Global, regional, and national burden of diabetes from 1990 to 2021, with projections of prevalence to 2050: a systematic analysis for the Global Burden of Disease Study 2021

Diabetes is a chronic disease characterized by hyperglycaemia due to abnormal β-cell biology and insulin action. Prior GBD 2019 estimates ranked diabetes as the eighth leading cause of combined death and disability, with nearly 460 million cases in 2019. International Diabetes Federation (IDF) estimates suggested 537 million people had diabetes in 2021, with global health expenditures nearing US$966 billion. The NCD-RisC Study projected a very low probability of halting the rise in diabetes prevalence by 2025. Diabetes is also a major risk factor for ischaemic heart disease and stroke. Type 1 often develops in childhood, whereas type 2 has strong genetic and lifestyle components, particularly obesity and physical inactivity. Despite established prevention and management strategies, substantial disparities persist, especially in low- and middle-income countries (LMICs), in risk factor profiles, access to screening and treatment, and health-system capacity. This study applies the updated GBD analytical framework to produce comprehensive, location-, age-, and sex-specific estimates of diabetes prevalence and burden (1990–2021), quantify the proportion due to type 1 vs type 2 in 2021, estimate the contribution of 16 risk factors to type 2 diabetes, and forecast prevalence through 2050 to inform policy and planning.

The manuscript situates its analysis within prior global assessments: IDF diabetes atlases reporting prevalence and costs; NCD-RisC pooled analyses showing rising global prevalence and the low likelihood of meeting global targets; and the 2020 Lancet Commission and WHO Global Diabetes Compact highlighting the disproportionate burden in LMICs and calling for high-quality, granular data. The paper discusses known associations of type 2 diabetes with obesity, diet, and physical inactivity, and notes sex, socioeconomic, and racial/ethnic disparities reported in multiple settings. It also contrasts methodological choices with IDF and NCD-RisC (e.g., excluding self-reported diabetes without biochemical confirmation, use of MR-BRT and DisMod-MR), emphasizing the added granularity and forecasting to 2050.

Design: Systematic analysis as part of GBD 2021. Data from 27,193 sources were integrated for 204 countries/territories, 25 age groups, and both sexes (1990–2021). Main metrics were prevalence, years of life lost (YLLs), years lived with disability (YLDs), and disability-adjusted life-years (DALYs). Age-standardisation was applied to enable comparisons across differing population structures. Mortality: 25,666 location-years of vital registration and verbal autopsy data since 1980 were modelled using the Cause of Death Ensemble model (CODEm) in separate models for total diabetes, type 1, and type 2. Garbage codes were corrected; approximately 11.1% of deaths coded to diabetes were reassigned from garbage codes. Unspecified diabetes deaths (>50%) were redistributed to type 1 or type 2 using a log-linear regression with obesity prevalence as a covariate; all diabetes deaths under age 15 were assumed to be type 1. Non-fatal prevalence: The reference case definition was fasting plasma glucose (FPG) ≥7.0 mmol/L (126 mg/dL) or use of insulin/diabetes medication. Population-representative data using FPG, HbA1c, OGTT, PPG, and mean FPG from 1,527 location-years in 172 countries were included, with insurance claims from the USA and Taiwan where available. MR-BRT was used to adjust non-reference case definitions; mean FPG was converted to prevalence where necessary. DisMod-MR 2.1 generated internally consistent estimates of incidence, prevalence, and excess mortality for total diabetes and type 1; type 2 prevalence was derived by subtracting type 1 from total. Covariates included maternal age ≥35 and maternal education for type 1 incidence, HAQ Index for type 1 excess mortality, and obesity prevalence and year for total diabetes prevalence. Remission was assumed to be 0% for type 1 and ≤1% annually for total diabetes. Health loss: YLLs were computed as deaths multiplied by standard life expectancy at age of death; YLDs were prevalence of sequelae (neuropathy, diabetic foot, lower limb amputation, and vision loss due to retinopathy) multiplied by disability weights, with comorbidity correction assuming independence and multiplicative effects. DALYs are YLLs + YLDs. Risk-attributable burden: Comparative risk assessment estimated population attributable fractions (PAFs) for 16 risk factors (ambient and household air pollution, smoking, second-hand smoke, high alcohol use, high BMI, multiple dietary risks, low physical activity, high/low air temperature) by age, sex, location, and year. Exposure distributions were estimated (DisMod-MR 2.1 or ST-GPR); dose–response functions used flexible meta-regression with regularized splines and least-trimmed squares outlier trimming; TMRELs were defined per evidence. PAFs were applied to DALYs to quantify risk-attributable type 2 diabetes burden. Forecasting: Regression-based forecasts to 2050 were generated by age–sex–location using SDI as predictor for type 1 prevalence and BMI for type 2, aligned to GBD 2021 levels for 2021. Forecasted prevalence was multiplied by forecasted populations to obtain counts. Uncertainty: At each step, 100 draws were propagated to generate 95% uncertainty intervals (UIs). Analyses used R 4.2.2; code is publicly available.

