EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

NEUROEPIDEMIOLOGY

0733-8619/96 $0.00

+

.20

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS Brian G. Weinshenker, MD

Multiple sclerosis (MS) is among the most intensively studied neurologic diseases from an epidemiologic point of view. Its inhomogeneous geographic occurrence has suggested that there may be important etiologic significance to its worldwide distribution and that the cause of MS may be discernible through analytic epidemiologic methods.& Studies of migrants have suggested that the risk of developing MS may be changed by migration, implying that environmental factors may be important.z,1s,46 Rare but intensively studied putative epidemics of MS have been interpreted to indicate that an exogenous and transmissible agent might be the cause of MS.47-j9Epidemiologic approaches used in detecting the causal agents in epidemics of infectious disease have been proposed as potentially rewarding strategies. In the face of evidence that MS is not transmissible between spouses,25however, and that adopted relatives of MS patients do not develop MS at a higher rate than expected in the general population,’3 the hypothesis that there is a transmissible MS agent has come under scrutiny. Furthermore, evidence is growing that suggests that genetic factors may influence the risk of developing MS substantially, although genetic factors do not appear to be sufficient and possibly may be unnecessary for the development of MS. Furthermore, the nature of these genetic factors at the molecular level is largely unknown, despite investigation of many candidate genes suspected to be involved in the pathogenesis of MS, primarily those related to immune function. The incidence and prevalence of MS has been changing in many countries in which repeated cross-sectional surveys or ongoing surveillance studies have been conducted. Among the most difficult factors to assess are whether and to what extent these changes might be explained by changes in the nosology of MS, in diagnostic modalities, and in physician and community awareness of MS, particularly of milder cases. Population-based surveys reveal that 20% to 40% of patients might be classified as having benign M S (minimal permanent disability after prolonged observation, generally 10 years or longer).6*,7n, lo3 The

From the Department of Neurology, Mayo Clinic and Mayo Foundation, Rochester, Minnesota

NEUROLOGIC CLINICS

-

-

VOLUME 14 NUMBER 2 MAY 1996

291

292

WEINSHENKER

natural history of MS may be changing as well, although this may be, to a large extent, a function of improved ascertainment, especially of milder cases. The last 50 years have been changing times for MS.4s Changes have occurred in incidence and prevalence rates, in the natural history of disease, in the consolidation of data pertaining to the genetic basis of MS, and in a reassessment of strategies to identify environmental factors that predispose to or trigger the onset of MS. This article reviews selected topics pertinent to both the descriptive and analytic epidemiology the epidemiology of MS.

NOSOLOGY OF MULTIPLE SCLEROSIS Case definition is an important prerequisite of any epidemiologic study. The starting point for the definition of MS is the establishment that the disease is disseminated in time and space. Numerous formal criteria have been proposed to document dissemination; the most widely accepted set of criteria also makes use of cerebrospinal fluid (CSF) immunoglobulin abnormalities to add to the security of diagn~sis.’~ The diagnosis of MS undoubtedly is more secure if these criteria can be fulfilled. Localized syndromes, such as nonstructural progressive myelopathy and optic neuropathy, often due to inflammatory demyelinating disease, may be mimicked by other entities. Nonetheless, formal diagnostic criteria, although improving specificity, may exclude entities that are idiopathic inflammatory demyelinating diseases, which may be indistinguishable pathogenetically from MS.For example, chronic progressive nonstructural myelopathy in many Western countries more commonly is due to idiopathic inflammatory demyelinating disease than to any other cause, even though relapses and remissions do not occur. Acute transverse Myelitis and optic neuritis generally are indistinguishable from the first presentation of MS. After suffering a second demyelinating attack, a patient‘s disease is reclassified as MS, even though it seems improbable that there is any fundamental change in the pathogenesis. Patients with malignant monophasic disease, such as Marburg’s variant of MS or acute disseminated encephalomyelitis, represent difficult problems of classification but, pathologically and by M R imaging, are not easily distinguished from MS.40. 61, 63 Devic’s disease or neuromyelitis optica, a syndrome of near simultaneous optic neuritis and transverse myelitis, is an inflammatory demyelinating disease. It often is excluded from the rubric of MS because of clinical and pathologic differences, including severe spinal cord necrosis with vascular thickening and hyalinization, negative MR image of the head, and absence of oligoclonal bands or immunologic markers of B-cell activation in spinal fluid.h0It may be a monophasic or relapsing remitting disease. It has not been established whether these differences are fundamental in the pathogenesis or simply reflect differences in the course and severity of the syndrome. Because Devic’s disease is an inflammatory demyelinating disease of the CNS involving the optic nerves and the spinal cord, it should be studied as a variant of MS until a more fundamental understanding of the cause of this disease is obtained that will clarify the distinction, if any, from MS. The prevalence and incidence of these variant inflammatory demyelinating diseases of the CNS generally are unknown. Often, these cases are not brought to the attention of MS clinics or MS society chapters, which are often the sources of case ascertainment in epidemiologic studies. Pending further classification of

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

293

TABLE 1. IDIOPATHIC INFLAMMATORY DEMYELINATING DISEASES OF THE CNS Localized or Restricted Distribution

Acute/Monophasic Optic neuritis Acute transverse myelitis Chronic/Recurrent Recurrent optic neuritis Recurrent transverse myelitis Chronic progressive myelopathy Neuromyelitis optics* Disseminated

Acute/Monophasic Acute disseminated encephalomyelitis Marburg’svariant of MS Chronic/Recurrent Relapsing remitting MS (with or without secondary progressive phase) Primary progressive MS (often presents as chronic progressive myelopathy) ‘May be acute/monophasic or chronic/recurrent.

these issues, the author has proposed that these disorders be considered under the general rubric idiopathic inflammatory disorders of the CNS (Table 1). DESCRIPTIVE EPIDEMIOLOGY Incidence and Prevalence

General Considerations

Descriptive epidemiologic studies are essential to identlfy populations with substantially different risk for MS. Such descriptive studies can generate hypotheses to be tested with analytic epidemiology. Often, descriptive studies can reveal important risk factors not evident in case-control studies, particularly those in which the risk factor is homogeneously distributed among cases and controls. In general, MS is most common in Europe, North America, Australia, and New Zealand, with prevalence rates generally greater than 50 per 100,000 population.a It is uncommon (prevalence rate less than 5 per 100,000) in the Far East,’3 and its frequency in Africa78and South America13 appears to be low, although only a limited number of studies have been performed in these areas. The distribution of MS is inhomogeneous in Europe, North America, and Australia. In Europe, the prevalence of MS is highest in the North, particularly in the 82 A series of United Kingdom, Denmark, and most areas of S~andinavia.~~, studies performed in Italy between 1960 and 1975 revealed that the prevalence rate was between 5 and 25 per 100,000. More recent studies conducted since 1980, however, reveal that the prevalence of MS in Italy is between 30 and 60 per 100,000, and the rates in Sardinia are substantially higher (between 60 and 100 per 100,000)?7~28~xz44~91 Recent studies in the south of Spain reveal rates of MS of approximately 50 per 100,000, whereas the rates of MS in some of the same areas surveyed less than 20 years earlier were one third this rate.” Clearly, improved ascertainment, particularly in southern Europe, has played a major role in the changing times of MS prevalence in Europe. A north-to-south gradient of declining prevalence in Europe still seems to be present, although this

