Global, regional, and national burden of multiple sclerosis 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016

1Multiple sclerosis is the most common inflammatory neurological disease in young adults. The mean age of diagnosis is approximately 30 years, with most patients presenting with periodic neurological relapses. One to two decades after onset, many patients with multiple sclerosis enter a progressive phase of the disease. Several environmental factors and genetic alleles might alter the risk of developing multiple sclerosis, but the underlying cause of the disease remains elusive., Common neurological manifestations of multiple sclerosis include optic neuritis, diplopia, sensory loss, limb weakness, gait ataxia, loss of bladder control, and cognitive dysfunction. 15 medications have been approved by the US Food and Drug Administration to reduce the number of relapses and attenuate progression of neurological disability. These drugs are partly effective, but whether they alter the long-term course of multiple sclerosis remains unclear.

2The epidemiology of multiple sclerosis has been the subject of many studies. Over the past five decades, prevalence has been rising across North America and Europe, high incidence has been seen among women and among African Americans, and persistent geographical risk gradients have been documented.,,,, Contemporary reviews have provided updates on the prevalence, incidence, and burden of multiple sclerosis in the Americas, Europe, and worldwide. The MS International Federation has compiled a useful international atlas of the epidemiology of multiple sclerosis with detailed country-level data, but this group did not use systematic modelling for disease morbidity, mortality, and risk factors. The lack of standardisation in the methods of epidemiological studies limits comparisons of findings and assessments of multiple sclerosis with those of other conditions. Additionally, few evaluations have been done at the country level, epidemiological data in different ethnic groups are incomplete, and many regions of the world remain to be studied.

3Evidence before this studyWe searched PubMed for studies published between 2011 and 2015 that were representative, population-based surveys, including retrospective case-report and hospital-report analyses and nationally representative health studies. The search string used was “multiple sclerosis AND epidemiology AND (2011/01/01: 2015/12/31)”. Additional searches for features such as prevalence, disability-adjusted life-years, years of life lost, and mortality are reviewed in the appendix. Global estimates of prevalence, incidence, and deaths from multiple sclerosis have been made in past iterations of the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD). Other epidemiological studies have focused on variations in prevalence and incidence of multiple sclerosis between individual countries and noted marked differences along latitude gradients. Similar data have been recorded on a map produced by the MS International Federation. The GBD study uses an integrated modelling approach to predict morbidity and mortality in countries and regions of the world from where there are insufficient data. Previously, estimates for multiple sclerosis have been summarised in GBD reports about other diseases, but no report has focused exclusively on this disease.Added value of this studyThis Article presents the methods and results for a focused GBD 2016 analysis of multiple sclerosis, making use of detailed country-level morbidity and mortality data. We have explored variation in the burden of multiple sclerosis by country development level, as measured by the Socio-demographic Index, which is a composite indicator of income per person, years of education, and fertility.Implications of all the available evidenceMultiple sclerosis is quite rare but is a potentially severe cause of neurological disability. Onset is typically in young adulthood, after which symptoms persist throughout adult life. Prevalence has increased in many world regions, due in part to improved survival. These findings are relevant to researchers, clinicians, and health policy makers. The health service needs of the increasing number of patients with multiple sclerosis must be considered. Several disease-modifying therapies have been developed that reduce the number of relapses and slow neurological disability, at least during the early phases of the disease. Although expensive, these medications are important components of comprehensive multiple sclerosis care, along with rehabilitation and access to multidisciplinary care. More national, epidemiological studies in multiple sclerosis are needed for GBD to generate more robust worldwide estimates in the future.

4Evidence before this study

5We searched PubMed for studies published between 2011 and 2015 that were representative, population-based surveys, including retrospective case-report and hospital-report analyses and nationally representative health studies. The search string used was “multiple sclerosis AND epidemiology AND (2011/01/01: 2015/12/31)”. Additional searches for features such as prevalence, disability-adjusted life-years, years of life lost, and mortality are reviewed in the appendix. Global estimates of prevalence, incidence, and deaths from multiple sclerosis have been made in past iterations of the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD). Other epidemiological studies have focused on variations in prevalence and incidence of multiple sclerosis between individual countries and noted marked differences along latitude gradients. Similar data have been recorded on a map produced by the MS International Federation. The GBD study uses an integrated modelling approach to predict morbidity and mortality in countries and regions of the world from where there are insufficient data. Previously, estimates for multiple sclerosis have been summarised in GBD reports about other diseases, but no report has focused exclusively on this disease.

