PPuurrppoossee:: Efficacy trials of preoperative erythropoietin therapy (PET) recommend a dosing schedule that cannot always be adhered to in everyday clinical practice. Consequently, we institut- ed a flexible dosing schedule and routinely offered it to anemic patients [hemoglobin (Hb)] # 130 g·L–1) undergoing total joint arthroplasty (TJA). The purpose of this observational, cohort study was to assess the effectiveness of this practice in reducing red blood cell (RBC) transfusion. MMeetthhooddss:: After obtaining Institutional Ethics Board approval, data were collected prospectively on all patients who underwent TJA at our institution from July 1999 to June 2003. Patients with baseline Hb # 130 g·L–1 were offered PET as follows: one to three sc injec- tions (20,000 IU for those # 70 kg, and 40,000 IU for those > 70 kg) every three to seven days before surgery. Since treatment was not randomly assigned, multivariable logistic regression analysis and propensity score case-control matching were used to adjust for baseline differences in patient demographics and perioperative risk factors for RBC transfusion. The adjusted relationship between PET and RBC transfusion was then determined. RReessuullttss:: Of the 1,782 patients that underwent TJA during the study period, 770 (47.9%) had a Hb < 130 g·L–1. Of these patients, 214 received PET and their RBC transfusion rate was 16.4%, whereas the transfusion rate was 56.1% in those who did not receive PET (P < 0.0001). The adjusted odds ratio of RBC transfusion with PET was 0.33 (95% confidence interval = 0.21–0.49). CCoonncclluussiioonn:: PET, used as part of routine clinical practice, is an effective blood conservation modality. Objectif : Les études de l’efficacité du traitement préopératoire à l’érythropoïétine (TPE) recommandent un dosage programmé difficile à observer au quotidien. Nous avons donc institué un programme flex- ible et l’avons offert automatiquement aux patients anémiques [hémoglobine (Hb)] # 130 g·L–1) devant subir une arthroplastie artic- ulaire totale (AAT). Notre étude observationnelle d’une cohorte veut évaluer l’efficacité de cette pratique. Méthode : Après avoir obtenu l’approbation du Comité d’éthique de l’institution, nous avons recueilli les données prospectives sur les patients qui ont subi une AAT entre juillet 1999 et juin 2003 à notre hôpital. Le TPE a été offert aux patients dont l’Hb de base était # 130 g·L–1 : de une à trois injections sc (20 000 UI pour un poids # 70 kg et 40 000 UI pour un poids > 70 kg) tous les trois à sept jours avant l’opération. Le traitement n’étant pas assigné au hasard, une analyse de régression logistique multivariée et un appariement sujet- témoin du score de propension ont été utilisés pour ajuster les dif- férences dans les caractéristiques des patients et les facteurs de risque périopératoires de transfusion de CG. La relation ajustée entre le TPE et la transfusion de CG a ensuite été déterminée. Résultats : Des 1 782 patients qui ont subi une AAT pendant la péri- ode étudiée, 770 (47,9 %) avaient une Hb < 130 g·L–1. De ces patients, 214 ont reçu le TPE et leur taux de transfusion de CG a été de 16,4 %, tandis qu’il a été de 56,1 % chez ceux qui n’ont pas reçu de TPE (P < 0,0001). Le risque relatif ajusté de transfusion de CG avec le TPE était de 0,33 (intervalle de confiance de 95 % = 0,21–0,49). Conclusion : Le TPE utilisé dans le cadre d’une pratique clinique courante est une modalité efficace de conservation du sang. 362 CAN J ANESTH 2005 / 52: 4 / pp 362–368 GGeenneerraall AAnneesstthheessiiaa Erythropoietin is an effective clinical modality for reducing RBC transfusion in joint surgery [Le traitement à l’érythropoïétine réduit efficacement les transfusions de culots globulaires en chirurgie orthopédique] Keyvan Karkouti MD,*† Stuart A. McCluskey MD,* Lucia Evans RN,‡ Nizar Mahomed MD,§ Mohammed Ghannam BSc,* Roderrick Davey MD§ From the Departments of Anesthesia* and Health Policy, Management, and Evaluation;† the Department of