294

WEINSHENKER

gradient is much less strong than previously believed. Furthermore, there are substantial regional differences in the occurrence of MS in a number of countries and S~itzerland.~, 39 In including N0rway,2~, 65 the United Kingdom,71,81,lul, lU6 some cases, disability rosters were used to generate the prevalence maps; although incomplete in ascertainment, comparable methodology over a broad area can generate important information about regional differences, as in studies of the Fenno-Scandinavian focus described by K ~ r t z k e ?Another ~ important example is Malta, where in 1978 the prevalence was 4 per 100,000, whereas 3 years earlier, the same senior investigator found that the rate in nearby Enna, Sicily was 53 per 100,000.18, Ioi The rate of MS in a number of communities surveyed in Canada appears to be relatively homogeneous and is approximately 100 per 100,000.31,77*100 In the United States, relatively few epidemiologic studies have been conducted. Kurtzke estimated the prevalence of MS from case-control ratios in World War I1 veterans as ranging from 90 per 100,000 in the northwestern United States to 17 per 100,000 in the southeastern United States.s1At the Mayo Clinic in Olmsted County, Minnesota, the prevalence of MS was tracked over 9 decades. As of December 1, 1991, the gender- and age-adjusted prevalence in the 1950 US white population was 167 per 100,000 total, 243 per 100,000 for women, and 88 per 100,000 in men (M. Rodriguez, personal communication, 1994). In the only prevalence study of MS in South America, the prevalence of MS in Sao Paulo, Brazil was reported to be 4 per 100,000, although the authors acknowledge that ascertainment is almost certainly not c0mp1ete.l~ In Australia, the prevalence of MS increased twofold between 1961 and 1981 in three cities: Perth on the west coast, Newcastle on the east coast, and Hobart, Tasmania in the South.”, The relative differences among the three cities persist, however, varying between 75 per 100,000 in Hobart to 18 per 100,000 in the state of Queensland in the North in the 1981 update. A similar south-to-north gradient of decreasing prevalence was observed in New Zealand?j Prevalence rates in the Far East are uniformly low, between 1 and 4 per 100,000 in Japan, Taiwan, Korea, Malaysia, and 37, 41, 43, -, lu2A recent study in Japan included cases of Devic’s disease in determining the prevalence The rates in Japan had not changed substantially since 1957. In mideastern Arab countries, the reported prevalence rates were between 6 and 8 per 100,000 in Kuwait, Libya, and Saudi Arabia.’,78,

Is the Incidence of Multiple Sclerosis Changing? Many incidence reports are based on cross-sectional ascertainment of cases in which the year of onset is determined by interview at the time of ascertainment. This method often is unreliable because many symptoms of MS are forgotten and patients may have long remissions. Furthermore, when symptoms develop insidiously, determining the onset of MS is difficult. One could imagine how cross-sectional ascertainment to derive incidence rates might have a tendency to produce artifactual epidemics of MS. Therefore, data on changes in incidence of MS are most reliable in areas where MS has been ascertained prospectively by an institution that provides ongoing care to its population base and where access to health care is not limited substantially by socioeconomic or other factors. Kurtzke refers to this approach as the ”spider” study, wherein an institution casts its web in a community for an indefinite peri0d.4~The longest serial follow-up occurred in Olmsted County, Minnesota. The incidence of MS increased by decade in Olmsted County. Over the last 3 decades, the incidence

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

295

rates were 7 to 8 per 100,OOO.lM(M. Rodriguez, personal communication, 1994.) In Sassari, Sardinia, where MS incidence has been tracked in triennia since 1965, the incidence rate increased from 2 per 100,000 from 1965 to 1977 to 5 per 100,000 from 1977 onward.%Doubling of the incidence rate has been reported in More and Romsdal Counties, Norwayz9and in south lower Saxony, Germany.75 In Gothenburg, Sweden, the incidence rates for definite and probable MS declined from a stable rate of 4.2 per 100,000 over successive 5-year periods between 1950 and 1964 to 2.7 per 100,000 and 2.0 per 100,000 over two 5-year periods ending in 1988.98A decline in incidence also was reported in the Orkney Islands between 1951 and 1968.15In Denmark, which has a high crude incidence rate of 4.5 per 100,000, the incidence rate declined between 1952 and 1967 in patients younger than age 35, after which it increased for women and for individuals over the age of 45.“ In summary, there is evidence that the incidence is increasing in some areas, although it is difficult to be certain if this is due to improved ascertainment. Several areas with high incidence rates have shown some decline over time. The significance of fluctuations in incidence encountered in several areas including Newfoundland, Denmark,Q western Norway,”O and northern Germany” is unclear. Does Migration Alter the Risk of Multiple Sclerosis?

If susceptibility to MS is determined by place (i.e., residence and factors therein), and if these factors could be changed before the critical susceptibility period to MS occurs, the incidence of MS might change accordingly. The susceptible period for MS has been defined primarily through migrant studies and based on observations in Israeli migrants and migrants from the United Kingdom to South Africa, in whom the period seems to be between the ages of 10 and 15.*,l7 This estimate of the susceptibility period obtained further support based on an analysis by Kurtzke et a1.&*49 of putative epidemics that occurred on the Faroe Islands, wherein the development of successive epidemics could be predicted retrospectively by assuming that the critical susceptibility period occurred in early puberty. The minimum period of exposure necessary to develop MS is uncertain, but Kurtzke has contended that it is 2 years, an estimate based on his observation of residents of the Faroe Islands, who migrated to Denmark and presumably were exposed to a critical environmental factor in Denmark; 2 years of exposure in Denmark seemed to be the minimum time necessary.*,49 The first migrant study was performed by Alter et a1,2 who studied Jewish immigrants to Israel. Dead7 studied British migrants to South Africa and suggested that migration under the age of 15 was able to shift the risk toward that of the adopted country. Immigrant studies in Hawaii, California, and Washington suggested that white immigrants and their offspring retain a relatively high risk of developing MS and that Asian immigrants retain their relatively low rate of developing MS.3, 2o A recent update on the Israeli migrant study confirms that differences in the occurrence of MS in the offspring of migrants to Israel depend on their ancestral origin; Israeli Jews with European fathers have a slightly, but significantly, higher rate than Israelis with African or Asian This observation provides some support for the importance of genetic factors. Nonetheless, in comparison with their ancestors, there was a significant increase in the rate of occurrence of MS in the offspring of Israelis with Asian or African fathers, suggesting an important environmental influence.= Elian et a P have reported recently that first generation descendants of those who have migrated to the United Kingdom from Africa, the Indian subcontinent, and the West

296

WEINSHENKER

Indies have prevalence rates of MS comparable to that of native-born residents of the United Kingdom. On the surface, this observation appears to argue against an important influence of ethnic and racial factors. Small numbers of subjects, uncertainties regarding the prevalence and precise ethnic mix of the parental generation, imprecise data on the size of the immigrant population at risk, as well as some insecurity about the diagnosis in all cases have prompted some cautionary statements, however." The message from migrant studies seems to be somewhat unclear, with some support for both genetic and environmental factors. Important methodologic considerations limit the interpretation of these studies, including the fact that migrant populations may differ from the populations from which they arise because socioeconomic, ethnic, religious, and other factors may prompt migration and may confound the variation in the prevalence rates. Differing access to medical care and its ascertainment in their home and adoptive countries also may confound the data.19,87 Clusters and Epidemics: Are They Real?