6Added value of this study

7This Article presents the methods and results for a focused GBD 2016 analysis of multiple sclerosis, making use of detailed country-level morbidity and mortality data. We have explored variation in the burden of multiple sclerosis by country development level, as measured by the Socio-demographic Index, which is a composite indicator of income per person, years of education, and fertility.

8Implications of all the available evidence

9Multiple sclerosis is quite rare but is a potentially severe cause of neurological disability. Onset is typically in young adulthood, after which symptoms persist throughout adult life. Prevalence has increased in many world regions, due in part to improved survival. These findings are relevant to researchers, clinicians, and health policy makers. The health service needs of the increasing number of patients with multiple sclerosis must be considered. Several disease-modifying therapies have been developed that reduce the number of relapses and slow neurological disability, at least during the early phases of the disease. Although expensive, these medications are important components of comprehensive multiple sclerosis care, along with rehabilitation and access to multidisciplinary care. More national, epidemiological studies in multiple sclerosis are needed for GBD to generate more robust worldwide estimates in the future.

10The Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) provides a systematic method of quantifying health loss in great detail for a given condition by demographic variables and geographical location. By using information from the literature, population statistics, and rigorous algorithms, a consistent picture is produced across the major regions of the world. Comparisons of multiple sclerosis with other common neurological conditions were reported by the GBD 2015 Neurological Disorders Collaborator Group. In 2015, multiple sclerosis ranked tenth for prevalence among neurological conditions measured, with 2 012 000 cases estimated globally. Prevalence of multiple sclerosis and disability-adjusted life-years (DALYs) were significantly higher in women than in men, and there were significant gradients in prevalence and incidence across different regions of the world.

11In this systematic analysis, we present new estimates of the global burden of multiple sclerosis from 1990 to 2016 (measured by prevalence, mortality, DALYs, years of life lived with disability [YLDs], and years of life lost [YLLs]), and assess the relation with development level, as measured by the Socio-demographic Index (SDI). The GBD approach provides a perspective on the burden of multiple sclerosis relevant to researchers, clinicians, and health policy makers.

12The general methods of GBD 2016 are summarised in the appendix, and we present here the methods specific to the estimation of multiple sclerosis. This study complies with the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) recommendations. Further details on GBD 2016 data, analysis, and results are available online.

13The International Classification of Diseases code for multiple sclerosis is 340 in the ninth edition and G35 in the tenth edition. Modelling of causes of death related to multiple sclerosis was based on 13 110 site-years of data (ie, a combination of GBD locations and calendar years of vital registration data) and analysed in the GBD cause of death ensemble modelling tool (CODEm). The model used predictive covariates, including population-weighted average latitude by country, mean cholesterol, the GDB Healthcare Access and Quality Index, the cumulative number of cigarettes consumed over the previous 5 and 10 years, education, lag-distributed income, and SDI (a composite indicator of income per person, years of education, and fertility; appendix). The choice of covariates was based on any reported putative relationship with multiple sclerosis mortality, but does not imply any causal relationships. CODEm is designed to assess which combination of covariates best helps to fit a model to the available input data by use of out-of-sample predictive validity testing.

14The reference case definition of multiple sclerosis for GBD is based on the 2005 McDonald criteria. We also incorporated data with alternative definitions, including the Poser criteria, Schumacher's criteria, Allison and Millar criteria, Bauer criteria, and McAlpine criteria (appendix). Other sources reporting on medical claims data using International Classification of Diseases codes or general practice codes from the UK were also used. From a systematic review, 167 unique sources informing the epidemiological model for multiple sclerosis were identified. We added US medical claims data from 3 years (2000, 2010, and 2012). We found 129 unique sources on prevalence and 65 unique sources on incidence, covering 13 of the 21 GBD world regions (appendix). No data were available for southeast Asia, Oceania, eastern Europe, the Caribbean, Andean Latin America, central sub-Saharan Africa, southern sub-Saharan Africa, and western sub-Saharan Africa. We used the GATHER checklist to define reporting practices for studies, like the GBD study, that calculate health estimates for multiple populations using multiple data sources (appendix).

