Global burden and strength of evidence for 88 risk factors in 204 countries and 811 subnational locations, 1990–2021: a systematic analysis for the Global Burden of Disease Study 2021

Summary Background Understanding the health consequences associated with exposure to risk factors is necessary to inform public health policy and practice. To systematically quantify the contributions of risk factor exposures to specific health outcomes, the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2021 aims to provide comprehensive estimates of exposure levels, relative health risks, and attributable burden of disease for 88 risk factors in 204 countries and territories and 811 subnational locations, from 1990 to 2021. Methods The GBD 2021 risk factor analysis used data from 54 561 total distinct sources to produce epidemiological estimates for 88 risk factors and their associated health outcomes for a total of 631 risk–outcome pairs. Pairs were included on the basis of data-driven determination of a risk–outcome association. Age-sex-location-year-specific estimates were generated at global, regional, and national levels. Our approach followed the comparative risk assessment framework predicated on a causal web of hierarchically organised, potentially combinative, modifiable risks. Relative risks (RRs) of a given outcome occurring as a function of risk factor exposure were estimated separately for each risk–outcome pair, and summary exposure values (SEVs), representing risk-weighted exposure prevalence, and theoretical minimum risk exposure levels (TMRELs) were estimated for each risk factor. These estimates were used to calculate the population attributable fraction (PAF; ie, the proportional change in health risk that would occur if exposure to a risk factor were reduced to the TMREL). The product of PAFs and disease burden associated with a given outcome, measured in disability-adjusted life-years (DALYs), yielded measures of attributable burden (ie, the proportion of total disease burden attributable to a particular risk factor or combination of risk factors). Adjustments for mediation were applied to account for relationships involving risk factors that act indirectly on outcomes via intermediate risks. Attributable burden estimates were stratified by Socio-demographic Index (SDI) quintile and presented as counts, age-standardised rates, and rankings. To complement estimates of RR and attributable burden, newly developed burden of proof risk function (BPRF) methods were applied to yield supplementary, conservative interpretations of risk–outcome associations based on the consistency of underlying evidence, accounting for unexplained heterogeneity between input data from different studies. Estimates reported represent the mean value across 500 draws from the estimate's distribution, with 95% uncertainty intervals (UIs) calculated as the 2·5th and 97·5th percentile values across the draws. Findings Among the specific risk factors analysed for this study, particulate matter air pollution was the leading contributor to the global disease burden in 2021, contributing 8·0% (95% UI 6·7–9·4) of total DALYs, followed by high systolic blood pressure (SBP; 7·8% [6·4–9·2]), smoking (5·7% [4·7–6·8]), low birthweight and short gestation (5·6% [4·8–6·3]), and high fasting plasma glucose (FPG; 5·4% [4·8–6·0]). For younger demographics (ie, those aged 0–4 years and 5–14 years), risks such as low birthweight and short gestation and unsafe water, sanitation, and handwashing (WaSH) were among the leading risk factors, while for older age groups, metabolic risks such as high SBP, high body-mass index (BMI), high FPG, and high LDL cholesterol had a greater impact. From 2000 to 2021, there was an observable shift in global health challenges, marked by a decline in the number of all-age DALYs broadly attributable to behavioural risks (decrease of 20·7% [13·9–27·7]) and environmental and occupational risks (decrease of 22·0% [15·5–28·8]), coupled with a 49·4% (42·3–56·9) increase in DALYs attributable to metabolic risks, all reflecting ageing populations and changing lifestyles on a global scale. Age-standardised global DALY rates attributable to high BMI and high FPG rose considerably (15·7% [9·9–21·7] for high BMI and 7·9% [3·3–12·9] for high FPG) over this period, with exposure to these risks increasing annually at rates of 1·8% (1·6–1·9) for high BMI and 1·3% (1·1–1·5) for high FPG. By contrast, the global risk-attributable burden and exposure to many other risk factors declined, notably for risks such as child growth failure and unsafe water source, with age-standardised attributable DALYs decreasing by 71·5% (64·4–78·8) for child growth failure and 66·3% (60·2–72·0) for unsafe water source. We separated risk factors into three groups according to trajectory over time: those with a decreasing attributable burden, due largely to declining risk exposure (eg, diet high in trans-fat and household air pollution) but also to proportionally smaller child and youth populations (eg, child and maternal malnutrition); those for which the burden increased moderately in spite of declining risk exposure, due largely to population ageing (eg, smoking); and those for which the burden increased considerably due to both increasing risk exposure and population ageing (eg, ambient particulate matter air pollution, high BMI, high FPG, and high SBP). Interpretation Substantial progress has been made in reducing the global disease burden attributable to a range of risk factors, particularly those related to maternal and child health, WaSH, and household air pollution. Maintaining efforts to minimise the impact of these risk factors, especially in low SDI locations, is necessary to sustain progress. Successes in moderating the smoking-related burden by reducing risk exposure highlight the need to advance policies that reduce exposure to other leading risk factors such as ambient particulate matter air pollution and high SBP. Troubling increases in high FPG, high BMI, and other risk factors related to obesity and metabolic syndrome indicate an urgent need to identify and implement interventions. Funding Bill & Melinda Gates Foundation.


Introduction
The COVID-19 pandemic revealed profound health disparities between individuals and across geographies. 1 These differential impacts reflect a combination of multiple contributing risk factors affecting individuals and the varying capacities of health-care systems to protect and treat their populations.To strengthen the ability of health systems to meet future challenges, there is a need to focus on primary prevention. 2,3In this context, and to regain momentum towards meeting UN Sustainable Development Goals, 1,4 identifying and quantifying the impact of key risk factors can help prioritise the use of scarce resources.
6][7] Effective risk-reduction policies and practices are dependent on location-specific and population-specific information about relationships between risk factors and health outcomes, trends in the prevalence of leading risk factors, and the proportion of disease-specific mortality and morbidity that can be attributed to particular risk factors.Rigorous, well-sourced risk factor meta-analyses can highlight areas of public health progress, provide insight into persisting or emerging risks and consequent health challenges, and inform further modelling of plausible risk-factor reduction scenarios-including cost-effectiveness-to galvanise effective risk-reduction policies and practices.To produce these vital risk factor data, the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) has, since 1996, systematically estimated exposure to risk factors, relative health risk by exposure, and attributable disease burden for comprehensive sets of risk factors. 8For a specified set of causes of death and disability, attributable burden metrics are calculated to quantify the proportion of burden-measured in disability-adjusted life-years (DALYs), representing the sum of years of life lost to premature mortality and years lived with disability-that can be attributed to a particular risk factor or combination of risk factors.6][17][18] However, only GBD produces systematic analyses of a comprehensive set of risk factors, identified on the basis of standardised data-driven criteria, in 204 countries and territories worldwide.
Here, we summarise GBD 2021 methods and present estimates of risk factor exposures and their relationships with health outcomes for 88 risk factors and combinations thereof included in the GBD 2021 hierarchical list of risk factors (appendix 1 table S1).Results are presented broadly within the Article and in more detail in appendix 2. Selected results are further accessible online through the Burden of Proof visualisation tool.This manuscript was produced as part of the GBD Collaborator Network and in accordance with the GBD Protocol. 19

GBD
publishes periodic updates, providing comprehensive estimates of risk exposure and riskattributable health loss worldwide using all relevant available data.GBD 2021 estimated relevant metrics for 23 age groups from birth to age 95 years and older; for males, females, and all sexes combined; and for 204 countries and territories grouped into 21 regions and seven super-regions.GBD regions are made up of countries and territories that are geographically close and epidemiologically similar, and regions are grouped into super-regions on the basis of cause of death patterns. 20he seven super-regions are central Europe, eastern Europe, and central Asia; high income; Latin America For the Burden of Proof visualisation tool see https:// vizhub.healthdata.org/burdenof-proof/and the Caribbean; north Africa and the Middle East; south Asia; southeast Asia, east Asia, and Oceania; and sub-Saharan Africa. 21GBD 2021 also includes subnational analyses for 21 countries and territories (see appendix 1 table S4 for the full GBD location hierarchy).Some results are presented stratified by Socio-demographic Index (SDI), a composite measure of lag-distributed income per capita, average years of education, and fertility rates among females younger than 25 years 22 (appendix 1 table S5).

Research in context
Evidence before this study The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) provides regularly updated estimates of risk factor exposure levels, relative health risk by exposure, and proportion of disease burden related to specific diseases or injuries that can be attributed to particular risk factors, categorised broadly into groups of environmental and occupational, behavioural, and metabolic risks.GBD has conducted analyses of risk-attributable burden since 1996, at which time ten risk factors were included in the analysis.GBD 2021 presents age-sex-location-yearspecific estimates for 88 risk factors at global, regional, and national levels from 1990 through 2021.Although several research organisations and initiatives, including the NCD Risk Factor Collaboration (NCD-RisC) in partnership with WHO and the Prospective Urban and Rural Epidemiological (PURE) study, have provided valuable population-level evidence about specific risk factors or groupings in selected populations, GBD stands out for its systematic evaluation of the health effects of a comprehensive selection of risk factors across all countries and territories worldwide.

Added value of this study
GBD 2021 advances previous GBD estimates of risk factor exposure levels, relative risks (RRs), and the risk-attributable burden in several meaningful ways.RR estimates were systematically updated for risk-outcome pairs with burden of proof meta-regression methods that accounted for differences in exposure ranges in different comparison groups by integrating across the risk function and used an ensemble spline method to capture the (potentially non-linear) shape of the risk-outcome relationship from the data rather than imposing log-linearity.For 211 risk-outcome pairs, evidence of association was further quantified with burden of proof risk function (BPRF) analyses, which account for unexplained between-study heterogeneity in the input data, yielding a conservative interpretation of the risk-outcome association.For ease in interpreting and comparing BPRF measures across risk factors, summary risk-outcome scores were computed and mapped onto a star rating system (from one to five stars) summarising the relationships between risks and outcomes.Of the 211 risk-outcome pairs analysed with the BPRF methodology, 80 (37•9%) received a rating of three to five stars, indicating a well established (moderate to very strong) relationship between risk and outcome, based on a conservative interpretation of the available evidence, while 131 (62•1%) received one to two stars, suggesting that existing evidence for a robust relationship is weak.Additionally, mediation methods used to address risk-outcome relationships involving risk factors that act indirectly on outcomes via intermediate risks (eg, an association between low fruit consumption and heart disease mediated through systolic blood pressure [SBP]) were updated and systematised, resulting in a total of 158 mediated risk-outcome relationships.Nitrogen dioxide air pollution was added as a new risk factor, which resulted in the addition of one associated risk-outcome pair: nitrogen dioxide air pollution-asthma.117 additional riskoutcome pairs were incorporated for risk factors already included in the study, based on new evidence, more detailed specification of outcomes, or refinements to mediation factors.Conversely, 25 risk-outcome pairs were excluded from GBD 2021 because they no longer met inclusion criteria.New or updated systematic reviews were conducted, as detailed in appendix 1 (section 2.1.3).Theoretical minimum risk exposure levels (TMRELs) were revised for 19 risk factors.
The GBD 2021 analytical framework for risk factors generated estimates for the period 1990-2021.][11][12][13][14] An international GBD Collaborator Network provides, reviews, and analyses the available data to generate these metrics, with the GBD 2021 round drawing on the expertise of more than 11 000 collaborators in more than 160 countries and territories.In each iteration of GBD, newly available data and improved methods are used to update the full time series of estimates from 1990 through the latest year of analysis.GBD 2021 estimates for the entire 1990-2021 time series therefore supersede all previously published estimates.

