AMERICAN JOURNAL or EPIDEMIOLOGY
VOL 131, No. 5
Copyright © 1990 by The Johns Hopkins University School of Hygiene and Public Health All nghu reserved
Pnnttd in U S A
Original Contributions MAGNETIC FIELD EXPOSURE FROM ELECTRIC APPLIANCES AND CHILDHOOD CANCER DAVID A. SAVITZ, ESTHER M. JOHN, AND ROBERT C. KLECKNER Savitz, D. A. (School of Public Health, U. of North Carolina, Chapel Hill, NC 27599), E. M. John, and R. C. Kleckner. Magnetic field exposure from electric appliances and childhood cancer. Am J Epidemiol 1990; 131:763-73. The effect on childhood cancer of prolonged exposure to 50-H magnetic fields from electric appliances was examined using interview data from a recently completed case-control study. Exposures of children aged 0-14 years whose incident cancers were diagnosed between 1976 and 1983 and who resided in the Denver, Colorado, Standard Metropolitan Statistical Area were compared with those of controls selected by random digit dialing, matched on age, sex, and telephone exchange area. Parents of 252 cases and 222 controls were interviewed at home about the use of electric appliances by the mother during pregnancy (prenatal exposure) and by the child (postnatal exposure). After adjustment for income, prenatal electric blanket exposure was associated with a small increase in the incidence of childhood cancers (odds ratio (OR) = 1.3, 95% confidence interval (Cl) 0.7-2.2) that was more pronounced for leukemia (OR = 1.7, 95% Cl 0.8-3.6) and brain cancer (OR = 2.5, 95% a 1.1-5.5). Postnatal exposure to electric blankets was also weakly associated with childhood cancer (OR = 1.5, 95% Cl 0.6-3.4), with a larger but imprecise association with acute lymphocytic leukemia (OR = 1.9, 95% Cl 0.6-6.5). Water beds and bedside electric clocks were unrelated to childhood cancer incidence. Results are limited by nonresponse and imprecision resulting from the rarity of appliance use, especially for subgroups of cases. Nonetheless, electric blankets, one of the principal sources of prolonged magnetic field exposure, were weakly associated with childhood cancer and warrant a more thorough evaluation. brain neoplasms; child; electromagnetic fields; leukemia
Editor's note: For a discussion of this paper and that by Verreault et at. immediately preceding, see page 774. Several studies suggest that exposure to magnetic fields in residences (1) and in the workplace (2, 3) may be related to the de-
velopment of cancer. The underlying hypothesis is that prolonged exposure to elevated levels of power-frequency (50-60 Hz) magnetic fields (and possibly electric fields) increases the incidence of certain cancers. The evidence is most supportive for leukemiaa and brain cancers in both children and adults.
Received for publication April 13,1989, and in final This study was part of a research program to deterform July 17, 1989. mine the possible adverse health effects resulting from Abbreviations: Cl, confidence interval; mG, milli- exposure to the electric and magnetic fields of overgauss; OR, odds ratio. head high-voltage transmission lines. The program From the Department of Epidemiology, CB #7400, was administered by the State of New York Depart School of Public Health, University of North Carolina, ment of Health and Health Research, Inc. Chapel Hill, NC 27599. (Send reprint requests to Dr. The authors acknowledge the helpful review of the David A. Savitz at this address.) manuscript by Dr. Dana Loomis. 763
764
SAVITZ ET AL.
