key: cord-0001055-hcbvfpg0 authors: Aarestrup, Frank M.; Brown, Eric W.; Detter, Chris; Gerner-Smidt, Peter; Gilmour, Matthew W.; Harmsen, Dag; Hendriksen, Rene S.; Hewson, Roger; Heymann, David L.; Johansson, Karin; Ijaz, Kashef; Keim, Paul S.; Koopmans, Marion; Kroneman, Annelies; Wong, Danilo Lo Fo; Lund, Ole; Palm, Daniel; Sawanpanyalert, Pathom; Sobel, Jeremy; Schlundt, Jørgen title: Integrating Genome-based Informatics to Modernize Global Disease Monitoring, Information Sharing, and Response date: 2012-11-03 journal: Emerg Infect Dis DOI: 10.3201/eid/1811.120453 sha: 37112d613a9cee556ddfdb361655451a5c538bef doc_id: 1055 cord_uid: hcbvfpg0 The rapid advancement of genome technologies holds great promise for improving the quality and speed of clinical and public health laboratory investigations and for decreasing their cost. The latest generation of genome DNA sequencers can provide highly detailed and robust information on disease-causing microbes, and in the near future these technologies will be suitable for routine use in national, regional, and global public health laboratories. With additional improvements in instrumentation, these next- or third-generation sequencers are likely to replace conventional culture-based and molecular typing methods to provide point-of-care clinical diagnosis and other essential information for quicker and better treatment of patients. Provided there is free-sharing of information by all clinical and public health laboratories, these genomic tools could spawn a global system of linked databases of pathogen genomes that would ensure more efficient detection, prevention, and control of endemic, emerging, and other infectious disease outbreaks worldwide. Annex C: Analysis categories and mortality data sources 119 What is new in this update for 2004? 3 Regional estimates for 2004 5 World Health Organization Part 1 Overview of the Global Burden of Disease Study A consistent and comparative description of the burden of diseases and injuries, and risk factors that cause them, is an important input to health decision-making and planning processes. Information that is available on mortality and health in populations in all regions of the world is fragmentary and sometimes inconsistent. Thus, a framework for integrating, validating, analysing and disseminating such information is needed to assess the comparative importance of diseases and injuries in causing premature death, loss of health and disability in different populations. The first Global Burden of Disease (GBD) Study quantified the health effects of more than 100 diseases and injuries for eight regions of the world in 1990 (1) (2) (3) . It generated comprehensive and internally consistent estimates of mortality and morbidity by age, sex and region (4) . The study also introduced a new metric -the disability-adjusted life year (DALY) -as a single measure to quantify the burden of diseases, injuries and risk factors (5) . The DALY is based on years of life lost from premature death and years of life lived in less than full health; more information is given in Box 1. Drawing on extensive databases and information provided by Member States, the World Health Organization (WHO) prepared updated burden of disease assessments for the years 2000-2002, the most recent version being published in the World health report 2004 (6) . Following a country consultation process, country-specific estimates for 2002 were also published on the WHO web site (7) . The GBD results for the year 2001 also provided a framework for cost-effectiveness and priority setting analyses carried out for the Disease Control Priorities Project (DCPP), a joint project of the World Bank, WHO and the National Institutes of Health, funded by the Bill & Melinda Gates Foundation (8) . The GBD results were documented in detail, with information on data sources and methods, and analyses of uncertainty and sensitivity, in a book published as part of the DCPP (9) . The production and dissemination of health information for health action at the country, regional and global levels are core WHO activities mandated by the Member States in the Constitution. In her speech to the World Health Assembly in May 2007, the WHO Director-General, Dr Margaret Chan, noted, "Reliable health data and statistics are the foundation of health policies, strategies, and evaluation and monitoring". She also noted, "Evidence is also the foundation for sound health information for the general public". World Health Assembly Resolution 60.27 (WHA60.27), adopted at the Assembly in 2007, requested the WHO Director-General to "… strengthen the information and evidence culture of the Organization and to ensure the use of accurate and timely health statistics in order to generate evidence for major policy decisions and recommendations within WHO". As part of the response to this request, the WHO Department of Health Statistics and Informatics has undertaken an update of the 1990 GBD study to produce comprehensive, comparable and consistent estimates of mortality and burden of disease by cause for all regions of the world in 2004. This update builds on the previous GBD analysis for 2002; revisions, new data and methods are summarized below. The standard DALYs reported here use 3% discounting and non-uniform age weights and differ from the discounted but nonage-weighted DALYs used in the DCPP (9) . The Bill & Melinda Gates Foundation has provided funding for a new GBD 2005 study to be published in late 2010. The study is led by the Institute for Health Metrics and Evaluation at the University of Washington, with key collaborating institutions including WHO, Harvard University, Johns Hopkins University and the University of Queensland (10) . The GBD 2005 study will develop improved methods to make full use of the increasing amount of health data, particularly from developing countries, and will include a comprehensive and consistent revision of disability weights. The study will also assess trends in the global burden of disease from 1990 to 2005. The disability-adjusted life year (DALY) extends the concept of potential years of life lost due to premature death to include equivalent years of "healthy" life lost by virtue of being in states of poor health or disability (3) . One DALY can be thought of as one lost year of "healthy" life, and the burden of disease can be thought of as a measurement of the gap between current health status and an ideal situation where everyone lives into old age, free of disease and disability. DALYs for a disease or injury cause are calculated as the sum of the years of life lost due to premature mortality (YLL) in the population and the years lost due to disability (YLD) for incident cases of the disease or injury. YLL are calculated from the number of deaths at each age multiplied by a global standard life expectancy for each age. YLD for a particular cause in a particular time period are estimated as follows: YLD = number of incident cases in that period × average duration of the disease × weight factor The weight factor reflects the severity of the disease on a scale from 0 (perfect health) to 1 (death). The weights used for the GBD 2004 are listed in Annex Table A6 of Mathers et al. (11) . In the standard DALYs reported here and in recent World Health Reports, calculations of YLL and YLD used an additional 3% time discounting and non-uniform age weights that give less weight to years lived at young and older ages (6) . Using discounting and age weights, a death in infancy corresponds to 33 DALYs, and deaths at ages 5-20 years to around 36 DALYs. This update for 2004 builds on previous analyses for 2002 (6) . It does not include a complete review and revision of data inputs and estimates for every cause. The methods and data sources are described in more detail in Annex B. The main changes in the 2004 estimates are listed below. • latest death registration data reported to WHO for 112 Member States; • updated country-level mortality estimates for all Member States for 17 specific causes: HIV/ AIDS, tuberculosis (TB), diphtheria, measles, pertussis, poliomyelitis, tetanus, dengue, malaria, schistosomiasis, trypanosomiasis, Japanese encephalitis, Chagas disease, maternal conditions, abortion, cancers, war and conflict; • incorporation of cause-specific and multicause models -developed by the WHO Child Health Epidemiology Reference Group (CHERG)for causes of child deaths under five years of age and for neonatal deaths (deaths within the first four weeks after birth), with model inputs updated for the year 2004; the resulting causespecific estimates were adjusted country by country for consistency with estimated total deaths for neonates, infants and children aged under five years; • revision of cause-of-death models for countries without usable death registration data; regional patterns for detailed cause-of-death distributions were updated for African countries using a greater range of information on cause-of-death distributions in Africa. • Estimates of years lost due to disability (YLD) were revised for 52 causes where updated information for incidence or prevalence was available. Revisions resulting in significant change are noted below. For other causes, YLD estimates from the GBD 2002 were projected from 2002 to 2004 (see Annex Section B5 for details). • Incidence, prevalence and mortality for HIV/ AIDS were based on the most recent estimates released by WHO and the Joint United Nations Programme on HIV/AIDS (UNAIDS) (12) . Advances in methodology, applied to an increased range of country data, have resulted in substantial changes in estimates. The global prevalence of HIV infections for 2004 was revised from the 38 million estimated in 2006 down to 32 million -a reduction of 16%. Similarly, the estimated global deaths due to HIV/AIDS were revised from 2.7 million to 2.0 million for 2004. YLD estimates for HIV/AIDS were also revised to take into account coverage of antiretroviral drugs and associated increased survival times. • Updated estimates for vaccine-preventable childhood diseases were prepared by the WHO Department of Immunization, Vaccines and Biologicals using estimates for vaccine coverage in 2004 prepared by WHO and UNICEF (United Nations Children's Fund). • Revised incidence and mortality estimates for all forms of malaria, and for Plasmodium falciparum specifically, were based on estimates and analyses prepared by the Roll Back Malaria (RBM) Partnership, CHERG and the Malaria Epidemiology Reference Group (MERG), together with data from national case reports. Estimates for mortality for ages five years and above were revised using a transmission-intensity-based model, resulting in an increased proportion of such deaths (21% globally in 2004, compared to 10% in the GBD 2002 estimates). • Estimates for tropical diseases, including dengue fever and Japanese encephalitis, were revised to take into account the latest WHO data on populations at risk, levels of endemicity, reported cases, treatment coverage and case fatality. • Recent WHO updates of country-level prevalences of underweight, stunting and wasting in children (based on the new WHO growth standards), and anaemia prevalence, were used to update estimates for protein-energy malnutrition and iron-deficiency anaemia. • Site-specific cancer incidence and mortality estimates were updated using revised estimates of site-specific survival probabilities for 2004, together with site-specific incidence distributions from the Globocan 2002 database of the International Agency for Research on Cancer (IARC). • Diabetes incidence and prevalence estimates were updated to take into account a number of recently published population surveys that used oral glucose tolerance tests and WHO criteria to measure diabetes prevalence. • Incidence and prevalence estimates for alcohol dependence and problem use were revised based on a new review restricted to studies conducted after 1990 that used one of three high-quality survey instruments. Disability weights for alcohol use disorders were revised downwards from 0.18 to 0.122-0.137 (depending on age and sex), based on analyses of the WHO Multi-country Survey Study. • Prevalence estimates for low vision and blindness due to specific disease and injury causes were revised to take into account WHO analysis of regional distributions for causes of blindness. A recent WHO analysis of surveys measuring presenting vision loss was used to estimate YLD for an additional cause -"refractive errors". Previous GBD estimates for vision loss based on "best corrected" vision did not include correctable refractive errors. • For the calculation of YLD for ischaemic heart disease, the model used to estimate the incidence and prevalence of angina pectoris was revised using recent analyses in national burden of disease studies. These revisions resulted in an increase in the estimated global prevalence of angina pectoris from 25 million in 2002 to 54 million in 2004, and a corresponding 78% increase in YLD and 7% increase in DALYs for ischaemic heart disease. • Data from two recent national burden of disease studies were used to recalibrate the long-term case fatality rates for stroke survivors, resulting in a reduction in the estimated prevalence of stroke survivors from 50 million to 30 million, and a 30% reduction in YLD for cerebrovascular disease. Annex C This report presents estimates for regional groupings of countries (including the six WHO regions) and income groupings, with the countries grouped as high, medium or low income, depending on their gross national income per capita in 2004. The classification most commonly used for low-and middle-income countries in the report is the six WHO regions, with the high-income countries separated off as a seventh group (see map). Regional and income groupings are defined in Annex C (Tables C1 and C2 ). Detailed tables of GBD 2004 results by cause, age, sex and region are available on the WHO web site a for a range of different regional groupings, including: In Africa, death takes the young; in high-income countries, death takes the old The distribution of deaths by age differs markedly between regions. In the African Region, 46% of all deaths were children aged under 15 years, whereas only 20% were people aged 60 years and over. In contrast, in the high-income countries, only 1% of deaths were children aged under 15 years, whereas 84% were people aged 60 years and older. There were also large differences in the Asia and Pacific regions. In the South-East Asia Region, 24% of deaths were of children aged under 15 years, compared with 8% in the low-and middle-income countries of the Western Pacific Region, where 67% of deaths were of people aged 60 years and older ( Figure 2 ). Out of every 10 deaths, 6 are due to noncommunicable conditions; 3 to communicable, reproductive or nutritional conditions; and 1 to injuries The GBD study classifies disease and injury, causes of death and burden of disease into three broad cause groups: • Group I -communicable, maternal, perinatal and nutritional conditions • Group II -noncommunicable diseases • Group III -injuries. Group I causes are conditions that occur largely in poorer populations, and typically decline at a faster pace than all-cause mortality during the epidemiological transition (in which the pattern of mortality shifts from high death rates from Group I causes at younger ages to chronic diseases at older ages). Among both men and women, most deaths are due to noncommunicable conditions (Group II), and they account for about 6 out of 10 deaths globally. Communicable, maternal, perinatal and nutritional conditions are responsible for just under one third of deaths in both males and females. The largest difference between the sexes occurs for Group III, with injuries accounting for almost 1 in 8 male deaths and 1 in 14 female deaths (Figure 3 ). Figure 4 shows the distribution of deaths at all ages for 12 major cause groups (groups responsible for at least 2% of all deaths, plus maternal conditions). This illustrates the relative importance of the respective causes of death and of male-female differences. Cardiovascular diseases are the leading cause of death in the world, particularly among women; such diseases caused almost 32% of all deaths in women and 27% in men in 2004. Infectious and parasitic diseases are the next leading cause, followed by cancers, but these groupings show much smaller overall sex differentials. The largest differences between men and women are observed for intentional injuries (twice as high among men) and unintentional injuries. Maternal conditions account for 1.9% of all female deaths. The respiratory infections are treated by the GBD as a separate cause group from infectious and parasitic diseases, and are to be distinguished from respiratory diseases, which refers to noncommunicable respiratory diseases (refer to Annex Table C3 ). This report uses 136 categories for disease and injury causes. The 20 most frequent causes of death are shown in Table 1 . Ischaemic heart disease and cerebrovascular disease are the leading causes of death, followed by lower respiratory infections (including pneumonia), chronic obstructive pulmonary disease and diarrhoeal diseases. HIV/AIDS and TB are the sixth and seventh most common causes of death respectively, and together caused 3.5 million deaths in 2004. As may be expected from the very different distributions of deaths by age and sex, there are major differences in the ranking of causes between highand low-income countries ( Table 2 ). In low-income countries, the dominant causes are infectious and parasitic diseases (including malaria), and perinatal conditions. In the high-income countries, 9 out of the 10 leading causes of death are noncommunicable conditions, including four types of cancer. In the middle-income countries, the 10 leading causes of death are again dominated by noncommunicable conditions; they also include road traffic accidents as the sixth most common cause. a This category also includes other non-infectious causes arising in the perinatal period, apart from prematurity, low birth weight, birth trauma and asphyxia. These non-infectious causes are responsible for about 20% of deaths shown in this category. b Self-inflicted injuries resulting in death can also be referred to as suicides. Part 2 a Countries grouped by gross national income per capita -low income ($825 or less), high income ($10 066 or more). Note that these high-income groups differ slightly from those used in the Disease Control Priorities Project (see Annex C, Table C2 ). b This category also includes other non-infectious causes arising in the perinatal period, which are responsible for about 20% of deaths shown in this category. The relative importance of the most common cancers, in terms of numbers of deaths at all ages, is summarized in Table 3 . Globally, lung cancers (including trachea and bronchus cancers) are the most common cause of death from cancer among men, and this is also the case in five of the seven regional groupings of countries. Lung cancers are the second most common cause of male cancer deaths in the low-and middle-income countries of the Americas, and the fifth most common cause in the African Region. For males, stomach cancer mortality is second overall, being a leading cause in all regions, whereas liver cancer is the second leading cause of cancer death in the African Region. Colon and rectum cancers are the fourth leading cause and oesophagus cancer the fifth leading cause globally. Prostate cancer is sixth globally, but is the leading cause of cancer deaths in the African Region and in the low-and middle-income countries of the Region of the Americas. In the South-East Asia Region, For women, 15 cancers are ranked for each of the regions. The most common cancer at the global level is breast cancer, followed by cancers of the trachea, bronchus and lung, and stomach cancer. Breast cancer is the leading cause in four of the seven regions, second in two regions and fifth in the Western Pacific Region. Stomach cancer is the main cause of cancer death among women in that Region, followed by lung cancer and liver cancer. Cervix uteri cancer is the number one cause of cancer deaths in the South-East Asia Region and the African Region. Other cancers of the