The distribution of multiple sclerosis

THE RATIONALE FOR EPIDEMIOLOGICAL STUDIES IN MULTIPLE SCLEROSIS

Morbidity statistics for multiple sclerosis have been surveyed in many places, on many occasions and for many reasons. Listing individuals in whom the diagnosis has already been made, or identifying those with one particular feature of the disease, is not a difficult exercise, only requiring clinical resources and limited epidemiological expertise. In practice, the aims of these surveys have varied. Few have proved definitive and methodological factors have often limited the extent to which useful conclusions can be drawn. This is particularly true when comparisons have been made between the frequency of multiple sclerosis in different places, or when one area has been surveyed repeatedly over time.

The commonest reason for carrying out an epidemiological survey is to use temporal and geographical gradients, and variations in risk depending on location over defined periods, to generate hypotheses for causation of the disease. In this situation, it is essential to know that claims for reported differences in morbidity statistics are reliable. No studies have contributed more to an understanding of the causation and aetiology of multiple sclerosis than those charting its global distribution. These epidemiological surveys have generated useful hypotheses and spawned secondary research seeking mechanistic explanations for the emerging patterns. But against a background of shifting statistics resulting from methodological variations and, perhaps, genuine changes in disease frequency and distribution both on the global and regional scales, many questions remain unanswered. Even the fundamental issue of whether an explanation for the distribution of multiple sclerosis should be forced into purely environmental or genetic constructs continues to be debated.

Whilst it is interesting to interpret the epidemiology of multiple sclerosis through the social history of populations in which the disease occurs, those that proved informative for epidemiologists working in the early part of the 20th century have been subjected to further changes in structure and distribution, making longitudinal or comparative studies difficult to perform and interpret. But despite not yet yielding a definitive account of the aetiology, the effort that has gone into serial identification of cases throughout the world using sound epidemiological principles, supports the concept of an interplay between environmental triggers and genetic susceptibility in the causation of multiple sclerosis. The challenge now is to nail the main factors in each category.

The second application of epidemiological work in multiple sclerosis is in public health medicine. Here, the priority is to assess local needs for the provision of services and allocation of resources. For this purpose, although accurate case definition is a prerequisite, issues relating to disease course and severity in the population as a whole are more important than trawling successfully for each and every affected individual using rigorous clinical and laboratory methods. For these purposes, patterns of disability, impairment and handicap provide the canvas on which the individual needs of prevalent patients are painted.

Lastly, epidemiological scrutiny of a fully ascertained, population-based register of patients can be used to define the natural history of multiple sclerosis, eliminating the bias inherent in description of the clinical course that inevitably contaminates cohorts derived from specialist clinical teams, groups of patients attending hospital clinics, or extrapolation of individual experience to the population at large. With the need to assess putative treatments that modify the course of multiple sclerosis, detailed description over time of large untreated population-based natural history cohorts has become even more important (see Chapter 4). Putative disease-modifying treatments for multiple sclerosis are now used routinely in many parts of the world. As a result, it is no longer considered ethical to withhold these medications and randomize patients to placebo preparations in clinical trials. The use of historical controls derived from natural history cohorts may therefore be the only comparator available for the assessment of new treatments.

DEFINITIONS AND STATISTICS IN EPIDEMIOLOGY

Although shrouded in the mysteries of professional vernacular, the definitions used in epidemiology are not complex. Statistics are used to protect investigators from believing in their own or other people's hypotheses, arising from random variation and confounding interactions between apparently independent events. Over time and with experience, agreed procedures have been adopted for the assessment of morbidity statistics in multiple sclerosis and for comparisons between factors that aim to distinguish individuals, or groups, having an unusually high or low risk of the disease (Kurtzke 1996).

The term frequency lacks precision but is useful in conveying a general impression of how often a particular event has occurred. The cumulative frequency, or lifetime risk, of multiple sclerosis, is the maximum chance that it will occur at any time during the life of each individual at risk, irrespective of when the condition first manifests. For northern Europeans, an approximate estimate of lifetime risk is 1:400, whereas it is 1:3 for the identical co-twin of an affected individual. Factors influencing the expected frequency of multiple sclerosis can properly be called risk factors. Relative risk is the product of the proportions of cases and controls with and without a particular risk factor. The odds ratio describes the ratio of incidence rates for multiple sclerosis in individuals who have, and have not, been exposed to the same risk factor.

Incidence (or more accurately cumulative incidence) measures the number of new events in a defined area over a given period. Events might be new diagnoses, stages reached in the course of the disease, or individual clinical features. The study area can be geographically or demographically defined but anything that deviates too far from a population base is likely to introduce ascertainment bias. The period over which observations are made can be long or short. The former introduces problems through social and age-related alterations in the population at risk. The latter leaves no real time to establish patterns free from natural oscillations in frequency of the disease or defined manifestations. In practice, incidence refers to the number of newly diagnosed patients per 10⁵ of the population at risk over one calendar year. Incidence, prevalence and mortality are all ratios of individuals with multiple sclerosis (numerators) amongst a population at risk (the denominator: see Figure 2.1 ). Errors in calculating the numerator or denominator will spuriously distort morbidity statistics but the impact will be greater for numerator errors, especially where a group of epidemiological interest has relatively few affected individuals. The odd case omitted, or diagnosed in error, amongst a minority ethnic group studied in a large mixed metropolis will make an enormous impact on estimates for the incidence and prevalence of multiple sclerosis in that community. Although incidence and prevalence can be used to make comparisons between populations, changes in either may arise from predictable cycles in these statistics (regression to the mean), and each is dependent on the structure of the population being surveyed (Figure 2.2 ).

Figure 2.1.

Percentage incidence, prevalence and mortality in all affected individuals in a population-based sample by decade. Incidence peaks in the 3rd decade, prevalence in the 4th and mortality in the 7th.

Figure 2.2.

Variations in structure of the population at risk (denominator) influence the numerator. Since multiple sclerosis is a disease of young age at onset but long duration, a population with a disproportionate number of young people will show inflated figures for incidence (I) by comparison with mortality (M), whereas a population with an old age structure will show lower prevalence but increased figures for mortality and lower incidence.

Prevalence defines the number of individuals in whom multiple sclerosis has been diagnosed, or those with a particular feature, in a population at risk on a given occasion. Because it includes all individuals who qualify for the outcome of interest, irrespective of the date of onset, it is numerically larger than incidence. Again, in practice, the event is usually the established diagnosis of multiple sclerosis. Estimates for the prevalence of multiple sclerosis are dependent on diagnostic accuracy. In turn, diagnostic precision depends on the choice and application of criteria for assessing the probability that each individual in the study does have demyelinating disease affecting the central nervous system. Here, there is bound to be an appreciable rate of error since demyelination has an extremely variable clinical phenotype. Prevalence is inevitably dependent on the degree of case ascertainment, and this tends to vary inversely with size and accessibility of the population at risk. It is almost invariable that first surveys underestimate prevalence (and incidence). Higher figures are obtained with second and subsequent assessments due to improved awareness and vigilance amongst both the population at risk and the investigator. It is for this reason that comparisons between neighbouring places and regions cannotnecessarily be used to define variations in the distribution of multiple sclerosis.

The category most likely to escape recognition, especially in first surveys, is the early or benign case. This artefact is especially problematic in areas with relatively underdeveloped health care systems. In this situation, there is also a likelihood that mortality will be higher, tending to reduce the number of individuals with advanced disability. Since duration is short when mortality occurs early, the overall number of prevalent cases will be further reduced. Thus, several factors may combine to favour under ascertainment and low estimates for prevalence in regions with poorly developed medical services. One can only guess at the extent to which this has contributed to apparent geographical gradients in the frequency of multiple sclerosis, and their erosion with serial updating of surveys performed in places enjoying improved health care facilities over time.

One proposed solution to the arbitrary use of date at diagnosis as the point of entry into an epidemiological survey is to adjust backwards to the perceived date at onset (C.M. Poser et al 1992). This is considered appropriate in studies probing the aetiology of multiple sclerosis since it gets closer to the point of origin for the putative risk factor – or so the argument goes. Onset, in this context, means the date of first symptoms and so even this statistic does not adjust for the unknown interval between development of the disease process and its first recognized clinical expression. It merely substitutes one imprecise date for another. Proponents of onset-adjusted prevalence argue that it is appropriate retrospectively to include in morbidity statistics those individuals who are later classified as having multiple sclerosis even if their whereabouts and diagnostic status were unknown at the time. They should have been identified and so ought retrospectively to be included. However, all surveys carry incorrectly diagnosed cases and omit individuals who would have been identified in a perfect and fully ascertained study. Assuming that these accumulate at a constant rate, onset adjustment does not improve accuracy in serial estimations of incidence or prevalence.

The denominator, or population at risk, should be demographically based, or selected to suit the hypothesis being tested. Thus, whereas most epidemiological studies will relate to census information for a given location, others will have as their denominator a stratified group – such as patients with the primary progressive phenotype. What must be avoided is bias that loads the probability of any one individual being a case, or having a defined set of clinical features. This is most likely to occur with hospital- or clinic-based samples, where referral is subject to the whimsical nature of access and timing of referral. Attending patients tend to represent the newly diagnosed and more severe end of the clinical spectrum. The occasion on which prevalence is assessed is conventionally a single date (prevalence day) but this can be extended to a period, such as a year. In fact, depending on the outcome of interest, it may be prudent to adopt the period prevalence approach so as to ensure that sufficient examples are identified from which to judge the frequency of the event of interest.

Mortality describes the number of individuals with multiple sclerosis, as a proportion of the population at risk living in a defined area, dying over a given period. It does not document individuals with multiple sclerosis as a proportion of those who actually died in the population at that time, but refers to the entire population who could die. More importantly, it can either identify the number dying from multiple sclerosis or the number dying with multiple sclerosis. With so many uncertainties, and the tendency in most countries for causes of death to satisfy administrative arrangements rather than clinicopathological precision, mortality is operationally a poor statistic for evaluating the epidemiology of multiple sclerosis.

Mortality should match incidence (if each has been accurately assessed and no changes are occurring in aetiological factors determining the frequency of the disease). This is often not the case since it is easier for clinicians and epidemiologists to document new cases than to establish who has died with or from multiple sclerosis. The relative insensitivity of mortality statistics is disappointing since an excess of incidence over mortality in individuals having multiple sclerosis is one way of suspecting that a recent change in aetiological conditions has occurred.

Superficially, mortality also reflects the impact of therapy since improvements in the management of potentially life-threatening complications of multiple sclerosis will reduce death attributable to multiple sclerosis. Conversely, increased longevity merely shifts the balance of mortality from multiple sclerosis to mortality with multiple sclerosis. Longevity upwardly adjusts mortality to a later age without altering the number of people with multiple sclerosis who eventually die from an unrelated cause. It is self-evident that incidence, prevalence and mortality have a close relationship since each relates to a separate phase of the same disease process. In situations where ascertainment has been complete, no changes in survival or diagnostic classification have occurred, and frequency of the disease has not changed, prevalence will be the product of incidence or mortality and disease duration.

Survival refers to the number of patients in a cohort who have not yet reached a defined end point, typically death. It is less easy to measure directly than might be expected. Survival can begin with date of the first symptom, but this often has to be identified retrospectively. Alternatively, onset can be taken as the time of diagnosis, but that may have been influenced by random events and personal attitude relating to medical attendance, over and above patterns of referral and general organization of health care. In fact, disease duration is usually calculated by default on the basis of prevalence and incidence, or taken as twice the interval from mean age of onset to mean age at prevalence (Poskanzer et al 1980a) – but this method has wide confidence intervals.

Like all statistics, survival is dependent on full ascertainment of the numerator and denominator. The difficulty lies in the fact that the denominator has continually to be adjusted to account for the shifting number of individuals in the cohort who remain at risk through movement out of the study area and death. Annual rates for survival, based on the number of deaths, are censored for the fraction of each year that those who moved away actually remained within the study area, and hence at risk, allowing the calculation of cumulative survival. This approach does not equate to disease duration, or survival from onset, unless the cohort has been ascertained at presentation and is fully prospective. In practice, it is only possible to include cases for assessment of survival from the time at which they first come under observation. (The problems and pitfalls of survival analysis have been amusingly dramatized by John Kurtzke in outlining the fictitious excursions into neuroepidemiology of Halvah Finster and his fistulous disease; Kurtzke 1989.) In order to provide a complete picture on morbidity statistics, in the context of incidence and prevalence in particular locations, some aspects of survival are described in the sections that follow: a more detailed account of survival in multiple sclerosis is in Chapter 4.

Perhaps the factor making morbidity statistics for multiple sclerosis most difficult to establish is the range of age at onset. Without a reliable disease marker, and with such variable modes of presentation, it is difficult to know whether and when an individual is entirely free from the risk of developing multiple sclerosis. This obstacle is partially overcome by calculating age- and sex-specific rates for incidence, prevalence and mortality. These relate the number of affected individuals to a denominator confined to that proportion of the population at risk, and having the same age and/or sex distribution (to the nearest decade: Figure 2.3 ). This goes some way towards dealing with the problem of comparing demographically different populations. Take, for example, the situation of multiple sclerosis in a small rural town attracting people at retirement but offering no employment for the young. The numerator for incidence will be small despite an adequate sample size, whereas prevalence may be proportionately much higher. This is especially likely if the area being surveyed offers excellent facilities for individuals with physical disability. Conversely, an area providing special opportunities for young people, such as a provincial university town, will have a relatively high proportion of individuals at risk, favouring higher statistics for incidence and prevalence but not mortality (Figure 2.2). In general, comparison of age-specific rates for multiple sclerosis in different populations eliminates confounding due to demographic structures.

Figure 2.3.

Age- and sex-adjusted figures for prevalence of multiple sclerosis in the south Cambridgeshire district of East Anglia in the United Kingdom.

Adapted from Mumford and Compston (1993).

© 2006

Major changes in birth rate, or substantial migration over the period in which morbidity statistics for multiple sclerosis are being collected, also affect estimates for incidence and prevalence. The usual way of dealing with these variations has been to relate age-specific rates to a single population, deriving the standardized prevalence ratio. Inevitably, the chosen population varies between regions and is rarely agreed. In the United Kingdom, standardized prevalence ratios have been referred to the population of Northern Ireland in the 1960s. Taking equivalence as 100, standardized prevalence ratios are then quoted as >100 or <100. In other disease contexts, an attempt has been made to create a standardized European population for age correction (Figure 2.4 : see Waterhouse 1976). Until recently, this reference has not been much used by epidemiologists studying multiple sclerosis.

Figure 2.4.

Standard populations: World and European.

Adapted from Waterhouse (1976).

© 2006

Confidence intervals provide an additional means of defining the extent to which a given result, or point estimate, is likely to be reproducible, and hence valid. Confidence has upper and lower limits. Its range represents possible results consistent with the observations that have been made. Wide confidence intervals invite caution. Narrow ones suggest that the point estimate is likely to be reproducible. The 95% confidence limits are usually quoted. These represent the range for which there is a 95% chance of including the correct value. Although confidence intervals can be used as surrogates for statistical significance, the confidence interval does much more than assess the extent to which the null hypothesis is fully consistent with the results of a particular study. Appropriate use of the confidence interval assumes that the sample studied is unbiased and representative of the population at risk. The method for calculating confidence intervals varies with size of the numerator. For the range represented in most epidemiological surveys of prevalence in multiple sclerosis, the limits are given as:

$$\begin{array}{l}\text{Upper limit}=\left(\text{numerator}/\text{denominator measured in}{10}^5\right)\times Z\\ \text{Lower limit}=\left(\text{numerator}/\text{denominator measured in}{10}^5\right)\times Z\end{array}$$

where Z is given as: exp (1.96 × [square root of 1/numerator]).

Case–control studies are used to identify risk factors. The contribution of each factor is expressed as an odds ratio or relative risk. These are often used interchangeably, and each is vulnerable to inappropriate selection of controls leading to poor matching between groups. However, matching inevitably excludes assessment of any risk contributed by the matched factor itself (such as age or sex) and events with which it happens to cosegregate. The contribution made by any one factor to the aetiology is the attributable risk. This component can potentially be modified, hopefully leading to a reduction in disease frequency. Confounders are factors influencing the assessment of another feature because the confounder has its own relationship with the population at risk. Each cofactor then appears to have an independent disease association but only one is biologically relevant.

Matching requires selection for each case of one or more controls from a similar group of individuals at risk who have the same frequency of confounding factors. Stratification groups subjects with one shared (and potentially confounding) factor in a particular category. This might be age, a clinical phenotype or a laboratory marker such as genotype. The difficulty with case–control studies in multiple sclerosis is lack of knowledge on potential confounders. In this situation, mismatching may persist – despite best efforts to stratify – through ignorance concerning aetiology of the disease. Matching is an important means of increasing the power of a case–control study. For example, if the hypothesis is that age of infection by Epstein–Barr virus (EBV) in genetically susceptible individuals determines the development of multiple sclerosis, matching both for EBV exposure and identified genetic risk factors will maximize the prospects of showing a difference in risk of multiple sclerosis depending on age at infection. Failure to match may obscure a biologically meaningful difference.

A cohort study is one in which the frequency of developing multiple sclerosis is studied in ≥2 groups of individuals who differ in their exposure to a putative aetiological factor, which is then assigned a risk. Cohort studies can measure absolute as opposed to relative risks and are preferable to case–control surveys. They are difficult to perform in a disease of low overall frequency but can explore several hypotheses simultaneously and relatively quickly. Cases should be representative of all affected individuals. In practice, they tend to be selected from prevalence rather than incidence registers. This runs the risk of introducing survival factors as confounders unless disease duration is uniform.

The threshold for concluding that a given relative risk or odds ratio identifies a factor increasing susceptibility, or conferring protection from the development of multiple sclerosis is, by definition, greater or less than 1. Interpretation depends on the 95% confidence intervals. If these are each >1, there is a >95% chance that the disease is associated with the factor in question. If the intervals straddle 1, the probability is that the factor is neutral. And if the intervals are both <1, the risk factor is considered to be protective. Since very few environmental factors conferring susceptibility or protection (in contrast to variables such as age, sex and race) have survived detailed scrutiny of the epidemiology of multiple sclerosis, it follows that the majority of case–control studies show confidence intervals that include 1, even though the odds ratio may itself be higher.

STRATEGIES FOR EPIDEMIOLOGICAL STUDIES IN MULTIPLE SCLEROSIS

In planning an epidemiological study in multiple sclerosis, the usual practice is to retrieve cases from those already known to be affected. This requires the coordination of lists from various sources. Very few surveys of multiple sclerosis are conducted by making new door-to-door enquiries. In many parts of the world, the diagnostic process is coordinated through hospital clinics to which all individuals with suspected symptoms of the disease are referred. Scrutiny of these departmental records provides the best single source of information when compiling a register of prevalent cases. Many, but not all, can also be identified from records held in primary care (general practice, in the United Kingdom). This is the source best placed to register individuals who have recently entered the area being surveyed, and those with established disease of long duration who may no longer be known to the hospital system. A few prevalent cases may be identified only as residents of homes for the chronic young sick although these rarely yield otherwise unascertained cases. Most epidemiologists find that morbidity registers for diseases that form part of general hospital admission or attendance records are a poor source of information.

The provisional register of prevalent cases must then be revised. Type A studies are those in which each case is assessed personally by the investigator in order to confirm or refute the diagnosis. Type B are those in which the diagnosis is assumed from available paper records. In practice, most studies combine both approaches. The fact that each registered case is domiciled within the study area needs to be ascertained for prevalence day from local health care systems, since movements in and out of the area contribute a significant source of error and will inevitably change between surveys in places providing serial estimations. It is wise to choose an area for which there is contemporary census information on the denominator. L.M. Nelson and Anderson (1995) have debated the relative merits of case finding using various sources in the United States. They conclude that the best yield is to be found through manual sorting of records held in neurology practices, and the files of local services for multiple sclerosis and care of the chronic young sick. They recommend the capture–recapture method. Borrowed from zoological and avian practice, this assesses how often the same individuals are identified in serial samples of a population, as the measure of ascertainment.

Serial change in incidence is generally considered a better marker than prevalence for detecting alteration in factors determining the aetiology of multiple sclerosis since these are presumed to be less vulnerable to case ascertainment and changing trends in classification. However, even incidence is not entirely secure. Raised level of awareness amongst physicians and patients, especially those who are acquainted with or related to patients with multiple sclerosis, increases the recruitment of incident cases. Sudden change in the provision of facilities for the disabled tends to inflate each statistic, as does the arrival of an investigator with a special interest in the epidemiology of multiple sclerosis. Morbidity statistics tend to plateau once ascertainment saturates after a period of steady increase. It follows that differential vigilance and efforts at complete ascertainment will tend to create spurious regional and temporal gradients.

Some questions relating to the epidemiology of multiple sclerosis can only be answered by performing studies in defined locations. It makes little sense to plan a study requiring significant numbers of individuals with a rare manifestation, such as twinning or familial multiple sclerosis, in a community that has a low overall prevalence of the disease. Similarly, ubiquitous but biologically important features may not differ between groups in places where multiple sclerosis is extremely common. The question also arises of whether it is appropriate to perform epidemiological surveys in locations with underdeveloped procedures for medical care. For multiple sclerosis, but not some other disorders, these tend to be regions of low prevalence in which the yield from population surveys is bound to be relatively unrewarding.

Opportunities for identifying risk factors making a significant contribution to the disease, but common in the normal population at risk, are improved in areas of relatively low prevalence. Conversely, risk factors for multiple sclerosis that are not over-represented in the normal population are identified more easily in high-prevalence regions. Even in these situations, large numbers and extra resources are needed to define the biologically less important risk factors, irrespective of their population frequency. Inevitably, decisions concerning the area to be included in a study represent a trade-off between administrative boundaries for which demographic information is available and a scale sufficient to include enough examples of the factor of interest. Decisions concerning the omission of suspected cases depend on the study aims. For surveys testing biological features, the ideal is to include all individuals who have the disease process even if this is not yet clinically declared. In other contexts, it is essential to restrict inclusions to those with definite disease in order to exclude doubtful cases and prevent contamination of the study aims. These decisions require judgement and knowledge of the disease but inevitably introduce difficulties for regional and cross-cultural comparison of surveys.

Generally, it is unwise to accept as proof the results of a single epidemiological study claiming involvement of any one aetiological factor. Cohort studies require careful selection of participants from a population base in order to avoid recruitment bias. The cohort can only be selected once the research aims have been defined and the population established. Many studies require a system for clinical classification and description of morbidity using well-validated rating scales. Groups defined on the basis of disability, impairment and participation are then selected randomly within diagnostic categories. No rating system addresses all these issues. The Kurtzke Expanded Disability Status Scale (EDSS; Kurtzke 1983a; see Table 6.1) has the merits of widespread use, familiarity amongst clinicians and extensive validation. But it is not ordinal and prevalent cases usually show a bimodal distribution. It is especially insensitive in certain ranges and excessively weighted towards ambulation. Some limitations of the EDSS are circumvented by using life-table analysis of the time taken for patients to reach strategic points on the scale, as part of natural history studies (see Chapter 4), or treatment trials (see Chapter 18). Population association studies often assess the frequency of a technically sophisticated clinical or laboratory marker, not easily applicable to a large sample. The merits of the study then depend entirely on the appropriate selection of cases and controls. Bias or inconsistency for either can invalidate the study. Sociohistorical factors may make for significant differences across even quite small regions, especially when comparing genetic risk factors and environmental exposures. Ethnicity should always be remembered at the planning stage of an epidemiological study. It is, for example, likely to remain an inviolate principle in the assessment of genetic susceptibility, whereas (for example) the potential confounding effect of ethnic heterogeneity is entirely appropriate in a study designed to assess resource requirements for the disabled.

