Global, regional, and national burden of Alzheimer's disease and other dementias, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016

1Alzheimer's disease and other dementias are a major and increasing global health challenge, with 40–50 million people currently living with dementia.,, Care and support of patients with dementia has wide-ranging consequences for families, health-care systems, and society as a whole. There is growing evidence of risk factors for dementia, which shows that lifestyle and other interventions might, if implemented effectively, contribute to delaying the onset and reducing the future number of people who have dementia.,,,, Changes in risk factor exposures might account for several cohort studies documenting a reduction in age-specific incidence rates in their study populations.,,,

2The Global Burden of Diseases, Injuries, and Risk Factors (GBD) Study uses a systematic method to analyse fatal and non-fatal health losses to facilitate comparisons across countries and diseases. Annual updates of results quantify mortality, prevalence, incidence, and non-fatal health losses for more than 300 diseases and injuries by age and sex from 1990 for 195 countries and territories and many subnational locations, such as Mexican states or provinces of China. Although other efforts to estimate dementia prevalence,, have used meta-analytic strategies to synthesise the available data, these analyses did not attempt to reconcile and make the combined best use of different types of data on incidence, prevalence, excess mortality, and causes of death reported in literature sources, claims databases, and vital registration systems. Furthermore, GBD assesses the strength of causal evidence for risks and includes estimates of attributable disease burden from risk factors using a comparative risk assessment framework.

3Evidence before this studyOver the past decade, there have been substantial research efforts to describe the global epidemiology of Alzheimer's disease and other dementias. Other efforts to summarise the evidence, including the World Dementia Report, have compiled evidence through systematic reviews and meta-analyses, but they report only on dementia prevalence by world region. The Global Burden of Diseases, Injuries, and Risk Factors (GBD) Study has published regular updates of estimates for dementia in 195 countries and territories and subnational locations, such as the provinces of China and states in countries such as Brazil, the USA, and Mexico. However, there has been no dedicated and detailed publication of GBD methods and estimates for dementia. GBD 2016 estimates for dementia incorporated 43 new data sources on the prevalence and incidence of dementia, updating a previous review that covered 1980 to 2015; new sources were identified through a systematic review of English-language articles published in PubMed from 23 Jan, 2015, to Oct 7, 2016, with the search terms “dementia”[Title/Abstract] AND (“prevalence” OR “incidence”)[Title/Abstract].Added value of this studyGBD 2016 added 5-year age groups from ages 80 to 95 years to replace the oldest category (≥80 years) used in GBD 2015, providing more detailed estimates where the burden from dementia was highest. Our report examined more closely the methods, results, and limitations specific to Alzheimer's disease and other dementias with the aim of making this information more accessible to clinicians and researchers. The data informing estimates were heterogeneous, with 230 different diagnostic procedures across 237 studies. Over time, the global age-standardised prevalence was fairly stable: 701 cases (95% uncertainty interval [UI] 602–815) per 100 000 population in 1990 and 712 cases (614–828) per 100 000 population in 2016, constituting an increase of 1·7% (95% UI 1·0–2·4). However, because of population ageing and growth, the number of people affected by dementia more than doubled since 1990 and almost 44 million prevalent cases were estimated globally in 2016. Age-standardised prevalence was 1·17 (1·17–1·18) times higher in females than in males. We also estimated that 22·3% (11·8–35·1) of the total global DALYs due to dementia in 2016 could be attributed to the four modifiable risk factors that met GBD criteria for assessment (high body-mass index, high fasting plasma glucose, a diet high in sugar-sweetened beverages, and smoking).Implications of all the available evidenceThis analysis identified substantial heterogeneity in case-ascertainment methods throughout the dementia literature, highlighting the need for more consistency in future research. Nevertheless, there is no doubt about the striking increase in the numbers of individuals living with dementia since 1990 due to ageing and population growth. Despite some evidence indicating small decreases in age-specific incidence, without a major scientific breakthrough the continuation of sweeping demographic trends in population ageing and growth will lead to further increases in the number of people living with dementia. With limited scope for prevention and the absence of an effective disease-modifying treatment, the burden on caregivers and the parts of health-care systems devoted to care of the elderly will continue to increase rapidly.