Clusters have been defined by RothmanV5as any excess of disease occurrence. Point prevalence studies are complicated by the fact that date of onset is difficult to establish, and the problem is compounded when patients with long standing MS are asked to recall the onset of early symptoms. Periodic crosssectional analyses often artifactually produce fluctuations in incidence because of factors related to differential recall accuracy of the date of onset according to duration of disease. Furthermore, immigration of the ill for medical care and emigration of the healthy because of job availability can distort the significance of high prevalence rates. Numerous populations have been reported to represent Hordaland, N0rway,5~ clusters of MS, including residents of Key West, and Henribourg, S a s k a t c h e ~ a nand ~ ~ workers in a zinc-related manufacturing plant,'7 among others. The numbers of cases in these clusters has tended to be small, and subsequent intensive investigation of geology, climate, and endemic infectious agents has contributed relatively little to our understanding of MS thus far. Intensive study of community-identified clusters has not been successful in identifymg causal agents. Such studies seem to be most valuable in identifylng new diseases or otherwise rare diseases and least valuable in identifying the cause of existing diseases that are relatively common and seem to OCCUT in apparent e x c e s ~One . ~ of the major difficulties inherent in the analysis of clusters is that they are identified from an infinite number of subgroups from which they are chosen.x7Furthermore, the observation period over which incident cases are identified is short, and the "background frequency" in the communities usually is undefined. Armon et a15 have reviewed how one might discriminate between bonafide clusters and chance by rigorous exclusion of the null hypothesis; this approach is based on correction of the P value for the number of implicit comparisons that are made. For example, if one observes an excess number of cases in a given county in the United States, one must adjust the P value for the number of estimated comparisons in other counties of the United States, where, presumably, a similar excess has not been found. As well as clusters in space, clusters in time have been studied. If one can demonstrate that there is excess temporal clustering in a cohort around a critical age of susceptibility compared with a suitable control group, it may be inferred that a common exposure might have caused or precipitated the disease. R i i ~ e ~ ~ has described such a clustering of a birth cohort around the age of 18 in Norway. Many confounding influences may influence such clustering; for example, indi-

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

297

viduals of higher socioeconomic status are more likely to attend a university; if socioeconomic status were associated with MS, one might expect to see a greater degree of clustering at approximately the age that individuals destined to develop MS enter a university. A small number of putative epidemics of MS have been described. Epidemic can be defined as the occurrence of a group of illnesses of a similar nature, clearly in excess of normal expectancy deriving from a common or propagated source.5xKurtzke et a15* contend that an epidemic occurred in Iceland after World War II. The most intensively studied epidemic is that described by Kurtzke et a14749on the Faroe Islands. The Faroe Islands lie between Iceland and Norway and were formerly a Danish possession. The initial report of 32 cases in native Faroese who had been off the Faroe Islands for less than 3 years before the onset of MS was striking in that none had developed MS before 1943. There have been obvious concerns about ascertainment, but the lead investigator believes that such cases would have been identified by virtue of the excellent neurologic care that was available to the Faroese in Denmark. The incidence rose to a rate of 10 per 100,000 in 1945, a rate that is comparable to that of a high incidence area. Subsequently there were two apparent peaks in the tail of the epidemic occurring in 1955 and 1965. There had been no reported cases of MS in native Faroese since 1973; however, seven new cases of MS recently were reported with symptom onset between 1984 and 1989.s0Kurtzke speculated that these recent cases may constitute a new epidemic. An analysis of successive birth cohorts on the Faroes revealed that the initial epidemic involved postpubescent individuals who were 11 years or older in 1941, and subsequent epidemics involved those who were prepubescent or younger than 11 years of age.4s,49 Several investigators including Poser et a17*,73 and Benedikz et al*,72 have argued that the apparent epidemic in the Faroes might be artifactual. They have argued that the date of clinical onset may be unreliable in determining the onset of disease. The small number of cases limits the power to differentiate between a true excess and a chance occurrence. Substantial concern exists regarding the ascertainment of cases before World War 11. The occurrence of epidemics would be strong evidence in favor of a transmissible agent as a cause of MS. The number of epidemics observed, however, has been very limited. The proponents of an epidemic on the Faroes contend that this is a unique or almost unique situation, in which a virgin, susceptible population was exposed to a ubiquitous infectious agent able to cause MS. Substantial uncertainty exists regarding whether intensive investigation in the Faroes ultimately will identify an infectious agent responsible for MS. Ethnic and Familial Studies

Substantial evidence has accumulated over the last decade in support of an important influence of genetics on susceptibility to MS." To address these issues fully, the following questions must be addressed: (1) Is susceptibility to MS inherited?; (2) What is the pattern of inheritance?; (3) Which genes are involved?; and (4) Are there gene-gene or gene-environmental interactions? Although the answer to the first question seems to be unequivocally "yes," no genetic polymorphism has been identified yet that is either sufficient or necessary to cause MS. Furthermore, reproducibility in identifying various genes as being linked with MS susceptibility in multiplex families or associated with MS risk in the general population has not been good, with the exception of class I1 major histocompatibility complex (MHC) genes and MS.