15Epidemiological outcomes for multiple sclerosis were modelled with DisMod-MR 2.1, a Bayesian meta-regression framework widely used in GBD epidemiological modelling. This framework combines data on prevalence, incidence, remission, and mortality into one model. We set incidence to zero among children younger than 4 years and assumed no remission (ie, no full cure). Study covariates were included to adjust the US claims data for 2000 and 2010, which we found to be systematically lower than claims data for 2012. SDI was included as a covariate on prevalence, average latitude as a covariate on prevalence and incidence, and the GBD Healthcare Access and Quality Index as a covariate on excess mortality (ie, excess deaths in people with multiple sclerosis compared with mortality in people without multiple sclerosis; appendix).

16We identified four separate health states of multiple sclerosis based on disability level: mild, moderate, severe, and asymptomatic. The latter category represents people who have a diagnosis of multiple sclerosis but no overt symptoms. In the GBD Disability Weight Surveys, short lay descriptions were written to capture the main symptoms, reported according to Kurtzke's Expanded Disability Status Scale and assuming that scores of 0 indicate asymptomatic disease, scores greater than 0 up to 3·5 indicate mild multiple sclerosis, those greater than 3·5 and up to 6·5 indicate moderate multiple sclerosis, and scores greater than 6·5 indicate severe disease (appendix). A systematic review of studies with severity data according to the Expanded Disability Status Scale identified 27 unique sources, covering ten of 21 GBD world regions. However, only 14 of these studies separated people with scores of 0, whereas the other studies included scores of 0 in the mild category. To make use of all available data on severity, a two-step meta-analytical approach was developed. First, we did a meta-analysis of studies with data on the proportion of people with asymptomatic multiple sclerosis to determine the proportions of those with asymptomatic or mild multiple sclerosis included in the overall mild category. We did the second meta-analysis on all available data on the proportion of the total mild category (asymptomatic and mild combined) and the proportions of moderate and severe multiple sclerosis. Finally, the proportions of asymptomatic and mild from the first meta-analysis were scaled to the total mild category that was generated in the second meta-analysis to estimate the final proportions for the four health state categories. The prevalence estimates for each severity state were multiplied by the corresponding disability weights to derive uncorrected YLDs. These initial YLD estimates were subsequently corrected for comorbidity with the comorbidity correction simulation developed for GBD.

17Relative risk data were pooled with a meta-analysis of cohort, case-control, and intervention studies. Risks and outcomes were paired, and for each we evaluated the evidence and judged whether the evidence fell into the categories of convincing or probable, as defined by the World Cancer Research Fund. From the prevalence and relative risk results, population-attributable fractions were estimated relative to the theoretical minimum risk exposure level. When we aggregated estimates for clusters of risks, such as metabolic or behavioural risks, we used a multiplicative function rather than simple addition, taking into account how much of each risk is mediated through another risk. The choice of covariates was based on any reported putative relationship with multiple sclerosis morbidity, but does not imply any causal relationships.

18Smoking was the only environmental risk of 84 risks quantified in GBD 2016 that we judged to have sufficient evidence for a causal relationship with multiple sclerosis as an outcome. Population-based surveys were the main source of data for the smoking exposure model. Population-attributable fractions were estimated based on these exposure data, along with data on relative risk and a theoretical minimum level of exposure, which for smoking was no lifetime exposure to tobacco smoke. Criteria for inclusion of risks into GBD include the availability of sufficient evidence for a causal relationship between a risk and one or more disease or injury outcome; evidence to support generalisability of an effect size beyond the populations included in epidemiological studies; availability of sufficient data and methods to enable estimation of exposure levels by country; and the likely importance of a risk factor to disease burden or policy considerations. Additional details on risk factor calculations have been described elsewhere.

19YLLs were calculated by multiplying the number of deaths in each age group by the remaining life expectancy for that age group, as determined by the GBD standard life table. DALYs were calculated as the sum of YLLs and YLDs. Uncertainty was propagated through all calculations by sampling 1000 draws at each step of the calculations, carrying through uncertainty from input data, corrections of measurement error, and estimates of residual non-sampling error. Uncertainty intervals (UIs) were defined as the 25th and 975th values of the ordered draws. The significance of changes between 1990 and 2016 was based on the 95% UI of the change estimate, not including zero.

