persons than you have -~ but we also had the one-and- one-half-hour point which appeared somewhat higher, The intra-class correlations were ,59 for one hour, .73 for one-and-a-half hours, and .56 for two hours, DR. BURCH: I think we are almost getting away from the strict detection survey into the population survey again on some of these comments, DR, WEST: Another minor advantage of the two=hour is the large amount of data which has been accumulated previously, I believe there is a little more data with which subsequent studies could be compared on the two-hour value for the one-hour. I wouldn't be prepared to say whether one is better than two, and with respect to Mrs, Fisher's comment, there is a real problem of correlating one value with the whole test, because the result of the whole test is decided very arbitrarily anyway. I don't think we know whether the one-hour or the two-hour level is more sensitive and specific, MRS, FISHER: I would agree completely with that but nevertheless, that is what we are attempting to do, predict those who would have a positive glucose tolerance test for referral to the physician, It is the only -~183~ objective measure that we use at present, as a re-test, I think, with our screening tests, we are trying to get the best predictor. The data I was referring to on the sensitivity are from a study done by Mr. Remein and Dr, Wilkerson where they did obtain sensitivity and specificity mea~- surements related to the prediction of the results of a glucose tolerance test, DR, WEST: But their end point or reference was not diabetes, Rather they compared results at the different times with an arbitrarily derived formula for determining the presence or absence of '"abnormal' tolerance, DR. MILMORE: I indicated my preference for the one-hour but there is another point in favor of the two hours, in a situation where one is in a detection program and referring to physicians, This has been brought out repeatedly in our dealings with local advisory groups. Practicing physicians are more familiar with the two-hour reading and its interpre- tation than they are with the one=hour. DR, KLIMI: I have pointed out before, but I would like to reiterate that in our hands and in accordance with the general recommendation of the Public Health Service, the one~hour test has the greater power to differentiate between known diabetics and the general population. The order of magnitude is something like 82% exclusion rate on the basis of the one-hour test and 65% on the basis of the two-hour test, And it has been long in print that if a choice has to be made, that the one-hour test is the more potent, DR, FAJANS: I hold no strong brief for one or the other, I think from a physiologic point of view, the one-hour level is more subject to variables of gastrointestinal absorption while the two~hour level, physiologically = 184 speaking, correlates better with glucose utilization. If it can be shown on a statistical basis that the one-hour level ranks people just as well as the two~hour level as far as overall carbohydrate tolerance is concerned, I cannot argue against it, But I wish to bring out that as far as physiologic happenings are concerned, the one-hour level will express a greater variable as far gs absorption is concerned. DR, LEONARDS: I would like to remind Dr. Klimt that we are not trying to differentiate between normal individuals and known diabetics. We are try- ing to differentiate between normal individuals and unknown diabetics and your statistics do not cover this point, And I would like perhaps to rephrase your question in asking the group what time they would choose by dividing the question up as to whether they use capillary blood or venous blood, In venous blood there might not be very much difference between the one~ and two-hour, but as Dr, Ricketts so admirably pointed out, if you use capillary blood the one-hour specimen is so variable that it probably would exclude it from consideration, DR, BURCH: It seems to me that the group here that has had more experience than anyone else with this type of program feels that two hours is best. So maybe we ought to go along with them, DR, ACHESON: Can I say that they are in a special situation. They are working in a city with a population of a million and not all diabetes detection drives are going to be done under those circumstances. For instance people may be coming to town for a fair or something like that and want to get away with- out waiting around for two hours, And from what I have heard from the evidence of the diabetologists here, none of them seems to be convinced one way or the other, So from the practical point of view because two hours of a person's time is twice as much as one hour, I favor the latter, -185- DR. KENT: The one hour blood sugar is very variable as Dr. Klimt has pointed out here. It is apt to be quite high, and we are going to have a great number of people identified that are borderline positives. And then we jeopardize our program psychologically because we have community physician adversaries instead of a cooperating group. DR. O'SULLIVAN: We examined 752 unselected pregnant women, and looked at the fasting, one-hour, two-hour, and three-hour values. We could show significance with respect to certain items at the one-hour a little better than at two hours. Both were almost equally good, but the one hour had the edge. The co-efficient of variation for the one and two-hour values was the same. MRS. FISHER: I think one of the things we should recognize in all blood testing programs that are going on today is that it is a very rare program that is using any kind of glucose loading. When we talk about the two- hour test we are talking about tests conducted in very special situations, and I think if we recommend that projects use blood tests obtained only two hours after glucose loading, many programs will be unable to consider such a possibility. DR. NEWILL: Can we come to the venous versus capillary? Dr. Leonards is all for the capillary method. Would anybody like to support the venous method for the large screening operation? DR. WEST: Many communities have the capacity to do venous determination and are not set up to do a capillary determination. We could push this and some would readily develop this capacity, but that is another consideration. =186~ DR, ALPERT: I would agree with Dr, West, You would have to leave this up to the local community. You have to provide the information which they can use, but the choice must be theirs, DR, RICKETTS: May I explain my position, I said there is not enough jgown about capillary blood behavior to use capillary blood at this point, I think we have to collect data on a large sequence of glucose tolerances before we know what a two-hour capillary means. And this 1s my reason for recommending venous now, DR, O'SULLIVAN: I would prefer the venous to the capillary if a choice had to be made between them, My preference at this time, however, is for the choice to be made locally at least until scientifically defensible reasons are available for a preference, but I didn't vote for either one. I voted for both, DR, MILLER: Dr, O'Sullivan, you have used venous blood in your studies? DR, O'SULLIVAN: Yes, DR, MILLER: How many of the practical difficulties that have been alluded to exist in your studies? Off the top of my head I would agree with Jack that on a large mass screening test if isn't feasible to try to get venous bloods. Do your technicians get it every time? DR, O'SULLIVAN: We don't have any trouble, We can take girls from school and train them to do it, As a matter of fact, the local nursing organ=~ izations don't allow their people to draw blood, yet we can pick school kids and they are allowed to draw blood. DR, MILLER: We