female reproductive system are the eighth (ovary) and thirteenth (corpus uteri) leading causes of cancer deaths globally. • acute respiratory infections, mainly pneumonia (17%) • diarrhoeal diseases (17%) • prematurity and low birth weight (11%) • neonatal infections such as sepsis (9%) • birth asphyxia and trauma (8%) • malaria (7%). The four communicable disease categories above account for one half (50%) of all child deaths. Undernutrition is an underlying cause in an estimated 30% of all deaths among children under five (14) . In this analysis, "undernutrition" refers to childhood malnutrition resulting in stunting and wasting, together with micronutrient deficiencies (iron, iodine, vitamin A and zinc). If the effects of suboptimal breastfeeding are also included, an estimated 35% of child deaths are due to undernutrition. a Includes other non-communicable diseases (1%) and injuries (0.3%). b ICD-10 codes Q00-Q99. Another 1.2% of neonatal deaths are due to genetic conditions classified elsewhere. c Other non-infectious causes arising in the perinatal period. d Includes all neonatal infections except diarrhoeal diseases and neonatal tetanus. Deaths in the neonatal period (0-27 days) account for more than one third of all deaths in children. Among neonatal deaths, three main causes account for 80% of all neonatal deaths: prematurity and low birth weight (31%), neonatal infections (mainly sepsis and pneumonia and excluding diarrhoeal diseases) (26%) and birth asphyxia and birth trauma (23%). Several analyses have shown that the decline in mortality in children aged under five years is falling behind the Millennium Development Goal 4 of reducing child mortality by two thirds from 1990 levels (15, 16) . For some causes -notably for measles and diarrhoeal diseases -there is evidence of a substantial decline. The GBD analysis by cause of death also shows that renewed efforts will be needed to prevent and control pneumonia and diarrhoea, and to address the underlying cause of undernutrition in all WHO regions (Figure 6 ). In the WHO African Region, increased efforts to prevent and control malaria are essential. Deaths in the neonatal period must also be addressed in all regions to achieve the Millennium Development Goal 4. In general, neonatal mortality becomes more important as mortality levels in children aged under five years decline. Cost-effective interventions are available for all major causes of death (17) . Deaths in the neonatal period -including prematurity and low birth weight, birth asphyxia and birth trauma, and other perinatal conditions based on the GBD cause list -represent between 42% and 54% of child deaths in all regions apart from the African Region, where the proportion of neonatal deaths (25%) is depressed by high numbers of postneonatal deaths, particularly those due to malaria (Figure 6 ). Among the 10.4 million deaths in children aged under five years worldwide, 4.7 million (45%) occur in the African Region, and an additional 3.1 million (30%) occur in the South-East Asia Region. The death rate per 1000 children aged 0-4 years in the African Region is almost double that of the region with the next highest rate, the Eastern Mediterranean, and more than double that of any other region (Figure 6 ). The two leading communicable disease killers in all regions are diarrhoeal diseases and respiratory infections. Deaths directly attributable to malaria occur almost entirely in the African Region, representing 16% of all under-five deaths in that region. HIV/AIDS and measles are important causes of death summarized in the "other" category. Globally, estimates suggest that 2.5% of all child deaths are associated with HIV infection. In the African Region, however -where more than 9 out of 10 of the total global number of child deaths due to HIV/ AIDS in 2004 occurred -5% of all child deaths are associated with HIV. Measles mortality, which has declined considerably in recent years, is estimated to be responsible for 4% of deaths among children aged under five years worldwide and also 4% of such deaths in the African Region. Further analyses of under-five deaths by cause show a burden distribution that is heavily skewed toward Africa ( Table 4 ). More than 9 out of 10 child deaths directly attributable to malaria, 9 out of 10 child deaths due to HIV/AIDS, 4 out of 10 child deaths due to diarrhoeal diseases and 5 out of 10 child deaths due to pneumonia occur in the WHO African Region. The ranking of regions by mortality rates among adults aged 15-59 years differs markedly from the rankings by child mortality. The European Region (low-and middle-income countries) is the WHO region with the second highest mortality level for adults aged 15-59 years; the mortality level is lower than for the African Region but higher than that for the South-East Asia Region (Figure 7) . The Eastern Mediterranean Region drops to fourth place for this age group. The difference between the high-income countries and other regions is less pronounced for adult mortality than for child mortality, due in part to the population structure -high-income countries have a higher proportion of people in the 15-59 years age group, and a higher proportion of people at the older end of this range, than lower income countries. These rankings are overshadowed by adult mortality in the African Region, which is 40% higher than for the next highest mortality region, and nearly four times higher than for high-income countries. The mortality rate due to noncommunicable diseases is highest in Europe, where nearly two thirds of all deaths at ages 15-59 years for low-and middleincome countries are associated with cardiovascular diseases, cancers and other noncommunicable diseases. Mortality rates due to noncommunicable diseases are second highest in the African Region, followed by the Eastern Mediterranean and South-East Asia regions, and lowest in the high-income countries. Injury mortality ranges from 0.5 (highincome countries) to 1.5 (European Region) per 1000 adults aged 15-59 years. The proportion of deaths in this age group due to injuries ranges from 22% (high-income countries) to 29% (the Americas) of all deaths at ages 15-59, except in Africa, where it is 13%. Group I causes of death -which include infectious and parasitic diseases, and maternal and nutritional conditions -account for more than one fifth of all deaths in adults aged 15-59 years in two regions: South-East Asia (29%) and Africa (62%). This includes 35% of the adult deaths due to HIV/ AIDS in Africa. In fact, the mortality rate among adults due to HIV/AIDS alone in Africa is higher than mortality at 15-59 years due to all causes in three other regions: high-income countries, the Americas and the Western Pacific Region. There are major differences in adult mortality by sex and major cause grouping (Figure 8) . Overall, mortality is highest among men and women in the African Region, mainly because of high mortality due to Group I causes. Men in the European Region (excluding high-income countries) had the second highest mortality rates at ages 15-59 years, considerably higher than mortality in South-East Asia, the Eastern Mediterranean and the Americas. In all regions, men had higher mortality rates than women. The largest differences were observed in Europe (male mortality 2.7 times as high as the female mortality rate), the Americas (2.0 times as high) and highincome countries (1.9 times as high). In the African Region, mortality among men is slightly higher than among women, due entirely to higher mortality through injuries. Women have higher mortality due to Group I causes. Figure 9 presents a more detailed look at the mortality rates in the African Region, by sex, for major cause groupings. At ages 15-59 years, women have much higher mortality than men for HIV/AIDS, which causes more than half of all deaths in Group I and 40% of all female deaths. Maternal conditions were associated with 14% of all deaths. In the South-East Asia Region, differences between male and female mortality were relatively small, with similar levels of mortality due to Group I causes, and somewhat higher mortality for men due to Group II and III causes. The Eastern Mediterranean Region presents a different picture, with much higher mortality among men, due almost entirely to Group III causes; that is, injuries. Figure 10 shows the distribution of male deaths due to Group III causes in the Eastern Mediterranean. War and violence caused almost 40% of these deaths, followed by road traffic accidents (31%). Injuries and cardiovascular diseases are leading causes of death among men in Europe Figure 11 illustrates the high levels of mortality among men in the low-and middle-income countries of the European Region. The main reason is the high mortality rates due to cardiovascular diseases and injuries, each associated with a mortality rate exceeding 2.5 per 1000 adults aged 15-59 years, and together being responsible for almost two thirds of overall male mortality in this age group. The most striking data from the low-and middle-income countries of the Americas relate to injury mortality, which is about 1.6 per 1000 men aged 15-59 years, making it the leading cause group (Figure 12) . Intentional injuries account for 57% of adult mortality due to injuries, while motor vehicle accidents account for 25% of adult mortality due to injuries. Road tra c accidents 30% Violence 7% War 31% Other intentional injuries 1% The years of life lost (YLL) measure is a measure of premature mortality that takes into account both the frequency of deaths and the age at which death occurs, and is an important input in the calculation of the DALYs for a disease or health condition (see Box 1, page 3). YLL are calculated from the number of deaths at each age multiplied by a global standard life expectancy for the age at which death occurs. Taking into account the age at death causes major shifts in the proportion of deaths occurring in each of the WHO Regions (Figure 13 ). Based on the distribution of the world's 58.8 million deaths in 2004, the South-East Asia Region has the highest proportion of deaths (26%), followed by the African Region (19%), the Western Pacific Region (18%) and highincome countries (14%). Based on the YLL, however, the African Region accounts for 32% of all YLL, followed by South-East Asia (30%), the Western Pacific (13%) and the Eastern Mediterranean (9%) regions. Using the YLL increases the relative importance of Africa and South-East Asia in the global picture, because people from these regions die at a relatively young age. The relative importance of the Eastern Mediterranean and the Americas change little, and the remaining three regions decline in relative importance. Figure 14 presents similar data on the proportional distribution of deaths and YLL for the leading causes of death. Taking the age at death into account causes major shifts in the relative importance of the major causes. The two most common causes of death -ischaemic heart disease (12.2% of all deaths) and cerebrovascular conditions (9.7% of all deaths) -are responsible for only 5.8% and 4.2% of YLL, respectively. The main causes of YLL are perinatal conditions (prematurity and low birth weight, birth Ageing of populations in low-and middle-income countries will result in significantly increasing total deaths due to most noncommunicable diseases over the next 25 years. Global cancer deaths are projected to increase from 7. The projected 28% increase in global deaths due to injury between 2004 and 2030 is predominantly due to the increasing numbers of road traffic accident deaths, and increases in population numbers are projected to more than offset small declines in age-specific death rates for other causes of injury. The four leading causes of death globally in 2030 are projected to be ischaemic heart disease, cerebrovascular disease (stroke), chronic obstructive pulmonary disease and lower respiratory infections (mainly pneumonia). Total tobacco-attributable deaths are projected to rise from 5.4 million in 2004 to 8.3 million in 2030, at which point they will represent almost 10% of all deaths globally. Apart from lower respiratory infections, the 10 main causes of death in 2004 included three other communicable diseases: diarrhoeal diseases, HIV/ AIDS and TB. HIV/AIDS deaths are projected to decrease by 2030, but will remain the tenth leading cause of death globally. Deaths due to other communicable diseases are projected to decline at a faster rate: TB will drop to the twentieth leading cause and diarrhoeal diseases to twenty-third. Population ageing will result in significant increases in the rankings for most noncommunicable diseases, particularly cancers. Increasing levels of tobacco smoking in many middle-and low-income countries will contribute to increased deaths from cardiovascular disease, chronic obstructive pulmonary disease and some cancers. Road traffic accidents are projected to rise from the ninth leading cause of death globally in 2004 to the fifth in 2030. Figure 16 shows projected trends in total numbers of global deaths for selected causes of death. This figure clearly illustrates the projected increases in numbers of deaths for important noncommunicable causes, and the projected declines for leading Group I causes. Projected changes in numbers of deaths may be due to changes in age-specific disease and injury death rates, or due to demographic changes that alter the size and age composition of the population, or both. Death rates are strongly age dependent for most causes, so changes in the age structure of a population may result in substantial changes in the number of deaths, even when the age-specific rates remain unchanged. The relative impact of demographic and epidemiological change on the projected numbers of deaths by cause is shown in Figure 17 . The change in the projected numbers of deaths globally from 2004 to 2030 can be divided into three components. The first is population growth, which shows the expected increase in deaths due to the increase in the total size of the global population, assuming there are no changes in age distribution. The second is population ageing, which shows the additional increase in deaths resulting from the projected changes in the age distribution of the population from 2004 to 2030. Both the population-related components are calculated assuming that the age-and sex-specific death rates for causes remain at 2004 levels. The final component, epidemiological change, shows the increase or decrease in numbers of deaths occurring in the 2030 population due to the projected change from 2004 to 2030 in the age-and sex-specific death rates for each cause. For most Group I causes, the projected reduction in global deaths from 2004 to 2030 is due mostly to epidemiological change, offset to some extent by population growth. Population ageing has little effect. For noncommunicable diseases, demographic changes in all regions will tend to increase total deaths substantially, even though age-and sex-specific death rates are projected to decline for most causes, other than for lung cancer. The impact of population ageing is generally much more important than that of population growth. For injuries, demographic change also dominates the epidemiological change. The total epidemiological change for injuries is small in most regions, because the projected increase in road traffic fatalities is offset by projected decreases in death rates for other unintentional injuries. a The dark blue bars show the total projected change in the annual numbers of deaths (in millions) from 2004 to 2030 for a given cause group. The dark orange bars show the change in the annual numbers of deaths that would have occurred due to epidemiological change only (changes in age-and sex-specific death rates) if the population size and age structure had remained unchanged. The light orange bars show the change that would have occurred due to population growth only, if the age structure had remained unchanged, and age-and sex-specific death rates had also remained unchanged. The purple bars show the change that would have occurred due to changes in the age distribution of the population only, if the size of the population had remained constant, and the age-and sex-specific death rates also remained unchanged. 9. How many people become sick each year? The "incidence" of a condition is the number of new cases in a period of time -usually one year ( Table 5) . For most conditions in this table, the figure given is the number of individuals who developed the illness or problem in 2004. However, for some conditions, such as diarrhoeal disease or malaria, it is common for individuals to be infected repeatedly and have several episodes. For such conditions, the number given in the table is the number of disease episodes, rather than the number of individuals affected. It is important to remember that the incidence of a disease or condition measures how many people are affected by it for the first time over a period of Episodes of illness. c Incidence of congestive heart failure due to rheumatic heart disease, hypertensive heart disease, ischaemic heart disease or inflammatory heart diseases. d Incidence of injuries severe enough to require medical attention. e An entry of 0.0 in the table refers to an incidence of less than 0.05 million (less than 50 000). Disease incidence, prevalence and disability . Incidence does not measure how many people have a disease at any given moment (this is "prevalence") or how badly their lives are affected. A health problem or disease can have a relatively low incidence but cause death or disability, and will therefore result in a high burden of disease or many life years lost. Conversely, some common illnesses may cause a much smaller burden of disease or fewer life years lost. Data on the contribution of various conditions and diseases to the burden of disease in a community are given in later sections. Of the diseases listed in Table 5 , diarrhoeal disease affects far more individuals than any other illness, even in regions that include high-income countries. Pneumonia and other lower respiratory tract infections are the second most common cause of illness globally, and in all regions except Africa. Other common illnesses -such as upper respiratory tract infections (including the common cold) and allergic rhinitis (hay fever) -have not been included in Table 5 . More cancers occur in high-income countries than in low-and middle-income countries. Cervix cancer is the only type of cancer more common in the African and South-East Asia regions than in highincome countries. In part, this is due to the age of the populations in different regions, because most cancers affect older adults; also, some cancers, such as prostate cancer, are much more common in older men than in younger men. Another factor contributing to the distribution of a type of cancer is the number of people exposed to causes, such as cigarette smoking in the case of lung cancer, and hepatitis B virus in the case of liver cancer. Globally, lung cancer is the most common cancer ( Table 6) , followed by breast cancer, then colon and rectum cancer, and stomach cancer. Lung cancer is also the leading cancer in the Western Pacific Region, but is less common than colon and rectum cancers or breast cancers in most other regions. Cervix cancer is the cancer with the highest incidence in the African and South-East Asia regions, even though it occurs only in women. Variations across regions in the risk of cancer are best shown using age-standardized incidence rates that apply the estimated age-and sex-specific incidence rates for cancers in each region to the WHO World Standard Population (22) . This estimates how many cases of cancer would occur in that population if it experienced the cancer incidence rates of a given region (Figure 18 ). Part 3 11 . How many people are sick at any given time? The prevalence of an illness or condition is the number of individuals who have the condition at any moment. In some cases, such as epilepsy or migraine, individuals will not have symptoms most of the time, but still have the condition. The effects of the illness and the loss of health will vary from one individual to another. The result may be serious impairments and