Parental Tobacco and Alcohol Use and Risk of Hepatoblastoma in Offspring: A Report from the Children's Oncology Group Free

Hepatoblastoma is a rare pediatric liver cancer affecting about 100 children each year in the United States (1). The incidence of hepatoblastoma has increased significantly over the period from 1992 to 2010 at a rate of approximately 3% per year (2). The increasing incidence of hepatoblastoma is contemporaneous to an increase in the prevalence of low birth weight (<2,500 g), which has been strongly associated with hepatoblastoma (3, 4). The biologic mechanism underlying the association between low birth weight and hepatoblastoma is unknown. Known risk factors for hepatoblastoma are limited to the inherited cancer syndromes Familial Adenomatous Polyposis and Beckwith–Wiedemann syndrome (3).

Evidence from case–control studies (the largest consisting of 155 cases) has recently accumulated that parental smoking may be associated with hepatoblastoma. Six studies examined maternal and/or paternal smoking before conception and during pregnancy and although the associations differed with respect to the time period and parent conferring risk (5–10), three studies reported significant ORs ranging from 2.1 to 4.7 (6, 8, 9). On the basis of results from some of these studies, the International Agency for Research on Cancer (IARC) recently classified hepatoblastoma as a tobacco-related cancer (11).

Although alcohol indisputably contributes to liver cancer in adults, little evidence has been generated about a potential association between maternal alcohol use and hepatoblastoma. Two small case–control studies and one case report have considered the possibility of an association, but results were not significant (7, 12, 13). Because byproducts and metabolites of alcohol are known or suspected carcinogens (14), and are known to cross the placenta (15, 16), further investigation in a larger study is warranted.

Here, we examined parental tobacco and alcohol use before and during pregnancy in the largest case–control study of hepatoblastoma to date.

Characteristics of the study population and design have been previously reported (17, 18). Cases were identified through the Children's Oncology Group (COG). Controls were identified through 32 U.S. birth registries and frequency-matched to cases on birth weight (<1,500, 1,500–2,500, and >2,500 g), gender, birth year, and region. After informed consent, participating mothers completed a computer-assisted telephone interview that gathered information about events and exposures around the time of pregnancy. Eligibility criteria for cases included having confirmed diagnosis of hepatoblastoma from a United States COG institution between 2000 and 2008, diagnosis before the age of 6 years, being born in the United States, and having an English- or Spanish-speaking birth mother available for a phone interview. Deceased cases were eligible for inclusion. Controls were eligible if born in the United States between 1994 and 2008, and if they had an English- or Spanish-speaking birth mother available for a phone interview.

We focused on interview questions about parental use of tobacco and alcohol. Specifically, we examined four questions assessing mothers' tobacco-smoking habits (any history of smoking one cigarette or more; smoking within the year before pregnancy; smoking early in pregnancy before knowing about pregnancy; and smoking during pregnancy, up until giving birth) and four questions assessing their alcohol-drinking habits (any history of consuming at least two drinks per month for 1 year or more; drinking within the year before pregnancy; drinking early in pregnancy before knowing about pregnancy; and drinking during pregnancy, up until giving birth). We derived new variables to examine dose response for maternal cigarette smoking. For three time periods (within the year before pregnancy; early in pregnancy before knowing about the pregnancy; and during pregnancy, up until giving birth), we considered the effects of no smoking, smoking fewer than 10 cigarettes per day, smoking between 10 and 14 cigarettes per day, and smoking 15 or more cigarettes per day. We also analyzed fathers' smoking habits that were reported by the study subject's mother (smoking within the year before the index pregnancy; smoking during the index pregnancy). We did not collect data on paternal drinking or paternal cigarettes per day. Finally, we derived new variables to consider the combined effect of both parents smoking 1 year before the index pregnancy and during the pregnancy.

All statistical analyses were conducted using Statistical Analysis Software version 9.3 (SAS Institute, Cary, NC). Adjusted ORs and 95% confidence intervals (CI) were calculated using logistic regression. Associations between hepatoblastoma and parental smoking and alcohol exposures were modeled separately. All models were adjusted for the matching variables (index child's birth weight, year of birth, and sex) as well as maternal race and educational attainment. Subjects with missing data were excluded from analyses. Effect modification on the multiplicative scale was evaluated in logistic regression models by including an interaction term for the exposure and the third variable of interest. Wald χ² tests were used to evaluate the statistical significance of main effects and interaction terms. Associations were considered statistically significant at a P value 0.05 or less.