298

WEINSHENKER

The evidence in support of genetic susceptibility to MS comes from a number of sources. Twin studies with relatively complete ascertainment have shown that there is roughly a tenfold increase concordance rate in monozygotic versus dizygotic twins, a conclusion that supports an important genetic influe n ~ eA. ~single ~ dissenting study from Francez6is complicated by the fact that ascertainment was based on self report and by inadequate ascertainment of disease status in a substantial proportion of the cotwins. In the Canadian study, even after prolonged observation when the majority of cotwins should have passed their risk period for developing MS, the concordance rate does not exceed 40%,8' suggesting that there is an important environmental factor responsible for the occurrence of MS. A number of resistant populations have been described. Many of these populations are the indigenous populations of countries now populated extensively by northern Europeans. For example, the Aborigines of Australia, the Maori of New Zealand, and Native Americans seem to be resistant to MS.22 Some European populations that are genetic isolates, such as the Hutterites of Saskatchewan, seem to be similarly resistant to MS, although they live in an endemic region for MS." Recently, it has been shown that Scandinavian background alone can account for most of the variation in the distribution of MS across the United A number of studies have been performed to identify the recurrence risk of MS in relatives. The most complete of these studies are those of Sadovnick et a1,88 which have detailed the empiric recurrence risk in relatives of different relatedness according to whether the index patient is male or female. For firstdegree relatives, the risk is approximately 4% for a sibling and 2% to 4% for a child, depending on the gender of the child. An analysis of the empiric recurrence risk eliminates autosomal recessive inheritance given the frequency of affected parents. X-linked recessive inheritance is eliminated by the female-tomale gender ratio of 2:l in affected individuals. Male-to-female transmission excludes mitochondria1 inheritance. X-linked dominant inheritance is not favored by the equal ratio of paternal-to-maternal affected uncles, whereas the disease would be expected to be twice as frequent in maternal as compared to paternal uncles if X-linked dominant inheritance were present.2z,86MS is suspected to be a multigenic disease, with an important component of genetic/ environmental interaction, which would account for the large number of discordant monozygotic twin pairs. An analysis of the monozygotic-to-dizygotic concordance ratio would suggest that several genes must be involved. For example, if a single dominant gene were to account for MS susceptibility, the ratio of concordance between monozygotic and dizygotic twins would be 2:1, and if a single recessive gene were operative, the ratio of concordance would be 4:l. In pooled data of same-gender twins, the observed ratio of monozygotic to dizygotic concordance is 8:1.86By dividing the concordance rate in relative pairs by the rate of occurrence of MS in the general population, RischSohas derived a ratio known as A. The rate at which the h ratio declines according to the degree of relatedness points to either a multiplicative model of a multigenic disease with epistatic interaction among different loci or an additive model without locus interaction. An analysis of the A ratio in MS points to epistatic interaction.% It is hoped that molecular genetic analysis will shed light on the genes responsible for MS. Although it is unclear how such studies will influence therapy, delineating genes responsible for MS susceptibility by such an analysis may help to clarify the pathogenesis of MS. Two major approaches have been undertaken. The first approach is based on an analysis of candidate genes suspected to be important based on their function and association with MS. For example, it has been known through serologic analysis that HLA DR2 is four to

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

299

five times more prevalent in northern European MS patients compared with northern European controls and perhaps is linked to MS susceptibility in multiplex families.35,67 A second, more recent approach is to use multiple anonymous markers spanning the entire genome without a specific hypothesis as to which candidate genes may be important. The advantage of the latter approach is that it provides a systematic, unbiased survey of the entire genome. Linkage studies can be model-dependent in that they require assumptions about the number of loci and mode of inheritance and penetrance (facts that generally are unknown for MS), or model-independent in that no assumptions are made about these factors. In a model-independent approach, the frequency of haplotype sharing between affected sibling pairs is studied. If the frequency of haplotype sharing departs from the expected, which is 1:2:1 for sharing 2, 1, or 0 haplotypes, the importance of the candidate gene is supported. Although the model-independent approach has the advantage of not specifying a model, each individual sibling pair offers relatively little information, and a great many sibling pairs are necessary. Furthermore, if a genetic marker is not highly informative (i.e., has few alleles), it may be difficult to distinguish whether siblings are identical by descent, which is the feature of interest, or merely identical by state, because the haplotype is indistinguishable at the marker locus due to inadequate polymorphism. Many factors have complicated this form of molecular epidemiologic analysis. There is substantial clinical heterogeneity of MS as discussed previously. It has been recognized recently that some individuals may be affected subclinically by MS, which can confound this form of analysis. This is evident by the frequency with which clinically unaffected cotwins manifest MR imaging changes typical of MS.s9 It is quite conceivable that MS is also genetically heterogeneous. The control groups also pose significant difficulties in this type of analysis. The frequency of a polymorphism in a control group is a major determinant of the odds ratio. Ethnic heterogeneity in the control group is, therefore, a substantial confounder in population studies. Finally, this approach is dependent on the presence of linkage. If several polymorphisms of a given gene predispose to MS and if these polymorphisms were linked with different allelic variations in a marker locus, linkage might not be observed even in the case of a bonafide susceptibility locus. Recently, molecular approaches have been developed that involve screening by direct genomic sequencing of highly suspect candidate genes for polymorphisms associated with disease susceptibility in multigenic diseases.96 ANALYTIC EPIDEMIOLOGY

Analytic epidemiology uses epidemiologic techniques to address etiologic hypotheses. There are several different approaches that have been used for MS. Ecologic or correlational studies have been used to identify causal factors by seeking to demonstrate a correlation between the occurrence of MS and various geoclimatic (e.g., natural radiation, climate, rainfall, soil content) and sociocultural (e.g., industry, occupation, diet) factors.56A second approach is the casecontrol study, in which cases are compared with a suitable control population for the risk factor of interest. A third approach is the cohort approach. Prospective cohort studies are the strongest methods for proving an association between a putative causal factor and a given disease. The frequency of disease in an exposed and an unexposed cohort is compared over time. Only a single exposure can be studied, however; exposure to many of the putative factors that may

300

WEINSHENKER

contribute to MS susceptibility is difficult to control. Because there is a long latent period until MS is expressed, prospective cohort studies tend to be unrealistic. A different cohort study, a historical (retrospective) cohort study, is feasible if a number of cohorts, among them an MS cohort, have been identified in a single community. Historical cohort studies search for an excess of overlapping cases between two cohorts, for example, a cohort exposed to a given risk factor and a cohort of MS patients. Several cohorts may be looked at in this way, but this approach is dependent on prior identification and follow-up of such cohorts. These four approaches are discussed, and examples of how they have been used in MS epidemiology are provided in the following sections. Ecologic Studies

Ecologic studies are particularly valuable in the setting of MS given its major global and within-country variability of occurrence. Ecologic studies are of particular value when the variables of interest have relatively small variation within a small subpopulation, a factor which limits the utility of a case-control approach?' There are several deficiencies of the ecologic approach. It is difficult to document exact exposure, and one often needs to rely on proxy measures instead of accurate exposure data. For example, to assess meat consumption, the density of butchers may be the measure considered. Numerous sources of confounding are present. For example, an association of solar exposure or temperature may be related to latitude. Ethnic background may be important in MS susceptibility. Individuals of a Scandinavian background tend to live in more northern, colder climates. A primary association of ethnic background may go undetected in a correlational study with climatic conditions. Ecologic studies are most persuasive if their conclusions can be replicated in several different studies using different populations. Demonstration of a dose-response gradient by dividing the measure of exposure into several categories may reveal correlations not otherwise apparent. Nonlinear effects of exposure nonetheless do occur. Multivariate techniques, including factor analysis and stepwise regression analysis, can be used to identify critical factors when multiple coassociations The methods and results of correlational studies in MS have been reviewed extensively by L a ~ e rAs . ~ summarized ~ by Lauer, numerous studies document a positive association between low temperature and the occurrence of MS. A majority of studies suggest a positive association between humidity and rainfall and MS. Analysis of winter conditions has suggested that winter precipitation is more strongly associated than total annual precipitation. Presence of coniferous forest and the content of peat in soil also have been correlated with the occurrence of MS. Industrialization, in general, has been positively associated in some studies with the occurrence of MS, but no specific industry has been associated strongly. Several different dietary associations have been recognized. A study by Swank et aly9suggested higher rates of MS in those living in meat consuming areas. A relationship also has been suggested between consumption of dairy food and the occurrence of MS.5' The ingestion of processed meats and smoke cured meats has been associated with the risk of acquiring MS in some ~tudies.5~

