Genetic susceptibility and childhood ALL: A numbers game

 

There have been several recent studies (some covered here see C3 Vol 14 no 2; Vol 12, no 2 ) that have explored the relation between specific genotypes of interest and risk of childhood leukemia.   In a report from Brazil, [Canalle R et al [Environ Mol Mutagen 2004; 43:100-109] provide results from a genotyping study of 113 cases of childhood ALL and 221 controls. Polymorphisms in GSTM1, GSTT1, GSTPi, CYP1A1 and CYP2E1 were explored. The glutathione S transferases (GSTs) are phase II detoxifying enzymes. The investigators hypothesized that children who possessed the null variant of GSTM1 or GSTT1, or the low activity GSTPi (105) valine allele, would be at an increased risk of developing ALL. The cytochrome p450 (CYP) genes are part of the phase I family of activating enzymes. The investigators hypothesized that children who possessed the high activity variant of CYP1A1 (*2 or *3) or CYP2E1*3 would be at increased risk. Overall, with the exception of CYP1A1*2, there was no statistically significant difference in the distribution of the various genotypes between cases and controls. Carriers of the CYP1A1*2/*2 genotype, however, were more frequent among cases than controls (5.3% vs. 1.4%, p=0.07). When the investigators combined the various genotypes to evaluate up to five at risk genotypes compared to the no risk genotype (i.e., GSTM1 and T1 present, GSTPi Ile/Ile, CYP1A1*2 (-/-) and CYP2E1*3 (-/-)), some interesting patterns were revealed.   Children who possessed at least one valine allele for GSTPi, but who had all other low risk genotypes, were at increased risk of developing leukemia (Odds Ratio=2.7, 95% CI=1.1.-6.8). Further, children with four at risk genotypes (albeit only 3 cases and 1 control fit this category) had a 10-fold increased risk of developing leukemia. The authors emphasize the need for larger studies and the incorporation of environmental exposure data.

 

COMMENT: This study is of interest as it explores a pathway driven approach to evaluating genetic polymorphisms in childhood cancer. However, this study is extremely hampered by small numbers. Other studies have had similar problems with small numbers. The Children's Oncology Group (COG) will be evaluating genetic polymorphisms in over 3000 children with ALL and AML to evaluate these types of relationships more systematically. However, it will still be important, as the authors acknowledge, to obtain environmental exposures to investigate gene-environment interactions. For example, laboratory data suggest that GSTPi is important in the detoxification of polycyclic aromatic hydrocarbons (PAHs), which are found in cigarette smoke [Salinas AE, Wong MG. Curr Med Chem 1999; 6:279-309] . It would be of interest to have available questionnaire data to determine whether the low activity valine allele of GSTPi (alone and in combination with other genes) interacts with reported exposure to PAHs to increase the risk of childhood ALL.

An expansion of the recent AE24 Epidemiology of Infant Leukemia study will soon be active throughout the COG. This study will double the number of cases enrolled (total N=480) and include the collection of exposure data as well as DNA from infants and   mothers (there is   evidence that maternal genotype during pregnancy might also play a role in cancer risk) [Labuda D, et al. Am J Hum Genet 2002; 71: 193-197] . Ultimately, it will be important to establish a large registry of all pediatric cancer in North America in order to adequately evaluate the role of gene-gene and gene-environment interactions. Until that time, we can look forward to more reports that are intriguing but simply cannot address the question sufficiently or comprehensively.

Julie A. Ross

Hepatoblastoma: a rare cancer well done

Hepatoblastoma (HB) is a cancer of the liver that strikes mainly infants and young children.   The extreme rarity of HB has in the past hampered epidemiological research into the disease.   Lately, however, investigators have been cooking up well-done studies of this rare cancer at a relatively fast pace. Here we review two recent studies of HB, birth characteristics [Reynolds P, et al. Cancer 2004; 100: 1070-1076] , and smoking [Sorahan T and Lancashire RJ. Br J Cancer 2004; 90:1016-1018] .

 

Reynolds et al conducted a case-control study of HB in California .   Using that state's cancer registry they identified 99 cases of HB diagnosed between 1988 and 1997 for whom there were California birth certificates available.   Four control children were randomly selected for each case, matched on month and year of birth and sex, and their birth certificates were obtained.   Data were abstracted from the certificates on demographic, pregnancy, and birth variables. The investigators then calculated odds ratios (ORs) and 95% confidence intervals (CIs) for each variable.   Most variables, including birth order, type of delivery, and parental age, were not significantly associated with HB in univariate analysis. However race, birth weight <1,500g, and gestational age <37 weeks were (p < 0.05).   These variables were then analyzed simultaneously.   Birth weight was still significantly associated with HB; the ORs comparing birth weight of <1,500g, 1,500-2,499g, and > 4,000g to 2,500-3,999g were 40.80 (95% CI: 4.21-395.5), 2.32 (95% CI: 0.75-7.21), and 1.13 (95% CI: 0.55-2.34).   However, the OR comparing gestational age <37 weeks to term pregnancies was 0.84 (95% CI: 0.31-2.32). Race also was not significant in the multivariate model. Lastly, the investigators correlated birth weight with age at diagnosis of HB.   They found that cases with birth weight <1,000g were significantly more likely to be diagnosed at older ages (p = 0.036), while there was no such tendency among cases with normal birth weight.