disability affecting a person's ability to work or take part in family and community activities, or only mild impairments or disability. Prevalence data therefore do not capture the burden of disease experienced by individuals in terms of lost health. The conditions that affect the largest number of individuals at any given moment are not dramatic, and are thus easily overlooked and underestimated ( Table 7) . Worldwide, at any given moment, more individuals have iron-deficiency anaemia than any other health problem. Even in high-income countries, iron deficiency anaemia is common. Other very common conditions, with varying levels of severity, include asthma, arthritis, vision and hearing problems, migraine, major depressive episodes and intestinal worms. The previous sections presented estimates of numbers of new and current cases for various diseases and injuries. A disease or injury may have multiple disabling effects of various levels of severity, and cause varying degrees of health problems. The GBD links average loss of health to disease and injury causes through the disability weights (see Box 1, page 3). The term disability has a number of different meanings and, in particular, is not seen by some as a synonym or proxy for "loss of health". However, the GBD uses the term disability to refer to loss of health, where health is conceptualized in terms of functioning capacity in a set of health domains such as mobility, cognition, hearing and vision. The original GBD study established severity weights for approximately 500 disabling sequelae of diseases and injury, in a formal study involving health workers from all regions of the world. These weights were then grouped into seven classes, where class I has a weight between 0 and 0.02, and class VII a weight between 0.7 and 1 ( Table 8) . Participants in the study estimated distributions across the seven classes for each sequela. Distributions across disability classes were estimated separately for treated and untreated cases where relevant; distributions could also vary by age group and sex. These distributions were applied to prevalence estimates from the GBD 2004 study to estimate the prevalence of disability by severity class in 2004. Results are presented here for the prevalence of: • "severe" disability, defined as severity classes VI and VII (the equivalent of having blindness, Down syndrome, quadriplegia, severe depression or active psychosis) -see Table 8 ; • "moderate and severe" disability, defined as severity classes III and greater (the equivalent of having angina, arthritis, low vision or alcohol dependence). Prevalence estimates were restricted to sequelae lasting, on average, six months or more. The GBD prevalence estimates cannot be added easily, because they were calculated without regard for multiple pathologies or comorbidities; thus, a given individual would be counted more than once if they had more than one diagnosis. Overall disability prevalence estimates presented here were adjusted for comorbidity using a method that takes account of the increased probability of having certain pairs of conditions (23) . Limited self-reported data were available on comorbidity levels in populations, so the adjusted disability prevalences presented here have quite high levels of uncertainty. Part 3 Of the world's population of nearly 6.5 billion in 2004, 18.6 million (2.9%) were severely disabled and another 79.7 million (12.4%) had moderate longterm disability, according to the definitions given above. Disability prevalences rise strongly with age ( Figure 19 ). The average global prevalence of moderate and severe disability ranges from 5% in children aged 0-14 years, to 15% in adults aged 15-59 years, and 46% in adults aged 60 years and older. At all ages, both moderate and severe levels of disability are higher in low-and middle-income countries than in high-income countries; they are also higher in Africa than in other low-and middle-income countries (Figure 19 ). Older people make up a greater proportion of the population in high-income countries, but have lower levels of disability than their counterparts in low-and middle-income countries. Disability is also more common among children in the low-and middle-income countries. Moderate disability rates are similar for males and females in high-income countries, but females have somewhat higher rates of severe disability. In low-and middleincome countries, male and female disability rates are similar, although females aged 15-59 years tend to have higher levels of moderate disability in Africa, the Eastern Mediterranean and the Western Pacific. The most common causes of disability globally are adult-onset hearing loss and refractive errors. Mental disorders such as depression, alcohol use disorders and psychoses (e.g. bipolar disorder and schizophrenia) are also among the 20 leading causes of disability ( Table 9 ). The pattern differs between the highincome countries and the low-and middle-income countries. In the lower income countries, many more people are disabled due to preventable causes such as unintentional injuries and infertility arising from unsafe abortion and maternal sepsis. The data also demonstrate the lack of interventions for easily treated conditions such as hearing loss, refractive errors and cataracts in low-income countries. Disability due to mental disorders is more common among people aged 0-59 years, whereas chronic diseases such as dementias, chronic obstructive pulmonary disease and cerebrovascular disease are more common in older populations. In low-income countries, disability due to unintentional injuries, among the younger population, and cataracts, among the older population, are far more common. The GBD prevalence estimates are based on systematic assessments of the available data on incidence, prevalence, duration and severity of a wide range of conditions. However, these assessments are often based on inconsistent, fragmented and partial data from different studies, meaning that there are still substantial data gaps and uncertainties. Improving the population-level information on the incidence, prevalence and states of health associated with major health conditions remains a major priority for national and international health and statistical agencies. Clinically and conceptually, it is not usual practice to infer disability from diagnoses. In future revisions of the GBD study, increased effort will be devoted to direct estimation of the prevalences of impairments and disabilities, and to ensuring consistency with the estimates for disease-and injuryspecific sequelae. Population survey data on disability prevalence are limited in availability and comparability. The estimates derived from the GBD have the virtue of comprehensiveness, and at least some grounding in disease prevalence. However, they are very much approximations, and are subject to very clear limitations in the way they were compiled. These estimates are presented to give an indication of the regional prevalences of long-term disability implied by the GBD analyses. The data presented in the sections above concern the number of new cases of diseases and injuries (incidence), and the number of individuals living with diseases or injuries and their sequelae (prevalence). These counts of incidence or prevalence of diseases in populations do not take into account the relative severity or health loss associated with different conditions, and hence do not capture the burden of disease experienced by individuals. The disability weights used in the GBD convert the years lived with various health conditions to equivalent lost years of full health. The disability weights used in the GBD 2004 are listed in detail elsewhere (24) . As explained in Box 1 (see page 3), YLD measure the equivalent years of healthy life lost through time spent in states of less than full health. When all the years of life with reduced capability for all the sufferers of each condition are added up and weighted by the disability weight, a total of YLD for each condition is obtained. YLD estimates are restricted to loss of health experienced by individuals, and do not take into account other aspects of quality of life or well-being, or the impacts of a person's health condition on other people (except as far as they experience directly assessed losses of health themselves). The 10 leading causes of YLD are shown in Table 10 for males and females, and in Table 11 for high-income and low-and middle-income countries. The overall burden of non-fatal disabling conditions is dominated by a relatively short list of causes, particularly a number of neuropsychiatric conditions and sense organ disorders. In all regions, neuropsychiatric conditions are the most important causes of disability, accounting for around one third of YLD among adults aged 15 years and over. The disabling burden of neuropsychiatric conditions is almost the same for males and females, but the major contributing causes are different. While depression is the leading cause for both males and females, the burden of depression is 50% higher for females than males. Females also have a higher burden from anxiety disorders, migraine and Alzheimer and other dementias. In contrast, the male burden for alcohol and drug use disorders is nearly seven times higher than that for females, and accounts for almost one third of the male neuropsychiatric burden. In both low-and middle-income countries, and high-income countries, alcohol use disorders are among the 10 leading causes of YLD. This includes only the direct burden of alcohol dependence and problem use. The total attributable burden of disability due to alcohol use is much larger. Curable disorders of vision (cataracts and refractive errors) cause 9% of YLD in men and women aged 15 years and over; adult-onset hearing loss accounts for another 6.5% in men and 5.6% in women. Adultonset hearing loss is extremely prevalent -more than 27% of males and 24% of females aged 45 years and over experience mild hearing loss or greater (hearing threshold of 26 decibels or greater in the better ear). The GBD 2004 has estimated only the burden of moderate or greater hearing loss (hearing threshold of 41 decibels or greater in the better ear). Childhood-onset hearing loss is not included in this cause category because most childhood hearing loss is due to congenital causes, infectious diseases, or other diseases or injury. It is included as sequelae for such causes in the estimation of burden of disease. Perhaps surprisingly, around 90% of global non-fatal health outcomes (as measured by YLD) occur in low-and middle-income countries, and nearly half (44%) of all YLD fall in low-income countries. Although the prevalence of disabling conditions such as dementia and musculoskeletal disease are higher in countries with long life expectancies, this is offset by lower contributions to disability from conditions such as cardiovascular disease, chronic respiratory diseases and long-term sequelae of communicable diseases and nutritional deficiencies. In Disease incidence, prevalence and disability but also live a higher proportion of their lives in poor health. The measures of ill-health used so far (incidence, prevalence and YLL) do not give a good indication of the burden of disease borne by individuals in different communities. The summary measure used to give an indication of the burden of disease is the DALY (see Box 1, page 3). One DALY represents the loss of the equivalent of one year of full health. Using DALYs, the burden of diseases that cause early death but little disability (eg. drowning or measles) can be compared to that of diseases that do not cause death but do cause disability (e.g. cataract causing blindness). As described in the Introduction, DALYs for 2004 combine the following: • YLL for years of life lost due to deaths in 2004 • YLD for equivalent healthy years of life lost through living in states of less than full health for cases of disease and injury incident in 2004. The global average burden of disease across all regions in 2004 was 237 DALYs per 1000 population, of which about 60% was due to premature death and 40% to non-fatal health outcomes. The contribution of premature death varied dramatically across regions, with YLL rates seven times higher in Africa than in high-income countries ( Figure 20) . In contrast, the YLD rates were less varied, with Africa having 80% higher rates than high-income countries. South-East Asia and Africa together bore 54% of the total global burden of disease in 2004, although they account for only about 40% of the world's population. The Western Pacific Region has the "healthiest" low-and middle-income countries, with countries such as China now having life expectancies similar to those of many Latin American countries, and higher than those in some European countries. The high levels of burden of disease for the WHO African, South-East Asia and Eastern Mediterranean regions compared to other regions are predominantly due to Group I conditions (communicable diseases, and maternal, perinatal and nutritional conditions), although injury DALY rates are also higher than in other regions ( Figure 21 ). European low-and middle-income countries have a substantially higher noncommunicable disease burden than high-income countries (Figure 21) . They also have a higher burden due to Group I causes and Group III causes (injuries). In fact, these countries have the highest proportion of burden due to injuries (16%) of all the regions, followed by the low-and middleincome countries of the Americas. Almost one half of the disease burden in low-and middle-income countries is now from noncommunicable diseases. Ischaemic heart disease and stroke are the largest sources of this burden, especially in the low-and middle-income countries of Europe, where cardiovascular diseases account for more than one quarter of the total disease burden. Injuries accounted for 17% of the disease burden in adults aged 15-59 years in 2004. In the low-and middleincome countries of the Americas, Europe and the Eastern Mediterranean Region, more than 30% of the entire disease and injury burden among men aged 15-44 years was from injuries. Burden of disease: DALYs Measured in DALYs, 36% of the total disease and injury burden for the world in 2004 involved children aged less than 15 years, and almost 50% involved adults aged 15-59 years. The disease burden for children falls almost entirely in low-and middle-income countries (Figure 22 ). While the proportion of the total burden of disease borne by adults aged 15-59 years is similar in both groups of countries, the remaining burden is predominantly among those aged 60 years and older in high-income countries. DALYs are attributed to the age at which the disease, injury or death occurred. Some of the YLD associated with DALYs for children will be lived at older ages. While the two leading causes of death -ischaemic heart disease and cerebrovascular disease -remain among the top six causes of burden of disease (Table 12) , four primarily non-fatal conditions are also among the 20 leading causes of burden of disease; these are unipolar depressive disorders, adultonset hearing loss, refractive errors and alcohol use disorders. This again illustrates the importance of taking non-fatal conditions into account, as well as deaths, when assessing the causes of loss of health in populations. The two leading causes of burden of disease in the world are infectious diseases -lower respiratory infections and diarrhoeal diseases. HIV/AIDS is now the fifth cause of burden of disease globally, and three other infectious diseases also appear in the top 15 causes ( Table 12) . The leading causes of burden of disease in lowincome countries were broadly similar to those for the world in 2004, apart from malaria and TB (Table 13) . Of the top 10 causes, 8 were Group I, but the leading causes in high-income countries were all noncommunicable diseases, with the exception of road traffic accidents (tenth leading cause). The leading causes in high-income countries included three diseases (unipolar major depression, adultonset hearing loss and alcohol use disorders) for which direct mortality is low. Unipolar depression makes a large contribution to the burden of disease, being at third place worldwide and eighth place in low-income countries, but at first place in middle-and high-income countries. Effective treatments for depression are available, suggesting that this burden could be reduced. Cigarette smoking is a major and entirely preventable cause of burden of disease in middle-and high-income countries. Chronic obstructive pulmonary disease is in fifth place in middle-income countries and seventh place in high-income countries, and lung cancer is in ninth place in high-income countries. Cigarette smoking also contributes to the burden of disease from ischaemic heart disease and cerebrovascular disease, and affects communities in low-income countries as well. Alcohol use disorders are another important preventable contributor to burden of disease in middle-and high-income countries. The WHO regions fall into two groups -those in which the burden of disease is dominated by infectious disease, and those in which the burden of disease is dominated by vascular disease and depression (Table 14) . In Africa, HIV/AIDS, lower respiratory infections and diarrhoeal disease are the leading causes of burden of disease, whereas in the Eastern Mediterranean and in South-East Asia, lower respiratory infections and diarrhoeal disease are the two leading causes. In all three of these regions, problems during pregnancy and childbirth are important and preventable causes of burden of disease. The role of road traffic accidents in these regions, and of war and conflict in the Eastern Mediterranean, should also be noted. Unipolar depression is one of the three leading causes of burden of disease in the WHO regions of the Americas, Europe and the Western Pacific. Ischaemic heart disease or cerebrovascular disease are also consistently leading causes of death in these World Health Organization Part 4 regions. The role of violence in the Americas as the second leading cause of burden of disease, and the role of chronic obstructive pulmonary disease in the Western Pacific as the third leading cause, are also notable. Alcohol use and road traffic accidents, consistently causing about 6% of DALYs, are also important in these regions. The leading 10 causes of the burden of disease in 2004 included 4 communicable diseases. HIV/AIDS was the fifth leading cause of the burden of disease globally in 2004 and the leading cause in the WHO African Region, where it was followed by lower respiratory infections, diarrhoeal diseases and malaria. The WHO regions of South-East Asia, the Eastern Mediterranean and Africa are affected by a dual burden of disease (Table 14) . These WHO regions are much more heavily burdened by infectious disease and conditions related to pregnancy and childbirth than other regions, but they also suffer severely from the problems that affect people in high-income countries -cardiovascular disease, depression and injury. Table C2 ). b This category also includes other non-infectious causes arising in the perinatal period apart from prematurity, low birth weight, birth trauma and asphyxia. These non-infectious causes are responsible for about 20% of DALYs shown in this category. a This category also includes other non-infectious causes arising in the perinatal period apart from prematurity, low birth weight, birth trauma and asphyxia. These non-infectious causes are responsible for about 20% of DALYs shown in this category. Part 4 Depression is the leading cause among young adult women Mental disorders are an important source of lost years of healthy life for women aged 15-44 years. They make up 3 of the 10 leading causes of disease burden in low-and middle-income countries, and 4 of the leading 10 in high-income countries; selfinflicted injuries are also in the leading 10 causes for low-and middle-income countries ( Figure 23 ). Depression is the leading cause of disease burden for women in both high-income and low-and middleincome countries. Injuries are also important for women aged 15-44 years, although road traffic accidents are the