4Evidence before this study

5Over the past decade, there have been substantial research efforts to describe the global epidemiology of Alzheimer's disease and other dementias. Other efforts to summarise the evidence, including the World Dementia Report, have compiled evidence through systematic reviews and meta-analyses, but they report only on dementia prevalence by world region. The Global Burden of Diseases, Injuries, and Risk Factors (GBD) Study has published regular updates of estimates for dementia in 195 countries and territories and subnational locations, such as the provinces of China and states in countries such as Brazil, the USA, and Mexico. However, there has been no dedicated and detailed publication of GBD methods and estimates for dementia. GBD 2016 estimates for dementia incorporated 43 new data sources on the prevalence and incidence of dementia, updating a previous review that covered 1980 to 2015; new sources were identified through a systematic review of English-language articles published in PubMed from 23 Jan, 2015, to Oct 7, 2016, with the search terms “dementia”[Title/Abstract] AND (“prevalence” OR “incidence”)[Title/Abstract].

6Added value of this study

7GBD 2016 added 5-year age groups from ages 80 to 95 years to replace the oldest category (≥80 years) used in GBD 2015, providing more detailed estimates where the burden from dementia was highest. Our report examined more closely the methods, results, and limitations specific to Alzheimer's disease and other dementias with the aim of making this information more accessible to clinicians and researchers. The data informing estimates were heterogeneous, with 230 different diagnostic procedures across 237 studies. Over time, the global age-standardised prevalence was fairly stable: 701 cases (95% uncertainty interval [UI] 602–815) per 100 000 population in 1990 and 712 cases (614–828) per 100 000 population in 2016, constituting an increase of 1·7% (95% UI 1·0–2·4). However, because of population ageing and growth, the number of people affected by dementia more than doubled since 1990 and almost 44 million prevalent cases were estimated globally in 2016. Age-standardised prevalence was 1·17 (1·17–1·18) times higher in females than in males. We also estimated that 22·3% (11·8–35·1) of the total global DALYs due to dementia in 2016 could be attributed to the four modifiable risk factors that met GBD criteria for assessment (high body-mass index, high fasting plasma glucose, a diet high in sugar-sweetened beverages, and smoking).

8Implications of all the available evidence

9This analysis identified substantial heterogeneity in case-ascertainment methods throughout the dementia literature, highlighting the need for more consistency in future research. Nevertheless, there is no doubt about the striking increase in the numbers of individuals living with dementia since 1990 due to ageing and population growth. Despite some evidence indicating small decreases in age-specific incidence, without a major scientific breakthrough the continuation of sweeping demographic trends in population ageing and growth will lead to further increases in the number of people living with dementia. With limited scope for prevention and the absence of an effective disease-modifying treatment, the burden on caregivers and the parts of health-care systems devoted to care of the elderly will continue to increase rapidly.

10Global data and cross-country comparisons might help further current understanding of complex and multifactorial diseases such as dementia. The capacity of GBD to examine patterns across countries creates a unique opportunity to identify populations with different trends that in turn could reveal clustered risk or environmental factors, providing empirical evidence about factors that affect neurodegenerative diseases. Such findings could provide insights that can supplement and aid ongoing research, and also allow estimates of the future effect of dementia expected from rapidly ageing populations across the world. To facilitate further exploration of these patterns, we present an analysis of GBD 2016 data with the aim of articulating the key methods, results, and limitations pertaining to dementia estimation.

11General methods of GBD, including methods for calculating the Socio-demographic Index (SDI), a composite indicator combining income per person, education, and fertility, can be found in the general methods overview (appendix) and in the GBD 2016 overview papers.,,, In this Article, we have highlighted the methods pertaining to processes specific to the estimation of Alzhemier's disease and other dementias (hereafter referred to as dementia).

12For the GBD analyses, the reference case definitions were either those from the Diagnostic and Statistical Manual of Mental Disorders (DSM; DSM-III, DSM-IV, or DSM-5), which are used in surveys and cohort studies, or those from the International Classification of Diseases (ICD; ICD-8, ICD-9, and ICD-10), which are used in vital registration and claims data sources., For GBD analyses of dementia, the relevant ICD-9 codes were 290, 291.2, 291.8, 294, and 331, and the relevant ICD-10 codes were F00, F01, F02, F03, G30, and G31. In GBD, for each disease a reference case definition was chosen that represents the most recent consensus or the most commonly used definition. Data based on alternative case definitions were adjusted if systematic bias was detected. We included 237 sources of data from the scientific literature, and 3 years of medical claims data from the USA. We identified 176 sources reporting on prevalence, covering 17 of 21 world regions, and 64 sources on incidence, covering ten of 21 world regions (appendix). No input data were available for Oceania, central Asia, eastern Europe, or southern sub-Saharan Africa.