GBD risk factor hierarchy
GBD classifies all GBD risk factors into a risk factor hierarchy with four levels, plus an overarching aggregate of all risk factors combined.At Level 1, risk factors are categorised as environmental and occupational, behavioural, and metabolic risks.These Level 1 categories are disaggregated at Level 2 into 20 risk factors or clusters of risk factors (eg, dietary risks and air pollution).At Level 3, nine of the Level 2 risks are further broken down into 42 additional risk factors or clusters of risks; Level 3 also includes the 11 Level 2 risks that are not further disaggregated.At Level 4-the most granular level-five of the Level 3 risks are further disaggregated into 22 additional specific risk factors; Level 4 also includes the 11 Level 2 risks that were not disaggregated at Level 3 and 37 Level 3 risks not further disaggregated at Level 4. This hierarchy allows for evaluation of individual risk factors, such as low birthweight, as well as groups of risk factors that are of policy interest, such as child and maternal malnutrition or behavioural risks.In total, GBD 2021 covers 88 total risks (one aggregation of all risks combined plus three Level 1 risks plus 20 Level 2 risks plus 42 additional Level 3 risks plus 22 additional Level 4 risks), including one Level 3 risk factor being reported in GBD for the first time: nitrogen dioxide, an additional air pollution measure strongly influenced by motor vehicle emissions. 29See appendix 1 (table S1) for the full 2021 GBD risk factor hierarchy, along with appendix 1 (section 6) and the Methods Web Portal for risk factorspecific definitions and modelling details.

Data sources
To generate relative risk (RR) estimates for risk-outcome pairs, GBD synthesises data from primary randomised controlled trials and cohort, pooled cohort, or casecontrol studies that report RRs of mortality or morbidity from a given health outcome as a function of risk exposure, in addition to meta-analyses summarising RRs (appendix 1 section 2.1.3).][32] 3359 distinct data sources from 124 countries were used in the estimation of RRs, 1176 of which were new for GBD 2021, supplementing those previously included in GBD 2019.To estimate mean exposure for each risk factor, systematic literature reviews were conducted to identify risk factor exposure studies published or identified since GBD 2019, and were combined with data from other sources, including household and health examination surveys and censuses, ground-sensing or remote-sensing data, and administrative records.51 337 distinct data sources from 204 countries and territories were used in estimating risk exposure, 14 252 of which were new, in addition to those previously included in GBD 2019.In total, the GBD 2021 risk factor analysis used 54 561 distinct data sources, which includes a small number of sources used to estimate both relative risk and risk exposure.
Available data sources for estimating RRs and exposure varied across risk factors; input data were highly heterogeneous, and quality varied across geography and time.See appendix 1 (section 2.1.3)for systematic review and bias assessment guidelines, and appendix 1 (section 6) for risk factor-specific details about data collection methods, systematic reviews, search strategies, data sources, bias assessment, and citations.The effort to systematically synthesise substantial quantities of heterogeneous data for large numbers of risk-outcome pairs in a comparable manner is ongoing, and protocols for performing systematic reviews and extracting and processing data will continue to be updated and integrated into methods in future GBD rounds.Detailed information on data sources used for risk factor estimation in GBD 2021 is also available online via the GBD 2021 Sources Tool in the Global Health Data Exchange (GHDx).

Risk factor estimation
For GBD 2021, we estimated relationships between 88 risk factors and selected health outcomes-comprising 155 outcomes across risk factors-for a total of 631 riskoutcome pairs analysed.Notably, the present analysis did not formally incorporate or quantify the impact of the COVID-19 pandemic across risk factors or health outcomes due to data limitations.GBD 2021 produced risk-specific estimates of summary exposure value (SEV), RR, population attributable fraction (PAF), riskattributable burden measured in disability-adjusted lifeyears (DALYs; the sum of years of life lost to premature mortality and years lived with disability), 33 and deaths. 14urthermore, a new method was introduced to complement RR estimates: burden of proof risk function (BPRF) analyses that account for unexplained betweenstudy heterogeneity in RR input data and yield an additional, conservative interpretation of the risk-outcome association and its underlying input evidence. 34The methods employed to generate the measures from past GBD rounds closely followed those used for GBD 2019 14 and have been extensively peerreviewed over previous GBD rounds [9][10][11][12][13][14] and concurrently as part of the peer review process for GBD 2021.Here, we provide a methodological overview with an emphasis on the main changes since GBD 2019.A more comprehensive description of the analytical methods for GBD 2021 is provided in appendix 1, with extensive source details for input data available online via the GBD 2021 Sources Tool in the GHDx.Each of these materials was included in the peer review process of the present Article.
Our analysis was based on the comparative risk assessment (CRA) framework (appendix 1 table S2) established to compute risk factor estimates 8,35 and included seven primary inter-related methodological components.The first step entailed estimating effect size by quantifying the RR of the specified health outcome occurring as a function of exposure to the specified risk factor (appendix 1 section 2 step 1).Estimates were generated for risk-outcome pairs already included in GBD 2019 (based on convincing or probable evidence of an association assessed following World Cancer Research Fund methods and criteria 36 ) and new pairs considered candidates for inclusion (based on informed judgements by GBD Collaborators and other subject experts on potential importance to disease burden or policy, in addition to sufficient data and appropriate methods to estimate key metrics) that met inclusionary criteria, described below (appendix 1 section 2.1.1).In our standard analytical process, the primary tool used to estimate RRs was meta-regression in the burden of proof approach, 34,37,38 which was used to synthesise data identified and extracted through systematic reviews conducted for each risk-outcome pair in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework. 39Guidelines about systematic reviews and bias assessment are provided in appendix 1 (section 2.1.3),with risk factorspecific information-including details about data sources, systematic reviews, data extraction, and modelling strategies-provided in appendix 1 (section 6) and in the Methods Web Portal (cited earlier).The burden of proof approach relies on an ensemble spline method to capture the (potentially non-linear) shape of the RR function from the data rather than imposing a log-linear relationship.The approach also incorporates differences in exposure ranges for different comparison groups by integrating across the RR function, tests and adjusts for systematic biases to account for identified heterogeneity across input study designs and characteristics, and trims potentially distorting outliers in the input data. 37ethodological details about splines, knot placement, monotonicity constraints, trimming strategies, and bias adjustment are provided in appendix 1 (section 2.1.4).RR estimates provide the basis for including new risk-outcome pairs in GBD 2021.Inclusion criteria defined by the GBD Scientific Council state that the RR estimate's 95% uncertainty interval (UI), conventionally calculated, without accounting for unexplained betweenstudy heterogeneity, must not cross the null RR value of 1 (ie, the mean RR estimate must be significantly higher [ for harmful risks] or lower [ for protective risks] than 1) for a risk-outcome pair to be included in GBD.On this basis, 118 new risk-outcome pairs were included in GBD 2021, for a total of 631 pairs.To maintain stability in included risk factors and risk-outcome pairs between GBD cycles, exclusion criteria for those pairs already included in GBD 2019 were less stringent; previously included pairs were excluded only if the conventionally calculated 90% UI crossed the null.On this basis, 25 risk-outcome pairs were dropped from GBD 2021.See appendix 1 (table S7) for a list of risk-outcome pairs included in GBD 2021 and details of pairs added or dropped since GBD 2019.New to GBD 2021, the burden of proof approach also evaluated potential publication or reporting bias (appendix 1 section 2.1.7)and quantified unexplained between-study heterogeneity (appendix 1 sections 2.1.5).Between-study heterogeneity was incorporated into estimates of uncertainty and used to generate BPRFs to complement mean RRs derived through our standard analytical process.BPRF metrics (ie, risk-outcome scores and star ratings) provide an additional, conservative interpretation of the riskoutcome effect and the consistency of underlying evidence (detailed below and in appendix 1 section 2.1.6).
The second step consisted of collecting exposure data and estimating the levels and distribution of exposure to each risk factor, primarily using two Bayesian statistical models (spatiotemporal Gaussian process regression [ST-GPR] and disease model meta-regression [DisMod-MR 2.1] 14,33 ) to pool heterogeneous data and to control and adjust for bias (appendix 1 section 2, step 2, and section 6).The third step involved determining theoretical minimum risk exposure levels (TMRELs; the counterfactual level of exposure that would minimise health risk) on the basis of epidemiological evidence 14 (appendix 1 section 2 step 3).In a fourth step, estimates of PAFs, 14 quantifying the proportional change in health that would occur if risk exposure was reduced to the TMREL, were independently computed for each risk-outcome pair with estimates of exposure, RR, and the TMREL (appendix 1 section 2 step 4).Fifth, SEVs, 14 representing the age-specific risk-weighted prevalence of exposure, were calculated for each risk.SEVs are reported on a 0 to 100 scale, where 0 equates to a scenario in which the entire population (in age groups included in the evaluation, eg, those aged 0-27 days for low birthweight) is exposed at the TMREL, and 100 indicates that the entire population is exposed at the maximum risk exposure level (appendix 1 section 2 step 5).Sixth, because some risk factors affect other risks that lie on the physiological pathway to an outcome, mediation factors were estimated and used to correct for PAF overestimation if independence between risk factors was assumed and to compute the burden attributable to combinations of risk factors (appendix 1 section 2 step 6; table S6 presents the full mediation matrix).Finally, estimates of attributable burden (ie, the proportion of disease burden attributable to the risk factor, as quantified by the product of the PAF and the DALYs or deaths associated with the outcome) were calculated for each combination of age group, sex, location, and year (appendix 1 section 2 step 7).The majority of risk-outcome pairs were evaluated with this standard set of analytical processes.For some pairs, other methods were used as dictated by the evidence available for those risks (appendix 1 section 2 step 1 and section 6).For example, non-optimal temperature RR estimation and TMREL identification was conducted through primary analysis of the relationship between temperature and cause-specific mortality. 40For some risk-outcome pairs, PAFs were assumed by definition to be 100% (eg, 100% of diabetes is assumed to be, by definition, related to high fasting plasma glucose [FPG]).For other pairs in which the outcome is specific to a risk factor (eg, mesothelioma and occupational exposure to asbestos), direct PAFs were used, calculated directly from the disease rather than based on an RR estimate generated with the standard set of analytical processes (appendix 2 table 6).
Methodological improvements for estimating risk exposure and risk-attributable burden in the current GBD round focused on standardisation of RR estimation as described above and application of new BPRF methods to generate conservative assessments of risk-outcome relationships and their underlying evidence incorporating between-study heterogeneity; improved specification of the mediation matrix; and re-evaluation of TMRELs with meta-regression or other methods to incorporate new data, resulting in revised TMREL values for 19 risk factors-primarily dietary risks and high systolic blood pressure (SBP), high LDL cholesterol, and high bodymass index (BMI; see appendix 1 table S9 for changes to 2019 TMREL values).Details of these improvements are provided below or in appendix 1 (section 2).

New for GBD 2021 Updates to the mediation matrix
To more fully and accurately account for mediated relationships involving distal risk factors that act indirectly on outcomes via intermediate risks (eg, an association between low fruit consumption and heart disease mediated through SBP), we reviewed and expanded the methods and evidence forming the basis of the GBD mediation matrix (appendix 1 table S6).A set of consistent rule-based inclusionary and exclusionary criteria were formalised and applied.First, a distal risk cannot be mediated by more than 100% through multiple mediators to the same outcome.Second, the full set of distal risks acting through a specific mediator should be applied to every outcome related to that mediator for all distal-mediator-outcome pathways previously included in GBD 2019 and new pathways that rated a three-star relationship or higher in the BPRF star rating system (exceptions to this included some pathways with smoking as a distal risk, and high FPG or high SBP as mediators).Last, outcomes previously absent from the mediation matrix in which a mediator has a direct causal effect in GBD should be added to the matrix (eg, chronic kidney disease due to diabetes was added as a mediated outcome for high FPG).Application of these criteria resulted in the addition of 87 new mediated riskoutcome pairs and the removal of 64 pairs previously in the matrix, resulting in a total of 158 pairs in the 2021 mediation matrix (appendix 1 table S8).See appendix 1 (section 2 step 6) for further details about GBD 2021 mediation methods.Specification of the matrix is ongoing and will be further updated for future GBD rounds.

