347 Million Diabetics Counted Worldwide
Download the pdf here|
By John Gever, Senior Editor, MedPage Today
Published: June 25, 2011
· Explain that the global prevalence of diabetes reached 347 million in 2008, more than doubling the number of people with diabetes worldwide since 1980.
· Point out that population growth and longer lifespans accounted for about 70% of the increase and the remaining 30% was likely related to changes in diets, exercise habits, and other factors that drive development of diabetes.
· Note that in 1980, the U.S. had one of the lowest diabetes prevalence rates in the world, but in 2008 it ranked in the upper third of the 199 nations included in the current study.
The global prevalence of diabetes reached 347 million in 2008, more than doubling the number of people with diabetes worldwide since 1980, researchers said.
Using data from some 280 health surveys and epidemiological studies from around the world, Majid Ezzati, PhD, of Imperial College London, and colleagues estimated an average, age-standardized, global prevalence of diabetes of 9.8% in adult men (95% CI 8.6% to 11.2%) and 9.2% in adult women (95% CI 8.0% to 10.5%).
Corresponding figures in 1980 were 8.3% in men and 7.5% in women, for a total worldwide diabetic population of 153 million, the researchers reported online in The Lancet.
They noted that population growth and longer lifespans in many countries accounted for about 70% of the increase in absolute numbers of diabetics.
Nevertheless, the remaining 30% was likely related to changes in diets, exercise habits, and other factors that drive development of diabetes.
The analysis showed that age-standardized levels of fasting plasma glucose averaged 99.2 mg/dL among men and 97.7 mg/dL in women globally in 2008. These represented increases of 1.3 mg/dL for men and 1.6 mg/dL for women since 1980.
Increases in mean fasting plasma glucose were seen in every region, with Pacific Ocean nations, "southern and tropical" Latin American countries, and the world's high-income nations showing the biggest rises.
The world's highest diabetes prevalence was seen in the Marshall Islands, at about 28.5% for both sexes combined. The lowest rates were in Cambodia and the Netherlands, where just over 5% of the adult population was considered diabetic.
Diabetes prevalence in the U.S. rocketed upward in the new analysis, as it has in CDC studies and others.
In 1980, the U.S. had one of the lowest diabetes prevalence rates in the world, but in 2008 it ranked in the upper third of the 199 nations included in the current study.
Prevalence in U.S. men soared from 6% in 1980 to 12.5% in 2008, the researchers calculated. A smaller but still very steep rise was seen in American women, among whom diabetes prevalence jumped to 9.5% from 5% in 1980.
More surprisingly, rapid industrialization and increasing wealth in China was not reflected in rising glucose levels or diabetes prevalence through 2008. Ezzati and colleagues found, for example, that 10% of Chinese men were diabetic in 1980 -- and that only about 9.5% were diabetic in 2008.
Similar findings were seen for a number of other countries with growing economies, such as Malaysia and Cambodia.
Elsewhere in the world -- such as in the U.S. -- increasing affluence has driven shifts to the high-carb diets and relatively sedentary lifestyles thought to promote type 2 diabetes.
Most regions with slower-than-average growth in diabetes were those with slower-than-average economic growth: central and eastern Europe, central Asia, and northern and sub-Saharan Africa.
Ezzati and colleagues also thought it was interesting that men in South Asia -- principally India -- "had the second smallest change in body mass index (almost zero)" among the 21 regions they studied, "but the sixth highest rise in mean fasting plasma glucose."
Such interregional variations during the 29-year study period probably reflects differences in "genetic factors associated with ethnic origin, fetal and early life nutritional status, diet quality, and physical activity" that affect glycemic values and metabolism.
Ezzati and colleagues had reviewed some 1,700 published studies of population-based diabetes prevalence and/or fasting plasma glucose measurements, as well as 154 national health surveys and epidemiological studies plus data retrieved from a World Health Organization database.
After excluding studies with duplicated data, those that did not break out findings by age and sex, and studies measuring capillary rather than plasma glucose levels, the researchers were left with a total of 282 studies, representing 370 country-years of data and 2.7 million individual participants.