13We could not adjust the data for different study designs or case definitions because of the extreme heterogeneity in case-ascertainment methods. We identified 230 different methods of screening and diagnosis in the 237 scientific literature sources. Data points that had an age range of greater than 20 years were split into new data points for the 5-year age groups within the age range, using the age pattern from the USA as the country for which we had the most detailed information.

14We did not use the typical GBD cause of death ensemble modelling (CODEm) approach for dementia because we noted large inconsistencies between cause of death data and prevalence data over time and between countries. Data from the US Vital Registration system showed that the age-standardised rates of deaths from dementia between 1990 and 2016 increased by a factor of five, whereas there has been no corresponding increase in the prevalence of dementia over the same period.,, Additionally, the highest age-standardised death rates were more than 50 times higher than the lowest age-standardised death rates across different locations for 2016, possibly indicating that the practice of coding a death to dementia as an underlying cause of death has not been consistent over time or between countries. To adjust for this bias, we jointly modelled cause of death and non-fatal outcomes for dementia. We first ran an initial cause of death model using CODEm, and an initial non-fatal model using DisMod-MR 2.1, the Bayesian meta-regression tool used in most non-fatal models for GBD., DisMod-MR 2.1 enforces consistency between the different parameters, because incidence determines the inflow into the pool of prevalent cases and excess mortality determines outflow via death. The rates of inflow and outflow then determine the average duration of disease. Both DisMod MR 2.1 and CODEm use covariates and borrow strength from locations in the same region that have data to make estimates for locations where there are no input data.

15The initial CODEm model included 16 226 site-years of data (ie, data for a unique combination of location and calendar year). The covariates used in the initial CODEm model included diabetes prevalence, mean cholesterol, and mean body-mass index (BMI; full list in the appendix). The initial DisMod-MR 2.1 model included settings of no remission (ie, no cure), and no incidence before age 40 years. We also excluded incidence data from the model, because data for incidence and prevalence were inconsistent. In many locations, the incidence data combined with mortality estimates suggested a prevalence higher than the available prevalence data, and the model fitted above the prevalence data and below the incidence data. Because measurement of prevalent cases of dementia probably has less error than measurement of incident cases, we decided to exclude incidence data and rely on prevalence data.

16We let DisMod-MR 2.1 adjust medical claims data to correct for under-reporting compared with survey data. We used average years of education in the population aged 15 years and older and smoking prevalence as predictive covariates, because these risk factors for dementia have a large evidence base., From the initial model results we identified the locations with high-quality vital registration systems that most often coded to dementia as a cause of death per prevalent case in the most recent year of estimates. For GBD 2016, these locations were the USA, Finland, Sweden, and Puerto Rico. We then used the log-transformed ratio between cause-specific mortality and prevalence for these locations to run a fixed-effects linear regression with dummy variables on 5-year age bins and sex. Because the ratio of cause-specific mortality and prevalence is an estimate of excess mortality rate, we used the results of this regression to predict data inputs for every GBD location and all years in a second DisMod-MR 2.1 model. As an exception, we retained the 2016 ratios of cause-specific mortality and prevalence for the four locations used in the regression analysis but assumed the 2016 ratios applied to the whole 1990–2016 period. Otherwise, the model included the same assumptions and covariates as the initial DisMod-MR 2.1 model. We used the prevalence outputs from this model to estimate years lived with disability (YLDs). We also used the cause-specific mortality estimated from this model for our final cause-specific mortality results.

17To derive YLDs for dementia, we first divided dementia prevalence into three severity categories. We used data from a systematic review of studies reporting on the Clinical Dementia Rating Scale (CDR) and pooled proportions of mild (CDR 1), moderate (CDR 2), and severe dementia (CDR 3) in random-effect meta-analyses. We did this separately for ages 40–69, 70–79, and 80 years or more. For GBD 2016, we included seven studies identified through systematic review published from 1980 to 2015, covering two world regions. To calculate YLDs, we multiplied the prevalence at each severity level by the corresponding disability weight and further corrected for comorbidity using a simulation assigning all non-fatal outcomes to hypothetical individuals for each age group, sex, location, and year.

