key: cord-0009782-vkueawox authors: Lauer, K. title: Environmental associations with the risk of multiple sclerosis: the contribution of ecological studies date: 2009-03-16 journal: Acta Neurol Scand DOI: 10.1111/j.1600-0404.1995.tb05861.x sha: b37ba6482e3f791298a03adbb525b321483a86c2 doc_id: 9782 cord_uid: vkueawox nan Although the role of genetic factors in the etiology of multiple sclerosis (MS) is presently undisputed there is a general consensus that exogenous factors play the major role. In a rather conceptual way one can subdivide the human environment into a number of fields which are by no means truly separated from each other. Commonly these are categorized into a geoclimatic and a sociocultural compartment, but even between these two compartments considerable overlaps exist. For example, a natural feature easily changes to a sociocultural one by exploitation and specific use by man. Substantial variables within the individual fields can operate by quite different mechanisms, including toxic-metabolic and/or immunological paths, but they might also be vectors of infectious agents. Thus a strict fixation on one or another of these two basic mechanisms of causation cannot be made by epidemiology, as long as no clear knowledge is available from clinical or experimental research. In addition it should be kept in mind that epidemiology is almost exclusively able to generate and to support hypotheses; final confirmation has to be left to the basic disciplines. Criteria for causal inference in epidemiological research have recently been reformulated by M. Susser (1) . Especially noteworthy is the important criterion of consistency in terms of replicability and of survivability, the latter term meaning the achievement of the same result under varying methodological conditions. Another important aspect is statistical coherence in the form of a doseresponse gradient, whereas specificity of the as-sociation is no longer considered a strict prerequisite for causality (2). Methodological problems of the ecological approach with special reference to MS will be treated in a subsequent paper (3) , and only some aspects shall be mentioned here. Beside easy performance, the lack of recall and selection bias for example, and high external validity, the fact that consistency is easy to be demonstrated is a clear advantage, and a test for a dose-response gradient is already inherent in the basic study design. There are, however, well known shortcomings: the individual risk is not tested, and many confounders are in play. As Greenland & Robins (4) pointed out, ecological confounding can be independent of individual confounding, and they mentioned some problems which were summarized under the term 'cross level bias' e.g. due to non-linear effects, effect modification and inhomogenous distribution within a given subarea. In the following those fields of environment, which were the subject of more systematic research into MS by the ecological approach shall be discussed. Regarding altitude, or mountain character of landscape, respectively, the literature is fairly inconsistent. In a large number of studies higher rates in more elevated regions in comparison with lowlands were emphasized (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) some of which included even a formal statistical analysis (8, 11 , 14, Lauer 17). There are, however, a somewhat smaller number of negative investigations (1 8-25 ; unpublished data for Albania). Thus the problem remains controversial, but the direction of the results is clearly biased: either no difference or a higher risk in more elevated areas,and, with the exception of Cyprus (24), not the reverse, was reported. This might point to environmental features associated with mountains in some, but not in other regions, and thus actually facilitate the search for true risk factors. This is another point of interest, especially under the aspect of a biological plausibility, regarding its possible impact on immune function. There are surprisingly few investigations giving only sparse data so far. Thus Barlow, in 1960, performed an extensive correlation analysis between the MS mortality rate and cosmic ray intensity for which geomagnetic latitude was used as a proxy measure (26). He found such a correlation worldwide and within individual countries such as the USA, Australia, and Italy. Resch (27) recently revived this view, and confirmed Barlow's findings on the basis of more recent prevalence data on the global level. In addition he emphasized the parallels of MS and geomagnetic latitude in the most northern parts of Europe. Apart from sun radiation (vide infra) natural types of radiation were not studied further. For example an interesting question might be a role of terrestric radiation due to radon decay which is high in mountainous areas (28). This might be an important subject for future research by both the ecological and the case-control approach. A classical comprehensive analysis was carried out by Acheson et al. in 1960 for the USA (29) where they clearly demonstrated a high correlation between MS mortality and both low temperature and low December solar irradiation, i.e. a lack of sunshine. This was later confirmed by Norman et al. (11) who found a positive association for a large number of climatic variables. Statistical adjustment for latitude removed, however, most of these associations, with the exception of a product term of rainfall and low temperature. The problem of such an adjustment will be treated below. In a greater number of further regions positive associations with MS were reported for low temperature (8, 14, 17, 18, (31) (32) (33) (34) . As far as humidity/ precipitation is concerned, our own investigations by the correlational approach in the USA (33), France and Australia point in particular to the winter conditions as more important in that respect. For example in the USA, beside the consistent correlation with low temperature, a positive one with snowfall but not with annual rainfall was found. The same was true for winter, but not for annual and summer humidity in Australia (Table 1 ). In France the MS mortality (35) showed a very similar northeast-to-southwest gradient as both January temperature and January humidity (36) ( Table 1 ). In the Australia and New-Zealand region MS prevalence data from the 1980's (37) showed a highly significant correlation with latitude but, using climatic tables (38) , also with low annual temperature (r,=0.883; p