Parental Occupational Exposures to Chemicals 
and Incidence of Neuroblastoma in Offspring 

American Journal of Epidemiology Vol. 154, No. 2 2001

Anneclaire J. De Roos, ^1,6 Andrew F Olshan,¹ Kay Teschke,² Charles Poole,¹ David A. Savitz,¹ Julie Blatt,³ Melissa L. Bondy,¹ and Brad H. Pollock⁵

To evaluate the effects of parental occupational chemical exposures on incidence of neuroblastoma in offspring, the authors conducted a multicenter case-control study, using detailed exposure information that allowed examination of specific chemicals. Cases were 538 children aged 19 years who were newly diagnosed with confirmed neuroblastoma in 1992-1994 and were registered at any of 139 participating hospitals in the United States and Canada. One age-matched control for each of 504 cases was selected through random digit dialing. Self-reported exposures were reviewed by an industrial hygienist, and improbable exposures were reclassified. Effect estimates were calculated using unconditional logistic regression, adjusting for child's age and maternal demographic factors. Maternal exposures to most chemicals were not associated with neuroblastoma. Paternal exposures to hydrocarbons such as diesel fuel (odds ratio (OR) = 1.5; 95% confidence interval (CI): 0.8, 2.6), lacquer thinner (OR = 3.5; 95% CI: 1.6, 7.8), and turpentine (OR = 10.4; 95% CI: 2.4, 44.8) were associated with an increased incidence of neuroblastoma, as were exposures to wood dust (OR = 1.5; 95% CI: 0.8, 2.8) and solders (OR = 2.6; 95% CI: 0.9, 7.1). The detailed exposure information available in this study has provided additional clues about the role of parental occupation as a risk factor for neuroblastoma. Am J Epidemiol 2001;154:106-14.

Key words: child; dust; hydrocarbons; metals; neuroblastoma; occupational exposure

Received for publication September 8, 2000, and accepted for publication January 31, 2001.
Abbreviations: CI, confidence interval; IH, industrial hygienist; OR, odds ratio.
¹ Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC.
² Department of Health Care Epidemiology, Faculty of Medicine, University of British Columbia, Vancouver, British Columbia, Canada.
³ Division of Pediatric Hematology/Oncology, School of Medicine, University of North Carolina, Chapel Hill, NC.
⁴ Department of Epidemiology, M. D. Anderson Cancer Center, University of Texas, Houston, TX.
⁵ Department of Health Policy and Epidemiology, School of Medicine, University of Florida, Gainesville, FL.
⁶ Current address: Occupational Epidemiology Branch, National Cancer Institute, Bethesda, MD.
Reprint requests to Dr. Anneclaire J. De Roos, Children's Cancer Group, PO. Box 60012, Arcadia, CA 910666012 (e-mail: deroosa@mail.nih.gov ).

Understanding of the molecular biology and genetics of neuroblastoma has advanced considerably in recent years, but comparatively little is known about its possible environmental etiology. Several studies have examined the association of neuroblastoma with parental occupation (1-5). In these studies, exposures were usually inferred from job titles rather than from descriptions of duties or self-reported exposures. Associations with paternal work in farming and parental (maternal or paternal) work as an electrician or in electronics assembly and repair were reported (2, 3). Jobs

with potential exposure to dusts, metals, or hydrocarbons were associated with increased incidence in some studies (2, 3, 5) but not in others (1, 4, 5). Although results from these studies are suggestive, the studies were limited with respect to sample size and assessment of occupational exposures.

To evaluate the effects of parental occupational exposures on neuroblastoma incidence in offspring, we conducted a large, multicenter case-control study. Our first analysis examined incidence by parental job title (6). It found an increased rate of neuroblastoma for parents working as electrical power installers or power plant operators, for mothers working as hairdressers, and for fathers working as firefighters, landscapers, mechanics, painters, glass and optical goods workers, or broadcast, telephone, and dispatch operators. In this paper, we report results from further analysis of the same study population. We employed detailed exposure assessment that allowed estimation of the effects of specific occupational chemical exposures, using information on job title as well as on self-reported duties and exposures.