THE GEOGRAPHY OF MULTIPLE SCLEROSIS

In the modern era (for earlier epidemiological work, see Chapter 1), John Kurtzke (Kurtzke 1980a; 1980b; 1983b; 1985b; 1993; 2000) has systematically updated his original landmark overview (Kurtzke 1975) in order to incorporate the results of new surveys. Others have attempted the same (see Bauer 1987; D.A.S. Compston 1990a; Rosati 2001). Earlier editions of this book benefited from the authoritative accounts of Donald Acheson (McAlpine et al 1965; 1972; W.B. Matthews et al 1985) and Christopher Martyn (W.B. Matthews et al 1991). Taking the big picture, Kurtzke concludes that whereas multiple sclerosis was geographically a regional disorder of males in the mid-20th century, subsequent experience shows more multiple sclerosis in places where the disease was already known to occur, diffusion into places where it was hitherto uncommon across even quite narrow confines within single land masses, and a steady increase amongst women and races other than Caucasians. For Kurtzke, the regional increases within Asia have been in the former Soviet Union but not Japan, Korea or China. In Europe, the southern littoral of the Mediterranean and Israel have shown conspicuous increases in prevalence. The same trends are apparent in Central and South America (Mexico, Argentina and Uruguay). But many of these focal trends are based on unpublished data, and hence vulnerable to all the factors that make for serial increase in morbidity statistics independent of an increase in incidence (Kurtzke 2000). A long list of best estimates for the frequency of multiple sclerosis makes dull reading but what emerges from an overview is that the main factor determining patterns in the observed frequency of multiple sclerosis (excluding differences dependent on hospital- or population-based surveys) is when, as much as where, the study was carried out. In following those who have successfully painted the big epidemiological picture on disease frequency and distribution, we have also tried comprehensively to collate studies that variably provide statistics for incidence, prevalence or mortality (Figure 2.5 ). We have not undertaken the arduous task of calculating confidence intervals for surveys where these are not quoted. Neither, unlike the admirable and pioneering work of John Kurtzke, do we claim to have been encyclopedic in listing each and every epidemiological survey. Inevitably, Europe is disproportionately represented since multiple sclerosis is relatively common and a focus for epidemiological studies (Figure 2.6 ). For surveys carried out during the heyday of epidemiological studies, readers are referred to two publications acting as important additional sources of primary information, although we suspect that each necessarily took surveys of differing quality in producing overviews. Helmut Bauer summarized the epidemiology of multiple sclerosis in Europe (Bauer 1987) on the basis of presentations held in association with the World Congress of Neurology in Hamburg. The Department of Neurology in Darmstadt celebrated its 20th anniversary with a workshop at which delegates presented recent epidemiological surveys from many parts of Europe (Firnhaber and Lauer 1994).

Figure 2.5.

Summary of epidemiological patterns in multiple sclerosis. Figures are estimates for prevalence. Solid lines with arrows represent migration vectors of northern Europeans. Open lines with arrows represent migration routes of Africans to the Caribbean and Mississippi delta and to the United Kingdom. In South Africa the numbers refer to English-speaking whites migrating as adults (60), English-speaking whites migrating as children (15), Afrikaners (10) and mixed race individuals (<5), as estimated in the mid-1980s.

Figure 2.6.

Distribution of multiple sclerosis in Europe. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

MULTIPLE SCLEROSIS IN SCANDINAVIA (Figures 2.7 and 2.8)

Figure 2.7.

Prevalence distribution of multiple sclerosis in Scandinavia. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

Figure 2.8.

Incidence distribution of multiple sclerosis in Scandinavia. Most recent figures are given for incidence/10⁵ population/year; a ‘best guess’ is given where local variations exist in the published literature.

The study of multiple sclerosis in Scandinavia has been of considerable importance to those who espouse the genetic doctrine and attribute causation to the Viking gene pool (see C.M. Poser 1994, and Chapters 3 and 5). This region featured prominently in the early studies by John Kurtzke (Kurtzke 1967; 1974; 1975) leading to definition of a Fennoscandian focus. As in most other places, estimates for prevalence outnumber those for incidence. Overall, the trend has been for a steady increase in prevalence, and for all the usual reasons – improved survival, earlier diagnosis, enhanced vigilance and more sensitive diagnostic criteria. But ascertainment saturation may have occurred allowing morbidity statistics to plateau. The early studies are summarized by Koch-Henriksen (1995).

In the modern era, the systematic study of multiple sclerosis in Denmark began with the pioneering work of Kay Hyllested (Hyllested 1956), initiated at the instigation of Torben Fog, and much dependent on the availability of Danish disease registers. Based on 2140 certain and 341 probable cases, rates were 4.4/10⁵/annum for incidence, and 58/10⁵ for prevalence on 1st October 1949. Age-specific prevalence peaked in the fourth decade and the sex ratio was 1.2F:M. Koch-Henriksen et al (1992) used the same Danish Multiple Sclerosis register established by Hyllested to review cases listed on 31st December 1986 whose onset was judged to postdate 1st January 1948. The interval between onset of symptoms and their recognition as manifestations of multiple sclerosis was >5, >10 and >20 years for 51%, 27% and 6%, respectively. The authors knew of 6478 cases incident between 1948 and 1986 (1.4F:M) of whom 83% had clinically definite disease. The ascertainment corrected incidence rate across this period was 4.4/10⁵/year in a population that increased from 4.2 to 5.1 million. Koch-Henriksen et al (1992) produced useful lifetime cumulative incidence rates allowing that proportion of the population expected to have developed multiple sclerosis by a given age to be identified (Figure 2.9 ). At 65 years, 0.31% or 1:314 Danes (1:271 females and 1:372 males) will have developed multiple sclerosis. Incidence rates for the three decades from 1950 were 5.1 (95% CI 4.9–5.3), 3.8 (95% CI 3.7–4.0) and 4.3 (95% CI 4.0–4.5), respectively. The fall in the second decade (1960–1970) depended on a reduced rate in people aged <35 years but this was followed by an increase for older incident cases, especially females, in subsequent years. The rate for the most recently evaluated quinquennium was 4.6/10⁵/year (95% CI 4.3–5.0; Koch-Henriksen et al 1994). After making corrections for the impact of laboratory investigations on the timing of diagnosis, the authors concluded that a change in environmental factors had altered the frequency of multiple sclerosis in the middle decade of their comprehensive survey. We remain to be convinced. By 1993, the number of registered cases on the database, prevalent in 1948 or incident between 1949 and 1993, was 12 070. Crude annual average incidence was 4.99/10⁵/year between 1980 and 1989, and prevalence 112/10⁵ on 1st January 1990 (Koch-Henriksen 1999). Of these 12 070 cases, 6068 had died between 1951 and 1993. Death from multiple sclerosis was registered in 55.4%. As we discuss in Chapter 4, standardized mortality ratios were increased for infectious and pulmonary disease (2.46, 95% CI 2.04–2.94), suicide (1.62, 95% CI 1.29–2.01) cardiovascular disease (1.34, 95% CI 1.22–1.48), and accidents (1.34, 95% CI 1.02–1.71). They were reduced for cancer (0.79, 95% CI 0.70–0.90), perhaps due to incomplete ascertainment and the reduced life expectancy (Koch-Henriksen et al 1998). In order to provide an account of survival in the post-antibiotic era, these data were recently reassessed to include only the 9881 cases incident between 1949 and 1996 (and not the cohort prevalent when the Danish register opened in 1948), of whom 4254 had died by January 2000. Multiple sclerosis was associated with a reduction in life expectancy of around 10 years, with 56.4% of deaths attributed to the disease itself and, as before, with an excess standardized mortality ratio for suicide and accidents and protection from cancer. But, as anticipated, the reduction in life expectancy had halved over the period 1948–2000 (Brønnum-Hansen et al 2004). Thus, leaving aside minor perturbations, Denmark has seen a two-fold increase in prevalence but stable rates for incidence over the last 50 years, providing clear evidence for the dominant impact of ascertainment and methodology over biological forces in changing the frequency of multiple sclerosis in a community having in place sophisticated nationwide procedures for ascertainment.

Figure 2.9.

Cumulative incidence rates (%) from birth to any specific age by sex and 5-year age entries.

Adapted from Koch-Henriksen et al (1992). Reproduced with permission from the BMJ Publishing Group.

© 2006 BMJ Publishing Group

A comprehensive analysis of multiple sclerosis over time has also been carried out in Gothenburg, Sweden. The incidence of definite and probable multiple disease dropped progressively from a stable rate of 4.2/10⁵ between 1950 and 1964 over successive 5-year periods between 1974 and 1988 to 2.0/10⁵ (Svenningsson et al 1990). Conversely, serial estimates of prevalence showed stable rates adjusted to the population structure of the city at 91, 91 and 96/10⁵ for 1978, 1982 and 1988, respectively (Svenningsson et al 1990). These authors point out that the reduction in incidence of multiple sclerosis in Gothenburg, with stable prevalence, over the two 15-year periods 1950–1964 and 1974–1988, coincided with the near complete eradication of measles by vaccination and (by 1990) no measles-vaccinated individual had developed multiple sclerosis in this community. In northern Sweden (Vasterbotten county), onset-adjusted incidence between 1987 and 1997 was 5.2 (95% CI 4.4–6.2) in women and 3.7 (95% CI 2.7–4.9)/10⁵/year (1.4F:M: Sundstrom et al 2003). Onset-adjusted prevalence was 154 (95% CI 139–170) in December 1997, with differential rates for women and men of 202 (95% CI 179–228) and 105 (95% CI 89–125), respectively (1.9F:M). Here, the annual increase in prevalence (2.6% compared with rates from Gothenburg between 1990 and 1997) is attributed to a real difference in incidence, compared with southern Sweden, and steady reduction in mortality leading to increased survival. As an approximation to incidence in the population, and to inform the epidemiology of health care needs, G.X. Jiang et al (1999) showed rates for first hospital admission of 4.46/10⁵/year (2.19F:M).

There have been rather few epidemiological studies of clinically isolated demyelinating syndromes. Jin et al (1998) reported an annual crude incidence for isolated optic neuritis of 1.46/10⁵/year between 1990 and 1995 (4F:M). Age adjustment indicated peaks in the fourth and fifth decades; rates were lower for those not born in Scandinavia; and the conversion to multiple sclerosis correlated with age at presentation.

Another attempt at painting the big Swedish picture on statistics for multiple sclerosis is a review of disability and mortality for 1952–1992 involving, respectively, 11 414 and 5421 cases, referred to equivalent statistics for Parkinson's disease (disability) and motor neuron disease (mortality). Disability attributable to multiple sclerosis and death certification with multiple sclerosis had each risen but showed regional variation (Landtblom et al 2002). Subsequently, Landtblom et al (2005) compared three sources of information: 1301 hospital admissions from 1925 to 1934, 5425 deaths between 1952 and 1992, and 11 371 disability pension receipts admitted between 1971 and 1994; the prevalence data were most heterogeneous, indicating higher rates, in seven counties surrounding major lakes in southern Sweden and others in regions north of the Bay of Bothnia and around Stockholm.

In western Norway, the incidence for definite or probable multiple sclerosis appeared to increase in Hordaland County from 1.1/10⁵ for 1953–1957 (Larsen et al 1984) to 4.9/10⁵ for 1978–1982, before falling to 3.4/10⁵ for 1983–1987 (Grønning 1994; Grønning et al 1991). Although incidence for the period 1978–1982 was higher than in the subsequent quinquennium, the relatively long interval from onset to diagnosis (approximately 4 years in most surveys) makes this short-term change difficult to interpret. Hordaland County has also seen a threefold rise in prevalence from 25/10⁵ in 1963 to 75.5/10⁵ in 1983 (Larsen et al 1984). In More, Romsdal County, the average annual incidence changed from 1.9/10⁵ during the period 1950–1954 to 3.8/10⁵ for 1975–1979 (Midgard et al 1991). Subsequently, Midgard et al (1996a) summarized serial observations of incidence as showing a steady increase to 1975, which became more marked over the next decade but has since declined. Prevalence for definite and probable multiple sclerosis increased from 24/10⁵ in 1961 to 75/10⁵ in 1985 (Midgard et al 1991). The Oslo Multiple Sclerosis Registry was used to derive morbidity statistics for eastern Norway based on 794 cases identified by 1999 (2F:M; Celius and Vandvik 2001). Incidence increased over a 20-year period from 3.7 in 1972–1976 to 8.7/10⁵/year in 1992–1996. These incremental changes depended especially on patients with relapsing–remitting disease and females, and were considered by the authors not to have resulted solely from improved ascertainment. Prevalence (based on 579 northern European patients identified by 1995) was 120/10⁵ (136/10⁵ amongst native Norwegians, originating from throughout the country, and other Scandinavians). These rates are much higher than those reported from a rural area on the west side of Oslo fjord (Edland et al 1996). In this survey, prevalence increased from 62 to 86/10⁵ between 1963 and 1993.

In the period 1953–1972, De Graaf (1974) identified cases of multiple sclerosis in the northern counties of Norway, amongst a population of approximately 213 000, of whom Lapps made up ≤10%, the majority being Finns and Nordic people. Cases were classified using the Allison and Millar criteria (see Chapter 1) following personal examination. Annual incidence rates varied from 0.9 to 2.5/10⁵/year but with no temporal trend. Prevalence for all cases was 21/10⁵. Similar incidence rates (3.0/10⁵/year) were reported from Troms and Finnmark working to a prevalence date of 1st January 1983 (Grønning and Mellgren 1985). Statistics for northern Norway have been updated to 1993 (Gronlie et al 2000). Incidence was 3.5/10⁵/year for the period 1983–1992 and, by 1993, prevalence had increased to 73/10⁵. By 1993, there were three individuals with both and three others with one Sami parent, compared with a single parental example in 1983. Nevertheless, regional differences still show lower prevalence in parts with the highest proportional Sami population. Prevalence in Nord-Trondelag County was 123/10⁵ on 1st January 2000. Age-adjusted incidence increased over the period 1974–1999 from 3.9 to 5.6/10⁵/year (4.6 to 6.3 and 2.2 to 4.4/10⁵/year for females and males, respectively; Dahl et al 2004).

The frequency of multiple sclerosis in Vaasa, western Finland, increased steadily from 1964 through 1972 to 1979, by which time prevalence was 93/10⁵ compared with 53/10⁵ in Uusimaa, southern Finland. Incidence rates, updated to 1979, were 3.3 and 2.2/10⁵/year, respectively (Kinnunen 1984). Most changes resulted from increased prevalence in younger age groups. Relaxation of diagnostic criteria or a change in age structure of the population at risk were not thought to be responsible. The authors suggested that the threefold increase in western Norway, matched by a rise in incidence (see above), could not solely be attributed to alterations in demography, survival or ascertainment. They argued that a change had occurred in biological factors determining the frequency of multiple sclerosis. A subsequent update from the same region (Sumelahti et al 2001) showed prevalences of 202, 111 and 108/10⁵ in Seinajoki, Vaasa and Uusimaa, respectively, in 1993. Rates for incidence were 11.6 and 5.2 and 5.1/10⁵/year in Seinajoki, Vaasa and Uusimaa, respectively (Sumelahti et al 2000). The 1.2-fold increase in Seinajoki (compared with 1983) is attributed to a change in incidence in men; in Uusimaa incidence remains stable in both sexes; and in Vaasa, prevalence is stable but incidence has decreased in both sexes. The authors attribute these changes to environmental conditions rather than differences in case ascertainment. Based on cases dying between 1964 and 1993, survival was >50% at 40 years from onset with an excess of deaths from suicide and neoplasia; complications of multiple sclerosis accounted for 70% of deaths (see Chapter 4; Sumelahti et al 2002). A more recent publication qualifies these prevalences: 219/10⁵ (95% CI 190–247) in Seinajoki-south; 136/10⁵ (95% CI 108–164) in Seinajoki-north; and 107/10⁵ (95% CI 90–124) in Vaasa (Tienari et al 2004). By comparison, prevalence in neighbouring central Finland in 1993 was lower at 59/10⁵ increasing to 105/10⁵ by 2000. Incidence increased from 3.8/10⁵/year in 1979–1993 to 9.2/10⁵/year in 1994–1998 (Sarasoja et al 2004). We discuss the genetic implications of the epidemiological observations from Vaasa in Chapter 3.

The prevalence of multiple sclerosis in the former Soviet Union is now being reported – in many areas for the first time. These statistics must be regarded as preliminary and many of the original reports are only to be found in the proceedings of local meetings published in Russian. The average figures quoted by A.N. Boiko (1994) and A.N. Boiko et al (1995) show a gradient of increasing frequency from east and south to northwest, with a low of <5/10⁵ in Uzbekistan, Kazakhstan, Turkmenistan and Kirghizstan to a high of >50/10⁵ in Latvia. Within these regions, there appears to be a difference in rate between native populations and Russians, suggesting that, apart from differences in ascertainment dependent on variations in access to specialist medical care, ethnic factors may determine the distribution of multiple sclerosis in the former Soviet Union.

The most parsimonious explanation for geographical and temporal trends throughout Scandinavia is increased survival due to improved symptomatic treatments, consequential shifts in age-specific mortality and more disabled prevalent cases, and a saturation effect of ascertainment. That said, some of the temporal and geographical trends are consistent with real changes in the impact of exogenous factors determining the frequency of multiple sclerosis. Other regions bordering the North Sea have seen a reduction in incidence of the disease (see below for discussion of the Orkney and Shetland islands). The epidemiology of multiple sclerosis relating to islands in the North Atlantic – Iceland and the Faroes – is discussed under the heading of ‘Epidemics and clusters of multiple sclerosis’.

MULTIPLE SCLEROSIS IN THE UNITED KINGDOM (Figure 2.10)

Figure 2.10.

Prevalence distribution of multiple sclerosis in the United Kingdom. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

We have already described the early history of epidemiological studies of multiple sclerosis in the United Kingdom (see Chapter 1), highlighting the separate activities of Richard Williamson, Russell Brain, Isabel Wilson, Sydney Allison and John Sutherland. Subsequent surveys were carried out in the Western Isles in 1954 (Sutherland 1956) and 1979 (Dean et al 1981a), Northumberland and Durham (Poskanzer et al 1963), Carlisle (Brewis et al 1966), Yorkshire (McCoubrie and Shuttleworth 1978), and Cornwall in 1950 (Campbell et al 1950) and 1958 (Hargreaves 1969). From the late 1980s, there was a return to systematic surveying of multiple sclerosis with studies reported from geographically disparate parts of England, Wales, Scotland and Northern Ireland. It is now possible, once again, to review what has been learned concerning the distribution of the disease and the lessons this provides for a general understanding of the aetiology. Prevalence has risen. Incidence may have reached a plateau or fallen in places previously showing very high frequencies of the disease. The move towards comparability between surveys through the use of standardized diagnostic systems and age- and sex-corrected denominators has eroded but not eliminated the apparent north–south gradient.

Sixty years after Allison (1931) first surveyed multiple sclerosis in the United Kingdom, the frequency was reassessed during 1984–1988 in Wales, this time in the industrial southeast (Swingler and Compston 1988). The tenfold increase in prevalence over 50 years reflects a pattern seen with respect to dissemination of the disease over time in practically every other part of the United Kingdom where serial studies have been performed (Swingler and Compston 1986). Over four decades, the number of newly diagnosed cases of multiple sclerosis showed marked variations in south Wales, the peaks often coinciding with the arrival of a new neurologist but otherwise revealing a slow increase from about 4.8 in 1947 to 8.2/10⁵/year by 1988. The temporal trends arose partly from changes in definition and classification, the availability of laboratory methods for supplementing the diagnosis, and increased clinical vigilance. Mostly, they depended on the steady reduction in mortality that occurred in the second half of the 20th century. Mean duration of disease from onset of symptoms, estimated at 8 years in the early part of that century (Bramwell 1917), had risen to >25 years by the mid-1980s (D.A.S. Compston and Swingler 1989). E.S. Williams et al (1991) subsequently reanalysed mortality statistics, calculating annual age- and sex-specific figures standardized to death rates in the United Kingdom for 1974. They found less evidence for the north–south gradient and considered (on the basis of a fall in mortality) that – as in some parts of Scandinavia – the incidence of multiple sclerosis had been falling in Scotland. They considered only those individuals with multiple sclerosis as the underlying cause of death and so underestimated absolute numbers of cases dying with the disease by up to 50%. Whilst confirming the general trend towards a reduction in death rates throughout the United Kingdom, they showed that this was more marked in Scotland and Northern Ireland (39%) than in England and Wales (10%) and highlighted the conspicuous reduction in mortality for Scots aged >65 years. Although the failure to show a correlation between temporal and geographical trends for mortality and prevalence invites the comment that these mortality returns were too crude for useful analysis, the same criticism can be levied at figures for prevalence – traditionally considered the more robust statistic.

The subsequent batch of surveys from new or previously studied parts of the United Kingdom confirms the high overall frequency of multiple sclerosis and, with notable exceptions, continues to show increasing prevalence in each newly surveyed district. Thus, to the previously reported rates of 117/10⁵ in southeast Wales (Swingler and Compston 1988; updated to 120/10⁵ by Hennessey et al 1989), 115/10⁵ in the urban area of Sutton close to London (E.S. Williams and McKeran 1986), 99/10⁵ in Southampton (M.H.W. Roberts et al 1991) and 178/10⁵ in northeast Scotland (J.G. Phadke and Downie 1987) were added figures of 122/10⁵ for Rochdale in Greater Manchester in 1989 (Shepherd and Summers 1996), 98/10⁵ for Trent in the Midlands (K.W. Allen 1994), 130/10⁵ in southeast Cambridgeshire (Mumford et al 1992 – updated to 152/10⁵ by N.P. Robertson et al 1996a), 119/10⁵ in north Cambridgeshire (N.P. Robertson et al 1995), 153/10⁵ in rural Suffolk (Lockyer 1991), 111/10⁵ in the mid-Downs region of Sussex in southern England (Rice-Oxley et al 1995) and 113/10⁵ and 87/10⁵ for Jersey and Guernsey, respectively (G. Sharpe et al 1995; these bailiwicks of the Channel Islands come within the administrative boundaries of the United Kindgom but, being 70–100 miles south of the English mainland, are latitudinally comparable to western France, lying 10–30 miles to the west, and may therefore be genetically unrepresentative of southern England). Evidence that Scots have a high risk of multiple sclerosis irrespective of domicile strengthened with the subsequent report of incidence and prevalence rates for the Lothian and Borders regions (southern Scotland), at 12.2 (95% CI 10.8–13.7) and 10.1/10⁵/year (95% CI 6.6–13.6), and 185 (95% CI 175–194) and 201/10⁵/year (95% CI 174–228), respectively. These were equal to contemporary figures for northeast Scotland and the offshore islands (Rothwell and Charlton 1998). This declaration triggered a brief competitive response from Northern Ireland claiming, at 230/10⁵ (95% CI 207–256), to have the highest contemporary prevalence rate attributable to Celtic ancestry (McDonnell and Hawkins 1998a) and representing a substantial increase from an earlier estimate for the same location of 137/10⁵ (Hawkins and Kee 1988). More accurate figures subsequently emerged of 191/10⁵ (95% CI 169–213) using the Poser criteria and 163/10⁵ (95% CI 143–183) based on the Allison and Millar classification. Incidence was estimated at 6.5/10⁵/year in the late 1990s (McDonnell and Hawkins 1999). In a more recent comparison within Ireland, the prevalence in Wexford was 121/10⁵ (95% CI 101–144) compared with 185/10⁵ (95% CI 162–210) in Donegal, Northern Ireland (McGuigan et al 2004). Taken together, no one would doubt that Northern Ireland has a relatively high frequency of multiple sclerosis.

