ABSTRACT--In a study of 261 male melanoma patients and age- and sex-matched controls, a strong positive univariate association between socioeconomic status, as determined by usual occupation, and risk of melanoma was detected. This association, however, was substantially explained by host constitutional factors and occupational, recreational, and vacation sunlight exposure. The study demonstrated an increased risk of melanoma in draftsmen and surveyors and a reduced risk of melanoma in construction workers and individuals employed in the finance, insurance, and real estate industry even after control for the effect of host factors and sunlight exposure.--JNCI 1987; 79:647-652.
Data from several descriptive and analytic studies of cutaneous malignant melanoma have shown an elevated risk for this tumor in individuals in upper socioeconomic occupations (1, 2). These upper socioeconomic occupations involve primarily indoor work, and there is some evidence to suggest that there is an inverse association in males between constant occupational sunlight exposure and risk of melanoma (3). This suggests that the socioeconomic gradient seen in melanoma risk may be largely due to sunlight exposure factors rather than to occupation.
A number of studies have suggested elevated risks for melanoma in chemists (4-6), telecommunication and electronic workers (7, 8), and a variety of other occupations as reported in the excellent review of Austin and Reynolds (9), but many studies have not been able to control for host pigmentation variables and sunlight exposure.
Further, an Australian study using the job-exposure matrix of Hoar et al. (10) detected an elevated risk of LMM in individuals exposed occupationally to ionizing radiation (11). Results with nodular melanoma were the reverse of those found for LMM and were not elevated for superficial spreading melanoma or unclassified melanoma.
We have conducted a case-control study of malignant melanoma to investigate its etiology and in this paper report on our results for socioeconomic gradient and the interaction of this variable with host factors and sunlight exposure. Occupational and industrial associations with melanoma are also discussed.
SUBJECTS AND METHODS
The Western Canada Melanoma Study identified all newly diagnosed, histologically confirmed cases of malignant melanoma in British Columbia, Alberta, Saskatchewan, and Manitoba from April 1, 1979, to March 31, 1981, as described in our previous reports (12, 13). Patients (904) with primary cutaneous lesions were ascertained; 801 of these patients were age eligible (20-79 yr) for interview and inclusion in the study. Of these, 665 patients (83%) were interviewed along with matched controls. The interviewed cases consisted of 56 LMMs, 415 superficial spreading, 128 nodular, 23 unclassified, and 29 borderline melanomas. An additional 14 patients of both sexes with acral lentiginous melanomas were interviewed but were not included in this report. Of the male cases, 159 were superficial spreading, 62 were nodular, 4 were unclassified, 28 were LMM, and 8 were borderline; this group of patients comprises the caseload described in this analysis.
Controls were selected at random from the provincial medical insurance plans from each province and were matched to cases by age +/-2 years and by sex. In Saskatchewan, Manitoba, and Alberta, where study coordinators were permitted to telephone control subjects, the response rate was 59%. In British Columbia a letter requesting participation was sent to each subject by the Medical Services Commission, and individuals were required to send back a completed consent form; no prior telephone contact was permitted. This procedure resulted in a control response rate of 48% in British Columbia.
Patients and controls were interviewed in their homes by trained interviewers by means of a standard questionnaire. Information was obtained on every job the respondent had held for more than 6 months. Details collected and coded included occupational title and industry, as well as duration of employment. Other variables recorded included skin, hair, and eye color, freckling, skin reaction to sunlight, and occupational, recreational, and vacation sunlight exposure.
Skin and hair colors were determined by the interviewers using direct comparisons with prostheses and wigmakers' samples made up especially for the study. Skin color was assessed twice on a non-sun-exposed area (upper inner aspect of each arm). In the event of hair dyeing or graying with age, subjects were asked to indicate, from the samples, their natural hair color as a child and as a young adult.
Each subject's sun exposure was assessed from responses to questions on occupational, recreational, and vacation activities. Assessment in these three areas was based on the hypothesis that occupational sunlight would represent chronic constant UV, whereas vacation and recreational exposure would be a surrogate for intermittent UV exposure. For each occupation the usual number of hours per week of outdoor work was recorded for both summer and winter seasons. In addition, data were collected on usual clothing preferences for each season while on the job.