MATERIALS AND METHODS

Study population

Eligible cases were patients under the age of 19 years who had a confirmed new diagnosis of neuroblastoma between May 1, 1992, and April 30, 1994, and were registered at any of 139 participating hospitals in the United States and English-speaking Canada. The hospitals were members of either of two pediatric collaborative clinical trials groups: the Children's Cancer Group or the Pediatric Oncology Group. A total of 741 potentially eligible case children were identified. The treating physicians provided permission to approach the parents for participation in the study. Following the physician's consent, further criteria for inclusion of eligible cases were that the biologic mother was available for interview in either English or Spanish and that the case household had a telephone. Of the families contacted, we enrolled 538 cases (73 percent of those eligible). Nonparticipation of cases was primarily due to physician refusal (12 percent) and maternal refusal (8 percent). The 538 enrolled neuroblastoma cases had ages at diagnosis that ranged from newborn to 17 years, with a mean of 2.2 years; 56 percent of case children were male. N-myc copy number, determined by Southern blot (7) or fluorescence in situ hybridization (8), indicated amplification of the N-myc oncogene in 71 of 408 tumors analyzed (17 percent).

One control was selected for each of 504 cases through random digit dialing. We were unable to identify controls for the remaining 34 cases after multiple attempts. Controls were individually caliper-matched to cases on date of birth (±6 months for cases < 3 years of age, ±1 year for cases >3 years of age). The response proportion for the random digit dialing screening was 74 percent, calculated as the percentage of households that participated out of the total number selected for screening (9), and recruitment among eligible controls was 71 percent.

Data collection

A telephone interview was conducted with each mother, as well as with the father when he was available. We obtained interviews from 472 case fathers (88 percent of enrolled cases; 405 direct interviews and 67 proxy interviews with the mother) and 445 control fathers (88 percent of enrolled controls; 304 direct interviews and 141 proxy interviews with the mother). The interview included questions on demographic characteristics such as parental age, race, education, and income. Information on occupational history was obtained from each parent, including dates of employment, names of employers, occupations, industries, job titles, specific duties, and hours per week. For each job held during the 2-year period prior to the child's date of birth, mothers and fathers were asked if they had been exposed to chemicals, dusts, fumes, gases, vapors, or oils. Respondents could report exposure to any of 65 chemicals, compounds, or broad categories of substances listed in an interview guide that was provided. For each chemical exposure reported in each job, data were collected on the form of exposure (gas or vapor, dust or powder, smoke, liquid, or solid), the route of exposure (air, skin, clothing, swallowing), the work activities being performed during exposure (e.g., working near exposure, cleaning, wiping, cutting), the number of hours of exposure per week, and the dates on which exposure began and ended.

Information on occupational chemical exposure was available for 1,040 mothers (537 case mothers and 503 control mothers; 503 matched pairs) and 707 fathers (405 case fathers and 302 control fathers; 232 matched pairs). These numbers are lower than the total number of interviews completed, primarily because proxy respondents for fathers' interviews were not asked questions from the occupational history section on chemical exposures. Proxy-reported information on occupational exposure is subject to misclassification due to low sensitivity and low positive predictive value (10, 11); therefore, we did not consider it reliable. Additionally, two mothers and three fathers refused to answer the questions on occupational chemical exposure.

Exposure assessment

Parental occupational chemical exposure was coded in several ways.

Self-reported exposures.
A parent was coded as having been exposed to a substance if he or she reported exposure to that substance in any job.

Industrial hygienist's review of self-reported exposures. 
We reviewed each reported exposure scenario carefully under the supervision of an industrial hygienist (K. Teschke), so that any situation in which exposure was improbable was recoded as unexposed. All exposure coding was conducted without knowledge of subject's disease status or parent's gender. In community-based case-control studies in which the prevalence of most exposures is low (< 15 percent), incorrectly classifying even small numbers of individuals as exposed results in bias that substantially attenuates effect estimates toward the null value, if the misclassification is nondifferential and there is a true positive association; whereas an equal proportion of individuals incorrectly classified as unexposed will contribute little bias to the estimate (12-14). In order to minimize bias resulting from exposure misclassification, our industrial hygienist (IH) review focused on increasing the specificity of exposure measures by reducing the number of false-positive exposures. Jobs in which no chemical exposure was reported were not reviewed. All reported information for each exposure was reviewed, including occupation, industry, hours of exposure per week, form of the substance, route of exposure, use of protective equipment or clothing, and specific work activities. A parent was coded as having been exposed to a chemical if the IH review determined probable exposure in any job.