After a busy period in the mid-1990s, rather few new epidemiological studies have since been reported from the United Kingdom. One new region of southeast Scotland was screened by Forbes et al (1999) providing a rate for prevalence of 184 (95% CI 171–198) and 222 (95% CI 210–240) depending on use of the Poser (727 definite and probable cases) or Allison and Millar (880 early, probable and possible cases), respectively. Use of the capture–recapture method suggested that ascertainment was >93%. Previously reported geographical gradients were considered largely to be methodological and did not hold up after age and sex corrections. A second study, in Fife to the north of Edinburgh, identified 508 patients (2.4F:M) by postal questionnaire providing a standardized prevalence ratio of 178/10⁵ (R.M. Grant et al 1998). A recent cross-sectional survey of 169 000 Glaswegians showed a prevalence of 145/10⁵ and incidence at 5.7/10⁵/year; the crude prevalence in Asians was 63/10⁵ (S. Murray et al 2004). As in Carlisle, England, when Brewis et al (1966) reported on the development of neurological disease between 1946 and 1961, finding a prevalence for multiple sclerosis of 82/10⁵, B.K. MacDonald et al (2000) prospectively surveyed a population of 100 230 patients registered in 13 general practices in Leeds for neurological diseases. The estimated incidence of multiple sclerosis was 7/10⁵/year (2.8F:M). Lifetime prevalence rates were reported for a subset of 27 658 in whom multiple sclerosis affected 1:500. By chance, a recent conventional prevalence study also comes from Leeds. H.L. Ford et al (2002) reported incidence for multiple sclerosis in 792 people with multiple sclerosis (2.3F:M) over the period 1996–9 at 6.1/10⁵/year (95% CI 5.1–7.2). The prevalence changed between October 1996 and 1999 from 93/10⁵ (95% CI 86–101: H.L. Ford et al 1998a) to 109/10⁵ (95% CI 101–116, including 14% suspected cases). Mortality, based on 57 deaths, was 3.2/10⁵/year, reflecting incomplete notifications of mortality and their poor standing as reliable indicators of disease frequency. C.M. Fox et al (2004) established prevalence for multiple sclerosis in a first survey from Devon. Using the capture–recapture method to confirm at least 94% ascertainment, they identified 446 cases in a population of 341 796 on 1st June 2001, providing prevalences of 117 (95% CI 106–129) and 118/10⁵ (95% CI 105–128), classifying cases using the criteria of either W.I. McDonald et al (2001) or C.M. Poser et al (1983).

Even now it is difficult to reliably map the distribution of multiple sclerosis across the United Kingdom. Diagnostic criteria, inclusion or omission of suspected cases, quotation of confidence intervals, variable citation of crude and age-adjusted figures, reworking of statistics for time of onset rather than diagnosis, and adjustment to a standardized population must all be taken into account when attempting a reliable overview. Until the mid-1980s, studies of multiple sclerosis in the United Kingdom used the system of classification suggested by Allison and Millar (1954). Adapting to the Poser criteria (C.M. Poser et al 1983) does not materially affect estimates for the total number of prevalent cases or standardized prevalence ratios. However, differences do arise when surveys are restricted to definite cases. Readers often fail to separate statistics that do and do not include the fringe of suspected cases. Although differences in numerator are apparent with the recent introduction of revised diagnostic criteria (W.I. McDonald et al 2001), comparing the frequency of multiple sclerosis using the Allison and Millar classification with more recent publications is more problematic. For example, the surveys from Northern Ireland and southeast Scotland show differences in prevalence of 28 and 38/10⁵, respectively, depending on which classification is used. But the main source of variation lies in the extent to which separate regions have been subjected to the same degree of epidemiological scrutiny.

Previously, the pattern appeared to show a marked difference in frequency between the northeast mainland and offshore islands of Scotland compared with other parts of the United Kingdom. This was most apparent when the estimates for prevalence of multiple sclerosis were being serially updated in northeast Scotland (Downie 1984; Phadke and Downie 1987; Shepherd and Downie 1978; 1980; Sutherland 1956) and in the Orkney Islands (Allison 1963; Fog and Hyllested 1966; Poskanzer et al 1980a). Over that period, the highest mainland prevalences were reported for Aberdeen in northeast Scotland, with figures of 127 (95% CI 116–137) in 1970, 144 (95% CI 133–156) in 1973, and 178 (95% CI 166–191) in 1980. Onset adjustment by C.M. Poser et al (1992) on these published figures raised the prevalences for northeast Scotland to 117 (95% CI 107–127), 117 (95% CI 107–127) and 139 (95% CI 129–147), respectively. By 1974, when very few other parts of the United Kingdom had been surveyed, the prevalence for Orkney stood at 309/10⁵ (95% CI 237–404; Poskanzer et al 1980a). S.D. Cook et al (1985) documented the annual incidence from 1941 to 1983 and suggested that there had been a steady reduction from 1964. By 1983, the quoted prevalence had also fallen from 309/10⁵ in 1974 to 224/10⁵ (the figures were 257/10⁵ and 193/10⁵ for probable cases only in 1974 and 1983, respectively).

With the steady rise in prevalence for southern parts of England and the more stable rates in northeast Scotland, there appears to have been a steady reduction in slope of the previously demonstrated gradient in frequency. Without knowing the extent to which surveys of multiple sclerosis in northeast Scotland have saturated prevalent cases, and with continuing uncertainty on whether other parts of the country have yet reached a steady state, it is difficult to predict how much further this gradient will collapse. Forbes and Swingler (1999) assessed the extent to which under-ascertainment, quantitated on the basis of capture–recapture comparisons, is sufficient to account for the latitudinal gradient. However, even after adjustment, they still found higher rates in northern parts of the United Kingdom (>180/10⁵) compared with the south (<160/10⁵). Despite the possibility of waning incidence in Orkney and Shetland, we also retain the view that the northeast of Scotland genuinely has a higher frequency of multiple sclerosis than other parts of the United Kingdom, although a systematic change in prevalence, correlating with latitude throughout the United Kingdom, now seems less likely. If the recent temporal trends are due to altered biological factors that determine the frequency of multiple sclerosis, it follows that these must be environmental, albeit affecting a fertile population, and not genetic. But our preferred explanation is that the epidemiological trends are due to saturation in northeast Scotland, and a catching-up effect elsewhere, resulting from incomplete ascertainment in the early surveys. Therefore, we retain the view that the distribution of multiple sclerosis in the United Kingdom is real and reflects differences in genetic characteristics of the population at risk (see Chapter 5).

MULTIPLE SCLEROSIS IN THE UNITED STATES (Figures 2.11 and 2.12)

Figure 2.11.

Prevalence distribution of multiple sclerosis in the United States. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

Figure 2.12.

Case–control ratios (× 100) for multiple sclerosis by state of residence at entry into active duty for white male veterans of the Second World War.

Adapted from Brown JR et al (1979). © 1979, reprinted with permission of Lippincott Williams & Wilkins (lww.com).

© 2006 Lippincott Williams & Wilkins

The regional frequency of multiple sclerosis in continental North America was first mapped in detail by Limburg (1950) and subsequently by Kurtzke (1993). The highest prevalence is seen in the northwest of the United States (and southern parts of Canada). The lowest figures for prevalence occur in the southeastern states. The combination of a large land mass, and a complex cultural history with substantial and ethnically diverse immigration over the last two centuries, makes it difficult to see clearly the epidemiological picture of multiple sclerosis in North America and to interpret the pattern.

The point that systematic change in prevalence, but not necessarily incidence, is to be expected over time in an area subjected to repeated scrutiny is more than clear in the studies of multiple sclerosis from North America. The surveys from Olmsted County, Minnesota, organized through the Mayo Clinic, are exceptional in terms of the high case ascertainment and sequential assessments of disease frequency made over a period of 100 years (Wynn et al 1990). Incidence appeared to remain stable in Rochester at about 3.6/10⁵/year from 1905 to 1974 (men 2.8/10⁵ and women 6.8/10⁵). However, re-evaluation of all patients classified as having possible multiple sclerosis over the period 1905–1984 (Wynn et al 1990) revealed a trend towards increasing incidence, especially for women, in whom crude incidence rose from 3.4 to 7.7/10⁵. It remains uncertain whether this increase in incidence merely reflects alterations in application of diagnostic criteria and improved ascertainment. These factors, together with improved survival, largely account for the systematic rise in prevalence from approximately 46/10⁵ in 1915 (Percy et al 1971) to 108/10⁵ in 1978 (Kranz et al 1983), 173/10⁵ in 1985 (Wynn et al 1990) and (based on retrospective analysis) 160/10⁵ in 1990 (Wynn et al 1990). At the last count, the crude prevalence was 177/10⁵ on 1st December 2000 and incidence was 7.5/10⁵/year over the period 1985–2000. Sex and age adjustment standardized to the 1950 United States Caucasian population showed that the rates have been stable over the last 20 years (Mayr et al 2003). The analysis does suggest a small but real increase in incidence from the early 1900s (accounting to some extent for the rise in prevalence) but with interruptions in the otherwise steady rise. Cyclical changes were apparent in the 1910s, 1930s, 1950s and 1980s. Despite the care with which these statistics were gathered, difficulties remain in developing elaborate aetiological hypotheses based on small trends in incidence and, in Rochester, the rate has been steady at 7–8/10⁵ over the last two decades (Weinshenker and Rodriguez 1995). Broadly similar figures are quoted for neighbouring Mower County, Minneapolis (106/10⁵ in 1978; Kranz et al 1983). By comparison with other parts of the world, rather few more recent surveys on the epidemiology of multiple sclerosis have been reported from the United States. In northern Colorado, the prevalence was 65/10⁵ in 1982 (L.M. Nelson et al 1986). Hopkins et al (1991) estimated the prevalence in Galion, Ohio, at 112/10⁵ (95% CI 64–174) on 1st June 1987 amongst a population of 15 161.

The use of defined cohorts has formed the basis for mapping trends in the epidemiology of multiple sclerosis in the United States. John Kurtzke first studied 5305 United States army personnel who served in the Second World War or Korean conflict and were judged by the Veterans Administration to have symptoms of multiple sclerosis recorded during or within 7 years of military service (Kurtzke and Page 1997; Kurtzke et al 1979; 1985; 1992; Norman et al 1983; W.F. Page et al 1993; 1995; Wallin et al 2000). The key finding on which all subsequent analyses have been predicated is that prospective incidence series showed high and low rates for multiple sclerosis above and below the 37th parallel, respectively. The gradient is not strictly latitudinal in that it also shows a west–east polarity. Kurtzke has demonstrated an overall distribution for incidence that matches that derived for mortality (Figure 2.12). Comparisons of location at birth and entry into military service, ancestry, demographics and natural history have allowed John Kurtzke and colleagues subsequently to assemble a picture of multiple sclerosis in the United States orientated towards understanding the aetiology from the perspective of epidemiology. From these various opportunities has emerged the principle of an effect on disease frequency of migration between regions of differing risk (see below). W.F. Page et al (1993) analysed the distribution of multiple sclerosis in these 5305 veterans with respect to ancestry, having noted differences in frequency of the disease within the northern and southern tiers of high and low risk, respectively. Multiple sclerosis appeared more common in Louisiana, Arizona and southern California – with a high proportion of French and Spanish ancestors – than in other southern states. The authors confirmed the link between Scandinavian (and in some analyses Scottish, French and Italian but not English or Dutch) ancestry and risk of multiple sclerosis. Ancestry accounted for 45–60% of state-by-state variance and correlated more closely with the risk of multiple sclerosis than univariate analysis of personal ethnicity (W.F. Page et al 1995). But, however the issue is approached, the correlation of genetic history and contemporary risk of multiple sclerosis is apparent. Subsequent analysis of the United States veterans cohort drew attention to higher socioeconomic status and years of education as risk factors for multiple sclerosis in whites and black women (Kurtzke and Page 1997), although effects of ethnicity and geography may have confounded this complex multivariate analysis. As discussed in Chapter 4, survival in around 2500 veterans from the Second World War who had developed service-connected multiple sclerosis by 1956 was estimated in 1996 (Wallin et al 2000). Median survival times from onset were 43, 30 and 34 years for white females, black males and white males, respectively. Crude 50-year survival frequencies were 32%, 22% and 17%, for these same groups. Reduced survival correlated with male sex (proportional hazard, 1.57), older age at onset (risk ratio, 1.05/year), and high socioeconomic status (risk ratio, 1.05/socioeconomic status category). Factors such as race or place at entry into the military did not influence survival. Standardized mortality ratios showed an excess attributable to multiple sclerosis that declined over the period 1956–1996.

Wallin et al (2004) subsequently described 4951 Vietnam era veterans in whom comparisons were made with 9378 matched controls. The cohort is considered representative of contemporary multiple sclerosis in the United States and is additionally useful as a comparator for the previous series of veterans with multiple sclerosis. Differences from the Second World War/Korean conflict series included an increase in the number of women and blacks at risk, as a result of demographic changes within the military. Cases were classified, as before, using the Schumacher and Poser criteria. The main findings were an increase of multiple sclerosis amongst blacks and in women, with partial erosion of the north–south gradient in disease frequency for state of residence at birth or service entry. The possibility exists of confounding effects on these apparently independent findings. Arguing that genetic factors could not have changed over the period during which altered risks for multiple sclerosis were documented in the two series of veterans, the authors promote the primary influence of environmental factors. Others may yet feel that these epidemiological trends illustrate the interplay of altered environmental exposures in populations that have intrinsically different risks. Not least, cultural change has led to significant but geographically uneven Afro-American admixture in this generation.

Hernán et al (1999) also adopted the powerful approach of studying a professional group in whom health records made for easy determination of numerator and denominator. They identified new cases of multiple sclerosis amongst the Nurses Health Study (carried out during 1976–1994 on nurses born between 1920 and 1946), and the Nurses Health Study II (carried out during 1989–1995 on nurses born between 1947 and 1964). Based on 181 cases, adjusted incidence rates showed a ratio of 3.5 (95% CI 1.1–11.3) for the northern and 2.7 (95% CI 0.8–8.9) for the middle compared with the southern states, respectively, in the first cohort. However, this gradient was not confirmed in Nurses Heath Study II in which 131 cases were identified providing adjusted incidence rate ratios of 0.8 (95% CI 04–1.6) and 0.9 (95% CI 0.4–1.8) for the northern and middle banded states compared with the south, respectively. This attenuation of the gradient for disease frequency matches the erosion reported for veterans (see above) but has the potential confounder that the age ranges for nurses in the first cohort were 30–74 years, compared with 25–48 years in Nurses Heath Study II. Thus, incidence may not yet be fully declared in the latter.

Taken together, and using a variety of adjustments from the available morbidity statistics, D.W. Anderson et al (1992) estimated that there were about 350 000 physician-diagnosed patients with multiple sclerosis in the United States in 1990, compared with 123 000 in 1976. This defines the health care needs, but can anything be learned about aetiology from the patterns defined in the United States? After Davenport (1921; 1922), Bulman and Ebers (1992) re-emphasized the importance of population genetics as a risk factor in North America by correlating the frequency of multiple sclerosis (mapped by Kurtzke) with the distribution of people having Scandinavian ancestry. Of course, social habits shared amongst ethnic communities introduce a potential confounding factor, and the similarities in distribution of multiple sclerosis and ancestry do not amount to evidence per se for a genetic mechanism. Our own view is that the distribution of multiple sclerosis in the United States has been shaped by patterns of immigration. High-risk areas in the mid-West were originally mainly populated from northern Germany and Scandinavia. The low-risk areas of the Mississippi delta had a high density of people of African descent. With time, these original groups have moved and mixed thus eroding the differential rates of multiple sclerosis. Despite the genetic stance, this analysis also requires that environmental events triggered the disease process in individuals at risk. It does not exclude geographical and temporal trends in those exposures.

MULTIPLE SCLEROSIS IN CANADA (Figure 2.13)

Figure 2.13.

Prevalence distribution of multiple sclerosis in Canada. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

The prevalence of multiple sclerosis has been systematically assessed in several Canadian provinces, and some have been serially updated. A first survey of Winnipeg, Manitoba, based on information from case notes and death certificates for the years 1939–1948, reported a prevalence on 1st January 1951 of 40/10⁵ compared with 6/10⁵ in New Orleans (Westlund and Kurland 1953). Two years earlier, the rates in Winnipeg were 42/10⁵ compared with 13 and 41/10⁵ in New Orleans and Boston, respectively (Kurland 1952). Prevalences of 64 and 53/10⁵ were cited from other authors for Rochester and Kingston, Ontario. At reassessment in 1961 (Stazio et al 1964), the diagnosis was reviewed in 149 individuals included in the first Winnipeg survey. This exercise was the first to scrutinize diagnostic inaccuracy in detail. Of 112 cases classified as having probable multiple sclerosis in 1951, this degree of certainty held up in only 85. Apart from 3 who were no longer prevalent, 3 others were known not to have the disease, 7 remained as possible cases and in 14 the diagnosis was now thought to be unlikely. In the category of 22 possible cases from 1951, 13 were either known not to have multiple sclerosis or this diagnosis was subsequently considered unlikely. Conversely, of the 15 with dubious diagnostic status in 1951, only 2 had matured clinically and were thought to have multiple sclerosis at follow-up. With the retrospective application of these corrections, prevalence in 1951 should have been 36/10⁵, and the new (1960) prevalence almost identical at 35/10⁵. This is surprising since practically every other serial estimate of prevalence has shown an increased rate, due to improved ascertainment and survival, even when methods and diagnostic criteria are standardized.

Published prevalence rates in Canada have since shown a systematic increase in frequency. This has been most obvious in the study from Saskatoon, Saskatchewan, where estimates for incidence at 4.8/10⁵/year and prevalence of 134/10⁵ (Hader 1982) prompted the reaction that Saskatoon should be scrutinized in detail to determine why the frequency was so high. However, it soon became clear that, after allowing for the contributions of improved survival and high case ascertainment resulting from arrangements for the provision of health care, these statistics more accurately reflected the overall prevalence of multiple sclerosis in southern Canada. Subsequent studies, ranging from both coasts of Canada and including the province of Ontario, have shown comparable figures. In British Columbia, the rate was 93/10⁵ for probable multiple sclerosis and 130/10⁵ for all cases in 1982 (V.P. Sweeney et al 1986). Prevalence in Ottawa was 68/10⁵ in 1975 (Bennett et al 1977) and 94/10⁵ in the first survey from London, Ontario, for 1984, at which time 70% of the population was from the United Kingdom and 23% from continental Europe (Hader et al 1988). The rate in Saskatoon levelled out at a more representative frequency of 111/10⁵ by 1977 (Hader 1982). Conversely, in Newfoundland the first estimate of prevalence was only half that reported for other Canadian centres (55/10⁵; Pryse-Phillips 1986) but this was subsequently adjusted to 95/10⁵ in 2001, with stable incidence rates over the period 1994–2001 at 5.6/10⁵/year (Sloka et al 2005a). Regional rates were those predicted from contemporary estimates of disease frequency for places in the United Kingdom from which these immigrants had originated (Sloka et al 2005b).

Other reports for Canada were provided from southern Alberta where the prevalence in Cardston and Crowsnest Pass was 87/10⁵ and 202/10⁵, respectively, on 21st June 1989 (G.M. Klein et al 1994). Incidence was 4.2/10⁵/year between 1980 and 1989, an increase from 1.3/10⁵/year for 1950–1959 in Barrhead County, and prevalence 196/10⁵ in 1990 (S. Warren and Warren 1992). Intensive scrutiny of a smaller area bordering Barrhead County showed comparable figures at 200/10⁵ and 7.3/10⁵/year for prevalence and incidence, respectively, suggesting that ascertainment was more or less saturated in the larger survey (S. Warren and Warren 1993). Mortality data relating to multiple sclerosis in the period 1965–1994 were highest in the provinces of Quebec (4.4/10⁵/year) and Ontario (3.9/10⁵/year), intermediate in the western provinces (2.1/10⁵/year), and lowest on the eastern Atlantic seaboard provinces (1.2/10⁵/year; S. Warren et al 2003). As expected, rates were higher in females than males, and in those aged >65 years. Taken together, average annual mortality rates fluctuated without any directional trend. These data do not match the distribution of statistics for prevalence, perhaps because of demographic differences in the age and sex structures of the populations at risk throughout Canada. Seven aboriginal people with multiple sclerosis were identified by Mirsattari et al (2001) in a study from Manitoba giving a prevalence of 40/10⁵ between 1970 and 1996, <50% of the rate expected for Canadians of northern European origin living in the mid-western provinces. However, the difference may be even greater since five of the seven cases had a phenotype dominated by relapsing spinal cord and optic nerve disease – as seen in several other racial groups with a low overall frequency of multiple sclerosis (see Chapter 5)

MULTIPLE SCLEROSIS IN AUSTRALIA AND NEW ZEALAND (Figure 2.14)

Figure 2.14.

Prevalence distribution of multiple sclerosis in Australia and New Zealand. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature. Minor differences from figures appearing in the original publications reflect our recalculations.

The surveys carried out in Australia and New Zealand over the last 30 years have been especially influential since they have maintained sufficiently consistent methodology over time, allowing temporal and regional comparisons to be made with reasonable confidence. The geographical area under scrutiny is large and incorporates considerable variation in latitude, climate and racial origins. Caucasians have been domiciled in Australasia for over a century, and new migrant groups – Mediterranean and Oriental – have since arrived. However, in other respects the data do not lend themselves to internal comparisons. Four different systems for classification have been used. There has been inconsistency in the choice of denominator within defined geographical regions, for example with respect to the inclusion of aboriginal peoples (notably Maoris in New Zealand). Standardized prevalence ratios have not been adjusted to the same population. The account of morbidity statistics for 1981 (S.R. Hammond et al 1987) contains a transcription error from 1961 (McCall et al 1968) which, when corrected, reduces the apparent latitudinal gradient.

In the first comprehensive survey in Australia (McCall et al 1968), rates for 1961 were based on the Allison and Millar criteria. A south–north gradient in age-standardized prevalence was apparent, ranging from 34/10⁵ in Hobart, Tasmania, to 19/10⁵ in Perth, Western Australia, and 18/10⁵ in Newcastle, New South Wales. By 1981, these rates had risen to 74, 29 and 38/10⁵, respectively (S.R. Hammond et al 1988). On that occasion, cases were classified using the criteria of Rose et al (1976). In surveys carried out by the same investigators (S.R. Hammond et al 1987), prevalence in Queensland was 11 and 21/10⁵, north and south of the tropic of Capricorn, respectively. These figures compared with prevalences of 7 and 12/10⁵ for the same areas in 1961 (Sutherland et al 1966), using the World Federation of Neurologists diagnostic criteria. Later, McLeod et al (1994) added prevalence figures for South Australia (29/10⁵) and reiterated the conclusion that, in the absence of an obvious genetic difference accounting for the sevenfold reduction in prevalence in tropical Queensland compared with Hobart, Tasmania – and despite the lack of any affected Aborigines or Torres Strait Islanders – the main determinant of the Australian gradient was environmental. Using the criteria of Rose et al (1976), prevalence in the Australian Capital Territory was 57/10⁵ (95% CI 43–74) on national census day in 1996 (R.D. Simmons et al 2001). The sex ratio was 2.4F:M. Reclassification using the C.M. Poser et al (1983) criteria altered the prevalence to 49/10⁵ (95% CI 42–58). Prevalence remained disproportionately high when corrected to the 1981 population by direct comparison with earlier surveys from other parts of Australia. Because disease duration in this cohort was short, longstanding cases may have been missed thereby leading to underestimate of disease frequency. The statistics have been updated for Newcastle, New South Wales, in 1996 with incidence and prevalence at 2.4/10⁵/year and 59/10⁵, respectively. The increase was attributed to altered incidence in females, aged 20–29 years, and improved survival, also in females (M.H. Barnett et al 2003).