"Our systematic analysis shows that glycemia and diabetes are a rising global hazard, with the number of adults with diabetes having more than doubled over nearly three decades," Ezzati and colleagues wrote.
In an accompanying editorial, an official with New Zealand's health ministry called the findings "stark," but also called for more reliable and comprehensive data collection to get a better picture of the global situation.
"Country-level survey-based surveillance for dysglycemia and diabetes should use HbA1c as the test of choice, because it does not necessitate a fasting blood sample, is readily calibrated against an international reference standard, and is accepted by WHO as a diagnostic test for diabetes," Martin Tobias, MBBCh, said in the editorial.
He also indicated that many countries could easily do away with health surveys. Data from their health information systems can include far more detail and cover more of the population than is feasible with survey instruments, Tobias suggested.
He recommended that data on the degree of glycemic control and other parameters important to diabetes management, such as blood pressure and kidney function, should be included in future studies of diabetes prevalence.
Ezzati and colleagues acknowledged that they faced several hurdles and limitations in assembling data for their study.
Most notably, they indicated that data were sparse for some regions in the 1980s and for certain low- and middle-income countries throughout the study period.
Also, the included surveys and studies used a variety of glycemic metrics, including mean postprandial glucose, HbA1c, and diabetes prevalence with 18 different definitions.
However, Ezzati and colleagues wrote, their reported confidence intervals reflected the uncertainties introduced by these differences and other defects in the data.
National, regional, and global trends in fasting plasma glucose and diabetes prevalence since 1980: systematic analysis of health examination surveys and epidemiological studies with 370 country-years and 2·7 million participants
Data for trends in glycaemia and diabetes prevalence are needed to understand the effects of diet and lifestyle within populations, assess the performance of interventions, and plan health services. No consistent and comparable global analysis of trends has been done. We estimated trends and their uncertainties in mean fasting plasma glucose (FPG) and diabetes prevalence for adults aged 25 years and older in 199 countries and territories.
We obtained data from health examination surveys and epidemiological studies (370 country-years and 2·7 million participants). We converted systematically between different glycaemic metrics. For each sex, we used a Bayesian hierarchical model to estimate mean FPG and its uncertainty by age, country, and year, accounting for whether a study was nationally, subnationally, or community representative.
In 2008, global age-standardised mean FPG was 5·50 mmol/L (95% uncertainty interval 5·37-5·63) for men and 5·42 mmol/L (5·29-5·54) for women, having risen by 0·07 mmol/L and 0·09 mmol/L per decade, respectively. Age-standardised adult diabetes prevalence was 9·8% (8·6-11·2) in men and 9·2% (8·0-10·5) in women in 2008, up from 8·3% (6·5-10·4) and 7·5% (5·8-9·6) in 1980. The number of people with diabetes increased from 153 (127-182) million in 1980, to 347 (314-382) million in 2008. We recorded almost no change in mean FPG in east and southeast Asia and central and eastern Europe. Oceania had the largest rise, and the highest mean FPG (6·09 mmol/L, 5·73-6·49 for men; 6·08 mmol/L, 5·72-6·46 for women) and diabetes prevalence (15·5%, 11·6-20·1 for men; and 15·9%, 12·1-20·5 for women) in 2008. Mean FPG and diabetes prevalence in 2008 were also high in south Asia, Latin America and the Caribbean, and central Asia, north Africa, and the Middle East. Mean FPG in 2008 was lowest in sub-Saharan Africa, east and southeast Asia, and high-income Asia-Pacific. In high-income subregions, western Europe had the smallest rise, 0·07 mmol/L per decade for men and 0·03 mmol/L per decade for women; North America had the largest rise, 0·18 mmol/L per decade for men and 0·14 mmol/L per decade for women.
Glycaemia and diabetes are rising globally, driven both by population growth and ageing and by increasing age-specific prevalences. Effective preventive interventions are needed, and health systems should prepare to detect and manage diabetes and its sequelae.