The patterns of human exposure to magnetic fields are incompletely understood, but substantial contributions to the timeintegrated "dose" appear to arise from three sources: 1) background levels in residences, which produce only modest elevations (with levels around 2 milligauss (mG) considered relatively "high") but extend over prolonged periods of time (4); 2) occupational exposures to electrically energized equipment, with somewhat less protracted exposures to substantially more elevated fields (10-50 mG) (5-7); and 3) electric appliances, including a few with prolonged exposure and relatively high fields (10-20 mG), such as electric blankets, and many others with more highly elevated magnetic field strengths (50 mG and higher) but very brief periods of exposure (hair dryers, electric razors) (8, 9). The focus of epidemiologic studies (including the present one) on markers of cumulative exposure is a default decision based principally on a lack of knowledge regarding the biologically relevant measure (if any) rather than on empirical evidence from laboratory or epidemiologic studies.
der a series of assumptions regarding background field levels in the home, field strength of the blanket, duration of use, and a duty cycle of 0.5 (i.e., the proportion of time current is being drawn). For magnetic fields, they estimated that an electric blanket user would have, on average, 1.8 times the exposure of a nonuser, with a range of 1.3-4.5 under varying assumptions. The estimated contribution to electric fields is somewhat more modest, with users predicted to have 1.4 times the exposure of nonusers and a range of 1.1-2.6. Given the need for further evaluation of magnetic fields and cancer (1, 2, 11), exposures from those appliances that make substantial contributions to cumulative exposure are worthy of examination. Since the strongest evidence for an effect of residential magnetic fields exists for childhood cancer (1), an examination of appliance exposures in relation to these diseases is especially important and has not previously been reported. A recently completed case-control study of childhood cancer and residential magnetic field exposures (12) included information on prenatal and postnatal electric appliance exposures.
Each of these sources is a potentially important contributor to exposure and worMATERIALS AND METHODS thy of consideration in relation to health outcomes, but exposures from appliances Methodological details of this study are have not received much attention in past available elsewhere (12, 13), so they will be studies. This is unfortunate, given that only briefly reviewed here. All incident such exposures may be much more easily cases of childhood cancer occurring among evaluated than background residential or children who lived in the Denver area at workplace exposures. Appliances are ame- the time of diagnosis, were under age 15, nable to self-report with at least as much and were diagnosed during the period 1976accuracy as many other exposures exam- 1983 were considered eligible. They were ined by epidemiologists like diet, tobacco identified principally through the Colorado smoking, and use of medications. Cancer Registry with supplemental identification through area hospitals (13). ReOnly Preston-Martin et al. (10) have sults are presented for total cancer cases conducted a detailed examination of an apand leukemias, with subgroups of acute pliance-baaed exposure and cancer (speciflymphocytic leukemias, lymphomas, brain ically, electric blankets in relation to adult cancers, and soft tissue sarcomas presented myeloid leukemia). Though they found when sufficient numbers were available. odds ratios very close to the null, they provided some useful calculations that esControls were identified by random digit timate the potential contribution of electric dialing (14), starting with the case's teleblankets to total (cumulative) exposure un- phone number and then randomly replac-
ELECTRIC APPLIANCES AND CHILDHOOD CANCER