eighth leading cause globally, followed by self-inflicted injuries in ninth place. Although injuries become more important for boys beyond infancy, the causes of burden of disease are broadly similar for boys and girls. However, striking sex differences emerge in adulthood (ages 15-59 years). The burden of reproductive problems is almost entirely confined to low-and middle-income countries, but it is so great that maternal conditions make up 2 out of the 10 leading causes of disease burden in women aged 15-44 years. Together with HIV/AIDS, maternal conditions are a major contributor to the high burden of disease for women in Africa relative to other regions. The burden of maternal conditions in the African and South-East Asia regions is responsible for 8% of the total global burden of disease for women aged 15-59 years. Almost all of this loss of healthy years of life is avoidable. Worldwide, and particularly in low-income countries, better care for women in pregnancy and childbirth could make a large contribution to reducing the burden of disease. The Millennium Development Goal of giving all women access to a skilled birth attendant when they give birth is directed at substantially reducing the burden of disease by avoiding preventable maternal and neonatal deaths. HIV/AIDS is the most important single cause of burden of disease for women aged 15-59 years in Africa (Figure 24) , and the per capita burden of HIV is 40% higher for women than for men. Neuropsychiatric conditions are responsible for 22% of global DALYs for women aged 15-59 years, the largest cause group in all regions outside Africa. Sense organ disorders are another important cause group, responsible for 8% of global DALYs for women aged 15-59 years. Causes of vision loss are responsible for more than two thirds of the DALYs for sense organ disorders in women; causes of hearing loss account for most of the rest. The burden of noncommunicable diseases now accounts for nearly half of the global burden of disease (all ages). Surprisingly, almost 45% of the adult disease burden in low-and middle-income countries globally is now attributable to noncommunicable disease. Population ageing and changes in the distribution of risk factors have accelerated the noncommunicable disease share of total disease burden in many developing countries. Part 4 Noncommunicable diseases dominate the disease burden of high-income countries, and in the past they have often been seen as a health priority mainly for high-income countries. In part this reflects the older population structure of the high-income countries, because noncommunicable disease risks generally increase with age. If the effects of different age distributions of populations are controlled for through age-standardization of DALY rates, it becomes apparent that noncommunicable disease risks, as measured by age-standardized DALY rates, are higher in low-and middle-income countries than in high-income countries (Figure 25 ). This is mainly due to cardiovascular diseases, principally ischaemic heart disease and stroke, whose agestandardized burden is substantially higher in lowand middle-income countries than in high-income countries. The burden of sense disorders, principally vision impairment and hearing loss, is also greater in low-and middle-income countries than in highincome countries. Injuries accounted for 17% of the disease burden in adults aged 15-59 years in 2004. In the low-and middle-income countries of the Americas, Europe and the Eastern Mediterranean Region, more than 30% of the entire disease and injury burden among men aged 15-44 years was from injuries. Globally for both sexes, road traffic accidents are the third leading cause of burden in that age-sex group, preceded only by HIV/AIDS and unipolar depression. The burden of road traffic accidents is increasingespecially in the developing countries of sub-Saharan The category of intentional injuries includes selfinflicted injuries and suicide, violence and war. This type of injury accounts for an increasing share of the burden, especially among economically productive young adults. In developed countries, suicides are the largest source of intentional injury burden, whereas in developing regions violence and war are the larger source. Countries of the former Soviet Union and other high-mortality countries of Eastern Europe have rates of injury, death and disability among males that are similar to those in sub-Saharan Africa (Figure 26) . The death rate due to poisoning is much higher in the low-and middle-income countries of Europe than in any other region of the world. Alcohol overdose deaths are likely to be a primary contributor to this situation. The death rate for injuries due to fire is much higher for women in South-East Asia than for men or women in any other region of the world. Global burden of disease per capita is projected to decrease Global DALYs are projected to decrease from 1.53 billion in 2004 to 1.36 billion in 2030, an overall decline of about 10%. Since the population increase is projected to be 25% over the same period, this represents a significant reduction in the global per capita burden. The DALY rate decreases at a faster rate than the overall death rate because of the shift in age at death to older ages, associated with fewer YLL. Even assuming that the age-specific burden for most non-fatal causes remains constant into the future, and hence that the overall burden for these conditions increases with the ageing of the population, there is still an overall projected decrease in the global burden of disease per capita of 30% from 2004 to 2030. This decrease is largely driven by projected levels of economic growth in the projection model. If economic growth is slower than in recent World Bank projections, or risk factor trends in lowand middle-income regions are adverse, then the global burden of disease will fall more slowly than projected. The proportional contribution of the three major cause groups to the total disease burden is projected to change substantially. Group I causes are projected to account for 20% of total DALYs lost in 2030, compared with just under 40% in 2004. The noncommunicable disease (Group II) burden is projected to increase to 66% in 2030, and to represent a greater burden of disease than Group I conditions in all income groups, including low-income countries. Figure 27 shows the changes in the leading causes of DALYs globally from 2004 to 2030. The three leading causes of DALYs in 2030 are projected to be unipolar depressive disorders, ischaemic heart disease and road traffic accidents. Lower respiratory infections drop from leading cause in 2004 to sixth leading cause, and HIV/AIDS drops from fifth leading cause in 2004 to ninth leading cause in 2030. Lower respiratory infections, perinatal conditions and diarrhoeal diseases are all projected to decline substantially in importance. On the other hand, diabetes mellitus, road traffic accidents, chronic obstructive pulmonary disease, hearing loss and refractive errors are all projected to move up three or more places in the rankings. Ischaemic heart disease, cerebrovascular disease and unipolar depressive disorders move up two places in the rankings to become three of the four leading causes of disease and injury burden in 2030. These projections represent a vision of an improving future for population health under: • an explicit set of assumptions • specific projections of income and human capital • specific projections of future trends in tobacco smoking, HIV/AIDS transmission and survival, and overweight and obesity. Under these projections, people in all regions of the world will live longer and with lower levels of disability, particularly from infectious, maternal, perinatal and nutritional conditions. But if there is no sustained and additional effort to address Millennium Development Goals, neglected tropical diseases, tobacco smoking and other chronic disease risks, or if economic growth in low-income countries is lower than the forecasts used here, then the world may achieve slower progress and experience widening of health inequalities. Burden of disease: DALYs 8 0 8 97 69 28 26 14 12 Bladder cancer 21 0 21 62 37 25 33 8 25 Lymphomas and multiple myeloma 25 2 24 79 55 24 55 19 37 Leukaemia 20 1 19 67 39 29 71 11 60 Other neoplasms 22 0 22 41 32 9 33 12 20 Diabetes mellitus 61 6 55 155 98 58 210 33 176 Endocrine disorders 29 2 27 37 30 8 65 11 55 Neuropsychiatric disorders 95 2 93 287 211 76 197 38 159 Unipolar depressive disorders 1 0 1 2 2 0 0 0 0 Bipolar affective disorder 0 0 0 0 0 0 0 0 0 Schizophrenia 2 0 2 3 1 2 7 1 6 Epilepsy 11 0 10 15 7 8 27 2 25 Alcohol use disorders 3 0 3 26 13 14 16 2 14 Alzheimer and other dementias 14 0 14 137 125 12 74 19 55 Parkinson disease 3 0 3 29 25 4 36 6 30 Multiple sclerosis 1 0 1 8 4 3 2 0 2 Drug use disorders 34 0 34 15 4 Injuries 485 19 466 789 186 604 1 196 119 1 077 Unintentional 321 16 305 564 131 432 846 68 778 Road traffic accidents 146 7 139 129 38 91 336 21 314 Poisonings 17 0 17 107 6 101 59 2 57 Falls 24 2 23 79 40 38 134 12 122 Fires 29 0 29 23 3 20 16 2 14 Drownings 30 2 28 34 4 30 139 7 132 Other unintentional injuries 76 5 71 191 40 152 163 23 140 Intentional 163 3 160 226 54 171 348 51 297 Self-inflicted injuries 36 2 35 151 49 102 286 49 236 Violence 25 1 24 65 5 60 57 2 3 0 3 3 2 1 0 0 0 Diphtheria 17 0 17 1 0 1 1 0 1 Measles 1 533 0 1 533 12 1 12 252 0 252 Tetanus 856 2 854 3 0 2 343 0 343 Meningitis 1 068 12 1 056 333 50 284 910 13 897 Hepatitis B e 281 6 275 136 32 104 487 30 457 Hepatitis C e 121 2 119 71 34 37 229 40 188 Malaria 1 412 1 1 411 4 1 3 169 0 169 Tropical-cluster diseases 574 11 563 7 0 7 128 0 128 Trypanosomiasis 62 0 62 0 0 0 0 0 0 Chagas disease 0 0 0 0 0 0 0 0 0 Schistosomiasis 145 8 136 0 0 0 13 0 13 Leishmaniasis 281 3 278 6 0 6 51 0 51 Lymphatic filariasis 75 0 75 1 0 1 65 0 65 Onchocerciasis 11 0 11 0 0 0 0 0 0 Leprosy 22 0 22 0 0 0 13 0 13 Dengue 28 6 22 0 0 0 169 0 169 Japanese encephalitis 0 0 0 4 0 4 186 2 183 Trachoma 208 0 208 0 0 0 419 0 419 Intestinal nematode infections 269 0 269 0 0 0 955 22 933 Ascariasis 162 0 161 0 0 0 308 5 303 Trichuriasis 61 0 61 0 0 0 269 6 0 0 0 35 31 3 389 389 0 Tetanus 0 0 0 20 15 3 143 132 10 Meningitis 3 1 1 73 43 21 263 176 64 Hepatitis B e 7 0 3 42 2 28 56 9 36 Hepatitis C e 13 0 5 20 1 13 21 3 14 Malaria 0 0 0 32 27 4 857 817 36 Tropical-cluster diseases 0 0 0 21 2 10 131 43 74 Trypanosomiasis 0 0 0 3 1 1 50 21 and multiple myeloma 119 0 20 117 5 51 97 11 41 Leukaemia 78 2 15 134 18 69 64 13 34 Other neoplasms 61 1 7 63 3 26 40 5 19 Diabetes mellitus 224 0 27 523 1 125 393 5 118 Endocrine disorders 76 3 16 132 23 43 94 22 32 Neuropsychiatric disorders 453 4 55 405 22 172 404 50 145 Unipolar depressive disorders 3 0 0 1 0 1 10 0 8 Bipolar affective disorder 0 0 0 0 0 0 0 0 0 Schizophrenia 2 0 1 9 0 7 18 0 12 Epilepsy 11 1 5 53 8 38 78 28 37 Alcohol use disorders 23 0 15 47 0 36 18 0 13 Alzheimer and other dementias 277 0 2 101 1 9 5 3 3 2 0 0 2 0 0 Measles 424 220 204 218 2 0 202 1 0 Tetanus 163 105 58 95 8 2 52 5 1 Meningitis 340 181 159 107 54 19 113 32 14 Hepatitis B e 105 74 31 5 51 17 5 15 10 Hepatitis C e 54 36 18 2 24 10 2 8 8 Malaria 889 456 433 434 19 2 410 20 3 Tropical-cluster diseases 152 94 58 27 54 13 18 30 9 Trypanosomiasis 52 33 19 14 18 1 8 10 1 Chagas disease 11 7 5 0 3 4 0 1 3 Schistosomiasis 41 26 15 1 18 7 1 11 4 Leishmaniasis 47 29 18 12 15 2 9 Injuries 509 339 171 11 212 116 6 65 99 Unintentional 343 214 128 9 120 85 5 37 86 Road traffic accidents 116 83 32 4 61 18 2 19 11 Poisonings 31 21 10 0 19 2 0 8 2 Falls 76 38 37 0 10 28 0 2 35 Fires 9 5 3 0 2 2 0 1 2 Drownings 17 12 5 1 7 4 1 1 3 Other unintentional injuries 95 55 40 3 21 31 2 5 33 Intentional 166 124 42 1 92 31 1 28 13 Self-inflicted injuries 137 102 35 0 73 29 0 23 12 Violence 27 20 7 1 18 1 1 5 1 War and conflict Injuries 2 954 2 003 952 217 1 457 329 177 533 242 Unintentional 1 922 1 308 614 164 922 222 127 322 165 Road traffic accidents 674 504 170 33 398 72 19 114 37 Poisonings 173 121 52 5 100 17 5 34 12 Falls 199 129 70 6 75 48 4 22 44 Fires 62 33 29 5 20 8 5 15 8 Drownings 202 137 65 46 76 15 25 27 13 Other unintentional injuries 613 384 229 69 253 62 68 110 51 Intentional 819 594 225 13 488 93 10 156 21 15 7 2 10 3 2 4 1 Malaria 857 439 418 420 17 2 397 18 2 Tropical-cluster diseases 131 82 49 26 48 9 18 26 5 Trypanosomiasis 50 31 18 13 17 1 8 10 1 Chagas disease 0 0 0 0 0 0 0 0 0 Schistosomiasis 36 22 13 1 15 6 1 9 3 Leishmaniasis 45 28 17 12 15 2 9 257 161 96 9 86 66 14 44 38 Appendicitis 11 6 4 1 3 3 0 1 3 Diseases of the genitourinary system 353 205 148 11 80 114 10 56 82 Nephritis and nephrosis 298 165 133 10 67 88 9 49 74 Benign prostatic hypertrophy 22 22 0 0 6 16 0 0 0 Skin diseases 29 12 17 1 4 7 1 5 11 Musculoskeletal diseases 29 13 16 1 3 9 1 4 11 Rheumatoid arthritis 6 2 4 0 1 2 0 1 3 Osteoarthritis 2 1 1 0 0 1 0 0 1 Congenital abnormalities 253 127 126 118 8 1 119 7 1 Oral diseases 2 Injuries 2 318 1 469 849 244 1 018 207 205 473 171 Unintentional 1 640 997 643 216 622 158 183 317 143 Road traffic accidents 485 356 128 64 249 43 44 63 22 Poisonings 142 80 62 14 53 13 9 29 24 Falls 149 92 57 13 37 42 10 16 31 Fires 240 82 157 27 47 8 35 104 18 Drownings 169 114 55 38 68 9 25 23 8 Other unintentional injuries 455 272 183 60 169 43 61 82 40 Intentional 656 462 195 23 391 48 18 151 26 Self-inflicted injuries 294 179 115 4 154 22 3 99 13 Violence 255 198 58 11 168 19 9 40 9 War and conflict 100 79 20 7 66 6 5 11 4 a See Annex 34 19 15 11 7 1 11 3 1 Diphtheria 174 86 88 83 3 0 86 1 0 Measles 14 853 7 699 7 154 7 624 75 0 7 102 52 0 Tetanus 5 283 3 409 1 875 3 188 209 12 1 741 126 8 Meningitis 11 426 5 891 5 536 4 314 1 451 126 4 472 959 104 Hepatitis B e 2 068 1 437 630 192 1 125 121 184 376 70 Hepatitis C e 955 653 302 73 507 73 72 172 58 Malaria 33 976 17 340 16 636 16 595 719 26 15 851 754 31 Tropical-cluster diseases 12 113 8 264 3 850 3 075 5 050 138 1 582 2 159 109 Trypanosomiasis 1 673 1 041 631 515 518 9 320 305 5 Chagas disease 430 231 199 0 205 27 0 176 22 Schistosomiasis 1 707 1 021 686 503 467 51 360 292 34 Leishmaniasis 1 974 1 227 748 565 644 17 407 327 14 Lymphatic filariasis 5 941 4 521 1 420 1 459 3 045 17 463 935 22 Onchocerciasis 389 223 166 33 172 18 31 123 12 Leprosy 194 116 78 32 74 11 28 45 4 Dengue 670 336 334 320 15 1 316 17 1 Japanese encephalitis 681 330 351 279 49 2 300 49 2 Trachoma 1 334 338 997 2 233 102 4 647 346 Intestinal nematode infections 4 013 2 052 1 961 1 654 373 24 1 574 361 26 Ascariasis 1 851 943 908 942 2 0 906 1 0 Trichuriasis 1 012 525 487 525 0 0 487 0 0 Hookworm disease 1 092 551 541 160 369 23 157 359 25 Respiratory infections 97 786 51 266 46 520 39 303 7 819 4 144 36 890 5 702 3 929 Lower respiratory infections 94 511 49 542 44 969 37 967 7 523 4 052 35 633 5 487 3 850 Upper respiratory infections 1 787 952 835 601 262 89 574 183 78 Otitis media 1 488 772 716 735 34 3 682 32 2 Maternal conditions 38 0 0 0 0 0 0 0 0 0 Measles 3 2 2 1 0 0 2 0 0 Tetanus 4 2 1 2 0 0 1 0 0 Meningitis 108 56 52 29 22 5 28 18 6 Hepatitis B e 82 57 26 0 41 15 0 12 13 Hepatitis C e 153 98 56 0 72 26 0 26 29 Malaria 5 3 2 2 1 0 1 1 0 Tropical-cluster diseases 15 9 6 3 6 0 2 3 0 Trypanosomiasis 0 0 0 0 0 0 0 0 0 Chagas disease 2 1 1 0 1 0 0 1 0 Schistosomiasis 9 5 3 2 3 0 2 2 0 Leishmaniasis 4 2 2 1 1 0 1 1 0 Lymphatic filariasis 0 0 0 0 0 0 0 0 0 Onchocerciasis 0 0 0 0 0 0 0 0 0 Leprosy 0 0 0 0 0 0 0 0 0 Dengue 6 3 3 3 0 0 2 0 0 Japanese encephalitis 2 1 1 1 0 0 1 0 0 Trachoma 0 0 0 0 0 0 0 0 0 Intestinal nematode infections 24 12 12 7 5 1 6 5 1 Ascariasis 6 3 3 3 0 0 3 0 0 Trichuriasis 6 3 3 3 0 0 3 0 0 Hookworm disease 12 6 6 0 5 0 0 5 1 Respiratory infections 1 374 742 632 123 194 425 127 121 383 Lower respiratory infections 1 220 663 557 69 178 417 76 106 375 Upper respiratory infections 54 28 27 4 16 8 4 15 8 Otitis media 99 51 48 50 0 0 4 3 1 0 2 0 0 1 0 Diphtheria 11 6 5 5 0 0 4 1 0 Measles 1 238 654 584 581 73 0 535 50 0 Tetanus 576 385 192 319 59 7 171 16 4 Meningitis 2 412 1 259 1 153 896 329 34 847 271 34 Hepatitis B e 748 574 174 35 482 58 32 110 32 Hepatitis C e 325 246 78 12 207 28 11 47 19 Malaria 1 177 598 579 519 75 4 503 72 4 Tropical-cluster diseases 1 271 796 475 177 582 38 97 349 29 Trypanosomiasis 86 53 32 26 27 0 16 16 0 Chagas disease 423 228 195 0 202 26 0 174 22 Schistosomiasis 204 121 83 54 61 7 41 38 4 Leishmaniasis 110 73 37 27 45 1 17 20 1 Lymphatic filariasis 444 319 125 69 246 3 22 101 3 Onchocerciasis 4 2 2 0 2 0 1 1 0 Leprosy 34 20 14 4 11 5 4 9 1 Dengue 266 134 133 125 8 0 122 10 1 Japanese encephalitis 198 95 103 78 16 1 86 16 1 Trachoma 436 106 330 0 69 37 1 207 122 Intestinal nematode infections 1 293 661 632 442 204 14 420 196 16 Ascariasis 420 215 205 214 1 0 204 0 0 Trichuriasis 344 178 167 177 0 0 167 0 0 Hookworm disease 502 254 247 38 202 14 36 196 15 Respiratory infections 17 565 9 612 7 953 5 915 2 485 1 211 5 518 1 344 1 091 Lower respiratory infections 16 319 8 947 7 372 5 429 2 340 1 178 5 046 1 263 1 064 Upper respiratory infections 642 351 291 182 137 32 190 74 27 Otitis media 604 313 291 303 9 1 1 236 806 430 157 602 48 151 254 25 Hepatitis C e 476 308 168 61 228 19 61 98 9 Malaria 32 766 16 725 16 041 16 062 641 21 15 335 680 27 Tropical-cluster diseases 10 823 7 456 3 367 2 895 4 461 100 1 482 1 805 79 Trypanosomiasis 1 586 987 598 488 490 9 304 289 5 Chagas disease 3 2 2 0 1 0 0 1 0 Schistosomiasis 1 493 893 600 446 403 44 317 253 29 Leishmaniasis 1 861 1 152 709 537 599 16 389 306 13 Lymphatic filariasis 5 496 4 202 1 295 1 390 2 798 13 442 834 19 Onchocerciasis 383 220 163 33 170 18 30 121 12 Leprosy 160 96 64 28 62 6 25 36 3 Dengue 397 199 198 191 7 1 191 7 1 Japanese encephalitis 481 234 247 200 33 1 213 33 1 Trachoma 897 231 666 2 164 66 3 438 224 Intestinal nematode infections 2 694 1 378 1 316 1 205 164 9 1 147 160 10 Ascariasis 1 424 725 700 724 1 0 699 1 0 Trichuriasis 661 344 317 344 0 0 317 0 0 Hookworm disease 578 291 287 121 161 8 120 (11) . Country groupings and regions used in this report are defined in Tables C1 and C2 (Annex C). GBD 2004 estimates are also available for other regional groupings on the WHO web site. a Table C3 (Annex C) lists the GBD cause categories and their definitions in terms of the International classification of diseases, tenth revision (ICD-10). The GBD cause categories are grouped into three broad cause groups: Group I (communicable, maternal, perinatal and nutritional conditions), Group II (noncommunicable diseases) and Group III (injuries). General methods and data sources for the GBD estimates have been documented elsewhere (11) ; this part of the report documents the specific revisions for the 2004 estimates. Apart from the incorporation of new epidemiological data for specific causes, the GBD 2004 has incorporated: • more recent death registration data for many countries; • new African mortality data using verbal autopsy methods to assign cause of death; • improvements in methods used for the estimation of causes of child deaths in countries without good death registration data. For these reasons, and also because of revisions to the United Nations population estimates, the GBD estimates for 2004 are not directly comparable with the previous GBD 2002 estimates. Life tables for the 192 WHO Member States in 2004 were published in the World health report 2006 (25) , with data and methods described in the Statistical Annex Notes. Age-and sex-specific all-cause mortality rates from these life tables were further adjusted for revisions in estimates of human immunodeficiency virus (HIV) mortality and deaths due to conflict and natural disasters. These cause-specific revisions and updates are outlined in Section B5. Total deaths by age and sex were estimated for each Member State by applying these death rates to the estimated 2004 de facto resident populations prepared by the United Nations Population Division in its 2006 revision (13) . Total deaths in the neonatal period (the first four weeks of life) for 2004 were estimated country by country using methods documented elsewhere in collaboration with the Department of Making Pregnancy Safer (26) . For China, recorded age-specific death rates for ages over five years from the 2000 census were adjusted for estimated underreporting of 11.3% for males and 18.1% for