S.R. Hammond et al (2000a) have again summarized the distribution of multiple sclerosis in five regions of Australia, focusing on individuals who had migrated from the United Kingdom and Republic of Ireland. There is a significantly higher prevalence ratio of multiple sclerosis amongst Australians leaving school at an older age and achieving a higher educational level (S.R. Hammond et al 1996). As before, a latitudinal gradient is apparent but the geographical differences are much influenced by the high prevalence in Hobart, Tasmania. Elsewhere, the frequency of multiple sclerosis appears lower than is apparent from contemporary studies in the places of origin within the United Kingdom and Ireland.

In New Zealand, the prevalence of multiple sclerosis was estimated at 69/10⁵ in Wellington (south of the North Island) in 1984 (D.H. Miller et al 1986a). Comparable figures were reported for the South Island (69/10⁵ in Otago and Southland; Skegg et al 1987) with much lower rates in northern parts of the North Island (e.g. 24/10⁵; 95% CI 18–30 in Waikato). These prevalence studies were consistent with an earlier report of nationwide hospital admission and mortality statistics: both measures suggested that the disease was more common in the South Island and the southern North Island when compared with the northern half of the North Island (Hornabrook 1971). Recently, however, prevalence has been estimated at 50/10⁵ (95% CI 40–62) on 15th January 2001 for a northerly part of the North Island (Bay of Plenty: Chancellor et al 2003). Fawcett and Skegg (1988) reviewed mortality rates for first admission to hospital in New Zealand excluding individuals with partial or full Maori ancestry. Incidence showed a south–north gradient, decreasing from 6.4/10⁵/year in Otago, in the south of the South Island, to 2.7/10⁵ in Auckland, which is in the north of the North Island. Age-adjusted mortality varied in parallel from 1.2 to 0.7/10⁵. Admissions and deaths were rare in Maoris (20 and 2 cases, respectively). These observed frequencies represented 1.2% of admissions compared with 8.8% expected from the size of the Maori population.

Surveys reporting the frequency of multiple sclerosis in Australia and New Zealand were collated by D.H. Miller et al (1990a). The numerators for both countries were compared with an age-adjusted denominator excluding aboriginal peoples only in New Zealand – potentially an important confounder since 16% of the North Island population is of Maori origin. Evidence was presented for a sevenfold difference in frequency between Queensland and Otago. Although the quoted figures were age corrected, results were not adjusted to account for the use of different methods of case classification (usually Allison and Millar or McDonald and Halliday). In New Zealand, the main step in morbidity seems to occur across the North Island. By comparison, Tasmania (Australia) has a much higher prevalence than Waikato even though both are of comparable southerly latitude. Our impression is that the whole Australasian region falls into two clusters: Hobart (Tasmania), Wellington and Otago (South Island) with rates of >75/10⁵ and Queensland, Newcastle (New South Wales), Adelaide (South Australia), Perth (Western Australia) and Waikato (North Island) with rates of <40/10⁵. Furthermore, multiple sclerosis remains rare in the small community of Aborigines living in mainland Australia and in the Maoris of New Zealand.

Several interpretations of these patterns have been offered. Using the indirect methods adopted by Skegg et al (1987) and Swingler and Compston (1986), D.H. Miller et al (1990a) argued that the latitudinal gradient is not explained by genetic clines. To us, the methods used do not exclude significant heterogeneity in the distribution of white populations in Australia or New Zealand (Compston 1990b). Sorting ancestry on the number of Mc/Mac prefixes in the telephone directory (as used by Skegg et al 1987) does not differentiate Nordic from Celtic peoples, although it is recognized that the predominant settlement of the southern half of the South Island of New Zealand – where multiple sclerosis is common – was Scottish. A proportion of the population from Waikato who declared themselves white nevertheless had up to 50% Maori ethnicity. (In recent times, claiming Maori ancestry has carried social and financial benefits.) More speculative is the suggestion that immigrant groups head for and settle in places having climatic similarities with their homeland, thus tending to maintain localized genetic clines until equilibrium is established. That said, it is worth pointing out that the highest rates for multiple sclerosis seen in Australia are still only half those reported from parts of northern Europe having the same ethnic constitution. This suggests that relative protection is afforded by the environment of the Australian continent for people of European origin.

MULTIPLE SCLEROSIS IN CONTINENTAL EUROPE (Figure 2.15, Figure 2.16, Figure 2.17, Figure 2.18, Figure 2.19, Figure 2.20)

Figure 2.15.

Prevalence distribution of multiple sclerosis in central Europe (Belgium, the Netherlands, Switzerland, Austria, Germany, Hungary, Slovakia, the Czech Republic, Poland). Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

Figure 2.16.

Prevalence distribution of multiple sclerosis in France. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

Figure 2.17.

Prevalence distribution of multiple sclerosis in Spain and Portugal. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

Figure 2.18.

Prevalence distribution of multiple sclerosis in Greece and the Balkans (Croatia, Romania, Bulgaria). Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

Figure 2.19.

Prevalence distribution of multiple sclerosis in Italy. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

Figure 2.20.

Incidence distribution of multiple sclerosis in Italy. Most recent figures are given for incidence/year/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

Working versions of geographical differences in the frequency of multiple sclerosis have been subjected to recent revision more in Europe than most other parts of the world. The trend has been for gradual erosion of latitudinal gradients as countries not previously subjected to detailed epidemiological scrutiny, but lying within or adjacent to the high and medium prevalence zones of Kurtzke's 1975 analysis, were studied in detail.

In southern Lower Saxony, Germany, mean annual incidence increased from 2.6/10⁵ to 4.6/10⁵ (S. Poser et al 1989b), and prevalence increased from 51 to 118/10⁵, between 1969 and 1989, for all the usual reasons (Poser and Kurtzke 1991). Elsewhere, prevalence ranges somewhat erratically both with respect to latitude (between 49° and 54°N) and time. In 1982, the rate was 67/10⁵ for Rostock (Meyer-Rienecker and Buddenhagen 1988) and 54/10⁵ in Darmstadt (Prange et al 1986). In the mid-1980s, it was 62/10⁵ in Stralsund and the Isle of Rugen (Meyer-Rienecker 1994). By the early 1990s, rates of 98/10⁵ were reported from Bochum (Haupts et al 1994). In 1992, prevalence was 85/10⁵ and 108/10⁵ (excluding patients with possible multiple sclerosis) in southern Hesse and southern Lower Saxony, respectively (Lauer and Firnhaber 1994, S. Poser 1994). Extrapolation to a large region, based on sampling a proportion of the population at risk, has been used to derive a more contemporary prevalence of multiple sclerosis in Germany of 150/10⁵ (Hein and Hopfenmuller 2000).

In western Poland, prevalence fluctuated between 51 and 43/10⁵ between 1965 and 1981 (Wender et al 1985) but the most recent estimate is 62/10⁵ for Szczecin (Potemkowski et al 1994). The fall in prevalence between 1965 and 1981 is probably explained by a change in the population structure due to a higher birth rate, artificially expanding that proportion of the population at risk but below the usual age of onset for multiple sclerosis (Cendrowski et al 1969, Wender et al 1985). Other contemporary surveys include the estimate of prevalence for native Estonians, Russians and other nationalities of 55, 29 and 42/10⁵, respectively, in southern Estonia (Gross et al 1993). Inthe Czech Republic, prevalence was 51 and 89/10⁵ in east and northwest Bohemia, respectively (Jedlicka et al 1994). In Hungary, rates were initially reported as ranging from 32 to 79/10⁵ (Guseo et al 1994; Palffy et al 1994). However, in Szeged, Hungary, at 31st December 1996 incidence was 7/10⁵/year and prevalence 65/10⁵ (Bencsik et al 1998), and prevalence in Csongrad County, Hungary, reported as 62/10⁵ (Bencsik et al 2001). The issue of whether Gypsies are protected from developing multiple sclerosis remains important in debating the relative contributions of race and environment in determining risk, since comparisons can be made with the indigenous peoples of central Europe. Palffy et al (1994) reported a prevalence of 5/10⁵ amongst a Gypsy population of 22 000 including two autopsy-proven cases. This compared with a frequency of 32/10⁵ in Hungarians. Previously, Palffy (1982) had shown population differences in gene frequency for markers of susceptibility to multiple sclerosis, offering an explanation for the apparent resistance of Gypsies to multiple sclerosis. However, even these carefully conducted studies have examined the frequency of multiple sclerosis in only a small proportion of the Gypsy population of central Europe. On 31st March 1998, the prevalence of multiple sclerosis in two regions of Bulgaria was 45 and 44/10⁵ compared with rates of 19 and 18/10⁵ in Gypsies from the same locations (Milanov et al 1999).

In Switzerland, the most recent prevalence was 110/10⁵ for the Canton of Berne (Beer and Kesselring 1994). Using the capture–recapture method and comparing cases identified from 30 multiple sclerosis clinics over a 4-week period and records of the Multiple Sclerosis Society, Baumhackl et al (2002) calculated that there were around 7900 affected individuals in Austria providing a nationwide prevalence of 99/10⁵. Whilst this does not reveal regional patterns throughout Austria, an earlier survey had reported a more or less similar rate for Lower Austria (Baumhackl 1995). In the Netherlands, incidence increased from 2 to 5.9/10⁵/year between 1982 and 1992, over which period prevalence rose from 54/10⁵ to 76/10⁵ (Minderhoud and Zwanniken 1994; Prange et al 1986). In Flanders, the most recent estimate for prevalence was 74/10⁵ (95% CI 59–89; van Ooteghem et al 1994). In general, figures for incidence in other parts of Europe show the same temporal trends, ranging from 0.8 to 3.9/10⁵/year for the regions discussed above and correlating, as expected, with differences in prevalence.

France appeared for some time to be a region having prevalences for multiple sclerosis that were genuinely lower than expected from its geographical location within Europe, if the gradient is latitudinal and not shaped by social and genetic history. Rates for Chalon-sur-Saône (southeast), Avignon (south) and Pyrénées-Atlantiques (southwest) were 58 (95% CI 46–71), 49 (95% CI 39–60) and 38/10⁵ (95% CI 33–44) in the mid-1980s, respectively (Confavreux et al 1987; M-P. Roth et al 1994a). These figures compare with 25/10⁵ for Brittany in 1978 (Gallou et al 1983). But the most recent epidemiological study provides rates for incidence that suggest comparability with other parts of continental Europe. Based on 21 identified cases, Moreau et al (2000) showed that the incidence in Dijon, Burgundy, was 4.3/10⁵/year (2.5F:M) between 1993 and 1997.

Morbidity statistics for Spain first reported prevalence rates of <10/10⁵ (Oliveras De Lariva et al 1968). Subsequently, estimates for incidence were 3.7/10⁵/year (95% CI 1.4–6) in Gijon (Uria et al 1998), approximately 2.2/10⁵/year in Teruel (Modrego Paedo et al 1997), and 3.0/10⁵/year (95% CI 1.8–4.2) in Alcoi (Matias-Guiu et al 1994). Corresponding rates for prevalence increased from much lower estimates made in the 1960s to 23/10⁵ (95% CI 25–74) for Gijon (northern Spain), later updated to 65/10⁵ (95% CI 38–92; Uria et al 1997); 57/10⁵ (95% CI 40–75) in Barcelona (Catalonia, western Spain); 58/10⁵ in Osana (which lies to the north of Barcelona; Bufill et al 1995); 32/10⁵ (95% CI 23–41) in Teruel to the southwest of Barcelona (Modrego Pardo et al 1997); 53/10⁵ (95% CI 32–83) in Malaga (southern Spain; O. Fernandez and Bufill 1994; O. Fernandez et al 1994); and 17/10⁵ in Alcoi (eastern Spain; Matias-Guiu et al 1990; 1994). The most recent wave of surveys includes prevalence for northern Spain (Valladolid) at 58/10⁵ (95% CI 44–76; Tola et al 1999). In Bajo Aragon, prevalence rose between 1994 and 2002 from 35/10⁵ (95% CI 20–50) to 75/10⁵ (95% CI 52–97) with a corresponding increase in incidence, between 1984–1993 and 1994–2003, of 3/10⁵/year (95% CI 1.6–4.5) to 4.6/10⁵/year (95% CI 2.8–6.5) representing an increased standardized incidence ratio of 1.4 (95% CI 0.95–2.17; Modrego and Pina 2003). In Calatayud, northern Spain, annual incidence was 2.6/10⁵/year between 1980 and 1989, and prevalence also 58/10⁵ (95% CI 39–78) on 1st April 1995 (Pina et al 1998). Incidence in Mostoles, central Spain, was 3.8/10⁵/year (95% CI 2.7–5.3) between 1992 and 1997, and prevalence 43/10⁵ (95% CI 35–54; Benito-Leon et al 1998) on 1st February 1998. Incidence was 3.4/10⁵/year (95% CI 2.2–5.3) in the Balearic island of Menorca and prevalence 69/10⁵ (95% CI 50–92) in 1996 (Casquero et al 2001).

In the Canary Island province of Las Palmas, prevalence was 6/10⁵ (95% CI 1–22) in the early 1980s (Sosa Enriquez et al 1983). This was adjusted to 15/10⁵ (95% CI 8–25) on the neighbouring island of Lanzarote in 1987 (Garcia et al 1989). By 15th December 1998, incidence was 2.25/10⁵/year and prevalence had increased in Las Palmas to 42/10⁵ (Hernandez 2002), eliminating the earlier apparent fivefold difference in frequency between the Canary Islands and southern parts of the Iberian Peninsula. The most recent survey (to 31st December 2001) provides estimates of 74/10⁵ (95% CI 56–95) using the criteria of McDonald et al (2001), with rates of 62/10⁵ (95% CI 47–79) and 71/10⁵ (95% CI 55–89) adjusted to the standardized world and European populations, respectively (Aladro et al 2005). A review of mortality trends in Spain between 1951 and 1997 shows a reduction from 3.1 to 0.6/10⁵/year before and after 1967, respectively, attributable in part to improved life expectancy but also to changes in codification practice (Llorca et al 2005).

In considering other regions of Spanish ancestry, Alvarez et al (1992) identified 68 cases of multiple sclerosis in Chile and noted that this constituted the largest cohort yet surveyed but did not calculate prevalence. A rate of 4.3/10⁵ (2F:M) was estimated by Callegaro et al (1992) for São Paolo in Brazil in 1990. This was updated for 1st July 1997 to 15/10⁵ (Callegaro et al 2001). The population is typically Caucasoid and mainly of Portuguese origin but with recent additions of Spaniards, Italians, central Europeans and Japanese.

Whilst momentum is gathering for Iberian epidemiological studies, Greece and the Balkan states have yet to show the same epidemiological commitment. This may explain why their estimates have until recently remained lower. Where available, incidence rates largely correlate with prevalences, being higher in Croatia (5.9/10⁵/year; 95% CI 4.3–7.8; Sepcic et al 1989) and Slovenia (2.9/10⁵/year; Koncan-Vracko 1994) than in Greece (1.8/10⁵/year; 95% CI 1.6–1.9; Milonas 1994). Prevalence in Belgrade (Serbia) was 42/10⁵ on 31st December 1996 (1.9F:M; Pekmezovic et al 2001). Survival in the same sample has been estimated at 38 years with a 73% probability of 25-year survival for the prevalent population (Pekmezovic et al 2002). Prevalence was 83/10⁵ in Slovenia (Koncan-Vracko 1994), 26/10⁵ in Romania (Petrescu 1994), 30/10⁵ (95% CI 27–33) in Bulgaria (Georgiev and Milanov 1994), 124/10⁵ (95% CI 89–169) in Gorski Kotar, Croatia (Sepcic et al 1989), and 29/10⁵ (95% CI 27–32) in Macedonia and Thrace, Greece in 1984 (Milonas 1994). Serial update of the prevalence for Evros in northeastern Greece showed increases from 10/10⁵ in 1984, to 30/10⁵ in 1990, and 39/10⁵ on 31st December 1999 (Piperidou et al 2003). Because incidence increased between 1974–1978 and 1994–1999 from 0.66 to 2.36/10⁵/year, Piperidou et al (2003) conclude that the increase in frequency reflects a change in aetiological conditions over and above the impact of methodological factors.

Multiple sclerosis has been surveyed in three Greek-speaking regions of Cyprus. The coastal areas under scrutiny included more refugees from northern Cyprus (Turkish speaking) than the mountain villages. Overall, the prevalence was 39/10⁵. Only one case of multiple sclerosis was observed in >16 500 refugees from the northern part of the island living in study areas in the south. Trends favouring a higher prevalence in Paphos than Famagusta, or the mountains of Troodos and Kyperounda, were not significant. A tendency towards consanguinity may have determined the relatively high incidence of multiplex families, especially in one village (L.T. Middleton and Dean 1991). In a subsequent assessment, prevalence was higher in the Greek community (51/10⁵) and among Turkish Cypriots born in Cyprus (56/10⁵) than in those migrating from mainland Turkey (24/10⁵; Dean et al 1997). These differences were especially marked in men.

In southern continental Europe, Italy stands out as a region that has been studied in great epidemiological detail and, perhaps, consequently has noticeably higher rates than some of its neighbours. Between 1962 and 1982, approximately 30 epidemiological studies were reported from peninsular Italy and prevalence varied from 7 to 27/10⁵. Many of these surveys were based on denominators >300 000 and were carried out in both northern and southern Italy; but standardization to a single population was not usually attempted and demographic differences make the studies difficult to compare. The conclusion, expressed in earlier editions of this book, that Italians have a low risk of multiple sclerosis compared with other Europeans, was first questioned by the findings of Dean et al (1976), who showed equivalent admission rates to hospital in London for Italian and English patients. This seemed correct when Dean et al (1979) reported that, based on findings from Enna (Sicily), morbidity statistics for Italy might generally have been underestimated. Their work was followed by studies from many other parts of Italy that also upwardly adjusted the overall rates and systematically eroded the latitudinal gradient.

There are several reasons why the study of multiple sclerosis in Sardinians may be of special importance. Disease frequency has been studied repeatedly and determinedly so that contemporary figures for morbidity statistics are fully informative. Sardinia is one of only a few regions where a class II HLA antigen association exists that is specifically different from DR(15)2. The island has a colourful history (see Chapter 5), which could offer explanations for those aspects of its genetic epidemiology that are of general importance for understanding multiple sclerosis.

In Alghero, northern Sardinia, incidence was 4.1/10⁵/year for 1971–1980, with a prevalence of 59/10⁵ (Rosati et al 1987). The incidence for native-born Sardinians during 1965–1985 in Sassari (northern Sardinia) was 3.4/10⁵/year (95% CI 2.7–4.2), and serial estimations indicate that this changed from about 1.3–2.0/10⁵/year before 1977 to around 5/10⁵/year thereafter, with prevalence stabilizing at 69/10⁵ (95% CI 55–86). Surprisingly, no familial cases were identified at that time. Working to the new prevalence day of 1st December 1991, Rosati (1994) estimated prevalence at 103/10⁵ (95% CI 92–115) in Sassari with no difference between rural and urban rates (97 and 108/10⁵, respectively). The frequency of familial multiple sclerosis was now 13%. Incidence was 3.7/10⁵/year (95% CI 3.3–4.1) and quinquennial rates had steadily increased from 1.3–2.0/10⁵/year around 1962 to 5/10⁵/year, with the highest rate seen in the most recent 5-year period. The most recent updates for Sassari, northern Sardinia, show a steady rise in quinquennial rates for incidence from 2.0 to 6.8/10⁵/year between 1968–1972 and 1992–1997, with prevalence reaching 144/10⁵ (onset adjusted, 150/10⁵) on 31st December 1997 (Pugliatti et al 2001a). The rate of increase in prevalence has been steeper in Sassari than Ferrara, northern Italy, commensurate with differential increases in incidence, leading Pugliatti et al (2001a) to conclude that the Sardinian focus represents a real increase in disease frequency. Cluster analysis within this well-surveyed region, working to the same prevalence day, identified a hot spot for multiple sclerosis in individuals resident in southwestern communes between the ages of 5 and 15 years, with evidence for a west–east gradient (Pugliatti et al 2001b). This region has a distinct cultural history, being predominantly Logudorese by comparison with Catalan areas showing lower rates.

In Barbagia, central Sardinia, incidence for the years 1961–1980 was 2.9/10⁵/year (adjusted to 3.2 when standardized to the Italian population) based on a numerator of 31 cases. Prevalence in 1975 was 41/10⁵ (95% CI 25–62; standardized prevalence ratio 48.5), and this had risen to 65/10⁵ (CI 44–93; 78/10⁵ standardized to the Italian population) by 1981 (Granieri and Rosati 1982). Serial studies in Macomer suggest that multiple sclerosis was not diagnosed prior to 1952. Between 1952 and 1981, the overall incidence was 6.3/10⁵/year (95% CI 3.4–10.8), with a change from 2.8/10⁵/year for the period 1952–1956 to 10.1, 7, 4.2, 3.9 and 1.8/10⁵/year in sequential quinquennia up to 1981. Prevalence was 62/10⁵ (95% CI 20–143) in 1961, 73/10⁵ (95% CI 30–151) in 1971 and 72/10⁵ (95% CI 31–141) in 1981. The province of Nuoro, central Sardinia, had incidence rates for multiple sclerosis of 4.3/10⁵/year across the period 1955–1995, increasing from 1.95 to 6.6/10⁵/year in the quinquennium 1985–1989 but stabilizing thereafter. Prevalence increased from 103/10⁵ in 1985 to 144 and 157/10⁵ in 1993 and 31st December 1994, respectively (Cassetta et al 1998; Granieri et al 2000). In a detailed epidemiological survey from the most archaic parts of central Sardinia, the overall estimate for prevalence was 157/10⁵ with variations from 143 to 262/10⁵, the highest regions having an excess of familial cases (Montomoli et al 2002a).

In summary, Sardinia had an incidence of around 3.0/10⁵/year in the east, north, central and northwest regions in the 1980s (Granieri et al 1983; Rosati 1989; 1994; Rosati et al 1986; 1987; 1988) but estimates have since increased to 6.6/10⁵/year in some regions. Although these results have been interpreted as showing that multiple sclerosis was first introduced into the island immediately following the Second World War, with the rise of industrialization and population mixing, the facts are no less consistent with the rival theory that these represent epidemics of recognition rather than the impact of aetiological factors altering the risk of multiple sclerosis. Nevertheless, these figures show that, throughout Sardinia, the main rise in incidence occurred in the 1970s. Cases have all been restricted to those with Sardinian names whose grandparents were born on the island, although some prevalent cases were born elsewhere.

This substantial increase in frequency of multiple sclerosis is not exclusive to Sardinia. In the 1980s, prevalences in the republic of San Marino (Morganti et al 1984), the central Sicilian cities of Enna (53/10⁵) and Caltanissetta (51/10⁵; Savettieri et al 1986), and other parts of Sicily (Monreale 43/10⁵ and Agrigento 32/10⁵; Dean et al 1979, 1981b, Savettieri et al 1981) were strikingly higher than in neighbouring Malta (4/10⁵; Vassallo et al 1978). As for other parts of the world subjected to serial surveys, more recent estimates show a substantial rise from these figures derived in the mid-1980s. Incidence was 3.3/10⁵/year (95% CI 1.5–6.2) for Monreale over the period 1981–1991 and prevalence was 73/10⁵ (95% CI 44–113) on 31st December 1991 (Savettieri et al 1998). Incidence increased to 5.7/10⁵/year and prevalence reached 120/10⁵ by 1995 in Enna, Sicily (Grimaldi et al 2001). In the coastal city of Bagheria, the high Sicilian incidence was again demonstrated with an increase from 3.5 to 5.3/10⁵/year between 1985 and 1994 (Salemi et al 2000a). Incidence during 1974–1995 in Catania, Sicily, was 2.3/10⁵/year (5% CI 2.0–2.6) with increases from 1.3 to 3.9/10⁵/year across this period, and prevalence on 1st January 1995 of 59/10⁵ (1.2F:M; Nicoletti et al 2001). But significantly, the differential between Sicily and Malta persists. An update for Malta on 1st January 1999, based on 63 living cases, showed that annual incidence was 0.7/10⁵/year and prevalence 17/10⁵ (13/10⁵ for clinically definite disease; 1.4F:M). By contrast, prevalence in the small community of immigrants was 166/10⁵ (Dean et al 2002). The case for differences in disease susceptibility based on genetic origins – Maltese speak a Semitic language and have mainly north African Arabic ancestry following the invasion by Habasa in 869/870 ad, whereas Monreale and Enna have strong Norman (northern European) origins – seems strong. A recent genetic analysis of multiple sclerosis in Malta supports this interpretation (see Chapter 3).