- Global prevalence: In 2021, 529 million (95% UI 500–564) people lived with diabetes; global age-standardised prevalence was 6.1% (5.8–6.5). Adults 20–79 years: 485 million (456–517). - Geographic variation: Highest age-standardised super-region prevalence in north Africa and the Middle East (9.3% [8.7–9.9]); highest regional prevalence in Oceania (12.3% [11.5–13.0]). Eastern sub-Saharan Africa had the lowest (2.9% [2.7–3.1]). Age-specific highlight: In ages 75–79 years, global prevalence peaked at 24.4% (22.3–26.2); Qatar had 76.1% (73.1–79.5) in this age group. Age-standardised prevalence exceeded 10% in 43 countries/territories in 2021. - Sex differences: Global age-standardised prevalence was higher in males (6.5% [6.2–7.0]) than females (5.8% [5.4–6.1]); sex ratio 1.14 (1.13–1.15), with regional variation. - Type-specific composition: Type 2 diabetes accounted for 96.0% (95.1–96.8) of cases and 95.4% (94.9–95.9) of diabetes DALYs worldwide in 2021; in all super-regions, >90% of prevalence was type 2. - Burden: 2021 YLLs 37.8 million (35.4–40.2); YLDs 41.4 million (29.5–55.4); DALYs 79.2 million (67.8–92.5). Global age-standardised DALY rate 915.0 (782.6–1067.5) per 100,000; highest regionally in Oceania (3577.0 [3157.0–4120.5]). Fiji had the highest country DALY rate (7333.9 [6066.7–8776.7] per 100,000). - Risk factors: 58.9 million (44.2–73.9) type 2 diabetes DALYs (76.5% [58.0–87.5]) were attributable to the 16 assessed risks. High BMI was the leading risk, accounting for 52.2% (25.5–71.8) of type 2 DALYs globally; its contribution increased by 24.3% (18.5–30.4) from 1990 to 2021. High BMI contributed >60% of type 2 DALYs in several super-regions and >50% in 167 (81.9%) countries/territories; in south Asia it accounted for 39.5% (17.1–58.4). - Projections to 2050: Global age-standardised total diabetes prevalence is projected to rise 59.7% (54.7–66.0) from 6.1% to 9.8% (9.4–10.2), yielding 1.31 billion (1.22–1.39) people with diabetes by 2050. About 49.6% of the increase is driven by obesity trends, 50.4% by demographic shifts. Age-standardised prevalence >10% is projected in two super-regions: north Africa and Middle East (16.8% [16.1–17.6]) and Latin America and Caribbean (11.3% [10.8–11.9]); 89 (43.6%) of 204 countries/territories will exceed 10%, and 24 (11.8%) will surpass 20%. Type 2 prevalence will rise 61.2% (56.2–68.1) to 9.5% (9.0–9.9), affecting >1.27 billion (1.19–1.35); >70% increases are projected in six regions (including north Africa and Middle East, east Asia, central and southern sub-Saharan Africa, central Latin America, and Australasia). No country is projected to experience a decline.

Findings demonstrate that diabetes—overwhelmingly type 2—continues to grow in every country, age group, and sex, with particularly high and rising burdens in Oceania, north Africa and the Middle East, and parts of Latin America and sub-Saharan Africa. The dominant role of high BMI in type 2 diabetes DALYs, and its increasing contribution since 1990, underscore the centrality of global obesity trends driven by changes in food systems, urbanisation, sedentary lifestyles, and socioeconomic factors. The results highlight disparities in sex-specific prevalence and treatment coverage, especially in LMICs, reflecting differences in obesity patterns, biology, and health-system performance. Given projections that nearly half of future increases will be demographic, health systems will need substantial capacity expansion to diagnose, treat, and manage complications. While type 2 diabetes is in many cases preventable or potentially reversible if addressed early, sustained population-level reductions in obesity have proven elusive, suggesting the need for coordinated, multisectoral, long-term policies addressing diet, physical activity, urban design, and access to effective treatments. Early detection and comprehensive care can mitigate downstream complications and broader impacts on cardiovascular and renal disease. The study’s granular, type-specific estimates and forecasts are intended to guide targeted prevention and health-system planning.

This study provides comprehensive, location-, age-, and sex-specific estimates of diabetes prevalence and burden from 1990–2021, quantifies type-specific composition and risk-attributable fractions, and offers forecasts to 2050. It identifies high BMI as the leading and growing driver of type 2 diabetes burden and projects a doubling in the number of people living with diabetes by mid-century. The findings reinforce the urgency of implementing multifaceted strategies to prevent obesity and improve diabetes care, especially in LMICs, alongside investments in early detection and access to effective therapies. Future research should incorporate post–COVID-19 dynamics as data mature, refine forecasting with additional modifiable risk factors, improve classification of diabetes types in younger populations, and deepen investigation into risk factors for type 1 diabetes.

- Data sources: Excluded self-reported diabetes without biochemical validation, which may differ from other global estimates but reduces bias from changing diagnostic and screening practices. - COVID-19: Estimates and forecasts do not include the impact of the COVID-19 pandemic on diabetes burden for 2020–2021; integration is planned as data become available. - Cause-of-death coding: Variability and misclassification in death certificate coding may over- or understate diabetes as the underlying cause; many records did not specify type, requiring redistribution using models informed largely by higher-income settings. - Diagnostic criteria: Many population studies relied on a single abnormal glucose measurement, potentially overestimating prevalence; about half of sources lacked details on blood collection method (capillary vs venous), which can affect glucose readings by up to ~20%. - Scope: Gestational and rarer forms (e.g., monogenic) were not explicitly modelled. - Type assignment in youth: All diabetes under age 15 was assumed to be type 1, though youth-onset type 2 is emerging in some regions; lack of population-based data precluded differentiation. - Risk assessment: Comparative risk framework includes modifiable risks for type 2; type 1 risk factors are not captured. Forecasts emphasized high BMI (type 2) and SDI (type 1); other covariates (e.g., physical activity, smoking) were not included and cohort effects were generally not modelled beyond tobacco. - Generalisation: Redistribution of unspecified deaths and some covariate relationships may not fully generalise across all settings.