18We made estimates for high BMI, high fasting plasma glucose, smoking (including all smoked tobacco products), and diet high in sugar-sweetened beverages as risk factors for dementia. For each of these risk factors we set a theoretical minimum exposure level (TMREL) at which the risk of health outcomes is lowest. The TMREL was set to greater than 20 and less than 25 kg/m2 for BMI; greater than 4·5 and less than 5·4 mmol/L for high fasting plasma glucose; zero for smoking; and less than 5g/day for sugar-sweetened beverages.

19Criteria for inclusion as a risk factor in GBD included sufficient evidence of a causal relationship, availability of exposure data, and potential for modification. Although physical inactivity and absence of education were not assessed for inclusion as risk factors for dementia in GBD 2016, these will be considered in upcoming rounds. We calculated population-attributable fractions based on relative risk data, exposure data, and a theoretical minimum level of exposure. When aggregating risks, we assumed a multiplicative function and accounted for instances where one risk was mediated through another. Additional details on risk factor calculations are in the GBD 2016 risk factor overview paper.

20Years of life lost (YLLs) were calculated by multiplying the reference life expectancy at each age, taken from the GBD reference life table, by the number of deaths in each age group. Disability-adjusted life years (DALYs) were then calculated as the sum of YLLs and YLDs. Age-standardised rates were calculated using the GBD world population standard. Uncertainty was propagated through all calculations by sampling 1000 draws at each step of the calculations, which enabled us to carry through uncertainty from multiple sources, including input data, corrections of measurement error, and estimates of residual non-sampling error. 95% uncertainty intervals (95% UI) were defined as the 25th and 975th values of the ordered draws. Any report on significant differences was based on the 95% UI of the difference not including zero.

21The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

22Between 1990 and 2016, the number of prevalent dementia cases increased by 117% (95% UI 114–121), from 20·2 million (17·4–23·5) in 1990 to 43·8 million (37·8–51·0) in 2016, whereas there was an increase of only 1·7% (95% UI 1·0–2·4) in age-standardised prevalence, from 701 (602–815) per 100 000 population in 1990 to 712 (614–828) per 100 000 population in 2016. For all-age prevalence over the same period, there was an increase of 54·7% (95% UI 52·1–57·5), from 383 (330–447) per 100 000 population in 1990 to 593 (511–690) per 100 000 population in 2016. The percentage change from 1990 to 2016 in age-standardised prevalence was highest for the high-middle SDI quintile (table). Age-standardised prevalence varied by a factor of three across countries in 2016. Turkey had the highest age-standardised prevalence (1192 [95% UI 1007–1405] cases per 100 000 population), followed by Brazil (1037, 882–1220). Nigeria (397, 335–462) and Ghana (406, 342–483) had the lowest age-standardised prevalence estimates (figure 1).

23The number of deaths due to dementia increased by 148% (140–157) between 1990 and 2016. Globally in 2016, dementia was the fifth-largest cause of death (2·4 million [95% UI 2·1–2·8] deaths) after ischaemic heart disease, chronic obstructive pulmonary disease, intracerebral haemorrhage, and ischaemic stroke. In 2016, deaths due to dementia accounted for 4·4% (95% UI 3·8–5·1) of total deaths but 8·6% (7·4–10·1) of deaths in individuals aged more than 70 years (2·2 million [1·9–2·6] deaths), making dementia the second largest cause of death in this age group after ischaemic heart disease. Globally, dementia caused 28·8 million (24·5–34·0) DALYs, making it the 23rd largest cause of DALYs globally in 2016, up from 41st in 1990. Dementia accounted for 1·2% (95% UI 1·0–1·4) of DALYs across all ages. Over the age of 70 years, this increased to 6·3% (5·4–7·5) of DALYs (23·9 million DALYs, 20·1–28·6).