An increase in incidence to 2.2/10⁵/year (95% CI 1.8–2.6) was also reported for Ferrara in northern Italy (Granieri and Tola 1994; Granieri et al 1985). Prevalence was 27/10⁵ in 1978 (Rosati et al 1981). Later, this was corrected to 37/10⁵ and updated in 1981 to 46/10⁵ (95% CI 40–53) based on the identification of 128 cases (Granieri et al 1985). The most recent figures are 2.3/10⁵/year (95% CI 2.0–2.6) and 69 (95% CI 61–79) for incidence and prevalence, respectively (Granieri et al 1996). Incidence and prevalence have been studied serially in Padova (northeast Italy) over 30 years. In 1970, the rates were 0.9/10⁵/year and 16/10⁵, respectively. Incidence increased from 2.2 to 3.9 to 4.2/10⁵/year over the periods 1980–1989, 1990–1994 and 1994–1999, respectively. Prevalence altered from 18 to 46 to 81/10⁵ (95% CI 70–91) in 1980, 1990 and 1999, respectively (Ranzato et al 2003). Here, the authors emphasize the rise in incidence during the 1980s with subsequent stabilization as an indicator that these changes in morbidity reflect the impact of improved diagnostic techniques and reduced latency from onset to diagnosis rather than a real increase in disease frequency. At 2.1/10⁵/year and 33/10⁵ the annual incidence and prevalence of multiple sclerosis in one culturally and genetically isolated Alpine region of northern Italy (Valle d′Aosta) were higher than previously reported but comparable to contemporary studies from neighbouring parts of Italy, with no obvious temporal trends between 1971 and 1985 (Sironi et al 1991). Solaro et al (2005) surveyed a population of 913 218 people living in the province of Genoa, northwest Italy, to identify 857 affected individuals, providing a rate for prevalence of 85/10⁵ adjusted to Italian standard population; crude (unadjusted) rates were 67/10⁵ (95% CI 60–76) in men and 118/10⁵ (95% CI 108–128) in women. L′Aquila district of central Italy had a prevalence of 56/10⁵ (95% CI 45–62) on 31st December 1996 standardized to the European population (1.9F:M; Totaro et al 2000). Standardized mortality rates for Italy between 1974 and 1993 were 4.1 and 5.0/10⁵ for males and females, respectively, with a north–south gradient, excluding Sardinia. Across this period, mortality had reduced in the north but increased in the south, presumably reflecting health care trends superimposed on increased ascertainment and, perhaps, incidence (Tassinari et al 2001).

Commenting on the overall picture of the epidemiology of multiple sclerosis in Italy, Rosati (1994) traced the history of Italian epidemiology in multiple sclerosis and lamented the difficulties faced by epidemiologists working within the health care system of the early postwar years – a situation that altered for the better from 1975. Publications on the epidemiology of multiple sclerosis dating from 1980 were considered under the headings of mainland and insular surveys. Prevalences, then ranging from 32 to 69/10⁵, had all increased over time, no longer showing latitudinal gradients or differences between continental Italy and Sicily. By contrast, the figures for Sardinia varied between 59 and 103/10⁵. A centralized survey of incidence (Comi et al 1989) based on data for 1971–1980 confirmed this distribution and showed rates, in the 1980s, of 1.1 and 1.9/10⁵/year for the mainland, or Sicily, compared with 4.2/10⁵/year in Sardinia. Others who have scrutinized the frequency of multiple sclerosis in Italy over time (Granieri et al 1993) also consider that, with the exception of Sardinia, the apparent gradient in frequency is mainly an artefact arising from differences in case ascertainment between the better and less developed parts of the country. The marked differences in population structure and failure to quote age-corrected figures have also contributed to the creation of a spurious gradient. Studies based on small denominators, so as to maximize case ascertainment, have since shown higher rates than previously claimed, and without regional variations.

With a certain amount of catching up still to occur, we cannot reach a final conclusion concerning the distribution of multiple sclerosis in southern Europe. The trends suggest that, with increased vigilance, estimates for prevalence will continue to rise, but not to the levels seen in northern Europeans. However, we anticipate that some island populations will retain disproportionately high (e.g. Sardinians) and low (e.g. Maltese) figures. Our position is that this reflects differences in genetic susceptibility. Whilst the facts seem clear, others may prefer alternative interpretations and explanations.

MULTIPLE SCLEROSIS IN THE MIDDLE EAST (Figure 2.21)

Figure 2.21.

Prevalence distribution of multiple sclerosis in the Middle East (Israel, Saudi Arabia, Kuwait, Iraq, Jordan and Oman). Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

There are many fewer epidemiological studies from outside Europe, Australia and North America. In serveral of these locations, our figures are taken from Kurtzke (1993), who himself made the calculations using assumptions about factors such as the place of origin and domicile of the index cases. Few of the figures are contemporary and none can be compared with other population-based surveys. Taken together, they give the impression of a low disease frequency in these places. Multiple sclerosis has been studied in Jordan (7/10⁵ based on 32 cases in Amman; Kurdi et al 1977) and Saudi Arabia (8/10⁵ by comparison between the case ratios for multiple sclerosis and amyotrophic lateral sclerosis; Yaquib and Daif 1988). Using the Schumacher criteria, Hamdi (1975) derived a figure of 3.4/10⁵ for Iraq by estimating the relative frequency of multiple sclerosis (11 cases) compared with motor neuron disease in a hospital-based sample for the years 1967–1969, relating this to a global figure for prevalence of the latter condition. However, this figure represented a threefold increase from an earlier estimate using similar methods (Shaby 1958). The majority of the cases were from northern Iraq where the population is predominantly Kurdish (Indo-European) by comparison with the middle and southern parts, inhabited more or less entirely by Arabs. It remains uncertain whether this pattern reflects genetic background or administrative arrangements for referral of cases to Baghdad in the 1960s.

By contrast, multiple sclerosis has been carefully surveyed in Kuwait. In the initial survey, prevalence was estimated at 8/10⁵ (Al-Din 1986). In 1988, the previously suspected difference in prevalence between Kuwaitis and Palestinians living in Kuwait was confirmed. Evidence was provided to support the hypothesis that the higher rate in peoples originating from Palestine is genetically determined, resulting from admixture of Caucasian genes as part of that region's turbulent social history. Prevalence was 10/10⁵; 186 of the 201 prevalent cases were Arabs and of these 72 were Palestinians (prevalence 24/10⁵) and 51 Kuwaitis (prevalence 9/10⁵; Al-Din et al 1990). By 2000, incidence had risen to 2.6/10⁵/year from 1.1/10⁵/year in 1993, the main increase being seen in women, with a change in prevalence from 7 to 15/10⁵; by comparison with earlier assessments, prevalence was higher in Kuwaitis (31/10⁵) than in non-Kuwaitis (6/10⁵: Alshubaili et al 2005).

Studies of the prevalence of multiple sclerosis in Israel have been particularly influential in developing ideas on the aetiology of multiple sclerosis (Alter et al 1962; 1978; Leibowitz et al 1970). Interpretation is made more complicated by the admixture in recent decades of peoples whose geographical origins were not necessarily in the Middle East. The prevalence for native born Israelis, age-adjusted to the 1960 United States population, was 13/10⁵ in 1965. Kahana et al (1994; see below for studies of migration) updated these figures to 1983. The incidence of multiple sclerosis was 1.4/10⁵/year and prevalence 32/10⁵ for all Israelis. Higher rates were reported for Jerusalem (2.4/10⁵/year and 61/10⁵, respectively). Karni et al (2003) have compared the frequency of multiple sclerosis in Jewish and Arab populations of greater Jerusalem. As expected, prevalence had increased in Israelis by December 1995 to 64/10⁵ in European/American Jews, and 52/10⁵ in African/Asian Jews, compared with earlier estimates. The rates for incidence and prevalence in Arabs were again lower (0.7/10⁵/year and 19/10⁵, respectively) but these observations need to be taken in the context of differential rates amongst Israelis depending on place of birth and African/Asian or European/American Jewish ancestry (see below). Most recently, Tharakan et al (2005) collected cases of multiple sclerosis in Oman, which lies in the eastern Arabian peninsula at 10–30°N and has an estimated population of 1.5 million. Health care is sophisticated and the 34 ethnic Omanis in whom the diagnosis of multiple sclerosis was made between 1990 and 2000 represent a moderately complete sample allowing a minimum estimate of period prevalence at 3/10⁵.

MULTIPLE SCLEROSIS IN AFRICA (Figure 2.22)

Figure 2.22.

Prevalence distribution of multiple sclerosis in Africa. Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature. For interpretation of South Africa see text and Figure 2.5.

The frequency of multiple sclerosis in the African continent has, until recently, been studied in little detail, since it was apparent that this is an uncommon disorder in native peoples from these regions. The continent is populated by African blacks throughout – but especially in the west, central and southern parts – with a significant admixture of Arab ancestry in the north and east, where Asian people are also settled, and white immigrants both to the offshore islands and mainland. It is because of the rarity of the disease amongst African blacks compared with the higher rates seen in immigrants that the study of multiple sclerosis on the African continent, and in African blacks elsewhere, has proved so influential. Despite the overall paucity of the disease, there are some reports suggesting a higher than expected frequency of multiple sclerosis in coastal regions of the African continent, in addition to some offshore islands, where social history has perhaps encouraged more genetic admixture than in continental Africa.

Multiple sclerosis has an appreciable prevalence in Libya (6/10⁵ based on 21 patients in Benghazi, age-adjusted to the population of the former West Germany; Radhakrishnan et al 1985) and Tunis (10/10⁵ based on 73 definite or probable cases; Ben Hamida 1977). Dean (1967) provided morbidity statistics for multiple sclerosis in white South African-born individuals and immigrants to South Africa but did not identify any affected African blacks. Multiple sclerosis was subsequently reported both in mixed race South Africans (Ames and Louw 1977) and in blacks (Bhigjee 1987). In a subsequent survey from Cape Town, the prevalence of multiple sclerosis was 3/10⁵ in pigmented people (B. Kies 1989; Brian Kies, personal communication). Case reports were published, and anecdotes went into circulation, of multiple sclerosis occurring inter alia in individuals from Zimbabwe, Kenya, Uganda, Cameroon, Ethiopia and Senegal (Adam 1989; Foster and Harries 1970; Goldstein 1946; Haimanot 1985; Kanyerezi et al 1980; Lisk 1991; Mbonda et al 1990). In some of these cases the distinction from acute disseminated encephalomyelitis and the exclusion of retroviral infection was assumed rather than proven, and there are uncertainties relating to ethnicity. But with increased world travel, neurologists routinely exposed to cases of multiple sclerosis have encountered patients in Africa meeting diagnostic criteria for the disease. It has been our own experience, mainly based on cases seen at Baragwanath Chris Hani Hospital, Johannesburg-Soweto (through the kindness of Dr David Saffer) that the clinical features are distinct, matching the relapsing Devic's disease phenotype that is rare in Caucasians (see Chapter 5).

It is against this background that Dean et al (1994) collated all reports of multiple sclerosis occurring in black South Africans and Zimbabweans. Of the 12 patients, six had disproportionate involvement of the visual system and the frequency of a progressive course from onset was higher than expected in a Western series of cases. A number were lost to follow-up and histological confirmation was not obtained in those who died. Several had evidence for previous, but now inactive, treponemal infection. Nevertheless, the laboratory abnormalities usually associated with multiple sclerosis were present and many had signs rarely having other explanations in Caucasians with neurological disease. Some, especially the cases from Zimbabwe, had a typical history of relapsing neurological symptoms with signs indicating multifocal involvement in episodes. Although not based on an epidemiological study, Kioy (2001) described nine individuals with clinically definite multiple sclerosis amongst 2831 referrals to a private electrodiagnostic clinic (3.5F:M). Two were Asian and seven African Bantus, leading the author to conclude that multiple sclerosis is becoming more frequent in native Africans.

Multiple sclerosis is diagnosed in African Caribbeans (Cruikshank 1973; Cruikshank et al 1961). With no cases reported until the 1970s, C. Poser and Vernant (1993) described 11 cases in blacks from Martinique and Guadeloupe who had no evidence for infection by human T-cell leukaemia virus (HTLV)-1. More recently, Cabre et al (2001) identified clinically definite multiple sclerosis in 51 individuals of mixed French-African ancestry living in Martinique, providing a prevalence of 14/10⁵ (95% CI 10–18). The clinical features were broadly equivalent to multiple sclerosis as seen in western Europe but with a lower frequency of oligoclonal bands and higher rate of the Devic phenotype (25%). Leaving those with mixed ancestry, it would be churlish to reject all these cases affecting Africans on the basis of their clinical details. The case for multiple sclerosis, defined by an appropriate clinical and laboratory phenotype, occasionally occurring in African blacks seems good. (We offer an analysis of the origins of multiple sclerosis that leans heavily on the epidemiology and phenotype seen in continental Africa, Asia and the Orient in Chapter 5.)

MULTIPLE SCLEROSIS IN ASIA AND THE FAR EAST (Figure 2.23)

Figure 2.23.

Prevalence distribution of multiple sclerosis in India and the Far East (Malaysia, Thailand, China and Japan). Most recent figures are given for prevalence/10⁵ population; a ‘best guess’ is given where local variations exist in the published literature.

There have been several surveys of multiple sclerosis in India and these show differences in risk depending on ethnicity for individuals living in the same geographical region. The prevalence of multiple sclerosis in Parsees from Bombay and Poona was 26/10⁵ (95% CI 13–40; age adjusted to 24/10⁵) and 58/10⁵ based on 14 and two prevalent cases, respectively, in 1988 (Wadia and Bhatia 1990). These rates are higher than those previously quoted for Indians in Bombay (1/10⁵; Singhal and Wadia 1975) but comparable to the earlier estimation of prevalence in Parsees from a door to door survey carried out in Bombay (Bharacha et al 1988). Multiple sclerosis has also been reported from southern India (Gourie-Devi and Nagaraja 1982). Reviewing the experience of several decades, Jain and Maheshwari (1985) comment on the greater frequency of multiple sclerosis in northern India, populated by Indo-Europeans, compared with the south, which has a higher density of Tamils and Dravidians. Multiple sclerosis in Chinese and Japanese people is evidently a rare disease with a markedly different phenotype from that seen in the Western world. However, this may be changing. As with the studies in Africans, an additional point of interest lies in the study of the disease amongst Orientals not living in China or Japan. But unlike Africans, who seem to show a changing frequency with relocation to other environments, the evidence suggests that Orientals are not placed at a higher risk of multiple sclerosis through exposure to a novel environment. Y.L. Yu et al (1989) studied multiple sclerosis amongst Chinese in Hong Kong and reported a prevalence of 0.9/10⁵ – with a similar clinical phenotype to other affected Orientals. Hung (1982) reported no discernible change in frequency of the disease in Taipei, northern Taiwan, during the 1970s. Prevalence rates for 1975 and 1980 were 0.8 and 0.9/10⁵ in a population of 1.9–2.2 million, respectively. In contrast to other parts of the world, there had been very little change when this region was again surveyed in 2004 (1.9/10⁵: Tsai et al 2004). Hou and Zhang (1992) carried out a door to door survey in the Yunnan Province of mainland China using an 11% sample of the population (just under 0.5 million), identifying only one patient and providing a prevalence estimate of 1.4/10⁵ (95% CI 0–8, age-adjusted to the United States population).

In Japan, neuromyelitis optica was once the common clinical picture but the so-called Western phenotype has increasingly emerged as the expected disease appearance in affected individuals born after the 1960s. Prevalence for all clinical forms was first estimated on the basis of 65 cases identified in ten cities, ranging from latitudes 26 to 44°N, at between 0.9 and 4/10⁵ with a mean of 2.1/10⁵ (Kuroiwa et al 1983; see also C.M. Poser 1994). Itoh et al (2003) updated an earlier prevalence estimate for Asahikawa, northern Japan, from 2.5/10⁵ in 1975 to 10.2/10⁵ in 2002. As in other series, there had been a change over the 27 years from an optico-spinal to more typically Western phenotype with only 3% of the more recent prevalent population presenting with optic nerve involvement. Houzen et al (2003) reported a prevalence of 8.6/10⁵ (2.9F:M) on 31st March 2001 in an area of northern Japan having a stable population at risk and showing very little inward migration or expansion of the denominator, based on postal questionnaire of clinics serving the population. There was a low frequency (16%) of the Devic (optico-spinal) phenotype in the 31 reported cases. An early estimate for the prevalence of multiple sclerosis in Malaysia was around 2/10⁵ (C-T. Tan 1988). Histological confirmation of the diagnosis is claimed for cases from Thailand (Vejjajiva 1982).

Informal conversations suggest that these figures for disease frequency are underestimates. Many neurologists working in Africa, Asia, the Middle East and the Orient have priorities other than documenting numbers of patients with neurological diseases that are rare in their communities. Conferences focusing on the global distribution of multiple sclerosis periodically present an opportunity for bringing these numbers up to date. No such meeting has been held for several years and is overdue.

MULTIPLE SCLEROSIS IN MIGRANTS (Figure 2.24, Figure 2.25, Figure 2.26)

Figure 2.24.

Routes taken by migrants proving informative with respect to changes in the frequency of multiple sclerosis with movement from high- to low-risk zones.

Figure 2.25.

Routes taken by migrants proving informative with respect to changes in the frequency of multiple sclerosis with movement from low- to high-risk zones.

Figure 2.26.

(A) Age-specific incidence (/10⁵/year) and (B) prevalence (/10⁵) of multiple sclerosis in Israel (1st January 1981). = native-born Israelis whose fathers were born in Europe or America; = native-born Israelis whose fathers were born in Africa or Asia; = immigrants to Israel born in Africa or Asia. The y axis is a logarithmic scale.

Adapted from Kahana et al (1994). ©1994, reproduced with permission of Springer-Verlag GmbH.

© 2006 Springer-Verlag GmbH

We have already developed the argument that social and historical events that led people of European, African, Asian, Oriental and Aboriginal stock to mix, move or remain in isolated communities as genetic relics shaped the global distribution of multiple sclerosis. However, especially over the last two centuries, founder populations have been neither geographically nor socially stable and there have been many migrations, involving relatively large numbers of people, which appear also to have influenced the distribution of multiple sclerosis. Migration between high- and low-risk areas has occurred in both directions and involving individuals of all ages. This led to the formal comparison of disease frequency between racial groups living in the same geographical region and, conversely, cohorts of individuals with the same ethnic origins living in different parts of the world. Overall, migration studies emphasize multiple sclerosis as an exogenous disorder, acquired some years before clinical expression and probably in childhood, whereas studies of indigenous peoples provide more compelling evidence for genetic effects on disease frequency and distribution.

One of the earliest and most influential studies of migrants emerged from South Africa in the decades following the Second World War. Arriving by chance in Cape Town, Geoffrey Dean was struck by the markedly different frequencies of the disease in people of African origin and those moving to South Africa during the 20th century (see Chapter 1). Dean (1967) provided annual incidence, prevalence and mortality statistics for multiple sclerosis in white South African-born individuals and immigrants to South Africa. The age-corrected frequency of the disease was highest in immigrants from Europe, lowest in Afrikaaners and intermediate in South African English, both with respect to incidence and prevalence (Kurtzke et al 1970b). The absence of multiple sclerosis in African blacks was confirmed but a slightly higher rate was seen in the mixed race population, in whom African and Caucasian genes are shared. However, within the English-speaking white group, there was a marked difference in frequency of multiple sclerosis depending on age at arrival in South Africa. Those moving as adults to South Africa from the areas of northern Europe where multiple sclerosis is common took with them the high frequency of the country of origin, whereas those migrating ≤15 years showed the lower rates characteristic of native-born South Africans. For the 114 northern European immigrants with multiple sclerosis resident in South Africa by 1960, the main risk factor appeared to be migration at ≥15 years (Dean and Kurtzke 1971). Age-adjusted prevalence for persons aged 15–19 years at immigration was 66/10⁵ for those migrating from the United Kingdom, compared with an overall figure for northern Europeans of 51/10⁵, whereas prevalence was 13/10⁵ for those migrating at ≤14 years.

Whilst few would now feel confident about confining risk to a particular calendar age, these studies proved enormously influential in generating concepts on the aetiology of multiple sclerosis. In an earlier edition of this book, we emphasized the danger of developing elaborate hypotheses based on a small numerator and with indirect methods of case ascertainment. The original South African studies depended on only six and 12 identified cases, respectively, in the two most informative groups. Inaccurate ascertainment or errors of diagnosis affecting a single individual would have significantly altered the conclusions of this study, and at least a 10% error rate for classification is to be expected even in areas surveyed using the most direct methods. That said, the principles first established in studies amongst migrants to South Africa are supported by a number of other surveys. But there are exceptions.

The effect on risk of multiple sclerosis for individuals of Dutch origin migrating to a low-risk region has been reported for the Antilles, where prevalence for native-born persons was 3/10⁵ compared with 59/10⁵ for Dutch-born migrants (Moffie 1966). Migration of Europeans to Australia and New Zealand has taken place in several phases over the last 150 years. Superficially, this should have provided another opportunity to examine the effect of migration and the ages over which the effects occur. Unlike the African continent, there is no reference rate for prevalence amongst indigenous peoples against which to compare disease frequency in immigrants. Aboriginal people in Australia and New Zealand were rapidly corralled by white settlers, or otherwise disadvantaged. That unglorious history is not our business, but it has reduced the epidemiological opportunity for comparison of disease frequency in diverse ethnic groups occupying the same geographical location. Amongst immigrants to Australia and New Zealand, people originating from groups with a disproportionately high or low risk of multiple sclerosis may, by chance, have settled in particular areas and remained separated from other settlers, thus creating local pockets of susceptibility and resistance. Tasmania stands out as an area of exceptional disease frequency. As a former penal colony, it was initially populated from 1803 by 24 000 individuals, 19 000 of whom were convicts. The present population of 500 000 is descended from the 10 000 founding couples providing a population retaining some genetic isolation. But even here, the prevalence for multiple sclerosis is half that currently ascertained for people of equivalent ethnicity in northern Europe, strongly suggesting that environmental factors have influenced expression of the disease in susceptible individuals. More generally, and in contrast to the findings from South Africa, S.R. Hammond et al (2000a) showed that the risk of developing multiple sclerosis was uninfluenced by age at migration – taking 15 years as the point of reference.

Considering migration in the reverse (low to high) direction, no less provocative than the surveys carried out in South Africa has been the study of United Kingdom-born children of immigrants from the Indian subcontinent, Africa and the West Indies (Elian and Dean 1987; Elian et al 1990). In claiming that the prevalence of multiple sclerosis in these children approximates to that seen in similar age groups amongst indigenous people, the authors provide powerful ammunition for the environmental doctrine. There are several reasons why the ascertainment of patients prevalent on 1st April 1986, aged >15 years and born in the United Kingdom of parents who were migrants from the West Indies, Africa or Asia, was probably an underestimate. That said, the possibility arises that in accepting the documented diagnosis and not reviewing the evidence, individuals with other diseases may have been included. There would be fewer anxieties about diagnoses in these cases if every patient had been tested for HTLV-I status. Whilst both age at presentation and clinical severity were not typical of multiple sclerosis in white English populations, the authors acknowledge that only the most severe and early onset cases from amongst the cohort at risk were identified in this initial screen, suggesting that the frequency of multiple sclerosis in this group will rise with time. Deaths and significant demographic shifts may have occurred in the immigrant population of London and the West Midlands between the 1971 census and prevalence day in 1986, making the denominator applied to index cases inappropriate. In this respect, it is only partially reassuring that all but one of the 28 West Indian patients was known not to have moved since 1971.