can't succeed in the hospital in getting bloods every time. ~187~ DR, McDONALD: Let me make this statement to the group that there are some states in the country where they won't even talk about capillary blood, They say, 'No, venous blood," There are others that say, "Our people won't accept capillary; they want venous." I don't think there is that much difference in the people involved, I think it is a question of what becomes the custom in those communities and with the proper presentation the population will accept either one. DR, BURCH: I have worked on two populations of Indians and I can assure you that with the Blackfeet Indians, who have a horror of bloodletting unless they are doing it, the capillary system would be far more acceptable than the venous, But the other group of Indians in Arizona -- it wouldn't make any difference. DR, MOURATOFF: In our case it was the other way around, They loved it because they thought it had a therapeutic effect, DR, ACHESON: There is a legal aspect which we meet in Connecticut, In Connecticut, bloodletting is looked upon as the practice of Medicine and you must have a State Board license to take blood from a vein, but I don't think you have to have a State Board license to do it from a finger prick. On our survey we have to have a physician present when we do a venipuncture outside the hospital. DR. McDONALD: I am very relieved that you still have both in there, myself, ~188~ RESOLUTION THE CONFEREES FORMALLY ADOPTED A RESOLUTION THAT FOR THE PRESENT AND UNTIL COGENT, EXPERIMENTALLY ESTABLISHED ARGUMENTS ARE AVAILABLE WHICH CLEARLY SUPPORT ONE OR THE OTHER, LARGE CABE~FINDING OPERATIONS SHOULD USE EITHER CAPILLARY OR VENOUS BLOOD AT THE DISCRETION OF THE DIRECTORS OF THE OPERATIONS, THEY FURTHER AGREED NOT TO CHOOSE BETWEEN THE ONE-AND TWO~HOUR POST LOADING INTERVALS AT THE PRESENT TIME FOR REASONS WHICH ARE REFLECTED IN THE DISCUSSION, DR, MILLER: What about Jack Leonards' point that if you do capillary you ought to do it at two hours? Could it be one hour for the venous and two hours for the capillary? DR, KLIMI: Both, DR, KENT: I think it is important to put it in because the variability we have at one hour is very marked, What Mrs, Fisher was talking about, I think, is venous blood, DR, ALPERT: I think we are mixing two things here, One is the desirability of the time factor, The other is: What are the criteria at those times? I don't think we have come to that point yet, If you define -- if you are able to =~ this latter factor, then you can leave the choice up to the local organization. DR, NEWILL: This was the next question I wanted to ask: Does the group want to recommend criteria for deciding the cut-off point for screening? Do you want to speak to it, Dr, McDonald? -189~ DR, McDONALD: Yes, let me speak to it, The only appropriate referral level, which is what your criteria for screening are going to be, is negotiated with the practicing physicians. If they won't accept as diabetics some that are too low, then there is no point in referring them, So I would rather think that the criterion for referral should not be recommended on a broad basis but left to negotiate with the practicing medical community in the agency doing the screening in each community. DR. NEWILL: So that you could say that the criteria are not going to be specified by this group but determined locally by the community where it is carried out, MRS, FISHER: I want to add another point, If you are just going to do an initial screening test and refer positive cases, you must set your point higher than if you can incorporate a retest into your program. With a retest, it would be desirable to use a lower screening level so that you would obtain more of the people who should be diagnosed. The retest would avoid the referral of large numbers of persons. DR, NEWILL: This should be added to this recommendation as a suggestion, that if there is to be no repeat test then the cut-off point certainly should be higher than if there is to be a repeat test. DR, WEST: I think this point is closely related and I can't recall whether it is in Dr. Klimt's suggestion, but I think we might discuss it. It was only alluded to before, the question of the relationship of the load to the meal, In other words, are we prepared to say that it doesn't make = 190 - any difference, that we can give the load any time during the day, or are we going to take the position, on the other hand, that a certain arbitrarily assigned time should elapse between the meal and administration of the load? Dr, Hayner has done a lot of work on this and there is quite a bit of data available, I think we could make recommendations, For example, if we are to assume it didn't make much difference it would be enormously helpful logistically to say 'Fellows, you can give the load whenever the customer comes in.'" On the other hand, if it makes a lot of difference, I don't think we should do this. DR, BURCH: I think Dr. Leonards made the point it wouldn't make much difference what you said{ that the people coming in are going to do what they feel like, And that is certainly what my experience has been. You can't be making too many stipulations. DR, NEWILL: It would seem to me that rather than adding this to the recom- mendation it might be a suggestion, You are really suggesting that a specific time be specified for giving the load? DR, WEST: No, no, I am convinced on the basis of the evidence that it probably doesn't make any difference, and I would personally favor just encouraging people to go ahead and give the load any time, but others may feel we should study this a lot further before we make any recommendation, and there may be a third group who believes that loading after a meal so complicates the result of the test that we ought to recommend that a certain time elapse between the two, DR, HAYNER: It might be worth including a suggestion that a simple question be asked af the respondent coming in for the test: 'When was your last meal or snack?" Then all persons under a 4-hour interval could be treated - 194 in one way, and those over a 4-hour interval treated in another way, using the approximately 25 mgs. per cent difference between these groups. DR. KLIMI: Dr. Newill, aren't we really making up two kinds of standards here? We are talking, on the one hand, of standards for case-finding drives, and on the other hand we are discussing standards for surveys to be used in a scientific analysis or for clinical trials. Those two are a different kettle of fish. DR. NEWILL: I thought we were discussing the first rather than the second, and were going to come to the second in the terms of what Dr. O'Sullivan stated for us. DR. KLIMI': Then I stand corrected. DR. BURTON: But the only thing is I agree with Dr. Klimt that this last consideration is more appropriate for the other type of study which we are not yet discussing. DR. NEWILL: This is why we are discussing his recommendation. How many feel this should be added to the recommendation for detection? DR. KENT: I think this is a very appropriate thing to add so somebody doesn't say, "IL can't test if they have eaten." Isn't that your point, Dr. West? DR. WEST: We have found in some communities we are able conveniently to test large numbers if we do not have to give the load in some specific temporal relationship to the last meal. On the other hand, if we say that everybody who reports has to fast at least two hours or four hours, then we may not be able to test nearly so many. And at other times, this doesn't make any difference. But it is a factor. And again I am relating my remarks only to case-finding. —-192- RECOMMENDAT ION FOR CASE-FINDING SURVEYS ONLY, THE CONFEREES RECOGNIZE AS BOTH PRACTICAL AND CLINICALLY ACCEPTABLE THE PRACTICE OF