months in New Orleans, La., I have not seen a single case of multiple sclerosis in a native-born individual, either clinically or at autopsy. Clinically, I have seen other diseases of the central nervous system but only one case of

multiple sclerosis. This was a former patient of my own, now resident in New Orleans, who lived in Germany up to 3 years ago. My short experience would be valueless without the corroborating reports from prominent physicians who have been resident here for a long time. All agree that in New Orleans multiple sclerosis is a very rare finding. One ophthalmologist told me that retrobulbar neuritis is very rare here also. Two pathologists who have performed autopsies here for the past 6 years have never seen a patient with multiple sclerosis on the autopsy table." Steiner reported that he found no scarcity of patients in New York City. Because of the great influence that Steiner has had on the thinking in this field, his conclusions on geographic distribution are presented in detail: "In the northern states of Europe multiple sclerosis is frequent; it is rare in Italy, in Spain (according to my pupil A. Borreguero), on the Mediterranean coast of France, and in Rumania (according to the observation of Marinesco). It is also rare in South Africa (according to Russell Brain, personal communication of Dr. Kooy), and in Brazil.

In China, Woods found an incidence two-fifths of that in the United States (1929). Dr. Beringer, the neuropsychiatrist of the German-Russian syphilis expedition, found that in the Burjato-Mongol Republic of Soviet Russia multiple sclerosis is not uncommon. In Japan, the disease is exceedingly rare. It can be stated that in America the disease is very infrequent in the countries below the 35° parallel, in Europe below the 40° parallel. This difference between northern and southern regions cannot be due to a different distribution of races, but some clues to etiology may be hidden in it." The evidence which was available to Steiner, other than the few personal observations noted, was the same as that described earlier in this paper.

Comment. It is obvious that conclusions relating to the distribution of multiple sclerosis over wide latitudes or in different countries have generally failed to recognize that the reported uneven distribution may be due to the lack of uniformity in diagnostic criteria at different times and in different places; in variations of the method and completeness of case finding, and, in developing rates, in the failure to relate cases to the appropriate population.

Despite the inadequacy of the evidence, these statements and reports have influenced the treatment of patients and views on etiology. Putnam (30) has gone so far as to state that not only are high temperatures and low humidity associated with low prevalence of multiple sclerosis, but also that the principal treatment of the disease is the migration of patients to warm, dry climates. The reported variations in geographic distribution have been used to support hypotheses of causation such as insect-borne infection (29), nutritional deficiency or excess (31), bacterial allergy, and vasoconstriction due to cold (32). Shields (33) has even contended that since multiple sclerosis is "rare in China" where organic waste matter is used to fertilize the soil, and is "common in Germany and Scandinavia" where manufactured fertilizer is used, multiple sclerosis is a deficiency disease due to "incomplete soil fertilization."

The review of the literature on morbidity studies of multiple sclerosis leaves the impression that there has been no valid evidence of a variation in geographic distribution of the disease. A large variety of studies have been conducted but in the analysis of the results little consideration has been given to the effect of time on the diagnosis of multiple sclerosis, to differences in the methodology employed, or to the comparability of medical facilities available in different communities.

SECTION II. THE ANALYSIS OF MULTIPLE SCLEROSIS MORTALITY RATES IN THE UNITED STATES AND CANADA, 1947, WITH SPECIAL REFERENCE TO GEOGRAPHIC

DISTRIBUTION

A. Evaluation of errors of mortality rates

One approach to the problem of multiple sclerosis frequency and geographic distribution is through the review of mortality statistics. After the Fifth Revision of the International List of Causes of Death went into effect in 1939, multiple sclerosis deaths were tabulated for the first time in a separate category (under rubric 87d "Disseminated sclerosis"), and routinely published data or special tables on deaths could be obtained with relative ease.

The first published report on these mortality statistics was presented in 1948 by Limburg (34), who compared multiple sclerosis crude death rates for various parts of the world. He reported that higher rates were found in the countries

more distant from the Equator (table 1). He also calculated death rates in Canada by Provinces and in the United States by States, finding that the Provinces and Northern States had higher rates than did the Southern States (map I). On the basis of the crude death rates, he concluded that "mortality from multiple sclerosis varies with the mean annual temperature of the reporting area; the colder the climate, the higher the crude death rates for multiple sclerosis." Limburg's inference did not take into consideration differences in the accuracy, completeness, and classification of mortality statistics, nor in the age or racial composition of the populations compared.

TABLE 1.-Crude rates per 100,000 population for deaths classified as multiple sclerosis for selected countries and specified years

Source: Limburg, C. C.: "The geographical distribution of multiple sclerosis and its prevalence in the United States," the Association for Research in Nervous and Mental Diseases, XXVIII: 15-24, 1950.