Burden of proof risk function and star ratings
To complement our standard estimates of risk-outcome relationships, we further applied BPRF methods introduced by Zheng and colleagues 34 that generate alternative metrics combining effect size and consistency of evidence.The motivation behind this methodology is to highlight risk factors for which the currently available data suggest there is either or both a large effect on health outcomes (and potentially high attributable burden) and robust evidence for the effect, in addition to risk factors that show large effects on outcomes but for which the evidence is less consistent, underscoring a need for additional research.For GBD 2021, BPRFs were generated for 211 risk-outcome pairs (ie, for most metabolic risks; all environmental but no occupational risks; and some behavioural risks such as dietary risks and high alcohol use; see appendix 2 table S6) to complement conventional estimates of RR used to calculate PAFs and attributable burden.
The BPRF is related to the mean RR relationship between exposure and health outcome, relying on 95% UIs inclusive of heterogeneity across estimates of effect from individual studies not accounted for by study design covariates (eg, confounding, selection bias, and exposure measurement; appendix 1 section 2.1.5). 41These 95% UIs are used to derive the BPRF, defined for harmful risks as the 5th quantile risk curve closest to null and for protective risks as the 95th quantile risk curve closest to null (RR=1; the function representing a relationship in which a change in risk exposure has no effect on health outcome).The BPRF therefore represents a conservative estimate, consistent with the available evidence, of the change in health outcome at each level of risk exposure.BPRF estimates are used to compute the risk-outcome score, defined as the signed value of the average log BPRF between the 15th and 85th percentiles of risk exposure levels observed across included studies. 34A higher positive risk-outcome score corresponds to either or both a greater average effect size (as represented by RRs) and stronger, more consistent evidence (as reflected in narrower www.thelancet.comVol 403 May 18, 2024 95% UIs), less distorted by spurious confounders or bias, for the specific risk-outcome relationship.For ease of interpretation and comparability across risk-outcome pairs, risk-outcome scores are mapped onto a star rating system (table 1; see appendix 2 table S6 for risk-outcome scores and star ratings for all risk-outcome pairs analysed using BPRF methods).All risk-outcome pairs receiving a one-star to five-star rating are eligible for inclusion in GBD.Application of the BPRF methodology might in some cases lead to 95% UIs including negative attributable burden estimates (eg, lower 95% UI <1) for one-star pairs; this is a result of values for the RR less than 1 in the 95% UIs, a consequence of including between-study heterogeneity in RR estimates.In these cases, the uncertainty includes the possibility of no effect or even protective effects of the exposure on the outcome.Although there might be biological plausibility for the protective effects for some risk factors (eg, metabolic and dietary), this is less likely for others (eg, air pollution and tobacco).In these cases, wide uncertainty suggests poorly understood or weak risk-outcome relationships.We report the full uncertainty distribution for transparency.
The BPRF methodology provides a structured analytical framework applied across the diversity of GBD risk factors to evaluate effect size and consistency across the underlying data.Although our core results are presented for all included risk-outcome pairs, BPRF metrics also allowed us to highlight risk factors with the strongest evidence of disease burden by re-calculating attributable burden estimates for three-star, four-star, and five-star risk-outcome pairs only.For further details on BPRF methods, see appendix 1 (sections 2.1.5and 2.1.6),the paper by Zheng and colleagues 2022, 34 and other publications associated with the methodology. 38,42evelopment of BPRF methods and their application to GBD risk factor analyses are ongoing and will continue to be refined in future GBD rounds.

Presentation of estimates
Risk-attributable burden estimates for 2021 are given as counts and age-standardised rates per 100 000 population, calculated with the GBD standard population structure to account for variation in age structures across populations. 22SEVs are given as age-standardised rates on a 0-100 scale.For changes over time, we present percentage changes during 2000-21 (see appendix 2 table S1 and table S3 for estimates for 1990-2021) and report annualised rates of change (ARCs) as the difference in the natural log of the values at the start and end of the time interval divided by the number of years in the interval.Estimates for all metrics are computed with the mean estimate across 500 draws, and 95% UIs are given as the 2•5th and 97•5th percentiles of that distribution.To reduce computing power and time, the number of computations per process was reduced from 1000 in previous GBD iterations to 500 for GBD 2021 based on simulations that revealed that estimates and uncertainty were not affected by this reduction.

GBD research and reporting practices
GBD 2021 complies with the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) statement (appendix 1 table S3). 43Analyses were completed with Python (version 3.10.4),Stata (version 13.1), and R (version 4.2.1).The statistical code used for GBD estimation is publicly available online.

Role of the funding source
The funder of this study had no role in study design, data collection, data analysis, data interpretation, the writing of the report, or the decision to submit the manuscript for publication.

Overview
Detailed estimates are available in appendix 2, which provides supplementary figures and links to tables in downloadable form through the Global Health Data exchange.All risk-related estimates are also available in searchable and downloadable form through the GBD Results tool and via visual exploration through the online tool GBD Compare and the Burden of Proof visualisation tool.Two-page summaries of results for each risk factor included in the analysis are also available online.

Summary exposure values (SEVs)
Quantifying risk exposure with age-standardised SEVs, which account for both the severity and proportion of the population exposed and are comparable across risks with different patterns of exposure, global Level Unsafe sanitation Other environmental risks Occupational exposure to asbestos Occupational exposure to arsenic Occupational exposure to benzene Occupational exposure to polycyclic aromatic hydrocarbons Occupational exposure to sulphuric acid

Behavioural
Low birthweight and short gestation Diet low in nuts and seeds 42 Childhood sexual abuse and bullying We categorised trends in SEVs between 2000 and 2021 as either increasing substantially (ARCs of >0•5%), declining substantially (ARCs decreasing >0•5%), or as neither (ARC values between -0•5% and 0•5%).At Level 1 of the risk hierarchy, only metabolic risk factors increased in exposure between 2000 and 2021, with an For panel C, ∩ refers to a burden that is attributed to two or all three Level 1 risk factors (ie, the intersecting set of DALYs that belong to both or all three risk factors).Mean estimates in panels A and B are aggregated to include all DALYs attributable exclusively to the specific Level 1 risk factor plus those attributable to the intersection of that risk and one or both of the other Level 1 risk factors (ie, for a single year, the DALY counts combined across the three lines sum to more than the total number of attributable DALYs for that year).DALYs due to COVID-19 were estimated as part of a separate GBD 2021 analysis by the GBD 2021 Diseases and Injuries Collaborators.They have been separated in this figure from the DALYs unattributed to a risk factor because attribution of COVID-19 DALYs to risk exposure was not conducted as part of this analysis.In GBD 2021, 41•4% of total global DALYs-or 44•7% excluding COVID-19 DALYs-were attributable to risk factors (see also appendix 2 figure S4); whereas in GBD 2019, 14 S3) for location-specific and sex-specific agestandardised SEVs and percentage change in SEVs over time for all risk factors over the period 1990-2021.

Risk-attributable burden (DALYs)
For Level 1 risks, the attributable global disease burdenas measured in DALYs reflecting both premature death and years lived in poor health-was highest in 2021 for behavioural risks, followed by metabolic risks, then environmental and occupational risks (figure 1).1), 33 for which there was no risk factor attributable burden estimation included in        S2a-e) and sex (appendix 2 figure S1a-b).Among children aged 0-5 years, low birthweight and short gestation, child growth failure, particulate matter pollution, unsafe water source, and unsafe sanitation were the leading five Level 3 risk factors.There were no metabolic risks in the top ten risk factors for this age group.For children and adolescents aged 5-14 years, iron deficiency, low birthweight and short gestation, and the three risks falling under the category of unsafe WaSH were the leading five risk factors, with only one metabolic risk-kidney dysfunction-in the top ten.Behavioural risk factors had a greater impact in those aged 15-49 years, with high alcohol use, unsafe sex, and smoking among the top five risks.Metabolic risk factors also gained influence, with high BMI, high FPG, high SBP, and high LDL cholesterol all in the top ten risk factors.For those aged 50-69 years, these same metabolic risks dominated the top ten ranking, and dietary risks (diets high in sodium, low in fruits, and low in whole grains) also gained in prominence.For those aged 70 years and older, metabolic risks remained dominant, and lead exposure and low temperature joined particulate matter pollution as environmental risks among the top ten risk factors.
Age-standardised DALY rates for all outcomes combined attributable to all GBD risk-outcome pairs in 2021 varied by geography (appendix 2 figure S3), as did rates attributable to the five leading global Level 2 risk factors in 2021: child and maternal malnutrition, air pollution, high SBP, tobacco, and dietary risks (figure 3A-J).Age-standardised DALY rates for child and maternal malnutrition were highest in the GBD super-regions of sub-Saharan Africa; south Asia; areas of north Africa and the Middle East; and parts of southeast Asia, east Asia, and Oceania (figure 3B).Regionally, low birthweight and short gestation-a component of child and maternal malnutrition-was ranked as the leading Level 3 risk factor based on attributable all-age DALYs in eastern, central, and western sub-Saharan Africa (with a country-level high of 5668•0 [95% UI 4317•1-7321•4] agestandardised DALYs per 100 000 in South Sudan) and the second-leading risk factor in southern sub-Saharan Africa and Oceania (figure 4, appendix 2 table S1).Low birthweight and short gestation was the leading risk in the low SDI group and second-leading risk in the lowmiddle SDI group (figure 4).
The second-leading global Level 2 risk factor-air pollution-contributed attributable age-standardised DALY rates that were highest in south Asia, sub-Saharan Africa, and parts of north Africa and the Middle East (figure 3D).At a regional level, particulate matter pollution was the leading Level   S1).Particulate matter pollution was ranked as the leading Level 3 risk factor in the low-middle SDI group and the second-leading risk factor in both the low SDI and middle SDI groups (figure 4).World maps of the 2021 global burden attributable to the third-leading, fourth-leading, and fifthleading global Level 2 risk factors-high SBP, tobacco, and dietary risks, respectively-are shown in figures 3F,H,J, and country-level estimates of the burden attributable to the associated Level 3 risks are shown in appendix 2 (table S1).
All-age PAFs-used to calculate the attributable burden-for all risk factors combined are provided in appendix 2 (table S4).Detailed information on estimates related to the attributable burden-PAFs, DALYs, and deaths-for each risk factor and outcome, across geography and time are provided in appendix 2 (table S1).
Over the period 2000-2021, all-age DALY counts attributable to behavioural risks declined by 20  2).These increases stand in contrast to decreases in age-standardised riskattributable global DALY rates over the same time period

Unsafe water source Unsafe sanitation
No access to handwashing facility

Unsafe sex
High systolic blood pressure Iron deficiency High fasting plasma glucose

Unsafe sex
Low birthweight and short gestation

High body-mass index
High systolic blood pressure

High systolic blood pressure
High systolic blood pressure High systolic blood pressure High systolic blood pressure

High systolic blood pressure
High systolic blood pressure

High systolic blood pressure
High systolic blood pressure High systolic blood pressure

High systolic blood pressure
High systolic blood pressure

High systolic blood pressure
High

Low bone mineral density
High body-mass index

Smoking
Kidney dysfunction High alcohol use High LDL cholesterol Diet high in sodium Diet low in fruits