Recreational exposure was recorded for each decade of the subject's life, and summer and winter seasons were considered separately. Exposure was assessed on the basis of the subject's activity in 4 broadly defined groups of activities: a) activities in which a bathing suit is normally worn (sunbathing, swimming); b) activities in which light clothing would be worn (baseball, track and field sports, camping, gardening in warm weather); c) activities in which normal or heavier clothing would be worn (fishing, hiking in moderate weather); and finally d) activities on snow and ice (skiing, mountaineering). Each type of recreation was assessed separately, and the usual number of hours per week and months per year of participation were recorded.
For vacations, subjects were asked about all holidays of 1 week's duration or more during which they spent more time engaged in outdoor activities than usual or were exposed to more intense sunlight than normal. For each vacation, the length of time, season, location of the vacation spot, and usual number of hours spent on activities in each of the 4 recreational groups described above were recorded.
On the basis of these measurements, lifetime whole-body exposure hour scores were compiled for each subject for chronic sunlight exposure (occupation) and intermittent sunlight exposure (recreation and vacation).
Crude associations between melanoma and occupation groups and industry were first examined with the use of matched pairs odds ratios as an estimate of RR. To control for other variables known to be significantly related to melanoma and to examine interaction effects particularly with occupational, recreational, and vacation sunlight exposure, and host pigmentation factors we fitted a conditional logistic model according to the procedures of Breslow and Day (14) using the EPILOG program (Epicentre Software, Pasadena, CA).
Initial examination of the occupation and industry data had revealed that most of the female respondents, both cases and controls, were predominantly employed as housewives. Since spouse's occupation was not available, it was decided to restrict the present analysis to males.
Socioeconomic Gradient and Risk of Melanoma
Because several studies have demonstrated a gradient of risk in melanoma from high in upper socioeconomic status professional workers to low in low-status unskilled laborers, we investigated this gradient with our data set.
Each male subject's occupational history was examined, and the occupation of longest duration was defined as the individual's usual occupation. Occupations were then assigned, by means of the Blishen scale for Canadian occupations (15), to one of five social class or socioeconomically graded groups: professional and scientific (I), management and administrative (II), sales and service (III), skilled manual workers (IV), and unskilled workers including farmers (V).
Evaluating the risk of melanoma in males in these 5 occupational groups on univariate analysis showed a significant trend with higher risk in the higher socioeconomic status groups similar to that detected in previous studies of melanoma (table 1). When host factors were controlled, the trend became less pronounced but remained strong and statistically significant. Occupational sunlight exposure, our surrogate for chronic sunlight exposure, was then added to the model, which resulted in a further weakening in the trend of higher melanoma risk in upper socioeconomic occupational groups. Finally, recreation and vacation sunlight, our surrogate for intermittent sunlight exposure, was added. Addition of the UV factors appeared to substantially explain the univariate association detected between melanoma and socioeconomic gradient (table 1). Only the professional and scientific group maintained a residual elevated risk of melanoma in the final model, and this was insufficient to maintain the significant trend in risk seen on univariate analysis.
Because of concern with the relatively low response rate of the controls in our study, we compared the occupational distribution of the most recent job held by control subjects 19-64 years of age with the distribution of current occupations in the active labor force of the four western Canadian provinces as recorded in the 1981 Census of Canada (16). The results of the comparison are shown in table 2 and indicate that the distributions are quite similar.
Occupation and Industry
Univariate analysis suggested that several occupations were significantly related to melanoma when evaluated on an "ever employed" versus "never employed" basis (table 3). Where numbers of subjects enabled us to divide groups by time employed in an occupation, this was done. Significantly elevated risks were seen for surveyors and draftsmen, and as well, a suggestion of an increase was seen in architects and engineers. Significantly reduced risks of melanoma were seen in farmers and woodworkers, with a suggestion of a decreased risk in construction workers. Table 4 demonstrates results for industry of employment with significant associations being seen for men in the agriculture industry and the finance, insurance, and real estate industry, as well as the community and personal service industries. A total of 85 different occupations and 32 industrial groups was examined.
Our previous analyses have demonstrated significant associations between risk of melanoma and host factors, ethnic origin, and indices of individual sunlight exposure (3, 12, 13). We therefore fitted a logistic model to the occupation and industry data, incorporating known significant host factors, including hair color, skin color, freckling, skin reaction to sunlight, and ethnic origin as well as sunlight variables (occupational, recreational, and vacation sunlight) to evaluate whether occupation and industry associations detected on univariate analysis would persist after controlling for the known strong confounding factors.