In deriving the denominator, individuals at risk of European ancestry returning from Africa in the 1950s or 1960s were not distinguished from African black immigrants. Census information merely recorded that both parents had been domiciled in one of the New Commonwealth countries. Whereas white repatriates will have spuriously increased the denominator, their affected offspring did not feature in the numerator. However, the effect of these biases was probably small and would have led to the risk of multiple sclerosis in the first-generation children of West Indian, African and Asian immigrants being underestimated. Marta Elian and Geoffrey Dean referred their identification of cases amongst the offspring of New Commonwealth immigrants to age- and sex-specific rates from the 1985 Sutton (Surrey) prevalence study (E.S. Williams and McKeran 1986) and assumed an equal risk for immigrants in order to calculate the number of children expected to develop multiple sclerosis. Much depends on the validity of these expected case numbers. The Sutton study was carried out close to a large urban metropolis where accurate definition of numerator and denominator are more difficult to establish than in population-based surveys and where non-white cases were not excluded. There are difficulties in accepting evidence for an increased risk of multiple sclerosis in the children of immigrants when contemporary figures are not available for the parental generation based on comparable methods of assessment and ascertainment. All the rates cited by Elian and Dean (1987; Elian et al 1990) necessarily depend on small numbers and are therefore subject to large confidence intervals. A few errors will have made a large impact on the quoted rates and their interpretation. The authors made little of the greater risk seen for the children of Asian populations by comparison with black populations, although this difference also reflects global trends in the distribution of multiple sclerosis.

To assess the effect of environmental factors in changing the risk of disease for the offspring of non-Caucasian immigrants to the United Kingdom, Dean and Elian (1997) subsequently reported on 76 ethnic Asian individuals with a clinical diagnosis of multiple sclerosis (58 Indians, 17 Pakistanis and 1 Bangladeshi) who were born in the Indian subcontinent, east Africa, Fiji or Malaysia. They showed a higher than expected number of people with multiple sclerosis arriving in the United Kingdom aged <15 years, in each quinquennium under consideration, and this difference was especially marked for females. Conversely, there was no difference in observed and expected age of immigration amongst 60 patients with multiple sclerosis born in the Caribbean.

The danger of extrapolating from studies involving a small numerator in an unusual environment is well illustrated by the study of multiple sclerosis amongst immigrants from Vietnam to Paris, France. Three cases were identified in a cohort of around 3400 persons born of Vietnamese mothers who came to France aged <20 years. However, an essential criterion for immigration was mixed parentage with a French-born father (Kurtzke and Bui 1980). The fact that the cumulative 18-year risk of multiple sclerosis, by 1975, was 89/10⁵ (95% CI 18–260), with an age-specific prevalence of 169/10⁵ (95% CI 94–135) in the third decade, tells us nothing about the shift in risk of multiple sclerosis consequent upon movement of peoples from southern Asia to northern France, because admixture of genes was required in order to establish eligibility for migration. Presumably the majority of Vietnamese not migrating to France did not have French fathers.

As part of a national survey of multiple sclerosis in France, with cases recruited through an appeal on television, Delasnerie-Laupretre and Alperovitch (1992) identified – amongst 8000 cases ascertained overall in France − 246 individuals who had migrated from north Africa in the first quinquennium of the 1960s following the Algerian war for independence. Excluding the 27 patients who had multiple sclerosis before, or at the time of migration, 86% of these 246 probands were European in origin and the remainder Arab or Berber. There was no apparent age- or sex-adjusted difference in frequency or mean age at onset between these individuals and native French cases. Delasnerie-Laupretre and Alperovitch (1992) interpreted this first analysis of north African migration to France as indicating that the provocative exogenous factors are ubiquitous and multiple sclerosis is acquired by the same age irrespective of location. Kurtzke et al (1998) subsequently reported on 260 individuals – many presumably also included in the earlier survey – amongst 7507 respondents to a nationwide survey of multiple sclerosis who had immigrated from Algeria, Morocco or Tunisia between 1923 and 1986 (the majority during 1956–1964, and two-thirds from Algeria). Compared with French-born patients, those with multiple sclerosis migrating from north Africa were younger and with an earlier year of onset. Assumptions were made in concluding that prevalence was higher in the immigrants from north Africa developing multiple sclerosis after arrival in France (77/10⁵; 95% CI 67–87) than those who migrated having already manifested the disease (17/10⁵; 95% CI 11–24), or native-born French cases (estimated at 50/10⁵). Cabre et al (2005) studied the population of Martinique and Guadeloupe, which includes large numbers of individuals who have returned to the West Indies after living for several years in France. Against a crude background prevalence of 15/10⁵ (95% CI 12–18; with higher rates in Martinique than Guadeloupe) and incidence of 1.4/10⁵/year (95% CI 1.0–1.8) on 31 December 1999, there were 33 observed cases amongst migrants whereas 19 were expected (standardized incidence ratio 1.7 (95% CI 1.2–2.4). Incidence was even higher for individuals living in France at <15 years of age (4.1: 95% CI 2.2–6.8). Since the genetic background of the permanent residents and those who spent time in France is similar, these differences are interpreted as reflecting the influence of different environmental conditions in the West Indies and in France.

The studies of multiple sclerosis in United States Army veterans recruited during and following the Second World War, and still under observation by John Kurtzke (see above), provide essential information on which to judge issues surrounding a change in risk of multiple sclerosis with migration between regions of differing prevalence. Kurtzke et al (1971; 1985) compared mortality amongst northern-born United States citizens dying in the south, and vice versa. The aim was to compare the frequency of multiple sclerosis in individuals who entered military service from zones differing from those in which they were born. The evidence for a north–south gradient in frequency depended on a mortality rate for southern-born patients dying with multiple sclerosis in the north of 0.7/10⁵/year compared with 0.5/10⁵/year for those remaining in the south. The mortality ratio for veterans born in the high-frequency zones (northern states) and entering service from the middle band dropped from 1.5 to 1.3, compared with 0.7 for those entering middle states but born in the southern states having lower frequencies of multiple sclerosis. Those born in the medium-risk band showed a ratio increase to 1.4 if entering military service in the northern states, and a reduced ratio of 0.7 if relocated before enlistment from the south. However, those born in the south and migrating to either the middle or northern zones showed rates of 0.6 and 0.7, respectively. Taken together, the difference between observed and expected rates showed a significant effect of migration for white males serving during the Second World War that was not apparent in black males or white females (in whom fewer migrations occurred). The gradient in disease frequency associated with state of residence at entry into active service was associated, inversely, with a younger age at onset for the states having higher disease frequencies (Kurtzke et al 1992). By contrast, Detels et al (1972; 1977) did not report increased rates for mortality or prevalence for southern-born migrants to the northwestern United States (Washington). Kurtzke (1993) has interpreted the apparent high prevalence of multiple sclerosis in Los Alamos County, New Mexico (76/10⁵; CI 42–108), as reflecting the immigrant population of non-Hispanic whites in that community because of military activities during and after the Second World War (R.E. Hoffman et al 1981).

Against these transcontinental shifts, movement within one relatively small country has previously been associated with alterations in mortality for multiple sclerosis. A reduction was reported in Norway for those moving from high- to low-risk regions with a corresponding increase after migration between low- and high-frequency zones – each compared with individuals having a stable domicile (Westlund 1970). Another informative series of epidemiological studies compared the prevalence of multiple sclerosis amongst Japanese living in Hawaii with inhabitants from the west coast of North America and native Hawaiians. In Hawaii, the prevalence amongst Japanese was 7/10⁵ compared with 11/10⁵ in Caucasian Hawaiians and 34/10⁵ in Caucasian immigrants to Hawaii (Alter et al 1971; see also Lauer 1994). These rates were virtually identical for Japanese and Caucasians living in California (7 and 30/10⁵, respectively; Detels et al 1977) and can be compared with the expected rate of 2/10⁵ for native Japanese (Kuroiwa et al 1983). Here, the evidence favours a strong protective effect for Japanese irrespective of environment but with some modification of risk on relocation to an area associated with higher rates of multiple sclerosis for the indigenous population.

Israelis are another group in which migration has occurred on a sufficient scale to show important epidemiological principles relating to the risk of developing multiple sclerosis. The original study (Alter et al 1962; 1978) in Israeli immigrants showed a difference in prevalence between those migrating from northern Europe (Ashkenazi) and from Asia and Africa (Sephardis). The higher frequency in Ashkenazi than Sephardic Jews also revealed an age at migration effect, in that there were very few affected Ashkenazis in the cohort migrating to Israel before adolescence. Although crude rates retained the difference seen in the parental groups, prevalence in the Israeli-born children of Ashkenazis and Sephardic Jews was the same after age adjustment to the population of the United States. Kahana et al (1994) subsequently compared native Israelis whose fathers were born in Israel, Europe or North America, and Africa or Asia, with immigrants (Figure 2.26). Depending on place of paternal birth, prevalence (age-adjusted to the Israeli population of 1960) was estimated at 32/10⁵ (fathers born in Israel), 38/10⁵ (Europe or North America) and 29/10⁵ (Africa or Asia) compared with 14/10⁵ in immigrants. Higher rates were observed in Jerusalem (61, 68 and 51/10⁵, for these three categories, respectively) than in other parts of the country. In the most recent update of these important Israeli studies, Karni et al (2003) reported a lower frequency of disease in Jewish immigrants from Africa/Asia (22/10⁵) than in native-born African/Asian Israelis (52/10⁵). Rates were similar in native-born and immigrant European/American Israelis (64/10⁵ in each, respectively). The implication is that, at least for Ashkenazi and Sephardic Jews, racially determined differences in risk for multiple sclerosis are modified by environment.

Having established from these studies of migration the probability that environmental factors modify inherent risks of multiple sclerosis dependent on ethnicity, it becomes important to establish at what age these influences occur. The studies from South Africa provide a starting point for opening the debate and suggest 16 years as the critical cut-off. The United States veterans’ survey also demonstrates that the north–south gradient for place at birth or entry into military service is no longer apparent for place of residence at clinical onset, indicating that acquisition of the disease has occurred earlier, and in all probability during childhood (Beebe et al 1967). The study of New Commonwealth immigrants to the United Kingdom is also consistent with an effect of exposure at ≤15 years (Dean and Elian 1997). Studies from Norway have shown maximal clustering between the ages of 13 and 20 years, suggesting that events in adolescence correlate with the development of multiple sclerosis in general and age of onset in particular (Riise and Klauber 1992; Riise et al 1991).

Kurtzke (1993) considered that matching by age in the study of north African migration to France (Delasnerie-Laupretre and Alperovitch 1992) had introduced a confounding factor since this restricted the study to individuals with the same age at onset. Subsequent observations on altered risks for north Africans in Algeria, Morocco or Tunisia or after arrival in France led Kurztke et al (1998) to conclude that multiple sclerosis occurs latently in a small but susceptible proportion of individuals having prolonged exposure to an environmental agent sometime between the ages of 10 and 40 years. Although prevalence rates are lower in Australians than in people of comparable ethnicity in northern Europe (see above), the recent analysis of age range for risk of developing multiple sclerosis in Australians showed (for all ages) that prevalence is similar in Australian-born and United Kingdom-born individuals for all locations except Hobart, Tasmania. Here the prevalence of multiple sclerosis was 68/10⁵ in Australian-born compared with 122/10⁵ in United Kingdom-born patients. But taken as a whole, there was no significant difference in the frequency of multiple sclerosis in the 331 immigrants from the United Kingdom, aged 29–41 years at prevalence in 1981, who developed multiple sclerosis depending on whether they arrived before (30%) or after (70%) the age of 15 years (S.R. Hammond et al 2000a). The key observation that individuals moving as adults to Australia and subsequently developing multiple sclerosis do not have a higher disease frequency than native-born Australians or juvenile immigrants suggests that the risk of exposure spans a much wider age interval than originally proposed from the South African studies.

Clearly, an interval of several years exists between exposure of individuals at risk and the development of clinical symptoms. Typically the environmental event(s) occur in childhood and onset is as a young adult. This situation epitomizes the interplay between genes and the environment in multiple sclerosis. However, exposure may be delayed and the latency prior to development of clinical manifestations can be short or long, making for a wide spectrum both in the age at which susceptibility converts to disease and in the year of onset.

EPIDEMICS AND CLUSTERS OF MULTIPLE SCLEROSIS

Protagonists of the environmental doctrine for causation are naturally enthusiastic about evidence for multiple sclerosis occurring in epidemics. Claims have been made for clusters in the Faroe Islands, Iceland, Orkney and Shetland, and Key West, apart from the more restricted groupings that occasionally feature in regional epidemiological studies (see above).

The most comprehensive survey of multiple sclerosis conducted by John Kurtzke over several decades forms the basis for the claim that epidemics have occurred in islands located in the North Atlantic. The arguments for a point source epidemic in these parts, especially the Faroe Islands, undoubtedly have merit although the evidence is not universally accepted. Others take the view that these are epidemics of recognition, reflecting the arrival of specialist medical services in the islands, rather than a real change in incidence arising from the introduction of transmissible aetiological factors into virgin populations.

In the first survey of Iceland, 168 cases of multiple sclerosis were identified with onset between 1900 and 1975 (Kurtzke et al 1982). Annual incidence rates appeared to rise in about 1922, and then stabilize until a further increase occurred in 1945, heralding a steady decline from the mid-1950s. Average annual incidence rates for 1923–1994 were around 1.6/10⁵/year increasing to 3.2/10⁵/year between 1945 and 1954 and falling to 1.9/10⁵/year between 1955 and 1974. Each quinquennial rate from 1900 was lower than that for 1945–1954, during which age at onset was also younger than before or after this period. These observations led S.D. Cook et al (1980) and Kurtzke et al (1982) to conclude that there had been a postwar epidemic of multiple sclerosis in Iceland, but opinions differ both with respect to the facts and their interpretation. Subsequently, Benedikz et al (1994), for the purposes of assessing the question of an epidemic in Iceland, considered 323 patients with onset of symptoms attributed to multiple sclerosis after 1st January 1900, of whom 252 were still living in December 1989. All but four were born and raised in Iceland and had maternal and paternal families going back several generations. Incidence rates were generally <1/10⁵ up until the 1930s but then increased to 2.5/10⁵, coinciding with the arrival of two neurologists. With waning enthusiasm, so the analysis goes, there was then a lull until 1945–1955, when incidence increased to 3.3/10⁵/year following the first systematic survey of the disease. With nine neurologists in practice from 1975, incidence peaked at 4.1/10⁵/year, and this rate has since been maintained. Prevalence in Iceland, based on cases ascertained by year of diagnosis, was 33 (95% CI 25–43), 33 (95% CI 26–43), 52 (95% CI 43–63) and 70/10⁵ (95% CI 60–82) in 1955, 1965, 1975 and 1985, respectively. Onset adjustment changed these rates for 1950 to 1990, by decade, to 50, 52, 62, 85 and 100/10⁵, respectively. The most recent figure for prevalence is 119/10⁵, the slower rate of increase (despite the luxury now of 15 neurologists) suggesting that ascertainment has saturated and steady state is now reached (Benedikz et al 2002).

It is significant that John Benedikz, a neurologist who has worked in Iceland for many years and takes a particular interest in the epidemiology of multiple sclerosis, has questioned whether there has been a genuine increase in incidence. He favours the view that any change in frequency of multiple sclerosis in Iceland during the 20th century reflects improved recognition and the development of more sophisticated diagnostic procedures (Benedikz et al 1991; 1994; C.M. Poser et al 1992). In the era before 1950, case ascertainment was far from adequate, as shown by the long interval between year of onset and diagnosis at that time. Comparisons of disability in the affected cohorts showed an excess of those with severe multiple sclerosis before 1950. The arrival of neurologists in Iceland (two of whom, Kjartan and Gunnar Guomundsson, were prime movers in identifying patients with multiple sclerosis) may have been entirely responsible for the increased number of cases through improved vigilance and recognition. The abrupt reduction in interval between onset of symptoms and diagnosis after 1940 is further evidence for the impact of improved neurological services on case ascertainment.

The 18 Faroe Islands lie in the North Atlantic at latitude 62°N. The population increased from around 15 000 in 1900 to 48 000 in 1990. The islands have had a strong administrative association with Denmark, which has loosened somewhat since 1948. Observations on multiple sclerosis were made originally by R.S. Allison (1963) and Fog and Hyllested (1966). They found fewer cases than expected from comparisons with neighbouring Orkney and Shetland (see Chapter 1). A national multiple sclerosis register was established in 1947 against the background of a long-established health care system, including routine transfer of cases to the Neurology Department of the Rijkshospital in Copenhagen. These and other sources were repeatedly searched for diagnoses of multiple sclerosis by John Kurtzke in his zealous ascertainment of multiple sclerosis in the Faroes, attempting to trace cases from before the Second World War. The Schumacher criteria were used to assign the diagnosis of multiple sclerosis, supplemented by laboratory investigations when these became available. By 1986, 41 cases had been ascertained of whom nine lived abroad for 3 or more years. These were not considered directly to have been part of the epidemic. Neither in the initial survey by Fog and Hyllested (1966), nor in the later assessments by John Kurtzke and colleagues (Kurtzke 1993; 2005; Kurtzke and Hyllested 1979; 1986; 1987; 1988) was any patient identified with an estimated date of onset earlier than 1943. In the initial analysis, there were 16 cases with onset during 1943–1949, and a further 16 developing clinical manifestations during 1950–1973 but none thereafter despite a steady increase in the population. These observations led John Kurtzke and Kay Hyllested to conclude that a 30-year epidemic of multiple sclerosis occurred on the Faroe Islands following the Second World War. The details of this claim were argued in minute detail with comprehensive summaries of the evidence by Kurtzke (1993).

Starting from the premise that multiple sclerosis is acquired at around the time of puberty, John Kurtzke and Kay Hyllested first separated incident cases into those who were pre- and post-pubertal in 1943. This suggested that three epidemics occurred between 1943 and 1973. A fourth was predicted (at approximately a 13-year interval) and was later claimed (Figure 2.27 : Kurtzke et al 1995). The first was dominated by individuals who were postpubertal in 1943. Conversely, peaks in incidence constituting subsequent epidemics were made up of individuals who were prepubertal at that time. These cases also had a younger age at onset of the disease. John Kurtzke concluded that the critical factor determining the Faroes’ experience of multiple sclerosis was occupation by British troops between 1940 and 1945. He proposed that the distribution of multiple sclerosis showed both a temporal and spatial relationship. Villages where individuals lived who contributed to each of the incidence peaks (especially those occurring in the late 1950s and 1960s) within the overall epidemic were also those where troops were billeted. This observation led John Kurtzke to conclude without reservation that the cause of multiple sclerosis (in the Faroes) is a transmissible infection not producing neurological symptoms in the majority of carriers. He argued that the pool of individuals at risk who might have acquired the factor in the early 1940s was that 75% of the population having close contact with British troops. They then represented the source from which others were later affected to create the second phase of the epidemic, and so on. Eventually, the number of susceptible individuals was no longer sufficient to sustain the disease so that the epidemic terminated in the late 1970s. For each case, a minimum exposure of 2 years was needed to acquire the disease.

Figure 2.27.

(A) Annual incidence (/10⁵) for clinical multiple sclerosis in native resident Faroese, calculated as 3 year centred moving averages, to 1998. Upper panel, total series; lower panel, rates for each of the four epidemics. (B) The Faroe Islands from a local picture postcard (c.mid-1990s).

Adapted from Kurtzke and Heltberg (2001).

© 2006

As a corollary to this argument, it is necessary to suppose that the potency of transmissibility diminished with time. Affected individuals acquired the agent at around the age of 11 years and infected others between, but not after, the ages of 20–25 years (Kurtzke et al 1995). Furthermore, susceptibility to infection did not persist beyond the age of 45 years. Overall, about 1:500 individuals exposed to the agent developed clinical manifestations of multiple sclerosis. To some extent the arguments advanced by John Kurtzke have had to be qualified retrospectively through the recognition of additional cases. Updating the story to 1998, Kurtzke and Heltberg (2001) reported prevalence of 66/10⁵ (95% CI 45–93) amongst Faroese age-adjusted to the 1969 United States population. There had been no appreciable change from rates recorded in 1960 (68/10⁵), 1970 (62/10⁵) and 1980 (63/10⁵). Seven new affected individuals with onset at around 21 years were identified between 1986 and 1990 and these have since been accommodated within a fourth epidemic (Kurtzke et al 1995). Pulling together a complex story, Kurtzke and Heltberg (2001) conclude that multiple sclerosis is an infectious disease introduced to the Faroese by British troops in the Second World War. The first wave of disease spread from a single individual born in 1919, exposed for 2 years, affected by 1943 and dead by 1971. Twenty others were caught up in the first epidemic, providing a reservoir of the primary multiple sclerosis infection handed on to others exposed for a minimum period of 2 years between the ages of 11 and 45 years. These contributed to the remaining affected individuals grouped in the four subsequent epidemics. Of the 83 cases prevalent at any time, 41 were fully native Faroese and 14 had lived away for short periods. These 55 constitute the epidemic cases; 15 who had lived away for longer and 13 foreign-born cases were excluded. Median survival has been 29 and 34 years in females and males, respectively. In ‘hanging up his Faroese epidemiological boots’ in 2001, John Kurtzke settled on the following numbers: epidemic one, 21 – exposed between 1941 and 1944; epidemic two, 10 – exposed between 1945 and 1951; epidemic three, 10 – exposed between 1958 and 1964; epidemic four, 13 – exposed between 1971 and 1977; and a single individual in epidemic five, exposed between 1984 and 1990.

On the nature of the transmissible agent, Kurtzke et al (1988) examined the possibility that British troops introduced canine distemper virus, pursuing the hypothesis developed by S.D. Cook et al (1978) to explain the epidemiology of multiple sclerosis in Orkney and the Shetland Islands, and in Iceland. In the Faroes, there was no evidence in affected individuals for previous infection by this virus. One patient only had owned a dog with distemper; and there was no correlation between outbreaks of canine distemper and the distribution of residence in incident cases of multiple sclerosis.

Proponents have speculated on dynamics of the transmission hypothesis, whilst accepting the overall concept of an epidemic. Cooke (1990) attached special significance to the absence of individuals, born in 1941–1945, contributing to the second and third phases, and took this to mean that whilst contact with the transmissible agent was universal amongst Faroese, exposure in the first 3 years of life provided absolute protection from the later development of multiple sclerosis. Any differences from the age at exposure analysis based on epidemiological observations elsewhere were attributed to the immunologically naive nature of Faroese encountering the provocative agent during the early 1940s. Sceptics remain to be convinced by any aspect of this analysis. As in Iceland, the availability of improved diagnostic expertise is offered by a resident neurologist (Poul Joensen, personal communication, 1998) to explain the Faroe Islands epidemic, and the lack of such services for the paucity of multiple sclerosis in Greenland. In his part retrospective and prospective survey of neurological disease in the Faroes, 28 cases of multiple sclerosis were identified as causes of disability from 1939 to 1975 – the distribution being 6, 17 and 4 in sequential decades from 1939 and 1 in the 1970s (Joensen 1992). Specific criticisms concerning validity of the diagnoses, exclusions, case ascertainment, definition of epidemics, and the putative role of the British occupation in generating this outbreak of multiple sclerosis (C.M. Poser and Hibberd 1988; C.M. Poser et al 1988) are robustly resisted by the main protagonists (Kurtzke 1993; Kurtzke and Hyllested 1988).

