|Home | About | Journals | Submit | Contact Us | Français|
Several case-control studies have evaluated associations between maternal smoking, alcohol consumption and illicit drug use during pregnancy and risk of childhood leukemia. Few studies have specifically focused on infants (<1 year) with leukemia, a group that is biologically and clinically distinct from older children. We present data from a Children’s Oncology Group case-control study of 443 infants diagnosed with acute leukemia (including acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML)) between 1996–2006 and 324 population controls. Mothers were queried about their cigarette, alcohol, and illicit drug use one year before and throughout pregnancy. Odds ratios (OR) and 95% confidence intervals (CI) were calculated using adjusted unconditional logistic regression models. Maternal smoking (>1 cigarette/day) and illicit drug use (any amount) before and/or during pregnancy were not significantly associated with infant leukemia. Alcohol use (>1 drink/week) during pregnancy was inversely associated with infant leukemia overall [OR = 0.64; 95% CI = 0.43, 0.94], AML [OR = 0.49; 95% CI = 0.28, 0.87], and leukemia with mixed lineage leukemia (MLL) gene rearrangements (“MLL+”) [OR = 0.59; 95% CI = 0.36, 0.97]. While our results agree with the fairly consistent evidence that maternal cigarette smoking is not associated with childhood leukemia, the data regarding alcohol and illicit drug use are not consistent with prior reports and are difficult to interpret. It is possible that unhealthy maternal behaviors during pregnancy, some of which carry potential legal consequences, may not be adequately measured using only self-report. Future case-control studies of childhood leukemia that pursue these exposures may benefit from incorporation of validated instruments and/or biomarkers when feasible.
With an annual incidence of about 40 cases per million children in the United States, leukemia is the second most common cancer in infants (<1 year of age) after neuroblastoma1. Infants with leukemia commonly have genetic translocations involving the mixed lineage leukemia (MLL) gene in their leukemia cells. MLL rearrangements (MLL+) are found in approximately 80% of infant ALLs and 60% of infant AMLs, but in only 5% of all childhood leukemias diagnosed after 1 year of age and 10% of adult AMLs.2–4 MLL status is believed to correspond to distinct etiologies, partly because MLL+ infant ALL cases typically have much poorer survival than MLL− cases.5, 6 Studies of identical twins7 and stored neonatal blood spots8, 9 have indicated that MLL rearrangements are initiated in utero. Therefore, studies of infant leukemia have focused on in utero exposures as potential contributing factors in leukemogenesis.
Many previous epidemiological studies of childhood leukemia have investigated possible connections between maternal cigarette smoking and/or maternal alcohol consumption. Most studies of maternal smoking and childhood leukemia have found no association.4, 10 For maternal alcohol consumption, results from a 2010 meta-analysis11 indicated that the risk of childhood AML, but not ALL, may increase with use during pregnancy. While there was heterogeneity between studies, the summary estimate for AML [OR = 2.7; 95% CI = 1.9, 3.9] was considerably more homogenous when restricted to children ages 0–4 years (P heterogeneity= 0.76). The relationship between maternal illicit drug use and childhood leukemia is less clear. A positive association between maternal marijuana use during pregnancy and childhood AML, observed in 1989,12 was later contradicted by results from a 2006 study13 that detected an inverse association.
Very few studies have specifically explored infant leukemia. Shu et al.14 found an inverse association for maternal smoking and a positive association for alcohol drinking, while Alexander et al.15 reported null findings for both. We present data from the largest case-control study of infant leukemia conducted to date in order to add to the limited amount of literature for these exposures.
Infants diagnosed with acute leukemia at <1 year of age were identified at Children’s Oncology Group (COG) institutions between January 1, 1996 and October 13, 2002 (phase I) and between January 1, 2003 and December 31, 2006 (phase II). Cases were eligible if they did not have Down syndrome, if their biological mother was available for a telephone interview in English or Spanish (phase II only), and if they were diagnosed or treated at a participating COG institution in the U.S. or Canada. Mothers of deceased cases were eligible for the study. Further aspects of subject recruitment and data collection have been described previously.16–18
The control recruitment strategies differed between the two phases described above. During phase I, controls were selected from the U.S. and Canada using a standardized random digit dialing (RDD) procedure 19, 20 and were frequency matched to cases on year of birth. Phase II controls were randomly selected from 15 U.S. state birth registries and were frequency matched to cases on year of birth and region of residence based on the geographical and annual birth distribution of phase I cases.17 Eligible controls did not have Down syndrome and had a biological mother who was available for a telephone interview in English or Spanish (phase II only).
Case and control mothers provided exposure data via structured, computer-assisted telephone interviews. Interviews were completed for 443 (64%) eligible cases (264 ALL, 172 AML, and 7 biphenotypic and acute undifferentiated leukemia) and 324 (47%) eligible controls. Participation rates and reasons for non-participation in phases I and II have been published elsewhere.16, 17, 21 Interview questions included items on demographics, reproductive history, family history of disease, and exposures during pregnancy with the index (case or control) child. For cases, pathology and cytogenetic reports from diagnosis were acquired, and MLL gene rearrangement status (MLL+, MLL−, or indeterminate) was ascertained through central review by three independent reviewers, as previously reported.21
The interview queried current and past use of cigarettes, alcohol, and illicit drugs such as marijuana, heroin, and cocaine. Questions focused on use in the year prior to pregnancy, in early pregnancy before mothers knew they were pregnant, and from the time mothers found out they were pregnant until the index child was born. By distinguishing between the latter two time periods, we aimed to decrease underreporting of socially undesirable behaviors, at least during early pregnancy.
