A smoking gun?

Maternal smoking during pregnancy is of interest in the etiology of childhood leukemia because the toxic byproducts of tobacco smoke are known to cross the placenta and a proportion of leukemia is known to originate in utero. Research on the subject consists mainly of case-control studies, which are susceptible to recall bias. Smoking during pregnancy is particularly difficult to study in this manner since social disapproval of the practice may lead to misreporting by mothers. Cohort studies are preferable but require following far more children than is usually practical. However, the Nordic nations are in an enviable position because of their ability to link population-based data from birth and disease registries [see Finnish Envy; C3 Vol. 15, No. 1]. In the latest example of the power of this resource Mucci L and colleagues [Cancer Epidemiol Biomarkers Prev 2004; 13(9): 1528-1533] describe the relationship between maternal smoking during pregnancy and incidence of hematological malignancies in a cohort of nearly 1.5 million Swedish children.

The study population consisted of children born in 1983-1997, inclusive, with records in the Swedish Medical Birth Registry (i.e. >99% of Swedish births). Their birth records were merged with data from the population-based Swedish Cancer Registry and the National Cause of Death Registry. About 90% of the birth cohort, or 1,440,542 children, were available for analysis after exclusions of infants who died in the first week of life, children with Down syndrome, and children with missing data. The length of observation for each child was calculated from the date of birth to the occurrence of a hematological malignancy, the occurrence of another cancer, death, or the end of the study period. Maternal smoking status was determined at the first prenatal visit and classified into non-smokers and smokers of 1-9 or >10 cigarettes/day. Other information taken from the birth record and considered as potential confounders was maternal age, maternal education, parental cohabitation status, rural residence, maternal Nordic ethnicity maternal parity, birth year, and child’s gender. Birth weight and gestational age were investigated separately, since they may be related to cigarette smoking and therefore could be in the causal pathway. The researchers used proportional hazards modeling to obtain the hazard ratios (HR) and 95% confidence intervals (CI) relating maternal smoking whilepregnant to the incidence of specific hematological malignancies.

There were 505 acute lymphoblastic leukemias (ALL), 48 acute myeloid leukemias (AML), 81 non-Hodgkin’s lymphomas (NHL), and 45 other hematological malignancies diagnosed during follow-up. There was a significantly reduced rate, adjusted for confounders, of ALL among children whose mothers smoked during pregnancy (HR = 0.73; 95% CI: 0.58-0.91). The data was consistent with an inverse linear trend (p for trend = 0.012), although the HRs comparing maternal smokers of 1-9 cigarettes/day (HR = 0.68; 95% CI: 0.52-0.89) and >10 cigarettes/day (HR = 0.80; 95% CI: 0.58-1.10) to non-smoking mothers were not indicative of a trend. Further adjustment for birth weight and gestational age changed HRs very little. The researchers also stratified the analysis of ALL incidence by age and gender. The HRs for any maternal smoking while pregnant were 0.56 (95% CI: 0.31-1.01), 0.83 (95% CI: 0.62-1.11), and 0.64 (0.42-0.97), respectively, for ALL diagnosed at 0-1, 2-4, and >5 years of age. Though the HRs for any maternal smoking while pregnant appeared different for males (HR = 0.63; 95% CI: 0.46-0.86) and females (HR = 0.85; 95% CI: 0.62-1.16), there was no statistically significant interaction between smoking and gender (p = 0.32).

The rate of AML, adjusted for confounders, was not significantly higher among children of mothers who smoked during pregnancy (HR = 1.41; 95% CI: 0.74-2.67). However, the HR comparing maternal smoking of >10 cigarettes/day to maternal non-smoking was 2.28 (95% CI: 1.05-4.94); that for 1-9 cigarettes/day was not significant (HR = 0.91; 95% CI: 0.38-2.21). Adjustment for birth weight and gestational age did not meaningfully alter these results. No increased rate of NHL with any maternal smoking (HR = 1.25; 95% CI: 0.76-2.04), maternal smoking of 1-9 cigarettes/day (HR = 1.21; 95% CI: 0.68-2.18), or maternal smoking of >10 cigarettes/day (HR = 1.30; 95% CI: 0.65-2.60) was apparent.

COMMENT: The finding that maternal smoking while pregnant might prevent later ALL in offspring is counterintuitive, but, due to the study’s strengths, should not easily be brushed aside. The finding that heavy maternal smoking while pregnant increased the rate of AML also seems robust. Previous literature (and possible biological hypotheses) offer little guide to interpretation. Results of previous case-control studies have been inconsistent regarding smoking and leukemia. Meanwhile the only other independent cohort study of the subject (barring an earlier analysis of the Swedish birth cohort with incomplete follow-up) found no significant association but was many times smaller than the present study. Confirmation of these findings awaits further studies. In the meantime it is clear that the risk of common, known adverse outcomes of maternal smoking far outweighs any protective effect of smoking for risk of ALL.

Logan G. Spector

Benzene of the crime

The aromatic hydrocarbon benzene is a known carcinogen. Workers in rubber manufacturing plants exposed to high levels of benzene develop acute non-lymphocytic leukemia (ANLL) at much higher rates than expected. Children may be exposed to benzene, in concentrations many times smaller than in industrial settings, through cigarette smoke, residue carried home by parents from their jobs, automobile exhaust, and contact with petrochemicals. These proxy measures of benzene exposure are typically investigated because case-control study design, which comprises most studies of childhood leukemia etiology, precludes direct measurement. Thus, a recent French case-control study [Steffen C et al. Occup Environ Med 2004; 61: 773-778] investigated parental occupation, proximity to roadways, and residence near gas stations, as potential risk factors for acute leukemia (AL).