ADMINISTERING A GLUCOSE LOAD AT ANY CONVENIENT TIME WITHOUT REGARD TO THE TIME OF THE PREVIOUS MEAL, DR, NEWILL: Let's go back to what Dr. O'Sullivan started us on a little while ago. Do you want to put this in terms of a recommendation? DR, O'SULLIVAN: I can see from back here I didn't draw my diagram very well, Yes, I would like to propose that this type of flexible design and quality control procedures be recommended for epidemiologic studies of diabetes. DR, McDONALD: I am wondering about the recording on this, Is it going to be sufficient that the actual recommendation can be drawn? DR, NEWILL: No. That is why I wanted him to make a specific recommenda- tion with regard to this, *Let me review a couple of things here. It seems to me that what we are getting into is the kind of information we would like to have made available to us in terms of specific population studies that are being done =-- this is changing it a bit, But in the reporting or in the description of a population study we want to have the specific question stated that people are attempting to answer, and we know they are going to have to tailor all their methods to that specific question. ~193 ~ Then, when we report on the methods, we want separate sections for the population methods, the measurement methods and the statistical analytic methods, In the population methods we want to know what population is necessary to answer the specific question, what the size ought to be =-- and these things should be reported not only in terms of apply- ing for a research grant but in terms of the write-up or description of the report that is submitted for publication. We ought to know the sampling frame out of which the cases were chosen and what the sampling methods have been, We want to know how the comparison group was chosen, why it was chosen the way it was chosen, and what we intend to show with this comparison group. We want to build in the kinds of methods that are necessary. In the measurement section, we should outline the various measures we are going to use. We should pre- sent the variability of these measures as is ascertain- ed by other people who have used the particular methods, and we should be able to say that this method is a method that has been tested, or we should set out to get the kind of information that is necessary to give the variability of the particular measure, This is the pre- testing phase or the pilot~testing phase of such a project. =194 - Whether we are going to use questionnaires should be outlined, expert help is frequently needed in setting these up and in pre-testing them, as well as the specific end points for data evaluation. Also, the methods of statistical analysis should be specified. It is necessery to record all of this information so that future investigators can make adequate comparison with some degree of assurance that their comparisons are valid, (*Editors Note: The above statement is discussed in detail below, but it represents the consensus finally reached by the conferees.) DR. O'SULLIVAN: Let me say this: I couldn't agree more with what you have said. You have been talking in terms of the principles of an epidemiologic study, I presume that anyone doing an epidemiologic study knows what he is going to do, why he is going to do it, and what the population is supposed to represent. DR. NEWILL: I hate to say it, but it is an unjustified assumption. DR, O'SULLIVAN: My concern is that even people who know the principles may be unaware of the very special difficulties with respect to blood glucose, The only other comment I would like to make is that it is not sufficient to report the method being used. While the study is proce- ding, quality control should measure the variation of the method over time, DR. KLIMI: Dr, O'Sullivan, is that not all part of what we call good design? You measure, within variations, the degree of repeatability -195- right in the study, and I think any major study nowadays would have to have these components, DR, O'SULLIVAN: It is difficult to name one, except for the National Health Survey. There isn't one that I know of, DR. KLIMI': Well, we use it, DR, FAJANS: We haven't published it but we do this daily in our laboratory. DR, KLIMI': In a way, it is right, These assumptions need to be re- stated, You are quite right, DR. NEWILL: I think many of us think about these things but don't do them, DR, ACHESON: I think if we are making recommendations about population studies, we should make a specific recommendation first of all that quality control is built into all studies for all measurable variables -- and there are so many it is difficult to count them; and secondly, publication should describe the nature and variation within these quality controls =-- because comparability does depend to a large ex~ tent on this very thing, even though it is only a footnote or it is put in the National Library and Dr, Lazarow gets it out for us, Some- how or other this information should be available. DR. NEWILL: I think your recommendation holds, I would, for one, like to see it more specific because I don't think everybody can read into those few sentences what you do, DR, ACHESON: Dr, Klimt gave us a very detailed description of it this morning, If you would like me to make it more specific I would like to recommend that we do what he says. -196~ MR. GORDON: I would like to speak to the suggestion of Dr, O'Sullivan here, that where you are trying to determine prevalence, or whatever your purpose, you perform a standard glucose tolerence test on some sub-sample. I think it is an excellent idea and I wish we could have done it, but I think many other people get in the same difficulty that we got into. We felt that in order to assure response, the demands which we should make on our study group should be restricted and care- fully delimited, and so we had them once and we said that was all we were going to do with them. The result was, whether because of this or because of other things, that we had a very high response rate for an examination survey, Now, you can do a lot of things on study groups, but I think you have to consider whether, in the course of doing these things, you are going to wind up with a population which doesn't represent any defined group at all, If you put too many demands upon the population you study -- and this will depend on the population studied -- you will wind up with a high non-response rate and the result is you will say a great deal about people you can no longer define, So this leaves the study groups with the option of doing this sort of thing on people whom they don't study, and then they are left with the dilemma that they don't know whether their calibration studies in fact represent the population they study or whether it is a peculiar group, as the prisoner group that we worked with, I am not suggesting a solution; I am not making a recommendation, but I am suggesting that there are some dilemmas you meet in practice which are not easily resolved by recommendations. -197= DR, WEST: We have run across a couple of concrete examples, One is that we were able, in a couple of populations, to establish under certain conditions the sensitivity and specificity of the urine glucose, finding that the actual prevalence rate as determined by performing the blood glucose after oral glucose was five times greater than the apparent prevalence rate as determined by pre- screening with urine glucose tests, Therefore the sensitivity of the urine glucose under this condition was 20%. That worked very nicely in population A and B and C, but when we got over to Pakistan we found this was not true at all, The specificity of the urine glucose in Pakistan was quite high because the nature of the disease in the population was different. So