ing the last two digits. The selected number was called up to six times in an attempt to locate a child of similar age (± 3 years) and sex who resided in the case's telephone exchange area at the time of the case's diagnosis and remained in that area up to the time of the interview. Though controls were sought for individual cases, they were not considered individually matched, since the selection criteria were broad and efficiency is lost in an analysis that retains the matching under those conditions (15). After obtaining physician consent to contact cases, data collection proceeded in the same manner as for identified controls. A letter describing the study was sent to eligible respondents, and they were called to arrange a time for an in-home interview and, if still occupying a home which was occupied prior to diagnosis, measurement of electric and magnetic fields in the residence. The 1-hour interview covered potential childhood cancer risk factors including demographic background, family cancer history, cigarette smoking, and alcohol consumption by the mother during pregnancy; occupational history of both parents; and diet, medications, radiographs, and electric appliance use by the mother during pregnancy and by the case or control child. A wiring configuration code to approximate average magnetic fields in the home (16) was obtained during a separate visit. To assess exposure from electric appliances, a precoded list was first discussed that included electric blanket, portable radiant heater, portable electric space heater, heated water bed, heating pad, color television, black and white television, hair dryer, curling iron or hot curlers, electric clock next to the bed, radio or stereo, home' computer or video game, electric typewriter, room fan, room air conditioner, and room humidifier or vaporizer. Respondents were then invited to name other appliances used for prolonged periods of time. Reported use of electric blankets and heated water beds was followed by questions concerning the setting of use (high,
765
medium, or low) and whether the window was left open, both intended as markers of the duty cycle (proportion of time the device was drawing current and thus producing elevated magnetic fields). More detailed questions were asked about all appliances used during pregnancy, assessing the timing of use (by trimester) and the amount of time per day the appliance was in use. Children's reported use of an appliance was followed by questions concerning the age at which the use began and ended, number of months used during the year (for seasonal items), and time used per day. Based on the frequency of use and potential for contribution to total exposure (determined largely by proximity and duration of use), a subset of appliances was selected for detailed analysis. At the time the original list was produced, primary consideration was given to the prevalence of use in the population and duration of contact with the device, with very limited data available on the possible contribution to exposure. The majority of the named appliances were too rarely reported for meaningful analysis. Many of the items are now known not to be likely to contribute to exposure (radio or stereo). Other items are potentially important sources, but are very commonly used, and exposure is highly dependent on the user's distance from the appliance (televisions). Mothers' use of some appliances is unlikely to expose the fetus, including hair dryer and curling iron or hot curlers. Primary consideration was given to the mother's and child's use of electric blankets, heated water beds, and electric clocks by the bed (all potentially capable of providing a significant exposure). Selected results for the mother's use of a heating pad and the child's use of a hair dryer were also reported. Appliances of interest were first examined with exposure assigned as ever or never used, separately for the mother's use during pregnancy (prenatal exposure) and the child's use after birth but prior to diagnosis (postnatal exposure). Potential
766
SAVITZ ET AL.
confounders were considered, including child's sex, age at diagnosis, year of diagnosis, family per capita income, mother's age, father's education, mother's smoking during pregnancy, and wire code at the time of diagnosis, all identified in these data as possible risk factors for childhood cancer (12,17). In addition, the possible confounding of appliance exposures by one another was examined. Several of these were also of interest as possible effect modifiers, with results presented separately by child's sex, age at diagnosis, and wire code at diagnosis. The setting of electric blankets and heated water beds was evaluated as well as the timing of use during each trimester of pregnancy and nightly duration of use for prenatal exposures. For the child's appliance use, the electric blanket and heated water bed setting, months of use per year, age started, nightly duration of use, number of years of use, time from initial use to diagnosis, and time from cessation of use to diagnosis were considered. Odds ratios relating exposures to appliances to cancer incidence (total cancers and subgroups) were calculated with test-based 95 percent confidence intervals (18). Stratified analyses were used to assess the presence of confounding and effect modification, with selected stratum-specific and adjusted odds ratios results (19) presented. Multiple confounders were examined in logistic regression analyses. RESULTS