Institutional review boards at the University of Minnesota and participating COG institutions approved all study procedures. The study was also approved by health departments for states providing birth certificates, as applicable. All mothers provided informed consent prior to participation.
Differences between cases and controls in potential confounders and other baseline characteristics were evaluated using Pearson’s chi-square test for categorical variables and the t-test for two means for continuous variables. We compared controls to all infant leukemias combined and to subgroups defined by leukemia type (ALL and AML) and MLL status (MLL+ and MLL−).
Unconditional logistic regression was used to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for the associations between maternal cigarette, alcohol, and illicit drug use and development of infant leukemia. Smokers were defined as those who smoked >1 cigarette per day, drinkers were those who consumed >1 alcoholic drink per week, and illicit drug users were those who reported any illicit drug use during the time period of interest. Dichotomous variables were created as there was no linear trend (i.e., dose response) for exposure frequency. Mothers who met the minimum criteria for smoking or drinking in at least one portion of pregnancy (i.e. before and/or after knowledge of their pregnancy) were classified as smokers or drinkers during pregnancy overall. Drug use prior to pregnancy and during pregnancy were not evaluated separately due to small cell counts. Reference groups included both never smokers/drinkers and ever smokers/drinkers not reporting use during the relevant time periods; similar results were obtained when only never smokers/drinkers were used as reference groups.
All analyses were repeated upon stratification by leukemia type and MLL status. The regression models for smoking were adjusted for maternal age (continuous), education, race/ethnicity and alcohol use during pregnancy (yes or no), household income, and child’s birth year (ordinal), a matching factor. For alcohol consumption, models were adjusted for maternal education and race/ethnicity and child’s birth year, while for drug use, models were adjusted for maternal age and child’s birth year. Child’s sex, maternal smoking, maternal household use of pesticides, and region of residence were also considered as potential confounders but were not included in final models because their inclusion did not change the natural log odds ratio estimates by >10%. All analyses were conducted using SAS software version 9.2 (SAS Institute, Inc., Cary, NC).
Table 1 presents descriptive characteristics for controls, all cases combined, and for ALL and AML subgroups. Cases and controls were similar with respect to child’s sex, household income, and mother’s education level. A greater percentage of cases were non-white (24% of cases vs. 16% of controls), and mothers of ALL cases were somewhat younger at the time of the child’s birth compared to mothers of controls (meancases = 28.7 years vs. meancontrols = 29.8 years).
Adjusted ORs and 95% CIs for maternal prenatal cigarette, alcohol, and illicit drug use are shown in Table 2. Cigarette smoking and drug use were not associated with infant leukemia in any of the models. Alcohol use during pregnancy was inversely associated with infant leukemia overall [OR = 0.64; 95% CI = 0.43, 0.94], AML [OR = 0.49; 95% CI = 0.28, 0.87], and MLL+ leukemia [OR = 0.59; 95% CI = 0.36, 0.97]. Similar associations were observed amongst AML cases for alcohol use in the year prior to pregnancy [OR = 0.67; 95% CI = 0.44, 1.00] and use anytime in the year prior to and/or during pregnancy [OR = 0.66; 95% CI = 0.44, 0.99].
Maternal cigarette, alcohol, and illicit drug use before and during pregnancy were evaluated as potential risk factors for infant leukemia. No statistically significant associations were observed for cigarette smoking and drug use. In contrast, alcohol use during pregnancy was inversely associated with infant leukemia overall and specifically among AML and MLL+ cases.
The absence of any association between maternal cigarette smoking and leukemia mirrors most previous findings, including those in infants.15 Only six of 25 recently-reviewed10 studies reported significant but inconsistent associations between maternal smoking and childhood leukemia. In light of this evidence, the International Agency for Research on Cancer recently concluded that maternal cigarette smoking is not causally related to childhood leukemia.22
For maternal alcohol use, a recent meta-analysis11 of 21 case-control studies reported a positive association with childhood AML [OR = 1.56; 95% CI = 1.13, 2.15] but not ALL [OR = 1.10; 95% CI = 0.93, 1.29] nor “grouped leukemias” [OR = 1.11; 95% CI = 0.88, 1.40]. The increased risk with AML was particularly strong for cases diagnosed at <5 years of age. Two studies observed inverse associations similar to ours, although children were 0–1423 or 0–924 years of age instead of exclusively infants. One of these studies speculated that chance likely played a role,23 while the other suggested two distinct potential biological mechanisms.24 First, alcohol may inhibit growth hormone/insulin-like growth factor (IGF-1), which could lead to lower birthweights and a reduced risk of childhood leukemia. Including birthweight in our models did not change effect estimates. A second plausible mechanism involves the cancer chemopreventive activity of flavonoids (i.e., antioxidants) contained in red wine and hopped beer.25–27 We performed a post-hoc analysis stratified by alcohol type (data not shown). Drinking two or more servings of beer and/or red wine per week during pregnancy was inversely associated with all of the leukemia strata, while liquor/spirits showed no associations. Thus, our overall results may lend some support to the proposed flavonoid hypothesis.