one must be careful in making general- izations for all populations from studies in a population or even from more than one population, Another example of this may be given by indicating that when we tried to measure the effect of changing the dose of glucose in re- lation to body weight in medical students, we found there was no difference because all the medical students had excellent glucose tolerance and whether you gave them .,8 or 1.6 the two-hour value was the same, On the other hand, when we went down to the State Mental Institution where they have mediocre tolerance and older people, we found out we could bring out quite a bit of difference when we changed the load; that it made a substantial difference in the response of the individual, So this just brings out again the problem of trying to relate calibration studies in one group to the general population, “195 DR, MILLER: In other words, you would suggest calibrating methods you are going to use for that particular group to see if it is still relevant, DR, WEST: Whenever possible, You can only go so far, but I think it should be kept in mind, DR, NEWILL: Are you going to enlarge on your recommendation, Dr, Acheson? Will we have a specific recommendation? DR, ACHESON: What I think is that quality control is essential; that there are various qualities of quality controlj that we should always try to use it, I don't think it is for us to decide now the exact techniques but those who know how to do it should use it and describe the kind of quality control they have used. Those who are not familiar with the relevant techniques should be asked to familiarize themselves with them and employ them, In general there are two forms. One is within a lab and the other is between labs, and I think there is a great deal of reason in doing both, if possible, DR, WEST: There are tw things Dr, O'Sullivan mentioned that I think are of such great importance they might be mentioned specifically. One is, to the extent feasible, by the design of the survey to minimize the number of non-responders; and second, when the rate of non-response is relatively substantial, whenever possible to try to design a specific method of examining a sample of the non-responders; and thirdly, if there is a screening test followed by a definitive test, the definitive test should be performed on a sample of the negative screenees. I think these things come up so much over and over again that they might be mentioned specifically as a recommendation. -199- I would add one other with respect to sample, and that is that the very fact that the population tested knows it is a diabetes survey induces an automatic and possibly substantial bias. While this can't sometimes be eliminated, I think some attention should be given to its elimination whenever possible by not telling the population they are being tested for diabetes -- and again we are not talking about screening surveys but systematic studies of prevalence rates, We have done this, Of course, it is sometimes absolutely necessary to advertise the nature of the survey, But it is important to recognize this possible source of bias in a sample collected under these conditions. DR, NEWILL: Are there other recommendations about surveys that people would like to make? How about the presentation of data and frequency distributions? DR, KLIMI: I think it has been repeatedly reiterated that we should give a complete description of the results rather than only an interpretation of the results, And while I think you said that every individual result ought to be read, that may be more than is possible, But I think there are more descriptive characteristics that can and need be added than are usually added. We need not only means} we need frequency distributions; we need a measure of the range, and a description of the results as far as they can be described, This does not mean every single glucose tolerance curve prior to the event when we might have a national library. DR, NEWILL: I think one of the things that was discussed was the fact that a great deal of the individual detailed data could not be published because the journals will not accept it, and one of the -200~ questions raised in private conversation was whether or not the group felt that it was so important to have the detailed data available that some specific recommendation could be made to either the National Institute of Arthritis and Metabolic Diseases or to the Diabetes and Arthritis Program of the Bureau of State Services that they would set up some kind of central repository for this kind of data and it would be available to other people who would want it for re-interpretation in terms of their feelings about this particular matter, whatever they happen to find, MR. GORDON: I just want to make a point about keeping a repository of all the detailed information. In the Milan study where we studied 24 men and did 16 tests on them, and where we had control of the publication ourselves =-- that is, we could print what we wanted to print ~- we printed most of the specific data we got. Not all of it but a lot of it, I am willing to wager that there are going to be very few people who are ever going to go back and recompute any of the material which is in there. Now, if you get into a very large study, the volume of that material becomes enormous, and it doesn't make too much difference whether it is in a published version or an unpublished version, except that an unpublished version is less accessible. I don't think there are too many people who have either the time or incldnation to go back and re-do or plow through the original data that have been collected by somebody else, In principle, I think this is an excellent idea and I have, in ~201- fact, gone back and recomputed == and Dr, West has done the same thing == some of the data which did appear in print on a small series of cases with respect to variability and instead of using the measure of variability that was reported we recomputed these things so they would be comparable with the way we were computing our own data, But this must be very rare and I wonder, if, for a large amount of data, the large amgunt of effort which is required to maintain this material, But if you are going to come back to this ten years from now, I would feel almost certain that these data ought to be discarded because I don't think there are very many data which are being collected today which are going to be of immense use to the general research population ten years from now. Now Dr, O'Sullivan will go back to the study group he is going to be concerned with and re-analyze that, but that is within a specific study. To make this a concern for people who are involved in general with these kinds of data, I think is unrealistic, DR. BURCH: TI don't really agree with Mr, Gordon. I have re-analyzed data from the articles, from the literature, that is, on those very, very few which present enough data so that you can, And I would have loved to have been able to have re-analyzed some of the others in order to see if their data might support or refute some of the ideas which I have had, However, it seems to me, though, that this would be a prodigious task, and it certainly would need something of the order of what Dr. Lazarow was talking about last night to be able to store these things, MRS, FISHER: I just wanted to say that I personally knew about 6 or 7 people who have obtained detailed statistical data and looked closely at the information, I think there are people intimately involved in =202~ certain aspects of these problems who will make use of the data available {n some detail, And it helps them, We try to do this any time we have a geries of data, to check with those other people collecting similar material