Nursing, Perioperative Blood Conservation;‡ the Department of Surgery, Division of Orthopedic Surgery§; University Health Network, University of Toronto, Toronto, Ontario, Canada. Address correspondence to: Dr. Keyvan Karkouti, Department of Anesthesia, Toronto General Hospital, Eaton North, 3-405, University Health Network, 200 Elizabeth Street, Toronto, Ontario M5G 2C4, Canada. E-mail: keyvan.karkouti@uhn.on.ca Dr. Karkouti is supported by a CIHR/CBS New Investigator Award. Drs McCluskey and Karkouti have received speaker’s fees and research support from Ortho-Biotech. Accepted for publication May 11, 2004. Revision accepted January 14, 2005. Karkouti et al.: PREOPEREATIVE EPO 363 HE efficacy of preoperative erythropoietin (Epo) therapy (PET) for increasing patients’ hemoglobin (Hb) concentration and reduc- ing exposure to allogeneic red blood cell (RBC) transfusion in orthopedic surgery has been demonstrated by several double-blinded randomized clinical trials.1–5 Feagan and his associates, in a multi- centre trial in which 201 patients undergoing primary total hip arthroplasty (THA) who had a pre-treatment Hb concentration of less than 137 g·L–1 and did not predonate blood were randomized into three groups to receive four weekly doses of Epo (20,000 or 40,000 IU) or placebo, found that PET increased the preoper- ative Hb concentration by more than 15 g·L–1 and reduced the frequency of RBC transfusion by more than 50%.1 Laupacis et al.5 combined the results of three of the other trials (including a total of 684 patients)2–4 and found that the overall odds ratio of RBC transfusion in patients who received PET was 0.36 [95% confidence interval (CI) 0.24–0.56]. Because of this clear efficacy, PET gained regulato- ry approval in 1996 for use in anemic patients (pre- treatment Hb concentration of > 100 g·L–1 to # 130 g·L–1) undergoing major surgery to reduce allogeneic RBC transfusions. The use of Epo for this indication, however, remains quite limited. A multinational sur- vey published in 1999 found that PET was used in fewer than 10% of hospitals in all surveyed countries except in Canada and Japan, where it was used in 13% and 51%, respectively.6 The results of a survey sent to 1,000 hospitals in the United States in 1997 showed that PET was used routinely in only 2% of the hospi- tals.7 A more recent practice review of Canadian hos- pitals (2001) found that PET was used in fewer than 2% of eligible patients undergoing total joint arthro- plasty (TJA).8 One reason for this limited use of PET in clinical practice may be the impracticality of the dosing sched- ule as recommended by the efficacy trials, particularly the required lead-time of four weeks. Another reason may be the high-cost of the recommended dosage – the full recommended dose of four 40,000 IU injec- tions of recombinant Epo costs about CAN $2,500. Because of these limitations, starting in July 1999, we implemented a more practical PET dosing sched- ule for eligible anemic patients undergoing TJA at our institution. Instead of the recommended four injec- tions of Epo, we treated patients with a maximum of three injections. Instead of using a dose of 600 IU·kg–1, we used a total of 20,000 IU for patients less than 70 kg and 40,000 IU for those more than 70 kg. In cases where the lead-time was less than three weeks, we reduced the number of injections (allowing one or two injections rather than three) or the frequency of injections (to as little as every three days apart as opposed to weekly), or both. The objective of this observational study was to assess the effectiveness of this more practical dosing schedule for PET in reducing allogeneic RBC transfu- sion in patients undergoing TJA. MMeetthhooddss Patient population and management Since July 1999, all patients who were scheduled for elective TJA (hip or knee) were assessed by our Perioperative Blood Conservation Program (PBCP) where, depending on their baseline Hb concentration (Hbbase), they were offered one of two blood