females (27) , projected forward to 2004 assuming an annual rate of mortality decline of 1.5%, based on the estimated rates of mortality for the two intercensal periods, 1982-1990 and 1990-2000. The projection of child mortality rate takes into account other sources of data such as the Child Mortality Surveillance System (28) . For India, the all-cause mortality envelope was derived from a time series analysis of age-specific death rates from the sample registration system for years 1990-2002, after correction for underregistration (88% completeness) (29) . Child and adult mortality rates were then projected to 2004 using a regression model that gave more weight to recent years of observation. The child mortality rate was projected independently taking into account other sources such as the Demographic and Health Survey 2005/2006 (30). Death registration data containing usable information on cause-of-death distributions were available for 112 countries; the majority of these countries were in the high-income group, Latin America and the Caribbean, and Europe and Central Asia. The data sources and methods for estimation of mortality are outlined in Table C4 Cause-of-death statistics are reported to WHO on an annual basis by country, year, cause, age and sex. The number of countries reporting data using ICD-10 has continued to increase. For the GBD 2002 estimates, a total of 72 countries had complete data (defined as coverage of 85% and more), but only 35 countries were reporting data coded to the third or fourth character of ICD-10. For the GBD 2004 estimates, in the same group of 72 countries, there are now 55 countries reporting data to the third or fourth character of ICD-10. This increase of 20 countries reporting data at the detailed level of ICD-10 enhances the comparability of the 2004 estimates. However, 10 countries with complete data still report to WHO using the ninth revision of ICD (ICD-9). GBD correction algorithms (11) were also applied to resolve problems of miscoding for cardiovascular diseases (mainly involving redistribution of deaths coded to heart failure or ill-defined heart disease), cancer (involving redistribution of deaths coded to secondary sites or ill-defined primary sites) and injuries (involving redistribution of deaths coded as due to events of undetermined intent). Coding of natural causes of death for neonates varies a great deal among countries. Some countries code these deaths to the 'P chapter' (i.e. codes that refer to conditions originating in the perinatal period) while others use a combination of P codes and other codes as well. In some instances the age of death is not always taken into account. Some conditions, such as septicaemia and pneumonia, have specific codes within P00-P96 which should be used for neonates (0-27 days). For countries with vital registration data, we have recoded all the deaths aged 0-27 days from natural causes that were initially coded outside the 'P chapter' to codes in the 'P chapter' whenever possible. In a number of countries, neonatal septicaemia (P36) is frequently assigned to A40 and A41 (septicaemia). In this case we have recoded them back to P36, thus identifying more deaths due to causes originating in the perinatal period. Since the GBD 2002 analyses, the number of countries with complete data (coverage of 85% or more) has remained stable at 76. Vital registration data were used to estimate deaths by cause for 78 countries, including two very small countries where coverage was lower than 85%. Where the latest available year was earlier than 2004, death registration data from 1980 up to the latest available year were analysed as a basis for projecting recent trends for specific causes, and these trend estimates were used to project the cause distribution for 2004 from the latest available year. When estimating cause-of-death distributions for very small countries, an average of the three last years of data were used to minimize stochastic variation. Adjustments for deaths due to HIV, drug use disorders, war and natural disasters were based on other sources of information as described in Section B6 below. For 34 countries with less than 85% coverage, cause-of-death modelling (CodMod) was used to adjust the proportions of deaths occurring in Groups I, II and III by age and sex as described elsewhere (11) . The regional distribution of the 112 countries for which complete or incomplete vital registration data were used is shown in Table B1 . To estimate deaths by cause for the remaining 78 countries without usable death registration data, CodMod was applied at country level for estimating the proportion of deaths in Groups I, II and III by age and sex, based on all-cause mortality levels (excluding HIV, war and natural disasters), gross national income per capita and region. The statistical model for cause-of-death composition was estimated using a substantially larger data set (1613 country-years) than that used for the 1990 GBD study (11, 31) . Detailed proportional cause distributions within Groups I, II and III were based on death registration data from within each region (see Table C4 , Annex C, for more details). Specific causes were further adjusted on World Health Organization Annex B the basis of epidemiological evidence from registries, verbal autopsy studies, disease surveillance systems and analyses from WHO technical programmes. The evidence used for the revision of specific causes is described in Section B5. For all regions except the WHO African Region, the choice of death registration data for the estimation of within-group cause distributions was consistent with that used for the GBD 2002. For countries in the WHO African Region, the GBD 2002 used cause-of-death distributions based on 1996 death registration data from rural provinces of South Africa, and grouped countries into two mortality strata labelled D and E (6) . For the 2004 update, the regional pattern was based on a greater range of information on cause-of-death distributions in Africa. These included the South African 2004 vital registration data, the Zimbabwe National Burden of Disease Study 1997 (32), INDEPTH verbal autopsy data from seven sites in Africa for 1999-2002 (33) , data from Antananarivo in Madagascar for 1976-1995 (34) , and Mozambique Maputo Central Hospital Mortuary data for 1993-2004 (35) . For Groups I, II and III, the proportional distributions of total deaths (excluding HIV and war) for urban populations were based on averages of the Madagascar urban death registration data, South African death registration data for 2004, the Zimbabwean 1997 estimates and CodMod predictions for 2004. As all the African INDEPTH sites were rural, the INDEPTH data were taken as representative of rural populations. The Group II fraction of deaths in the African INDEPTH data flattened out above age 50 years, and was lower than in all other African data sources. We thus used an average of INDEPTH and South African data for rural provinces, giving less weight to the INDEPTH data with increasing age for rural populations. For countries in the D mortality stratum, 44% of the population were assumed to be urban, and 29% for the E mortality stratum. Proportional cause distributions for Groups I, II and III for individual African countries were adjusted away from regional average distributions using CodMod together with country-specific inputs on mortality levels and income per capita. Figure B1 compares the final resulting African regional cause distributions for Groups II and III for 2004 with those for the GBD 2002. China Cause-specific mortality data for China continued to be available from two sources -the sample vital registration system (VR) monitored by the Ministry of Health, and the Disease Surveillance Points system (DSP) monitored by the Chinese Center for Disease Control and Prevention. The following table gives a brief summary of the characteristics of the most recently available data from these systems. Both data sets were assessed for suitability in estimating 2004 cause-specific mortality for China at the national level. According to Chinese authorities, the sampling distribution of sites in the DSP was more nationally representative than the VR (36), but they advised that DSP data for 2004 were not yet finalized, with further adjustments to cause attributions pending. We therefore based the update of the broad Tehran Province, which is the most populous province (population 12 million), was the only province not covered by the death registration system (38) . Coverage has substantially improved compared to earlier years. In 1999, the system was capturing deaths in only four provinces with coverage of 5% of all deaths in the entire country. In 2001, the system further expanded by recording deaths in 18 provinces and one district with coverage of nearly 40% of all deaths in the country. The 2004 data were coded to a condensed list of 318 cause categories, using the ICD-10 classification system. As coverage was partial, CodMod was used to predict the proportionate mortality distributions for Groups I, II and III, and specific cause mortality distributions adjusted within these groups. Supplementary information from WHO technical programmes and UNAIDS was also used in estimating specific causes of death. Cause-specific mortality for Mexico was estimated using death registration data for 2004, assessed to be nearly 100% complete. As well as the standard redistributions for ill-defined cause codes, further corrections for miscertification of deaths due to diabetes and cardiovascular diseases were carried out based on in-depth analyses undertaken by the Harvard Initiative for Global Health (39) . The completeness of the 2004 death registration data for South Africa was assessed to lie in the range of 75-89%. Approximately 13 000 deaths were reported to be due to HIV/AIDS, although UNAIDS has estimated that HIV/AIDS was responsible for 290 000 deaths in 2003 and 320 000 deaths in 2005 (40) . infectious origin, respiratory TB and herpes zoster, causes which the national statistical office had also found to be often associated with HIV/AIDS (41) . In addition, deaths classified as ill-defined showed spurious peaks in the same adult age groups as for deaths from HIV/AIDS. AIDS-defining diseases such as Kaposi sarcoma were also examined. Averaged age distributions for cause-specific mortality rates for 1993 and 1996 were used to remove the embedded misdiagnosed HIV/AIDS deaths in the 2004 data, in order to obtain HIV/AIDS-free sexage-cause distribution patterns. A total of 224 000 deaths were reassigned from other causes to HIV/ AIDS, resulting in an estimated national total for HIV/AIDS deaths of 292 000 in 2004, close to the UNAIDS estimate of 295 000, derived independently from prevalence data. The HIV/AIDS-free sex-age-cause distribution patterns thus obtained were then proportionately adjusted to the WHO estimated number of deaths by sex and age for South Africa in 2004. Supplementary information from WHO technical programmes for some specific diseases and causes was also used to adjust final estimates by cause. Nearly 70% of all injury deaths in the 2004 death registration data were classified as due to events of undetermined intent, and less than 10% of all injury deaths were classified as caused by homicides. We used separate estimates of the external cause distribution for injury deaths based on data from the National Injury Mortality Surveillance System prepared for the revised South African Burden of Disease study (42) . Death registration data were available for the year 2002, with an estimated coverage of about 85%. However, the proportion of ill-defined conditions was nearly 50%, since many deaths in Thailand occur at home, and the cause of death is reported by lay people. In order to improve the usability of the death registration data, the Thai Ministry of Health conducted a re-test survey on a sample of about 33 000 deaths, using verbal autopsy methods, to ascertain the true cause of death (43) . This included a sample of 12 000 deaths with ill-defined causes. The reallocation algorithm for ill-defined causes from the verbal autopsy study was used to correct the high proportion of ill-defined deaths from the death registration data. Deaths from septicaemia and pneumonia (approximately 27 000 in total in the 2002 data) were also reallocated based on the outcomes of the verbal autopsy study. The resultant cause-specific proportionate mortality was inflated to the national mortality envelope derived from the life table analysis. Supplementary information from WHO technical programmes and UNAIDS was also used in estimating specific causes of death. Death registration data for 2003 were only available for urban areas of Turkey, with an estimated national coverage of 45%. Causes of death were coded using the condensed list of the ICD eighth revision. Instead of using these data, we used a detailed analysis of causes of death for the year 2000, from a national burden of disease study conducted by the national authorities in Turkey (44) . Proportional distributions for Groups I, II and III in 2004 were derived using CodMod. Supplementary information from WHO technical programmes and UNAIDS was also used in estimating specific causes of death. In (46, 47) and for neonatal deaths (48, 49) were rerun with updated inputs for the year 2004. Outputs were adjusted and mapped to GBD cause categories as described below. Together with causespecific inputs from WHO technical programmes and UNAIDS, the resulting cause-specific inputs were adjusted country by country to fit neonatal, infant and child under-five death envelopes. For 76 countries with death registration data covering 85% or more of deaths, deaths by cause were estimated using the most recent death registration data, after adjusting for deaths coded to ill-defined World Health Organization Annex B categories (see above). For the other 116 countries, available mortality data together with CodMod (see above) were used to estimate the broad cause-ofdeath patterns (Groups I, II and III) for infant deaths (under one year) and child deaths at ages one to four years. Total estimated deaths in the neonatal period were distributed to specific causes using a model developed by the Child Health Epidemiology Reference Group (CHERG) for countries without usable death registration data (48, 49) . The CHERG neonatal working group undertook an extensive exercise to derive global estimates for seven programme-relevant causes of neonatal death, including preterm birth, asphyxia, severe infection, neonatal tetanus, diarrhoea, congenital malformation and other causes, based on 56 studies of neonatal deaths from 29 countries that met inclusion criteria. Multinomial models developed to estimate simultaneously the distribution of these seven causes of death by country were rerun for 2004 using updated input data for covariates, and then mapped to GBD cause categories. The GBD 2002 residual cause category "Other conditions" for the cause group "Conditions arising in the perinatal period" includes a number of infectious disease categories as well as some non-infection causes. For the GBD 2004, this cause category has been renamed "Neonatal infections and other conditions" (arising in the perinatal period). Under ICD rules, some deaths due to neonatal infections should be coded to specific infectious causes (e.g. meningitis, syphilis) and others to codes within the perinatal conditions chapter (e.g. neonatal sepsis). Clinically, it may be difficult to distinguish pneumonia, sepsis and other causes of neonatal infection. The cause category "Severe infection" was mapped to GBD cause categories based on an analysis of available death registration data and expert advice on the evidence from neonatal mortality studies, as summarized in Table B3 . The GBD cause category "Neonatal infections and other conditions" also includes other non-infectious causes arising in the perinatal period (such as deaths resulting from maternal factors such as maternal hypertension and surgery, umbilical and other neonatal haemorrhage, haemolytic disease, neonatal jaundice, hypothermia). Mapping of the CHERG "Other neonatal category" to GBD categories resulted in 21% of the "Neonatal infections and other conditions" deaths being due to non-infectious causes, reasonably consistent with the low end of the proportions in death registration data. The GBD cause category "Prematurity and low birth weight" (formerly labelled "Low birth weight" in the GBD 2001 and 2002) includes the CHERG neonatal category "Preterm birth" as well as deaths of infants born at term but small for gestational age. Based on expert advice, together with an analysis of available death registration data, term but small-forgestational age deaths were assumed to be 0.4% of total neonatal deaths, and this fraction was added to the CHERG estimate for deaths of preterm infants. The CHERG neonatal category "Congenital abnormalities" includes the GBD cause "Congenital anomalies" plus other Group II causes which comprise genetic conditions (e.g. inherited neurological and metabolic conditions and miscoded congenital heart conditions). Non-genetic Group II conditions include cancers, endocrine and skin conditions. In mapping the CHERG neonatal congenital category to GBD cause groups, it was assumed that 90% of the deaths in the CHERG group were in the GBD congenital cause group, based on an analysis of previous mortality estimates for GBD Group II conditions in infants. Outputs from the CHERG neonatal model were used for all causes except neonatal tetanus. Outputs for neonatal tetanus deaths were compared country by country with estimates prepared by the WHO Department of Immunization, Vaccines and Biologicals (see Section B5). Final estimates from these models were reasonably consistent and the estimates prepared by this department were used for the GBD 2004. causes exceeded the CodMod-derived Group I fraction. For these 63 countries, the estimated cause-specific child deaths exceeded the child death envelopes by 9% or 937 000 deaths. For these countries, deaths for all Group I causes except HIV were adjusted pro-rata to match the CodMod-derived Group I fraction. HIV/AIDS deaths were excluded, as they have been treated as outside the mortality envelopes for countries with high HIV death rates. Adjustment of estimated cause-specific deaths to fit the under-five envelopes differentially affected specific causes. For example, the total estimated malaria deaths before adjustment were 932 000 and after adjustment 812 000, a 12% reduction. Years lost due to disability (YLD) estimates were revised for a number of causes where updated estimates for incidence or prevalence for national or regional populations were available (see Section B5 for details Age-sex-specific YLD rates for leprosy were assumed to have declined by 4% from 2002 to 2004. For disease and injury causes where mortality was responsible for a significant proportion of the total burden, regional estimates of YLD/YLL ratios by age and sex for 2002 were used together with 2004 YLL to estimate 2004 YLD. The causes for which this ratio method were used included meningitis, hepatitis B and C, lower respiratory infections, endocrine disorders, rheumatic heart disease, hypertensive heart disease, inflammatory heart disease, chronic obstructive pulmonary disease, peptic ulcer disease, cirrhosis of the liver, and nephritis/nephrosis. For certain causes, regional age-sex-specific YLD/ YLL ratios for the overall cause group were used for estimating YLD for 2004. These causes included maternal haemorrhage and sepsis, hypertensive disorders of pregnancy, causes arising in the perinatal period, congenital malformations excluding cleft lip and palate and Down syndrome, and all the injury causes apart from war and conflict. Estimates of incidence, average case duration for treated and untreated cases, and deaths due to TB (excluding HIV-infected individuals) for all countries in 2004 formed the basis of estimates of TB prevalence in 2004 (50) . For countries with VR data for TB deaths, incidence estimates have been revised to be consistent with estimated deaths. Country-specific estimates of HIV/AIDS mortality are revised annually by UNAIDS and WHO to take into account new data