Chemical categories.
Chemicals were grouped into five broad categories to allow evaluation of the effect of chemicals with similar physicochemical characteristics and for comparability of results with the findings of previous studies. Data on exposures to sufficiently common specific chemicals were retained for analyses, but these substances were also grouped into the following categories: halogenated hydrocarbons; nonvolatile hydrocarbons; volatile hydrocarbons; paints, inks, and pigments; and metals, alloys, and solders (see table 1).

Coding of covariates

Demographic characteristics of interest included maternal education, maternal race, and maternal age at the birth of the index child. These factors were coded using indicator variables with the following categories: for maternal education, less than high school graduation, high school graduation and/or some college, and college degree or more (referent); for maternal race, White (referent), Black, Hispanic, and other; and for maternal age at the birth of the index child, < 18 years, 18-39 years (referent), and > 40 years.

TABLE 1. Categories of chemicals* evaluated in a study of parental occupational exposures and neuroblastoma in offspring, United States and Canada, 1992-1994

* Other chemicals analyzed but not included in categories: plastics, synthetics, or resins (NOS); animal fur or feather dust; cardboard dust; flour or starch dust; grain dust; rubber dust; and wood dust.
NOS, not otherwise specified.

Statistical analyses

Separate analyses were conducted for maternal and paternal exposures to each chemical. Where the overall frequency of IH-reviewed exposure among cases and controls was five or greater, we calculated exposure odds ratios to estimate the incidence rate ratios for neuroblastoma associated with self-reported and IH-reviewed exposure. Each chemical exposure was evaluated in a separate model, using persons unexposed to the chemical as the reference group. Exposure odds ratios were calculated using unconditional logistic regression modeling so that all matched and unmatched subjects could be included to enhance precision. Each model included variables for the matching factor, child's age, which was coded as a set of indicator variables (6-month intervals for ages <3 years; 2-year intervals for ages >3-11 years; one variable for ages > 11 years), as well as variables for the set of demographic covariates. We conducted separate analyses to evaluate associations between parental occupational chemical exposures and specific subtypes of neuroblastoma, as defined by age at diagnosis (<1 year vs. > 1 year) or N-myc oncogene status (amplified vs. normal).

We considered confounding among certain groups of paternal occupational chemical exposures that frequently occurred together (i.e., 10 or more fathers in the study were exposed to both of two chemicals). Frequent co-occurrence was common among the following groups of chemicals, which were evaluated in three separate models: 1) hydrocarbons with paints; 2) wood dust with paints and hydrocarbons; and 3) four types of steel alloys and solders (not otherwise specified) with lubricating and cutting oils. Confounding was examined by modeling all relevant exposures simultaneously, along with variables for child's age and the demographic covariates. Maternal exposures were too infrequent to allow examination of confounding by specific chemicals.

RESULTS

The distributions of selected characteristics in the entire study population and in the subset of fathers with exposure information are shown in table 2. Matching ensured an equitable distribution of child's age between cases and controls, although the distribution differed in the subset of fathers with chemical exposure information; specifically, the controls were slightly younger than these cases. Cases were more likely than controls to be male. Mothers of cases were more likely to be younger than 18 years or older than 40 years at the time of the child's birth, and they were less likely to be college graduates than were mothers of controls. Maternal race was similar for cases and controls.

The frequencies and effect estimates for maternal and paternal occupational chemical exposures are shown in tables 3 and 4. Chemicals with IH-reviewed exposure frequencies below five were not analyzed further and are therefore not listed in the tables, although they were included in the chemical category frequencies. Very few mothers reported having occupational exposure to chemicals. In addition, where maternal exposures were reported, they frequently occurred in activities that were less likely to have high exposure; therefore, these mothers were recoded as unexposed. For both mothers and fathers, the frequency of reporting of individual chemical compounds (e.g., benzene) was not as high as the reporting of chemical products with well-known common names (e.g., gasoline). Given that individual chemical compounds such as benzene constitute a small component of chemical products such as gasoline, these individual chemical compounds are likely to have been underreported. The exposure frequencies were adequate for analyses of 46 paternal exposures and 17 maternal exposures to chemicals or categories. Because there were virtually no differences between crude and adjusted effect estimates, only adjusted estimates are presented in the tables. Effects of parental occupational chemical exposures on the incidence of neuroblastoma in offspring were evaluated primarily by means of the adjusted odds ratios for IH-reviewed exposure.

TABLE 2. Characteristics of the study population, by child's disease status, in a study of parental occupational chemical exposures and neuroblastoma in offspring, United States and Canada, 1992-1994

* Numbers in parentheses, standard deviation.