Case-control Studies Although ecologic studies are ideally suited to address putative agents or effects that alter the risk for an entire population (or subpopulation within a

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

301

given country or region), a case-control study is best suited to address risk factors that differ among individuals within a population. The case-control approach is very robust and can be used to address a broad variety of factors, including residence, diet, ethnic background, recent or remote infections, and occupational and toxic exposures. Furthermore, a case-control study can be used to address several putative risk factors simultaneously, assuming that there is no bias for one of the risk factors being assessed in the selection of the control group. There are many deficiencies of a case-control study, however. Selection of the control group and accounting for systematic differences such as ethnic background are critical. Generally, several control groups are necessary, including those matched for ethnic background and others matched for other neurologic or autoimmune diseases. There is a well-known bias for cases to have selectively better recall than controls, in that cases commonly search for an explanation for their illness and are more likely to identify the presence of a random risk factor than are controls. Finally, if many risk factors are assessed simultaneously, some correction or allowance should be made for the number of factors being analyzed. Rigorous methodology can overcome some difficulties inherent in casecontrol studies. Rigorous specification of the protocol is essential; ideally, the interview should be administered by trained interviewers who are blinded to the study hypothesis. Questions regarding exposure should be verified for their accuracy. For example, prior to embarking on a study in which patients are asked to recall the age at which they had measles, there should be some method to verify that this approach generally produces valid answers. The time at which important exposures should be evaluated also is problematic. Clearly, an exposure is relevant only if it occurs prior to the onset of MS. Current thinking suggests that there is a latent period of a decade or longer from exposure to onset of MS, based on studies of migrants that estimate that critical exposure occurs at approximately the age of 15 years. Restricting inquiry to an appropriate risk period might improve the specificity of a case-control study if one accepts the correctness of assumptions about the critical period of exposure. Generally, the factors that have been assessed in case-control studies in MS have been focused on infections, primarily because of a hypothesis, highly favored by some, that MS is triggered by an infectious disease. Such hypotheses have been formulated initially based on both observation of spontaneous inflammatory demyelinating diseases in various animal models and on interpretations about the descriptive epidemiology of MS, in particular the geographic variation among and within different countries and the existence of putative epidemics. Canine distemper virus has been and still is under investigation, particularly by Cook et a1.I6 Distemper can produce a spontaneous CNS inflammatory demyelinating disease in a variety of species. Some serologic evidence has pointed to the occurrence of distemper infection in humans, particularly those with MS, but because of cross-reaction with measles antibodies, this evidence has been difficult to interpret. Indoor dog density apparently is greater in Europe and North America than it is in various parts of Asia, where MS is less common. Furthermore, dogs are kept indoors more often in colder weather, and an excess number of MS relapses and new cases of MS have occurred in winter months. In several geographic areas, MS has been reported to increase in incidence following attacks of canine distemper. In some areas, such as the Faroes, KurtzkeS3has not been able to substantiate a relationship between canine distemper and MS. Based on these correlational studies, 21 case-control studies have been

302

WEINSHENKER

performed, 7 of which show more exposure to dogs, indoor dogs, small dogs, or dog exposure 5 to 10 years before the onset of MS in cases versus controls. Cook et all6 has argued that these studies are relatively insensitive because of the high baseline rate of exposure to dogs in the population. Case-control studies using dog exposure as a proxy for canine distemper virus could be improved substantially if a reliable method were developed to identify canine distemper infection in humans and to distinguish this from measles virus infection. Also, analysis of the occurrence of MS in areas where dog exposure is somewhat less common may result in a greater degree of sensitivity in case-control studies. Elevated titers of antibodies to the measles virus were detected in the blood of MS patients; however, no convincing differences were found when appropriate controls, including family members, were studied.66A number of studies have suggested that individuals who develop MS acquire measles at a later age than do those who do not develop MS. Alter et a14have hypothesized that the epidemiology of MS suggests that a viral infection early in life may protect against MS, whereas later development of a viral infection may predispose to autoimmune disease, perhaps because the host’s immune response is aberrant. This hypothesis is analogous to the one proposed by P ~ s k a n z e r ~ ~ known as the “polio hypothesis,” a name that refers to the resistance to severe paralytic poliomyelitis when infection is acquired in early life. Some have speculated that effective measles vaccination may decrease the incidence of MS. Svenningsson et aP8 reported a decrease in the incidence of MS in the Gothenburg region of Sweden over 30 years and speculated that measles eradication may account for this observation. A recent report by Bansil et a16 documented that a vast majority of young MS patients had been immunized to measles virus. Genetic studies examining the role of markers linked to candidate genes essentially are case-control studies methodologically. The approach used in such studies was discussed previously. Cohort Studies

Few prospective cohort studies have been conducted in MS. One cohort study in patients already diagnosed with MS, conducted by Sibley et a1,93 reported that minor respiratory infections preceded 27% of MS relapses and that this was the only significant factor associated with occurrence of MS exacerbations. Two prospective studies that address the effect of pregnancy on MS’, y(l dispute previous observations that the frequency of MS exacerbations is increased in the postpartum period.’” These recent clinical observations are supported by prospective MR imaging studies in a small number of patients; such studies show a marked decrease in the number of new lesions during pregnancy, using M R imaging.Io4 Studies addressing the risk factors that ultimately may lead to the development of MS cannot be studied realistically in a prospective cohort study, given the suspected long interval between exposure and onset, the lack of widely accepted risk factors, and the difficulties in controlling exposures. Historical cohort studies can be accomplished in centers that have a number of defined cohorts, in particular, a cohort of MS patients and a cohort(s) of individuals who have been exposed to a putative risk factor. Such a study has been conducted recently at the Mayo Clinic to address the hypothesis that trauma is able to cause or trigger attacks of MS?* Two mechanical trauma cohorts in Olmsted County, Minnesota, were analyzed at the Mayo Clinic. The first consisted of a

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

303

head injury cohort (n = 819), and the second consisted a cohort undergoing lumbar disc surgery (n = 942). None of the patients in the head injury cohort developed MS within 6 months of their trauma. Of five patients in the disc surgery cohort with MS, MS preceded the surgery in four of five. Fifty-four episodes of trauma occurred in 39 patients in the MS prevalence cohort; these traumatic episodes were primarily small bone fractures. Comparing the group with trauma to the group without trauma, there was no difference in the mean disability score, 'nor was there any difference in the exacerbation rate in the individuals who experienced trauma in the 6 months prior to the traumatic episode compared with the 6 months following the episode. Although the power of such historical cohort studies is limited, the absence of even one case of MS in 800 individuals in a susceptible age within 6 months after significant head trauma suggests that head trauma is not a major causal factor for MS.