Adjustment for host factors and sunlight exposure decreased the strength of the associations between melanoma and several of the occupations including architecture, engineering, and woodworking. In particular, the strong inverse association between risk of melanoma and employment as a farmer detected on univariate analysis was weakened when host and sunlight exposure variables were incorporated in the model and became nonsignificant (table 3). The incorporation of these variables in the model, however, strengthened slightly the positive association with surveyors and draftsmen and left unaltered the negative association between melanoma and work in construction.
A similar effect for these adjustments was seen for industries initially found to be related to risk of melanoma on univariate analysis (table 4). The associations with agriculture and for community and personal services industries were weakened but that with the finance insurance, and real estate industry was essentially unaltered
Occupational and Industrial Exposure Matrix
To determine whether we could identify specific carcinogens associated with significant occupations and industries found in the study and to search for other possible associations between chemical exposures and risk of melanoma, we applied the Hoar et al. (10) job exposure matrix to our data after reconciling the U.S. occupation and industry information in the matrix with the Canadian codes used in our study. By measuring the number of years worked in an industry or occupation and weighting by the intensity of exposure, we created a gradient of exposure for each of several groups of substances.
No significant positive associations between melanoma and chemical exposures were detected, and in particular no association was seen with exposure to ionizing radiation. A fairly strong inverse association was seen initially with pesticide exposure and exposure to ethylene glycol; however, exposure to both of these compounds occurred largely in farmers, and the negative association with farming was satisfactorily explained by the primary association with host factors and sunlight exposure.
Melanoma Subtype Analysis
In an earlier analysis we detected similar etiologic patterns in the relationship of superficial spreading and nodular melanomas to host and sunlight exposure variables, but the pattern for LMM appeared to be somewhat different (17). The LMMs were, in general, less strongly related to host pigmentation factors and sun exposure than the other two main melanoma subtypes. Because of this we reanalyzed separately our occupational data using male cases with superficial spreading melanoma, nodular melanoma, and where numbers permitted, LMM. Using a test of heterogeneity, we saw no significant differences in melanoma subtype as compared to all melanomas combined for any occupation or industry. The socioeconomic gradient was likewise compared for superficial spreading and nodular melanoma, and a test of heterogeneity demonstrated no significant difference between the two subtypes. These two subtypes were then combined and compared with the LMMs. No statistically significant differences were found. The numbers of both LMMs and nodular melanomas were small, making it very difficult to detect differences between melanoma subtypes for any of the variables examined in this paper, particularly with relatively insensitive tests of heterogeneity.
The positive gradient of risk between social class or socioeconomic status and melanoma has been noted in a number of descriptive and analytic data sets including mortality and incidence data from England (18), Australia (19), New Zealand (20), and the United States (2).
The New Zealand study, the most definitive descriptive review, was based on occupation as recorded on death registrations and cancer registry records. The authors assigned individual cases to categories of predominantly outdoor, predominantly indoor, and intermediate exposure categories based on independent assessment of sunlight exposure according to job title. Each case was assigned to a socioeconomic group, again on the basis of job title. No differences were seen when standardized mortality or standardized incidence ratios were calculated for outdoor workers versus indoor workers within socioeconomic status levels. However, when socioeconomic status was evaluated within outdoor and indoor exposure categories, individuals in the upper social classes were found to have higher standardized mortality ratios and standardized incidence ratios than those in the lower social classes. The New Zealand analysis was unable to control for important host pigmentary differences between individual subjects, factors known to be strongly related to risk of melanoma; in addition, occupational sunlight exposure had to be implied from job title.
Our analysis of the usual occupation of subjects revealed a gradient of risk from the highest to the lowest Socioeconomic class. When adjusted for skin and hair color, degree of freckling, skin reaction to sunlight, and ethnic origin, the gradient was reduced; when adjusted for both occupational sunlight (our surrogate for long-term chronic UV exposure) and recreational and vacation sunlight (our surrogates for intermittent UV exposure), the gradient was substantially reduced.
There is, however, still some residual socioeconomic effect in professional and scientific workers even after all sun and host adjustments have been made. If, as seems likely, sunlight exposure as recorded in our study is a relatively inaccurate measure of actual exposure, then the true effect of sunlight on the perceived gradient in our study might be more substantial with better data. Further, this inaccuracy might be expected to be most pronounced in men with none of the more easily remembered type of exposure, that is, occupational. This line of reasoning is reinforced by the observation that the perceived socioeconomic gradient is reduced much more by addition of the occupational sunlight measure to the logistic model than by addition of both recreational and vacation sunlight variables combined. More accurate sunlight data might therefore have further reduced the suggestion of socioeconomic effect in the professional and scientific workers in our study.