There was little evidence that any of the maternal exposures were associated with an increased incidence of neuroblastoma in offspring (see table 3), although there was an excess of case mothers who had been exposed to animal fur or feather dust (five cases and no controls).

Paternal occupational exposures to several chemicals were associated with increased incidence of neuroblastoma in offspring (see table 4). Odds ratios for several hydrocarbons were 1.5 or above, including alcohols, benzene, cutting oil, diesel fuel, lacquer thinner, mineral spirits, paint thinner, and turpentine. Odds ratios for the categories of volatile hydrocarbons (odds ratio (OR) = 1.5; 95 percent confidence interval (CI): 1.0, 2.1) and nonvolatile hydrocarbons (OR = 1.5; 95 percent CI: 1.0, 2.2) were elevated as well. The data indicated positive associations between neuroblastoma and paternal exposures to certain metals, alloys, and solders, including stainless steel, mild steel, galvanized steel, high-speed steel, brass, metals (not otherwise specified), and solders (not otherwise specified), although exposure to metals, alloys, and solders as a category was not strongly associated with neuroblastoma. There was a 50 percent increased incidence of neuroblastoma associated with paternal exposure to wood dust. Of the other dust exposures analyzed, odds ratios were elevated for grain dust and rubber dust, although the estimates were extremely imprecise. When separate analyses were conducted by N-myc oncogene status or age at diagnosis, there did not appear to be a specific subtype of neuroblastoma with which paternal occupational chemical exposures were more or less strongly associated (results not shown).

TABLE 3. Estimated effects of maternal occupational exposures to chemicals on the incidence of neuroblastoma in offspring,* 
United States and Canada, 1992-1994

* The analysis included 537 case mothers and 503 control mothers. 
Main entries indicate categories of chemicals analyzed as groups (from table 1). 
IH, industrial hygienist; CI, confidence interval. 
§ Adjusted for child's age, maternal race, maternal age, and maternal education.

Analyses of paternal exposures that frequently occurred together revealed some confounding between different groups of chemicals. When data were simultaneously adjusted for paternal exposures to frequently co-occurring hydrocarbons and paints, odds ratios for alcohols (OR = 1.6; 95 percent CI: 0.7, 3.4), diesel fuel (OR = 2.0; 95 percent CI: 1.0, 4.3), lacquer thinner (OR = 2.9; 95 percent CI: 1.0, 8.3), mineral spirits (OR = 1.7; 95 percent CI: 0.6, 4.5), and turpentine (OR = 12.0; 95 percent CI: 2.2, 65.9) remained elevated, whereas the odds ratio for paint thinner was diminished (OR = 0.9; 95 percent CI: 0.4, 2.1). The odds ratio for paternal exposure to wood dust was elevated (OR = 1.6; 95 percent CI: 0.8, 3.0), even after adjustment for potential confounding by exposure to hydrocarbons or paints. When four steel alloys and solders (not otherwise specified) were included in the same model as cutting and lubricating oils, the effect estimate for solders remained moderately elevated (OR = 2.5; 95 percent CI: 0.8, 7.6), whereas the estimates for the steel alloys were too imprecise for evaluation of confounding.

DISCUSSION

There are several mechanisms by which parental chemical exposures might increase the risk of neuroblastoma in offspring. According to Knudson's two-stage model of carcinogenesis, a minimum of two mutations or inactivating events in a crucial gene are required for carcinogenesis (15, 16). These mutations could occur during gametogenesis in the mother or father and would then be inherited by the child, or they could occur sporadically in target tissues of the developing fetus or child. Exposures of concern to the mother or father are likely to occur at different time points relative to conception. During the period of gametogenesis prior to conception, paternal exposures could cause a germline mutation in sperm DNA. There are human data demonstrating that paternal exposures can cause mutations in sperm DNA, and animal models provide some evidence that paternal exposures can increase the risk of cancer in offspring via mutations in the germ cell line (17-19). Maternal exposures are of particular concern during pregnancy, because in-utero exposure of the fetus could occur by means of chemicals crossing the placenta. Although we wished to examine the effects of parental occupational chemical exposures during the specific time periods of preconception and pregnancy, exposures overlapped the different periods and therefore did not support comparative analyses.

TABLE 4. Estimated effects of paternal occupational exposures to chemicals on the incidence of neuroblastoma in offspring,* United States and Canada, 1992-1994