SUMMARY The inhomogeneous distribution of MS may represent an important clue about the cause of the disease. The prevalence of MS appears to be increasing in many regions of the world, although it is likely that in a majority of regions surveyed, this increase is attributable to differences in ascertainment over the relatively short period that M!3 has been surveyed worldwide, which is almost exclusively the last 50 years. MS occurs most frequently in regions populated by northern Europeans, and in these areas, it is much more common in the northern European population than in the indigenous populations. MS occurs in relatives of MS patients at a rate 10- to 50-fold greater than in the general population. On average, the absolute risk to a first-degree relative of an MS patient is 2% to 5%, depending on the exact degree of relatedness. Although the mode of transmission is not certain, a multigenic pattern best fits the observed pattern of transmission. A majority of monozygotic twins are discordant for MS, indicating that a major component of MS susceptibility is environmentally determined, although recent observations in adopted relatives suggest that this risk is not transmissible. Rare epidemics of MS involving small numbers of individuals in geographically isolated regions have been reported. A number of migrant studies suggest that the risk of acquiring MS may be altered by migration, although some also support genetic factors. Analytic approaches to address the role of various risk factors include ecologic studies, case-control studies, and cohort studies. Ecologic studies are best suited for addressing risk factors that influence overall population risk, whereas case-control studies are best suited for addressing risk factors that pertain to individuals within a population. If a risk factor is distributed homogeneously in a given population, a case-control study is insensitive to its effect. Improved methodology to diminish the biases inherent in case-control studies and identification of other important risk factors from basic scientific studies undoubtedly will be important for analytic epidemiologic studies of the future. Furthermore, analysis of highly informative populations, such as discordant identical twins and adoptive siblings of MS patients, likely will improve the specificity of case-control studies by minimizing the vast number of potential differences between cases and controls.

304

WEINSHENKER

ACKNOWLEDGMENTS Theresa Hanson typed the manuscript.

References 1. Al-Din ASN: Multiple sclerosis in Kuwait: Clinical and epidemiological study. J Neurol Neurosurg Psychiatry 49:928-931, 1986 2. Alter M, Leibowitz U, Speer J: Risk of multiple sclerosis related to age at immigration to Israel. Arch Neurol 15234-237, 1966 3. Alter M, Okihiro M, Rowley W, et al: Multiple sclerosis among Orientals and Caucasians in Hawaii. Neurology 21:122-130, 1971 4. Alter M, B a n g ZX, Daoanipour 2, et al: Multiple sclerosis and childhood infections. Neurology 3633861389,1986 5. Armon C, Daube JR, OBrien PC, et al: When is an apparent excess of neurologic cases epidemiologically significant? Neurology 41~1713-1718,1991 6. Bansil S, Troiano R, Dowling PC, et al: Measles vaccination does not prevent multiple sclerosis. Neuroepidemiology 9:248-254, 1990 7. Bartsclu-Rochaux W MS in Switzerland-Canton Wallis. In Bauer JH, Poser S, Ritter G (eds): Progress in Multiple Sclerosis. Berlin, Springer-Verlag, 1980, pp 535-538 8. Benedikz J, Magnusson H, Gudmundsson G: Multiple sclerosis in Iceland, with observations on the alleged epidemic in the Faroe Islands. Ann Neurol 36(suppl 2):S175-S179, 1994 9. Bernard S, Grasso MG, Bertolhi R, et al: The influence of pregnancy on relapses in multiple sclerosis: A cohort study. Acta Neurol Scand 84:403406, 1991 10. Birk K, Ford C, Smeltzer S, et al: The clinical course of multiple sclerosis during pregnancy and the puerperium. Arch Neurol47738-742,1990 11. Bufill E, Blesa R, Galan I, et a1 Prevalence of multiple sclerosis in the region of Osona, Catalonia, Northern Spain. J Neurol Neurosurg Psychiatry 58:577-581, 1995 12. Bulman DE, Ebers GC: The geography of MS reflects genetics susceptibility. Journal of Tropical and Geographical Neurology 2:66-72, 1992 13. Callegaro D, de Lolio CA, Radvany J, et a1 Prevalence of multiple sclerosis in the city of Sao Paulo, Brazil, in 1990. Neuroepidemiology 11:ll-14, 1992 14. Compston DAS: Risk factors for multiple sclerosis: Race or place [editorial]. J Neurol Neurosurg Psychiatry 54:821-823,1991 15. Cook SD, Cromarty JI, Tapp W, et al: Declining incidence of multiple sclerosis in the Orkney Islands. Neurology 35:545-551, 1985 16. Cook SD, Rohowsky-Kochan C, Bansil S, et al: Evidence for multiple sclerosis as an infectious disease. Acta Neurol Scand Suppl 161:3442, 1995 17. Dean G Annual incidence, prevalence, and mortality of multiple sclerosis in white South African-born and in white immigrants to South Africa. BMJ 2:724-730,1967 18. Dean G, Grimaldi G, Kelly R, et al: Multiple sclerosis in southern Europe: I. Prevalence in Sicily in 1975. J Epidemiol Community Health 33:107-110, 1979 19. Delasnerie-Laupretre N, Alperovitch A Migration and age at onset of multiple sclerosis: Some pitfalls of migrant studies. Acta Neurol Scand 8540W11, 1992 20. Detels R, Visscher B, Malmgren RM, et al: Evidence for lower susceptibility to multiple sclerosis in Japanese-Americans. Am J Epidemiol 105:303-310, 1977 21. Ebers G C Genetic factors in multiple sclerosis. Neurol Clin 1:645-654, 1983 22. Ebers GC, Sadovnick A D The role of genetic factors in multiple sclerosis susceptibility. J NeuroimmunoI54:1-17, 1994 23. Ebers GC, Sadovnick AD, Risch N: A genetic basis for familial aggregation in multiple sclerosis. Nature 377:150-151, 1995 24. Elian M, Nightingale S, Dean G. Multiple sclerosis among the United Kingdom born children of immigrants from the Indian Subcontinent, Africa and the West Indies. J Neurol Neurosurg Psychiatry 53:906-911, 1990 25. Fredrikson S, Michelsberg J, Hillert J, et al: Conjugal multiple sclerosis: Immunoge-

