Global, regional, and national burden of stroke and its risk factors, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019

Background Regularly updated data on stroke and its pathological types, including data on their incidence, prevalence, mortality, disability, risk factors, and epidemiological trends, are important for evidence-based stroke care planning and resource allocation. The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) aims to provide a standardised and comprehensive measurement of these metrics at global, regional, and national levels. Methods We applied GBD 2019 analytical tools to calculate stroke incidence, prevalence, mortality, disability-adjusted life-years (DALYs), and the population attributable fraction (PAF) of DALYs (with corresponding 95% uncertainty intervals [UIs]) associated with 19 risk factors, for 204 countries and territories from 1990 to 2019. These estimates were provided for ischaemic stroke, intracerebral haemorrhage, subarachnoid haemorrhage, and all strokes combined, and stratified by sex, age group, and World Bank country income level. Findings In 2019, there were 12·2 million (95% UI 11·0–13·6) incident stroke, 101 million (93·2–111) cases of 143 DALYs due to stroke, and 6·55 million deaths from stroke. the cause of death total deaths) the third-leading cause of death and DALYs) the absolute of incident strokes 70·0% prevalent strokes 85·0% (83·0–88·0), deaths from stroke 43·0% and DALYs to stroke by 32·0% (22·0–42·0). grains, low physical activity, and high LDL cholesterol were not assessed for intracerebral haemorrhage. Diet low in whole grains, alcohol use, low physical activity, high LDL cholesterol, and kidney dysfunction were not assessed for subarachnoid haemorrhage. DALY=disability-adjusted life-year. GBD=Global Burden of Diseases, Injuries, and Risk Factors Study.


Introduction
Disease and population distribution patterns, life expectancy, mortality, causes of death, and sociodemographic factors continue to change across the world, including ageing of populations and changes in the prevalence of risk factors for non-communicable disorders. Timely estimates of the burden of stroke and its pathological types, the burden attributable to risk factors, and trends in the burden over time are necessary at the global, regional, and national levels to guide evidence-based health-care policy, planning, and resource allocation for stroke.
The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2017 showed that stroke was the third-leading cause of death and disability combined (as measured by disability-adjusted life-years [DALYs]) and the second-leading cause of death in the world in 2017. 1,2 A GBD 2017 stroke analysis found that, although age-standardised mortality rates for stroke decreased sharply from 1990 to 2017, 2 the decrease in age-standardised incidence was much less steep, suggesting that prevention efforts have been less successful than treatment efforts. The results from GBD 2016 3 showed that 87·9% of ischaemic stroke DALYs and 89·5% of haemorrhagic stroke DALYs were due to potentially modifiable risk factors measured in GBD, demonstrating the enormous potential to reduce the burden of stroke through reductions in risk factor exposure. According to WHO, effective stroke prevention strategies include reducing the risk associated with hypertension (high systolic blood pressure), elevated lipids, diabetes (high fasting plasma glucose), smoking, low physical activity, unhealthy diet, and abdominal obesity (high body-mass index [BMI]), 4 which is similar to the findings from GBD 2016 3 and GBD 2017. 5 In this study, we estimated the global, regional, and national burden of overall stroke, ischaemic stroke, intracerebral haemorrhage, and subarachnoid haemorrhage in terms of their incidence, prevalence, mortality, and DALYs, as well as stroke-related DALYs associated with 19 potentially modifiable behavioural, environmental and occupational, and metabolic risk factors or groups of risk factors. We present data for 204 countries and territories, 21 GBD regions, and four World Bank income level groups from 1990 to 2019, by age group and sex. This manuscript was produced as part of the GBD Collaborator Network and in accordance with the GBD Protocol.
or retinal infarction. Intracerebral haemorrhage was defined as stroke with a focal collection of blood in the brain not due to trauma. Subarachnoid haemorrhage was defined as non-traumatic stroke due to bleeding into the subarachnoid space of the brain. The GBD methods for assigning cause of death to stroke and stroke subtypes in regions where neuroimaging was not available have been previously described. 9 GBD classifies causes into four levels, from the broadest (Level 1; eg, non-communicable diseases), to the most specific (Level 4; eg, intracerebral haemorrhage). Stroke is a Level 3 cause, within the Level 2 category of cardiovascular diseases, while its subtypes are Level 4 causes.

