We reported previously that the risk of childhood leukemia was increased among mothers with a job involving sewing at home during pregnancy. 1 Given that sewing machine operators are among the workers with the highest exposure to extremely low frequency magnetic fields (ELF-MF), 2,3 we hypothesized that such an exposure could explain the observed results. 4
There is little information on a possible relation between prenatal occupational exposure to ELF-MF and childhood leukemia. A study of 136 leukemia cases did not find an association with average maternal occupational exposure to ELF-MF in jobs held 2-26 months before the child’s birth. 5 Exposure has been assigned based on a matrix developed for men’s occupations.
We recently developed a matrix for occupational exposures to ELF-MF from jobs held by women with young children. 2 We report results from a case-control study on the relation between maternal occupational exposure to ELF-MF during pregnancy and childhood acute lymphoblastic leukemia (ALL) by using individual exposure estimations derived from our previously published matrix. 2
Case and Control Selection
The study methodology has been reported elsewhere. 6,7 Briefly, all cases of ALL between zero and 9 years of age diagnosed from 1980 to 1990 in the Province of Québec were eligible. To reduce costs, we limited the inclusion of cases between 1991 and 1993 to all those from the Greater Montreal area, which includes approximately 60% of the population in the province. Cases were recruited from tertiary care centers designated by government policy to treat and hospitalize children with cancer in the province; tracing cases from these hospitals is equivalent to a population-based ascertainment. An oncologist in a tertiary care center determined the diagnosis of ALL (International Classification of Diseases −9 code 204.0) on the basis of clinical and biologic standard criteria. All available hospital and provincial databases were reviewed to identify cases. We matched population-based controls (one per case) on sex, region of residence, and age at the calendar date of diagnosis; these were concurrently selected. Controls were identified through family allowance files, the family allowance being a government stipend awarded to all families with children living legally in Canada. This source of data was the most complete census of children for the study years. We excluded children who were adopted, who lived in foster families, whose family spoke neither French nor English, who were not resident in Canada, or whose parents were both unavailable for interview. We identified 510 eligible cases and interviewed the parents of 491 (96.3%); we contacted 588 eligible control families and 493 (84%) participated. Among controls, 74 were the second control on the list of eligibles, 9 were the third choice, and 1 was the fourth. Reasons for nonparticipation were a confidential telephone number, refusal, or the family could not be traced. Because 2 controls were interviewed for whom cases were not available, 2 strata without cases were rejected, leaving 491 cases to be used in the analysis and 491 controls. The study was approved by each hospital’ s ethics committee and informed consent was obtained from all study participants.
Soon after a mailing to the parents introducing the general purpose to the study, trained interviewers contacted the parents to schedule an appointment for the interview. Interviews were administered by telephone using structured questionnaires. One questionnaire pertained to potential confounding factors; another collected a detailed job history from the age of 18 years, as well as an in-depth description of each job held by the mother in the 2 years before and during pregnancy. Among case mothers, 98.8% answered for themselves whereas 97.4% of control mothers did. Occupations were coded according to the seven-digit Canadian Classification and Dictionary of Occupations 1971 8 and industries were coded according to the three-digit Standard Industrial Classification Manual. 9 For each job, the mother was asked about the company, its products, the nature of the worksite, her main and subsidiary tasks, and any additional information (eg, use of protective equipment, activities of coworkers) that could provide clues about possible exposures. For some occupations, additional questionnaires were used for more detailed probing. 10 The interviewers were trained for the occupational interview, and feedback was provided throughout by the team of chemists who would subsequently analyze the interviews to determine potential exposures.
An industrial hygienist blind to the case/control status reviewed the general and job-specific questionnaires to extract information on the following potential determinants of exposure to ELF-MF: job title, work environment, magnetic field sources, duration of use or exposure, total hours of work per week, and type of work shift; the detailed questionnaires were further evaluated for information on electrical equipment in the worker’ s vicinity, hours of use or of exposure per week, or type of work environment. Exposure levels were assigned to jobs based on ELF magnetic field sources (primarily electrical equipment), work environments, and duration of exposure. Specific sources of data used to assign ELF-MF intensity values to sources and work environments are given in our previous report. 2 For each job held by a subject during the 2 years previous to pregnancy up to birth, a weekly time-weighted average (TWA) exposure was calculated by multiplying the ELF magnetic field intensity of each identified source by the weekly duration of use. Any remaining work time was multiplied by the background field level assigned to the specific work environment. The products (source times duration, and environment times duration) were summed and divided by the total weekly hours spent at work. For 54% of the 1002 jobs that were held by case and control mothers during the target period defined above, local magnetic field sources were not expected to contribute substantially to exposure (eg, most waitresses and agricultural workers). In these cases the exposure was presumed to be similar to the background workplace level and that level was assigned. For jobs in which local sources of magnetic fields were expected to increase exposures substantially above background levels from the work environment, but where the sources or their duration of use had not been reported in the questionnaire, we assigned the arithmetic mean value of the calculated weekly time-weighted average exposure from other workers in the same job title who had reported durations of use of local sources. In the present report, only exposures occurring during the pregnancy period were analyzed.