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

305

netic characterization and analysis of T- and B-cell reactivity to myelin proteins. Neurology 42:577-582, 1992 26. French Research Group on Multiple Sclerosis: Multiple sclerosis in 54 twinships: Concordance rate is independent of zygosity. Ann Neurol32:724-727,1992 27. Granieri E, Rosati G, Tola R, et al: The frequency of multiple sclerosis in Mediterranean Europe: An incidence and prevalence study in Barbagia, Sardinia, insular Italy. Acta Neurol Scand 68:84-89, 1983 28. Granieri E, Tola R, Paolino E, et al: The frequency of multiple sclerosis in Italy: A descriptive study in Ferrara. Ann Neurol 1780-84, 1985 29. Gronning M, Mellgren SI: Multiple sclerosis in the two northernmost counties of Norway. Acta Neurol Scand 72321-327, 1985 30. Gronning M, Rise T, Kvale G, et al: Incidence of multiple sclerosis in Hordaland, Western Norway: A fluctuating pattern. Neuroepidemiology 10:53-61, 1991 31. Hader WJ, Elliot M, Ebers G C Epidemiology of multiple sclerosis in London and Middlesex County, Ontario, Canada. Neurology 38:617-621, 1988 32. Hader WJ, Irvine DG, Schiefer HB: A cluster-focus of multiple sclerosis at Henribourg, Saskatchewan. Can J Neurol Sci 7391-394, 1990 33. Hammond SR, de Wytt C , Maxwell LC, et al: The epidemiology of multiple sclerosis in Queensland, Australia. J Neurol Sci 80:185-204, 1987 34. Hammond SR, McLeod JG, Millingen KS, et al: The epidemiology of multiple sclerosis in the three Australian cities: Perth, Newcastle and Hobart. Brain 111:l-25, 1988 35. Hillert J, Kall T, Vrethem M, et al: The HLA-DW2 haplotype segregates closely with multiple sclerosis in multiplex families. J Neuroimmunol5095-100, 1994 36. Hung T-I? Multiple sclerosis in Taiwan: A reappraisal. In Kuroiwa Y, Kurland LT (eds): Multiple Sclerosis: East and West. Kyushu, Japan, Kyushu University Press, 1982, p p 8S96 37. Hou JB, Zhang ZX: Prevalence of multiple sclerosis: A door to door survey in Lan Cang La Hu Zu Autonomous County, Yunnan Province of China. Neuroepidemiology 11:52, 1992 38. Kahana E, Zilber N, Abramson JH, et al: Multiple sclerosis: Genetic versus environmental etiology in Israel updated. J Neurol214:341-346, 1994 39. Kesselring J: High prevalence of multiple sclerosis in Switzerland [abstract]. Neurology 37(suppl 1):151, 1987 40. Kesserling J, Miller DH, Robb SA, et al: Acute disseminated encephalomyelitis: MRI findings and the distinction from multiple sclerosis. Brain 113:291-302, 1990 41. Kim SW, Kim SY: Multiple sclerosis in Busan, Korea: Clinical features and prevalence. In Kuroiwa Y , Kurland LT (eds): Multiple Sclerosis: East and West. Kyushu, Japan, Kyushu University Press, 1982, pp 57-69 42. Koch-Henriksen N, Bronnum-Hansen H, Hyllested K: Incidence of multiple sclerosis in Denmark 1948-1982 A descriptive nationwide study. Neuroepidemiology 11:l10, 1992 43. Kuroiwa Y, Shibasaki H, Ikeda M: Prevalence of multiple sclerosis and its north-tosouth gradient in Japan. Neuroepidemiology 262-69, 1983 44. Kurtzke JF Epidemiology of multiple sclerosis. In Vinken PJ, Bruyn GW, Klawans HL (eds): Handbook of Clinical Neurology, vol 3, no. 47. Amsterdam, Elsevier, 1985, pp 259-287 45. Kurtzke JF Multiple sclerosis: Changing times. Neuroepidemiology 101-8, 1991 46. Kurtzke JF, Bui QH: Multiple sclerosis in a migrant population: 11. Half-Orientals immigrating in childhood. Ann Neurol8:25&260,1980 47. Kurtzke JF, Hyllested K Multiple sclerosis in the Faroe Islands I. Clinical and epidemiological features. Ann Neurol5:&21, 1979 48. Kurtzke JF, Hyllested K Multiple sclerosis in the Faroe Islands 11. Clinical update, transmission, and the nature of MS. Neurology 36:307-328, 1985 49. Kurtzke JF, Hyllested K: Multiple sclerosis in the Faroe Islands 111. An alternative assessment of the three epidemics. Acta Neurol Scand 76317-339,1987 50. Kurtzke JF, Hyllested K. Multiple sclerosis in the Faroe Islands and the lack of protection by exposure in infancy. Neuroepidemiology 11:90-99, 1992

306

WEINSHENKER

51. Kurtzke JF,Beebe GW, Norman JR Epidemiology of multiple sclerosis in US veterans: 1. Race, sex and geographic distribution. Neurology 29:1228-1235, 1979 52. Kurtzke JF, Gudmundsson KR, Bergmann S: Multiple sclerosis in Iceland 1. Evidence of a post-war epidemic. Neurology 32143-150,1982 53. Kurtzke JF, Hyllested K, Arbuckle JD, et al: Multiple sclerosis in the Faroe Islands IV.The lack of a relationship between canine distemper and the epidemics of multiple sclerosis. Acta Neurol Scand 78:484-500, 1988 54. Larsen JP, Aarli JA, Riise T: Western Norway, a high risk area for multiple sclerosis: A prevalence/incidence study in the county of Hordaland. Neurology 341202-1207, 1984 55. Larsen JP, Riise T, Nyland H, et al: Clustering of multiple sclerosis in the county of Hordaland, Western Norway. Acta Neurol Scand 71:390-395, 1985 56. Lauer K Environmental associations with the risk of multiple sclerosis: The contribution of ecological studies. Acta Neurol Scand Suppl 161:77-88, 1995 57. Lauer K Multiple sclerosis in relation to meat preservation in France and Switzerland. Neuroepidemiology 8:308-315,1989 58. MacMahon B: Epidemiological methods. In Clark DW, MacMahon B (eds): Preventitive Medicine. Boston, Little Brown, 1967, pp 81-104 59. Malosse D, Perron H, Sasco A, et a1 Correlation between milk and dairy product consuption and multiple sclerosis prevalence: A worldwide study. Neuroepidemiology 11:304-312,1992 60. Mandler RN, Davis LE, Jeffrey DR, et a1 Devic’s neuromyelitis optica: A clinicopathological study of eight patients. Ann Neurol 34:162-168, 1993 61. Marburg 0:Die sogenannte “akute multiple sklerose.” Journal of Psychiatric Neurology 27211-312, 1906 62. McAlpine D: The benign form of multiple sclerosis: A study based on 241 cases seen within 3 years of onset and followed up until the 10th year or more of the disease. Brain 84186-203, 1961 63. Mendez MF, Pogacar S: Malignant monophasic multiple sclerosis or “Marburg’s disease.” Neurology 381153-1155, 1988 64. Meyer-Rienecker H, Buddenhagen F Incidence of multiple sclerosis: A periodic or stable phenomenon. J Neurol235:241-244, 1988 65. Midgard R, Riise T, Nyland H Epidemiologic trends in multiple sclerosis in More and Romsdal, Norway: A prevalence/incidence study in a stable population. Neurology 41:887-892, 1991 66. Norrby E: Viral antibodies in multiple sclerosis. Prog Med Virol 24:l-39, 1978 67. Olerup 0, Hillert J: HLA class 11-associated genetic susceptibility in multiple sclerosis: A critical evaluation. Tissue Antigens 38:l-15, 1991 68. Page WF, Kurtzke JF, Murphy FM, et al: Epidemiology of multiple sclerosis in US. veterans: V. Ancestry and the risk of multiple sclerosis. A m Neurol33:632-639,1993 69. Park CS: Multiple sclerosis in Korea. Neurology 16:919-926, 1966 70. Phadke JG: Clinical aspects of multiple sclerosis in the North East Scotland with particular reference to its course and prognosis. Brain 1131597-1628,1990 71. Phadke JG, Downie AW: Epidemiology of multiple sclerosis in the north-east (Grampian Region) of Scotland: An update. J Epidemiol Community Health 415-13, 1987 72. Poser CM, Hibberd PL: Analysis of the ”epidemic” of multiple sclerosis in the Faroe Islands 11. Biostatistical aspects. Neuroepidemiology 7181-189, 1988 73. Poser CM, Hibberd PL, Benedikz J, et al: Analysis of the “epidemic” of multiple sclerosis in the Faroe Islands I. Clinical epidemiological aspects. Neuroepidemiology 7168-180, 1988 74. Poser CM, Paty DW, Scheinberg L, et al: New diagnostic criteria for multiple sclerosis: Guidelines for research protocols. Ann Neurol13:227-231,1983 75. Poser S, Stickel 8, Krtsch U, et al: Increasing incidence of multiple sclerosis in South Lower Saxony, Germany. Neuroepidemiology 8207-213,1989 76. Poskanzer DC, Schapira K, Miller H Multiple sclerosis and poliomyelitis. Lancet ii:917-921, 1963 77. Pryse-Phillips WEM: The incidence and prevalence of multiple sclerosis in Newfoundland and Labrador, 1960-1984. Ann Neurol20323-328,1986