24More women than men died from dementia in 2016 (1·5 million, 95% UI 1·3–1·8 vs 0·8 million, 0·7–1·0). The age-standardised death rates in women were also higher than in men, in line with a higher prevalence in women than in men, indicating the female predominance was not simply due to the longer lifespan of women. Age-standardised global prevalence in females was 1·17 times (1·17–1·18) the age-standardised prevalence in males in 2016, with more women globally affected by dementia (27·0 million, 95% UI 23·3–31·4) than men (16·8 million, 14·4–19·6).

25Both YLLs and YLDs increased sharply with age (figure 2). However, YLL rates increased faster with age and were much higher than YLD rates at the oldest ages. Prevalence also increased substantially with age in both men and women, approximately doubling every 5 years between the ages of 50 and 80 years, after which the increase slowed owing to the high prevalence in the oldest ages (figure 3).

26There was no clear pattern between age-standardised DALY rates for the 21 GBD world regions and SDI over the 1990–2016 estimation period (figure 4). At each SDI level, there was a large amount of heterogeneity between DALY rates due to dementia. There was no uniformity between regions in the changes over time in DALY rates, with some regions increasing and others decreasing over the 1990–2016 period.

27Overall, 6·4 million (95% UI 3·4–10·5) DALYs (22·3% [95% UI 11·8–35·1] of DALYs due to dementia) in 2016 could be attributed to risk factors analysed as part of GBD: 3·5 million (1·4–6·7) DALYs (12·2% [4·7–23·1]) were attributed to high BMI; 2·1 million (0·4–4·9) DALYs (7·4% [1·6–16·1]) to high fasting plasma glucose; 1·3 million (0·6–2·0) DALYS (4·4% [2·0–6·7]) to smoking; and 18·9 thousand (7·8–34·1) DALYs (0·07% [0·03–0·12]) to a diet high in sugar-sweetened beverages. There was no significant difference in the proportion of all dementia DALYs in 2016 explained by risk factors in men (24·7%, 12·7–37·6) compared with women (20·8%, 8·6–35·7). More detailed estimates, including values for all years from 1990–2016, and all locations, all age groups, and both sexes, can be found in our online results tool.

28We estimated that in 2016, 27·0 million (95% UI 23·3–31·4) women and 16·8 million (14·4–19·6) men lived with dementia in the 195 countries and territories that were included in the 2016 round of GBD. The number of prevalent cases of dementia more than doubled from 1990 to 2016, contrasting with relatively minor changes in age-standardised prevalence and pointing to population ageing and growth as the main drivers of the increase. The numbers of DALYs and age-standardised DALY rates showed similar patterns to the prevalence figures. Dementia was the fifth leading cause of cause of death in 2016.

29Our overall global estimate of 43·8 million people living with dementia in 2016 is close to the estimate in the World Alzheimer Report 2015 of 46·8 million for 2015. Additionally, the GBD estimate of a doubling in number of prevalent cases and a 148% (140–157) increase in dementia deaths over the 26-year period from 1990 to 2016 is of the same order as the doubling time of 20 years previously reported., However, the previous studies reported results only at the region level and did not use data across regions to generate estimates for all countries.

30The increase in the number of cases of dementia is of even more importance given that there is currently no effective disease-modifying cure or treatment for the disease. Additionally, even when clinical trials are initiated, many more end in failure than success in bringing a new disease-modifying drug to market, with ratios of more than 100:1 compared with the 14·6:1 pharmaceutical industry average., Without potential treatments, increasing numbers of cases will pose undue burden on individuals who have dementia, their caregivers, and health-care systems more generally.

31In GBD 2016, only four risk factors were judged to have sufficient evidence for a causal link to Alzheimer's disease and other dementias: high BMI, high fasting plasma glucose, smoking, and high intake of sugar-sweetened beverages. In GBD, the effect of high intake of sugar-sweetened beverages was posited to be mediated through BMI on the basis of scientific literature linking BMI with dementia, but sugar-sweetened beverages as such explained only a negligible fraction of dementia burden attributed to risks. GBD is continuously reviewing new evidence for risk-outcome pairs and will update estimates of dementia burden attributable to risks in upcoming cycles for risk factors that meet GBD criteria for causality. A Lancet Commission Report suggested that modifiable risk factors including hearing loss, education, smoking, depression, physical inactivity, social isolation, diabetes, and obesity could account for as much as 35% of dementia burden. In the context of our finding of a doubling of the prevalence of this terminal disease every 5 years over age 50 years, and the absence of a cure, the impetus to examine risk factors is clear.