and improved methods (12) . For the most recent round of estimates released in November 2007, advances in methodology applied to an increased range of country data have resulted in substantial changes in estimates of incidence, prevalence and mortality for HIV. The estimated For the 2004 incremental update, the disease model for estimating YLD for HIV was updated to include three sequelae: HIV cases (not progressed to AIDS), AIDS cases not receiving antiretroviral therapy and AIDS cases receiving antiretroviral therapy. Average durations for these three sequelae were revised in line with the most recent UNAIDS estimates of survival with and without antiretroviral therapy (20) . The disability weight for AIDS cases receiving antiretroviral therapy was set to be the same as that for HIV cases, apart from the terminal stage assumed to be have the same duration and disability weight as AIDS cases not receiving antiretroviral therapy. For countries with usable death registration data, deaths due to diarrhoeal diseases were estimated directly from these data. For other countries, a regression model was used to estimate proportional mortality from diarrhoea for children aged under five years (51) . The final regression model for the GBD 2004 included the logit of the proportional mortality from diarrhoeal diseases in children aged 0-4 years as a dependent variable and total underfive mortality rates, mid-year of study and WHO regions as explanatory variables. The regression data were drawn from more than 60 community-based studies since 1980 with study durations of multiples of 12 months. This model was validated and supplemented with vital statistics from developing countries where coverage was high. Estimates of child deaths due to diarrhoeal disease for 2004 from this model were assumed to include the neonatal diarrhoea deaths estimated using the CHERG neonatal model. Pertussis Pertussis cases and deaths were based on a natural history model using vaccine coverage and age-specific case fatality rates from community-based studies (52) . Updated estimates for 2004 were prepared by the WHO Department of Immunization, Vaccines and Biologicals using WHO-United Nations Children's Fund (UNICEF) estimates for vaccine coverage in 2004, interpolated for missing data (53, 54) . Incidence estimates for polio were based on reported cases in 2004 (55) with adjustments for underreporting. Mortality estimates were derived from incidence estimates assuming a 10% case fatality rate. Updated incidence estimates for 2004 were prepared by the WHO Department of Immunization, Vaccines and Biologicals using WHO-UNICEF estimates for vaccine coverage in 2004, interpolated for missing data (53) . A case fatality rate of 10% was assumed for diphtheria in countries without high vital events coverage (56) . Measles incidence and mortality for 2004 were estimated using a revised natural history model whose Snow et al. (69) that the age distribution in Africa was 65% for ages under five years, 19% for ages 5-14 years, and 16% for ages over 15 years. Recent verbal autopsy data from the INDEPTH network for malaria (excluding fever of unknown origin) also gave around 62% of deaths for children aged under five years, but only 7% for those aged 5-14 years (33) . For the 2004 estimates of malaria mortality, the relationship between malaria mortality rates at ages 5-19 years and 20-39 years and child malaria mortality rates for ages 0-4 years was estimated using a model of age-specific malaria mortality by transmission intensity (70) . The revised death rates for malaria at ages five years and over result in an estimated 10% of malaria deaths at ages five years and over in the African region, and 13% globally. For the lower transmission areas outside Africa, the death rates are lower for children and higher for adults. At very low transmission, the death rates are similar across all ages, but slightly higher for younger children and older adults. The resulting global age distribution of malaria deaths is very similar to that in the original GBD study for 1990 (3) . The GBD 2002 disease model for human African trypanosomiasis (HAT) did not distinguish the two forms of HAT (infection with Trypanosoma brucei gambiense or Trypanosoma brucei rhodesiense) and assumed an average five-year duration with close to 100% case fatality. For GBD 2004 calculations, separate estimates were made for incidence of cases of infection with T.b. gambiense and T.b. rhodesiense. An average duration of five years was assumed for T.b. gambiense and one year for T.b. rhodesiense, so giving a lower prevalence than 2002 estimates. For the mortality estimates, 85% of cases were assumed to be untreated with 100% case fatality, and 15% treated with 5% case fatality. WHO programme data on populations at risk and levels of endemicity in African countries, the form of HAT present in each country, and numbers of reported cases of disease were used to develop estimates of the incidence of the two forms of HAT for 2004. Completeness of reporting of cases was assumed to be around 33% for T.b. gambiense and 5% for T.b. rhodesiense (71) . The resulting total incidence of HAT in 2004 is estimated at 60 300, consistent with recent WHO estimates of global incidence in the range 50 000-70 000 (72). Updated Chagas disease incidence and prevalence estimates for 2005 were provided by the Pan American Health Organization (73) . These estimates were supplemented with and validated against vital statistics from Latin American countries where coverage was high. Over 9000 deaths due to Chagas disease were recorded in death registration data for the region and the implied annual case fatality rate for prevalent cases in countries with good registration data was 0.18%. For a number of countries where the number of deaths in the death registration data for 2004 was much lower than that implied by the prevalence and case fatality estimates, a conservative estimate of numbers of Chagas disease deaths was based on an average of the number estimated from the prevalence estimates and the death The prevalences of onchocerciasis-caused low vision and blindness were updated using country-specific estimates provided by the WHO Programme for Prevention of Blindness and Visual Impairment. The GBD 2002 estimates for dengue were based on country-level estimates of dengue mortality in the year 2000 from the WHO Department of Control of Neglected Tropical Diseases and on an earlier review of nearly 300 population-based incidence studies (79) . For the GBD 2004, country-specific mortality estimates were revised, drawing on a review of dengue mortality studies and on deaths reported by DengueNet (80) . The incidence of dengue haemorrhagic fever or dengue shock syndrome was back-estimated from the mortality estimates assuming case fatality rates of 0.025 in China and the Americas, and 0.04 in other regions. The incidence of dengue fever was then estimated assuming that 6% of symptomatic cases were dengue haemorrhagic fever and 94% were dengue fever (81) .The disease model was revised to include two sequelae: 1. Dengue fever with an average duration of 5.5 days, and an average disability weight of 0.2 (from the GBD 1990 weight for dengue haemorrhagic fever). 2. Dengue haemorrhagic fever/dengue shock syndrome with an average duration of 11 days and an average disability weight of 0.5 (adjusted so that disability weight multiplied by duration is the same as for GBD 1990 (0.2 multiplied by 1 month)). The prevalences of trachoma-caused low vision and blindness were updated using country-specific estimates provided by the WHO Programme for Prevention of Blindness and Visual Impairment (82). A single-cause proportional model was used to estimate child mortality due to acute lower respiratory infections (mainly pneumonia) for countries without usable death registration data. Communitybased studies with durations of one year or longer, and published since 1980, were used to estimate the proportional mortality from acute respiratory infections in children aged 0-4 years in developing countries (83 Abortion-related mortality occurs mainly as a result of unsafe induced abortion. GBD estimates for 2004 were updated using WHO estimates of incidence and mortality rates due to unsafe induced abortion in 2003 based on published and unpublished reports for more than 130 countries together with other information on legal and social contexts (85, 86) . The prevalences of underweight, stunting and wasting were based on analysis of 388 nationally representative studies for 139 countries from the WHO Global Database on Child Growth and Malnutrition a . These were used to estimate prevalence of child stunting and wasting for each country in the world according to the new WHO Child Growth Standards (87) (88) (89) (90) . In 2004, an estimated 32% of children aged under five years were stunted (height-for-age more than two standard deviations below the standard) and 20% were wasted (weight-for-age more than two standard deviations below the standard). The WHO Department of Nutrition for Health and Development has prepared global, regional and national estimates of anaemia prevalence in pre-school-aged children, pregnant women and non-pregnant women, using data collected for the Vitamin and Mineral Nutrition Information System (VMNIS) b . These estimates were based on the most recent national and subnational surveys measuring blood haemoglobin concentration carried out in the years 1993-2005 (91, 92) . These surveys covered around two thirds of the world's population and were used to calculate regional age-and sexspecific prevalence estimates for mild, moderate and severe anaemia. According to these estimates, 42% of pregnant women and 47% of preschool children worldwide have anaemia. As for the GBD 2002 estimates, it was assumed that 60% of anaemia was due to iron deficiency in non-malaria areas and 50% in malaria areas. Previous GBD estimates for the site-specific distribution of cancer mortality for Member States without good vital registration data, and for site-specific incidence of cancers, were based on a model for relative interval survival applied to country-level estimates of cancer incidence distributions by site published by the International Agency for Research on Cancer (IARC) in its Globocan 2000 database (93) (94) (95) . For the GBD 2004, regional site-specific survival probabilities were updated based on regional trends in gross domestic product per capita (in international dollars) and applied to revised site-specific incidence distributions for 2002 published by IARC in Globocan 2002 (96) . Diabetes prevalence estimates were updated to take account of a number of recently published population surveys that used oral glucose tolerance tests and WHO criteria to measure diabetes prevalence (97) (98) (99) (100) (101) (102) (103) (104) (105) (106) (107) (108) (109) (110) . For countries and regions for which more recent prevalence studies were not available, previous diabetes prevalence estimates (111) were projected from those for 2000-2004 using projected trends in population body mass index distributions (112) together with recent estimates of the relative risk of diabetes associated with each unit increase in body mass index from a large meta-analysis of the Asia Pacific Cohort Studies (113) . These revisions increased the estimated prevalence of diabetes mellitus from 191 million (if GBD 2002 prevalence rates are applied to the 2004 population) to 220 million in 2004. Point prevalence estimates for episodes of unipolar major depression were derived for the GBD a http://www.who.int/nutgrowthdb/ b http://www.who.int/vmnis (114) . Variations in the prevalence of unipolar depressive disorders in some European countries, Australia, New Zealand and Japan were estimated directly from relevant population studies (115) . In the GBD 2002, all comorbid cases of depression and alcohol use disorder were included in the burden of depressive disorders. For the GBD 2004, one half of these comorbid cases were included in depressive disorders and the other half in alcohol use disorders. Age-sex-specific prevalence rates from the GBD 2002 were adjusted downwards by one half of the estimated prevalence rate of comorbid cases of depression and alcohol use disorder (see following section) and then applied to the 2004 population estimates to calculate the prevalence and burden of depressive disorders. Previous global estimates for alcohol dependence and harmful use (alcohol use disorders) were based on a range of assessment instruments with varying quality including screening instruments such as the CAGE or the AUDIT, which may not be completely comparable across populations (116) . A new review was carried out for the GBD 2004 update using only studies conducted after 1990, which used ICD-10, DSM-IIIR or DSM-IV criteria for alcohol dependence and one of the following diagnostic instruments: the Composite International Diagnostic Interview (CIDI), the Schedules for Clinical Assessment in Neuropsychiatry (SCAN) or the Alcohol Use Disorder and Associated Disabilities Interview Schedule-Alcohol/Drug-Revised (AUDADIS-ADR) (117) . Population estimates of the point prevalence of alcohol use disorders in the age range 18-64 years were obtained from 37 studies (118) . Published data on alcohol production, trade and sales, and adjusted for estimates of illegally produced alcohol were used to estimate country averages for volumes of alcohol consumed. These preliminary estimates were then further adjusted on the basis of the survey data on alcohol consumption to estimate prevalence of alcohol use disorders for countries where recent population-based survey data were not available. Estimated regional prevalence rates for alcohol dependence by age and sex were then adjusted upwards by 39% for men and 61% for women to account for the additional prevalence of harmful use of alcohol. To correct for comorbidity with depression, prevalence was reduced by a proportional subtraction of 9% (males) and 21% (females) for the WHO regions of Europe and the Americas. For all other regions, prevalence was reduced by 5% (males) and 10% (females) to correct for comorbidity with depression. Incidence rates and average durations for alcohol use disorders were estimated from prevalence, relative risk of mortality and remission rates using DISMOD II (118) . Instantaneous remission rates of 0.175 were assumed for ages 15 years and over; the relative risk of mortality averaged 1.8 for males and 3.8 for females. The original GBD 1990 disability weight of 0.18 for alcohol dependence was applied to both alcohol dependence and harmful use in the GBD 2000-2002. Recent analyses of disability associated with harmful use of alcohol from the Australian Burden of Disease Study (119) and of health state valuations collected in the WHO Multi-country Survey Study (MCSS) (120) suggest that the average disability weight for harmful use of alcohol is much lower than for alcohol dependence. Weights derived from the multicountry study were used for the harmful use proportion of alcohol use disorders resulting in average disability weights for alcohol use disorders of 0.137 (males) and 0.132 (females) in people aged under 30 years, and 0.134 (males) and 0.122 (females) for those aged 30 years and over. Applying these weights, the overall YLD for alcohol use disorders for 2004 are 22.0 million compared to 19.1 million if GBD 2000 assumptions and estimates were maintained for the 2004 population. However, the revised prevalence estimates have resulted in increases in the estimated burden for China, India and countries of the former Soviet Union, and decreases for high-income countries, Latin America and Africa. Estimating mortality directly attributable to illicit drug use, such as overdose death, is difficult because of variations in the quality and quantity of mortality data. For some regions where there is known to be a Annex B substantial prevalence of illicit drug dependence, no deaths are recorded in available data sources as being due to drug dependence. As a result, it is necessary to make indirect estimates, involving estimates of the prevalence of illicit drug use and case fatality rates, and adjustments to death registration data (121, 122) . For the GBD 2004, estimates of deaths due to drug use disorders for 2002 were updated using regional trends in the use of illicit opiate drugs reported by the United Nations Office on Drugs and Crime (123). Estimates of YLD for mental retardation caused by environmental lead exposure prepared for the GBD 2000 were updated to 2004 using projected trends in the incidence of lead-caused mental retardation (124, 125) . Prevalence rates for blindness and low vision due to all causes were updated for the GBD 2002 based on data from available population-based surveys since 1980. Low vision and blindness were defined in terms of measured visual acuity in the better eye with best possible correction (126, 127) . These revisions were carried out before the regional distributions for causes of vision loss were available. For the GBD 2004, prevalence estimates for causes with low vision and blindness sequelae were revised to take account of the regional distributions for causes of blindness published by Resnikoff et al. (127) . Relevant causes included gonorrhoea, onchocerciasis, trachoma, vitamin A deficiency, injury, diabetic retinopathy, glaucoma, cataract, and macular degeneration and other age-related causes of vision loss not correctable by provision of glasses or contact lenses. The GBD 2002 cause category "Age-related vision disorders" was renamed "Macular degeneration and other" (age-related causes of vision loss). The GBD 2002 did not include refractive errors (myopia, hyperopia and astigmatism) correctable with spectacles or other refractive corrections. The definition of vision impairment based on "best possible correction" excludes these impairments in people who do not normally use spectacles or other refractive corrections. For the GBD 2004, an additional cause "refractive errors" was added to the sense organ disorders cause group. Prevalence estimates were based on recent WHO estimates of the regional prevalences of uncorrected refractive errors for all ages over five years (128) . The original GBD 1990 weight for low vision (visual acuity less than 6/18 and greater than 3/60) was quite high at about 0.24, and low vision was only included as a sequela for a few infectious causes, but not for glaucoma or cataracts. For the GBD 2004, the disability weight for low vision has been revised to 0.17 based on the Netherlands disability weights study (129) . The disability weight for blindness has been maintained at 0.60 for blindness with best possible correction, but the Netherlands study weight of 0.43 used for blindness due to uncorrected refractive errors. The GBD 2000-2002 study developed a model for ischaemic heart disease YLD which included three sequelae: acute myocardial infarction (AMI), angina pectoris and congestive heart failure following AMI (130) . The GBD 1990 study estimated a regression relationship between the prevalence of AMI survivors and the prevalence of angina pectoris (both preand post-AMI) using data from developed country studies, and then applied this regression relationship in other developing regions. For the GBD 2000-2002 revisions, the same methodology was followed, but the regression relationships were re-estimated using data from the US Burden of Disease Study for 1996 (131) . Two recent national burden of disease studies have made estimates of the prevalence of angina pectoris for Australia in 2003, a developed country population (119), and for Thailand in 2004, a developing country population (132) . These were used to re-estimate the regression equations, in order to estimate angina pectoris prevalence for each region from updated estimates of the prevalence of 24-hour AMI survivors for 2004. These revisions resulted in an increase in the estimated global prevalence of angina pectoris from 25 million in 2002 to 54 million in 2004. DISMOD II was used to estimate the incidence and average duration of angina pectoris for the calculation of YLD from estimated prevalence rates, relative risks of mortality and remission rates (due to revascularization