20The funder of the study had no role in study design, data collection, data analysis, data interpretation or the writing of the report. The corresponding authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

21In 2016, an estimated 2 221 188 people worldwide had multiple sclerosis (95% UI 2 033 866–2 436 858), corresponding to a prevalence of 30·1 cases (95% UI 27·5–33·0) per 100 000 population. Age-standardised prevalence estimates increased by 22·47 cases (95% UI 20·5–24·61) per 100 000 population or 10·4% (9·1–11·8) between 1990 and 2016 (table). Age-standardised prevalence was greater than 120 cases per 100 000 population in North America and some northern European countries, moderate (60–120 per 100 000) in some countries in Europe and Australasia, and lowest (<60 per 100 000 population) in North Africa and the Middle East, Latin America, Asia, Oceania, the Caribbean, and sub-Saharan Africa (figure 1). Of note, in low prevalence regions, many countries had no data and modelling was used to generate estimates (appendix). Detailed estimates on morbidity and mortality by year (1990–2016) and geographical region can be found on the GBD 2016 Global Health Data Exchange. Age-standardised prevalence of multiple sclerosis changed most in the east Asia region (44·8% increase) and Canada (81·9% increase) between 1990 and 2016. The highest age-standardised multiple sclerosis prevalence estimates per 100 000 population were in high-income North America (164·6, 95% UI, 153·2–177·1), western Europe (127·0, 115·4–139·6), and Australasia (91·1, 81·5–101·7), and the lowest were in eastern sub-Saharan Africa (3·3, 2·9–3·8), central sub-Saharan Africa (2·8, 2·4–3·1), and Oceania (2·0, 1·71–2·29)

22We found a gradient for prevalence from low to high between low and high SDI quintiles. The greatest change in multiple sclerosis prevalence between 1990 and 2016 was seen in the middle SDI quintile, with an increase in prevalence of 41·3% (changes in other quintiles were increases of 30·6% in the high quintile, 18·7% in the high-middle quintile, 33·1% in the low-middle quintile, and 23·0% in the low quintile). We found a significant association between prevalence and latitude. The coefficient of average latitude in our DisMod-MR model was 1·03 (95% UI 1·03–1·04) per degree of latitude which translates to an almost nine times difference in prevalence between countries at the equator and the highest population-weighted average latitude of 74·7°.

23The global prevalence of multiple sclerosis differs substantially by sex (figure 2). Among preteen children, the prevalence of multiple sclerosis is similar in boys and girls. During adolescence, the curves start to diverge, with the prevalence increasing more among girls than boys. This pattern continues until around the end of the sixth decade of life, when the sex ratio is 2:1 in favour of women. In older people, prevalence generally continues to climb for women, but a slow attenuation in prevalence is seen for men.

24Globally, there were 18 932 deaths due to multiple sclerosis (95% UI 16 577–21 033) in 2016 (table). Most deaths occurred in high SDI countries (10 021 95% UI 7358–10 873), and mortality in the USA was highest among all countries (3862 deaths, 2743–4226). Between 1990 and 2016, the age-standardised mortality rate for multiple sclerosis globally decreased by 11·5%, but changes by region and country were mostly not significant because of wide uncertainty intervals (table).

25The effect on YLLs due to premature death and disability was greatest in the sixth decade of life (22·05, 95% UI 19·08–25·34) rising steeply beforehand and dropping substantially afterwards (figure 3). For YLDs, the curve rises to a peak at age 55 years, stabilises, then climbs slightly higher during the eighth decade of life and more steeply thereafter (figure 3).

26The global DALYs for multiple sclerosis in 2016 totalled 1 151 478 (95% UI 968 605 to 1 345 776) representing a non-significant decrease of 4·2% (95% UI −16·4 to 0·8) from 1990 (table). Countries in the SDI highest SDI quintile accounted for half of the DALYs in 2016 (577 314 DALYs [50%] of 1 151 478 worldwide). Overall, multiple sclerosis made up 0·04% (95% UK 0·04 to 0·05) of the overall DALYs from all neurological disorders, as shown on the GBD 2016 website.

27The relationship between age-standardised DALYs and SDI over time for each of the 21 GBD regions, represented as annual time series from 1990 and 2016, are shown in figure 4. Age-standardised DALYs rose substantially over time for high-income North America, western Europe, and Australasia. By contrast, in central Europe and eastern Europe age-adjusted DALYs declined while SDI increased. However, among all changes only the increase in high-income North America was significant (table). Among high-income regions, high-income Asia Pacific had DALYs that were much closer to those of low-income and middle-income regions.