and see if their results verify ours. For example, in the plasma whole blood paper, we use some of Dr. O'Sullivan 's data, and some from other sources, too, DR. ACHESON: I hope I am not being too naive about this but with my MPH class I introduce this with four short words in the first lecture of the course. I tell them they must answer four questions before they ever start at all, First of all, WHY they are doing the survey, For our purposes this can be broken down into two parts. One 1s prevalence studies which are never an end in themselves. They are always a means to an end, For instance, the identification of un- known cases, as a basis for planning heal th services. Another use is what Dr. West was talking about, the comparison of two communities using identical techniques with a view to identifying systematic differences between the two communities, Another is the study of natural history, a fourth is to look at variables in a community and try to relate those systematically to the disease in question, Here the design will be obviously different, So the next thing is to decide WHO you are going to study. You don't know WHO you are to study until you know WHY you are going to study them, Dr. Klimt is studying a completely different kind of sample from the one I am studying or that which Dr, O'Sullivan is studying because the answer to his "why' is different. -203- So if we want to find out something about the variability of the blood sugar curve, for instance, and relate this to qualities in a community we want a different sort of study than if we are trying to look at kinds of treatment. Now, the next thing is WHAT you are going to measure, and this, in our present context, is whether you are going to look at urine only or urine and blood sugar. And then the next question is HOW you are going to measure it. I don't see how you can start talking about the HOW until you have answered the other three and that is why I think every study must be different from every other study because the answer to WHY is different. DR. NEWILL: This is why I was suggesting that the resolution about surveys should, in fact, be a broad one that covered the actual description of the survey so the information would be available as to WHY, WHO, WHAT, and HOW. But I don't think we need to go any further on this. I think we are going to have to worry about the specific working or recommendations and we are going to have to get back to how we handle that later. Do we have any more comment. on this business about the central re- pository of data or the way data should be published; that sort of thing? Does the group want to make any recommendation on this or not? DR, ALPERT: Yes. DR. HAYNER: In reference to this, the American Heart Association and the Heart Institute have been working on this type of problem in regard to coronary heart disease studies and there is an operation going on to collect the data from these studies. ~204 - RECOMMENDAT ION THE GROUP AGREED THAT THE RETENTION OF STUDY DATA FOR SUBSEQUENT REEXAMINATION IS DESIRABLE, AND RECOMMENDED THAT THE ENTIRE QUESTION BE LOOKED INTO TO DISCOVER A PRACTICAL WAY OF HANDLING THE PROBLEM, DR, WEST: Are we going to say anything about size of dose? DR, NEWILL: We should come back to the recommendation Dr, Leonards made for a discussion that three people have come back to already. DR, ALPERT: I was discussing with Dr. Sharkey this noon that really it is not a difficult decision to make if you look at it in this way, if you accept two things, namely, that with a 50-gram loading dose you are not sure you are going to pick up all of the potential or incipient diabetics, and at the 100-gram loading dose, you are going to have at least a variable problem with regard to acceptance in terms of nausea and vomiting, I say this from a pharmaceutical standpoint, Then, if you want to hit all aspects of this, you might say, in terms of a compromise, you just go to the median point, which is 75. This is the way I would try to sort of rationalize this. I would like to make this as a proposal. DR, CRAMPTON: We have gone back to talking simply about screening? DR, NEWILL: We are talking simply about screening now and it is the specific recommendation that Dr. Leonards made at length for us. DR, McDONALD: I would like to say it seems reasonable to me leaving it as is, because what we are hoping for is to put enough stress to -205~ bring out as much of the hidden diabetes as possible, or hidden dis=- turbed glucose tolerance as possible, but not enough to excite nausea, and it seems from the discussions that you have glven that 75 seems a reasonable amount that might cover this comfortably. DR, MILMORE: Particularly in view of Dr, Ricketts’ comments, we are now speaking of detection which means, for the most part, capillary blood, I would suggest any recommendations we are in a position to make here today be regarded as quite tentative, because really much more work is needed before we can make firmly-backed recommendations. DR. NEWILL: Your point is well taken. I hope it is understood that Dr, Leonards' recommendations were really an interim~type recommendation until we had more specific information available. DR. ALPERT: This is purely pragmatic, I want that very clear, I am not making any implications in any other direction. DR, O'SULLIVAN: There was no comment about dropping any loading dose at all, I think it was specifically recommended to load, DR. NEWILL: It means this group would be recommending a loading dose for the blood detection programs, DR, ALPERT: Here again I think you are going to have to allow some leeway, but I think on your point here, What are we to recommend -- I don't think there is any doubt in anyone's mind here that you get more and better information with a loading dose than you do without. DR. NEWILL: I don't think this precludes people doing detection programs without loading but the recommendation of this group is that it is pre- ferred, =206" RECOMMENDATION THE GROUP RECOMMENDED, AS AN INTERIM PROCEDURE PENDING THAT FURTHER STUDY, GLUCOSE LOADING BE INCLUDED AS A DESIGN FEATURE OF DIABETES DETECTION PROGRAMS, AND THAT THEY BE BASED ON A DETERMINATION OF BLOOD GLUCOSE. AGREEMENT WAS REACHED THAT THE PRACTICAL LEVEL FOR LOADING IS 75 GRAMS OF GLUCOSE OR ITS EQUIVALENT, DR. BURCH: The recommendation concerning the loading dose was specifically limited to the detection survey. Are we going te have any recommenda - tions, for instance, in regard to the population survey as discussed by Dr. O'Sullivan? DR. KLIMI': If and when we discuss the population surveys, I think we should. DR, O'SULLIVAN: I think it should be restricted. That is why I didn't specify the test, referring simply to "first test" etc, You can gather useful information from surveys which can then be reapplied. So I think it should be up to the investigator to have firm ideas about what he is going to do. DR, BURCH: Here again, regardless of what recommendations come out from here, the investigators will do what they decide to do, but it would be nice, it seems to me, if there were at least some recommendations on this as a guide, DR. ALPERT: Isn't this one of the problems that has to be investigated: What is a proper loading dose? We haven't had any real agreement on that. DR. NEWILL: Yes. And the reason I wasn't pressing for this, Dr, Buxch, is that when you get to the surveys you want to know why the person is doing this and every other decision depends on it, why the study is -207 ~ being done. I think the investigator has to be given the right at that particular time to vary the things he does in terms of getting an answer to the question he asks. I, for one, would hate to see us try to impose anything on top of the survey type study, DR, LEONARDS: Except for one thing, Dr, Ricketts made the recommenda- tion this morning that loading doees for population surveys be based on surface area, It was pointed out that Dr, Klimt is already doing this, that Dr, Fajans is very nearly doing this, I indicated that I thought it was an excellent idea and maybe the whole group might support this, that is, this recommendation for a specific amount per surface area, DR, WEST: I am sorry I wasn't able to be here this morning, but I think I can make a pretty good argument that the ''glucose space' and other factors influencing a test are such that the dose could be based on welght alone rather than square meters of body surface, And the second thing is that weight would make dose calculation much easier. DR, MILLER: I think the point has been made clear that further work will have to be done to define this, but I would like to suggest in this recommendation Dr. Andres' very excellent suggestion that three dose levels might be used, This might be sufficient for practical field studies, practical epidemiologic field studies, You don't have to do it down to the third significant figure, depending on whether it is 1,63 square meters or 1,64, But there might be certain broad categories, DR, KLIMI: It might be broad enough, as Dr. Andres suggested, if we have bottles at 10 gram increments, for example. The dose would depend on body weight or body surface or some other correlate, blood volume, if you want -- -208. DR. NEWILL: Obviously this depends on the studies regarding the glucose tolerance test. It seems to me we might make this recommendation just as an interim one. DR. KLIMI: Could it be worded so that the glucose loading dose ought to be scaled on the basis of some appropriate physiological measurement of the person, so that we agree on scaling but we might use different ways of scaling it, Body surface is one we have picked. DR, RICKETTS: I am afraid I have to take exception to that because of the great error you may introduce in examining very obese individuals, Here it is perfectly clear that body surface area comes much closer to representing general metabolic functions, if you consider oxygen consumption, than does weight. And when you realize, as somebody pointed out this morning, surface area may vary 100% from one individual to another, it does seem to me that it makes good sense to use a rough correlation with surface area as your criterion for load. And I would emphasize the "rough." I think Dr. Andres' suggestion is entirely feasible and adequate, DR, MILLER: You would go along with that, wouldn't you, Dr, Klimt? DR. KLIMI':t Indeed. That was the first choice we had, I was trying to accommodate Dr, West. Because if we compromise on one side by using ten-gram increments rather than taking the exact dose, the body weight in turn may be an adequate correlate. DR, RICKETTS: It may be possible but I doubt it. DR. O'SULLIVAN: I had understood Dr, Ricketts to suggest that we study this. Now we have gone a stage further and we are recommending it for epidemiologic studies, -209- DR, RICKETTS: It was intended as a recommendation, Dr, O'Sullivan. It needs to be studied, too, DR, O'SULLIVAN: There are other recommendations we could also add. For example, if postprandial testing is done, the period of time should be confined to a period as narrow as possible; and secondly, you could say that testing should be confined to either afternoon, evening, or morn- ing, because there is a diurnal variation in blood glucose, So, these items might also be taken into consideration. DR, KLIMI't But it needs to be specified here, I believe ~- it is an important point and we have studied it, too -- that in a given study the time of the day needs to be specified in which the test is applied. It makes a difference in several tests, if not in most, when you apply it, DR, WEST: I promised not so say anything but I do object to the notion that fat tissue is not metabolically active, (In a sense this is implied when the loading dose is based on square meters of surface. As adipose tissue is added the dosage is increased somewhat but not in proportion to the weight of the adipose tissue added. Thus fat people receive loading doses which are relatively low in propoxgion to their weight and the weight of their adipose tissue.) In fact, one could make an argument that fat tissue tends to sop up more of this stuff (glucose) than anything else. DR, RICKETTS: Nobody said that fat tissue is metabolically inactive. DR, WEST: That is the impression I got. DR, KLIMI': No, fimst the opposite. DR, MILLER: Dr, Ricketts suggestion incorporates that, -210. DR, WEST: So the dose would be ~- yes, I see what you mean, Second, the consideration of metabolically active tissue is not the only consideration, Another is the size of the pool and the amount of extracellular space in adipose tissue as opposed to muscular tissue and other tissue. But I think if you figured this out you would find square meters of body surface is as good as weight and maybe better, but I want to make another plea that basing dosage on weight is possibly as appropriate and a lot simpler. MR GORDON: I am not sure whether we are working our way down to a reductio ad absurdum here or in fact we have a set of recommendations which are genuinely being put in the record, I think it is desirable, if you are conducting a study where you can do this sort of thing, to record these variables with respect to each measurement you make: some anthropometric measurements you are interested in, time of last meal, interval since last meal, perhaps content of last meal, the time of day, and almost anything else that seems appropriate -- and treat these analytically in presenting your data. But I think if we are going to consider holding studies to some very rigid set of procedures, for example, always examining at the same time of the day, at a fixed period after a meal which is more or less constant -- I think we sme getting ourselves into a position where most studies which are going to produce useful results cannot be undertaken, And I think the point that is being made, or at least the moral I want to draw from the remarks that have just been made, is that these factors, which you really don't try to control in most studies, are as important as the factor of body surface in dosage, and that what I really would like to see is a -211- recommendation to study the contribution of these various factors so that we can later on justify analytical techniques in evaluating our data, rather than constraining studies to a particular set of molds which happen to seem appropriate at the present time, DR, McDONALD: I would like to say first that we are all agreed that more surveys on populations are indicated, and if we make it too complicated we will discourage more, so I would like to second what Mr. Gordon said over there about trying to get down to a definition of just one thing when other variables are equally important. RECOMMENDATION THE GROUP RECOMMENDED THAT FOR POPULATION SURVEYS, THE LOADING DOSE BE SCALED ACCORDING TO BODY SURFACE AND THAT THE SIZE OF THE LOAD AND THE PRECISION OF ITS CALCULATION BE STATED, DR. FAJANS: We have made a recommendation to relate dose of glucose to surface area, We haven't said anything about what order of magnitude this dose should be once we use this factor of surface area. I don't think I am ready, or anybody else is ready, to state an absolute dose by which to multiply the surface area, but maybe we could come to a consensus of opinion. Should this come out in the neighborhood of 50 grams, 75 grams, 100 grams, or 125 grams? We haven't accomplished enough by only specifying that we are going to make the dose related to surface area, DR, NEWILL: This is in terms of the survey, not