conser- vation modalities as part of routine clinical care: pre- operative autologous donation (PAD) if Hbbase was > 130 g·L–1, or PET if Hbbase was # 130 g·L –1. The Hbbase cut-off was used as a general guideline: a few patients were offered PAD or PET outside the stated criteria, and in certain cases [such as revision THA or bilateral total knee arthroplasty (TKA)] both modali- ties were offered to patients. Patients were not offered PET if they had contraindications to Epo (allergy, severe coronary artery disease, recent thromboembol- ic events, uncontrolled hypertension, carotid artery disease (> 50% stenosis)), if they had active infection or inflammation, or if they had anemia other than ane- mia of chronic diseases. During the study period, the PBCP was supported by Ortho-Biotech, the Canadian supplier of Epo. This support consisted of 50% reim- bursement for the costs of Epo (with the remaining 50% provided by University Health Network) to any patient who did not have full third-party insurance coverage (approximately 60% of eligible patients) and allowed us to offer patients a more organized approach to perioperative blood conservation. Ortho- Biotech was not involved in any other aspect of the study, including treatment algorithm design, data analysis, or manuscript preparation. Patients who met the Hbbase cut-off, had sufficient time before surgery for PET therapy, and opted for treatment received one to three injections of sc Epo (20,000 IU for those # 70 kg, and 40,000 IU for those > 70 kg) every three to seven days before surgery, depending on the amount of time available to surgery. In addition, they received dietary iron supplementation (various formulations providing 100 mg elemental iron per day). These patients were considered as the treat- ment group for the purposes of this study. Patients who met the Hbbase cut-off but did not have sufficient time for PET therapy or opted against therapy were consid- ered the control group. These patients were also T advised to take dietary iron supplementation. Patients who did not meet the Hbbase cut-off or who opted for PAD were excluded from the study. Other than the above interventions, patient care was not modified in any way. The personnel in the PBCP had no input regarding any other aspect of the patients’ perioperative care, including blood product transfusion decisions. The clinical team, however, was aware of patients’ preoperative therapy. The RBC transfusion guidelines in use at the hospital during this time period are presented in Table I. Data collection After obtaining Institutional Ethics Board approval, data were collected on all patients who underwent elective TJA at our institution from July 1999 to June 2003. The following information was collected: patient demographics; American Society of Anesthesiologists’ (ASA) classification of medical sta- tus; baseline serum creatinine concentration; Hb con- centration before PET (baseline), before surgery, nadir post surgery, and at the time of hospital dis- charge; duration of surgery; exposure to blood prod- ucts; and postoperative length of hospital stay. Data analysis SAS™ version 8.2 (SAS Institute, Inc., Cary, NC, USA) was used for the statistical analyses. Categorical variables were summarized as frequencies and percent- ages, continuous variables as means and standard devi- ations if normally distributed and medians and 25th and 75th percentiles or range if not normally distrib- uted. The one-way ANOVA with Duncan post hoc analysis was used to determine the relationship between the number of Epo injections and the increase in Hb concentration. As noted above, only patients whose Hbbase was 130 g·L–1 or less and did not receive PAD were includ- ed in the analysis. The characteristics and outcomes of those who received PET were compared to those who did not receive PET using the t test or Wilcoxon rank sum test for continuous variables and the Chi-squared or Fisher’s exact test for categorical variables. Since PET was not