Fatal disease modelling
We used vital registration and verbal autopsy data as inputs into the Cause of Death Ensemble modelling (CODEm) framework to estimate deaths due to overall stroke and stroke subtypes. CODEm is a flexible modelling tool that utilises geospatial relationships and information from covariates to produce estimates of death for all locations across the time series . Deaths from vital registration systems coded to impossible or intermediate causes of death or unspecified stroke were reassigned by use of statistical methods (appendix sections 1.4, 1.7, 1.8). 10

Non-fatal disease modelling
Estimates of the incidence and prevalence of stroke were generated with the DisMod-MR 2.1 (disease-model-Bayesian meta-regression) modelling tool. 3 DisMod-MR is a Bayesian geospatial disease modelling software that uses data on various disease parameters, the epidemiological relationships between these parameters, and geospatial relationships to produce estimates of prevalence and incidence (appendix section 3). All available high-quality data on incidence, prevalence, and mortality were used to estimate non-fatal stroke burden. We modelled first-ever ischaemic stroke, intracerebral haemorrhage, and subarachnoid haemorrhage from the day of incidence through 28 days and separately modelled survival beyond 28 days.

Risk factor estimation
To analyse the attributable burden of stroke due to 19 risk factors currently available for such analysis in GBD 2019, we calculated population attributable fractions (PAFs) of DALYs (appendix section 2). 7 This work was done within the comparative risk assessments framework of GBD by use of four datasets: the burden estimates for stroke and its three pathological types; the exposure level for each risk factor; the relative risk of stroke as an outcome of exposure to the risk factor; and the theoretical minimum risk exposure level (TMREL), which is the level of exposure that minimises risk for each individual in the population. 11 The relative risks included in this analysis were generated from meta-analyses of epidemiological studies reporting associations between the risk factors of interest and stroke; these analyses are not stroke-type specific. The PAF (estimated independently for each risk factor) is the proportion of the cause that would be decreased if the exposure to the risk factor in the past had been reduced to the counterfactual level of the TMREL.
Risks included in the analysis were ambient particulate matter pollution; household air pollution from solid fuels; non-optimal temperature-ie, low temperature (daily temperatures below the TMREL) and high temperature (daily temperatures above the TMREL); lead exposure; diet high in sodium; diet high in red meat; diet low in fruits; diet low in vegetables; diet low in whole grains; alcohol consumption (any dosage); low physical activity (only for ischaemic stroke burden); smoking; secondhand smoke; high BMI; high fasting plasma glucose; high systolic blood pressure; high LDL cholesterol (only for ischaemic stroke burden); and kidney dysfunction, as measured by low glomerular filtration rate (GFR; not assessed for subarachnoid haemorrhage burden). As with causes, GBD organises risk factors into four levels, from the broadest (Level 1) to the most specific (Level 4). In addition to the specific risk factors above, we assessed the Level 1 groups of risks: behavioural, environmental and occupational, and metabolic. The PAFs of risk factor groups took into account interactions between risk factors included in the group, as explained elsewhere. 12 Percentages and number of DALYs are not mutually exclusive. The crude sum of the PAF of the risk factors might exceed 100% because the effects of many of these risk factors are mediated partly or wholly through another risk factor or risk factors. Definitions of risk factors and risk groups and further details of risk factors are provided in the appendix (section 2.1).

Data sources and presentation
For GBD 2019, we used data from 3686 vital registration sources, 147 verbal autopsy sources, 368 incidence sources, 117 prevalence sources, 229 excess mortality sources, 7753 risk factor exposure sources, and 2733 risk factor relative risk sources. Further details of the data sources used in this analysis are available on the Global Health Data Exchange website.
Estimates in this Article are presented in absolute numbers and as age-standardised rates per 100 000 pop ulation (with 95% uncertainty intervals [UIs]) and are stratified by age, sex, 21 GBD regions, seven GBD super-regions (appendix figure 6.1), and four income levels (as determined by the World Bank). 13 Count data are presented in tables to two decimal places (and rounded to one decimal place in the text), and percentage data (including percentage change) are presented to one decimal place.