procedures). Mortality relative risks were revised for high-income countries based on a study of mortality by cause following first-ever AMI in the Danish MONICA study (133) . Based on the ratio of the regional 28-day case fatality rates, it was assumed that the excess mortality risk was 40% higher in B and C strata, and 50% higher in D and E strata. These letters denote mortality strata used to group countries in the GBD 2002 (6) . Disability weights for angina pectoris were also revised based on the revisions in the Australian and Thai studies (119, 132). The revised disability weights ranged from 0.104 in high-income countries to 0.141 in the WHO African Region, compared with a previous range of 0.108-0.207. The proportion of incident AMI cases who go on to develop congestive heart failure was assumed to be 0.17 in A strata, 0.20 in B and C strata, 0.20 in D strata other than countries in the African Region, and 0.25 in D and E strata in the African Region. Case fatality rates for A strata were based on those estimated in the United States Burden of Disease and Injury Study (131) . Case fatality rates were assumed to be 10% higher in B and C strata, and 20% higher in D and E strata. The GBD 2000-2002 study developed a model for stroke based on available population data on case fatality within 28 days for incident cases of first-ever stroke and on long-term survival in cases surviving this initial period in which the risk of mortality is highest (134) . For the GBD 2004, estimated prevalence rates for stroke survivors from this model were compared with results from available population studies. GBD estimates for stroke survivor prevalence rates were generally around 10-30% higher than prevalence rates reported in available studies from developing countries, but most of the studies dated from the 1980s or 1990s, and may not fully identify survivors of mild strokes without noticeable neurological problems. Two recent national burden of disease studies have made estimates of the prevalence of stroke survivors for Australia in 2003, a developed country population (119), and for Thailand in 2004, a developing country population (132) . The Australian stroke estimates were based on detailed analysis of linked databases for Western Australia to identify incidence of first-ever stroke, and mortality for cases. Provisional Thai stroke prevalence estimates were based on data from the Third National Health Examination Survey 2004 (135) . Data from these studies were used to recalibrate the long-term case fatality rates for first-ever stroke survivors across all regions for the GBD 2004 estimates, resulting in a reduction in the estimated prevalence of stroke survivors from 50 million to 30 million, and a 30% reduction in YLD for cerebrovascular disease. Regional prevalence estimates for edentulism (loss of all natural teeth) were updated using data from the World Health Surveys for 60 countries in 2002-2004 (136) , together with a review of other published prevalence estimates (137) . The GBD 2004 includes estimates of deaths occurring in 2004 for which the underlying cause (following ICD conventions) was an injury due to war, civil insurrection or organized conflict, whether or not that injury occurred during the time of war or after cessation of hostilities. The GBD estimates include injury deaths resulting from all organized conflicts, including organized terrorist groups, whether or not a national government was involved. They do not include deaths from other causes (such as starvation, infectious disease epidemics, lack of medical intervention for chronic diseases), which may be counterfactually attributable to war or civil conflict. Country-specific estimates of war and conflict deaths were updated to 2004 using information on conflict intensity, time trends and mortality obtained from a variety of published and unpublished war mortality databases (138) (139) (140) (141) (142) . Murray et al. (143) have summarized the issues with estimation of war deaths, and emphasized the very considerable uncertainty in the GBD estimates for conflict deaths. Additional information from epidemiological Annex B studies and surveys was also used for certain specific conflicts in the Democratic Republic of the Congo (144), Iraq (145) and the Sudan (146) (147) (148) (149) . Deaths due to landmines and unexploded ordinance were estimated separately by country (150, 151) . Deaths due to major natural disasters in 2004 are included in the external cause of injury category "Other unintentional injuries". Such deaths will not be included in mortality data sources and estimates except for countries with death registration data B6. Prevalence of long-term disability GBD 2004 updates were used to estimate the prevalence of people with long-term disability by age, sex and region for two severity thresholds of disability. The GBD links loss of full health to disease and injury causes through the concepts of cases and sequelae. For incident cases of a given disease or injury in the population, there will be a distribution of current and future health states (conceptualized in terms of functioning capacity in a set of health domains) in the population, aggregated across individuals. The term disability was used in the naming of the DALY to stress the vision of health that goes beyond the absence of disease and the emphasis on difficulties in functioning. Although the term disability has a number of different meanings and connotations and is not seen by some as a synonym or proxy for "loss of health", the GBD uses "disability" to refer to a subconstruct of "diminution of health", where health is conceptualized in terms of functioning capacity in a set of health domains. The GBD 1990 study established severity weights for 22 sample "indicator conditions" using an explicit "person trade-off " protocol in a formal exercise involving health workers from all regions of the world. These weights were then grouped into seven classes where class I has a weight between 0 and 0.02 and class VII a weight between 0.7 and 1 (see Table 8 ). To generate disability weights for the remainder of the approximately 500 disabling sequelae in the study, participants in the study were asked to estimate distributions across the seven classes for each sequela. Distributions across disability classes were estimated separately for treated and untreated cases where relevant, and distributions could also vary by age group and sex. These severity distributions for GBD sequelae, together with updated severity distributions for those sequelae for which disability weights have been revised in the GBD 2004, were used to estimate the prevalence of long-term disability by severity class in 2004. Results are presented here for the prevalence of "severe" disability, defined as severity classes VI and VII (the equivalent of having blindness, Down syndrome, quadriplegia, severe depression or active psychosis) and for "moderate and severe" disability, defined as severity classes III and greater (severity class III is the equivalent of having angina, arthritis, low vision or alcohol dependence). The disability prevalence estimates also excluded conditions with average durations of less than six months. The GBD prevalence estimates cannot be simply added, because they were calculated without regard for multiple pathologies or comorbidities (that is, it is possible for a given individual to fall within more than one disability level if they have more than one diagnosis). Murray & Lopez adjusted for comorbidity assuming that all conditions were statistically independent and that the probability of having two conditions was the product of the probability of having each condition alone (3) . For the disability estimates presented here, a comorbidity adjustment method was used that takes account of the increased probability of having certain pairs of conditions (23) . Only very limited data were available on comorbidity, so the resulting disability prevalences have an additional level of uncertainty above the cause-specific prevalence estimates. Data sources and methods Clinically and conceptually it is not usual practice to infer disability from diagnoses. Disabilities are limitations or problems in the performance of actions or tasks. They are identified and assessed in their own right -and in some cases may only subsequently be explained in terms of pathology, or may never be adequately explained at all. In future revisions of the GBD study, increased effort will be devoted to the estimation of the prevalences of impairments and disabilities directly, and to ensuring consistency with the disease-and injury-specific sequelae estimates. The GBD disability prevalence estimates have the virtue of comprehensiveness, and at least some grounding in disease prevalence. However, they are very much approximations, and are subject to very clear limitations in the way they were compiled and in the way that comorbidity was addressed in adding across causes. WHO recently published mortality projections from 2003 to 2030 using methods similar to those applied in the original GBD study and starting from the GBD 2002 estimates (19) . A set of relatively simple models were used to project future health trends for baseline, optimistic and pessimistic scenarios, based largely on projections of economic and social development, and using the historically observed relationships of these with cause-specific mortality rates. Updated projections to the year 2030 have been prepared using the updated GBD 2004 results as a starting-point. The methods used are essentially the same as those previously published (19) , with the following changes: • Projections of income per capita were revised to take account of World Bank revisions to purchasing power parity (PPP) conversion rates (157) and latest World Bank regional projections of real growth per annum in income per capita (21); • Countries for which low-income regression coefficients were used were revised to take account of revised country income levels and projections; • The projection regression equations were recalibrated so that back projections of child mortality rates from 2004 to 1990 matched observed trends for World Bank regions. In the recalibrated projections, the regression coefficient for human capital was left unchanged and the regression coefficient for time (a proxy for technological change) was set to zero for low-income countries in the African, European, South-East Asia and Western Pacific Regions; • Revised UNAIDS projections for HIV mortality were used, based on projections of current trends in scale-up of antiretroviral therapy coverage; • Relative risks for diabetes mellitus mortality associated with increasing levels of body mass index (kg/m 2 ) were revised downwards based on latest information from the Asia Pacific Cohorts study. There remain substantial data gaps and deficiencies, particularly for regions with limited death registration data. The GBD 2004 includes results for these regions, albeit with wider uncertainty ranges, based on the best possible assessment of the available evidence. Uncertainty ranges for all-cause mortality rates for WHO Member States were published in the World health report 2006 (25) . Uncertainty analysis for the GBD 2004 cause-specific estimates has not been carried out, but uncertainty ranges are likely to be similar to those assessed for the GBD 2001 (158) . Ninety-five per cent uncertainty ranges for regional cause-specific mortality estimates were calculated using simulation methods based on estimated uncertainty ranges for input data (158) . Uncertainty in estimated all-cause mortality for 2001 ranged from ±1% for high-income countries to ±15-20% for sub-Saharan Africa, reflecting differential data availability. Uncertainty ranges were generally larger for deaths from specific diseases. For example, the relative uncertainty for deaths from ischaemic heart disease ranged from ±12% for high-income countries to ±25-35% for sub-Saharan Africa. Assessments of 2001 YLD uncertainty for specific causes took into account not only typical levels of measurement error in the input data sets, but also expert judgment about the degree of uncertainty arising from the lack of representativeness of the available data for each region. The resulting uncertainty varied considerably across causes, ranging from relatively certain estimates for diseases such as polio, for which intensive surveillance systems are in place, to highly uncertain estimates for those such as osteoarthritis, where for some regions no usable data source was found, and for others the latest available data were decades old. Typical uncertainty for regional prevalence estimates ranged from ±10% to ±90%, with a median value of ±41%, among a subset of diseases for which uncertainty analysis was carried out (158) . Although the GBD 2004 estimates have similarly large uncertainty ranges for some causes and some regions, they continue to provide useful information on broad relativities of disease burden, on the relative importance of mortality and disability, and on regional patterns of disease burden. The analysis of the levels of uncertainty in the GBD 2001 estimates reinforces the need for caution when interpreting global comparative epidemiological assessments and the need for increased investment in population health measurement systems and in improved methods for analysing available population health data. The projected declining age-specific death rates for most chronic diseases reflect the observed declines in age-specific chronic disease death rates with increasing levels of development in the available death registration data for 107 countries between 1950 and 2002. Adverse trends for some chronic disease risk factors such as overweight and physical inactivity were probably more than offset in these countries by improved control of other risk factors such as high blood pressure, high blood cholesterol and tobacco smoking, and improved access to effective treatment interventions. Projected chronic disease mortality rates are not highly sensitive to a reasonably broad range of assumptions about future economic growth and trends in the tobacco epidemic. The projected decline in communicable disease mortality rates is more sensitive to these assumptions, and may be optimistic if future trends in economic growth are not as high as projected. The mortality and burden of disease projections are less firm than the base year assessments, and provide "business as usual" projections under specified assumptions that do not specifically take account of trends in major risk factors apart from tobacco smoking and, to a limited extent, overweight and obesity. If risk factor exposures do not generally decline with economic development and with improving health systems in developing countries, then these projections may underestimate future deaths in low-and middle-income countries. A00-B99, G00-G04, N70-N73, J00-J06, J10-J18, J20-J22, H65-H66, O00-O99, P00-P96, E00-E02, E40-E46, E50, D50-D53, D64.9, E51- Leading causes of death Causes of death among children aged under five years Years of life lost: taking age at death into account Zimbabwe Region of the Americas High Bahamas, Canada, United States of America Region of the Americas LMIC Antigua and Barbuda Viet Nam Non-Member States or territories American Samoa Western Sahara a Income categories for 2004 as defined by the World Bank (159). Countries are divided among income groups according to 2004 gross national income (GNI) per capita. The groups are low income (US$825 or less), lower middle income (US$826-3255), upper middle income STD, sexually transmitted diseases Symptoms, signs and ill-defined conditions" (780-799 in ICD-9 and R00-R99 in ICD-10) are distributed proportionately to all causes within Group I and Group II b Cancer deaths coded to ICD categories for malignant neoplasms of other and unspecified sites including those whose point of origin cannot be determined, and secondary and unspecified neoplasms (ICD-10 C76, C80, C97 or ICD-9 195, 199) were redistributed pro-rata across the footnoted malignant neoplasm categories within each age-sex group, so that the category "Other malignant neoplasms These include heart failure, ventricular dysrhythmias, generalized atherosclerosis and ill-defined descriptions and complications of heart disease. Proportions of deaths coded to these causes were redistributed to ischaemic heart disease as described in (23) e For countries with 3-digit ICD10 data, for "Road traffic accidents" use: V01-V04, V06, V09-V80, V87, V89 and V99. For countries with 4-digit ICD10 data, for "Road traffic accidents" use: V01.1-V01.9, V02.1-V02.9, V03.1-V03.9, V04.1-V04.9, V06.1-V06.9, V09.2, V09.3, V10.3-V10.9, V11.3-V11.9, V12.3-V12.9, V13.3-V13.9, V14.3-V14.9, V15.4-V15.9, V16.4-V16.9, V17.4-V17.9, V18.4-V18.9, V19.4-V19.9 Evidence-based health policy -lessons from the Global Burden of Disease Study The global burden of disease: a comprehensive assessment of mortality and disability from diseases, injuries and risk factors in 1990 and projected to 2020. Cambridge, Harvard School of Public Health on behalf of the World Health Organization and the World Bank Global health statistics. Cambridge, Harvard School of Public Health on behalf of the World Health Organization and the World Bank The global burden of disease. Cambridge, Harvard School of Public Health on behalf of the World Health Organization and the World Bank World health report 2004: changing history. Geneva, World Health Organization Death and DALY estimates for 2002 by cause for WHO Member States Disease control priorities in developing countries Global burden of disease and risk factors Global Burden of Disease 2005: call for collaborators The burden of disease and mortality by condition: data, methods and results for UNAIDS, World Health Organization. AIDS epidemic update United Nations Population Division. World population prospects -the 2006 revision Maternal and child undernutrition: global and regional exposures and health consequences Can we achieve Millennium Development Goal 4? New analysis of country trends and forecasts of under-5 mortality to Can the world afford to save the lives of 6 million children each year? Preventing chronic diseases: a vital investment: WHO global report. Geneva, World Health Organization Projections of global mortality and burden of disease from Resource needs for AIDS in low-and middle-income countries: estimation process and methods. Methodological Annex II: Revised projections of the number of people in need of ART Global economic prospects Age standardization of rates: a new WHO standard. Geneva, World Health Organization Adjusting for comorbidity in the calculation of health-adjusted life expectancies Global burden of disease in 2002: data sources, methods and results. Geneva, World Health Organization World health report 2006: working together for health Neonatal and perinatal mortality: country, regional and global estimates 2004. Geneva, World Health Organization, Department of Making Pregnancy Safer Mortality in China Mortality registration and surveillance in China: history, current situation and challenges Completeness of India's sample registration system: an assessment using the general growth balance method The epidemiologic transition revisited: compositional models for causes of death by age and sex. Population and Development Review The burden of disease in Zimbabwe in 1997 as measured by disability-adjusted life years lost Cause-specific mortality rates in sub-Saharan Africa and Bangladesh La Mortalité à Antananarivo de 1984 à 1995 Maputo Central Hospital, Ministry of Health Evaluating national cause-of-death statistics: principles and application to the case of China Prospective study of one million deaths in India: rationale, design, and validation results Mortality in the Islamic Republic of Iran Characterizing the epidemiological transition in Mexico: national and subnational burden of diseases, injuries, and risk factors Joint United Nations Programme on HIV/AIDS AIDS in South Africa in the light of death registration data: In search of elusive estimates Revised burden of disease estimates for the comparative risk factor assessment A critical assessment of mortality statistics in Thailand: potential for improvements Estimating mortality and causes of death in Turkey: methods, results and policy implications World health report 2005: child and maternal survival. Geneva, World Health Organization WHO Child Health Epidemiology Reference Group. WHO estimates of the causes of death in children Predicting the distribution of underfive deaths by cause in countries without adequate vital registration systems 4 million neonatal deaths: when? where? why? Estimating the causes of 4 million neonatal deaths in the year 2000 Global tuberculosis control: surveillance, planning and financing: WHO report Estimating the child mortality due to diarrhoea in developing countries How best to estimate the global burden of pertussis? WHO/UNICEF estimates of national immunization coverage. Geneva, World Health Organization WHO ICE-T: immunization coverage estimates and trajectories, version 4.0. Geneva, World Health Organization, Department of Immunization Vaccine-preventable diseases Maternal and neonatal tetanus A modelbased approach to monitoring global progress in the elimination of neonatal tetanus. Paper presented to WHO Quantitative Immunization and Vaccines Related Research Advisory Committee WHO Department of Immunization. Meningitis estimates. Paper presented to the expert review of estimates of morbidity and mortality associated with Haemophilus influenzae type b (Hib) and Streptococcus pneumoniae (SP) General analytic methods. Paper presented to the expert review of estimates of morbidity and mortality associated with Haemophilus influenzae type b (Hib) and Streptococcus pneumoniae (SP) Malaria incidence estimates at country level for the year 2004 -proposed estimates and draft report. Geneva, World Health Organization. Roll Back Malaria Monitoring and Evaluation Reference Group & MERG Task Force on Malaria Morbidity Roll Back Malaria, World Health Organization Malaria risk: estimating clinical episodes of malaria Malaria Epidemiology Reference Group. Minutes of the MERG Taskforce Meeting on malaria morbidity The burden of malaria mortality among African children in the year 2000 Conquering malaria An epidemiologic model of severe morbidity and mortality caused by Plasmodium falciparum Quantifying the level of under-detection of Trypanosoma brucei rhodesiense sleeping sickness cases Pan American Health Organization, WHO Program on Neglected Tropical Diseases. Estimación cuantitativa de la enfermedad de Chagas en les Américas Pan American Health Organization The global epidemiology of infectious diseases Helminth infections: soil-transmitted helminth infections and schistosomiasis Morbidity and infection with schistosomes or soil-transmitted helminths A longitudinal study of Schistosoma haematobium infection in Qena governorate, Upper Egypt. 