randomly assigned, two statisti- cal methods were used to adjust for baseline differ- ences in patient demographics and perioperative risk factors for RBC transfusion. One method was the use of multiple logistic regression to adjust for any signif- icant (P # 0.3; a standard criterion for selection of can- didate variables to be used in multivariable analysis)9 between-group differences. A second method was the use of propensity analy- sis to match patients who received PET to similar patients who did not receive PET.10 The propensity score derivation model was constructed using multi- variable logistic regression to model for probability of RBC transfusion without inclusion of PET as a pre- dictor variable. This model was then used to calculate the propensity score of RBC transfusion for each patient. Using a SAS macro,A patients receiving PET were individually matched to unique control patients based on their propensity score using a 5→1 greedy matching technique (cases were first matched to con- trols that had an identical propensity score to five dig- its. Those that did not match were then matched to controls on four digits of the propensity score. This continued down to a one-digit match on the propen- sity score for those that remained unmatched).A A Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. Proceedings of the twenty-sixth annual SAS Users Group International conference, Cary, NC, USA, 2004. 364 CANADIAN JOURNAL OF ANESTHESIA TABLE I Institutional red blood cell transfusion guidelines in effect during the study period* Hemoglobin (g·L–1) Risk Strategy > 100 Very low Avoid 80–100 Low Avoid; may occasionally require transfusion based on clinical evaluation† 60–80 Moderate Try to avoid; base transfusion decision on clinical evaluation† < 60 High Usually requires transfusion †Clinical evaluation: volume, respiratory, cardiac, and cerebrovas- cular status, duration of anemia, symptoms, ongoing blood loss, presence of coagulopathy *Adapted with permission from: Winslow RM. A physiological basis for the transfusion trigger. In: Spiess BD, Counts RB, Gould SA (Eds). Perioperative Transfusion Medicine, 1st ed. Baltimore: Williams & Wilkins; 1997: 27–43. TABLE II Procedure and transfusion breakdown in the total sample (n = 1,611) Procedure n % transfused THA, primary 653 28.8% THA, revision 170 53.5% TKA, primary 679 28.1% TKA, revision 83 24.1% TKA, bilateral 26 50.0% THA = total hip arthroplasty; TKA = total knee arthroplasty. RReessuullttss During the study period, 1,782 patients underwent elective TJA at our institution; 171 of them received PAD and were excluded from analysis. Of the remain- ing 1,611 patients, 503 (31.2%) received RBC transfu- sions (see Table II for breakdown according to procedure). Upon presentation to the blood conserva- tion program, 770 (47.9%) were anemic (Hbbase # 130 g·L–1), 45.1% (347/770) of whom received RBC trans- fusions. In comparison, 18.3% (153/837) of non-ane- mic patients received RBC transfusions (P < 0.0001; four had missing Hb). Of the 770 anemic patients, 214 (27.8%) received PET. Of those who did not receive PET, 12 patients were excluded due to contraindica- tions: eight who were considered at high-risk for thromboembolic events, and four for having an active inflammatory condition. The remaining patients who did not receive PET either were not assessed early enough before their surgery or opted against treatment. The number of doses given to patients and their Hb response is presented in Table III. In the 770 anemic patients, the rate of RBC trans- fusion was 16.4% (35/214) in those who received PET compared with 56.1% (312/556) in those who did not (P < 0.0001; odds ratio = 0.15, 95% CI = 0.10–0.23). The number of RBC units transfused in the PET and no PET groups were: mean ± SD = 0.3 ± 0.8 and 1.5 ± 1.6; median (range) = 0 (0–4) and 2 (0–5), respectively; P < 0.0001. The characteristics of the two patient groups are presented in Table IV. The Hb concentration increased by 16.5 ± 9.6 g·L–1 