Role of the funding source
The funder had no role in study design, data collection, data analysis, interpretation of the study results, writing For more on the Global Health Data Exchange see http://ghdx. healthdata.org/ of the report, or the decision to submit the manuscript for publication.
Although the absolute number of DALYs due to stroke in males (    Intracerebral haemorrhage and subarachnoid haemorrhage showed larger reductions in age-standardised rates from 1990 to 2019 than ischaemic stroke (table 1; appendix section 4·2, tables S3-5, and figure S8). There were substantial between-country variations in the agestandardised incidence (figures 1B-D), prevalence, mortality, and DALY rates (appendix figures S8-11) of these three pathological types of stroke by GBD regions, country income level, and sex (appendix section 4.2), with an almost two-fold greater proportion of intracerebral haemorrhage in World Bank low-income to upper-middle-income countries compared with highincome countries (29·5% [28·4-30·3] vs 15·8% [15·5-16·2]), but a lower proportion of subarachnoid haemorrhage in low-income to upper-middle-income countries compared with high-income countries (7·9%

Stroke-related DALYs attributable to risk factors
In other words, if high BMI exposure were reduced to its TMREL, there would be a 24·3% reduction in stroke in 2019, compared to just a 15·4% reduction in 1990. Other risk factors with an increasing age-standardised stroke PAF from 1990 to 2019 included high systolic blood  In 2019, there were moderate between-country (1·3 times), regional (as measured by 21 GBD regions), and country economic development level (as measured by the World Bank income groups) variations in the proportion of stroke-related DALYs and its DALYs related to stroke pathological types that were attributable to risk factors. Between-country variations were more pronounced for subarachnoid haemorrhage (figure 3; appendix tables S6-12 and figures S12-14), and the highest Household air pollution from solid fuels

Physiological factors
High body-mass index proportion of stroke-related DALYs was observed in the World Bank low-income to upper-middle-income groups (ranging from 85·9% [95% UI 83·2-88·6] in the World bank low-income group to 87·3% [84·4-89·9] in the World Bank upper-middle-income group). From 1990 to 2019, there was an increase in the total number of stroke-related DALYs due to high BMI, high fasting plasma glucose, high LDL cholesterol, kidney dysfunction, a diet high in red meat, alcohol consumption, and second-hand smoking, but a decrease in DALYs due to smoking and a diet low in fruits and vegetables (appendix figure S15). There were also moderate variations in the ranking of risk factors by pathological types of stroke (figure 4; appendix figures S16-18). In 2019, the five leading specific risk factors contributing to stroke death and disability combined (DALYs) were high systolic blood pressure ( figure 5). For risk factors by pathological type of stroke and changes in risk factor rankings from 1990 to 2019 by GBD regions, see the appendix (section 4.3 and figures S19-26).

Discussion
In 2019, stroke remained the second-leading Level 3 cause of death and the third-leading Level 3 cause of death and disability combined in the world, and its burden (in terms of the absolute number of cases) increased substantially from 1990 to 2019. Our findings indicate that the bulk of the global stroke burden (86·0% [95% UI 85·9-86·9] of deaths and 89·0% [88·9-89·3] of DALYs) is in lower-income and lower-middle-income countries. Globally, over the past three decades, the total number of stroke-related DALYs due to risk factors increased substantially (by 33·5 million, from 91·5 million in 1990 to 125 million in 2019), with diverging trends in World Bank high-income countries and low-income to upper-middle-income countries: a relatively small decrease in the high-income group and large increases in the low-income to upper-middle income groups. The large increase in the global burden of stroke was probably not only due to population growth and ageing but also because of the substantial increase in exposure to several important risk factors such as high BMI, ambient particulate matter pollution, high fasting plasma glucose, high systolic blood pressure, alcohol consumption, low physical activity, kidney dysfunction, and high temperature (appendix figure S26). 7,14 This study is also the first systematic analysis to determine the effect of non-optimal temperature on