Response Among the 356 eligible cases, 252 (70.8 percent) were interviewed, with 61 (17.1 percent) untraceable, 31 (8.7 percent) respondent refusals, and four (1.1 percent) physician refusals. Losses were greater for nonwhites and children diagnosed earlier, both due to more untraceable subjects. Through random digit dialing, 278 potential controls were identified, with an estimated response of 78.6 percent to the tele-
phone screening (12). Among identified controls, 222 agreed to be interviewed (79.9 percent), with essentially all the others (19.8 percent) lost because of refusal. Further losses were incurred for specific items when the respondent failed to provide the desired data; e.g., adoptive parents were missing data on exposures during pregnancy. Complete appliance use data for mothers were provided for 233 cases (92.5 percent of those interviewed) and 206 controls (92.8 percent). Appliance use information was available for a greater proportion of children: 244 cases (96.8 percent) and 216 controls (97.3 percent). Prenatal exposure Unadjusted associations of the mother's use of selected appliances (resulting in prenatal exposures) and childhood cancer are presented in table 1. None of the appliances under consideration produced notably elevated odds ratios for total cancers. For some cancer subgroups, elevated odds ratios were observed, though there is no clear pattern to the finding (e.g., electric blankets and brain cancer, heated water beds and soft tissue sarcomas, heating pads and lymphomas, soft tissue sarcomas). Of these, the largest and most precise elevation in the effect estimate was observed for electric blanket use and brain cancer (odds ratio (OR) = 1.8, 95 percent confidence interval (CI) 0.9-4.0). Electric blanket use was examined in more detail since it is known to be a substantial source of magnetic field exposure (10). Potential confounding was considered for total cancers, leukemias, and brain cancers, and the adjusted estimate differed only slightly from the crude odds ratio (less than 10 percent deviation) except for per capita income. Stratum-specific odds ratios by income and income-adjusted odds ratios are presented in table 2. Adjusted odds ratios for total cancer, leukemia, and brain cancer were 1.3 (95 percent CI 0.7-2.2), 1.7 (95 percent CI 0.8-3.6), and 2.5 (95 percent
767
ELECTRIC APPLIANCES AND CHILDHOOD CANCER TABLE 1
Unadjusted association of prenatal appliance exposure with childhood cancer Denver, Colorado, 1976-1983 No. exposed
No. not exposed
OR-
95% CI*
31 38 13 10 4 11
175 195 57 46 20 34 21 63
1.0 1.1 1.3 12 1.1 1.8 08 06
0.7-1.8 0.7-2.6 0.6-2.7 0.4-3.6 0.9-4.0 0.3-3.0 0.3-1.5
0 3 6 9
182 213 67 53 24 42 19 61
1.0 0.7 0.3 0.4 0.0 2.4 1.1
0.2-2.0 0.9-6.6 0.5-2.5
Total cancer Leukemia Acute lymphocytic leukemia Lymphoma Brain Soft tissue Other cancers
149 157 48 39 14 31 19 46
57 76 21 17 10 14 6 25
1.0 0.8 0.9 0.9 0.5 0.8 1.2 0.7
0.5-1.2 0.5-1.6 0.5-1.6 0.2-1.2 0.4-1.7 0.5-2.9 0.4-1.2
Heating pad Controls Total cancer Leukemia Acute lymphocytic leukemia Lymphoma Brain Soft tissue Other cancers
24 29 7 7 5 5 5 7
182 202 60 48 19 40 20 63
1.0 1.1 0.9 1.1 2.0 0.9 1.9 0.8
0.6-1.9 0.4-2.2 0.5-2.8 0.7-5.9 0.4-2.7 0.7-5.6 0.4-2.1
Appliance Electric blanket Controls
Total cancar Leukemia Acute lymphocytic leukemia Lymphoma
Brain
3
Soft tissue Other cancers
7
Heated water bed Controls Total cancer Leukemia Acute lymphocytic leukemia Lymphoma Brain Soft tissue Other cancers
Bedside electric clock Controls
24 21 3
3
0.5
0.4-1.4 0.1-1.2 0.2-1.6
* OR, odds ratio; CI, confidence interval
CI 1.1-5.5), respectively, reflecting moderate confounding toward the null. There wasno additional effect of adjusting the mother's electric blanket use for other prenatal (heated water bed, electric clock) and postnatal (electric blanket, heated water bed, electric clock) appliance exposures. Income-adjusted, stratum-specific results for electric blanket use by sex, age at diagnosis, and wire code (as a marker of
residential background magnetic fields) (table 2) suggested slightly stronger associations for males, older leukemia cases (OR = 3.0 for 10- to 14-year-olds), and younger brain cancer cases (OR = 3.7 for 0- to 4-year-olds) (table 2). For leukemias, an enhanced effect was found for residents of higher wire code (presumably higher magnetic field) homes though the opposite tendency was found for brain cancer.