A valuable point of departure for intensive clinical and laboratory investigations would be reached if it could be demonstrated that multiple sclerosis frequency varied consistently with some geographic or climatic feature. Furthermore, even though mortality statistics may be subject to a variety of errors, they provide the only available extensive data on geographic distribution of multiple sclerosis and should be utilized, when possible, as an index of prevalence. For multiple sclerosis there may be some question whether the mortality rates afford a valid index of the relative prevalence in different areas. Theoretically, an excess in mortality of a disease in area A as compared with area B might be due to a higher prevalence, a more severe form of the disease or a combination of the two. Thus, it is not possible from a difference in mortality alone to state that prevalence of the disease is necessarily also higher, unless there is also evidence that the proportion of cases to deaths is parallel in the two areas. As yet there is no positive evidence that severity or duration of multiple sclerosis differs by region, although the possibility of such differences cannot be ignored.

Before comparing mortality rates for different groups or for different places, it is desirable to study the possible sources of error in the statistics and to make corrections wherever feasible. Several questions that relate to mortality statistics in general and to the study of multiple sclerosis in particular should be considered: (1) To what extent are diagnostic errors present in death certificates reporting multiple sclerosis? Do the diagnostic errors vary so much geographically that comparisons of specific mortality rates may be misleading?

(2) To what extent is multiple sclerosis as a cause of death deliberately concealed?

(3) Are clerical or tabulating errors of unusual consequence?

(4) Are there errors of terminology or classification in this category which affect comparisons of the rates in different geographic areas?

The first three questions can be considered briefly. The last question was the subject of a special investigation and will be dealt with in greater detail. Following the study of errors of terminology and classification, a comparison of the geographic distribution of multiple sclerosis death rates, corrected for sex, race, and age differences will be made.

1. Diagnostic error.-The reliability of the diagnosis of multiple sclerosis at time of death is not known. However, because multiple sclerosis is a chronic, generally progressive disease, it is likely that the diagnosis will be more correct in the later than in the carlier stages. The diagnostic error in a selected group of hospitalized autopsied cases in northeastern United States was described in the introduction and was found to be large. Whether the size of the error varies in different areas within the United States and Canada is unknown.

The availability of specialist consultation and adequate diagnostic facilities depends so largely upon local resources that, for this reason alone, comparisons of crude rates for the variety of countries as in Table I must be interpreted with caution. At least comparison should be limited to countries with similar medical resources, such as the United States and Canada. In these two countries, the similarity of language, medical training, terminology and practice, plus the migration of many teachers and practitioners from one area to another provide some assurance of comparability in the diagnostic criteria and skills of physicians. However, variations of medical facilities as well as of the diagnostic ability and the accuracy of reporting by physicians in different parts of these countries must remain an imponderable source of error in comparisons.

2. Errors in reporting.-Errors may arise from the misstatement of a known diagnosis to protect the family of the deceased from stigma thought to be attached to certain diseases or to aid the survivors in securing certain death benefits. The extent of errors from this source for multiple sclerosis cannot be determined but they are probably negligible.

3. Clerical errors.-Clerical errors can be controlled by a strict verification procedure. This is done in the National Office of Vital Statistics and in the Canadian Bureau of Vital Statistics and such errors, too, are probably negligible. 4. Errors of terminology and classification.—A special study was made to determine whether the mortality statistics in the United States and Canada might have been affected by the inclusion in the 87d cateogry of diseases other than multiple sclerosis. It was believed that a certain proportion of "disseminated sclerosis" death certificates listed such causes of death as "cerebral sclerosis," "general sclerosis" and "generalized sclerosis" and did not refer to multiple sclerosis at all. This sort of misclassification had been suspected in the course of surveys made in New Orleans, Winnipeg and other cities which involved the study of rubric 87d death certificates for the years 1939 to 1948.

83289-51-4

Each of the rubric 87d death certificates in the United States and Canada was examined to determine whether terms other than multiple or disseminated sclerosis on the death certificates referred to other causes of death. This study was made using 1947 death certificates for both the United States and Canada, together with certain additional data.

(a) Classification under rubric 87d.-Certain discrepancies were noted in local multiple sclerosis mortality studies. It was observed that the average age at death and the contributory causes of death appearing on certificates reporting "multiple sclerosis" or "disseminated sclerosis" as the cause of death were markedly different from those reporting "cerebral sclerosis," "general sclerosis," or "generalized sclerosis." Furthermore, in the local studies described in section III, multiple sclerosis morbidity reports were obtained from hospitals and physicians on individuals whose deaths were reported as due to "multiple sclerosis" or "disseminated sclerosis" but were lacking where the death was certified under one of the other terms.

The decision by registrars of vital statistics to include certain synonyms in a particular category is usually based on the accepted definition of the term. The American Illustrated Medical Dictionary defines cerebral sclerosis as "multiple sclerosis of the brain," although the same term when qualified by "MerzbacherPelizaeus," "Scholz" or "Krabbe" signifies one of the rare cerebral heredofamilial degenerative conditions of children which are not considered synonymous with multiple sclerosis.