We used 3 different exposure metrics for the pregnancy period: cumulative exposure in microtesla-days (μT-days), average exposure, and maximum exposure for a given occupation. For example, working 50 days in job A at a TWA of 0.3 μT and 100 days in job B at 0.1 μT gives a cumulative exposure of (0.3 × 50 + 0.1 × 100) = 25 μT-days. Average exposure in this case is (25 μT-days/150 days) = 0.16 μT. Both measures were dichotomized at the 90th percentile based on the distribution for all study subjects. Maximum exposure for an occupation was measured as weekly TWA and dichotomized at 0.4 μT, a level above which residential magnetic fields were associated with ALL. 11
Conditional logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs). In the comparison of working with nonworking women, a healthy worker effect may result because only healthy women are at work; to control for confounding by work status, we carried out separate analyses including only working women. Thus 3 exposure metrics (cumulative exposure, average exposure, and maximum exposure) were used in the analyses, which included either all study women or only working women. In each of these analyses, we used 3 models: the first accounts only for the matching factors (child’ s age and sex); the second accounts for the matching factors and maternal age; and the third model adjusts for the matching factors, maternal age, maternal diseases during pregnancy, and the presence of other children at the time of pregnancy.
The distribution of baseline characteristics (Table 1) was quite similar between cases and controls. A similar number of women in both groups had no gainful employment but more case mothers already had children at the time of pregnancy. More control mothers were exposed to ELF-MF at work but cumulative levels were somewhat higher in cases. More case mothers were exposed at maximum levels for ELF-MF for a given occupation. Use of bedroom electric appliances such as electric blanket or mattress cover or heated water bed was slightly more frequent among control mothers.
By using the cumulative exposure metric and including all women in the analysis (Table 2), the risk of ALL was increased by 70% in the first model which accounted only for the matching variables. The increase was only slightly less in the 2 other models which adjusted for additional variables. In a separate analysis that included only working women, the OR was 1.8 (95% CI = 1.1-3.0) in the first model and only slightly less in the models with additional control variables. By using the average exposure metric among all study women, the odds ratio for ALL in the first model was 1.4 (95% CI = 0.95-2.2) and similar in the other models. The odds ratios were slightly higher when using this metric among working women only. Women whose jobs put them in the category of maximum level attained for an occupation (≥ 0.4 μT) had more than double the risk of having a child with ALL (risk slightly increased in models with additional control variables). In the fully adjusted model limited to working women the OR was 2.5 (95% CI = 1.3–5.0).
The cumulative and average metrics were also categorized according to multiple dose groups: <50th percentile (used as the baseline), 50–<75th percentile, 75–<90th percentile and ≥90th percentile. With the cumulative metric in the fully adjusted model among all women, we found the following ORs: 0.8 (95% CI = 0.6–1.2), 0.9 (0.6–1.3), and 1.4 (0.8–2.2), respectively. With the average metric these results were:1.0 (0.7–1.5), 1.1 (0.7–1.7) and 1.5 (0.9–2.4), respectively. Results were similar with both metrics when the analyses were restricted to working women.
Maternal occupations with the highest estimated exposures were sewing machine operators in factories (weekly TWA of 0.68 μT) and electronics workers (0.43 μT);2 there were 27 case mothers reporting these jobs in comparison with 11 control mothers. To determine whether the pattern of results was determined solely by these jobs, we repeated the analyses of Table 2 taking out these cases and controls. We found results that are still suggestive of increased risk at high levels of exposure. By using the cumulative metric in the first model among all women, the OR was 1.5 (95% CI = 0.9–2.4); the OR was 1.3 (0.8–2.2) with the average metric and 2.4 (1.0–5.8) by using the maximum exposure metric.
Finally, we controlled in the analyses for occupational exposure to any kind of polycyclic aromatic hydrocarbons during pregnancy; the results for ELF-MF did not materially change.
Our results suggest that children whose mothers were exposed during pregnancy to the highest levels of ELF-MF from sources in their work environment were at moderately increased risk of ALL. Including all study women in the analyses or restricting the analyses to working women gave similar results, although they were somewhat more suggestive of an association among working women. Adjusting for potential confounding variables, such as maternal health status during pregnancy, and the presence of other children at the time of pregnancy did not substantially change the results. With the average exposure metric, the increases in risk were less than with the cumulative or maximum exposure metrics.
As previously reported, 2 2 job titles (sewing machine operator in factories and electronics worker) were assigned the highest levels of exposure. These may be the ones at risk; there were more case mothers with these titles than control mothers. However, when mothers with these job titles were taken out of the analysis, results were still indicative of increased risk at the highest exposure levels, suggesting that the association does not rest exclusively on these job titles. By using the same cutoff point (≥90th percentile) on the cumulative and average distributions but comparing results in that group with those in the group below the 50th percentile (as opposed to the rest of the distribution as in Table 2) again suggested that risk was increased at the highest levels of exposure. However, the cumulative metric also suggested that study subjects below the 50th percentile for ELF-MF possibly carried a higher risk than those between the 50th and 90th percentile in all women (which includes women at home) and in working women only. The reason for this is not clear although this may be related to the metric used because the average metric did not show this pattern.
