Geographic variation of MS incidence in two prospective studies of US women

Methods.

Results.

In the NHS, 446 participants reported a new diagnosis of MS during the 18-year follow-up period (2,119,277 person-years). Of the 349 participants who we were able to contact and who confirmed their diagnoses, 83% gave us permission to contact their treating physicians. We obtained information from 249 physicians, who confirmed 98% of the self-reports as MS cases. Of these, 211 patients (127 definite, 54 probable, 30 possible) were diagnosed between enrollment in the study and June 1994. There were 129 physician-confirmed MS patients (79 definite, 29 probable, 21 possible) with first symptoms during the follow-up period.

In the NHS II, 324 participants reported a new diagnosis of MS during the 6-year follow-up (649,583 person-years). Of the 214 participants who we were able to contact and who confirmed the diagnosis, 88% gave us permission to contact their treating physicians. We obtained information from 170 physicians, who confirmed 97% of the self-reports as MS patients. Of these, 150 patients (102 definite, 29 probable, 19 possible) were diagnosed between enrollment in the study and June 1995. There were 94 physician-confirmed MS patients (62 definite, 18 probable, 14 possible) with first symptoms during the follow-up period.

Age-specific incidence rates of definite and probable MS for the NHS and the NHS II combined are shown in figures 1 and 2⇓ according to date of diagnosis and date of first symptoms respectively. The highest incidence of diagnosis occurred among women in their 30s and 40s, whereas the incidence of first symptoms peaked before age 30. Based on these rates, the expected lifetime risk of having a diagnosis of definite or probable MS would be 4.89 per 1,000 women.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Incidence of MS diagnosis in the Nurses’ Health Study (NHS) and the Nurses’ Health Study II (NHS II).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Incidence of MS clinical onset in the Nurses’ Health Study (NHS) and the Nurses’ Health Study II (NHS II).

The percent of women with a Scandinavian ancestry was highest in the northern tier at birth, and in the southern tier at age 30 (table 1). The percent of Scandinavians was lowest, and that of southern Europeans was greatest, in the eastern zone at all ages, whereas the percent of nonwhites was greater in the Pacific and mountain zones at all ages.

In the NHS, incidence of MS appeared to increase with latitude of residence at age 15. The rate in the north was 3.5 times greater than in the south, after adjustment for age, ancestry, and longitude (table 2). In the NHS II, the rate in the north was not greater than in the south. The p value of a Wald test for the heterogeneity of the study-specific RRs was 0.05. The incidence of MS was not greater for those residing in the north at age 30. (The age-adjusted RR for the north compared with the south was 0.99 in the NHS and 0.95 in the NHS II). We could not study whether the north–south gradient was constant across longitude due to the small numbers of individuals in zones other than the eastern zone. When using date of first symptoms, the relative rates were qualitatively similar, but with broader CIs. A suggestion of a lower risk of MS was found among women born in the Pacific zone compared with the eastern zone: The multivariate RR was 0.7 (95% CI, 0.2, 2.3) in the NHS and was 0.3 (95% CI, 0.1, 1.2) in the NHS II, but these results are based on small numbers and are not significant (pooled RR, 0.5; 95% CI, 0.2, 1.2). After adjusting for latitude, the association between risk of MS and residing in the Pacific zone at age 15 was greatly attenuated.

When we restricted the analyses to NHS women born in the period 1920 to 1934, the age-adjusted relative rate for the northern tier versus the middle tier was 1.9. There were no patients in the southern tier (age-adjusted RR for north versus middle and south combined was 2.1). Among NHS women born from 1935 to 1946, the age-adjusted RRs were 1.3 for the middle and 1.5 for the north, compared with the south.

The relative rates of MS diagnosis by ancestry are shown in table 3. Although in the NHS women of southern European ancestry and other whites had a similar risk of being diagnosed, in the NHS II southern Europeans had the highest risk. Results for Scandinavians, although not significant, suggest an increased risk for Scandinavians compared with all others in the NHS, and a risk similar to other whites in the NHS II. The p value of the test of heterogeneity was 0.05. Note that the results for Scandinavians in the NHS II are based on a small number of participants. Relative risks for nonwhites are presented for completeness only and are not discussed further because the small sample size and the heterogeneity of the group (including a greater proportion of Hispanics, with mixed southern European descent, in NHS II) make difficult the interpretation of the estimates. The relative rates of MS first symptoms by ancestry are shown in table 4. Compared with the results in table 3, there is a marked increase in risk for Scandinavians.

Discussion.

In these large prospective studies among US women, we found that the incidence of MS increased with latitude, as reported previously,1-7 but the north–south gradient diminished over time. Women with Scandinavian ancestry had a higher risk of MS in the earlier cohort but not in the later cohort, in which the highest risk was associated with southern European ancestry.

The incidence of MS has been reported from several surveys conducted in North America,13-18 but only one3 reported age-specific incidence rates of MS in the United States. The risk of MS in the NHS and the NHS II is of similar magnitude to the one derived from this study (3.9 per 1,000 women).

Because we did not include the self-reported MS diagnoses that we were unable to confirm in the rate calculations (some participants denied us permission to contact their treating physician, others could not be contacted, and others may have not even reported their diagnosis to us), and because we assumed no cases occurred by age 25, the reported risk of MS in the NHS and the NHS II is probably an underestimate. On the other hand, this risk might have been overestimated if MS was overdiagnosed. Overdiagnosis seems unlikely, however, because the majority of the treating physicians were certified neurologists (89% NHS, 94% NHS II), and the proportion of definite and probable cases of MS (86% NHS, 87% NHS II) were similar to those reported in the literature.19-22 Furthermore, none of the MS diagnoses made by the treating physicians was rejected by the study neurologists who reviewed medical records of a sample of patients, and the diagnoses made by the treating physicians were also consistent with the criteria of Poser et al.11 applied to signs and symptoms reported in the questionnaire. Substantial agreement between observers in classifying MS according to these criteria has been reported before.23 If women with undiagnosed MS had been more likely to join the study, we would have expected to find a greater incidence of MS diagnosis during the first years of follow-up, which did not occur (data not shown).