Because of the low response rate among controls in the Western Canada Melanoma Study, there was some concern that our controls might have been grossly atypical, in terms of occupational distribution, by comparison with the male population of the four western provinces. However, an actual comparison of occupational information from controls with that from the 1981 Census of Canada indicates that our control group occupational profile did not differ materially from that of the population. Thus we are relatively confident that our findings concerning the socioeconomic gradient are not due to a control group with a markedly atypical occupational pattern.
Our findings appear to indicate that the socioeconomic or social class gradient for melanoma may be due to a combination of host pigmentation factors and sunlight exposure. In general the subjects in our study who were of upper socioeconomic status were of northern European origin and tended to have lighter shades of skin and hair coloring. It seems likely that this situation may hold for other countries in which a socioeconomic gradient has been detected, including England, Australia, New Zealand, and the United States.
Results of our study also indicate that UV or sunlight exposure is important in explaining melanoma risks in males; both the protective or negative effect of chronic (occupational) sunlight exposure and the positive effect of strong intermittent (recreational and vacation) sunlight are important factors in melanoma risk.
The subtype analysis for socioeconomic gradient in males was generally unhelpful because of the small numbers of nodular and LMM cases available. The risk ratios for the socioeconomic gradient in nodular melanoma were similar to those for the superficial spreading variety, whereas those for lentigos revealed no clear trend. These findings are in line with subtype differences found for host variables in our study (17).
Descriptive and analytic studies have suggested that a number of occupational groups are at elevated risk of malignant melanoma including chemists, administrators, accountants, and other groups (9, 21). Our study does not confirm previous findings of an excess of melanoma in chemists nor did we detect an elevated risk in any melanoma subtypes in those occupationally exposed to ionizing radiation. However, relatively few subjects were involved in the chemical industry or had major exposure to ionizing radiation.
A highly significant inverse association between farming and risk of melanoma was substantially reduced when pigmentation and sunlight exposure were controlled for.
After allowance was made for the effects of host pigmentation factors, as well as sunlight, only one occupational group, surveyors and draftsmen, had a significant positive association with risk of melanoma. No industries were positively related to melanoma risk, but one industry, that of finance, insurance and real estate, and one occupational group, construction workers, demonstrated a significant inverse association with melanoma.
Of the cases working in the surveyor and draftsmen group, the dominant occupations were architectural technicians (5/17) and surveyors (6/17). It is possible that the technicians are exposed to unusual light sources in the performance of their duties including blueprint copying lights, and such an association has been reported in another study (22). However, the surveyors likely would not have been exposed to such sources.
Why construction workers and individuals in finance, insurance, or real estate should be at reduced risk of melanoma after controlling for host variables and sunlight exposure is open to question; of course, there is always the possibility that with multiple testing of occupations and industries in this study, the associations arose by chance. Finally, it is possible that even the occupational associations remaining after control of potential confounding factors may be substantially altered if more accurate sunlight exposure data than those used in our study were obtainable.
In summary, future occupational analyses of malignant melanoma in males should be controlled for the effect of known risk factors. Further work needs to be done to determine if control for sunlight exposure influences the socioeconomic gradient for melanoma in other countries.
1 Received December 8, 1986; revised May 14, 1987; accepted May 28, 1987.
2 Supported by Health and Welfare Canada (6610-1203-53), the National Cancer Institute of Canada, and the Alberta Heritage Trust Fund.
3 This paper is presented on behalf of the Western Canada Melanoma Study. Participants include--Coordinators: M. Grace,[a] S. Kemel,[b] H. Colls (deceased), C. Leinweber,[a] Diane Robson,[c] J. Moody,[d] and M. Beagrie.[d] Pathologists: A. J. Worth[d] and W. S. Wood.[d] Consultants: M. L. Jerry,[a] D. I. McLean,[d] P. Rebbeck,[d] and H. K. B. Silver.[d] Secretaries: Shirley Morton[d] and Karen Anderson.[d]
4 Division of Epidemiology, Biometry, and Occupational Oncology, Cancer Control Agency of British Columbia, 600 West 10th Ave., Vancouver, BC V5Z 4E6.