Hyperglycaemia and diabetes are important causes of mortality and morbidity worldwide, through both direct clinical sequelae and increased mortality from cardiovascular and kidney diseases.1-6 With rising overweight and obesity,7 concern has risen about a global diabetes epidemic, with harmful effects on life expectancy and health-care costs.8, 9 A few studies have examined global patterns of glycaemia and diabetes, finding substantial variation between regions.1,10-13 Others have assessed trends in specific countries.14-24 Findings from these studies have helped to show that hyperglycaemia and diabetes are important worldwide and regional issues, but these studies also have limitations.
First, diabetes definitions have varied by expert committees and over time.25-27 Most current studies have either used all definitions without adjustment for incomparability, or selected one definition and excluded data based on other definitions. Second, these studies pooled national, subnational, and community data, regarding them as equally representative of countries' populations. Third, some data included in these studies used random (non-fasting) glucose measurement; other data were from specific occupational groups, communities with high obesity prevalence, health-care facilities and practitioners, registries, or self-reported diabetes. These sources were probably biased because obesity is a risk factor for hyperglycaemia, occupational groups might differ from the general population in their health risks, and some diabetes cases are undiagnosed.28, 29 Fourth, previous analyses assigned estimates to countries without data based on geographical proximity and ad-hoc expert opinion about similarity to countries with data without a formal analytical model. Fifth, these studies pooled data from different years without adjustment for underlying trends. Finally, these studies did not account for all sources of uncertainty including missing and older country data, leading to overly confident estimates.
These shortcomings have hindered our ability to systematically examine trends. In recent years, health examination surveys have measured different glycaemic indicators, providing an opportunity to systematically assess trends by country. We reviewed and accessed unpublished and published studies and collated comprehensive data for different glycaemic metrics. We applied statistical methods to systematically address measurement comparability, missing data, non-linear time trends, age patterns, and national versus subnational and community representativeness. With these data and methods, we estimated trends and associated uncertainties by country and region.
Our final dataset included 370 country-years with 2·7 million participants (figure 1). Of the included studies, 71% had reported mean FPG or diabetes prevalence based on FPG, with others using postprandial glucose or HbA1c (webappendix pp 7-24). 128 country-years were from 22 high-income countries and 242 from 85 low-income and middle-income countries. Japan had the most nationally representative data with 8 years of national data since 1980, followed by the USA and Singapore (webappendix pp 46-48). We could not identify any population-based data for 92 countries. Central Asia, central and eastern Europe, and sub-Saharan Africa had the largest proportion of countries without data (webappendix pp 49-50). National surveys contributed 29% of all data, subnational studies 19%, and community studies 52% (webappendix pp 49-50).
Globally, age-standardised mean FPG was 5·50 mmol/L (95% uncertainty interval 5·37-5·63) for adult men and 5·42 mmol/L (5·29-5·54) for women in 2008 (figure 2), having risen by an estimated 0·07 mmol/L per decade (-0·02 to 0·15; posterior probability of the observed increase being a true increase=0·94) for men and 0·09 mmol/L per decade (0·00-0·17; posterior probability=0·98) for women since 1980. Age-standardised prevalence of diabetes was 9·8% (8·6-11·2) in men and 9·2% (8·0-10·5) in women in 2008 (figure 3), leading to an estimated 173 (151-197) million men and 173 (151-197) million women with diabetes. 40% (about 138 million) of people with diabetes were from China and India, 10% (about 36 million) from the USA and Russia, and 12% (about 42 million) from Brazil, Pakistan, Indonesia, Japan, and Mexico. In 1980, age-standardised prevalence was 8·3% (6·5-10·4) in adult men and 7·5% (5·8-9·6) in women, yielding 77 (60-97) million men and 76 (58-97) million women with diabetes. Of the nearly 194 million additional cases of diabetes between 1980 and 2008, 70% (52-98) were attributable to population growth and ageing and the other 30% (2-48) to a rise in age-specific prevalences. Across regions, the epidemiological share of the change ranged from -2% in east and southeast Asia to 60% in Oceania (data not shown).