in the PET group. The nadir and discharge Hb concentra- tions were also higher in the PET compared with the no PET group. When multivariable logistic regression was used to adjust for potential confounders, the following vari- ables remained in the model: ASA classification, age, weight, sex, renal dysfunction, baseline Hb concentra- tion, PET, duration of surgery, surgeon, and proce- dure (categorized as primary THA, revision THA, pri- mary TKA, and revision or bilateral TKA). The logis- tic regression model – which fitted the data well (Hosmer-Lemeshow Goodness-of-fit test = 4.6; P = 0.8) and was accurate (receiver operator characteristics area under the curve = 0.81) – showed that the adjust- ed odds ratio of RBC transfusion with PET was 0.33 (95% CI = 0.21–0.49; P = 0.0002). Using propensity analysis, 198 of the patients who received PET were successfully matched to 198 con- trol patients. The characteristics of the cases and con- trols are compared in Table V. Patient characteristics were similar in the two groups except for age: cases were younger than controls. The number of RBC units transfused in the cases and controls were: mean ± SD = 0.3 ± 0.8 and 1.1 ± 1.4; median (range) = 0 (0–4) and 0 (0–5), respectively; P < 0.0001. The rate of RBC transfusion was 15.7% (31/198) in the cases and 42.9% (85/198) in the controls (P < 0.0001; odds ratio = 0.24, 95% CI = 0.15–0.40). DDiissccuussssiioonn In this observational cohort study, we found that PET was highly effective in reducing perioperative RBC transfusions when it was used in anemic (Hb # 130 g·L–1) patients undergoing elective TJA as part of rou- tine practice that employed a more flexible dosing reg- imen than those used in randomized clinical trials. The most recent clinical trials used four weekly sc injections of 600 IU·kg–1 (or approximately 40,000 IU) of Epo starting three to four weeks before surgery.1,5 In our study, however, patients received one to three sc injections of 20,000 IU or 40,000 IU of Epo given every three to seven days. Despite this lower, more flexible dosing regimen, our results were comparable to those of the randomized clinical trials. In our study, PET was associated with a 16.5 ± 9.6 g·L–1 increase in the average Hb concentration and was associated with a 40% reduction in the absolute risk of RBC exposure. After adjustment for potential confounders, the odds ratio for RBC transfusion with PET was 0.33 (95% CI = 0.21–0.49). These findings were similar to those of randomized controlled trials, which found that PET increased the preoperative Hb concentration by approximately 15 g·L–1 and reduced the frequency of RBC transfusion by approximately 50%,1 and that the odds ratio of RBC transfusion with PET was 0.36 (95% CI 0.24–0.56).5 Since this was an observational study, the validity of our results depends on the appropriateness of our management of important sources of bias. One impor- tant bias is susceptibility bias, which occurs if variables that are associated with a better outcome occur more Karkouti et al.: PREOPEREATIVE EPO 365 TABLE III Hemoglobin response to Epo Number of doses of Epo n Hb increase (g·L–1) (mean ± SD)* 0 1372 0.0 1 23 8.4 ± 6.5 2 93 14.3 ± 8.0 3 105 19.8 ± 10.0 Epo = preoperative erythropoietin; Hb = hemoglobin. *Difference in response significant (P < 0.05) between each group using two- way analysis of variance and Duncan’s post hoc analysis. frequently in the treatment group.11 To control for this bias, we used multivariable modelling and propen- sity case-control matching to adjust for potential con- founders previously shown to be related to RBC transfusion in patients undergoing TJA as well as other surgical procedures.12–20 This does not, howev- er, correct for the effects of unmeasured confounders on the observed association. An example of a potential confounder is coronary artery disease, the presence of which may lead to a higher RBC transfusion trigger. Although we adjusted for patients’ ASA classification, which is a general