Few studies have explored maternal illicit drug use and childhood leukemia. Maternal use of marijuana prior to or during pregnancy was positively associated with childhood AML in one study12 and ALL in another.28 More recently, a study designed to specifically test the hypothesis that marijuana use was associated with an increased risk of AML actually observed an inverse association.13 While our study results did not reach statistical significance, there was a suggestion of an inverse association with any illicit drug use across most subtypes.
It is difficult to interpret our results in the context of other studies, especially in light of discrepant results. Notably, a case-control study of infant leukemia (< 18 months of age) conducted in the 1980s reported a positive association with maternal alcohol consumption14. In either study, misclassification of exposure could be an issue. Of note, misclassification of alcohol use was reported in 45% of pregnant women recalling first trimester use in their 7th month of pregnancy and at delivery.29 This percentage could be even greater in our study population since the maternal interviews took place nearly three years on average after the index child’s birth. We attempted to reduce misclassification by focusing on the periods prior to and after knowledge of pregnancy, although results were similar. It may be fruitful for future case-control studies like ours to consider consistency in questionnaires (ideally they would be validated questionnaires) in order to more readily compare results across studies.
There is also concern regarding recall bias, especially with regard to exposures that may be perceived as harmful. In several states, alcohol and illicit drug use are subject to mandatory reporting by health professionals during pregnancy.30 Although our observational study was protected by a Certificate of Confidentiality from the federal government, mothers of infants with leukemia may have been reluctant to report certain exposures for fear of repercussions. In support of this contention, one study found that mothers of sick infants might be more apt to deny alcohol and cigarette use compared to mothers of healthy children.31 This differential misclassification could lead to odds ratios that are biased in unpredictable directions.32 Due to concerns regarding reporting of these types of exposure, it may be useful to validate certain exposures in future studies, at least in a subset of cases. For example, dried blood spots, collected shortly after birth to test for metabolic and other disorders,33 are kept in long-term storage by several health departments across the United States, potentially permitting their use in etiologic research.34 Several analytes, including cotinine in cigarette smoke35 and cocaine,36 have been measured in dried blood spots. Thus, with proper consideration of storage conditions, analyte half-life and stability, etc., newborn dried blood spots could potentially be used to provide an independent measure of certain exposures in case-control studies such as ours.37
Selection bias is another potential concern, however, the use of COG institutions for case ascertainment likely minimized selection bias among cases since these institutions treat nearly all leukemia patients diagnosed in the U.S. at <5 years.38 With a response rate of 64%, however, participating cases may be fundamentally different than cases who chose not to participate. The same concern exists among controls, who had a response rate of only 47%. Previously, both the RDD and birth certificate controls from our study were compared to U.S. National Center for Health Statistics data for all births in the year 2000.17 Compared to the U.S. population, control mothers were older, more often white, married, and had more years of education. Control children were more likely to be born at term and to weigh more at birth. Case and control mothers in our study were similar to each other on all measured demographic characteristics except race/ethnicity, as control mothers were more often white. This difference is of interest because, in general, white women consume alcohol more often than women of other racial/ethic groups. In the 2009 National Survey on Drug Use and Health,39 60% of white women reported drinking alcohol in the past month; only 45% of black and Hispanic women and 40% of Asian, American Indian, and Alaska Native women reported drinking. While we adjusted for race/ethnicity, residual confounding may remain.
In an ad hoc sensitivity analysis, we evaluated whether any differences occurred between the two study phases. Although the associations for alcohol use before pregnancy did not appear to change over time, the association between alcohol use during pregnancy and infant leukemia was attenuated in phase II. It is unclear whether this pattern is due to an actual trend, but data from the National Survey on Drug Use and Health (formerly called the National Household Survey on Drug Abuse)39 do not suggest any obvious trends in the prevalence of alcohol use during pregnancy over the past decade. These findings may instead be due to chance, especially given small cell counts. Nevertheless, our overall study results were driven primarily by the first phase of the study.
In conclusion, we found no evidence of an association with cigarette smoking and infant leukemia. In contrast to some previous reports, however, we found evidence of an inverse association between maternal alcohol use during pregnancy, and no evidence of an association with illicit drug use. Due to conflicting findings across the literature and the potential for recall bias, it is important for future studies to consider consistency in questionnaires (ideally validated), as well as an independent measure of exposure, such as prediagnostic biological samples when feasible.
The authors would like to thank Michelle A. Roesler for her helpful comments and suggestions.
Funding: Research supported by National Institutes of Health Grants R01 CA79940, T32 CA99936, U10 CA13539, and U10 CA98543, U10 CA98413, P30 CA77598 (University of Minnesota Masonic Cancer Center shared resource: Health Survey Research Center), and the Children’s Cancer Research Fund, Minneapolis, MN.