In 2008, mean FPG in men was lowest in sub-Saharan Africa (5·27 mmol/L, 4·96-5·58), followed by east and southeast Asia and high-income Asia-Pacific (figure 2). Women in high-income Asia-Pacific had the lowest mean FPG (5·17 mmol/L; 4·94-5·39) followed by sub-Saharan Africa and east and southeast Asia (figure 2). Oceania had the highest mean FPG and diabetes prevalence of any region in 2008, for both men (6·09 mmol/L, 5·73-6·49; and 15·5%, 11·6-20·1) and for women (6·08 mmol/L, 5·72-6·46; and 15·9%, 12·1-20·5). Mean FPG and diabetes prevalence were also high for both sexes in south Asia, Latin America and the Caribbean, and a region consisting of central Asia, north Africa, and the Middle East (figure 2 and figure 3). Men in the high-income region consisting of Australasia, North America, and western Europe also had relatively high FPG and diabetes prevalence (figure 2 and figure 3). In high-income subregions, mean FPG and diabetes were lower in Asia-Pacific and western Europe than in Australasia and north America, with the difference between the highest and lowest means and prevalences about 0·4 mmol/L and 4-5 percentage points, respectively (webappendix pp 51-56).
With few exceptions, countries with the highest FPG and diabetes prevalence in 2008 were in Oceania, north Africa and the Middle East, and the Caribbean, with age-standardised mean FPG 6·5 mmol/L or higher in the Marshall Islands, Kiribati, Saudi Arabia, the Cook Islands, and Samoa in both sexes (figure appendix 1); age-standardised diabetes prevalence in these countries ranged 21-25% in men and 21-32% in women (figure appendix 2). Countries in southeast Asia, east Africa, and Andean Latin America had the lowest mean FPG in 2008 (as low as 5 mmol/L or less; figure appendix 1). Of high-income countries, mean FPG and diabetes were highest in the USA, Greenland, Malta, New Zealand, and Spain, and lowest in the Netherlands and Austria, for both sexes and in France for women (figure appendix 1). Mean FPG in these western European countries was lower than in Japan and South Korea, despite having higher BMIs.7
FPG increased or at best remained unchanged in almost every region between 1980 and 2008 (figure 2). FPG increased the most in Oceania, by 0·22 mmol/L per decade (-0·02 to 0·47; posterior probability=0·97) in men and 0·32 mmol/L per decade (0·08-0·55; posterior probability >0·99) in women. This rise led to an estimated increase in age-standardised diabetes prevalence of 5·9 percentage points for men and 7·8 percentage points for women in this region (figure 3). Large increases in mean FPG were also recorded in southern and tropical Latin America and south Asia for men, and in south Asia and the combined region of central Asia, north Africa, and the Middle East for women (all >0·15 mmol/L per decade; posterior probabilities ≥0·87). In men and women, we recorded almost no change in mean FPG in east and southeast Asia and in central and eastern Europe during these 28 years (figure 2). Male FPG trend in sub-Saharan Africa was indistinguishable from no change (posterior probability=0·69), but women had an increase of 0·13 mmol/L per decade (-0·07 to 0·34; posterior probability=0·91). In high-income subregions, FPG increased the least in western Europe, by 0·07 mmol/L per decade (-0·08 to 0·21; posterior probability=0·82) in men and by 0·03 mmol/L per decade (-0·13 to 0·18; posterior probability=0·63) in women, which was statistically indistinguishable from no change. Conversely, in high-income North America, FPG rose by 0·18 mmol/L per decade (0·00-0·36; posterior probability=0·98) in men and 0·14 mmol/L per decade (-0·03 to 0·31; posterior probability=0·94) in women.