Cases of AL were eligible for the study if they were diagnosed at <15 years of age between 1995-1999, inclusive, at four referral hospitals in Nancy, Lille, Lyon, and Paris. Controls were children being treated at the same hospitals as cases for reasons other than cancer and congenital deformities. Cases and controls were included in the study only if they resided in the same administrative region as the hospital. Exposure history was determined through face-to-face interview of the mothers of children. The researchers used unconditional logistic regression to obtain odds ratios (ORs) and 95% confidence intervals (CIs) relating benzene exposure proxies to overall incidences of AL, ANLL, and acute lymphoblastic leukemia (ALL).

There were 280 cases of AL (240 ALL and 40 ANLL) and 285 controls recruited into the study. Controls were significantly older than cases but were not significantly different in ethnicity, socioeconomic status (SES), or rurality of residence. The ORs comparing residence adjacent to a gas station or garage to residence with no adjacent businesses were 2.2 (95% CI: 0.9-5.7) in the in utero period and 4.0 (95% CI: 1.5-10.3) for the postnatal period for overall AL. Having a residence adjacent to a business other than a gas station or garage was not associated with AL. The association of residence near a gas station or garage in the postnatal period appeared to be stronger for ANLL (OR = 7.7; 95% CI: 1.7-34.3) than for ALL (OR = 3.6; 95% CI: 1.3-9.9). There appeared to be trend of increasing risk of AL with increasing length of residence adjacent to a gas station or garage (p for trend < 0.05). There were no significant associations of residence near heavily trafficked roads or of maternal occupational exposure to benzene and AL.

COMMENT: Case selection is considered strongest when it is population-based rather than hospital-based. Only in the former situation can a researcher be sure that cases and controls were derived from the same population. Although the researchers took care to limit analysis to cases and controls that resided near the participating hospitals, they did not appear to seek out cases that resided near the hospitals but were treated elsewhere. As the findings regarding residence adjacent to a gas station or garage were based on very few exposed cases and controls, even a small number of missed cases could influence ORs. Nevertheless, the rationale for investigating benzene in the etiology of leukemia is clear and the findings indicate a dose-response relationship between the length of exposure and AL. These considerations give the finding credibility despite the study’s flaws.

Logan G. Spector

Noonan syndrome, PTPN11 and leukemia

The PTPN11 gene encodes the protein SHP-2 and germline mutations in the gene cause Noonan syndrome (short stature, dysmorphic face, congenital heart disease, skeletal abnormalities). SHP-2 is a signaling molecule and positively regulates RAS. Tartaglia et al (Blood 2004;104:307-313) report somatic mutation in PTPN11in 23 of 317 B-cell precursor childhood ALL cases. No mutations were seen in T-lineage cases. In the B-cell precursor cases, mutations were seen largely in cases without a TEL-AML translocation and with a CD19 positive, CD10 positive, cytoplasmic IgM negative immunophenotype. PTPN11, NRAS and KRAS mutations were largely mutually exclusive, as one would expect. Among children with AML, PTPN11 mutations occurred in 4 of 12 cases with acute monocytic leukemia only. The mutations identified were missense and were predicted to result in SHP-2 gain of function (ie activation of the RAS pathway). The authors suggest these data indicate a wide role for PTPN11 mutation in leukemogenesis that is lineage–related and differentiation stage-related.

COMMENT: This study adds to previous reports that have described mutations of PTPN11 in myeloid malignancies, largely JMML [Tartaglia M et al, Nat. Genet. 2003;34:148-150; Loh ML et al, Blood 2004;103:2325-2331]. A Children’s Cancer Group Study also reports that mutations in AML occur more frequently in children with monocytic (FAB M5) morphology [Loh ML et al, Leukemia 2004, 18:1831-1834]. This study screened PTPN11 in 278 pediatric patients and identified mutations in 11, 4 of whom were M5 cases. The association of this gene with Noonan syndrome would predict an increased frequency of leukemia in this disorder, and JMML and ALL have been reported in association Noonan syndrome [Bader-Meunier B et al, J Pediatr. 1997;130:885-889; Fukuda M et al, J Pediatr. Hematol. Oncol. 1997;19:177-178; Choong K et al J Pediatr. Hematol. Oncol. 1999;21:523-527; Piombo M et al Med. Pediatr. Oncol. 1993;21:454-455; Attard-Montallo SP et al, Med. Pediatr. Oncol. 1994; 23:391-392; Johannes JM et al Pediatr. Hematol. Oncol. 1995;12:571-575]. These reports indicate that leukemia occurs in Noonan syndrome, but lacking a denominator, it is not possible to know the magnitude of increased risk of JMML and acute leukemia might be.

Stella M. Davies

Briefly Noted

Guthrie (neonatal heelstick) cards are a valuable resource for investigating early events in the blood. Isa et al (Pediatr. Blood Cancer 2004; 42:357-360) examined 54 cards from Swedish children who subsequently developed leukemia for presence of parvovirus B19 DNA using nested PCR. All cards yielded amplifiable DNA, but no parvovirus DNA was found.

Stella M. Davies

C3 Quarterly Newsletter
Children's Cancer Research Fund
Epidemiology Research Unit
Division of Pediatric Epidemiology
Clinical Research
University of Minnesota
420 Delaware St. SE, Box 422
Minneapolis, MN 55455
pedsepi@umn.edu

Editors: 
Stella M. Davies, MD, PhD, and Julie A. Ross, PhD