EPIDEMIOLOGY OF MULTIPLE SCLEROSIS

307

78. Radhakrishnan K, Ashok PP, Sridharan R, et al: Prevalence and pattern of multiple sclerosis in Benghazi, North-Eastern Libya. J Neurol Sci 7039-46, 1985 79. Riise T, Klauber MR Relationship between the degree of individual space-time clustering and age at onset of disease among multiple sclerosis patients. Int J Epidemiol 21:528-531, 1992 80. Risch N: Linkage strategies for genetically complex traits. I. Multilocus models. Am J Hum Genet 46:222-228, 1990 81. Roberts MH, Martin JP,McLellan DL, et al: The prevalence of multiple sclerosis in the south Hampton and South West Hampshire Health Authority. J Neurol Neurosurg Psy&iatry 5455-59,1991 82. Rosati G Descriptive epidemiology of multiple sclerosis in Europe in the 1980s: A critical overview. AM Neurol36(suppl2): 5164-974, 1994 83. Rosati G, Aiello I, Granieri E, et al: Incidence of multiple sclerosis in Macomer, Sardmia, 1912-1981: Onset of the disease after 1950. Neurology 3614-19, 1986 84. Rosati G, Aiello I, Pirastru MI, et al: Incidence of multiple sclerosis in the town of Sassari, Sardinia, 1965-1985: Evidence for increasing occurrence of the disease. Neurology 38:384388, 1988 85. Rothman KJ: Clustering of disease. Am J Public Health 7713-15, 1987 86. Sadovnick AD: Genetic epidemiology of multiple sclerosis: A survey. Ann Neurol 36(suppl2): S194-S203,1994 87. Sadovnick AD, Ebers GC: Epidemiology of multiple sclerosis: A critical overview. Can J Neurol Sci 20:17-29, 1993 88. Sadovnick AD, Baird PA, Ward RH: Multiple sclerosis: Updated risks for relatives. Am T Med Genet 29:533-541,1988 89. Sadovnick AD, Armstrong H, Rice GPA, et a1 A population-based study of multiple sclerosis in twins: Update. Ann Neurol33:281-285, 1993 90. Sadovnick AD, Eisen K, Hashimoto SA, et al: Pregnancy and multiple sclerosis: A prospective study. Arch Neurol51:1120-1124, 1994 91. Savettieri G, Elian M, Giordano D, et a1 A further study on the prevalence of multiple sclerosis in Sicily: Caltanisseta city. Acta Neurol Scand 7371-75, 1986 92. Sheremata WA, Poskanzer DC, Withum DG, et al: Unusual occurrence on a tropical island of multiple sclerosis. Lancet ii:618, 1985 93. Sibley WA, Bamford CR, Clark K Clinical viral infections and multiple sclerosis. Lancet t1313-1315,1985 94. Siva A, R a d h a h h n a n K, Kurland LT, et al: Trauma and multiple sclerosis: A population-based cohort study from Olmsted County, Minnesota. Neurology 13:18781882, 1993 95. Skegg DCG, Corwin PA, Craven RS, et al: Occurrence of multiple sclerosis in the north and south of New Zealand. J Neurol Neurosurg Psychiatry 50:134-139, 1987 96. Sobell JL, Heston L, Sommer S S Delineation of genetic predisposition to multifactorial disease: A general approach on the threshold of feasibility. Genomics 12:l-6, 1992 97. Stein EC, Schiffer RB, Hall J, et a1 Multiple sclerosis and the workplace: Report of an industry-based cluster. Neurology 37:1672-1677, 1987 98. Svenningsson A, Runmarker B, Lycke J, et al: Incidence of MS during two fifteen-year periods in the Gothenburg region of Sweden. Acta Neurol Scand 82161-168, 1990 99. Swank R, Lerstad 0, Strom A, et al: Multiple sclerosis in rural Norway: Its geographic and occupational incidence in relation to nutrition. N Engl J Med 246:721-728, 1952 100. Sweeney VP, Sadovnick AD, Brandejs V: Prevalence of multiple sclerosis in British Columbia. Can J Neurol Sci 1347-51, 1986 101. Swingler RJ, Compston DAS: The prevalence of multiple sclerosis in South East Wales. J Neurol Neurosurg Psychiatry 51:1520-1524, 1988 102. Tan C-T Multiple sclerosis in Malaysia. Arch Neurol45:624-627, 1988 103. Thompson AJ, Hutchinson M, Brazil J, et al: A clinical and laboratory study of benign multiple sclerosis. Q J Med 58:69-80, 1986 104. van Walderveen MAA, Tas MW, Barkhof F, et al: Magnetic resonance evaluation of disease activity during pregnancy in multiple sclerosis. Neurology 44:327-329, 1994 105. Vassallo L, Elian M, Dean G: Multiple sclerosis in southern Europe, 11.: Prevalence in Malta in 1978. J Epidemiol Community Health 33:111-113, 1979

308

WEINSHENKER

106. Williams ES, McKeran R O Prevalence of multiple sclerosis in a south London borough. BMJ 293:237-239, 1986 107. Wynn DR, Rodriguez M, OFallon WM, et al: A reappraisal of the epidemiology of multiple sclerosis in Olmsted County, Minnesota. Neurology 40780-786, 1990 108. Yaquib BA, Daif A K Multiple sclerosis in Saudi Arabia. Neurology 38:621-623, 1988

Address reprint requests to Brian G. Weinshenker, MD Department of Neurology Mayo Clinic 200 First Street, SW Rochester, MN 55905