High LDL cholesterol
High   Risk-deleted DALY rates are DALY rates after removing the effect of a risk factor or combination of risk factors on overall rates.They are calculated as the overall DALY rate multiplied by one minus the PAF for the risk or set of risks; this isolates the underlying changes in DALY rates unattributable to risk factors.Broadly grouped into three categories, category I risk factors are those for which the risk-attributable burden declined due in large part to decreased risk exposure, but in some cases also due to proportional declines in young populations due to population ageing.Category II risk factors are those for which the risk-attributable burden increased moderately despite decreased risk factor exposure, due largely to population ageing.Category III risk factors are those for which the risk-attributable burden increased considerably, due to both increased risk factor exposure and population ageing.DALY=disability-adjusted life-year.PAF=population attributable fraction.
for all other leading 25 Level 3 risk factors (figure 2 S1).Decomposition analysis quantifying the drivers underlying the change in attributable burden indicates that exposure to household air pollution decreased over time but exposure to ambient particulate matter pollution increased, and further shows that population ageing played a larger role in increasing the burden attributable to ambient particulate matter pollution (figure 5; details on decomposition methods are provided in appendix 1 section 3).Decomposing time trends of the number of global DALYs attributable to Level 4 risk factors revealed three main groups of risks (figure 5).The first group (category I) generally includes risk factors for which the global attributable burden decreased between 2000 and 2021 due in large part to declines in risk exposure (eg, occupational injuries, diet high in trans fatty acids, household air pollution from solid fuels, and diet low in fibre), and in some cases these exposure declines were enhanced by the positive effects of population ageing (ie, a given risk has a proportionally greater burden in younger individuals and therefore the risk-attributable burden declines in ageing populations).Category I risk factors in this latter class mainly fall under the broader umbrellas of child and maternal malnutrition and unsafe WaSH.The second group (category II) mainly comprises risk factors for which global attributable DALYs increased moderately between 2000 and 2021, in most cases despite decreasing risk exposure and in almost all cases related to the negative effects of population ageing.This group includes numerous dietary factors (eg, diets low in calcium, low in fruits, and low in vegetables), smoking, and risks related to environmental or occupational factors (eg, occupational exposure to asbestos and lead exposure).
The third group (category III) includes risk factors for which the attributable global disease burden rose considerably over the study period due to both increasing risk exposure and the effects of population ageing.This group comprises many metabolic risks (eg, high BMI, high FPG, low bone mineral density, kidney dysfunction, and high SBP), occupational risks (occupational exposure to trichloroethylene, diesel engine exhaust, chromium, cadmium, and others), and some dietary risks (diets high in sugar-sweetened beverages, high in red meat, and low in milk).For nearly all risk factors across all three groups, risk-deleted global DALY rates (ie, change in DALYs not attributable to a risk factor included in our assessment, to population growth, or to ageing) exerted a downward effect on trends in risk-attributable global DALY counts between 2000 and 2021.
Trends in the risk-attributable burden varied both by SDI level and by location.Figure 6 provides a high-level overview of ARCs between 2000 and 2021 in agestandardised DALY rates attributable to Level 1 risk factors, by SDI.For behavioural risks, the attributable burden declined at a lower rate in higher SDI areas than in lower SDI areas over this period.Conversely, the attributable burden for metabolic risks generally declined at a higher rate with increasing SDI.There was minimal association between SDI and all environmental and occupational risks combined.Disaggregating to a more specified level of risk factors, figure 4 presents ARCs between 2000 and 2021 for age-standardised DALYs attributable to the ten leading Level 3 risks, stratified by SDI and GBD region.Age-standardised rates of attributable DALYs declined at high rates for behavioural risk factors related to child and maternal malnutrition in the low and middle SDI quintiles; that is, child growth failure showed declines in ARCs of more than 3•2% and low birthweight and short gestation showed declines in ARCs of 1•6-3•2%.The DALY burden attributable to unsafe sex, another behavioural risk factor, also declined at high rates (ARC decrease of >3•2%) in the low SDI quintile.Although higher SDI quintiles had higher rates of decline (ARC decrease of 1•6-3•2%) in the burden attributable to the behavioural risk factor of smoking than did lower SDI quintiles, DALY rates attributable to drug use notably increased (ARC increase of >1•2%) in the high SDI quintile.With respect to metabolic risk factors, figure 4 shows that although the burden attributable to high SBP and high LDL cholesterol decreased over time across SDI strata, rates of decline were highest in the high-middle and high SDI quintiles (ARC decreases of 2•2-3•2%).Notably, the burden attributable to high BMI increased in the low-middle, middle, and high SDI quintiles, showing the highest rates of increase (ARC increase of >1•2%) in the low-middle and middle SDI quintiles, and decreased in the high-middle SDI quintile.For high FPG, attributable DALY rates increased across almost all SDI strata, with the highest rates of increase (ARC increase of 0•7-1•2%) in the low-middle SDI quintile.However, the high-middle SDI quintile was again an exception, showing slight declines between 2000 and 2021 in the burden attributable to high FPG.With respect to Level 3 environmental risk factors, the burden attributable to particulate matter pollution decreased between 2000 and 2021, with the highest rates of decline in the high, high-middle, and middle SDI quintiles (ARC decrease of >3•2%), and the burden attributable to all three Level 3 risks related to unsafe WaSH decreased at high rates (ARC decline of >3•2%) in the low SDI quintile.
Between 2000 and 2021, reductions in agestandardised DALY rates attributable to child and maternal malnu trition, the leading level 2 risk factor in 2021, were broadly evident in Latin America and the Caribbean, central Asia, and parts of east Asia and southeast Asia (figure 3A,B).Risk-attributable DALY rates also decreased in parts of sub-Saharan Africa, but child and maternal malnutrition remained a challenge in 2021 throughout this super-region, in addition to northern India and countries such as Pakistan, Afghanistan, Yemen, and Papua New Guinea.Locationspecific reductions in Level 3 risks related to the child and maternal disease burden between 2000 and 2021 are reflected in decreases in age-standardised DALY rates attributable to low birthweight and short gestation, with high rates of decline (ARC decrease of >3•2%) in the regions of Andean Latin America and north Africa and the Middle East.Similarly, rates of decline in DALYs attributable to child growth failure were highest in the Caribbean; Oceania; and in central, eastern, southern, and western regions of sub-Saharan Africa (figure 4).
Declines in age-standardised DALY rates attributable to air pollution, the second-ranked Level 2 risk factor, between 2000 and 2021 can be seen in Andean Latin America, central Asia, eastern Europe, parts of east Asia, southeast Asia, and sub-Saharan Africa, but challenges remain in south Asia, sub-Saharan Africa, and parts of east Asia (figure 3C,D).Rates of decline between 2000 and 2021 for the age-standardised burden attributable to Level 3 particulate matter pollution were especially high (ARC decrease of >3•2%) in the regions of central Asia, central Europe, eastern Europe, western Europe, Andean Latin America, central and tropical Latin America, and east Asia.The burden attributable to particulate matter pollution declined between 2000 and 2021 in other regions, including south Asia, but at a slower rate (figure 4).Declines over time in the burden attributable to high SBP, the third-ranked Level 2 risk factor, appear relatively limited in geographical scope, occurring primarily in central Asia, eastern Europe, and tropical Latin America (figure 3E,F).Although age-standardised DALY rates attributable to high SBP declined in 20 of 21 GBD regions (rising slightly in Andean Latin America; figure 4), high rates of decline in high SBP (ARC declines of >3•2%) occurred only in Australasia, high-income Asia Pacific, and western Europe, with increases or low declines elsewhere.As with SBP, the burden attributable to high LDL cholesterol declined in many regions between 2000 and 2021, with the highest rates of decline in Australasia and western Europe.Conversely, the burden attributable to high BMI and high FPG rose over this period in many regions, with the highest increases for high BMI (ARC increase of >1•2%) in south Asia, east Asia, southeast Asia, and central sub-Saharan Africa.The highest increases in the burden attributable to high FPG (ARC increase of >1•2%) occurred in north Africa and the Middle East, but increases were also high in other regions.
The burden attributable to tobacco, the fourth-ranked Level 2 risk factor, decreased between 2000 and 2021 throughout the world, notably in eastern Europe, central Asia, and east Asia, but remains a challenge in those and other areas (figure 3G,H).Decreases over the 2000-21 period in Level 3 risk-attributable DALYs to smoking-an important component of tobacco risk-were likewise high (ARC declines of 2•2-3•2%) in eastern Europe, central Asia, and east Asia, in addition to Australasia, high-income Asia Pacific, and western Europe.In east Asia, declines in the burden attributable to secondhand smoke were higher than the declining burden attributable to smoking.
The burden attributable to dietary risks, the fifthleading Level 2 risk factor, declined in large areas of central Asia and eastern Europe between 2000 and 2021 (figure 3I,J), but the dietary burden remains concerning in these areas and parts of central Europe, and in countries such as Afghanistan and Yemen.High rates of decline (ARC decrease of >2•2%) in age-standardised risk-attributable DALY rates for many of the 15 diverse Level 3 dietary risk factors, from diet low in whole grains to diet high in sodium (figure 4) were seen in central and eastern Europe, in addition to Australasia, southern Latin America, and western Europe.High rates of decline (ARC decrease of >2•2%) in the burden attributable to diet high in sodium occurred in central Asia, eastern Europe, and high-income Asia Pacific.Other classes of risk factors related to unsafe WaSH and to unsafe sex also declined at high rates (ARC decrease of >3•2%) throughout sub-Saharan Africa (figure 4).
Detailed information on the change over time in estimates related to the attributable burden-DALYs and deaths-for each risk factor and outcome, across geography, are shown in appendix 2 (table S1).

Burden of proof risk function assessments of effect size and strength of evidence
To complement conventional estimates of risk, we calculated risk-outcome scores, which quantify the effect size of an association and the strength of evidence for the effect (ie, the extent of between-study heterogeneity), and evaluated the relationship between attributable DALYs and risk-outcome scores.There was a positive relationship between the two values, indicating that risk-outcome pairs contributing the most to the overall attributable burden also had a stronger evidence of a risk-outcome association (figure 7; appendix 2 table S6).Of 211 riskoutcome pairs analysed by BPRF methods for GBD 2021, 12 (5•7%) were identified as having very strong (ie, fivestar) relationships (appendix 2 table S6).The three fivestar pairs responsible for the most risk-attributable DALYs in 2021 were high SBP-ischaemic heart disease (IHD; which contributed >75 million DALYs), 44 high SBP-stroke (which contributed >75 million DALYs), and smokinglung cancer 32 (which contributed 25-50 million DALYs).13 (6•1%) risk-outcome pairs were assessed as having strong (ie, four-star) associations and evidence.Five of those 13 risk-outcome pairs each contributed 25-50 million risk-attributable DALYs, with high BMItype 2 diabetes generating the highest number of DALYs in the four-star group.55 (26•0%) pairs were estimated as being three-star associations with moderate effect sizes and evidence strength; of those, high LDL cholesterol-IHD and particulate matter pollution-IHD contributed the most (50-75 million) attributable DALYs, and six others each contributed 25-50 million DALYs.131 (62•1%) pairs were found to have weak relationships: 79 (37•4%) with two-star associations and 52 (24•6%) with one-star associations.None of the pairs in either of these groups contributed more than 25 million attributable DALYs.Relatively few risk-outcome pairs with more than 12•5 million attributable DALYs were rated as one-star or two-star pairs (appendix 2 table S6).
To highlight risk factors for which there is a strong and consistent evidence base and those for which there is not a strong or consistent evidence base of robust health effects, according to available data, a secondary analysis was conducted that excluded one-star and twostar risk-outcome pairs.Because most of these pairs do not contribute large numbers of risk-attributable DALYs, the overall effect of excluding them was relatively modest, with the number of risk-attributable DALYs as a proportion of total DALYs (with total DALYs inclusive of estimated COVID-19 DALYs 33 ) decreasing from 41•4% to 37•5% and DALYs unattributed to any risk factors rising from 51•4% to 55•2% when excluding one-star and twostar pairs (appendix 2 figure S4).This analysis found that various Level 3 risk factors in the top 20 for riskattributable DALY counts in 2021, when calculated inclusive of all pairs, dropped considerably in ranking after removing one-star and two-star pairs, suggesting an absence of strong evidence for at least some of the health effects associated with these risk factors (figure 8).These include high BMI, for which attributable DALY counts declined from 128 Estimates related to the attributable burden-PAFs, DALYs, and deaths for each risk factor and outcomethat were made by excluding risk-outcome pairs rated as one star or two stars based on BPRF analysis are shown in appendix 2 (table S2).Maps of age-standardised DALY rates attributable to all risk factors combined, by location, based on datasets including all risk-outcome pairs and on all pairs exclusive of those rated with one or two stars using BPRF analysis are also presented in appendix 2 (figure S3).