28Of the global DALYs due to multiple sclerosis, 99 076 (95% UI 62 164–140 650) or 8·6% (95% UI 5·6–11·7) were estimated to be due to smoking.

29With the rigorous standardised GBD approach for morbidity assessments, we estimated that there were around 2·2 million cases of multiple sclerosis worldwide in 2016. In that year, prevalence was 10·4% higher than in 1990. Multiple sclerosis also contributed 0·04% (95% UI 0·04–0·05) of total DALYs and 0·05% (0·04–0·07) of all YLLs in 2016. Although neurological disability progression is variable, DALYs peaked in the sixth decade of life. Because onset is most frequently in early adulthood and because survival has been improved, people with multiple sclerosis are affected throughout adult life, leading to the high number of YLDs. The disability weights for multiple sclerosis are generally high (appendix) and YLDs begin to increase steeply early in the second decade of life.

30We found a strong latitude gradient for the prevalence of multiple sclerosis, with an increase in prevalence of 1·03 times per degree of latitude. A north to south decrease in prevalence by latitude gradient has been recognised in North America and western Europe,, and a reverse south to north increase in gradient has been reported in Australia. Temporal trends in prevalence show that these gradients were weakening in the 20th century, but the distribution of multiple sclerosis can still be generally described as having three zones of frequency or risk (figure 1), as originally proposed by Kurtzke. In 2016, northern European countries and North America made up the high-risk prevalence zone, with estimates of 100 or more cases of multiple sclerosis per 100 000 population. These regions are bounded by areas of medium frequency (prevalence 30–100 cases per 100 000). Low-frequency areas are centred around the equator, and the prevalence in Asia is less than 30 cases per 100 000 population. Geographical location before onset of multiple sclerosis remains a risk factor for acquisition. Clear gradients from low to high prevalence between low and high SDI quintiles are reported on the GBD 2016 website. The burden of multiple sclerosis was greatest in the regions with the highest socioeconomic status. These gradients point to environmental risk factors that modulate risk based on location.

31Prevalence of multiple sclerosis differed substantially over the adult life of women and men. Rising multiple sclerosis morbidity among women in the later 20th century has been reported in studies of multiple sclerosis prevalence and incidence., An analysis of data from the Danish Multiple Sclerosis Registry showed that incidence in women doubled between 1950 and 2009, whereas increases among men have been more modest. By contrast, excess mortality among patients with multiple sclerosis in Denmark has declined since 1950. Environmental changes that might be contributing to the rapid change in incidence among women include the rise in obesity, increased cigarette smoking, and changes in the frequency of breastfeeding infants.,

32Our results are consistent with some previous reports of multiple sclerosis morbidity from around the world. Canadian studies have indicated the highest multiple sclerosis prevalence estimates so far. In the province of British Columbia, prevalence was 179·9 cases per 100 000 population in 2008. This value is close to our estimate for all of Canada, which fell into the band of 150–180 cases per 100 000 population. However, an estimate of 266·9 cases per 100 000 population in Nova Scotia from 2010 is substantially higher. The prevalence of multiple sclerosis reported from European regions has been variable, with values being higher in northern than in southern regions. For example, national prevalence of 154·5 cases per 100 000 was reported in Denmark in 2005, and an estimate of 230·6 cases per 100 000 population in Northern Ireland was reported in 2008. A meta-analysis suggests prevalence below 100 cases per 100 000 population in France and southern and eastern Europe. Studies of multiple sclerosis prevalence in South America have largely been done in small regions, but a study from Panama produced a national crude prevalence of 5·2 per 100 000, which is notably lower than the estimates for North America. Studies in Africa have been sparse and those from Asia have had variable quality. Nevertheless, our world map of multiple sclerosis prevalence is similar to the 2013 MS International Federation world atlas for multiple sclerosis morbidity by country and region.

33Incidence of multiple sclerosis has been relatively stable or slightly increased over the past four to five decades in white populations, but has been higher in selected racial groups.,, Therefore, the rising prevalence estimates for multiple sclerosis across high-income regions and countries might mostly reflect improved survival. In addition, the diagnostic criteria for multiple sclerosis have evolved and earlier diagnosis is possible with the use of neuroimaging, and these factors are likely to be contributing to the increased prevalence observed.