The present study is the first to relate individual estimates of occupational exposure to ELF-MF during pregnancy and ALL. Mechanisms for the effect of ELF-MF on childhood leukemia (if any) are not known 11 and therefore the relevant exposure metric for the effect of ELF-MF is speculative. 12 All measures we used gave approximately the same results, although the average metric was somewhat less sensitive. Results from residential exposures have shown an association with leukemia only among children exposed at the highest levels (ambient residential ELF-MF at or above 0.4 μT). 11 At such levels, the pooled relative risk from 9 studies carried out in different countries was 2.0 (95% CI = 1.3–3.1). Our data are compatible with this observation.
In this study, controls were selected to be at risk at the time cases were diagnosed. In addition, participation rates were high in both study groups. Both factors tend to reduce the possibility for selection bias. Limiting the analyses to working women controls for the potential confounding effect by work status. We did not measure home exposure to ELF-MF. However, high exposure levels in the home are rare;13 in addition, the proportion of nonworking women was the same in both groups, as was the socioeconomic level. These factors would tend to reduce the potential for marked differential home exposures between the mothers of cases and controls. We also controlled for maternal health status during pregnancy and the presence of other children at the time of pregnancy, because both factors could be associated with work status as well as with the studied outcome.
The validity of our results hinges mainly on the validity of the exposure assessment. It is unlikely that differential misclassification would have affected results. On the one hand, job titles and work environments are unlikely to be differentially recalled by mothers of cases and controls and, on the other hand, the hygienist developing the exposure estimates was blind to the case/control status. However, nondifferential misclassification is likely. In addition, the exposure estimations we are proposing 2 have not been used before and will need to be applied by others to determine their validity.
Pregnancy exposures could initiate the first events leading to childhood ALL. 14 This is why it is reasonable to target the pregnancy period for exposure to suspected agents even though at this time carcinogenic effects of ELF-MF have not been established.
In conclusion, we found a relation between the highest levels of maternal occupational exposure to ELF-MF and childhood leukemia. The exposure estimations we are proposing will need to be used by others before any firm conclusion can be reached. In addition, mechanistic evidence for the carcinogenic potential of ELF-MF will be necessary to consider that the association as causal.
1. Infante-Rivard C, Mur P, Armstrong B, et al. Acute lymphoblastic leukemia in Spanish children and mother’s occupation: A case-control study. J Epidemiol Comm Health. 1991; 45: 11–15.
2. Deadman J, Infante-Rivard C. Individual estimation of exposures to extremely low frequency magnetic fields in jobs commonly held by women. Am J Epidemiol. 2002; 155: 368–378.
3. Sobel E, Davanipour Z, Sulkava R, et al. Occupations with exposures to electromagnetic fields: A possible risk factor for Alzheimer’s disease. Am J Epidemiol. 1995; 142: 515–524.
4. Infante-Rivard C. Electromagnetic field exposure during pregnancy and childhood leukemia. Lancet. 1995; 346: 177.
5. Feychting M, Floderus B, Ahlbom A. Parental occupational exposure
to magnetic fields and childhood cancer (Sweden). Cancer Causes Control. 2000; 11: 151–156.
6. Infante-Rivard C, Fortier I, Olson E. Markers of infection, breast-feeding and childhood acute lymphoblastic leukemia. Br J Cancer. 2000; 83: 1559–1564.
7. Infante-Rivard C, Krajinovik M, Labuda D, et al. Childhood acute lymphoblastic leukemia associated with parental alcohol consumption and polymorphisms of carcinogen-metabolizing genes. Epidemiology. 2002; 13: 277–281.
8. Department of manpower and Immigration. Canadian classification and dictionary of occupations 1971. 1974, Vol 1. Ottawa: Information Canada.
9. Dominion Bureau of Statistics. Standard Industrial classification manual. 1971. Ottawa: Information Canada.
10. Gérin M, Siematycki J. The occupational questionnaire in retrospective epidemiologic studies. Appl Occup Environ Hyg. 1991; 6: 495–501.
11. Ahlbom A, Feychting M. Current thinking about risks from currents. Lancet. 2001; 357: 1143–1148.
12. Savitz DA. Magnetic fields and miscarriage. Epidemiology. 2002; 13: 1–3.
13. Deadman JE, Armstrong BG, McBride ML, et al. Exposures of children in Canada to 60-Hz magnetic and electric fields. Scand J Work Environ Health. 1999; 25: 368–375.
14. Greaves M. Science, medicine, and the future: Childhood leukaemia. BMJ. 2002; 324: 283–287.