detection programs? DR, FAJANS: Yes, DR, KLIMI: My understanding of Dr. Ricketts was that he rightfully recommended a somewhat higher dose than we are using, namely 30 grams -212- per square meter, be considered. Our scale leads to 60 grams per adult male and this could be adjusted so it leads to 75 grams on the average per adult male, DR. NEWILL: Should this be added as one of the items to be tested in the first recommendation? DR. FAJANS: If we are talking about specifics, I would say on the basis of information that I have available, such a small dose is not optimal, The only comparative studies that are available are with the one Moorhouse has gone from % gram to 2 grams of glucose per kg body weight and has shown quite unequivocally that when % gram per kg was used there was overlap between known diabetics and healthy control subjects. But he found no overlap when he used the large dose, DR, MILLER: What about a gram per kilo? DR, ALPERT: Excuse me, Dr, Fajans. This was a statement of principle, if I understood Dr, Ricketts correctly, that we would change from using a weight standard, so to speak, to a surface area standard. But by no means is there any implication that this gives you the dose. This now becomes a matter of investigation. DR. RICKETTS: That is a wholly new subject. DR, FAJANS: That is what I am discussing. DR, ALPERT: How can we decide that? You have to say this is a project which would be worth looking into, which is the optimal ratio to use. DR, FAJANS: At the moment we don't have the information that surface area is a better criterion either. We think that it is the best, and this is the rational interim way of doing it. We have to consider recommending a rational dose in the interim until we find the best. This will take 5 years and something has to be recommended in the meantime. -213- DR, RICKETTS: May I ask Dr. Klimt what surface area he is finding now? DR. KLIMI: A little more than one-and-a-half. It is 30 grams per square meter and for adult males it comes out to an average of a little less than 60 in this way. DR. RICKETTS: Purely arbitrarily I would like to throw out the sugges- tion that for the immediate future, and before final determinations are made, we try 50 grams per square meter, so that for a 1.64 meter person you get 80 grams. This is purely a suggestion. You can knock it down if you like. DR. FAJANS: I would knock it up. DR. MILLER: Forty grams would be closer. DR, KLIMI: Forty grams would be a 33% increase over our dose. That would bring it in the neighborhood of 75 on the average for adult males. DR. ALPERT: Mr. Chairman, I did want to say that my recommendation of 75 grams was pragmatic, I don't think it should be accepted for research investigation. You have to find out what is the optimal dose. DR, KLIMI: It was an interim suggestion in view of the fact that without a dose the body surface isn't going to help anyone. DR. NEWILL: The way I think this would be worded is that the recommenda- tion is that we use square meters of surface area and it was suggested that a dose of so much -- 40 grams or 50 grams per square meter would be used as the loading dose. I have no strong feeling about this, I think we have the greater number of people voting for 40 grams per square meter, so the suggestion then would carry the 40-gram level for now, but this is certainly something I think needs to be investigated and would only be an interim suggestion added to the recommendation Dr. Ricketts made. -214~ DR, BURCH: Probably in that same statement it should be stated that this should be the subject of further investigation. DR, MILLER: Let me say this again. This is sort of redundant but we have to get the answer to these questions. As I understand the discussion, the plea has been made that we can't wait until all these exact facts are obtained; therefore, we ought to make some general recommendations for people to use in this interim. And I don't think it commits the group to saying that this is the gospel truth, but here is something that investigators who want to survey now, for whatever reason have as a frame of reference, that they might be in the right ball park and wouldn't be too far off when the correct facts were obtained, DR, HAYNER: I would like to ask if the Cleveland group has any experi- ence with giving an equivalent of, 100 grams in males at least with this special solution you have developed, Dr. Leonards -- whether it tends to produce nausea when you go to that high dosage. DR, LEONARDS: No, it does not. DR. HAYNER: Would it be the recommendation of this group that this type of approach, modified challenge, be used in population investigations, or not? DR, MILLER: In some sort of more palatable solution rather than pure glucose? DR, HAYNER: Yes, MR. GORDON: I have a question because I have no feeling at all as to what this means practically. 215~ Can somebody give me just a vague notion as to what the dosage would be for a male of average height who was, say, 20% overweight. How much would such a person be given? And if this is a large amount, let's say if it is 120 grams, what the frequency, whatever the pro- cedure used ~-- what the frequency of nausea would be as a factor of increasing obesity. Because I can envisage a population study where you have fewer and fewer respondents as you go up the scale of obesity, and as a consequence you would lose a large amount of data which would be of considerable interest to you. Does somebody have a guess as to what the effect is when you use this dosage on a really obese individual? DR, MILLER: That is why I think Dr, Klimt suggested the 40 grams, We will not get into excessive doses, It will be somewhere around 75. DR, KLIMI: You would have to be extremely obese to get over 100. DR. MILLER: Fifty would get you over it considerably. DR, KLIMI': The 50 on the average would be 100, DR. ; This is another subject for investigation. DR, NEWILL: Obviously all of you have many things to go home and gear up to do. DR, HAYNER: May I ask that we suggest that population investigations or population surveys, as the word has been used, use a modified glucose challenge which does not produce nausea so frequently. DR, MILLER: Or less nausea. DR, ACHESON: You mean Leonards' mixture? DR. KLIMT: This is really a problem with children where they vomit and are in real trouble, Adults may complain. =-216~- DR, NEWILL: Is there any objection to this? DR, FAJANS: In relation to the discussion about nausea and vomiting, I would like to state that we have been doing glucose tolerance tests for fourteen years on some of the same individuals. They come back yearly and some complain of nausea, but they come back, and as far as I am aware we haven't lost a single one because of nausea, In fourteen years I can remember very few individuals who have vomited, I cannot give an exact percentage for those who have nausea. I think the frequency of vomiting has been greatly overstressed., In my experience it occurs very infrequently. DR. MILLER: This is another subject that needs careful investigation, I have asked Dr, Leonards to study this systematically, but this was not within the framework of his studies at the time and it is clear that this study will have to be done. What percentage of people get nausea at 75 grams, 100 grams, 125 grams, and so forth, and how does this differ if you use the modified glucose challenge. DR, KENT: I think there is quite a bit of difference here, Dr, Fajans is a very attractive personality and probably sees his patients him- self or they know him, whereas we don't know the patient we try to screen, They are seen by technicians and volunteers and the volunteer 1s not about to be persuasive, If somebody says "I am nauseated," they run out the other side of the room. So