1. Initial epidemiological findings The Brazilian program for schistosomiasis control Dengue and dengue haemorrhagic fever DengueNet: WHO's internet-based system for the global surveillance of dengue fever and dengue haemorrhagic fever References 82. Global elimination of trachoma documents. Geneva, World Health Organization Estimates of worldwide distribution of child deaths from acute respiratory infections Maternal mortality in 2005: estimates developed by WHO, UNICEF, UNFPA and the World Bank. Geneva, World Health Organization Induced abortion: estimated rates and trends worldwide Unsafe abortion: global and regional estimates of the incidence of unsafe abortion and associated mortality in 2003. Geneva, World Health Organization The World Health Organization Global Database on Child Growth and Malnutrition: methodology and applications Methodology for estimating regional and global trends of child malnutrition WHO child growth standards: length/height-for-age, weight-for-age, weightfor-length, weight-for-height and body mass index-for-age: methods and development WHO child growth standards Worldwide prevalence of anemia in preschool aged children, pregnant women and nonpregnant women of reproductive age Globocan 2000: cancer incidence, mortality and prevalence worldwide, version 1.0. Lyon, IARCPress Global and regional estimates of cancer mortality and incidence by site: I. Application of regional cancer survival model to estimate cancer mortality distribution by site Global and regional estimates of cancer mortality and incidence by site: II. Results for the Global Burden of Disease Study GLOBOCAN 2002: cancer incidence, mortality and prevalence worldwide Prevalence of diabetes and impaired fasting glucose in adults in the U.S. population: national health and nutrition examination survey Characteristics of patients with type 2 diabetes in Mexico: results from a large population-based nationwide survey Diabetes and hypertension increases in a society with abdominal obesity: results of the Mexican National Health Survey Prevalence of diabetes mellitus and impaired glucose tolerance in the urban population aged 30-69 years in Ribeirão Preto The epidemiology of diabetes in a large Israeli HMO The epidemiology of type 2 diabetes mellitus in Greek adults: the ATTICA study Prevalence, incidence and mortality of type 2 diabetes mellitus revisited: a prospective population-based study in The Netherlands Prevalence of type II diabetes mellitus in population of Krakow The prevalence of type II diabetes mellitus in rural urban population over 35 years of age in Lublin region (Eastern Poland) The prevalence and management of diabetes in Thai adults Prevalence of diabetes and impaired fasting glucose in the Chinese adult population: International Collaborative Study of Cardiovascular Disease in Asia Type 2 diabetes in rural and urban population: diverse prevalence and associated risk factors in Bangladesh The burden of diabetes and impaired glucose tolerance in India using the WHO 1999 criteria: prevalence of diabetes in India Study (PODIS). Diabetes Research and Clinical Practice Global prevalence of diabetes: estimates for the year 2000 and projections for 2030 Surveillance of chronic disease risk factors: country level data and comparable estimates World Health Organization Body mass index and risk of diabetes mellitus in the Asia-Pacific region Global burden of depressive disorders in the year 2000 Depressive disorders in Europe: prevalence figures from the ODIN study Global burden of alcohol use disorders in the year 2000: summary of methods and data sources. Global burden of disease 2000 working paper. Geneva, World Health Organization WHO Collaborating Centre at the Research Centre for Public Health and Addiction A generic model for the assessment of disease epidemiology: the computational basis of DisMod II Health systems performance assessment: debates, methods and empiricism. Geneva, World Health Organization Comparative quantification of health risks: global and regional burden of disease attributable to selected major risk factors. Geneva, World Health Organization The costs of substance abuse in Canada. Ottawa, Canadian Centre on Substance Abuse Estimating the global burden of disease of mild mental retardation and cardiovascular diseases from environmental lead exposure Comparative quantification of health risks: global and regional burden of disease attributable to selected major risk factors Global data on blindness Global data on visual impairment in the year 2002 Global magnitude of visual impairment caused by uncorrected refractive errors in 2004 Disability weights for diseases in the Netherlands Global burden of ischaemic heart disease in the year 2000 The burden of disease and injury in the United States Survival and cause of death after myocardial infarction: the Danish MONICA study Prevalence and management of diabetes and associated risk factors by regions of Thailand: third National Health Examination Survey Health systems performance assessment: debates, methods and empiricism. Geneva, World Health Organization The global burden of oral diseases and risks to oral health. Bulletin of the World Health Organization Conflict barometer 2003: 12th annual conflict analysis. Heidelberg, Department of Political Science Conflict barometer 2004: 13th annual conflict analysis. Heidelberg, Department of Political Science Project Ploughshares. Armed conflicts report 2005. Waterloo, Canada, Project Ploughshares Project Ploughshares. Armed conflicts report Peace and conflict 2005: a global survey of armed conflicts, self-determination movements, and democracy Armed conflict as a public health problem Mortality in the Democratic Republic of Congo: a nationwide survey Iraq Family Health Survey Study Group. Violence-related mortality in Iraq from Death in Darfur Darfur: counting the deaths. Mortality estimates from multiple survey data What are the trends? Brussels World Health Organization, European Programme for Intervention Epidemiology Training. Retrospective mortality survey among the internally displaced population Landmine monitor report 2005: toward a mine-free world Landmine monitor report 2006: toward a mine-free world EM-DAT: the OFDA/CRED international disaster database Centre for Research on the Epidemiology of Disasters Damage and seismic intensity of the 1996 Lijiang earthquake, Vhina: a GIS analysis Public health impact and medical consequences of earthquakes Who died as a result of the tsunami? Risk factors of mortality among internally displaced persons in Sri Lanka: a retrospective cohort analysis Tsunami mortality in Aceh Province Sensitivity and uncertainty analyses for burden of disease and risk factor estimates Deaths assessed by medical personnel in city hospitals 1973-1992 (159) Mortality and causes of death in Jordan 1995-96: assessment by verbal autopsy Third national health family planning and social welfare plan 1992-1995. Ministry of Health Family Planning and Social Welfare Demographic and Health Survey Demographic and Vital Statistics Survey Deaths assessed by verbal autopsy, Niakhar, Senegal 1983-1990. Dataset provided by Centre Population et Développement Report of the Minister of Health Congenital abnormalities 25 280 12 853 12 427 12 187 638 28 11 892 507 29 Oral diseases 7 875 3 878 3 997 1 138 2 010 730 1 074 2 015 907 Dental caries 4 882 2 476 2 406 1 122 1 180 174 1 057 1 148 201 Periodontal disease 320 160 160 0 150 9 2 147 11 Edentulism 2 555 1 191 1 364 0 651 540 0 679 685 III. Injuries 187 614 123 366 64 249 26 529 92 220 4 617 21 819 38 823 3 607 Unintentional 138 564 87 130 51 434 24 340 59 345 3 445 20 211 28 318 2 905 Road traffic accidents 41 223 29 240 11 983 5 308 22 974 958 4 174 7 266 542 Poisonings 7 447 4 893 2 554 678 3 979 236 528 1 739 287 Falls 17 157 10 447 6 710 3 312 6 336 799 2 984 2 933 793 Fires 11 271 4 534 6 738 2 051 2 367 116 2 514 4 038 185 Drownings 10 728 7 354 3 374 3 094 4 079 181 1 (58, 59) . Assumptions and inputs for both sets of estimates were revised to take account of most recent available data on numbers of susceptible births (not protected by vaccination of the mother) and from elimination status surveys. Final estimates were reasonably consistent and WHO Department of Immunization, Vaccines and Biologicals estimates for neonatal and postneonatal tetanus incidence and mortality were used for the GBD 2004. Haemophilus influenzae type B (HiB) meningitis incidence, together with incidence for meningitis due to Streptococcus pneumoniae and Neisseria meningitidis were estimated by the Department of Immunization, Vaccines and Biologicals using an incidence-based model whose inputs included vaccine coverage and estimates of age-specific case fatality rates (60, 61) . These models estimated meningitis incidence and mortality in the 1-59 month age group. Neonatal meningitis deaths were separately estimated, as described in Section B3. For ages 5 years and over, GBD 2002 estimates were applied to 2004 populations. Revised incidence estimates for episodes of illness due to all forms of malaria, and the proportions due to Plasmodium falciparum, were based on estimates prepared by the Roll Back Malaria Partnership in 2004 and 2005 (62-64) . These incidence estimates were adjusted as follows. First, case reports data from national health information systems were updated to 2004. Second, the risk-based estimates were compared to national case reports and adjusted downwards where necessary to be consistent with a minimum reporting completeness of 2% rather than the 1% used in the Roll Back Malaria estimates. This reduced estimated cases for certain large countries in South-East Asia to levels more consistent with country studies and advice. Third, for other countries outside Africa with reporting completeness greater than 2%, incidence estimates were reduced by 10% to account for reductions in endemicity levels associated with increased urbanisation (65) . Fourth, incidence and mortality estimates for 2004 were compared country by country with estimates for 2006 prepared for the Global Malaria Report 2008, and adjusted for countries where improved 2006 estimates were available based on an analysis of health facility and survey data (66) . Finally, implied case fatality rates were calculated using the mortality estimates (described below) and for a few countries where case fatality rates were zero or very high, adjustments made to mortality or incidence estimates -these were mostly countries with low levels of malaria. Global and regional estimates of the total number of malaria cases in 2004 are shown in Table B5 .Revised mortality estimates for children aged under five years in sub-Saharan Africa were based on the analysis by Rowe et al. (67) . Cause fractions for malaria and other major child causes of death were adjusted to fit postneonatal child death envelopes for 2004 country-by country as described in Section B2. Countries with larger numbers of malaria deaths tended to be those where cause-specific fractions required greater adjustment, so the input estimates for malaria deaths (804 600 for sub-Saharan Africa) were adjusted to 724 400. This is quite similar to the implied figure of 744 000 estimated by Breman et al. in the Disease Control Priorities project (68) . For countries outside sub-Saharan Africa, child mortality due to malaria was based on either recent death registration data, or on the GBD 2002 estimates adjusted for consistency with postneonatal child death envelopes and for plausibility of implied case fatality rates.The GBD 2002 applied an assumption that 90% of malaria deaths were children aged under five years in all regions of the world, based on some information Annex B registration number. The GBD 1990 estimated 1.5 million DALYs and 8000 deaths due to schistosoma infections. Schistosomiasis cases were treated as a single sequela, with an average disability weight of 0.006 (74) . The GBD 2002 used the same disease model with updated prevalence estimates to estimate that 1.7 million DALYs and 15 371 deaths were due to schistosomiasis in 2002.There are five different types of schistosoma species: Schistosoma haematobium (the most prevalent and widespread species in Africa and the Middle East), S. mansoni (found in Africa and the only species seen in Latin America), S. japonicum (restricted to the Pacific region including China and the Philippines), S. mekongi (found in limited areas of the Lao People's Democratic Republic and Cambodia) and S. intercalatum (found in 10 countries in the rainforest belt of Africa). For the 2004 revision, the incidence and prevalence of cases of infection were separately estimated by country for S. mansoni, S. haematobium and S. japonicum plus S. mekongi. Severe renal damage (S. haematobium) and severe liver damage (other forms) were also modelled, assuming the incidence of these sequelae was the same as the estimated numbers of deaths, and the disability weights were the same as those for end-stage renal disease. Cancer outcomes were not included as sequelae for schistosomiasis, but are included with overall cancer burden estimates. A separate calculation would be required to estimate attributable cancer burden for schistosomiasis.Estimates of prevalence for infection with S. mansoni and S. haematobium in African countries were based on information provided by the WHO programme for Control of Neglected Tropical Diseases for [2003] [2004] . For countries in the Eastern Mediterranean region, the previous 2002 prevalence estimates were used. For the 2004 revision, incidence rates were estimated assuming zero remission rates (except for a period in early adulthood when prevalence rates decline). The estimated durations are much longer than the one year assumed in the GBD 2002 and GBD 1990, and as a result of time discounting, the YLD estimated for the 2004 revision (0.88 million globally) are lower than those published for 2002 (1.5 million).The GBD 2002 estimated that schistosomiasis was responsible for around 15 000 deaths globally (excluding attributable cancer deaths), although others have argued that the figure should be much higher (75) . Van der Werf et al. (76) , using limited data from Africa, estimated that schistosomiasis caused 210 000 deaths annually. A literature review found limited data from studies with small sample sizes, limiting ability to extrapolate to population level. In the absence of usable studies, a back-calculation method was employed to estimate approximate case fatality rates for two populations with significant numbers of schistosomiasis deaths recorded in death registration data.Egyptian death registration data for 2000 gave an estimate of 3303 schistosomiasis deaths in 2000 after adjustment for incompleteness of reporting. Reported deaths did not distinguish between S. mansoni and S. haematobium. Estimates of population prevalence for the 1980s and 1990s vary widely (20-40%); see for example Mansour et al. (77) . Assuming a prevalence rate of 30% in the 1980s, this gives an approximate case fatality rate of around 0.03%. This method was also applied to Brazil, where only S. mansoni is endemic. An analysis of Brazilian death registration data for 1990 to 1995 found that annual numbers of deaths due to schistosomiasis averaged around 570. Based on a conservative estimate of infection prevalence of around 8 million in 1980 (78) , this gives a case fatality rate of around 0.007%.These estimates are generally consistent with those reported in the literature, and also with the known lower fatality rates for S. mansoni compared with S. haematobium. For the GBD 2004 update, annual case fatality rates for prevalent cases were conservatively assumed to be 0.01% for S. mansoni, 0.02% for S. haematobium, and 0.03% for S. japonicum and S. mekongi. Applying these case fatality rates to the revised prevalence estimates gave a revised global estimate for deaths due to schistosomiasis (excluding cancers caused by schistosomiasis) of 41 000 for 2004. The sex ratio of schistosomiasis deaths was assumed to be similar to the 1990 estimates (62.5% male in the African and Eastern Mediterranean regions, 66% in the Americas and Asia, and 80% male in China).The Global Burden of Disease: 2004 Update provides a comprehensive assessment of the health of the world's population in 2004. Consistent and comparative description of the burden of diseases and injuries is an important input to health decision-making and planning processes. The Global Burden of Disease provides a framework for integrating the available information on mortality and health in populations to assess the comparative importance of diseases and injuries in causing premature death, loss of health and disability in all regions of the world.This report draws on the extensive databases of the World Health Organization and information provided by Member States to provide detailed estimates of premature mortality, disability and loss of health for 135 causes by age and sex for the world, for regions of the world, and for countries grouped by average income per capita. It builds on previous assessments for the years This report is part of ongoing efforts by WHO to monitor and analyse the global health situation and trends and to foster increased comparability of data between countries. The report builds upon the work of technical programmes within WHO at country, regional and global levels, and collaboration with partners, including UN agencies, the private sector and academic institutions. ISBN 978 92 4 156371 0