Combined risk factors
All factors 125·00 (115·00-134·00) Data in parentheses are 95% uncertainty intervals. Count data in millions are presented to two decimal places and percentage data are presented to one decimal place. Percentages and number of DALYs are not mutually exclusive: the sum of percentages and number of DALYs in the columns exceeds the totals for all risk factors combined because of overlap between various risk factors. The crude sum of population attributable fraction (PAF) of the risk factors might exceed 100% because the effects of many of these risk factors are mediated partly or wholly through another risk factor or risk factors. DALY=disabilityadjusted life-year. PM 2·5 =particulate matter with a diameter of <25 μm. *Air pollution cluster includes ambient PM 2·5 pollution and household air pollution from solid fuels. †Tobacco smoke cluster includes smoking and second-hand smoking. ‡Dietary risks cluster includes diet high in sodium, diet low in fruits, diet low in vegetables, diet high in red meat, and diet low in whole grains, and alcohol consumption. §Behavioural risks cluster includes smoking (including second-hand smoking), dietary risks (diet high in sodium, diet low in fruits, diet low in vegetables, diet high in red meat, diet low in whole grains, and alcohol consumption), and low physical activity. ¶Environmental risks cluster includes air pollution cluster, low ambient temperature, high ambient temperature, and lead exposure. ||Metabolic risks cluster includes high body-mass index, high fasting plasma glucose, high LDL cholesterol, high systolic blood pressure, and kidney dysfunction.   stroke burden. The greater age-standardised burden of stroke in World Bank low-income to upper-middleincome countries than in the high-income countries might also relate to poorer acute health care for stroke, 15 poorer stroke awareness, 16 and greater prevalence or effect of some risk factors (eg, tobacco use, poor diet, diabetes, hypertension, cardiovascular disease, rheumatic heart disease, dyslipidaemia, and obesity) in low-income countries than in upper-middle-income countries, 17,18 which highlights the inadequacy of primary prevention efforts in these settings.
For the first time, we have presented the global, regional, and national burden of stroke and its risk factors by its major pathological types. Although ischaemic S o u t h e rn L a t in A m e ri c a W e s t e rn E u ro p e A n d e a n L a t in A m e ri c a C a ri b b e a n C e n t ra l L a t in A m e ri c a T ro p ic a l L a t in A m e ri c a N o rt h A fr ic a a n d M id d le E a s t

S o u t h A s ia E a s t A s ia
O c e a n ia S o u t h e a s t A s ia C e n t ra l s u b -S a h a ra n A fr ic a E a s t e rn s u b -S a h a ra n A fr ic a S o u t h e rn s u b -S a h a ra n A fr ic a W e s t e rn s u b -S a h a ra n A fr ic a   stroke continues to constitute the largest proportion of all new strokes (comprising 62·4% of all incident strokes in 2019), followed by intracerebral haemor rhage (27·9%), and subarachnoid haemor rhage (9·7%), the relative proportions of each pathological type varied substantially by income group. For example, a new stroke case was nearly twice as likely to be intracerebral haemorrhage in the World Bank low-income to upper-middle-income groups combined than in the high-income group (29·5% of all incident strokes in 2019 vs 15·8%), whereas a new stroke case was more than twice as likely to be subarachnoid haemorrhage in the World Bank high-income group than in the low-income to upper-middle-income groups combined (19·7% vs 7·9%). The increased risk of intracerebral haemorrhage in low-income and uppermiddle-income countries might be related to the high relative clinical significance and population-attributable risk of hypertension in these countries. 18 Our finding that a greater proportion of incident strokes in low-income to upper-middle-income countries are intracerebral haemor rhages in males than in females (appendix figure F6.9) are in line with previous observations, 19,20 and might be explained by lower levels of awareness and control of hypertension in low-income  S o u t h e rn L a t in A m e ri c a W e s t e rn E u ro p e A n d e a n L a t in A m e ri c a C a ri b b e a n C e n t ra l L a t in A m e ri c a T ro p ic a l L a t in A m e ri c a N o rt h A fr ic a a n d M id d le E a s t S o u t h A s ia E a s t A s ia O c e a n ia S o u t h e a s t A s ia C e n t ra l s u b -S a h a ra n A fr ic a E a s t e rn s u b -S a h a ra n A fr ic a S o u t h e rn s u b -S a h a ra n A fr ic a W e s t e rn s u b -S a h a ra n A fr ic a