Main findings
This study presents comprehensive estimates of the riskattributable burden for 204 countries and territories from 1990 to 2021.At the global level, particulate matter pollution, high SBP, smoking, low birthweight and short gestation, high FPG, and high BMI were the largest

Environmental and occupational risks Behavioural risks Metabolic risks
All risk-outcome pairs All risk-outcome pairs excluding 1 and 2 star-rating risk-outcome pairs metabolic syndrome-particularly high FPG and high BMI-due to a combination of population ageing and increasing risk exposure.

Reducing the risk-attributable burden
Reducing the risk-attributable burden requires understanding the differing impact of specific risk factors, such as those under the broader umbrella of air pollution.The two forms of particulate matter pollutionthe leading Level 3 contributor to the burden in 2021are household air pollution and ambient particulate matter air pollution.Although global SEVs and riskattributable DALYs for household air pollution decreased substantially between 2000 and 2021, they increased for ambient particulate matter pollution and ambient ozone pollution.Consequently, steep declines in the burden attributable to household air pollution in south Asia and China, for example, have been accompanied by increases in ambient particulate matter air pollution and ozone pollution, with some populations in these locations facing the substantial burden attributable to all three risk factors.Although it is tempting to suggest that these trends simply reflect a transfer of attributable burden from one risk factor to another, in the case of residential energy sources, the introduction of coupled policies has shown that success is possible on multiple fronts simultaneously.For example, SEVs decreased in China between 2015 and 2021 for both household (nearly 40%) and ambient particulate matter air pollution (approximately 9%; appendix 2 table S3).Indeed, one approach to improve ambient air pollution in Beijing has been a ban on the use of coal for residential energy in the surrounding region. 45As success with reducing both forms of particulate matter air pollution has been observed in China, there is also emerging evidence suggesting a recent peak in ambient air pollution in India coinciding with success in reducing household air pollution. 46,47Introduction of a new air pollution risk factor in GBD 2021-nitrogen dioxide, a marker for motor vehicle pollution-adds another dimension to the air pollution picture.In contrast to other air pollution risk factors, SEVs for nitrogen dioxide are highest but decreasing in high, high-middle, and middle SDI locations, while they are increasing in low and lowmiddle SDI locations.This pattern reflects the high levels of vehicle use in higher SDI locations relative to low and low-middle SDI locations, while SEV trends in rapidly developing locations reflect the combination of overall increasing vehicle use with differing levels of adoption of lower-emissions vehicles.
The second-leading Level 3 contributor to the riskattributable burden in 2021-high SBP-rose from fourth position in 2000, and along with high FPG and high BMI, shows a concerning trend of substantial burden attributable to key metabolic risks.The burden attributable to high SBP represents a continuing challenge and remains particularly impactful outside of most high-income countries, with reductions in burden limited in geographical scope.High FPG and high BMI stand out as risks for which both exposure and burden have increased considerably in almost all regions of the world, with the burden attributable to high BMI increasing with increasing SDI.Given projected increases in rates for outcomes such as type 2 diabetes 48 and those related to high BMI (eg, musculoskeletal disorders 33,49 ), concerted policy actions addressing obesity and metabolic syndrome should be high priorities.Such actions could include evidence-based prevention efforts, treatment, or upstream socioeconomic policies to reduce underlying DALY rates. 50Balancing risk exposure reduction with other approaches might be more beneficial in cases where a single risk factor contributes to multiple outcomes (eg, low physical activity, high SBP, high BMI, and high FPG), especially when clear and compelling evidence exists of the effectiveness of specific interventions.
Smoking was the third-leading risk factor for all-age disease burden globally and a leading risk factor across most geographies and sociodemographic levels in 2021.Although actions by governmental agencies, multilateral institutions, and non-governmental organisations focused on tobacco control have contributed to a nearly 35% reduction in the age-standardised rate of global DALYs attributable to smoking over the period 2000-21, persistent and sustained action is needed to further reduce the burden of smoking, a highly consequential risk factor that can be addressed with proven interventions such as tobacco control policies, enforcing bans on tobacco advertising and sponsorship, encouraging current smokers to quit, and smoke-free policies. 51ow birthweight and short gestation remained the fourth-leading Level 3 contributor to all-age DALYs in 2021, despite major improvements in the riskattributable burden related to child and maternal malnutrition and health over the study period, including a more than 70% reduction in the rate of age-standardised DALYs attributable to child growth failure; a nearly 35% decrease in the age-standardised DALY rate attributable to low birthweight and short gestation; and declines of more than 65% for risk factors related to unsafe WaSH.This reflects the major role of child and maternal risk factors in the overall global disease burden.In sub-Saharan Africa, south Asia, and parts of southeast Asia, child and maternal malnutrition remained the leading Level 2 risk factor for attributable DALYs in 2021.In these locations, there is a high need for evidencebased, locally relevant policies addressing the many complex factors that affect malnutrition, in order to maintain and further accelerate reductions in child and maternal malnutrition exposure. 52First, countries must implement or expand food and nutrition policies that have been developed with the best available evidence to maximise effectiveness. 53][56][57][58] No specific Level 3 dietary risks were among the leading contributors to the burden either globally or in specific SDI quintiles or super-regions, yet-in aggregate-Level 2 dietary risks were a leading risk factor globally throughout the study period.This pattern was sustained despite broad reductions in the attributable burden due to improvements in eastern and central Europe and central Asia.Among specific leading dietary risk factors such as diets low in fruit, whole grains, and vegetables and diet high in sodium, there was a consistent pattern of increases in attributable DALY counts but an opposing decrease in age-standardised DALY rates, due to population growth and ageing.Specific dietary risks also contribute, via mediation, to multiple metabolic risk factors such as high FPG and high SBP that are leading contributors to the global burden.
By decomposing temporal patterns in risk factor exposure, population growth and ageing, and trends in underlying disease burden, we were able to broadly group risk factors into three categories.Category I includes risks such as maternal and childhood risk factors, household air pollution, and diets high in trans fatty acids for which the demonstrated reductions in riskattributable burden have been substantial but not equitably distributed.There have been especially large decreases in the SEV for trans-fat, highlighting the effectiveness of trans-fat bans that have been implemented in a growing number of locations.For this and other category I risk factors, existing effective actions should be maintained, although the ability to sustain improvements could be challenged by other forces such as economic instability, conflict, absence of trust, and climate change.Category II includes many dietary risk factors for which exposure reduction actions have been successful but not sufficient to overcome demographic trends, particularly in older populations, and more action is required to counter demographic shifts.Category III entails those risk factors for which exposure has continued to increase since 2000, contributing to large increases in the risk-attributable burden.When combined with demographic and disease burden trends, insufficient actions to date to reduce exposure to this category of risk factors portend concern for the future.Category III risk factors include ambient particulate matter air pollution, drug use, and a group of risk factors related to the ongoing epidemics of obesity and metabolic syndrome: metabolic risks (including high BMI, high FPG, and high SBP), low physical activity, and diet high in sugar-sweetened beverages, which have been shown in decomposition analyses to be associated with the highest percentage increase in risk-attributable DALYs between 2000 and 2021.Taxing sugar-sweetened beverages is one strategy used to reduce the associated burden, but other policies such as public health campaigns to increase awareness of health risks might also be appropriate. 59,60 new methodological component introduced in GBD 2021, BPRF analysis, offers an additional lens through which to prioritise actions and brings nuance to GBD estimates.For the first time, we present two distinct views of GBD results: one consistent with previous releases in which all considered risk-outcome pairs have been included, and an alternative view in which those with low risk-outcome scores-indicating weaker evidence or lower effect sizes, or both, based on a conservative interpretation of the evidence-have been excluded (one-star and two-star pairs).This secondary analysis suggests that for most of the leading risk factors with the highest attributable burden-such as particulate matter pollution, high SBP, smoking, high FPG, high LDL cholesterol, kidney dysfunction, child growth failure, and high alcohol use-the supporting evidence of their effects on specific health outcomes is also strong, strengthening the case that continued action is necessary.For some of these risk factors (eg, particulate matter pollution), actions will largely be related to public policy, whereas for others (eg, smoking), a combination of clinical guidance, individual action, and public policy are warranted.More research is needed on risk factors and risk-outcome pairs with a high attributable burden but low risk-outcome scores.Research should focus on onestar, two-star, and three-star pairs with more than 12•5 million risk-attributable DALYs.Research could also focus on risk factors and risk-outcome pairs with low risk-outcome scores but of specific geographical importance, such as child wasting and diarrhoea, as well as particulate matter pollution and lower respiratory infection.BPRF analysis can also be used as a transparent methodology to assess potential expansion of additional risk factors, such as pesticides and other chemical pollutants, to consider for inclusion in future GBD cycles.
The GBD risk factor analysis covers a broad array of risk factors including a complex and interconnected web of distal (eg, environmental and social), proximal (eg, diet and smoking), and biochemical (eg, FPG) risk factors.The BPRF scoring for specific risk-outcome pairs can be used as a standardised method to compare impacts across this diverse range and to prioritise research.Distal risk factors might be more amenable to upstream systemic change, whereas more proximal risks might be targets for individual control and clinical guidance.Prioritisation of policy action should be informed on the basis of all outcomes paired with a risk factor and the underlying DALY rate for the associated outcomes, as described by the attributable burden.For example, a three-star risk-outcome pair (eg, unsafe water and diarrhoea) with a very common outcome is a greater priority for public health than a three-star pair for a less common outcome (eg, diet high in sodium and stomach cancer).Furthermore, application of the precautionary principle would indicate that public health policy should still apply to risk factors with lower star ratings (eg, diet high in sugar-sweetened beverages), especially those with multiple outcomes.Risk factors for which a high proportion of the population is exposed (eg, ambient particulate matter air pollution) could also be a higher priority for public health policy than risk factors for which exposure prevalence is lower (eg, secondhand smoke), while also considering temporal and spatial variation in risk factor exposure.Although our analysis of burden and risk factor prioritisation is primarily relevant to public policy actions, our BPRF analysis and risk-outcome scores could also be useful for clinical guidance and individual behavioural actions to reduce risk factor exposure by highlighting those relationships for which the evidence is strongest. 34