measure of medical status, we did not correct for presence of coronary artery disease. Another type of bias is proficiency bias, which occurs if the treatment being studied is supplemented by concomitant additional therapy.11 Additional ther- 366 CANADIAN JOURNAL OF ANESTHESIA TABLE IV Comparison of patients based on preoperative Epo therapy Variable* Class Epo† No Epo† P value (n = 214) (n = 556) Pre-treatment Hb (g·L–1) 119.2 ± 7.9 118.8 ± 10.1 = 0.6 Preoperative Hb (g·L–1) 135.6 ± 12.0 118.8 ± 10.1 < 0.0001 Nadir Hb (g·L–1) 95.2 ± 13.7 83.1 ± 11.4 < 0.0001 Discharge Hb (g·L–1) 100.0 ± 10.3 96.5 ± 10.6 < 0.0001 Procedure THA: primary 39.7% 38.0% = 0.4‡ THA: revision 8.4% 12.9% TKA: primary 44.9% 43.8% TKA: revision or bilateral 7.0% 5.4% ASA III or IV 26.5% 36.5% = 0.01 Gender Female 86.0% 76.4% = 0.004 Renal dysfunction Yes 8.9% 15.1% = 0.02 Age (yr) 64.4 ± 11.9 70.2 ± 11.9 < 0.0001 Height (cm) 159 ± 13 161 ± 10 = 0.1 Weight (kg) 75.4 ± 18.1 76.3 ± 18.3 = 0.5 Duration of surgery (min) 131 ± 31 137 ± 40 = 0.1 Length of stay (days) 7 (6,8) 7 (6,9) = 0.03 *See text for definitions. †Mean (± SD) if normal distribution; median (25th, 75th percentile) if not normal distribution; percent if categor- ical. ‡Primary THA or TKA vs other. Epo = preoperative erythropoietin; Hb = hemoglobin; THA = total hip arthroplasty; TKA = total knee arthroplasty. TABLE V Comparison of propensity matched cases and controls Variable* Class EPO† No EPO† P value (n = 198) (n = 198) Pre-treatment Hb (g·L–1) 119.6 ± 7.4 121.5 ± 8.0 = 0.01 Preoperative Hb (g·L–1) 135.9 ± 11.8 121.6 ± 8.1 < 0.0001 Nadir Hb (g·L–1) 95.4 ± 13.6 85.9 ± 11.1 < 0.0001 Discharge Hb (g·L–1) 100.0 ± 10.4 96.5 ± 10.3 = 0.001 Procedure THA: primary 39.4% 45.0% = 0.5‡ THA: revision 8.6% 5.6% TKA: primary 45.0% 41.9% TKA: revision or bilateral 7.1% 7.6% ASA III or IV 24.6% 26.9% = 0.6 Gender Female 84.9% 79.8% = 0.2 Renal dysfunction Yes 8.1% 5.6% = 0.3 Age (yr) 64.5 ± 11.9 69.2 ± 10.8 < 0.0001 Height (cm) 159 ± 13 161 ± 10 = 0.1 Weight (kg) 75.5 ± 18.4 77.2 ± 17.5 = 0.4 Duration of surgery (min) 131 ± 32 135 ± 36 = 0.3 Length of stay (days) 7 (6,8) 7 (6,9) = 0.2 *See text for definitions. †Mean (± SD) if normal distribution; median (25th, 75th percentile) if not normal distribution; percent if categor- ical. ‡Primary THA or TKA vs other. Epo = preoperative erythropoietin; Hb = hemoglobin; THA = total hip arthroplasty; TKA = total knee arthroplasty. apy in this study would be the use of additional blood conservation modalities in the PET group. Our pro- tocol for iron therapy may have introduced such a bias. Preoperative oral iron therapy has been shown to improve preoperative anemia independently of PET.21 In this study, it is possible that the likelihood of iron therapy was higher in patients who received PET com- pared to those who did not receive PET since the for- mer were prescribed iron therapy whereas the latter were only advised to take iron. This difference, if it exists, could account for some of the observed treat- ment effect. Proficiency bias would also result if those making transfusion decisions, who were aware of patients’ pre- operative therapy, used lower RBC transfusion triggers in patients who received PET. This, however, is unlikely given that the nadir and discharge Hb con- centrations were higher in the PET group. A third form of bias is detection bias, which occurs if the outcome of interest is more frequently diag- nosed or detected in the treatment group due to dif- ferences in patient monitoring and follow-up.11 Detection issues were not a concern in this study because perioperative RBC transfusion is an objective outcome, and we used the most accurate data source – the transfusion laboratory database – to identify the patients who received RBC transfusions. Having excluded or managed important sources of bias, we can therefore conclude that PET is effective in routine clinical practice. But does this mean that PET should be used as part of routine clinical care? The