Intracerebral haemorrhage
Subarachnoid haemorrhage C D to upper-middle-income countries than in high-income countries, 18,21 as well as increased exposure to risk factors predisposing a higher proportion of males to intracerebral haemorrhage compared with females. 20,22 Our study also adds to the body of research on the incidence of subarachnoid haemorrhage; a previous systematic review of population-based studies of subarachnoid haemorrhage incidence by Etminan and colleagues 23 had similar findings to ours, as the authors used many of the same sources, but they only used crude incidence rates, which is perhaps why we found smaller between-country variations in the age-standardised incidence of subarachnoid haemor rhage (approximately a tenfold variation in our findings [appendix table s5] compared with a >20-fold variation in the systematic review). 23 The size of between-country variations we observed in age-standardised incidence, prevalence, and DALY rates of other pathological types of stroke and stroke overall were in line with previous observations. 2,3,19,20,24 Despite the overall declines in age-standardised stroke incidence, prevalence, death, and DALY rates, three concerning trends have emerged. First, the greatest share of the global burden of stroke continues to be borne by low-income to upper-middle-income countries.
The proportion of DALYs attributable to GBD-modelled risk factors was also particularly high in low-income to upper-middle-income countries. Second, the pace of the global decline in age-standardised stroke incidence, death, and DALY rates was noticeably slower over the past decade  than in the previous decade , and global age-standardised prevalence significantly increased from 2010 to 2019 (appendix figure 6.8). There was a significant increase in stroke prevalence and incidence rates in people younger than 70 years between 1990 and 2019 with even faster increases from 2010 to 2019 (appendix figure 6.7). A trend towards plateauing or increasing stroke incidence or mortality rates, or both, in middle-aged people was recently observed in the USA, European countries, Brazil, and China. [25][26][27][28][29][30] This trend might be a reflection of the increased exposure to some risk factors for stroke, such as elevated blood pressure, high BMI, and high fasting plasma glucose, across most countries. [31][32][33] In the USA, a worrisome trend observed in recent years  is that awareness of hypertension in the population whose blood pressure is controlled is declining. 34 Third, most countries have not achieved sufficient declines in stroke incidence rates to offset the demographic force of  population growth and ageing, resulting in overall increases in the number of incident, prevalent, fatal, and disabling strokes over time. A linear interpolation shows that if current trends continue, by 2050 there will be more than 200 million stroke survivors and almost 300 million DALYs, 25 million new strokes, and 13 million deaths from stroke annually. This study was, to our knowledge, the first systematic analysis to provide estimates of the burden of stroke and its subtypes associated with non-optimal temperature (daily temperatures below or above the TMREL). Although previous studies have made ecological observations of the effects of ambient temperature on the risk of stroke, this study was the first to show the sizeable global effect of non-optimal temperature (primarily low temperature, at 8·36 million [95% UI 6·19-10·80] DALYs or a PAF of 5·8% [4·4-7·5]) on the burden of stroke and its pathological types (appendix tables T10b, T11b, and T12b). These findings were in line with a recent systematic review on ambient temperature and stroke occurrence. 35 Our estimates of geographical variations in the burden of stroke and its pathological types associated with non-optimal temperature and other risk factors suggest that country-specific and stroke type-specific priorities and strategies should be developed and implemented for reducing the burden of stroke in different geographical locations.
Our findings of the high proportion (87·0%) of age-standardised stroke-related DALYs associated with GBD risk factors are in line with previous observations 17,36 and highlight the potential to greatly reduce the stroke burden by addressing risk factor exposure. The increased contribution of certain metabolic risk factors in 2019 compared with 1990 (eg, an increase in the proportional contribution to stroke-related DALYs of 57·8% by high BMI and 40·3% by high fasting plasma glucose) and a decreasing contribution of certain environmental and occupational and behavioural risk factors to the strokerelated DALY burden over the same period (eg, a 38·2% decrease for household air pollution from solid fuels and a 6·1% decrease for a diet low in vegetables) might be related to a growing proportion of the global population reaching the final stages of the epidemiological transition, in which the risk burden has shifted towards metabolic risk factors and an increased proportion of the disease burden comes from stroke and other noncommunicable diseases. 37 This observation also means that guidance on reducing the risk of stroke by targeting certain risk factors will need to change to reflect changes in the risk-attributable profile.