However, it has been the experience of the National Office of Vital Statistics that physicians generally use the term "cerebral sclerosis" in the sense of cerebral arteriosclerosis, and State and Provincial offices of vital statistics have been requested to query physicians on their use of the term and reclassify deaths where the physician's reply indicates misuse. These offices, however, make such queries infrequently. The National Office of Vital Statistics, on reviewing cerebral sclerosis death certificates from the States, attempts to do so but has been able to query only a fraction of the physicians using the term and has not received replies from many of those queried.

Thus, in the past the majority of reports listing these synonyms have been classified as multiple sclerosis, since those not queried or those for which no reply was received remained in rubric 87d.

(b) Method of study. All 1947 death certificates classified in rubric 87d as the primary cause of death were reviewed, 1,583 from the United States and 186 from Canada. About 14 percent of all certificates containing a cause classifiable under rubric 87d are coded to other rubrics (35); but as those certificates could not be obtained, the study is limited to those certificates in which multiple sclerosis was coded as the primary cause of death.

It will be seen from table 2 that the distribution of deaths assigned to the disseminated sclerosis category by reported cause of death is about the same for the United States and Canada. About 82 percent of the certificates listed multiple or disseminated sclerosis; the remaining certificates reported cerebral, general, or generalized sclerosis.

To simplify matters in this study, all medical certificate listings of multiple or disseminated sclerosis will be referred to as "multiple sclerosis. All listings of cerebral, general, or generalized sclerosis will be referred to as "cerebral sclerosis." Comparisons of duration of illness and the associated or contributory causes of death were made between the deaths reported in the United States as multiple sclerosis and those reported as cerebral sclerosis. A comparison of the age at time of death for cerebral sclerosis and multiple sclerosis deaths was made for both the United States and Canada as this information was available for both countries. In addition, a number of physicians who had signed cerebral sclerosis death certificates were queried as to the meaning of the term.

(c) Results of study.-(1) Duration of illness: In the multiple sclerosis group, duration of illness prior to death was reported in 722 or 55.6 percent of the certificates, whereas it was given in only 80 or 28.2 percent of the cerebral sclerosis group. Based upon certificates which gave this information, the mean duration of illness when cause of death was listed as multiple sclerosis was 8.4 years, while the mean duration for cerebral sclerosis was found to be 3.6 years. (See table 3.) (2) Associated causes of death: Prior to 1949, the selection of the cause of death to be tabulated when more than one cause was reported was made by reference to the Manual of Joint Causes of Death (36). This manual gave the priority relationship between diseases on the basis of which the primary cause to be tabulated was selected.

In this study, it was found that an associated cause of death was listed for 801 or 61.7 percent of the multiple sclerosis death certificates and for 251 or 88.4 percent of the cerebral sclerosis certificates. The distributions of these causes in the two groups are very different (table 4). Though several of the causes appear in both lists, the order and percentages differ. In fact, three of the nine most frequent causes are not common to the two lists.

TABLE 2.-Reported causes of death on certificates classified as multiple sclerosis (rubric 87d), United States and Canada, 1947

TABLE 3. Duration of disease prior to death for deaths classified as multiple sclerosis (rubric 87d), based upon certificates in which a duration is stated, United States, 1947

It is not possible to compare associated causes of cerebral arteriosclerosis deaths with those of cerebral sclerosis because cerebral arteriosclerosis is not specifically coded or tabulated. The cerebral sclerosis deaths were classified under rubric 87d, and the rules for determining the primary cause, when more than one cause is listed on the death certificate, differ for the 87d (disseminated sclerosis) category and the 97 (arteriosclerosis) category.

However, it is possible to get an idea of the associated causes of death for arteriosclerosis by referring to the special study carried out by the Bureau of Vital Statistics on 1940 death certificate transcripts (35). The causes of death with which arteriosclerosis was associated most frequently (as a primary or as a secondary cause of death) were found in the following order: cerebral hemorrhage, diseases of the coronary arteries, cerebral embolism and thrombosis, senility, bronchopneumonia, chronic affections of heart valves and endocardium. Although this list differs from both the multiple sclerosis and cerebral sclerosis lists, it resembles the latter much more closely.

(3) Mean age at death: It may be seen from table 5 that in the United States groups, the mean age at death for multiple sclerosis is 53.9; for cerebral sclerosis, 72.5; and for arteriosclerosis, 79.0 years. The values for multiple sclerosis and cerebral sclerosis in the Canadian groups differ very little from the corresponding groups for the United States. The marked difference in the distribution of ages at death for cerebral sclerosis and multiple sclerosis indicates that these are two distinctly different groups.