768
SAVITZ ET AL. TABLE 2
Stratum-specific results for prenatal electric blanket exposure and childhood cancer Denuer, Cobrado, 1976-1983' Total cancer Stratification vanflhlfl
Not stratified* Unexposed Exposed Annual per capita income
VOL 131, No. 5
Copyright © 1990 by The Johns Hopkins University School of Hygiene and Public Health All nghu reserved
Pnnttd in U S A
Original Contributions MAGNETIC FIELD EXPOSURE FROM ELECTRIC APPLIANCES AND CHILDHOOD CANCER DAVID A. SAVITZ, ESTHER M. JOHN, AND ROBERT C. KLECKNER Savitz, D. A. (School of Public Health, U. of North Carolina, Chapel Hill, NC 27599), E. M. John, and R. C. Kleckner. Magnetic field exposure from electric appliances and childhood cancer. Am J Epidemiol 1990; 131:763-73. The effect on childhood cancer of prolonged exposure to 50-H magnetic fields from electric appliances was examined using interview data from a recently completed case-control study. Exposures of children aged 0-14 years whose incident cancers were diagnosed between 1976 and 1983 and who resided in the Denver, Colorado, Standard Metropolitan Statistical Area were compared with those of controls selected by random digit dialing, matched on age, sex, and telephone exchange area. Parents of 252 cases and 222 controls were interviewed at home about the use of electric appliances by the mother during pregnancy (prenatal exposure) and by the child (postnatal exposure). After adjustment for income, prenatal electric blanket exposure was associated with a small increase in the incidence of childhood cancers (odds ratio (OR) = 1.3, 95% confidence interval (Cl) 0.7-2.2) that was more pronounced for leukemia (OR = 1.7, 95% Cl 0.8-3.6) and brain cancer (OR = 2.5, 95% a 1.1-5.5). Postnatal exposure to electric blankets was also weakly associated with childhood cancer (OR = 1.5, 95% Cl 0.6-3.4), with a larger but imprecise association with acute lymphocytic leukemia (OR = 1.9, 95% Cl 0.6-6.5). Water beds and bedside electric clocks were unrelated to childhood cancer incidence. Results are limited by nonresponse and imprecision resulting from the rarity of appliance use, especially for subgroups of cases. Nonetheless, electric blankets, one of the principal sources of prolonged magnetic field exposure, were weakly associated with childhood cancer and warrant a more thorough evaluation. brain neoplasms; child; electromagnetic fields; leukemia
Editor's note: For a discussion of this paper and that by Verreault et at. immediately preceding, see page 774. Several studies suggest that exposure to magnetic fields in residences (1) and in the workplace (2, 3) may be related to the de-
velopment of cancer. The underlying hypothesis is that prolonged exposure to elevated levels of power-frequency (50-60 Hz) magnetic fields (and possibly electric fields) increases the incidence of certain cancers. The evidence is most supportive for leukemiaa and brain cancers in both children and adults.
Received for publication April 13,1989, and in final This study was part of a research program to deterform July 17, 1989. mine the possible adverse health effects resulting from Abbreviations: Cl, confidence interval; mG, milli- exposure to the electric and magnetic fields of overgauss; OR, odds ratio. head high-voltage transmission lines. The program From the Department of Epidemiology, CB #7400, was administered by the State of New York Depart School of Public Health, University of North Carolina, ment of Health and Health Research, Inc. Chapel Hill, NC 27599. (Send reprint requests to Dr. The authors acknowledge the helpful review of the David A. Savitz at this address.) manuscript by Dr. Dana Loomis. 763
764
SAVITZ ET AL.