32Furthermore, the timing of interventions and prevention efforts focused on modifiable risk factors for dementia warrants further investigation. The prodrome of disease is thought to be long, with evidence pointing to a length of 20–30 years, but potentially longer. The inability to identify individuals accurately in the prodromal stage complicates the study of risk factors. Efforts to explore the timing of risk factors during this period are further limited by the span of observational studies and the short length of randomised controlled trials, as well as uncertainty over whether exposures are causative, are bystanders of highly correlated factors, or are even early symptoms of disease. Because of these issues, there are so far no official international lifestyle guidelines for preventing dementia. However, attention on the effects of dementia risks is increasing, and WHO has created a group tasked with the development of risk-reduction guidelines.

33By 2050, the number of people living with dementia could be around 100 million. Tackling this will require training of health professionals, as well as planning and building facilities to cater to increasing numbers individuals with dementia. The cost of care for those living with dementia is also very high, especially in high-income countries. According to recent estimates, in the USA the cost was US$818 million in 2015, an increase of 35% since 2010. Since ageing is expected to continue, the only way to reduce burden and associated costs is to identify effective preventive or treatment measures. Despite the low return on research investment in dementia in the past, the size of the burden and its increasing trend warrant a continued effort to find effective means of intervening. Until such breakthroughs are made, dementia will constitute an increasing challenge to health-care systems across the globe.

34There were several limitations, separate from the broader limitations of GBD, relevant to the modelling process for dementia. First, to correct for changes in coding practices in cause of death data, we selected the countries that were most willing to code to dementia as a cause of death per prevalent case, and we assumed that certification and coding practices in these countries are correct. Although coding practices are probably not perfect in these countries, we selected them as the best available benchmark. Second, in this correction we then assumed that the excess mortality derived from these countries applied to all countries across the entire timeseries. Although this assumption was clearly approximate, it was necessary to address the changes in coding practices that led to large changes over the study period in cause of death data from countries with high-quality vital registration. Third, in correcting for the bias in cause of death data, we relied on prevalence data to determine patterns in geographical distribution of both prevalence and mortality. Although we had a large number of data sources from western Europe, east Asia, high-income Asia-Pacific, and high-income North America, for 13 of the 21 regions we had fewer than five prevalence sources. Fourth, there was a large amount of heterogeneity in the ways in which dementia was diagnosed within the available data. Of the 237 available data sources, 230 different diagnostic procedures were used.

35Although most of the data ultimately classified dementia cases using the DSM or ICD definitions, differences between different versions of the DSM criteria could have led to differences in prevalence estimates, as suggested in a meta-regression analysis of prevalence studies in China. Even if the same screening test was used, different studies often used different cutoff scores. For example, although 42% of studies used the Mini-Mental State Examination in the screening phase, the absolute cutoff scores that were used ranged from 18 to 28, and other studies used different cutoff scores by educational attainment level. In this round of GBD we did not find a way to correct for such bias because of the extreme variability in methods, so it is likely that part of the observed variation in prevalence was due to measurement bias rather than reflecting true geographical variation.

36A potential next step in the GBD is to consider dividing dementia into subtypes, as these might have different epidemiological features and potentially different prevention and treatment strategies. A first subdivision could be Alzheimer's disease dementia, vascular dementia, and remaining types. The challenges of subdividing include sparse data and the complication of how to handle mixed types of dementia. However, increased use of biomarkers in the classification of dementia and Alzheimer's disease might help facilitate subdivision., Additionally, the data on severity distributions over age rely on few data sources and can be strengthened. We also aim to expand our data coverage through increased use of claims data and other data types, including general practitioner data, which have been used to estimate dementia prevalence.,,

37Monitoring trends in dementia is difficult because of the extreme variation in cause of death coding practices and the large heterogeneity in case-ascertainment methods. Although previous guidelines have been developed to systematise the reporting of neurological disorders generally, because of the diagnostic challenges noted with dementia, disease-specific guidelines are warranted and resources should be directed towards creating and implementing more systematic methods. The GBD study will continue to update its estimates for dementia annually, and estimates might become more robust if data collection methods improve. Additionally, as new data become available on risk factors for dementia that meet GBD criteria for causal links, they can be incorporated into future iterations of GBD.