Our estimates of the global, regional, and national burden of stroke and its pathological types and risk factors are important for evidence-based health-care planning, priority setting, and resource allocation for stroke care, primary prevention, and research. The high and increasing stroke burden alongside stagnant or even increasing mortality rates from cardiovascular disease in some countries, 14 and increasing rates of exposure to many important stroke risk factors from 1990 to 2019, 7,14 suggest that current primary stroke prevention strategies and measures are not sufficient, and that efforts to implement population-wide primary prevention strategies more widely must be reinforced worldwide. 38 For every US$1 spent on prevention of stroke and cardiovascular disease, there is an estimated $10·9 return on investment. 39 Population-wide interventions for primary prevention of stroke and cardiovascular disease should include measures to reduce exposure to metabolic risk factors (eg, screening for and proper management of systolic blood pressure and weight), behavioural risk factors (eg, smoking cessation programmes and programmes to increase the accessibility and affordability of nutrient-rich foods), and environmental and occupational risk factors (eg, measures to reduce air pollution and lead exposure). The development and implementation of such populationlevel interventions, alongside efforts to reduce poverty and racial and socioeconomic inequities, through legislation, taxation, and other measures at the government level, must be the mainstream approach for reducing the risk of stroke, cardiovascular disease, and other non-communicable diseases, but the importance of primary prevention measures at the individual level should not be overlooked. In this respect, the emphasis should be on strategies that are appropriate for most people at risk of stroke and cardiovascular disease regardless of their level of risk exposure, 38 such as digital health technologies for affordable identification of people at increased risk of stroke and cardiovascular disease, universal health coverage, cheap and effective multidrug regimens (eg, polypills) for people at increased risk of cardiovascular disease, and involvement of health-care volunteers in primary prevention activities. For example, the World Stroke Organization recommends that all adults know their individual risk of having a stroke, their personal risk factors for stroke, and how to control these risk factors using the validated, internationally endorsed, and free Stroke Riskometer app, which is currently available in 19 languages for more than 70% of the global population. 40 A recent Cochrane systematic review showed the feasibility and potential effectiveness of several health promotion interventions targeting risk factors to achieve behavioural changes for primary prevention of cardiovascular disease in low-income to upper-middle-income countries. 41 Although knowledge of personal risk and management of behavioural risk factor activities is primarily the prerogative of individuals, health professionals have a responsibility to identify risk factors that require pharmacological and non-pharma cological treatment to reduce the chance of stroke occurrence (eg, elevated blood pressure, atrial fibrillation, diabetes, dyslipidaemia, or symptomatic carotid artery stenosis). Simple, inexpensive screening for cardiovascular disease risks (eg, elevated blood For more on the Stroke Riskometer & PreventS app see https://nisan.aut.ac.nz/Stroke-Riskometer/ pressure, smoking, and overweight) by health professionals in low-income and middle-income settings or more accurate screening for high cardiovascular disease risks (including blood lipid tests) by health professionals in higher-income locations can help to identify people who might require prophylactic drug therapy, in conjunction with behavioural interventions. 40 However, health professionals often do not have enough time to conduct detailed assessments of behavioural risk factors or to develop individually tailored recommendations for primary prevention of stroke and cardiovascular disease. To ameliorate this problem, data on stroke risk and risk factors from individuals should be integrated with the electronic patient management systems of health service providers. A study in Finland suggests that the quality of stroke prevention by primary health-care professionals could be improved by developing digital clinical decision-making tools and by implementing inter-professional teamwork 42 (eg, the PreventS web app currently being developed in New Zealand). All of these measures should be facilitated by ongoing, culturally appropriate health education campaigns (including coordinated activities of non-governmental organisations) and inclusion of such health education information into standardised educational curricula at all levels.