The patterns of human exposure to magnetic fields are incompletely understood, but substantial contributions to the timeintegrated "dose" appear to arise from three sources: 1) background levels in residences, which produce only modest elevations (with levels around 2 milligauss (mG) considered relatively "high") but extend over prolonged periods of time (4); 2) occupational exposures to electrically energized equipment, with somewhat less protracted exposures to substantially more elevated fields (10-50 mG) (5-7); and 3) electric appliances, including a few with prolonged exposure and relatively high fields (10-20 mG), such as electric blankets, and many others with more highly elevated magnetic field strengths (50 mG and higher) but very brief periods of exposure (hair dryers, electric razors) (8, 9). The focus of epidemiologic studies (including the present one) on markers of cumulative exposure is a default decision based principally on a lack of knowledge regarding the biologically relevant measure (if any) rather than on empirical evidence from laboratory or epidemiologic studies.
der a series of assumptions regarding background field levels in the home, field strength of the blanket, duration of use, and a duty cycle of 0.5 (i.e., the proportion of time current is being drawn). For magnetic fields, they estimated that an electric blanket user would have, on average, 1.8 times the exposure of a nonuser, with a range of 1.3-4.5 under varying assumptions. The estimated contribution to electric fields is somewhat more modest, with users predicted to have 1.4 times the exposure of nonusers and a range of 1.1-2.6. Given the need for further evaluation of magnetic fields and cancer (1, 2, 11), exposures from those appliances that make substantial contributions to cumulative exposure are worthy of examination. Since the strongest evidence for an effect of residential magnetic fields exists for childhood cancer (1), an examination of appliance exposures in relation to these diseases is especially important and has not previously been reported. A recently completed case-control study of childhood cancer and residential magnetic field exposures (12) included information on prenatal and postnatal electric appliance exposures.
Each of these sources is a potentially important contributor to exposure and worMATERIALS AND METHODS thy of consideration in relation to health outcomes, but exposures from appliances Methodological details of this study are have not received much attention in past available elsewhere (12, 13), so they will be studies. This is unfortunate, given that only briefly reviewed here. All incident such exposures may be much more easily cases of childhood cancer occurring among evaluated than background residential or children who lived in the Denver area at workplace exposures. Appliances are ame- the time of diagnosis, were under age 15, nable to self-report with at least as much and were diagnosed during the period 1976accuracy as many other exposures exam- 1983 were considered eligible. They were ined by epidemiologists like diet, tobacco identified principally through the Colorado smoking, and use of medications. Cancer Registry with supplemental identification through area hospitals (13). ReOnly Preston-Martin et al. (10) have sults are presented for total cancer cases conducted a detailed examination of an apand leukemias, with subgroups of acute pliance-baaed exposure and cancer (speciflymphocytic leukemias, lymphomas, brain ically, electric blankets in relation to adult cancers, and soft tissue sarcomas presented myeloid leukemia). Though they found when sufficient numbers were available. odds ratios very close to the null, they provided some useful calculations that esControls were identified by random digit timate the potential contribution of electric dialing (14), starting with the case's teleblankets to total (cumulative) exposure un- phone number and then randomly replac-
ELECTRIC APPLIANCES AND CHILDHOOD CANCER
ing the last two digits. The selected number was called up to six times in an attempt to locate a child of similar age (± 3 years) and sex who resided in the case's telephone exchange area at the time of the case's diagnosis and remained in that area up to the time of the interview. Though controls were sought for individual cases, they were not considered individually matched, since the selection criteria were broad and efficiency is lost in an analysis that retains the matching under those conditions (15). After obtaining physician consent to contact cases, data collection proceeded in the same manner as for identified controls. A letter describing the study was sent to eligible respondents, and they were