Climate change: direct and indirect impacts
GBD 2021 partially captures the disease burden attributable to climate change via the high temperature risk factor, the most direct pathway through which climate change can affect health. 61,62SEVs and DALYs attributable to high temperature increased modestly between 2000 and 2021.High temperature accounted for 0•5% of total global DALYs in 2021, contributing to 2•5% or more total DALYs in Saudi Arabia, Oman, Mauritania, and Iraq. 33Global attributable deaths due to high temperature in 2021 reached nearly 450 000 (appendix 2 table S1).The impact of high temperatures was modest in 2021 compared to that of other risk factors, including low temperatures, which had nearly double the number of attributable DALYs.While SEVs for high temperature increased between 2000 and 2021, the decrease in low temperature SEVs correspondingly decreased but to a much lesser degree.Beyond the direct impacts of temperature, climate change has been viewed as a crucial public health challenge and potentially presents an opportunity for health improvements if mitigation and adaptation actions are taken. 63Indirect effects of climate change, whether mediated by other GBD risk factors or via other pathways, have not yet been included in the GBD methodology, in part due to challenges associated with attribution of changes in risk factor exposure (eg, increases in ambient ozone air pollution and wildfires) or specific outcomes (eg, malaria and dengue), given their multifactorial nature, especially in the context of sociodemographic and health trends.For example, existing analyses of malaria suggest that climate-driven increases are likely to be restricted in geographical scope with relatively small impacts on burden. 64Furthermore, quantitative attribution of extreme weather events to climate change remains challenging, but if future warming leads to more frequent or severe events, or both, direct impacts on disease burden (eg, deaths due to drowning during floods) can be expected, whereas indirect effects such as reductions in safe drinking water and sanitation could be compromised.Additionally, climate change has disrupted and will continue to disrupt agricultural production. 65Although global undernutrition continues to decrease and this trend is unlikely to reverse, specific locations strongly affected by warm temperatures and large increases in extreme weather events (droughts or floods) could experience increases in wasting, but not stunting. 58The magnitude of future impacts on disease burden due to the direct effects of temperature, as well as those related to food security, extreme weather events, and rise in sea levels must also be considered in the context of population growth in areas most stressed by climate change and the extent to which populations may out-migrate from such areas.Along with climate change itself, actions taken to mitigate the emissions contributing to climate change could also indirectly affect disease burden and are the subject of substantial research.For example, actions to reduce fossil fuel combustion for energy and transportation are expected to lead to decreases in air pollution that would consequently decrease the attributable burden. 66Efforts to shift populations to more climate-friendly diets could also lead to reductions in the disease burden attributable to specific dietary factors such as diets low in whole grains and legumes. 67

Limitations
The GBD 2021 risk factor findings are limited by several considerations, including the omission of various potentially consequential risk factors and covariates.Importantly, the impact of the COVID-19 pandemic was not formally incorporated or quantified across risk factors or health outcomes, although COVID-19-related effects were included in the analyses for some riskoutcome pairs involving mental health, flu, pertussis, and malaria outcomes.As more information becomes available, future iterations of GBD could quantify risk factors for the COVID-19 burden, such as high BMI. 68urthermore, drug use as well as stress, anxiety, depression, and other mental health conditions correlated with drug use increased sharply during the pandemic, 33 but these changes have not yet been fully captured in the available data.Importantly, although there was no risk attribution for COVID-19, the overall numbers of deaths and non-fatal outcomes for risk attribution were smaller in 2021 than they would have been in the absence of COVID-19, as COVID-19 is likely to have accounted for a proportion of deaths that would have occurred due to other outcomes.
0][71] An additional limitation is that, although our methods to estimate RRs provide a standardised mechanism to code and test for bias, this functionality is constrained to the extent that differences across studylevel characteristics are not always fully known or accurately described in individual studies.Moreover, more nuanced bias coding (eg, using dummy variables to expand to multiple categories) is possible in the current methodology, but is dependent on the availability of sufficient data in the input studies.Standardisation of our bias adjustment processes is ongoing and will be updated for future GBD rounds.While we applied new BPRF methods to 211 risk-outcome pairs-a strength of this study-we were unable to apply this analysis to all applicable pairs for GBD 2021.The BPRF work is ongoing, and the methods will be extended to additional risk-outcome pairs in the future.For one-star and two-star risk-outcome pairs, the evidence base is less consistent across studies and is likely to change in response to ongoing research.
To assess the joint effects of risk factors, our analyses were adjusted to account for the assumption that RRs are multiplicative.Known pathways in which one risk factor (eg, diet low in fruits) was mediated through another risk factor (eg, diet low in fibre) were incorporated into the estimation process.We computed non-mediated RRs and then assumed that non-mediated RRs are multiplicative to avoid overestimation of joint effects.One limitation of this approach is that it does not capture the possibility that some combinations of RRs might be supermultiplicative or sub-multiplicative.Given the centrality of nutrition in public policy discourses, as well as the large evidence base that diet-based interventions can produce positive health outcomes, more detailed work is needed to strengthen the scientific understandings of mediation.On estimating TMREL, we were confronted with several limitations.We generally assumed that the TMREL is 0 for harmful risks with monotonically increasing risk functions.However, for protective risks such as fruit or whole grain intake, selecting the minimum risk level of exposure required more careful analysis because extrapolating the risk function outside the range of where the available literature supports the protective effect could lead to both exaggerated estimates of attributable burden and implausible levels of consumption.We therefore set the TMREL for protective risks to be equal to the 85th percentile of exposure in the available cohorts and trials.A population-level study on red meat consumption is an example of how these proposed improvements and modifications to TMREL estimations improve outputs across the risk factor estimation process. 31ast, we faced challenges in achieving the generalisability of estimated risk-outcome relationships across time and place.We assumed in most cases that RRs as a function of exposure are universal; based on this assumption, RR functions apply to all locations and time periods.However, we did not make this assumption for all risk functions.Temperature is one exception.The risk functions of temperature depend on the annual mean temperature.Additionally, we did not assume RR functions for all locations and time periods for high BMI and breast cancer because of known differences between Asian and non-Asian populations. 72GBD methods require clear and substantive evidence of significant differences in the RR for different subgroups; based on such rules, few cases met this standard.We continue to assess the evidence of the RRs for different populations and will identify and incorporate more location-specific or sub-group RRs.Indeed, cases where risk-outcome relationships indicate substantial between-study heterogeneity suggest a need for further evaluation of the sources of this heterogeneity (eg, location or sub-group).

Future directions
This analysis identifies a group of specific risk factors for which there is consistent evidence of strong riskoutcome relationships, which currently contribute considerably to the disease burden across all levels of SDI, and for which the risk exposure is either increasing or the declines are insufficient to reduce the attributable burden in the face of growing and ageing populations.These results therefore provide ample rationale for accelerated policy action on these risk factors.However, given that our analysis attributes 41•4% of total global DALYs to risk-included factors, a substantial proportion of currently unattributed disease burden remains.In addition to our eventual expansion of the BPRF methodology to all relevant risk-outcome pairs, future iterations of GBD will need to expand the scope of risk factors, especially for specific outcomes that are large and growing contributors to disease burden.For example, musculoskeletal disorders account for 5•6% of global DALYs, 33 but only 20•5% of this burden is attributable to risk factors currently included in GBD.Similarly, mental disorders are responsible for 5•4% of the global burden, but only 8•0% of mental disorders are attributable to risk factors. 33dding additional risk factors for these and other outcomes will require sufficient information on risk factor exposure and sufficiently strong evidence of a riskoutcome relationship.The BPRF methodology introduced in this GBD cycle provides a transparent and efficient approach for evidence scoring.Advances in artificial intelligence for summarising the literature and extracting relevant information to feed into this methodology could also accelerate the evaluation of new risk factors.Although many of the risk factors included here could mediate the impact of social determinants on disease burden, the inclusion of more distal social determinants of health as risk factors in GBD needs further development.For example, low educational attainment can be a strong determinant of health, at a level of attributable burden similar to the impacts of diet, physical activity, smoking, or alcohol. 73Accumulating data also suggest the plausibility of quantifying DALYs attributable to genetic risk factors for certain diseases, including major causes such as ischaemic heart disease, type 2 diabetes, and chronic obstructive pulmonary disease. 74urthermore, greater attention is now being placed on commercial determinants of health such as tobacco and alcohol companies, the fossil fuel industry, and producers of ultra-processed foods, all of which can affect risk factor exposure and which suggest additional aggregation approaches are needed. 75Inclusion of distal risk factors such as education or commercial determinants in future iterations of GBD is likely to provide additional guidance to policy makers, including those sectors outside of health.However, inclusion of additional risk factors and especially such upstream factors will be challenging given the need for detailed understanding of complex mediation relationships.Numerous other individual risk factors such as sleep-related disorders, stress, and exposure to UV radiation, environmental noise, and heavy metals have been considered but are not yet included in the analysis.Multimorbidity, particularly in older age groups in whom health effects from exposure to multiple risk factors are more likely to occur, is another important factor to consider in future iterations of GBD.