 

Sorahan and Lancashire examined data from the Oxford Survey of Childhood Cancers (OSCC), which attempted to enroll all childhood cancer cases diagnosed in the UK between 1955-1981. The purpose of the analysis was to determine if OSCC data corroborates a previous finding, in a more recent UK study, of an association between parental smoking and HB [Pang D, et al. Br J Cancer 2003; 88: 373-381] . History of parental smoking around pregnancy, collected by interview, was available for 43 and 5,734 cases of HB and any other cancer, respectively, and for their matched controls.   The investigators broke the matching in order to compare HB cases to the entire control group and thereby increase statistical power. The adjusted ORs comparing children whose parents both smoked to those whose parents didn't were 1.28 (95% CI: 1.16-1.42) and 2.69 (95% CI: 1.18-6.13) for other cancers and HB, respectively.

 

COMMENT: The increase in risk of HB associated with low birth weight has been known for some time, but the California study is the first to provide a population-based estimate of the increase.   Previously this had been estimated imprecisely by comparing the prevalence of low birth weight in case series to that in published figures.   The study is also the first to show that the increase in risk of HB due to low birth weight is independent of gestational age. That cases with low birth weight differ from those with normal birth weight in age at diagnosis suggests that they may also differ in etiology. The use of birth certificates for sampling and data collection meant that the study was perfectly population-based, data were recorded without respect to subsequent disease, and participation was complete.   Thus the findings were very credible.   The OSCC data were collected through interview and were therefore susceptible to both recall and selection biases. However, the presence of a strong association of parental smoking with HB but not with other cancers argues against these potential biases as an explanation for the results.   Confounding due to low birth weight, which is a known consequence of maternal smoking during pregnancy and a known risk factor for HB, probably does not explain the OSCC results either.   Pang et al reported, in response to a letter raising the possibility of confounding in their study [Spector and Ross 2003; Br J Cancer 89: 603], that the association of HB and smoking was independent of birth weight. The Epidemiology Committee of the COG has submitted a proposal to the National Cancer Institute for a case-control study of hepatoblastoma that will follow up on these promising leads.

Logan G. Spector

Ovarian stimulation and childhood cancer: more stimuli for a national pediatric cancer registry

Given the increasing number of women delaying childbearing until their thirties and early forties, investigations have focused on the potential association between artificial means of reproduction and risk of childhood cancer. We recently reported (see C3, vol 14, no 1) on suggested links between assisted reproductive technology and the occurrence of Beckwith-Wiedemann Syndrome [DeBaun M, et al. Am J Hum Genet 2003; 72:156-160] and retinoblastoma [Moll A, et al. Lancet 2003; 361:309-310] . Here, [Brinton LA, et al. [Fertil Steril 2004; 81: 1083-1091] explore whether ovulation-stimulation drugs might be associated with risk of childhood cancer. In a record linkage study from Denmark, a total of 30,364 women evaluated for infertility during the period 1960-1996 were identified and referred to as an ‘infertility cohort'. These women gave birth to 51,063 children. Of these children, 105 developed cancer before 20 years of age including   54 children who developed cancer prior to their mother's entry into the infertility cohort and 51 who developed cancer thereafter. The overall cancer occurrence in the infertility cohort was comparable to that expected in the general population (Standardized Incidence Ratio (SIR)=1.14, 95% CI=0.8-1.5)). For site-specific malignancies, there was a suggestion of an elevated risk of sympathetic nervous system malignancies (SIR=2.14, (95% CI=0.9-4.4), but this estimate was based on only 7 cases. Using a case-cohort approach, the investigators explored the use of ovarian-stimulating drugs prior to the birth of the child. Comprehensive medical and surgical interventions for fertility were compared between 47 (of 51) cases and 868 children randomly selected from the infertility cohort. Overall usage of any fertility drug was associated with a relative risk (RR) of 0.82 (95% CI=0.4-1.6). Specific use of clomiphene citrate was associated with a RR=0.77 (95%CI=0.4-1.6). Early life tumors and non-hematopoietic tumors (including neuroblastomas) were not associated with an increased risk. In contrast, a non-significant increased risk was observed for childhood hematopoietic malignancies (RR=2.3, 95% CI=0.8-6.6) and usage of any ovulation-stimulating drugs. The authors conclude that although this study provides reassurance that most childhood cancer is unlikely to be associated with ovulation-stimulation, the findings with hematopoietic malignancy warrant further study.

 

COMMENT: A strength of this study was its record linkage capabilities. A few previous epidemiologic studies (both   record review and case-control) have suggested a positive association with sex hormone use during pregnancy and risk of neuroblastoma (reviewed in Brinton LA et al). This study did not confirm this association. While there was a possibility of an increased risk of hematopoeitic malignancy (mainly leukemia) in the present study, the risks were based on small numbers and did not reach statistical significance. Even with a large prospective group of pregnant women, the overall childhood cancer incidence is low and the ability to detect differences among specific subtypes is hampered by small numbers. This again emphasizes the importance of being able to identify a population-based series of childhood cancer   cases in order to tease out potential associations with these types of exposures.

Julie A. Ross

 

 

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