called to arrange a time for an in-home interview and, if still occupying a home which was occupied prior to diagnosis, measurement of electric and magnetic fields in the residence. The 1-hour interview covered potential childhood cancer risk factors including demographic background, family cancer history, cigarette smoking, and alcohol consumption by the mother during pregnancy; occupational history of both parents; and diet, medications, radiographs, and electric appliance use by the mother during pregnancy and by the case or control child. A wiring configuration code to approximate average magnetic fields in the home (16) was obtained during a separate visit. To assess exposure from electric appliances, a precoded list was first discussed that included electric blanket, portable radiant heater, portable electric space heater, heated water bed, heating pad, color television, black and white television, hair dryer, curling iron or hot curlers, electric clock next to the bed, radio or stereo, home' computer or video game, electric typewriter, room fan, room air conditioner, and room humidifier or vaporizer. Respondents were then invited to name other appliances used for prolonged periods of time. Reported use of electric blankets and heated water beds was followed by questions concerning the setting of use (high,
765
medium, or low) and whether the window was left open, both intended as markers of the duty cycle (proportion of time the device was drawing current and thus producing elevated magnetic fields). More detailed questions were asked about all appliances used during pregnancy, assessing the timing of use (by trimester) and the amount of time per day the appliance was in use. Children's reported use of an appliance was followed by questions concerning the age at which the use began and ended, number of months used during the year (for seasonal items), and time used per day. Based on the frequency of use and potential for contribution to total exposure (determined largely by proximity and duration of use), a subset of appliances was selected for detailed analysis. At the time the original list was produced, primary consideration was given to the prevalence of use in the population and duration of contact with the device, with very limited data available on the possible contribution to exposure. The majority of the named appliances were too rarely reported for meaningful analysis. Many of the items are now known not to be likely to contribute to exposure (radio or stereo). Other items are potentially important sources, but are very commonly used, and exposure is highly dependent on the user's distance from the appliance (televisions). Mothers' use of some appliances is unlikely to expose the fetus, including hair dryer and curling iron or hot curlers. Primary consideration was given to the mother's and child's use of electric blankets, heated water beds, and electric clocks by the bed (all potentially capable of providing a significant exposure). Selected results for the mother's use of a heating pad and the child's use of a hair dryer were also reported. Appliances of interest were first examined with exposure assigned as ever or never used, separately for the mother's use during pregnancy (prenatal exposure) and the child's use after birth but prior to diagnosis (postnatal exposure). Potential
766
SAVITZ ET AL.
confounders were considered, including child's sex, age at diagnosis, year of diagnosis, family per capita income, mother's age, father's education, mother's smoking during pregnancy, and wire code at the time of diagnosis, all identified in these data as possible risk factors for childhood cancer (12,17). In addition, the possible confounding of appliance exposures by one another was examined. Several of these were also of interest as possible effect modifiers, with results presented separately by child's sex, age at diagnosis, and wire code at diagnosis. The setting of electric blankets and heated water beds was evaluated as well as the timing of use during each trimester of pregnancy and nightly duration of use for prenatal exposures. For the child's appliance use, the electric blanket and heated water bed setting, months of use per year, age started, nightly duration of use, number of years of use, time from initial use to diagnosis, and time from cessation of use to diagnosis were considered. Odds ratios relating exposures to appliances to cancer incidence (total cancers and subgroups) were calculated with test-based 95 percent confidence intervals (18). Stratified analyses were used to assess the presence of confounding and effect modification, with selected stratum-specific and adjusted odds ratios results (19) presented. Multiple confounders were examined in logistic regression analyses. RESULTS