5 Department of Community Health, University of Nottingham, Nottingham, England NG7 2UH.
6 Department of Epidemiology, Alberta Cancer Board, Edmonton, AB.
7 We are indebted to the referring physicians, interviewers, and subjects for their time and effort.
a Alberta Cancer Board, Edmonton, AB
b Manitoba Cancer Treatment and Research Foundation, Winnipeg, MB
c Saskatchewan Cancer Commission, Regina, SK
d Cancer Control Agency of British Columbia, Vancouver, BC
ABBREVIATIONS USED: LMM=lentigo maligna melanoma; RR=relative risk.
(1) Lee JA. Melanoma. In: Schottenfeld D, Fraumeni JF Jr, eds. Cancer epidemiology and prevention. Philadelphia: Saunders, 1982:984-995.
(2) Graham S, Marshall J, Haughey B, et al. An inquiry into the epidemiology of melanoma. Am J Epidemiol 1985; 122:606-619.
(3) Elwood JM, Gallagher RP, Hill GB, et al. Cutaneous melanoma in relation to intermittent and constant sun exposure--the Western Canada Melanoma Study. Int J Cancer 1985;35:427-433.
(4) Austin DF, Reynolds PJ, Snyder M, et al. Malignant melanoma among employees of Lawrence Livermore National Laboratory. Lancet 1981;2:712-716.
(5) Wright W, Peters J, Mack TM. Organic chemicals and malignant melanoma. Am J Ind Med 1983;4:577-581.
(6) Hoar SK, Pell S. A retrospective cohort study of mortality and cancer incidence among chemists. J Occup Med 1981; 23:485-494.
(7) Vagero D, Olin R: Incidence of cancer in the electronics industry: Using the Swedish Cancer environment Registry as a screening instrument. Br J Ind Med 1983; 40:188-192.
(8) Wagero D, Ahlbom A, Olin R, et al. Cancer morbidity among workers in the telecommunications industry. Br J Ind Med 1985; 42;191-195.
(9) Austin DF, Reynolds P. Occupation and malignant melanoma of the skin. In: Gallagher RP, ed. Epidemiology of malignant melanoma. Heidelberg: Springer-Verlag, 1986:98-108.
(10) Hoar SK, Morrison AS, Cole P, et al. An occupation and exposure linkage system for the study of occupational carciongenesis. J. Occup Med 1980; 22:722-726.
(11) Holman CD, Armstrong BK, Henan PJ. The causes of malignant melanoma: Results from the West Australian Lions melanoma research project. In: Gallagher RO, ed. Epidemiology of malignant melanoma.Heidelberg: Springer-Verlag. 1986: 18-37.
(12) Gallagher, RO, Elwood JM, Hill GB. Risk factors for cutnaeous malignant melanoma. The Western Canada Melanoma Study. In: Gallagher RO, ed. Epidemiology of malignant melanoma. Heidelberg: Springer-Verlag, 1986:38-55.
(13) Elwood JM, Gallagher RP, Hill GB, et al. Pigmentation and skin reaction to sun as risk factors for cutaneous melanoma: Western Canada Melanoma Study. Br Med J [Clin Res] 1984; 1:99-102.
(14) Breslow NE, Day NE. Statistical methods in cancer research. Vol. 1. The analysis of case-control studies. IARC Sci Publ 1980; 32;249-276.
(15) Blishen BR. A socio-economic index for occupations in Canada. Can Rev Sociol Anthropol 1967; 4:41-53.
(16) Ministry of Supply and Services, Canada. Population: Labour force occupational trends. 1981 Census of Canada statistics Canada. Catalogue #92-920, Ottawa 1983.
(17) Elwood JM, Gallagher RO, Wroth AJ, et al. Etiological differences between subtypes of cutaneous malignant melanoma: Western Canad Melanoma Study. JNCI 1987: 78:37-44.
(18) Lee JA, Strickland D. Malignant melanoma: Social status and outdoor work. Br J Cancer 1980; 41:757-763.
(19) Holman CD, Mulroney CD, Armostrong BK. Epidemiology of pre-invasive and invasive malignant melanoma in Western Australia. Int J Cancer 1980; 25:317-323.
(20) Cooke KR, Skegg DC, Fraser J. Socio-economic status, indoor and outdoor work on malignant melanoma. Int J Cancer 1984; 34:57-62.
(21) Gallagher RP, Elwood JM, Threlfall WJ. et al. Occupation and risk of cutaneous melanoma. Am J Ind Med 1986; 9:289-294.
(22) Elwood JM, Williamson C, Stapelton PJ. Malignant melanoma in relation to moles, pigmentation, and exposure to flourescent and other lighting sources. Br J Cancer 1986;53:65-74.