Apart from women in Singapore, for whom mean FPG decreased by 0·21 mmol/L per decade (-0·06 to 0·51; posterior probability=0·92), no country had a meaningful fall in FPG; the few countries with apparent decreases had posterior probabilities of 0·80 or less and hence were statistically indistinguishable from flat trends (figure 4). Countries with flat trends were in east and southeast Asia, sub-Saharan Africa, Andean and central Latin America, high-income Asia-Pacific, and, especially for women, in Europe (figure 4). Countries in Oceania and North Africa and the Middle East had the largest increase in mean FPG, by 0·5 mmol/L per decade or more in the Marshall Islands, Samoa, Kiribati, and Saudi Arabia.
Our model did well in external predictive validity tests. Specifically, the 95% uncertainty intervals of our model predictions included 95% of withheld study means for women and 96% for men (webappendix p 45). Our model also had good predictive validity in most regions and by year of data, gross domestic product, and age group. When we excluded all data for some countries (ie, created the appearance of no data when data were available), the uncertainty intervals of model predictions included 96% of the female study means that were known but excluded, and 98% of the male study means. Although data were sparse early in our analysis period, our model covered 99% of withheld values from 1980 to 1995, suggesting that our uncertainty estimates were slightly conservative when data were scarce.
Our systematic analysis shows that glycaemia and diabetes are a rising global hazard, with the number of adults with diabetes having more than doubled over nearly three decades. Although population growth and ageing are important contributors to this increase, there is also an important epidemiological component with age-standardised global mean FPG having increased by 0·07 mmol/L per decade or more.
Our estimate of 347 (314-382) million adults with diabetes is higher than Shaw and colleagues' estimate of 285 million for 2010 (panel).13 The differences between the estimates could be explained by the inclusion and exclusion criteria and the number of studies used, different age ranges of estimates, or different methods to deal with missing data, year of data, rural and urban data, and national versus subnational and community data. Recent narrative reviews have stated that diabetes is rising in Asia and Africa, without addressing whether the data were representative, whether age groups were the same in the included studies, and other aspects of data comparability.20,39-41 We screened all data sources used in these overviews and used additional sources. Although we generally recorded increasing mean FPG and diabetes prevalence, our quantitative results are not comparable with the previous reports because we had used a larger number of studies and different methods. Notably, with use of national studies from China, Taiwan, Thailand, Malaysia, Cambodia, and the Philippines, in addition to multiple subnational and community studies, we noted no increase in age-standardised diabetes prevalence in east and southeast Asia, although ageing and population growth led to an increase in the
number of people with diabetes.
Research in context
A few studies have examined global patterns of glycaemia and diabetes, but have not estimated past trends for all countries and regions. Other studies have assessed trends in specific countries or regions. A recent publication estimated that there were 285 million people with diabetes worldwide in 2010,13 but some of the data were from specific occupational groups, communities with high obesity prevalence, or health-care facilities and practitioners, registries, and self-reported diabetes. Recent narrative reviews20,39-41 have stated that diabetes is rising in Asia and Africa, without addressing incomparable age groups in the studies included and other aspects of data comparability. Previous studies (including one by some members of our study group1) also had not distinguished between data that are nationally representative and those that are subnational or from specific communities. We obtained data for the levels of different glycaemic metrics from the following sources: health examination surveys and epidemiological studies with anonymised data available to the Collaborating Group members; multicentre studies; review of published articles; and unpublished data sources identified through the WHO Global InfoBase. Our final dataset included 370 country-years with 2·7 million participants. We could not identify any population-based data for 92 countries. We had data for mean fasting and postprandial glucose, mean HbA1c, and diabetes prevalence with use of 18 different definitions. We systematically converted between different glycaemic metrics based on data sources that had measured multiple metrics.
Our estimate of 347 (uncertainty interval 314-382) million adults with diabetes in 2008, is higher than the previous 285 million estimate for 2010.13 The differences between the estimates could be due to the inclusion and exclusion criteria and the number of studies used; different age ranges of estimates; or different methods of handling missing data, differences in year of data, rural and urban data, and national versus subnational and community data. Although we generally found increasing mean fasting plasma glucose and diabetes prevalence, our quantitative results are not comparable with the previous reports because we had used a larger number of studies and different methods. Notably, with use of national studies from China, Taiwan, Thailand, Malaysia, Cambodia, and the Philippines, and multiple subnational and community studies, we recorded no increase in age-standardised diabetes prevalence in east and southeast Asia, although ageing and population growth led to an increase in the number of people with diabetes in these regions.