In Orkney and the Shetland Islands, the incidence and prevalence of multiple sclerosis were at one time higher, almost by an order of magnitude, than in other regions (Figure 2.28 ). In presenting their classical study, Poskanzer et al (1980a) reviewed certificates located in Edinburgh for deaths attributed to neurological disease in Orkney and Shetland, concluding that the first case of disseminated sclerosis occurred in an Orcadian who died in 1898. A second case was reported 10 years later, and thereafter the diagnosis became more common, coinciding with a general increase in awareness and the adoption of clinical criteria for the diagnosis of multiple sclerosis in neighbouring parts of Scotland. Changes in disease frequency in that era can reasonably be attributed to alterations in nosological fashion. Poskanzer et al (1980a) concluded that multiple sclerosis may have been no less frequent an illness in the early years of the 19th century than was later apparent. However, as we have indicated, estimates of prevalence carried out on four occasions between 1954 and 1974 showed a steady rise in frequency from 111/10⁵ in 1954 to 309/10⁵ in 1974 for Orkney, and from 134 to 184/10⁵ in Shetland over the same period.

Figure 2.28.

Serial change in the frequency of multiple sclerosis in Orkney and the Shetland Islands and northeast Scotland; figures are prevalence/10⁵ population with the year of ascertainment.

Recognizing that these changes did not necessarily reflect a genuine increase in incidence, Poskanzer et al (1980a) defined – as accurately as possible – the date of onset in 66 incident Orcadian and 53 Shetland cases between 1930 and 1969. They concluded that ascertainment would have been reasonably complete between 1940 and 1960, during which annual incidence of the disease was 2.2/10⁵/year in Orkney and 1.6/10⁵/year in Shetland. Quinquennial rates were relatively constant and they attributed a slight reduction towards the end of this period to incomplete ascertainment of undiagnosed cases, made more obvious by exclusion of individuals with possible multiple sclerosis using the criteria of Allison and Millar. Comparison of incidence and mortality rates confirmed the impression that there had been no significant alteration in statistics for the disease over this period other than those attributable to changes in classification and ascertainment. Several pioneering epidemiologists had previously identified factors tending to maximize the prevalence of multiple sclerosis in surveys conducted on islands (Allison 1963; Fog and Hyllested 1966; Poskanzer et al 1976; Sutherland 1956). Over the period of these studies, systematic depopulation in Orkney and the Shetland Islands resulted in an older population, at lower risk of multiple sclerosis. However, given stable incidence, the rise in prevalence was nevertheless attributed to increased survival, from 26 to 40 years in Orkney and 24 to 34 years in Shetland, between 1954 and 1974, together with improved case recognition.

S.D. Cook et al (1985) revisited the question of serial change in incidence for multiple sclerosis in Orkney, documenting the annual rate from 1941 to 1983 and claiming a steady reduction from 1964. Between 1941 and 1964, 53 incident patients with probable multiple sclerosis were identified in an estimated cumulative population of 506 541 individuals, whereas only 12 developed relevant neurological symptoms in the comparable group of 320 757 individuals occupying the islands between 1965 and 1982 and showing a different age structure. In a stable population, 33 incident cases would have been expected over this period. The prevalence had also fallen to 240/10⁵ in 1983 for probable and possible cases (compared with 309/10⁵ in 1974). The rates for probable cases were 193/10⁵ and 258/10⁵ in 1983 and 1974, respectively. Disease duration increased to 22 years (providing an estimate for life expectancy, using Poskanzer's method, of 44 years) and, at 52 years, the mean age at the time of prevalence was higher than for any other population recorded at that time. This overall change in disease frequency was attributed to a reduction in the number of older patients, suggesting that prevalence in the sixth decade and beyond had diminished as a result of fewer incident cases in younger age groups.

As part of the systematic attempt to survey multiple sclerosis in the Orkney and Shetland Islands, Stuart Cook and his local collaborators on the islands have serially updated the figures (S.D. Cook et al 1988) and last reported a prevalence for probable and definite cases of 287/10⁵ and 134/10⁵ for Orkney and Shetland, respectively, in 1994 (S.D. Cook, unpublished results). These statistics show that there has been significantly less multiple sclerosis in the Shetland Islands since 1965 but this is no longer true for Orkney.

The reduction in frequency of multiple sclerosis seen in northeast Scotland over the same period is reminiscent of epidemiological surveys from other regions that border the North Sea (see above). If these trends reflect a change in biological factors determining the frequency of multiple sclerosis, this can best be attributed, as John Kurtzke has emphasized, to a change in environmental conditions. Population genetics shape the distribution of disease more slowly. The alternative explanation is that the systematic increase in prevalence with time and gradual erosion of the latitudinal gradient in the United Kingdom are due the catching up of more recently surveyed areas with those studied repeatedly over a longer period and already epidemiologically saturated.

It is easy, especially for the individuals concerned, to assume that multiple sclerosis is contagious when a cluster occurs, apparently selecting a group with demographic or social links – for example attendance at the same school, domicile in the same street or membership of the same sports team. With small denominators and a disproportionately high numerator, these claims rarely stand up to detailed scrutiny. Statistical methods usually show that the clusters could as easily have arisen by chance.

Indeed, that was the preferred interpretation offered by Deacon et al (1959) to explain an apparent excess of cases in Duxbury, Massachusetts. Having been drawn to the survey by the suspicion of Dr Deacon that too many cases of multiple sclerosis had appeared in the community, the authors identified 16 cases in a population of 4900, of whom 8 were prevalent providing a rate of 224/10⁵ for all probable, and 163/10⁵ for more certain, cases; incidence was 13/10⁵/year (or 6.5/10⁵/year for the cases with physical signs) over the period 1945–1955. At that time, the highest reported prevalence in North America was 64/10⁵ for Rochester, Minnesota (MacLean et al 1950), with an incidence of 5/10⁵/year. Perhaps the authors’ commendable restraint in claiming a cluster was the disproportionate regional increase in prevalence, rather than incidence, the small sample, the potential impact of even one diagnostic error, the two familial cases, and the awareness that events happen by chance.

Amongst the 10 000 inhabitants of Mansfield, Massachusetts, word of mouth identified 13 probable cases and one possible example of multiple sclerosis, providing a prevalence rate of 141/10⁵ – three times higher than found elsewhere in New England at that time (Eastman et al 1973). All were white and 13/14 female. There had been no recent changes in demography. Eight affected individuals lived in close proximity and in a part of town long held to have a ‘suspect’ water supply.

Koch et al (1974) claimed as a time–space cluster, the six cases of multiple sclerosis known to have been born in the small town of Mossyrock, Washington State, between 1916 and 1921, where the population increased from an estimated 100 in 1920 to 415 in 1970. Another case, born in 1942, came to light through enquiry during the survey. A trawl for items of common exposure or comorbidity for infectious disease was suggestive. However, there was no documentary support for the belief, held amongst the cases themselves, that they all had smallpox in 1924; rather it seemed that an epidemic of measles might have occurred at around that time. Dressed up as rates for prevalence and incidence over the 24-year period of 1687/10⁵ and 82/10⁵/year, respectively, these look to be impressive distortions but, to us, the explanation lies in the familial aggregation: one family had two affected siblings; another had three affected siblings, and a first cousin and her son also with multiple sclerosis. In fact, only one case was sporadic.

Cook and Dowling (1982) reported three definite or probable and two possible cases of multiple sclerosis in Sitka, Alaska. Clinical onset occurred between 1967 and 1970. They linked this time cluster to an epidemic of canine distemper virus in dogs occurring in 1965. No cases were known to have been on the island during the period 1949–1979.

Murray (1976) identified ten patients from Nova Scotia living in close proximity during an outbreak of poliomyelitis in 1951. Although he sought a variety of environmental links between these cases, it is significant that six were related, and members of two extended families. A genetic explanation also seems probable for the cluster of 33 cases identified in a small rural community in northern Sweden of whom 21 were linked through kinship (Binzer et al 1994). Twenty people working in a New York industrial plant with c.6000 employees exposed to zinc developed multiple sclerosis between 1970 and 1989 (Schiffer et al 1994; E.C. Stein et al 1987). Here, an assessment of recurrence risks and screening for a number of candidate genes failed to confirm the investigators’ hunch that the cluster had a genetic explanation.

An epidemic has been claimed for Key West, a tropical island off the west coast of Florida, where 37 patients with peak onset in and around 1977–1979 were identified in 1984 (prevalence 140/10⁵). The change was not thought attributable to alterations in clinical vigilance or differential migration of symptomatic individuals to a more favourable climate (Sheremata et al 1985). Later, S.D. Cook et al (1987) checked out the rates of canine distemper on Key West and nearby Stock Island, using the recollections and records of local veterinarians, and claimed a close symmetry between outbreaks of canine distemper in dogs and clustering of multiple sclerosis in these remote parts of south Florida. However, referring to the same group of cases, MacGregor (1991) concluded that the cluster could not be explained by genetic susceptibility and exposure to canine distemper virus or measles. Rather, the vector was considered to be birds and Marek's disease virus the best microbial bet.

Eight of 283 children attending school in a small hamlet (Henribourg, population 75) in Saskatchewan over a 15-year period later developed multiple sclerosis (Hader et al 1990). All were affected by one of three measles epidemics to which the community was exposed, and each had consumed well water distributed by ladle from a large crock stationed near the school door. The authors were careful not to overstate the potential significance of this anecdote.

Haahr et al (1997) drew attention to eight Danes with multiple sclerosis who lived in a small isolated and stable community. All attended one school over a 7-year period and took part in the same activities, but two were siblings and two others second-degree relatives.

Nine new cases of multiple sclerosis occurred between 1971 and 1990 amongst the 1729 residents of DePue Township, Illinois (Schiffer et al 2001). Using various comparators, this was estimated to represent a relative risk of 4.1–17.1 (95% CI 3.1–20.0). No cases were identified before or after these dates. Concern had arisen that morbidity in this mining community was being influenced by exposure to mitogenic trace metals, especially zinc, over this period. Commenting on this report, Park (2002) added triaryl phosphate ester to the list of possible heavy metal culprits. Using a spatial statistic scan, Donnan et al (2005) searched a database of 772 cases identified in Tayside, Scotland, between 1970 and 1997 (providing an annual incidence of 7/10⁵/year) for clusters. The general increase in disease frequency showed some evidence for periodicity and peaked in the mid-1990s but with a time cluster between 1982 and 1995 throughout the region, and a spatial cluster in the rural area southwest of Perth between 1993 and 1995. The implication is that these patterns reflect the influence of exogenous aetiological factors, but the impact of changes in scrutiny, as the availability of neurological services has altered, cannot altogether be excluded.

Riise (1997) has brought definitions and common sense to the analysis of these clusters in north Atlantic Islands (Faroes, Iceland, Orkney and Shetland), the United States (Massachusetts, New York, Wyoming, Ohio, Key West and Alaska), Canada (Nova Scotia and Saskatchewan) and Scandinavia. Amongst the post hoc clusters, none is convincing with respect to a specific environmental factor and several seem more closely related to genetic influences. There is statistical evidence for clustering at 21–23 years of age and 2 years prior to onset in the Orkney cases but not Shetland. Conversely, the studies from Hordaland county, Norway, provide evidence for clustering of affected individuals as teenagers (13–20 with a peak at 18 years) but specifically not in the run up to disease presentation (Riise et al 1991).

THE ENVIRONMENTAL FACTOR IN MULTIPLE SCLEROSIS

It may seem coy endlessly to refer to those aspects of the aetiology and epidemiology of multiple sclerosis that seem not to be dependent on genetic susceptibility as environmental factors, without being more specific, but, despite much effort, these influences remain unidentified and still somewhat enigmatic. Although many would argue that the environmental contribution is microbial and, in all probability viral, the evidence is limited and alternative hypotheses continue to be explored both within the confines of mainstream multiple sclerosis research and in more maverick circles. Many analysts now conclude that multiple sclerosis does not have a single cause. We discuss later (see Chapter 14) the thorny issue of whether multiple sclerosis is a syndrome resulting from complex genetic traits, multiple causes and mechanistic heterogeneity, in which discrete pathological processes converge on a shared clinical phenotype, or a nosologically distinct but nonetheless complex disorder where definition has become blurred around the edges through confusion with related but recognizably different conditions.

Since it remains an assumption that the environmental factor in multiple sclerosis is microbial, some analysts have sought to implicate altogether different putative risk factors. Often, the evidence is anecdotal and with no attempt to test hypotheses using sound epidemiological principles. The distinction is not always made between factors that may initiate the disease process in multiple sclerosis, those that are claimed to expose latent disease, and events that alter the clinical course in individuals who have already experienced manifestations.

We review ‘life events’ that have been considered as candidates for affecting the course of the disease in Chapter 4: these include the effects of infections, vaccinations, pregnancy, trauma and stress. Here, these and some other extrinsic circumstances are reviewed for their potential contribution in causing multiple sclerosis – aetiological factors rather than modifiers of the disease process.

Infections

As in many situations where direct leads are lacking, one epidemiological approach is the non-prejudicial trawl for risk factors associated with multiple sclerosis. Hopkins et al (1991) examined many candidates – around the general theme of infection and immunity – in a case–control study involving 16 prevalent cases and 61 controls. They showed that a family history of neurological disease, a personal history of allergies, exposure to oral polio vaccine, owning a cat that died from unexplained causes, or completing high school and/or college education were independent risk factors for multiple sclerosis. Blood serology provided no additional clues to the development of multiple sclerosis amongst individuals included in this survey. Evidence for infectivity might be apparent from altered risk amongst those having close personal contact with affected individuals. The answer to the question of whether professionals who frequently meet patients with multiple sclerosis have an increased risk of developing the disease is ‘no’ but, for methodological reasons, the study by Dean and Gray (1990) cannot be regarded as definitive. Developing the theme that 7 of 307 nurses working in Key West (South Florida) were found to have multiple sclerosis, Dean and Gray (1990) compared expected and observed death certifications for multiple sclerosis in medical practitioners and nurses from the United Kingdom using Office of Population Censuses and Surveys statistics and documentary evidence available from the prospective British Doctors’ Smoking Study. In neither case did a trend towards increased risk emerge. However, we do not know the degree of contact between patients and these medical staff. Prospective records of the Danish Multiple Sclerosis Registry identified 60 cases of multiple sclerosis amongst 69 428 nurses registered from 1980 to 1996, where 69 would have been expected, representing a standardized incidence ratio of 0.87 (95% CI 0.6–1.12: Stenager et al 2003). The frequency of multiple sclerosis in the spouses of affected individuals (conjugal multiple sclerosis) is no higher than the population risk (Ebers et al 2000a; N.P. Robertson et al 1997). As we discuss in Chapter 3, the studies of recurrence risk in adoptees, half-siblings and twins also exclude an increased risk of multiple sclerosis through close personal contact with probands.

A survey of risk factors showed seasonal variation in the incidence of optic neuritis and suggested that the subsequent conversion to multiple sclerosis correlated with onset of optic neuritis in the winter or spring (Compston et al 1978). Jin et al (1999; 2000) have also shown that new episodes of demyelination, especially optic neuritis (ratio 1.84 compared with other disease manifestations; 95% CI 1.13–3.0) cluster in the spring (45% of all such episodes; 95% CI 35–55%), at least in northern Europe. Although this may tell us something about environmental factors causing optic neuritis, the seasonal association within risk of multiple sclerosis is almost certainly confounded by the generally high conversion rate of optic neuritis in adults. Others have shown seasonal variation in symptoms at onset (or during the course) of multiple sclerosis (Andersen et al 1993; Goodkin and Hertsgaard 1989). In Japan, 172 exacerbations occurring in 34 individuals with multiple sclerosis clustered in the six summer months but 123 (72%) were in those with optico-spinal disease – making it difficult to generalize these findings (Ogawa et al 2004). The increased springtime risk (but not in the winter) is confirmed in meta-analysis of nine studies reporting on optic neuritis (relative risk 1.19, 95% CI 1.16–1.23), six of onset in multiple sclerosis (relative risk 1.45, 95% CI 1.36–1.55) and nine of exacerbations in patients with pre-existing disease (relative risk 1.10, 95% CI 0.7–1.13; Jin et al 2000). Marrie et al (2000) used the General Practice Research Database to show an increased risk of respiratory infection in 5-, 12- and 52-week periods at risk before first onset of multiple sclerosis. Taken together, these indirect observations do no more than implicate environmental triggers for disease onset in multiple sclerosis but the issue has been more directly studied in the surveys of Sibley and colleagues (see Chapter 4).

Several investigators have attempted to correlate exposure to viral illness in childhood with the subsequent development of multiple sclerosis. Methodology has varied and the results are critically dependent on the choice of controls (for review, see Granieri and Casetta 1997). The picture to emerge from these studies is that the risk of developing multiple sclerosis is increased for individuals affected by a variety of exanthematous and other common viral disorders relatively late in childhood. The studies suggest that a narrow and age-linked period of susceptibility to viral exposure exists in those who are constitutionally at risk of developing the disease. Part of the evidence is provided by population serology but many prefer the interpretation that differences in the titres or distribution of antibodies in case–control series do not relate to specific viral antigens. Rather, these reflect a general enhancement in immune responsiveness. However, one candidate may yet hold some promise.

Compston et al (1986) first retrospectively compared historical and laboratory evidence for previous viral exposure in patients with multiple sclerosis and optic neuritis with a cohort of controls screened to have an increased frequency of HLA-DR2, so as to match for at least one marker of genetic susceptibility. Patients with demyelinating disease reported later age at infection by measles, mumps and rubella. Martyn et al (1993) subsequently re-analysed these data to show an even more marked effect of EBV infection. Those who reported having infectious mononucleosis at <18 years had a relative risk for multiple sclerosis of 7.9 (95% CI 2–38). Meanwhile, Lindberg et al (1991), looking at registries for infectious mononucleosis in Sweden, found a slight excess of multiple sclerosis following EBV infection compared with individuals not reporting exposure. Haahr et al (1995) used records from the Danish State Serum Institute register of EBV infections and the Danish Multiple Sclerosis Registry to identify 16 individuals who developed multiple sclerosis amongst 6853 experiencing EBV infection in the decade 1968–1978 (relative risk 2.8). Median age at infection was 17 years. This did not differ from the age at infection in heterophile antibody-positive individuals not developing multiple sclerosis. Goldacre et al (2004) reported an increased risk of multiple sclerosis ≥10 years after hospital admission in Oxford, England, with infectious mononucleosis (rate ratio 14.0: 95% CI 1.5–8.9). High population exposure to EBV makes for difficulty in resolving the disease-related status of these serological observations. Munch et al (1998) implicated infection due to an EBV subtype, identified by a 39-bp repeat of the EBNA 6 coding region, in a cluster of eight cases from Fjelso, Denmark. Vaughan et al (1996) used more discriminating laboratory markers and showed an enhanced immune response to EBV, involving the glycine/alanine repeat peptide (p62) epitope of EBV nuclear antigen-1, also expressed by other pathogens and cross-reacting with neuroglial cells, in patients with multiple sclerosis previously reported from Norway (Riise et al 1991). Subsequently, Marrie et al (2000) reported an odds ratio for multiple sclerosis of 5.5 (95% CI 1.5–19.7) for individuals with a clinical history of infectious mononucleosis based on their access to the United Kingdom General Practice Research Database. Hernán et al (2001) confirmed a moderate increase in risk of multiple sclerosis amongst nurses with a history of infectious mononucleosis (2.1: 95% CI 1.5–2.9) especially if the history was confirmed serologically (2.3: 95% CI 1.6–3.5). They also showed that late age at infection by mumps and measles conferred an increased risk of demyelinating disease. H.J. Wagner et al (2000) demonstrated universal infectivity by EBV in patients with multiple sclerosis and suggested, on the basis of reduced EBV nuclear antigen-1 antibody, that affected individuals may have defective control of persistent latent EBV carrier state and reactivation. Using samples collected over a 10-year period, Alotaibi et al (2004) provided serological evidence for remote EBV infection in 83% of 30 children with multiple sclerosis, compared with 42% of 90 matched emergency department controls and 53 healthy children (p = 0.001). Conversely, exposure to herpes simplex virus was less likely in affected individuals and the effects of cytomegalovirus, parvovirus B19 and varicella zoster were neutral. Ponsonby et al (2005) used indirect epidemiological methods to substantiate the hygiene hypothesis, whereby early exposure provides immunological protection, in the specific context of EBV infection: 136 Tasmanian patients, identified from a population-based sample and studied between 1999 and 2001, were compared with 272 matched controls for a history of contact with siblings aged <2 years during their own early childhood (<6 years). Taking 1 year of contact as the reference point, increased exposure correlated with a reduced risk of multiple sclerosis (1–<3 years: adjusted odds ratio 0.57, 95% CI 0.33–0.98; 3–<5 years: 0.40, 95% CI 0.19–0.92; ≥5 years: 0.12, 95% CI 0.02–0.88: p = 0.002). In controls with a history of increased exposure to infant siblings, IgG responses to EBV were reduced and there was a lower probability of infectious mononucleosis. However, in cases, although the reported frequency of mononucleosis and EBV IgG antibody responses were higher, these did not correlate with periods of exposure to younger siblings. Therefore, the simple notion that early contact provides protective immunity, a lower frequency of infectious mononucleosis (at the critical age, later in childhood), and a lower frequency of multiple sclerosis is not directly supported. But, as we discuss in Chapters 5 and 11, the story implicating EBV in the aetiology of multiple sclerosis gains credence from immunological studies showing molecular mimicry with epitopes of myelin basic protein (Lang et al 2002).

The list of other microbial agents that have at some time been implicated in the aetiology of multiple sclerosis on the basis of single but unconfirmed isolations is long. The catalogue includes rabies, herpes simplex viridae, scrapie agent, parainfluenza 1, measles, the ‘Carp’ and ‘bone marrow’ agents, cytomegalovirus and coronavirus (for a summary of the earlier studies, see R.T. Johnson 1982). Subsequently, the question arose of whether patients with multiple sclerosis had at one time been exposed to human retroviruses. The occasional reports of retroviral genomic material in the central nervous system tissue of patients with multiple sclerosis (Cosby et al 1989; Greenberg et al 1989; E.P. Reddy et al 1989) sustained expectations that a specific cause could be discovered. However, these anecdotes appear no more consistent (Bangham et al 1989; Nicholl et al 1993; Richardson et al 1989) or specific than the serological surveys and periodic viral isolates from cell cultures. Although the nervous system is an important reservoir and site for cytopathic retroviral infection, exhaustive search for evidence of retrovirus using reverse transcriptase activity in blood and cerebrospinal fluid of patients with multiple sclerosis has proved relatively unrewarding (Hackett et al 1996). Mogensen (1997) reviewed in detail the question of whether these analyses provide evidence for retroviral infection in multiple sclerosis. Whilst the hypothesis remains tenable, the facts are scanty and, of the newer candidate retroviruses, none appears better supported than its predecessors.

The story began with preliminary reports implicating HTLV-I and HTLV-3 (HIV-1; Koprowski et al 1985) in multiple sclerosis. But these were not confirmed (S.L. Hauser et al 1986; Karpas et al 1986). Sommerlund et al (1993) demonstrated retrovirus-like particles in a cell line producing EBV from a patient with a chronic myelopathy attributed to multiple sclerosis. Munch et al (1995) demonstrated EBV and a novel retrovirus in a greater proportion of B-cell lines from patients with multiple sclerosis than controls. Against the background of failure to implicate human endogenous retroviruses (HERV) in multiple sclerosis (Rasmussen et al 1995), Clerici et al (1999) showed enhanced peripheral blood mononuclear cell proliferation and cytokine responses to human HERV peptides in patients with active multiple sclerosis. Trabbattoni et al (2000) reported increased production of interleukin-2 and IFN-α, and decreased amounts of interleukin-10 during disease activity, after stimulation of peripheral blood mononuclear cells from individuals with multiple sclerosis with HERV peptides. In a survey of HERV expression in monocytes and brain tissue from individuals with a variety of inflammatory brain diseases, Johnston et al (2001) claimed more specific increased expression of HERV-W and HERV-K in brain tissue from individuals with multiple sclerosis (and selected other disorders). Christensen et al (2003) described increased levels of antibody to HERV-H/RGH-2 peptide sequences both in serum and cerebrospinal fluid, and detected HERV-H protease-env splice variant more frequently in mRNA extracts from peripheral blood lymphocytes. Most recently, a mechanistic explanation has been offered through demonstration that increased expression of the HERV-W encoded glycoprotein synactin in astrocytes recovered from acute lesions in multiple sclerosis proved toxic to oligodendrocytes through the release of reactive oxygen species; and antioxidants were effective in protecting oligodendrocytes and limiting the behavioural consequences of tissue injury, in a mouse model of inflammatory brain disease (Antony et al 2004).