in a mass survey a little nausea goes a long way. -217~ Before time runs out I would like to take this opportunity to thank our hosts, Dr, McDonald, Mrs, Fisher, and Dr. Burton for a very, very exciting and stimulating meeting, and I would like to propose this as a resolution at the end of the meeting, DR. McDONALD: Let me thank all of you for responding to our invitation, I, too, have enjoyed it and learned a great deal and look forward to some more discussions on these same areas in the future, TABLE II. III. IV. VI. Vili. FIGURE 1. APPENDIX NATURAL HISTORY OF DIABETES MELLITUS PROGRESSION OF "PROBABLE DIABETES" IN 15 SUBJECTS "MATURITY-ONSET TYPE" OF DIABETES IN CHILDHOOD GLUCOSE TOLERANCE TESTS IN TEN ASYMPTOMATIC CHILDREN OF A DIABETIC FATHER COMPARISON OF GLUCOSE TOLERANCE TEST IN 38 AMBULATORY SUBJECTS COMPARATIVE RESULTS BY DECADES OF POSITIVE SCREENEES COMPARISON OF SOUTH EUCLID TESTING STANDARD ORAL GLUCOSE TOLERANCE TEST TWO-HOUR POST-GLUCOSE (100 gm) BLOOD GLUCOSE LEVELS FREQUENCY DISTRIBUTION OF SUMMED BLOOD GLUCOSE VALUES OF STANDARD ORAL GTT FOR CONTROL SUBJECTS AND APPARENTLY HEALTHY RELATIVES OF DIABETIC PATIENTS UNDER AGE 50 COMPARISON OF STANDARD ORAL GLUCOSE TOLERANCE TEST IN TWO CLASSES OF INDIVIDUALS AGE 10-39 BLOOD GLUCOSE CONCENTRATION 120 MINUTES AFTER GLUCOSE ADMINISTRATION AMONG SUBJECTS WITH NEGATIVE FAMILY HISTORIES FOR DIABETES. NOMOGRAM FOR ESTABLISHING AN INDIVIDUAL'S PERCENTILE RANK COMPARED WITH AGE COHORTS, BASED ON TWO-HOUR BLOOD GLUCOSE CONCENTRATION. PAGE 220 221 222 223 224 225 226 227 228 229 230 231 232 TABLE I NATURAL HISTORY OF DIABETES MELLITUS Prediabetegs —2 Subclinical —~————3 Latent ——3» Overt Bgrrmi— Diabetes “4 Diabetes & —— Diabetes FASTING BLOOD SUGAR Normal Normal Normal or ( f GTT Normal Norma) Abnormal Not necessary Abnormal during for diagnosis pregnancy, stress Cortisone-GIT Normal Abnormal Not ILA; Synalbumin Antagonist Vascular Changes — I | | | + | | | | | | | | | | | 1 | | | | | | t | | | | IE a fe oe a ol —- —————— ln — ————] i ————— = I ! | | + | | | | | | | | | | necessary 1 | | | | | | 3 | | 1 | NOTE: Progression (or regression) from earliest stage(s) to next may: a) never occur b) proceed slowly over many years c) be rapid or even explosive -0Z2~ TABLE II PROGRESSION OF "PROBABLE DIABETES" TO DIABETES IN 15 SUBJECTS FOLLOW-UP OF 1 MONTH TO 11 YEARS (MEAN 3 YEARS) 12 = 47 (Mean 35.2) Years FBS Peak Two-hour Level B. S. mg/100 ml mg/100 ml mg/100 ml Initial test: Probable Diabetes 95 190 116 (80 - 105) (167 - 229) (112 - 118) Follow-up test: Diabetes 99 210 167 (82 - 126) (180 = 245) (136 - 208) = ZZ A,P.S., Female Date 1/12/54 11/29/54 5/8/58 5/19/59 11/22/59 1/16/61 3/6/62 3/4/63 3/7/63 & 10 10 14 15 15 17 18 19 19 TABLE III '"MATURITY-ONSET TYPE'' OF DIABETES IN CHILDHOOD F.H. Diabetes: 1) Father - died of vascular disease 2) Mother - died of vascular disease 3) 2 brothers and one sister Glucose Tolerance Test Oral Tolbutamide Test; % of FBS: 20 minutes - 97%, 30 minutes - 85%, 40 minutes - 80% F x 1 1% 2 2% J, mg/100 ml 94 146 214 207 161 160 135 85 --- 129 146 100 118 104 82 105 136 142 150 105 --- 92 144 185 157 156 164 133 91 125 132 106 136 120 102 Delivery, M, 8 1b. 6 oz. Delivery, F, 11 1b. 8 oz. 88 = 224 --- 207 162 --- Weight Height Pounds 4'9" 118 1400 Calorie Diet 110 52%" 177 178 1400 Calorie Diet 160 sy 205 -2%%- TABLE IV GLUCOSE TOLERANCE TESTS IN TEN ASYMPTOMATIC CHILDREN OF A DIABETIC FATHER Seven Diabetic ’ | Age Sex ; F x 1 1% 2 2% 3 3 : mg/100 ml 1 1 r : 30 ] F : 322 (4+ glucose and acetone) 1 1 : : tr +H 29 ES : 147 264 337 357 355 287 229 ' . : : + 27 : M ! 210 ! ' 3 8 : 4 0 H+ 25 : M : 113 236 274 267 259 216 175 i 4 : : 0 0 21 : M : 80 146 178 141 140 129 157 1 4 I ' 1 ' 0 0 16 : F 3 98 175 210 202 160 156 154 L A . : 0 » 11 i M : 109 183 219 188 161 170 128 ’ 4 1 -£ZC~ TABLE V COMPARISON OF GLUCOSE TOLERANCE TEST WITH ORAL TOLBUTAMIDE RESPONSE TEST IN 38 AMBULATORY SUBJECTS WITH STRONG FAMILY HISTORY OF DIABETES MELLITUS Oral Glucose Tolerance Test Oral Tolbutamide Response Test Blood Sugar mg/100 ml Number of Tests Abnormal Normal FBS 115-119 6 4 2 Grossly Abnormal GTT FBS <110 lh > 200 11 8 3 ort {1 h > 180 err {1b >180 21 8 13 2 h 140-180 lh > 160 GIT {1% nh 2140 12 3 9 2 h 120-140 Total 50 23 27 ~9ZT~ =22 5m TABLE VI COMPARATIVE RESULTS BY DECADES OF POSITIVE SCREENEES All Positive Positive Negative G.T.T. Subjects Screens G.T.T. G.T.T. Not Done No. % No. % No. No. 10-19 49 20 - 29 917 15 2.0 5 0.5 6 4 30 - 39 2285 77 3.5 23 34 20 40 - 49 2852 164 6.0 83 3.0 48 33 50 - 59 1671 181 18.0 108 6.5 33 36 60 - 69 741 176 23.5 118 16.0 28 32 70 - 79 228 74 32.5 50 22.0 4 22 80 - 89 40 18 45.0 16 40.0 2 90 - 99 7 4 57.0 4 57.0 0 0 TOTALS 8790 709 8.0 407 4.5 153 149 Legend: Positive Screens Positive G.T.T. G.T.T. Not Done - Subjects with a blood glucose above 140 mg.,/100 ml. two hours after "CL" load. Subjects with any one finger blood glucose level above the accepted normal. Fasting 120; 1 hour 190; 2 hours 140. Screenees did not appear for confirmatory glucose tolerance test. -226- TABLE VII COMPARISON OF SOUTH EUCLID TESTING PROGRAMS Urine Test Finger Blood (Glucose Oxidase) Test Total Screened 802 3810 Positive Screens 8 450 G.T.T. Done 5 382 G.T.T. Positive 4 242 Final Positive % 5% 6% Age Distribution in Samples Years Urine Test Finger Blood Test 1-20 27% «5% 20 - 50 42% 68 % 50 - 80 30% 0 % BLOOD GLUCOSE mg/I00ml -227- 200 1 = MEAN 28D. 160 150 T 140 120 100 50 - A 0 T T T T T 1 F Vid | | 1/2 2 22 3 HOURS FIGURE 1. STANDARD ORAL GLUCOSE TOLERANCE TEST. 111 subjects without F.H. diabetes or large babies. Ages 20-39. -228- 175 paeaics 58 2 8 134 No on ! S o 1 ~ o hens oa ? BLOOD GLUCOSE mg/100ml nN On 1 T T TT TT T T - —y AGE 20-29 30-39 40-49 50-59 60-69 70-79 80-88 NUMBER 43 384 210 164 83 30 9 FIGURE 2. TWO-HOUR POST-GLUCOSE (100gm) BLOOD GLUCOSE LEVELS. University of Michigan faculty members (without F.H. diabetes) undergoing routine health appraisal. PERCENT OF SUBJECTS IN EACH GROUP (%) S50 40 + 30 A 20 + | I 10 A Ix I ! | ~-229~- A CONTROL SUBJECTS (124) o==0 \ \ \ \ \ \ \ i { 1 RELATIVES OF DIABETIC PATIENTS (609) max / i f | ! i ! i I I i f | i ! 4 ! | { I i 1 | \ 1 ? \ \ \ \ \ \ \ \ \ 3 \ NX ~~ 0 ep T T T ¥ T | T T T ee ee ee nme eee pry, 200- 400- 600- 800- 1000- 1200- 1400- 1600- 1800- 2000- 2400- 2800- 3200- 3600- 299 499 699 899 1099 1299 1499 1699 1899 2099 2499 2899 3299 3699 SUMMED BLOOD GLUCOSE VALUES OF STANDARD GTT (mg/100 ml) FIGURE 3. FREQUENCY DISTRIBUTION OF SUMMED BLOOD GLUCOSE VALUES OF STANDARD ORAL GTT FOR CONTROL SUBJECTS AND APPARENTLY HEALTHY RELATIVES OF DIABETIC PATIENTS UNDER AGE 50. BLOOD GLUCOSE mg/lOOml ~230~- MEAN AGE (D116 CONTROL SUBJECT (C) 26 (2)57 NON-DIABETIC OFFSPRING OF CONJUNGAL DIABETIC PARENTS 26 160 1 /60 110 + 100 HOUR FIGURE 4. COMPARISON OF STANDARD ORAL GLUCOSE TOLERANCE TEST IN TWO CLASSES OF INDIVIDUALS AGE 10-39 400 p= 30C SUBJECT Le” o EXCLUDED ~® . + ” 200 p= 100 p= BLOOD GLUCOSE CONCENTRATION mg/100 ml oL—} J 1 l 1 I x 1 3 1 20 40 60 80 100 AGE -YEARS FIGURE 5. BLOOD GLUCOSE CONCENTRATION 120 MINUTES AFTER GLUCOSE ADMINISTRATION AMONG SUBJECTS WITH NEGATIVE FAMILY HISTORIES FOR DIABETES. Divergence of the two standard deviation lines from the regression line illustrates the wider range of levels in older subjects. TEL 2 301440 ONILNI¥d LNINNYIA0D SN » 021-02Z—0 9961 10 60 300 50 250 40 200 PERCENTILE 30 150 RANK 20 100 AGE 50 TWO HOUR BLOOD GLUCOSE (mg/100ml) FIGURE 6. NOMOGRAM FOR ESTABLISHING AN INDIVIDUAL'S PERCENTILE RANK COMPARED WITH AGE COHORTS, BASED ON TWO-HOUR BLOOD GLUCOSE CONCENTRATION. Nomogram was constructed from data on subjects shown in Figure 5. -ZCZ~ > S) Pad Public Health Service Publication No.1486 U.S. DEPARTMENT OF HEALTH, EDUCATION, AND WELFARE Public Health Service it LIBRARIES 029302129