The strengths and innovations of this study include analysis of trends; the large amount of data accessed and used; systematic conversion between different metrics of glycaemia and definitions of diabetes; application of a Bayesian hierarchical model to estimate mean FPG, which included non-linear time trends and age associations and used national income, urbanisation, food availability, and BMI as covariates; incorporation of study coverage as offset and variance components; and systematic analysis and reporting of uncertainty. Coverage-specific offsets and variances allowed our estimates to use all available data and to follow data from nationally representative studies more closely. Coverage-specific variance components were larger for less representative data sources, which led to larger uncertainty when we did not have nationally representative data, thus representing the true availability of information.
The main limitation of our study is that despite extensive data seeking, many country-years still did not have data, especially in the 1980s and in some low-income and middle-income countries. The absence of data is reflected in wider uncertainty intervals. Our external predictive validity assessment showed that the estimates and their uncertainty intervals are valid; importantly, applications of our results should use the full uncertainty intervals. Further, we noted substantial incomparability in metrics of glycaemia in published data. Specifically, we had data for mean postprandial glucose, mean HbA1c, and diabetes prevalence using 18 different definitions. Although we systematically converted between different glycaemia metrics, the conversions led to increases in uncertainty intervals. Similarly, to estimate diabetes prevalence, we relied on mean FPG as an intermediate step for conversion between different metrics and for handling of missing data. The association between mean and prevalence could vary between countries beyond what is measured by variables in webappendix p 31-eg, because of variations in quality of care. Such variability is shown by larger uncertainty in our prevalence estimates than the uncertainty of mean FPG. The persistent incomparability of glycaemic metrics is partly because definitions for clinical purposes are the subject of continuous debate, research, and refinement. Population-based surveillance, however, needs indicators that are easy to measure and are comparable across populations and over time. Although we incorporated information about study coverage into our model and excluded studies based on random (non-fasting) blood glucose, other quality indicators-such as duration of fasting, laboratory methods, calibration, and other sources of interassay and intra-assay variability-were not included in the model, potentially accounting for some of the uncertainty in our estimates.
As important as the global rise were the similarities and differences between regions and countries. Trends ranged from nearly flat in some regions to a rise of 0·2-0·3 mmol/L per decade in Oceania. This variation is undoubtedly partly attributable to regional BMI trends;7 the correlation between change in BMI and FPG across the 21 subregions was 0·71 for women and 0·57 for men. However, genetic factors associated with ethnic origin, fetal and early life nutritional status, diet quality, and physical activity might also affect glycaemic values and trends. Notably, men in south Asia had the second smallest change in BMI (almost zero) of the 21 subregions7 but the sixth highest rise in mean FPG; women in this region had the fourth smallest BMI change (0·4 kg/m2 per decade) but the sixth largest rise in FPG, about the same as in high-income North America where female BMI rose three times as much.7
The global and regional trends in mean FPG differ from those of other metabolic risks-namely, systolic blood pressure which decreased globally and in most subregions,37 and total cholesterol which decreased in Australasia, Europe, and North America, but rose in east and southeast Asia and Asia-Pacific, leading to relatively unchanged global mean.42 Because high BMI is a risk factor for all three metabolic indicators, these differences probably arise from other determinants, including dietary composition and medical treatment. Specifically, although effective drugs to lower blood pressure and cholesterol are increasingly used for primary prevention of cardiovascular disease in high-income countries, the use of specific drugs for primary prevention, and the targets and intensity of glycaemic management, are still being investigated.43 Therefore, primary prevention of dysglycaemia will need weight control, physical activity, and improved diet quality. Such interventions are difficult to implement within populations and will not affect diabetes incidence in the short term. Therefore, health systems in most countries will inevitably have to develop programmes to improve detection and management of diabetes to slow progression to microvascular and macrovascular complications.