Conclusion
Attribution of disease burden to risk factors can help guide prioritisation of actions.Considering both the overall contribution to disease burden, trends in attributable burden, and the strength of evidence relating risk factor exposure to specific outcomes, we identified a highly consistent group of risk factors for which actions have been insufficient.Ambient particulate matter air pollution, high SBP, smoking, and high FPG are not only among the five leading risk factors globally but are also in the top three ranking risk factors for nearly all levels of SDI, suggesting a need for renewed and increased attention to exposure reduction.Low birthweight and short gestation is the leading risk factor at the lowest level of SDI and requires continued action to extend the reductions in attributable burden observed since 2000.By contrast, high BMI is the leading risk factor at the highest level of SDI.With increasing risk exposurecompounded by interactions with metabolic risk factors such as high FPG, high SBP, low physical activity, and diet high in sugar-sweetened beverages-there is an urgent need for interventions focused on obesity and metabolic syndrome.More generally, among the ten leading risk factors globally-each contributing to at least 2•5% of total global DALYs-all except child growth failure and low birthweight and short gestation have shown risk factor exposure trends from 2000 to 2021 that indicate inadequate action has been taken to reduce the attributable burden.Furthermore, all risk factors contributing the most to the attributable burden are supported by strong evidence of their association with specific outcomes, although in the case of high BMI the magnitude of the attributable burden changes when the strength of evidence for individual risk-outcome pairs is considered, suggesting a high-priority need for additional research on BMI-outcome relationships.Future iterations of GBD will continue to track levels and trends in risk factors and their attributable burden, assimilate a growing literature into the burden of proof framework, and incorporate additional risk factors to aid in prioritisation of actions to reduce the disease burden.reports grants or contracts from Gilead and GSK; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Gilead and GSK; support for attending meetings or travel, or both, from Gilead; receipt of equipment, materials, support for attending meetings or travel, or both, from the European Renal Association; leadership or fiduciary roles in board, society, committee, or advocacy groups, unpaid with the Advocacy Group of the International Society of Nephrology and the Western Europe Regional Board of the International Society of Nephrology; and other financial or non-financial support from Scientific-Tools.Org; outside the submitted work.A Biswas reports consulting fees from Lupin Pharmaceuticals (India), Alkem Laboratories (India), and Intas Pharmaceuticals (India); payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Roche Diagnostics (India); outside the submitted work.C S Brown reports other financial or non-financial interests from ad-hoc one-off market research advisory, on a variety of infection topics, all anonymously conducted via market research companies with no direct communication nor any knowledge of any pharmaceutical companies or products (and none specifically related to global burden of disease); outside the submitted work.M Carvalho reports other financial or non-financial interests from LAQV/REQUIMTE, University of Porto (Porto, Portugal) and acknowledges support from FCT/MCTES under the scope of the project UIDP/50006/2020 (DOI 10.54499/UIDP/50006/2020); outside the submitted work.A L Catapano reports grants or contracts from Amryt Pharma, Menarini, and Ultragenyx; consulting fees from Eli Lilly, Menarini Ricerche, and Sanofi; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Amarin, Amgen, Amryt Pharma, AstraZeneca, Daiichi Sankyo, Esperion, Ionis Pharmaceuticals, Medscaper, Menarini, Merck, Novartis, NovoNordisk, PeerVoice, Pfizer, Recordati, Regeneron, Sandoz, Sanofi, The Corpus, Ultragenyx, and Viatris; participation on a data safety monitoring board or advisory board with Amarin, Amgen, Amryt Pharma, AstraZeneca, Daiichi Sankyo, Esperion, Ionis Pharmaceuticals, Medscaper, Menarini, Merck, Novartis, NovoNordisk, PeerVoice, Pfizer, Recordati, Regeneron, Sandoz, Sanofi, The Corpus, Ultragenyx, and Viatris; outside the submitted work.C R Cederroth reports support for attending meetings or travel, or both, from the Mérida Institute of Technology to present at the 18th international symposium from the AMCAOF at Mérida (Mexico), March 6-9, 2024; leadership or fiduciary roles in board, society, committee, or advocacy groups, paid or unpaid with the Professional Advisers Committee from the Tinnitus UK as a member, and with the Scientific Advisory Committee of the American Tinnitus Association as a member; outside the submitted work.receipt of equipment, materials, drugs, medical writing, gifts or other services from Allergan-AbbVie and NovoNordisk; outside the submitted work.P S Sachdev reports grants or contracts from the National Health and Medical Research Council of Australia and the NIH; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Alkem Labs; participation on a data safety monitoring board or advisory board with Biogen Australia Medical Advisory Committee in 2020 and the 2021 Roche Australia Medical Advisory Committee in 2022; leadership or fiduciary roles in other board, society, committee or advocacy groups, unpaid with VASCOG Society on the executive committee and the World Psychiatric Association on the planning committee; outside the submitted work.Y L Samodra reports grants or contracts from Taipei Medical University; leadership or fiduciary roles in other board, society, committee, or advocacy groups, paid or unpaid as co-founder of Benang Merah Research Center; all outside the submitted work.J Sanabria reports support for attending meetings or travel, or both, from the Continuing Medical Education section of the University of Marshall School of Medicine; one patent issued and one patent pending; participation on a data safety monitoring board or advisory board with the Marshall University Department of Surgery; leadership or fiduciary roles in other board, society, committee, or advocacy groups, paid or unpaid, with the American Society of Transplant Surgeons, Society of Surgical Oncology, American Board of Surgery, Americas Hepato-Pancreato-Biliary Association, and International Hepato-Pancreato Biliary Association; all outside the submitted work.A E Schutte reports grants or contracts from the National Health and Medical Research Council of Australia (Investigator Grant); consulting feeds from Abbott, Medtronic, Servier, and Skylabs; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Abbott, Medtronic, Servier, Skylabs, Aktiia, Sanofi, Omron, and Novartis; support for attending meetings or travel, or both, from Medtronic and Servier; leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid as Co-Chair of the National Hypertension Taskforce of Australia, Secretary of the Australian Cardiovascular Alliance, Board Member of Hypertension Australia; outside the submitted work.A Sharifan reports leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid from Cochrane; receipt of equipment, materials, drugs, medical writing, gifts, or other services from Elsevier; outside the submitted work.V Sharma acknowledges support from DFSS (MHA)'s research project (DFSS28(1)2019/EMR/6) at the Institute of Forensic Science & Criminology, Panjab University (Chandigarh, India); outside the submitted work.V Shivarov reports one patent pending and one utility model with the Bulgarian Patent Office; stock or stock options from RSUs with ICONplc; and a salary from ICONplc; outside the submitted work.S Shrestha reports support from the School of Pharmacy, Monash University Malaysia and the Graduate Research Merit Scholarship; outside the submitted work.C R Simpson reports grants or contracts from the Health Research Council of New Zealand, Ministry of Health (New Zealand), Ministry of Business, Innovation, and Employment (New Zealand), Chief Scientist Office (UK), and MRC (UK); leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid as Data Ethics Advisory Group Chair for the New Zealand Government; outside the submitted work.J A Singh reports consulting fees from Schipher, Crealta/Horizon, Medisys, Fidia, PK Med, Two Labs, ANI Pharmaceuticals/Exeltis USA, Adept Field Solutions, Clinical Care options, ClearView Healthcare Partners, Putnam Associates, Focus Forward, Navigant Consulting, Spherix, MedIQ, Jupiter Life Science, UBM, Trio Health, Medscape, WebMD, and Practice Point Communications; and the NIH and the American College of Rheumatology; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events as a member of the speaker's bureau of Simply Speaking; support for attending meetings or travel, or both, as a past steering committee member of OMERACT; participation on a data safety monitoring board or advisory board with the US Food and Drug Administration (FDA) Arthritis Advisory Committee; leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid as a past steering committee member of the OMERACT, an international organization that develops measures for clinical trials and receives arm's length funding from 12 pharmaceutical companies, Chair of the Veterans Affairs Rheumatology Field Advisory Committee, and editor/ Director of the UAB Cochrane Musculoskeletal Group Satellite Center on Network Meta-analysis; stock or stock options in Atai Life Sciences, Kintara Therapeutics, Intelligent Biosolutions, Acumen Pharmaceuticals, TPT Global Tech, Vaxart Pharmaceuticals, Aytu BioPharma, Adaptimmune Therapeutics, GeoVax Labs, Pieris Pharmaceuticals, Enzolytics, Seres Therapeutics, Tonix Pharmaceuticals Holding, Aebona Pharmaceuticals, and Charlotte's Web Holdings, and previously owned stock options in Amarin, Viking, and Moderna Pharmaceuticals; outside the submitted work.S T Skou reports grants or contracts from the European Research Council, European Union's Horizon 2020 research innovation programme, Region Zealand; Royalties or licences from Munksgaard and TrustMe-Ed; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Nestlé Health Science; other financial support as a co-founder of GLA:D; outside the submitted work.R Somayaji reports grants or contracts through clinical research funding from the Canadian Institutes of Health Research, Cystic Fibrosis Foundation, Vertex Pharmaceuticals, and the University of Calgary; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events from educational events from Vertex Pharmaceuticals; participation on a data safety monitoring board or advisory board with Oncovir and the Cystic Fibrosis Foundation; leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid, with the Canadian Pressure Injury Advisory Panel; all outside the submitted work.D J Stein reports personal fees from Discovery Vitality, Johnson & Johnson, Kanna, L'Oreal, Lundbeck, Orion, Sanofi, Servier, Takeda, and Vistagen; outside the submitted work.J H V Ticaolu reports leadership or fiduciary roles in other board, society, committee, or advocacy group, paid or unpaid, with the Benang Merah Research Center as co-founder, outside the submitted work.F Topouzis reports grants or contracts from Thea, Omikron, Pfizer, Alcon, AbbVie, and Bayer; consulting fees from Omikron, Thea, and Bausch and Lomb; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Omikron, AbbVie, and Roche; leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid, with the European Glaucoma Society, Greek Glaucoma Society, and World Glaucoma Association; all outside the submitted work.S J Tromans reports grants or contracts from NHS Digital via the Department of Health and Social Care; leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid, with the Neurodevelopmental Psychiatry Special Interest Group and the Royal College of Psychiatrists; all outside the submitted work.E Upadhyay reports patents planned, issued, or pending: "A system and method of reusable filters for anti-pollution mask" (published), "A system and method for electricity generation through crop stubble by using microbial fuel cells" (published), "A system for disposed personal protection equipment (PPE) into biofuel through pyrolysis and method" (published), "A novel herbal pharmaceutical aid for formulation of gel and method thereof" (published), "Herbal drug formulation for treating lung tissue degenerated by particulate matter exposure" (filed), "A method to transform cow dung into the wall paint by using natural materials and composition thereof" (filed); leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid as Joint Secretary of Indian Meteorological Society, Jaipur Chapter (India 49•5) 38•5 (24•7 to 44•5) 34•2 (21•0 to 40•0) 29•8 (17•3 to 35•3) -1•2% (-1•7 to -1•0) -1•2% (-1•7 to -0•9) -1•3% (-1•9 to -0•8) Unsafe water source 44•8 (33•7 to 56•7) 41•0 (28•8 to 55•0) 38•4 (25•5 to 54•0) 35•4 (22•3 to 50 Occupational asthmagens 17•9 (15•5 to 20•9) 18•3 (15•9 to 21•5) 18•1 (15•8 to 20•9) 17•6 (15•5 to 20•2) -0•1% (-0•2 to 0•1) -0•2% (-0•4 to 0•0) -0•2% (-0•6 to 0•1) Occupational particulate matter, gases, and fumes 10•4 (8•4 to 12•8) 10•5 (8•6 to 12•9) 10•4 (8•5 to 12•7) 9•9 (8•2 to 12•0) -0•1% (-0•2 to -0•1) -0•3% (-0•4 to -0•2) -0•4% (-0•5 to -0•3) Occupational noise 10•6 (10•2 to 11•2) 10•8 (10•4 to 11•4) 10•9 (10•5 to 11•5) 10•8 (10•4 to 11•3) 0

Figure 1 :
Figure 1: Global DALYs attributable to Level 1 risk factors, 1990-2021 (A) Global DALY counts attributable to Level 1 risks, 1990 to 2021.(B) Age-standardised DALY rates attributable to Level 1 risks, 1990 to 2021.(C) Global total DALY counts that were unattributed, due to COVID-19, or attributable to Level 1 risk factors, 2021.Mean estimates by Level 1 risk factor in panels A and B are represented by coloured lines; the shading indicates 95% uncertainty intervals.For panel C, ∩ refers to a burden that is attributed to two or all three Level 1 risk factors (ie, the intersecting set of DALYs that belong to both or all three risk factors).Mean estimates in panels A and B are aggregated to include all DALYs attributable exclusively to the specific Level 1 risk factor plus those attributable to the intersection of that risk and one or both of the other Level 1 risk factors (ie, for a single year, the DALY counts combined across the three lines sum to more than the total number of attributable DALYs for that year).DALYs due to COVID-19 were estimated as part of a separate GBD 2021 analysis by the GBD 2021 Diseases and Injuries Collaborators.They have been separated in this figure from the DALYs unattributed to a risk factor because attribution of COVID-19 DALYs to risk exposure was not conducted as part of this analysis.In GBD 2021, 41•4% of total global DALYs-or 44•7% excluding COVID-19 DALYs-were attributable to risk factors (see also appendix 2 figureS4); whereas in GBD 2019,14 47•8% of total global DALYs were attributable to risk factors.DALY=disability-adjusted life-year.Environmental risks=environmental and occupational risks.GBD=Global Burden of Diseases, Injuries, and Risk Factors Study.
Environmental and occupational risks only Metabolic ∩ behavioural risks Behavioural ∩ environmental risks Metabolic ∩ behavioural ∩ environmental risks Metabolic ∩ environmental risks Behavioural risks Metabolic risks Environmental and occupational risks Behavioural risks Metabolic risks Environmental and occupational risks Total DALY counts: unattributed, due to COVID-19, or attributable to Level 1 risk factors, 2021.

Figure 2 :
Figure 2: Leading 25 Level 3 risk factors by attributable DALYs, percentage of total DALYs (2000 and 2021), and percentage change in attributable DALY counts and age-standardised DALY rates from 2000 to 2021 Each column displays the top 25 risks in descending order for the specified year.Risk factors are connected by lines between time periods; solid lines represent an increase or lateral shift in ranking, dashed lines represent a decrease in rank.DALY=disability-adjusted life-year.UI=uncertainty interval.

C
Air pollution, 2000 D Air pollution, 2021 (Figure 3 continues on next page) (Figure 3 continues on next page)

Figure 4 :
Figure 4: Annualised rate of change in age-standardised attributable DALY rates, 2000-21, for the leading ten Level 3 risk factors in 2021, by SDI quintile and GBD region For each region and SDI quintile, Level 3 risk factors are ranked by attributable DALY counts from left (first) to right (tenth).Risk factors are coloured by their annualised rates of change in agestandardised rates of attributable DALYs from 2000 to 2021.DALY=disability-adjusted life-year.GBD=Global Burden of Diseases, Injuries, and Risk Factors Study.SDI=Socio-demographic Index.

Figure 5 :
Figure 5: Percentage change in global DALY counts attributable to Level 4 risk factors from 2000 to 2021, due to population growth, population ageing, changes in risk factor exposure, and changes in risk-deleted DALY rates (A) Category I risk factors.(B) Category II risk factors.(C) Category III risk factors.This decomposition analysis visualises changes in risk-specific attributable DALYs from 2000 to 2021 due to changes in risk exposure, population growth, population age structure, and risk-deleted DALYs.Risk-deleted DALY rates are DALY rates after removing the effect of a risk factor or combination of risk factors on overall rates.They are calculated as the overall DALY rate multiplied by one minus the PAF for the risk or set of risks; this isolates the underlying changes in DALY rates unattributable to risk factors.Broadly grouped into three categories, category I risk factors are those for which the risk-attributable burden declined due in large part to decreased risk exposure, but in some cases also due to proportional declines in young populations due to population ageing.Category II risk factors are those for which the risk-attributable burden increased moderately despite decreased risk factor exposure, due largely to population ageing.Category III risk factors are those for which the risk-attributable burden increased considerably, due to both increased risk factor exposure and population ageing.DALY=disability-adjusted life-year.PAF=population attributable fraction.