Response Among the 356 eligible cases, 252 (70.8 percent) were interviewed, with 61 (17.1 percent) untraceable, 31 (8.7 percent) respondent refusals, and four (1.1 percent) physician refusals. Losses were greater for nonwhites and children diagnosed earlier, both due to more untraceable subjects. Through random digit dialing, 278 potential controls were identified, with an estimated response of 78.6 percent to the tele-
phone screening (12). Among identified controls, 222 agreed to be interviewed (79.9 percent), with essentially all the others (19.8 percent) lost because of refusal. Further losses were incurred for specific items when the respondent failed to provide the desired data; e.g., adoptive parents were missing data on exposures during pregnancy. Complete appliance use data for mothers were provided for 233 cases (92.5 percent of those interviewed) and 206 controls (92.8 percent). Appliance use information was available for a greater proportion of children: 244 cases (96.8 percent) and 216 controls (97.3 percent). Prenatal exposure Unadjusted associations of the mother's use of selected appliances (resulting in prenatal exposures) and childhood cancer are presented in table 1. None of the appliances under consideration produced notably elevated odds ratios for total cancers. For some cancer subgroups, elevated odds ratios were observed, though there is no clear pattern to the finding (e.g., electric blankets and brain cancer, heated water beds and soft tissue sarcomas, heating pads and lymphomas, soft tissue sarcomas). Of these, the largest and most precise elevation in the effect estimate was observed for electric blanket use and brain cancer (odds ratio (OR) = 1.8, 95 percent confidence interval (CI) 0.9-4.0). Electric blanket use was examined in more detail since it is known to be a substantial source of magnetic field exposure (10). Potential confounding was considered for total cancers, leukemias, and brain cancers, and the adjusted estimate differed only slightly from the crude odds ratio (less than 10 percent deviation) except for per capita income. Stratum-specific odds ratios by income and income-adjusted odds ratios are presented in table 2. Adjusted odds ratios for total cancer, leukemia, and brain cancer were 1.3 (95 percent CI 0.7-2.2), 1.7 (95 percent CI 0.8-3.6), and 2.5 (95 percent
767
ELECTRIC APPLIANCES AND CHILDHOOD CANCER TABLE 1
Unadjusted association of prenatal appliance exposure with childhood cancer Denver, Colorado, 1976-1983 No. exposed
No. not exposed
OR-
95% CI*
31 38 13 10 4 11
175 195 57 46 20 34 21 63
1.0 1.1 1.3 12 1.1 1.8 08 06
0.7-1.8 0.7-2.6 0.6-2.7 0.4-3.6 0.9-4.0 0.3-3.0 0.3-1.5
0 3 6 9
182 213 67 53 24 42 19 61
1.0 0.7 0.3 0.4 0.0 2.4 1.1
0.2-2.0 0.9-6.6 0.5-2.5
Total cancer Leukemia Acute lymphocytic leukemia Lymphoma Brain Soft tissue Other cancers
149 157 48 39 14 31 19 46
57 76 21 17 10 14 6 25
1.0 0.8 0.9 0.9 0.5 0.8 1.2 0.7
0.5-1.2 0.5-1.6 0.5-1.6 0.2-1.2 0.4-1.7 0.5-2.9 0.4-1.2
Heating pad Controls Total cancer Leukemia Acute lymphocytic leukemia Lymphoma Brain Soft tissue Other cancers
24 29 7 7 5 5 5 7
182 202 60 48 19 40 20 63
1.0 1.1 0.9 1.1 2.0 0.9 1.9 0.8
0.6-1.9 0.4-2.2 0.5-2.8 0.7-5.9 0.4-2.7 0.7-5.6 0.4-2.1
Appliance Electric blanket Controls
Total cancar Leukemia Acute lymphocytic leukemia Lymphoma
Brain
3
Soft tissue Other cancers
7
Heated water bed Controls Total cancer Leukemia Acute lymphocytic leukemia Lymphoma Brain Soft tissue Other cancers
Bedside electric clock Controls
24 21 3
3
0.5
0.4-1.4 0.1-1.2 0.2-1.6
* OR, odds ratio; CI, confidence interval
CI 1.1-5.5), respectively, reflecting moderate confounding toward the null. There wasno additional effect of adjusting the mother's electric blanket use for other prenatal (heated water bed, electric clock) and postnatal (electric blanket, heated water bed, electric clock) appliance exposures. Income-adjusted, stratum-specific results for electric blanket use by sex, age at diagnosis, and wire code (as a marker of
residential background magnetic fields) (table 2) suggested slightly stronger associations for males, older leukemia cases (OR = 3.0 for 10- to 14-year-olds), and younger brain cancer cases (OR = 3.7 for 0- to 4-year-olds) (table 2). For leukemias, an enhanced effect was found for residents of higher wire code (presumably higher magnetic field) homes though the opposite tendency was found for brain cancer.
768
SAVITZ ET AL. TABLE 2
Stratum-specific results for prenatal electric blanket exposure and childhood cancer Denuer, Cobrado, 1976-1983' Total cancer Stratification vanflhlfl
Not stratified* Unexposed Exposed Annual per capita income