34Genetic susceptibility to multiple sclerosis is an important factor that influences risk for onset. Multiple sclerosis is considered to be a complex genetic disease, with over 200 alleles having been discovered to contribute small risk effects. Yet the rapid changes by sex, race, and ethnicity in incidence and prevalence over the past few generations give support to environmental factors as drivers of susceptibility to multiple sclerosis. Prime candidates include infection, such as with the Epstein-Barr virus or other organisms, as the initiator of multiple sclerosis, with the suggestion that infections start in the gut and spread to the CNS., Environmental exposures, such as smoking, lack of sunlight, diet, changes in the gut microbiome, and obesity, have also received support as risk factors for onset of multiple sclerosis.

35The GBD 2016 neurological data related to multiple sclerosis provide a pathway for priority setting and service planning in health care in relation to other disorders. The rising cost of multiple sclerosis disease-modifying medications is a major global concern. Ensuring access to disease-modifying medications as well as rehabilitation and multidisciplinary care will help to slow disability progression and support independence in daily activities for patients with multiple sclerosis. Ageing of the large multiple sclerosis population in North America and Europe will be important for health-care providers and policy makers to assess patients' future health-service needs. Ways to provide adaptive work environments and cost-effective nursing care options to patients will be important for policy planners in regions where prevalence is increasing.

36The GBD methods have limitations for the epidemiological assessment of multiple sclerosis. First, data are absent or extremely sparse for many regions of the world, including Latin America, sub-Saharan Africa, and Asia. As such, the models we used to predict prevalence, DALYs, and YLLs might lead to unusual changes in segments of the data. For example, the rapid increases in prevalence in women older than 80 years and the YLD curve we estimated should be viewed with caution because they are not typical of individual population-based studies. We cannot exclude that the relatively low burden of multiple sclerosis in less-developed countries was related to the underdiagnosis of the condition due to limited access to specialised medical care, imaging resources, and laboratory investigations. Additionally, even in high-income regions where multiple sclerosis is well studied, there are few national prevalence and incidence studies, and case-ascertainment infrastructures are limited. For example, in Greenland our estimate for multiple sclerosis prevalence was based on one study in a small community of fewer than 2000 inhabitants, in which no cases of multiple sclerosis were found between 1950 and 1974. The uncertainty around this mean incidence value of 0 per 100 000 incidence is so wide that it is compatible with a high prevalence, as predicted by our latitude covariate. We acknowledge that with additional epidemiological data this high estimate for Greenland might be altered, but until such information becomes available we maintain that the results from our model are valid.

37A second major limitation was the lack of robust predictive covariates for multiple sclerosis to aid in population-based risk assessments, which was due partly to the limited pool of longitudinal neurological disability data that is representative of the multiple sclerosis population. We found a significant relationships between multiple sclerosis, prevalence, latitude, level of development. However, by including both covariates we might have underestimated the effect of latitude because of collinearity between SDI and distance from the equator. Inclusion of SDI as a covariate might also have spuriously led to increasing estimates of prevalence over time. Latitude remains a proxy and, therefore, an uncertain predictor for multiple sclerosis as long as there is no established biological basis for the relationship. With sparse data and uncertain predictive covariates, we cannot exclude that some of the variation measured in prevalence is due to measurement error.

38One way forward for the assessment of the global burden of multiple sclerosis is to use a validated algorithm approach to estimate prevalence and incidence in population-based health care administrative datasets. This approach has been successfully applied to North American multiple sclerosis populations., The availability of medical claims data from the USA makes us more confident of the estimates for that country. GBD is actively seeking access to medical claims data in other countries to improve the accuracy of estimates for diseases such as multiple sclerosis, for which every patient can be expected to be in contact with the health-care system if there are no major barriers to accessing care. Through our network of collaborators, we expect future iterations of GBD to be able to add such sources from other countries.

39In summary, multiple sclerosis is an important cause of neurological disability throughout adult life. This report gives an integrated, contemporary understanding of the global multiple sclerosis disease burden. Prevalence has increased partly due to improved survival. The GBD approach to estimating multiple sclerosis morbidity and mortality is novel and can be repeated with relative efficiency. Our findings will be useful for resource allocation and health services planning for the growing numbers of patients with multiple sclerosis in ageing societies. More national multiple sclerosis epidemiological studies, especially from low-income and middle-income countries, are needed for the GBD Multiple Sclerosis collaborators to generate robust worldwide estimates in the future.