Of the 408 consenting cases identified from COG institutions, 383 completed a telephone interview. Of the 5,813 control mothers identified from birth registries, we were able to reach 754 of 1,718 control mothers with whom contact was attempted. Of the 754 control mothers that were reached, 387 completed the telephone interview.

Cases and controls were frequency matched on sex and birth weight resulting in a similar percentage of female cases and controls; however, the frequency matching worked less well for low birth weight resulting in a greater percentage of cases that had low birth weight (especially ranging from 1,500 to 2,499 g) than controls. Case mothers reported less education (29% of cases attaining high school or less vs. 22% in controls) and lower incomes (31% of cases earning <$30,000 vs. 23% in controls) than control mothers. A lower percentage of case mothers were White as compared with control mothers (69% vs. 75%), and a higher percentage of case mothers than control mothers were Hispanic (19% vs. 9%; Table 1).

We found no associations between maternal smoking and hepatoblastoma in the offspring for any history of maternal smoking (OR, 1.0; 95% CI, 0.7–1.4), smoking within the year before pregnancy (OR, 1.1; 95% CI, 0.8–1.6), smoking early in pregnancy (OR, 1.0; 95% CI, 0.7–1.6), or smoking up until giving birth (OR, 0.9; 95% CI, 0.5–1.6). We found no evidence of a dose–response relationship for maternal smoking and hepatoblastoma. There was no statistically significant association between the number of cigarettes smoked and hepatoblastoma, whether the smoking occurred in the year before pregnancy, early in the pregnancy, or throughout the pregnancy (data not shown). We also observed no positive associations between hepatoblastoma and maternal drinking for any of these time periods with nonsignificant ORs ranging from 0.8 to 1.0 (Table 2).

Paternal smoking within the year before the pregnancy was positively associated with hepatoblastoma in the offspring (OR, 1.4; 95% CI, 1.0–2.0; P = 0.06). Similarly, mothers reported that case fathers were more likely to be smokers during the pregnancy than control fathers (OR, 1.4; 95% CI, 0.9–2.0); however, neither of these associations reached statistical significance. Both parents smoking 1 year before pregnancy and during the pregnancy showed weakly positive nonsignificant associations with hepatoblastoma (Table 2).

Finally, we evaluated whether there was any evidence of effect modification by race between parental smoking and drinking during the peri-gestational period (Table 2). The results were stronger for most parental smoking associations in non-Hispanic Whites, most notably for both parents smoking within the year before pregnancy; however, the interaction term was not significant (P = 0.17).

Three of six prior studies have suggested that parental smoking increases risk of hepatoblastoma in offspring, although associations were not consistent with respect to parent or time period (Table 3). Significant increased risks for hepatoblastoma in the offspring were reported for mothers and for both parents who smoked before pregnancy in one childhood cancer study out of the United Kingdom (6). In contrast, a case–control study of 75 cases and 75 controls in the United States and Canada found no association between maternal or paternal preconception smoking and hepatoblastoma (7). Parental smoking of the mother (5–7) or father (7) during pregnancy was also not associated with hepatoblastoma in these studies. Significant positive associations were reported between hepatoblastoma and maternal smoking during an unspecified time period in three studies (8–10), and for paternal smoking and both parents smoking in one study (8). Our large study that comprehensively evaluated maternal, paternal, and both parents smoking at several different developmental time periods provides no evidence of a role of maternal smoking in hepatoblastoma, and only weak evidence for a role of paternal preconception smoking.

It is interesting to note from the small body of literature on this topic described in Table 3, some differences in results depending on study location and data source. In studies that collected exposure data by parental interview (UKCSS, OCSS, and CCSG), only the U.K. studies (UKCSS and OCSS) reported positive associations. Varying results between the U.K. studies and the North American CCSG study could stem from differential reporting by participants from North America versus the United Kingdom due to cultural differences in the acceptability of smoking during the peri-gestational period between the two geographic locations. With respect to data source, the results are inconsistent. Assessment of maternal prenatal smoking by birth certificate before knowledge of disease in the U.S. NYSCR study would presumably eliminate the concern that the reported positive association was an artifact due to differential reporting between case and control mothers (19). However, the larger Nordic study that also collected data on maternal smoking by birth certificate before disease did not show an association (5).

Biologic plausibility for a causal role of paternal preconception smoking in hepatoblastoma stems from known DNA mutagens contained in cigarette smoke (20) that could cause heritable genetic aberrations in sperm. Cigarette smoking has been associated with oxidative sperm DNA damage in some (21–23) but not all studies (24) and evidence from animal models have shown that mice exposed to mainstream tobacco smoke can develop mutations at the Ms6-hm tandem repeat locus that can be passed on to their offspring (25, 26). However, a declining U.S. population prevalence of smoking over the last couple of decades contemporaneous to an increasing hepatoblastoma incidence argues against a major role for paternal smoking in hepatoblastoma etiology (27).