answer to this question depends to a large extent on the cost-effectiveness of PET. In the one study that has for- mally examined the cost-effectiveness of PET, the esti- mated cost of PET per quality adjusted life year gained was about 50 million US dollars,22 a value that clearly does not favour the routine use of PET. This analysis, however, did not properly model for all the direct and indirect health-effects of RBC transfusion and its avoid- ance.23 For example, a review of the economic analyses of PAD shows that the cost-effectiveness estimates for PAD are also clearly unfavourable when these health- effects are not included in the analysis,24–26 but once they are included, the estimates become favourable.27–29 In addition, PET, by increasing Hb concentration throughout the hospital stay, may improve functional recovery as has been demonstrated in patients undergo- ing surgery for hip fracture.30,31 Moreover, the economic analysis of PET also did not account for the substantial cost-effectiveness improvements that can be achieved by modifying the dosing schedule or by improving patient selection cri- teria such that PET is offered only to those patients who are at high-risk of requiring RBC transfusion. For example, our patients received an average of about 76,000 IU of Epo, which is more than 40% lower than the dose used in the economic analysis of Epo.22 Despite this reduced dose, the treatment effect was similar to those of previous studies, the results of which were used in the economic analysis.22 This dose reduction alone, therefore, translates to a 40% improvement in the cost-effectiveness of PET. Further improvements in cost-effectiveness can be achieved if the third injection of Epo is eliminated for patients in whom the Hb concentration exceeds 130 g·L–1 after two injections. Thus, even though the ‘true’ cost-effectiveness of PET is yet to be determined, because of its excellent safety profile,32 and this study’s finding that it is effec- tive when used in routine clinical practice, PET may be the preferred method of blood conservation for anemic patients undergoing TJA. RReeffeerreenncceess 1 Feagan BG, Wong CJ, Kirkley A, et al. Erythropoietin with iron supplementation to prevent allogeneic blood transfusion in total hip joint arthroplasty. A random- ized, controlled trial. Ann Intern Med 2000; 133: 845–54. 2 Canadian Orthopedic Perioperative Erythropoietin Study Group. Effectiveness of perioperative recombinant human erythropoietin in elective hip replacement. Lancet 1993; 341: 1227–32. 3 Faris PM, Ritter MA, Abels RI. The effects of recombi- nant human erythropoietin on perioperative transfusion requirements in patients having a major orthopaedic operation. The American Erythropoietin Group. J Bone Joint Surg Am 1996; 78: 62–72. 4 de Andrade JR, Jove M, Landon G, Frei D, Guilfoyle M, Young DC. Baseline hemoglobin as a predictor of risk of transfusion and response to epoetin alfa in orthope- dic surgery patients. Am J Orthop 1996; 25: 533–42. 5 Laupacis A, Fergusson D. Erythropoietin to minimize perioperative blood transfusion: a systematic review of randomized trials. The International Study of Peri- operative Transfusion (ISPOT) Investigators. Transfus Med 1998; 8: 309–17. 6 Fergusson D, Blair A, Henry D, et al. Technologies to minimize blood transfusion in cardiac and orthopedic surgery. Results of a practice variation survey in nine countries. International Study of Peri-operative transfu- sion (ISPOT) Investigators. Int J Technol Assess Health Care 1999; 15: 717–28. 7 Hutchinson AB, Fergusson D, Graham ID, Laupacis A, Herren J, Hillyer CD. Utilization of technologies to reduce allogeneic blood transfusion in the United Karkouti et al.: PREOPEREATIVE EPO 367 States. Transfus Med 2001; 11: 79–85. 8 Arellano R, Karkouti K, Muirhead B. A review of peri- operative blood conservation practices in Canada. Can J Anesth 2002; 49: 75 (abstract). 9 Feinstein AR. Multiple logistic regression. In: Feinstein AR (Ed.). Multivariable Analysis: An Introduction. New Haven: Yale University Press; 1996: 297–330. 