The NHS studies cover a period during which MRI was used increasingly as a diagnostic aid. Although the diagnosis of MS remains largely clinical,11 MRI is the most sensitive test for demonstrating lesions that are not clinically detectable to satisfy the criterion of dissemination in space.24 Also, MRI is the best test for predicting a future diagnosis of clinically definite MS.25 Thus, generalized use of MRI would identify a larger proportion of patients in the earliest or mildest stages of the disease. Previously published incidence rates in European populations cover periods before the generalized use of MRI, which might contribute to their somewhat lower magnitude. This is also consistent with the increased incidence over time that we found in the NHS studies (data not shown).

Another intriguing possibility is that nurses might actually have a higher incidence of MS than other women. An increased risk among nurses was reported in the Key West cluster of MS,26 and in a survey in northeastern Scotland in 1973.27 However, no increased risk among nurses was detected in a subsequent survey of the same Scottish area in 1980,28 nor was MS listed as a cause of death more frequently than expected among British nurses and nursing administrators.29 Our findings do not support an important excess risk of MS among NHS participants compared with US women in 1970 to 1975.3 The slight elevation we found for MS diagnosis could be explained by the greater proportion of white women, and of women living in the north, along with an increased awareness of the disease and easier access to health care among nurses, so that even very mild MS cases would be detected. Higher MS rates have been found in two large epidemiologic studies of married British women, with duration of follow-up comparable with the NHS. In the Oxford/Family Planning Association study,30 an expected lifetime risk of 5.4 per 1,000 women (including possible cases) was estimated among 17,032 women attending family planning clinics between 1968 and 1974, and followed through 1991.31 In the Royal College of General Practitioners’ Oral Contraception Study, an expected lifetime risk of 6.1 per 1,000 was estimated among a cohort of 46,000 women followed between 1968 and 1996.

Kurtzke et al.1 found a north–south relative risk of 3.5 among white women (and 2.6 among white men) who served in the army during World War II or during the Korean conflict. In a national US survey, the relative incidence of MS during the period 1970 to 1975 was 2.7 for the northern tier relative to the south, both sexes combined (3.1 for women if the same proportions used in Kurtzke’s study apply).3 In the NHS, the incidence of MS was also 3.5 times greater among women residing in the northern tier by age 15 compared with those in the southern tier, after adjustment for age, ancestry, and longitude zone. The north–south difference was greater for women born from 1920 through 1934 than for those born from 1935 to 1946. For NHS II women, born between 1947 and 1964, no increase in the risk of MS was detected among those living in the north. Although borderline significant, these differences between cohorts are consistent with an attenuation over time of the north–south gradient of MS risk among women. If confirmed, this attenuation may provide an important clue to the etiology of MS because it suggests that the environmental factors that determined the north–south gradient of MS risk have also changed in the past few decades. This relatively rapid change is consistent with theories that involve infectious agents (e.g., viruses) in the etiology of MS, and suggests that genetic factors may have had a secondary role in the latitude gradient of the disease.

Since Davenport proposed that Scandinavians have a higher risk of MS and southern Europeans a lower risk than other whites living in North America,32 others have looked into the same issue. An ecologic analysis found a positive association between MS and percent of population reporting Scandinavian ancestry in the state of residence of more than 5,000 US war veterans.33 However, this association was not confirmed when the authors used an individual surname-derived ethnicity. Instead, individuals of southern European ethnicity were found to have a 1.25-fold higher risk of MS than those of other ethnicity.34 Researchers in Alberta, Canada, also found a positive correlation between single Scandinavian ancestry and MS prevalence at the census division level, but again this correlation was not confirmed at the individual level.35 Incidence rates of MS reported from Scandinavian countries36-39 are of similar or lower magnitude than those from the United Kingdom.19-22,40

Our results suggest that US women of southern European ancestry do not have a significantly lower MS risk than other white women. In fact, in the NHS II, southern Europeans have the highest risk. The discrepancy between the two cohorts regarding the relative risk of Scandinavian women may simply reflect the uncertainty associated with the small number of participants of Scandinavian descent in the NHS II. Nonetheless, some methodologic differences between the two studies may also account for part of the difference.

A bias could exist if MS had an earlier onset among southern Europeans compared with the other white groups. Because women in the NHS were older at baseline than those in the NHS II, earlier cases might have been missed more frequently. The net result would be an apparent increased risk for Scandinavians. This is consistent with the observed drop in risk for southern Europeans when only patients with first symptoms during the follow-up are included in the analyses (see table 3).

Another factor that might contribute to the discrepancy is that NHS participants were asked to report their ancestry after most of the MS patients had been already diagnosed, and their responses may therefore be affected by the widespread belief that Scandinavians have a greater risk of MS. Interestingly, the response rate of MS patients to this question is higher than that of non-MS patients (98% versus 83%). The results would be biased if women of Scandinavian descent were more aware of their ancestry, or more likely to report it, after an MS diagnosis. On the other hand, ancestry data from the NHS II were obtained before any MS diagnosis was made, so the results are not subject to this potential bias. Self-reports on ancestry may have some degree of misclassification, but is probably not differential between cohorts. In any case, the NHS II appears to have a more appropriate design to study the association between risk of MS and ancestry. Unfortunately, it contains too few women of Scandinavian descent to permit a precise estimation of their risk.