Figure 7 :
Figure 7: Global risk-attributable DALYs and risk-outcome score categorised by star rating for all risk-outcome pairs submitted to BPRF analysis, 2021 Risk-outcome score star ratings indicate a conservative assessment of the effect size and strength of evidence for each risk-outcome pair analysed using the BPRF framework.Each point represents a single risk-outcome pair, coloured by Level 1 risk factor category and shaped by type of PAF calculation.Risk-outcome pairs evaluated with direct PAFs and PAF=1 were not submitted to a BPRF analysis and thus did not receive a risk-outcome score or star rating.Risk-outcome pairs associated with more than 15 million attributable DALYs are labelled.BMI=high body-mass index.BPRF=burden of proof risk function.CKD=chronic kidney disease.COPD=chronic obstructive pulmonary disease.DALY=disability-adjusted life-year.Iron=iron deficiency.Diet iron def=dietary iron deficiency.FPG=high fasting plasma glucose.HHD=hypertensive heart disease.IHD=ischaemic heart disease.Larynx C=larynx cancer.LDL=high LDL cholesterol.LRI=lower respiratory infection.Occ injury=occupational injury.PAF=population attributable fraction.PM 2•5 =particulate matter pollution.SBP=high systolic blood pressure.
), Member Secretary of the DSTPURSE Program; outside the submitted work.P Willeit reports consulting fees from Novartis Pharmaceuticals; outside the submitted work.Y Yasufuku reports grants or contracts from Shionogi & Co; outside the submitted work.M Zielińska reports other financial or non-financial interests in AstraZeneca as their employee, outside the submitted work.A Zumla reports support for the present manuscript from the Pan-African Network on Emerging and Re-Emerging Infections (PANDORA-ID-NET) funded by the EDCTPthe EU Horizon 2020 Framework Programme, the UK NIHR Senior Investigator Award, Mahathir Science Award and EU-EDCTP Pascoal Mocumbi Prize Laureate; participation on a data safety monitoring board or advisory board as a member of the Scientific Expert Committee of the EC-EDCTP-Global Health Program; outside the submitted work.All other authors declare no competing interests.Lebanon 2016-17 STEPS survey, implemented by the Ministry of Public Health (Lebanon) with the support of WHO.This paper uses data from the Lesotho 2012 STEPS survey, implemented by the Ministry of Health and Social Welfare (Lesotho) with the support of WHO.This paper uses data from the Liberia 2011 STEPS survey, implemented by the Ministry of Health and Social Welfare (Liberia) with the support of WHO.This paper uses data from the Libya 2009 STEPS survey, implemented by the Secretariat of Health and Environment (Libya) with the support of WHO.This paper uses data from the Madagascar -Antananarivo and Toliara 2005 STEPS survey, implemented by the Ministry of Health and Family Planning (Madagascar) with the support of WHO.This paper uses data from the Malawi 2009 STEPS survey and the Malawi 2017 STEPS survey, implemented by the Ministry of Health (Malawi) with the support of WHO.This paper uses data from the Maldives 2011 STEPS survey, implemented by the Health Protection Agency (Maldives) with the support of WHO.This paper uses data from the Mali 2007 STEPS survey, implemented by the Ministry of Health (Mali) with the support of WHO.This paper uses data from the Marshall Islands 2002 STEPS survey, implemented by the Ministry of Health (Marshall Islands) with the support of WHO.This paper uses data from the Marshall Islands 2017-18 STEPS survey, implemented by the Ministry of Health and Human Services (Marshall Islands) with the support of WHO.This paper uses data from the Mauritania -Nouakchott 2006 STEPS survey, implemented by the Ministry of Health (Mauritania) with the support of WHO.This paper uses data from the Micronesia -Chuuk 2006 STEPS survey, implemented by the Department of Health and Social Affairs (Micronesia), Chuuk Department of Health Services (Micronesia), with the support of WHO.This paper uses data from the Micronesia -Chuuk 2016 STEPS survey, implemented by the Chuuk Department of Health Services (Micronesia), Department of Health and Social Affairs (Micronesia) with the support of WHO.This paper uses data from the Micronesia -Pohnpei 2002 STEPS survey, implemented by the Centre for Physical Activity and Health, University of Sydney (Australia), Department of Health and Social Affairs (Micronesia), Fiji School of Medicine, Micronesia Human Resources Development Center, Pohnpei State Department of Health Services with the support of WHO.This paper uses data from the Micronesia -Pohnpei 2008 STEPS survey, implemented by FSM Department of Health and Social Affairs, Pohnpei State Department of Health Services with the support of WHO.This paper uses data from the Micronesia -Yap 2009 STEPS survey, implemented by the Ministry of Health and Social Affairs (Micronesia) with the support of WHO.This paper uses data from the Micronesia-Kosrae 2009 STEPS survey, implemented by FSM Department of Health and Social Affairs with the support of WHO.This paper uses data from the Moldova 2013 STEPS survey, implemented by the Ministry of Health (Moldova) with the support of WHO.This paper uses data from the Mongolia 2005 STEPS survey, the Mongolia 2019 STEPS survey, and the Mongolia 2013 STEPS survey, implemented by the Ministry of Health (Mongolia) with the support of WHO.This paper uses data from the Morocco 2017 STEPS survey, implemented by the Ministry of Health (Morocco) with the support of WHO.This paper uses data from the Mozambique 2005 STEPS survey, implemented by the Ministry of
47•8% of total global DALYs were attributable to risk factors.DALY=disability-adjusted life-year.Environmental risks=environmental and occupational risks.GBD=Global Burden of Diseases, Injuries, and Risk Factors Study.
Data in parentheses are 95% uncertainty intervals.GBD=Global Burden of Diseases, Injuries, and Risk Factors Study.SEV=summary exposure value.

High systolic blood pressure, 2000 F High systolic blood pressure, 2021
(Figure3continues on next page)

Dietary risks, 2000 J Dietary risks, 2021
] of total DALYs).Of the 25 leading Level 3 risk factors in 2021, 13 (ie, more than half) were behavioural risks.Five of the leading 25 Level 3 risk factors were metabolic risks-all of which were ranked in the top ten, with high BMI ranked sixth, high LDL cholesterol seventh, and kidney dysfunction eighth-and the remaining seven were environmental or occupational risk factors.The contribution of Level 3 risk factors to global DALYs in 2021 varied considerably by age (appendix 2 figure 3 risk factor for the attributable number of all-age DALYs in south Asia (highest in Pakistan, at 6278•3 [95% UI 5312•0-7405•6] age-standardised DALYs per 100 000) and Oceania (highest in the Solomon Islands, at 8813•4 [6860•2-11 307•0] age-standardised DALYs per 100 000) and the second-ranking risk factor in central Asia, southeast Asia, and eastern and central sub-Saharan Africa (figure 4, appendix 2 table

Figure 8: Level 3 risk factors rank ordered by risk-attributable DALYs inclusive of all GBD risk-outcome pairs versus GBD risk-outcome pairs excluding one-star and two-star associations, 2021
Each column displays Level 3 risk factors in descending order by risk-attributable DALYs.Risk factors for which no risk-outcome pairs have a better than two-star association are indicated in the right column with lighter shading and no attributable DALYs.One-star and two-star associations are those that are either or both weakly associated or lacking strong evidence, based on BPRF analysis.Risk factors are connected by lines, with solid lines representing an increase or lateral shift in risk-attributable burden ranking and dashed lines representing a decrease in rank.A number of risk factorsincluding low birthweight and short gestation, low bone mineral density, childhood sexual abuse, intimate partner violence, suboptimal breastfeeding, and all occupational risks-have not yet been submitted to BPRF analysis, and therefore no associated DALYs were removed due to low star rating.BPRF=burden of proof risk function.DALY=disability-adjusted life-year.UI=uncertainty interval.
variation across ages, sexes, and locations.Broadly, the 2000-21 period saw sustained progress in reducing the number of global all-age DALYs attributable to environmental and occupational as well as behavioural risks, with approximately 20% reductions for both groups, while the number of DALYs attributable to metabolic risks increased by nearly 50% over the same period, reflecting global changes in demographics and lifestyle.The greatest declines in the risk-attributable burden occurred for risk factors related to maternal and child health and unsafe water, sanitation, and handwashing, due largely to decreases in risk exposure but also to proportionally smaller infant and youth populations.Among risk factors related to the leading Level 3 risks, the steepest increases in the risk-attributable burden occurred for ambient particulate matter air pollution and for risk factors related to obesity and Bhaskar reports grants or contracts from Japan Society for the Promotion of Science (JSPS), Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT), Grant-in-Aid for Scientific Research (KAKENHI), and JSPS and the Australian Academy of Science (JSPS International Fellowship); leadership or fiduciary roles in board, society, committee or advocacy groups, paid or unpaid with Rotary District 9675 as the District Chair, Diversity, Equity, and Inclusion; Global Health & Migration Hub Community, Global Health Hub Germany (Berlin, Germany) as the Chair and Manager; PLOS One, BMC Neurology, Frontiers in Neurology, Frontiers in Stroke, Frontiers in Public Health, and BMC Medical Research Methodology as an Editorial Board Member; and the College of Reviewers, Canadian Institutes of Health Research (CIHR), Government of Canada as a member; outside the submitted work.B Bikbov reports grants or contracts from the European Commission, University of Rome, and Politecnico di Milano;

403 May 18, 2024 or
J S Chandan reports grants or contracts from the National Institute of Health Research Youth Endowment Fund (Home Office) and the College of Policing, University of Birmingham (UK Research and Innovation); support for attending meetings or travel, or both, from the University of Miami; outside the submitted work.S Cortese reports grants or contracts from the National Institute of Health Research and the European Research Agency; and payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from CADDRA, Medice, BAP, and ACAMH; outside the submitted work.L L Dalli reports support for the present manuscript from the National www.thelancet.comVol contracts from NZ Health Research Council and NZ Ministry of Health; participation on a data safety monitoring board or advisory board with Phase II, Multicenter, Double-Blinded, Randomized, Placebo-Controlled, Parallel-Group, Single-Dose Study to Determine the Safety, Preliminary Efficacy, and Pharmacokinetics of ARG-007 in Acute Ischemic Stroke Patients; leadership or fiduciary roles in other board, society, committee, or advocacy groups, paid or unpaid with Australia and NZ Stroke Organization, World Stroke Organization, and NZ Stroke Foundation; outside the submitted work.L F Reyes reports grants or contracts from GSK; royalties for licences from GSK; consulting fees from GSK; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from GSK; payment for expert testimony from GSK; support for attending meetings or travel, or both, from GSK and Pfizer; and stock or stock options from GSK; outside the submitted work.S Sacco reports grants or contracts from Novartis and Uriach; consulting fees from Novartis, Allergan-AbbVie, Teva, Lilly, Lundbeck, Pfizer, NovoNordisk, Abbott, and AstraZeneca; payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Novartis, Allergan-AbbVie, Teva, Lilly, Lundbeck, Pfizer, NovoNordisk, Abbott, AstraZeneca; support for attending meetings or travel, or both, from Lilly, Novartis, Teva, Lundbeck, and Pfizer; leadership or fiduciary roles in other board, society, committee or advocacy groups, paid or unpaid as the President elect European Stroke Organization and Editor-in-Chief of Cephalalgia;