We found no evidence for an association between maternal drinking and hepatoblastoma. Two small studies (<100 cases each) also did not find any significant associations between parental alcohol consumption and hepatoblastoma (7, 12). In the CCSG study of 75 case–control pairs from the United States and Canada, a positive association was reported between maternal alcohol use at the time of pregnancy and hepatoblastoma (OR, 1.9; ref 7), but no association with paternal alcohol use at the time of pregnancy (OR, 1.0). A brief case report described a 27-month-old child diagnosed with hepatoblastoma whose mother drank heavily during pregnancy, but other substances were also ingested and the child had numerous concurrent birth defects and health complications (13). Taken together, there is no support for a major role for parental alcohol consumption in hepatoblastoma etiology.

Major strengths of our study include its large size, high participation rate by cases, and pathologically confirmed cases. In addition, data collection was uniform between cases and controls, minimizing the possibility of interviewer bias.

Methodologic limitations of case–control studies include the potential for selection, information, and overmatching biases. Selection bias was not evident in our study compared with U.S. population–based data on the prevalence of smoking and drinking around the time of pregnancy. The Centers for Disease Control and Prevention's (CDC) Pregnancy Risk Assessment Monitoring System estimated that between 2000 and 2008, 23% of women nationwide smoked during the 3 months before pregnancy (28), similar to the 22.1% prevalence of smoking among controls within the year before pregnancy in our study. Data from the Behavioral Risk Factor Surveillance System in 2010 found that 7.6% of pregnant women reported alcohol use in the past 30 days (29), which is also similar to the control mothers in our study in which 7.9% reported drinking during pregnancy. However, we note that evidence for selection bias was detected in a sensitivity analysis of birth certificate demographic data from nonparticipating eligible control mothers; participants were significantly older, more White, and had higher education levels than nonparticipants (data not shown). Although we did not obtain information on smoking status of nonparticipants, the prevalence of smoking has been reported to be lower among older and more educated moms (30). This would not be a likely explanation for our failure to find evidence for a positive association between maternal smoking on hepatoblastoma in the offspring, as the effect of higher participation by nonsmoking mothers would be to bias the ORs away from the null.

Case–control studies are also prone to information biases including reporting bias, particularly nondisclosure and recall bias. Nondisclosure of smoking among pregnant women has been reported to be high in the United States (31). The effect of nondifferential nondisclosure of smoking during the peri-gestational period by both cases and controls would tend to bias results toward the null. This could explain why the U.K studies found positive associations between parental smoking and hepatoblastoma, whereas ours and the other North American study that collected data by interview did not. In addition, recall bias could have influenced our results if case mothers reported very trivial amounts of smoking and drinking during the pregnancy with the index child compared with control mothers. Although either source of reporting bias may be present in our results, it is not possible to detect this type of bias with our study design.

Finally, it is also possible that overmatching bias may have impacted our findings on maternal smoking during pregnancy and hepatoblastoma, as we matched cases and controls on birth weight and both maternal smoking and hepatoblastoma have established associations with low birth weight (3, 32). The effect of this type of bias on the risk estimates for associations between maternal smoking and hepatoblastoma would be toward the null.

In conclusion, in the largest case–control study to date of hepatoblastoma, we found little evidence of an association between maternal smoking and hepatoblastoma in the offspring, and weak evidence of an association with paternal smoking. We also found no association with maternal alcohol use during pregnancy. Although our study does not support or entirely rule out IARC's conclusion that hepatoblastoma is a tobacco-related cancer, it remains wise to counsel prospective parents on the merits of smoking cessation.

No potential conflicts of interest were disclosed.

Conception and design: J.A. Ross, M.H. Malogolowkin, L.G. Spector

Development of methodology: K.J. Johnson, K.S. Williams, J.A. Ross, L.G. Spector

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.D. Krailo, G.E. Tomlinson, L.G. Spector

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.J. Johnson, K.S. Williams, J.A. Ross, J.H. Feusner, L.G. Spector

Writing, review, and/or revision of the manuscript: K.J. Johnson, K.S. Williams, J.A. Ross, G.E. Tomlinson, M.H. Malogolowkin, J.H. Feusner, L.G. Spector

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.D. Krailo, G.E. Tomlinson, L.G. Spector

Study supervision: L.G. Spector

This work was supported by NIH grants R01CA111355, K05 CA157439, U10CA13539, U10CA98543, and the Children's Cancer Research Fund, Minneapolis, Minnesota.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.