In addition to primary stroke prevention efforts, appropriate secondary prevention efforts and adequate acute treatment and rehabilitation are essential to improve stroke outcomes. Our findings of large geographical variations in stroke prevalence, mortality, and disability are a reflection not only of geographical differences in stroke incidence but also of major inequities in acute stroke care and rehabilitation across countries. 43 Even in European countries, only 7·3% of all patients with acute ischaemic stroke receive intravenous thrombolysis and only 1·9% receive endovascular treatment, with the highest country-level rates being 20·6% for intravenous thrombolysis (in the Netherlands) and 5·6% for endovascular treatment (in Malta), 44 and one in three patients discontinues using one or more secondary stroke prevention drugs about 1 year after stroke. 45 Treatment rates are even lower in many low-income and middleincome countries. 21,43 To reduce inequalities in stroke care, a roadmap for delivering quality stroke care and various action plans 46,47 have been suggested, with emphasis on the importance of applying culturally appropriate and context-appropriate strategies. There is a pressing need to implement evidence-based guidelines for stroke management and to reduce the gap in stroke care between high-income countries and low-income and middleincome countries. Recent evidence suggests that delivering an adequate level of stroke care 48,49 and preventive interventions 49 in low-income and middleincome countries are feasible. Attention should be paid to developing the workforce for stroke care and setting up affordable and accessible rehabilitation facilities. Promising results 50 suggest that self-management could be used as an adjunct strategy for ongoing rehabilitation at home or in other settings. The importance of countrybased ongoing stroke registries and stroke risk factors surveys, which are profoundly lacking in low-income and middle-income countries, should also be emphasised.
Although this study was, to our knowledge, the first and most comprehensive review of the global, regional, and national burden of stroke and its 19 specific risk factors by all three pathological types, it was not free from limitations common to all previous GBD estimates of stroke risk and risk factors, 2,3,11,36 particularly the absence of original, good-quality stroke epidemiological studies for most countries. We therefore were not able to include some important potential risk factors (eg, atrial fibrillation and substance abuse), or include different patterns in risk factor exposure (eg, different doses and types of alcohol consumption, pack-years of smoking) and doses of exposure, analyse stroke burden by ischaemic stroke subtypes, or do a decomposition analysis to attribute changes in stroke burden to changes in the population growth, ageing, and risk factors separately. Additionally, evidence for the selection of TMRELs for some risk factors was uncertain and based on non-experimental studies, although all TMRELs were discussed and approved by a team of risk epidemiologists and stroke experts. Despite these limitations, our results are broadly consistent with previous estimates from population-based and analytical epidemiological studies, thus supporting the validity of our results.
In summary, although strokes are largely preventable, as indicated by declining incidence rates globally, stroke remained the second-leading cause of death and thirdleading cause of death and disability combined worldwide in 2019. Without wider implementation of populationwide primary stroke and cardiovascular disease prevention strategies, the burden of stroke is likely to continue growing, disproportionally affecting lowincome and middle-income countries. As the 19 analysed risk factors for stroke are common for other major noncommunicable diseases, appropriate control of these risk factors will also reduce the burden of coronary heart disease, vascular dementia, type 2 diabetes, and even some types of cancer. Further research on the frequency, outcomes, and determinants of stroke and its pathological types in different locations and over time is warranted. Such research could include identifying populations at highest risk as well as further investigating differences in stroke pathological types and their geographical patterns, all of which would be useful for more targeted prevention and treatment efforts. Closing the gaps between high-income countries and low-income and middle-income countries in the adaptation and implementation of internationally recognised guidelines and recom mendations for reducing stroke morbidity and mortality, with an emphasis on primary prevention strategies, is crucial to addressing the global stroke burden.

Contributors
Please see the appendix (pp 7-10) for more detailed information about individual author contributions to the research, divided into the following categories: managing the estimation or publication process; writing the first draft of the manuscript; primary responsibility for applying analytical methods to produce estimates; primary responsibility for seeking, cataloguing, extracting, or cleaning data; designing or coding figures and tables; providing data or critical feedback on data sources; development of methods or computational machinery; providing critical feedback on methods or results; drafting the manuscript or revising it critically for important intellectual content; extracting, cleaning, or cataloguing data; designing or coding figures and tables; and managing the overall research enterprise. V L Feigin and G A Roth had access to and verified the data underlying this study. All authors had full access to all the data in the study, and V L Feigin had final responsibility for the decision to submit for publication.