10 Joffe MM, Rosenbaum PR. Invited commentary: propensity scores. Am J Epidemiol 1999; 150: 327–33. 11 Feinstein AR. An additional basic science for clinical medicine: III. The challenges of comparison and mea- surement. Ann Intern Med 1983; 99: 705–12. 12 Faris PM, Spence RK, Larholt KM, Sampson AR, Frei D. The predictive power of baseline hemoglobin for transfusion risk in surgery patients. Orthopedics 1999; 22(Suppl): s135–40. 13 Benoist S, Panis Y, Pannegeon V, Alves A, Valleur P. Predictive factors for perioperative blood transfusions in rectal resection for cancer: a multivariate analysis of a group of 212 patients. Surgery 2001; 129: 433–9. 14 Hatzidakis AM, Mendlick RM, McKillip T, Reddy RL, Garvin KL. Preoperative autologous donation for total joint arthroplasty. An analysis of risk factors for allo- genic transfusion. J Bone Joint Surg Am 2000; 82: 89–100. 15 Larocque BJ, Gilbert K, Brien WF. Prospective valida- tion of a point score system for predicting blood trans- fusion following hip or knee replacement. Transfusion 1998; 38: 932–7. 16 Nuttall GA, Santrach PJ, Oliver WC Jr, et al. The pre- dictors of red cell transfusions in total hip arthroplas- ties. Transfusion 1996; 36: 144–9. 17 Bierbaum BE, Callaghan JJ, Galante JO, Rubash HE, Tooms RE, Welch RB. An analysis of blood manage- ment in patients having a total hip or knee arthroplasty. J Bone Joint Surg Am 1999; 81: 2–10. 18 Vamvakas EC, Carven JH. Allogeneic blood transfu- sion, hospital charges, and length of hospitalization. A study of 487 consecutive patients undergoing colorec- tal cancer resection. Arch Pathol Lab Med 1998; 122: 145–51. 19 Feagan BG, Wong CJ, Lau CY, Wheeler SL, Sue-A- Quan G, Kirkley A. Transfusion practice in elective orthopaedic surgery. Transfus Med 2001; 11: 87–95. 20 Karkouti K, Cohen MM, McCluskey SA, Sher GD. A multivariable model for predicting the need for blood transfusion in patients undergoing first-time elective coronary bypass graft surgery. Transfusion 2001; 41: 1193–203. 21 Andrews CM, Lane DW, Bradley JG. Iron pre-load for major joint replacement. Transfus Med 1997; 7: 281–6. 22 Coyle D, Lee KM, Fergusson DA, Laupacis A. Economic analysis of erythropoietin use in orthopaedic surgery. Transfus Med 1999; 9: 21–30. 23 Karkouti K, McCluskey S. Pro: Preoperative autologous blood donation has a role in cardiac surgery. J Cardiothorac Vasc Anesth 2003; 17: 121–5. 24 Birkmeyer JD, AuBuchon JP, Littenberg B, et al. Cost- effectiveness of preoperative autologous donation in coronary artery bypass grafting. Ann Thorac Surg 1994; 57: 161–9. 25 Goodnough LT, Grishaber JE, Birkmeyer JD, Monk TG, Catalona WJ. Efficacy and cost-effectiveness of autolo- gous blood predeposit in patients undergoing radical prostatectomy procedures. Urology 1994; 44: 226–31. 26 Etchason J, Petz L, Keeler E, Reddy RL, Beck JR. The cost effectiveness of preoperative autologous blood donations. N Engl J Med 1995; 332: 719–24. 27 Healy JC, Frankforter SA, Graves BK, et al. Preoperative autologous blood donation in total-hip arthroplasty. A cost-effectiveness analysis. Arch Pathol Lab Med 1994; 118: 465–70. 28 Blumberg N, Kirkley SA, Heal JM. A cost analysis of autologous and allogeneic transfusions in hip-replace- ment surgery. Am J Surg 1996; 171: 324–30. 29 Roberts WA, Kirkley SA, Newby M. A cost comparison of allogeneic and preoperatively or intraoperatively donated autologous blood. Anesth Analg 1996; 83: 129–33. 30 Halm EA, Wang JJ, Broockvar K, et al. Effects of blood transfusion on clinical and functional outcomes in patients with hip fracture. Transfusion 2003; 43: 1358–65. 31 Lawrence VA, Silverstein JH, Cornell JE, Pederson T, Noveck H, Carson JL. Higher Hb level is associated with better early functional recovery after hip fracture repair. Transfusion 2003; 43: 1717–22. 32 Goodnough LT, Monk TG, Andriole GL. Erythropoietin therapy. N Engl J Med 1997; 336: 933–8. 368 CANADIAN JOURNAL OF ANESTHESIA