One component of the HERV-W family is designated ‘the multiple sclerosis associated retrovirus’ (MSRV). Perron et al (1989) had first isolated this retrovirus from the spinal fluid of a patient with multiple sclerosis. Subsequently, this was demonstrated in serum samples from >50% patients in relapse, and extracellular virions were recovered from cerebrospinal fluid containing MSRV pol sequence in 5/10 patients with multiple sclerosis but no controls with other neurological diseases (Perron et al 1991; 1997). Similarities with paramyxovirus were used to explain serological findings previously reported in multiple sclerosis on the basis of antigen cross-reactivity. Human herpes virus 6 (HHV-6; see below) may trigger expression of MSRV. In collaborative work, Garson et al (1998) extended these observations to show that MSRV could be detected in serum from 9/17 (53%) patients, usually untreated, compared with 4/44 controls (7%). Evidence for an association between multiple sclerosis and MSRV continues to accumulate, with differences reported between cases and controls from Sardinia (Dolei et al 2002) perhaps correlating with a progressive clinical course (Sotgui et al 2002), but whether this is a causal or epiphenomenal relationship still remains unclear.

Immunological methods have been used to provide indirect evidence for virus exposure through, for example, the demonstration of antigen-specific differences in cytotoxic T-cell responses to measles (S. Jacobson et al 1985). Wucherpfennig and Strominger (1995) claimed that myelin basic protein-specific T-cell clones from patients with multiple sclerosis are activated by peptides based on those required for class II binding and T-cell recognition and derived both from a bacterium (Pseudomonas aeruginosa) and from several viruses (herpes simplex, adenovirus, papillomavirus, influenzavirus and EBV). Since only one pathogen showed sequence homology with the activating peptides, the suggestion is that more than one trigger can activate antigen-specific T cells as part of a nonspecific response to infection (see also Chapter 11). An alternative approach has been to use similarities between the pathology and clinical course of virus-induced models of demyelination to support the hypothesis of viral causation in multiple sclerosis. Viral modelling of the pathology and course of human demyelinating disease remains attractive because of the ease with which these systems can be manipulated but the approach does not provide direct evidence for a viral cause in multiple sclerosis. For example, the same sequence of events characterizes the development of tissue injury in visna-maedi as that which occurs with evolution of the plaque in multiple sclerosis. A lentivirus-induced interferon response increases class II histocompatibility antigen expression, amplifying the inflammatory component, whilst also restricting viral replication and leading to persistent infection (Kennedy et al 1985). In Theiler's virus infection of the murine nervous system, the initial infection of neurons is followed by direct damage to oligodendrocytes, especially in the spinal cord, producing inflammation and demyelination ( Fiette et al 1993; Pena-Rossi et al 1991). The recognition of genetic restriction in susceptibility to Theiler's virus infection (Bureau et al 1993) serves to increase the similarities with multiple sclerosis.

The results of population serology attracted much attention in the 1960s and 1970s but the candidature of organisms implicated through the demonstration of increased antibody titres could not be corroborated using independent techniques. Subsequently, sero-epidemiological studies of Borrelia burgdorferi suggested a link with multiple sclerosis. Chmielewska-Badora et al (2000) reported an increased frequency of seropositivity in patients with multiple sclerosis (38%) compared with controls (20%). Brorson et al (2001) described changes in cerebrospinal fluid and serological reactivity indicating that patients with multiple sclerosis harboured a spirochaetal organism indistinguishable from B. burgdorferi. The two organisms attracting most attention since we last reviewed the topic are human herpes virus 6 (HHV-6) and Chlamydia pneumoniae.

Sriram et al (1999) first identified C. pneumoniae in a single patient with multiple sclerosis. This then led to the detection in cerebrospinal fluid of chlamydia in more cases (64%) than controls (11%) and detection of chlamydial genetic material in most patients (97% compared with 18% in controls). There was an excess of anti-chlamydia antibodies (86% in cases vs. 18% in controls). Others could not confirm these findings (Boman et al 2000; Saiz et al 2001; Treib et al 2000). Although individual reports continue to describe an excess of antibodies to C. pneumoniae in the cerebrospinal fluid of patients with multiple sclerosis (Hao et al 2002; Krametter et al 2001; Sotgiu et al 2001a), or those with particular features (Contini et al 2004), perhaps the most telling study is a comparison between three laboratories of polymerase chain reaction detection rates for Chlamydia pneumoniae in the same samples, one being the laboratory that reported the original findings. Whereas Sriram et al (1999) found the organism in the majority of patients but few controls, the other two were unable to detect chlamydias in any samples (M. Kaufman et al 2002). The story is not settled. Derfuss et al (2001) confirmed an increase in intrathecal anti-C. pneumoniae antibody synthesis. There was no correlation with disease duration or severity. The oligoclonal bands did not react with C. pneumoniae antigen and this could not be identified using molecular techniques. Together, the results suggested a nonspecific polyclonal antibody response rather than implicating an aetiological role for C. pneumoniae. A population serology survey using the Nurses Health Study cohort showed an increased odds ratio of 1.7 (5% CI 1.1–2.7) for C. pneumoniae seropositivity, especially in patients with progressive multiple sclerosis (7.3: 95% CI 1.4–37.2; Munger et al 2003). Grimaldi et al (2003) detected sequences of the major outer membrane protein from cerebrospinal fluid in 21% of 107 patients with multiple sclerosis but only 3% of 77 controls. Amongst patients, MRI markers of disease activity and younger age at onset correlated with the presence of C. pneumoniae in cerebrospinal fluid samples from patients with relapsing–remitting multiple sclerosis but not those who had entered the chronic progressive phase. Buljevac et al (2003a) correlated peaks in serologically defined C. pneumoniae (generally indicating positive polymerase chain reactions for chlamydia) with an increased risk of exacerbation. Dong-Si et al (2004) detected C. pneumoniae-specific DNA more frequently in the cerebrospinal fluid of individuals with relapsing–remitting multiple sclerosis than those with other neurological disease although the difference narrowed considerably in comparisons with cases having conditions other than multiple sclerosis but nevertheless with abnormalities of the spinal fluid – suggesting that activity rather than disease specificity is the critical trigger to this response.

Nonspecific features supporting the candidature of HHV-6 include the ubiquitous exposure early in childhood, neurotropism, ready reactivation, and the range of cell types that can be infected. But against that general background, Merelli et al (1997) failed to detect HHV-6 or HHV-8 DNA in all but a few samples of mononuclear cells from 56 patients with multiple sclerosis. Although HHV-8 could be detected in brain tissue from each of five patients, virus was also recovered from adult but not perinatal stillborn controls. Ablashi et al (1998) found that patients with multiple sclerosis had an excess of antibody to early HHV-6 protein (>68% vs. 28% in controls) in comparison with EBV and cytomegalovirus. J.E. Friedman et al (1999) reported an excess of HHV-6 structural protein, and antibody to HHV-6 in serum and brain tissue in multiple sclerosis compared with controls. Soldan et al (2000) showed increased peripheral blood lymphocyte reactivity of the neurotropic HHV-6A variant. Knox et al (2000) identified HHV-6 in 73% of brain and lymphoid tissue samples from patients with multiple sclerosis, especially those showing active demyelination (90%), more often than in controls (13%). In a comprehensive virological screen for HHV-6 infection, Chapenko et al (2003) correlated genomic sequences and viral mRNA transcription in peripheral blood mononuclear cells, plasma viraemia, and serum antibodies with clinical, radiological and immunological features of multiple sclerosis. They concluded that HHV-6 is associated with multiple sclerosis, and virus reactivation occurs during periods of disease activity perhaps through modulation of interleukin-12 (IL-12) synthesis. Goodman et al (2003) identified HHV-6 genome but not viral antigen in lymphocytes, oligodendrocytes and microglia from all of five biopsies in patients with multiple sclerosis. In pursuing a mechanistic hypothesis, Tejada-Simon et al (2003) identified a peptide having identical sequence in myelin basic protein (residues 96–102) and HHV-6 (residues 4–10). Patients had increased antibody titres to both antigens compared with controls and a high proportion of T cells recognized and reacted identically to either the myelin basic protein or HHV-6 derived peptides. But a serological study failed to confirm excess seropositivity for HHV-6 in patients with multiple sclerosis from Kuwait (Al-Shammari et al 2003). The most recent reports continue to provide conflicting evidence. Based on an uncontrolled study, Rotola et al (2004), reported evidence for HHV-6 (variant A) infection early in the course of multiple sclerosis. Conversely, Tuke et al (2004) showed no difference between cases and controls or between plaques and non-plaque regions in post-mortem brain tissue.

Reviewing the evidence from the position of sceptical virus watchers in multiple sclerosis, Swanborg et al (2003) conclude that the candidature of neither C. pneumoniae nor HHV-6 is yet secure. Review of the literature concerning varicella zoster virus reaches the same negative conclusion on its potential role in the pathogenesis of multiple sclerosis (Marrie and Wolfson 2001). Uncertainty breeds speculation. Those who develop hypotheses about the aetiology and mechanisms of multiple sclerosis exercise the luxury of picking selectively from the entire corpus of knowledge on the disease, any facts that decorate their particular interpretation without necessarily having to test the ideas experimentally. Although often a harmless enough armchair exercise, occasionally the ideas cause distress, undermine mainstream doctrines on the aetiology and mechanisms of tissue injury in multiple sclerosis, require time spent dealing with spurious media interest provoked by the more provocative claims, and distract from serious research. According to Hawkes (2002), multiple sclerosis is sexually transmitted – teenage-onset cases representing examples of child abuse. In the distress that followed this suggestion, the author claimed that his ideas were not intended for lay attention – but agreed to speak on national radio. An institution to which the article was linked dissociated itself from responsibility. Journalists waved the freedom of speech banner. Patients were upset. The Multiple Sclerosis Society was kept busy with enquiries and its professional advisers were required to assemble coherent arguments to offset the implausible. Our own view expressed at the time was that as no new facts were reported, this paper had little if any scientific value. The hypothesis fell down quickly and repeatedly in the face of known facts. The specific claim that multiple sclerosis in young people might result from child abuse seemed mischievous and deeply wounding.

The many unsuccessful attempts to identify causative factors that trigger the disease process leading to demyelination in patients with multiple sclerosis, have fuelled speculation on whether the environmental event is an unusual reaction to a ubiquitous agent or the ubiquitous response to a rare infection (P.G.E. Kennedy and Steiner 1994). The story of searches for a candidate that qualifies as the cause (or one cause) of multiple sclerosis seems much influenced by biomedical fashion and technical opportunities deploying increasingly sophisticated methods of investigation. As each new microbial hare is set running, chased and caught, we seem no closer to incriminating a particular agent.

Noninfectious environmental events

Pursuing reasons for the gender bias in the frequency of multiple sclerosis, Hernán et al (2000) used the Nurses’ Health Studies I and II (involving 315 index cases from amongst 121 700 aged 30–55 years in 1976 and 116 671 aged 25–42 years in 1989) to show that neither parity nor use of the oral contraceptive pill influence the risk of developing multiple sclerosis, even after adjustment for potential confounding effects. The relationship of pregnancy to multiple sclerosis mainly affects relapse rate in those with established disease (see Chapter 4) rather than the risk of developing multiple sclerosis. There is no evidence that the onset of multiple sclerosis clusters around pregnancy (Birk et al 1990; Roullet et al 1993; Worthington et al 1994). Coming at the problem from the position of whether the higher reported relative risk for multiple sclerosis should be used to modify protocols for hepatitis B vaccination, Tosti et al (1999) concluded that much would be lost and little gained in terms of overall public health measures by discontinuing the existing vaccination programme. Perhaps they need not have worried. Zipp et al (1999) did not observe any increase in monophasic demyelinating diseases or multiple sclerosis after hepatitis B vaccination in a cohort of 134 698 individuals enrolled in a United States health care database from 1988 to 1995. Using the Nurses Health Study databases, Ascherio et al (2001) showed that the risk of developing multiple sclerosis (and the risk of relapse in individuals with pre-existing disease, see Chapter 4) is not increased after hepatitis B vaccination (relative risk 0.9: 95% CI 0.5–1.6 for any exposure; and 0.7: 95% CI 0.3–1.8 for the previous 2 years) irrespective of the dose or use of recombinant vaccines. But all is not settled. Hernán et al (2004) used the General Practice Research Database (GPRD) in the United Kingdom to assess the frequency of immunizatons in the 3-year period before onset of symptoms in a cohort of 163 cases incident over a 7-year period compared with 1604 controls. There was an increased risk attributable to hepatitis B (odds ratio 3.1, 95% CI 1.5–6.3) but not tetanus or influenza vaccinations. Commentators have drawn attention to epidemiological deficiencies of this study.

There is a basic human need to explain unexpected illness and everyone feels unreserved sympathy for the person who has multiple sclerosis in their coming to terms with the clinical manifestations and adjusting to their social and domestic implications. It is understandable that an affected person should connect onset or a change in clinical course with life events, such as a recent road traffic accident, but this necessarily raises issues of general importance in which compassion cannot influence the interpretation of evidence. Although the overwhelming majority of medical opinion does not support the conclusion that there is any relationship between trauma and multiple sclerosis, this view is not universally held. The possibility of a relationship between trauma and multiple sclerosis began with the clinical observation that some patients recall an episode of trauma shortly before the onset of symptoms or an exacerbation of pre-existing manifestations of the disease. Clinical anecdotes sustained this belief in an era when evidence-based medicine was poorly developed. Even then, many neurologists concluded that the relationship was coincidental. Since there is consensus that the process leading to tissue damage in the central nervous system that forms the pathological basis for multiple sclerosis is established in childhood, no informed person would now claim that trauma causes the disease. But that leaves unanswered the subsidiary questions of whether trauma allows latent multiple sclerosis to manifest, causes clinically overt disease to relapse, or adversely affects the natural history of the disease (see Chapter 4).

In the 1970s, some patients with multiple sclerosis were advised to have all their amalgam dental fillings replaced on the basis that these contained elemental mercury at about 50% by weight. Mercury is absorbed from amalgam fillings and accumulates in tissues, including the central nervous system. However, we are not aware of any epidemiological evidence that correlates dental management with any aspect of multiple sclerosis, and share the view of the American Dental Association (see Lancet 3 August 2002, p. 393) that patients should not have their fillings replaced. Although no difference in number of amalgam fillings (and hence body mercury levels) was found in a case–control study from Leicestershire (England) involving 39 female prevalent patients, McGrother et al (1999) correlated dental caries with an increased risk of having multiple sclerosis. Cassetta et al (2001) have also formalized the status of this risk factor, showing no significant association between amalgam fillings and the risk of multiple sclerosis. Moving from teeth to industrial exposure, Mortensen et al (1998) used the Danish Multiple Sclerosis Register to exclude an effect on the risk of developing multiple sclerosis from presumed solvent exposure by nature of occupation. A national newspaper in the United Kingdom (Guardian, 7 June 2003) highlighted the plight of a small community in Scotland where 10 of 600 residents developed multiple sclerosis (period prevalence c.1700/10⁵), allegedly due to tributyltin exposure related to protecting the hulls of small ships from overgrowth of barnacles.

In a rambling review, Behan et al (2003) claimed that multiple sclerosis is a neurodegenerative and invariably clinically progressive trait driven by the product of a gene encoded on chromosome 17 that is influenced by sunlight and vitamin D activity. We provide a critical analysis of this hypothesis in Chapter 1. But is there substance to the sunshine theory? Acheson et al (1960) first linked variations in disease frequency to hours of annual and winter solar radiation taking the position that visible solar radiation protected individuals from developing multiple sclerosis. Lindstedt (1991) related this directly to the latitudinal gradient in disease frequency. Castigating others for not seeing the light with respect to the aetiology of multiple sclerosis, Hutter and Laing (1996) developed the hypothesis that illumination suppresses the immune activation properties of melatonin and the release of inflammatory leukotrienes – making the case for an immunosuppressant effect of environmental light at 2500 lux, and explaining the seasonal onset (winter) of symptoms due to demyelinating disease. Also backing melanin as a marker of the effect, Dumas and Jauberteau-Marchan (2000) proposed that sunlight inhibits the function of cutaneous antigen presentation (Langerhans cells). They suggested that this might account for the rarity of multiple sclerosis in blacks, whereas white immigrants to sunny places possess genes that cannot benefit from the protective effects of sunshine. Conversely, the offspring of these immigrants are advantaged by early exposure to sunlight since this sets their melanocyte and Langerhans cell repertoire more favourably. Mortality returns for multiple sclerosis documented in 24 North American states from 1984 to 1995 showed a negative correlation (i.e. protection) from residential and occupational sunlight (odds ratio 0.24; D.M. Freedman et al 2000). Previously, Norman et al (1983) had made the valid point that whilst air pollution, ground minerals, solar radiation, temperature, annual rainfall and humidity all correlated independently with birthplace in 4371 veterans having multiple sclerosis, these associations were all eliminated by correction for latitude suggesting that the correlations were confounded. van der Mei et al (2001) found a strong inverse relationship between ultraviolet radiation and prevalence of multiple sclerosis in six Australian regions, and offered an immunological hypothesis based on T-cell-mediated immunosuppression to account for these findings. Later, they provided an inverse correlation between exposure to sunshine in childhood and early adolescence, especially during the winter months, and the subsequent development of multiple sclerosis in Tasmania (odds ratio 0.31: 95% CI 0.16–0.59; van der Mei et al 2003). These epidemiological data have been used to support the hypothesis that 1α,25-(OH)₂ vitamin D3 exerts an anti-inflammatory effect mediated through enhanced suppressor cell function (Hayes et al 2003). Amongst 187 563 individuals registered with the Nurses Health Study I and II, prospective analysis of dietary vitamin D showed a protective effect of highest compared with lowest intake (age-adjusted relative risk 0.67: 95% CI 0.40–1.12) and for dietary supplementation (0.59: 95% CI 0.38–0.91; Munger et al 2004).

Oikonen et al (2003) correlated monthly hospital admissions for 1205 exacerbations in 406 patients with multiple sclerosis from southwestern Finland during 1985–1999 with ambient air quality, and correlated relapse rate with high concentrations of inhalable particulate matter (PM10). The confounding effect of PM10s on respiratory infection and relapse rate was not thought responsible for this effect. Bolviken et al (2003) mapped gradients in the distribution of multiple sclerosis in Norway with radon in indoor air, and suggested that this is modulated by exchangeable magnesium in the soil and rainfall – each reducing the availability of aerial radon – and showing a negative relationship with multiple sclerosis. Gilmore and Grennan (2003) supported the hypothesis that radon exposure at <15 years is a risk factor for multiple sclerosis. Others have examined the relationship between multiple sclerosis and exposure to ionizing radiation. Based on pooled analysis from two case–control series, Axelson et al (2001) reported an odds ratio of 4.4 (95% CI 1.2–2.6) for radiological work and X-ray examinations, noting that five cases in one series had been treated with ionizing radiation.

The possibility that something related to climate may yet be involved is supported by a recent epidemiologically sophisticated survey using 17 874 cases from the Canadian population-based register and 11 502 affected individuals from the United Kingdom, part population-based and part retrospectively identified from death certificates (Willer et al 2005). In both countries, being born in November was protective for the development of multiple sclerosis, and (in the United Kingdom) birth in May placed individuals at increased risk – each by comparison with unrelated and sibling controls. Combining these samples and datasets from Scandinavia showed that 8.5% of the 42 045 individuals were born in November compared with 9.1% in May [p corrected (p_c)<0.0001 for each comparison with controls], representing a 13% increase in risk for those born in May (95% CI 5–22%). The effect showed some regional variation, being most apparent in Scotland where the prevalence of multiple sclerosis is high. Despite the very discrete intervals of protection and risk, the authors do not doubt the validity of their findings (based on large numbers, replication and internal consistencies), and consider that these reconcile previous ambiguous findings on season of birth and risk of multiple sclerosis from these same countries (Sadovnick and Yee 1994; Salemi et al 2000b; Torrey et al 2000; Wiberg and Templer 1994). They offer maternal folate, correlates of infant birth weight, virus infection, and factors predisposing to schizophrenia (which evidently shows a similar effect) – but not signs of the Zodiac – as candidates, preferring seasonal reductions in levels of maternal vitamin D as the most plausible explanation.

With little progress made from the selection of candidate conditions that might account for the environmental factor in multiple sclerosis, less focused screens have been performed in the hope that a general trawl might yield an interesting catch. Such an approach is statistically fraught given the probability of 1:20 factors generating a ‘statistically significant’ result if considered alone. But, to date, screening populations of patients for antecedent events that can be implicated in the initiation of multiple sclerosis has not identified novel environmental triggers. Souberbielle et al (1990) found no difference in past history of specific infections or autoimmune disease in 230 cases and 230 controls although there was an excess amongst cases of hairdressers, and those having professional contact with pathology specimens. The previously reported claim that patients with multiple sclerosis have a greater number of childhood domiciles than controls is refuted by Savettieri et al (1991). S. Warren et al (1991a) reported a case–control study of environmental conditions at onset and during childhood in 173 Canadian patients with multiple sclerosis. Apart from a family history of diabetes (and multiple sclerosis) in patients with age of onset at <20 years, and some association with rural residence, no aetiological clues emerged from this study. In an exhaustive survey, Kurtzke et al (1997) revealed no clues regarding the cause of the putative series of multiple sclerosis epidemics on the Faroe Islands by comparing education, occupation, residences, bathing, sanitary or drinking facilities, domestic architecture or source of heating, diet, contact with animals, most vaccinations (those for smallpox, tetanus and diphtheria were less common), history of exanthematous illnesses, operations, hospitalizations and injuries, and age at menarche. Based on a comparison of 200 recent-onset cases with 202 controls, Ghadirian et al (2001) described an increased risk of multiple sclerosis through cigarette smoking, eye disease, a family history of cancer and autoimmune disease, trauma (see below) and contact with caged birds, whereas there was an inverse relationship between disease frequency and contact with cats. Riise et al (2003) also demonstrated an increased risk of multiple sclerosis (relative risk 2.7 for men and 1.6 for women) amongst 22 312 residents of Hordaland County, Norway, who had smoked for ≥15 years. Most recently, Hernán et al (2005) compared smoking habits amongst 201 people with multiple sclerosis registered with the General Practice Research Database between January 1993 and December 2000 and 1913 matched controls to show a modest increase in risk (odds ratio 1.3: 95% CI 1.0–1.7) that was more marked for those who had developed secondary progression at a mean follow-up interval of 5.3 years (hazard ratio 3.6 (5% CI 1.3–9.9).

Whether this catalogue includes environmental factors – microbes or conditions unrelated to infectious disease – that do genuinely determine the development of tissue injury in multiple sclerosis must, for now, be a matter of opinion. Many will remain to be convinced that anything of relevance has yet been identified. Others will have greater confidence in the observations made to date: age at infection by EBV; HHV-6; and climate-related alterations in vitamin D status will all have their advocates. But by any analysis, chasing the environmental factor(s) in multiple sclerosis has proved to be an even more perplexing and unrewarding task than practically any other branch of research into this enigmatic disease.