Materials and methods

The National Cancer Institute, the National Institute for Occupational Safety and Health, and the National Center for Health Statistics maintain a database of all deaths in 24 states, which coded occupation, state of residence at birth and at death, and other information from death certificates, as described by Figgset al.16 Cases included all underlying causes of deaths between 1984 and 1995 from multiple sclerosis (international classification of diseases, ninth revision (ICD-9), code 340) and non-melanoma skin cancer (code 173). A common set of controls was used across a series of case-control studies on neurological and cancer mortality (including breast, prostate, ovarian, and colon cancer) and solar radiation. Deaths from cancer (ICD 140–239), multiple sclerosis (ICD 340), and some diseases of the central nervous system (ICD 330–337), were excluded. Controls were frequency matched by five year age group to the combined group of multiple sclerosis, non-melanoma skin cancer, and other causes of death of the cases in the series. The controls represent a one to one ratio with the most common cause of death in the series (colon cancer), but a ratio of about 25 to one with multiple sclerosis and with skin cancer. The data were also analysed with an alternative subset of controls, excluding those with deaths due to heart disease or accidents, causes which may be related to residential region.17 18

Residential exposure to sunlight was assessed by state residence recorded on the death certificate. We assigned each state one of three levels of solar radiation based on United States Weather Bureau data.19 Subjects were limited to those who resided in the same solar radiation region at birth and at death to reduce misclassification in analyses based on lifetime residential solar exposure. This included 4282 cases of multiple sclerosis, 4929 skin cancer cases, and 115 195 controls (about 75% of each).

Occupational exposure to sunlight was based on usual occupation classified by an industrial hygienist (MD) into three categories: indoor work, work that combined indoor and outdoor exposure, and outdoor work. Occupation was also used to assess socioeconomic status based on a method of scoring occupation developed by Green20 and to create an index of occupational physical activity with four levels of physical activity. The effects of occupational exposure to solar radiation were assessed among all subjects and among the subjects in occupations categorised as requiring high or moderate levels of activity, so as to exclude those whose indoor jobs resulted from disabilities subsequent to onset of the disease. This resulted in 1562 (27%) cases of multiple sclerosis, 2849 (43%) cases of skin cancer, and 59 690 (39%) controls.

The effects of ethnic origin and race were assessed by classifying subjects based on national origin and race as recorded on the death certificate. Northern European ancestry in particular was assessed because of its possible role in the higher risk previously found with higher latitudes in the US.

We used multivariate models of potential exposure to solar radiation, adjusting for age, sex, race, and socioeconomic status. The models were applied to the entire population; sex and race subpopulations (adjusting only for age and socioeconomic status); as well as subpopulations of residents of eastern and western states (to examine whether possible ethnic differences between eastern and western regions accounted for any observed patterns).

The measures of association were the mortality odds ratio (OR) and 95% confidence interval (95% CI) derived by standard logistic regression methods in SAS.21

Results

Table 1 provides the characteristics of cases and controls for deaths from multiple sclerosis and non-melanoma skin cancer, as well as for those with the same residential region at birth and at death, and the group with active occupations. Of those with active occupations, about 10% of cases of multiple sclerosis had outdoor jobs.

Residence in states with the highest potential exposure to sunlight was associated with an adjusted OR for multiple sclerosis of 0.53 (table2). By contrast, the adjusted OR for skin cancer and residential exposure to solar radiation increased significantly to 1.24 in the highest exposure category. When the country was stratified into eastern and western regions, the risk of mortality from multiple sclerosis declined with increasing residential solar radiation in each region, with the east showing a stronger negative gradient (OR=0.28 (east) and OR=0.63 (west) for highest exposure.) The negative gradient was also found in the subpopulations of white men, white women, black men, and black women (adjusted OR=0.57, 0.49, 0.44, 0.59 respectively for highest exposure). Among white people, family origin from a northern European country was not associated with risk of mortality from multiple sclerosis, after controlling for residence, age, sex, and socioeconomic status (OR=1.0).

Occupational solar radiation (among active jobs) was associated with reduced risk of mortality from multiple sclerosis (table 2). This association was generally found in racial and sex subpopulations, although the numbers of cases were especially small for minorities (data not shown). By contrast, occupational solar radiation was positively associated with risk of skin cancer in the total population, and generally among sex and race subpopulations (data not shown). We also found that the residential and occupational solar radiation gradient in risk remained in analyses with the alternative subset of controls.

Table 3 shows the combined effect of residential and occupational solar exposure among those in active jobs. With the referent category comprised of indoor workers in residential areas receiving least solar radiation, increased potential exposure to sunlight was generally associated with reduced risk of multiple sclerosis in each residential and occupational category. In the highest combined (occupational and residential) sunlight category, the OR was 0.24 (95 % CI 0.15 to 0.38). By contrast, the risks for skin cancer increased with potential occupational exposure to sunlight in each residential category, and with residential exposure among indoor workers, although not consistently among outdoor workers.

Discussion

By contrast with the mortality from skin cancer, our study found that both residential and occupational solar radiation were inversely associated with mortality from multiple sclerosis, and that the associations generally characterised sex and race subpopulations. Ethnicity does not seem to explain the geographical gradient because northern European ancestry was not a risk factor and a negative gradient was found in both eastern and western regions of the country, although it was weaker in the west. By assessing the effects of potential occupational exposure to solar radiation only among those with the most active jobs, we also increased the likelihood that the inverse association reflects causal factors, rather than occupational changes after the onset of disease.

Death certificate studies, such as this, have recognised limitations, including potential misclassification of occupation and residential exposure, and lack of information on potential confounders—such as viral infections, environmental childhood exposures, genetics, and other sources of exposure to sunlight.22 As noted, residence was based simply on place of birth and residence at death, whereas occupational exposure was limited to usual occupation. If the misclassification of cases does not differ from those of controls, any error in the association will at most reduce the association found between exposure and mortality.

One of the most problematic limitations is the potential bias in use of deaths rather than population as the study base, which may result in biased associations, unless the distribution of exposure of the dead controls reflects that of the living source or base population from which the cases came.23 Although the similarity of our findings with an alternative subset of controls supports the findings, uncertainties remain about whether the dead controls reflect the underlying exposure profile of the base population.

Despite these limitations, our study confirmed many of the findings previously reported. The validity of the solar radiation indices was supported by the positive association we found between the exposure to solar radiation and mortality from non-melanoma skin cancer, which is consistent with previous studies linking incidence of skin cancer to cumulative exposure to sun.15 24 Also, although socioeconomic status was based on usual occupation, and not on other possible indicators, the positive relation we found between socioeconomic status and risk of multiple sclerosis is consistent with other studies.25 26 Our finding that northern European ancestry is not related to risk, after controlling for residence, is consistent with the results of Page et al 27 in another of the veteran series studies.

Several mechanisms have been presented about how sunlight may affect multiple sclerosis.10 11 13 Hayes et al 11 proposed that the role of sunlight on immune function may be mediated by hormonal vitamin D. Hutter and Laing13 suggested that the effect of sunlight may be mediated by the secretion of melatonin, which is both sensitive to light and may affect the cellular structure of the thymus. McMichael and Hall10 also suggested that sunlight may suppress immune functions, either by reducing early T cell sensitisation to antigens that mimic myelin or subsequent cell mediated responses to these antigens.

As our study was based on death certificates, our findings are necessarily exploratory. The negative associations between mortality from multiple sclerosis and both residential and occupational exposure to sunlight found here warrant future study based on incident cases and more refined assessments of exposure to sunlight.