Viruses and Leukemia: Finnish Envy

Epidemiologists in North America and elsewhere have long admired, and envied, the research infrastructure in Scandinavian countries. Births, deaths, major illnesses, addresses, and emigration are nearly universally recorded and public attitudes towards medical research are, to judge from participation rates, quite favorable. Together these factors enable the Scandinavian countries to conduct studies with relative ease, albeit within ethnically and culturally homogenous populations. A recent joint Finnish and Icelandic nested case-control study [Lehtinen et al. 2003 Am J Epidemiol 158 (3): 207-213] of maternal antibodies to viruses during pregnancy and childhood leukemia (<15 years age) once again summons the “green-eyed monster” here in the U.S.

In a nested case-control study subjects are chosen from within an established cohort study. The study design avoids having to perform expensive procedures, in this case antibody testing, on a large number of subjects while preserving a cohort study’s main advantage: the knowledge that an exposure preceded disease. The Finnish Maternity Cohort and the Rubella Screening Serum Bank and the University of Iceland collected serum at 12-14 weeks gestation from about 750,000 pregnancies in 1983-1997 and 75,000 pregnancies in 1975-1997, respectively, which represented >98% and >95% of pregnancies during those time periods. Lehtinen and colleagues linked cohort members to the Finnish and Icelandic national cancer registries to identify 403 cases of childhood leukemia (342 acute lymphoblastic leukemia (ALL) and 61 non-ALL) whose mothers had serum samples available. Controls (n = 1,216) were matched to cases 3:1 in Finland and 4:1 in Iceland on age of mother at serum sampling, date of specimen collection, and date of birth and gender of offspring. Sera were tested for immunoglobulins G (IgG) and M (IgM) and antibodies to cytomegalovirus (CMV), Epstein-Barr virus (EBV), and human herpesvirus 6 (HHV6). IgM indicates a recent infection while IgG indicates past infection.

There were no statistically significant associations of childhood leukemia and maternal IgG antibodies to any of the viruses. The odds ratios (ORs) and 95% confidence intervals (CIs) comparing IgG positives to negatives were 0.9 (95%CI: 0.7-1.2), 1.5 (95%CI: 0.8-2.8), and 0.8 (95%CI: 0.6-1.1) for CMV, EBV, and HHV6, respectively. While there was not a significant association comparing IgM positives to negatives for CMV (OR=1.1; 95%CI: 0.8-1.6) or HHV6 (OR=0.8; 95%CI: 0.6-1.2) there was a statistically significant association for EBV (OR =1.9; 95%CI: 1.2-3.0). This association was also significant at the more stringent 99% confidence level, used to account for multiple comparisons. In attempting to improve the sensitivity and specificity of EBV antibody tests, they restricted further analyses of IgG positivity to IgM negatives and of IgM positivity to CMV- and HHV6-IgG negatives. After adjusting for birth order, sibship size, and matching factors, the OR comparing IgG positives to negatives were 1.0 (95% CI: 0.1-10), 4.0 (95% CI: 0.5-33), 1.4 (95% CI: 0.5-3.8), infinity (95% CI: 0.0-undefined), and 1.9 (95% CI: 0.8-4.4) for subjects <1, 1, 2-6, >6, and 0-14 years old, respectively. Corresponding ORs comparing IgM positives to negatives were 9.2 (95% CI: 0.8-102), 4.4 (95% CI: 0.6-35), 2.9 (95% CI: 1.2-7.4), 1.6 (95% CI: 0.3-7.3), and 2.9 (95% CI: 1.5-5.8). The apparent inverse trend between age at diagnosis and the ORs comparing IgM positives to negatives was not significant (p = 0.6).

COMMENT: This study’s methodology was impeccable. Nevertheless, several explanations are possible for the finding that EBV IgM was associated with childhood leukemia. First, EBV reactivation in the mother while the child is in utero may be causally related to childhood leukemia. Reactivation, rather than primary infection, is suggested by the fact that nearly all mothers were positive for EBV IgG. The researchers implied a possible mechanism in noting that recombination activating gene 1 is expressed in lymphoblastic leukemia cells and its expression can be induced by EBV. Second, the association may have been due to chance observation, though it retained significance at the more stringent 99% confidence level. Third misclassification of EBV IgM positivity may have occurred, although restricting analysis to CMV- and HHV-6 IgG negatives lessened the possibility that these antibodies cross-reacted with the EBV IgM assay. Lastly, EBV reactivation may be associated with another factor that itself leads to leukemia. For instance, the researchers suggested that DNA topoisomerase II inhibitors such as quinone precursors might induce rearrangements of the MLL gene and coincidentally reactivate EBV. Interestingly, the association of EBV IgM with leukemia appeared to be the greatest in infancy, when MLL 11q23 rearrangements predominate. Few cohorts are large enough or have the right biological samples to address this topic and traditional case-control studies are ill-suited to detecting recent or recently reactivated infections. Hence these results are not likely to be replicated any time soon. However, determining the nature of the relationship between EBV reactivation and suspected risk factors for childhood leukemia may help to determine whether the former causes or is merely coincidental to subsequent illness.

Logan G. Spector

Racial differences in genomic imprinting?

As we have described previously [e.g. C3 Vol 11, No 3; Vol 11; No 2], genomic imprinting is the preferential expression of a gene depending on the parent of origin. Chromosome 11p15 is one area the contains several imprinted genes including insulin-like growth factor-2 (IGF2). With rare exceptions, almost all human tissue expresses only the paternally-inherited IGF2; the maternally-inherited allele is silenced. In several pediatric tumors, however, loss-of-imprinting (LOI) (or maternal expression) of IGF2 has been observed, which may be important in etiology. Few if any studies have considered known racial differences in pediatric tumors relative to their imprinting status. In this report from the Lancet [2004; 364:445-51], Fukuzawa R et al examined patterns of imprinting in Wilms’ tumor tissue collected from predominantly white children in New Zealand and east-Asian children from Japan. The investigators measured IGF2 imprinting as well as 11p15 loss of heterozygosity by extent of methylation. For the white children, 13 of 41 (32%) Wilms’ tumors demonstrated LOI of IGF2. In contrast, none of the 21 tumors from Japanese children demonstrated LOI. These data suggest that the etiologic mechanism(s) for developing Wilms’ tumor in Asian children may be different than in white children.

COMMENT: This provocative study utilizes information from many different sources to develop and test the hypothesis. As the investigators note, there is variation in the incidence of Wilms’ tumor worldwide, with rates at least two-fold higher in white children compared to Asian children. Further, the histological patterns are different between the two groups: nearly 25% of Wilms’ tumors in white children present in the perilobar nephrogenic rest, while this presentation is rare (1-8%) in Asian children. As demonstrated by a recent RFA, there is interest by the National Cancer Institute to understand how environmental factors might influence epigenetic mechanisms such as imprinting. Studies such as these that examine racial variations in molecular patterns can assist in the development of the next generation of molecular epidemiology studies.

Julie A. Ross

L'hypothèse de Greaves revisitée

Greaves’ hypothesis suggests that childhood leukemia, particularly common ALL (cALL; i.e. CD10+ B-lineage ALL), results from at least two genetic “hits”. The first hit is thought to occur in utero and the second postnatal hit is thought to occur more often among children whose exposure to infections is delayed. Investigations of this hypothesis have focused not only on the occurrence of infections in early life, but also on more easily measured proxies for infection. Examples include attendance at day care and having older siblings, both of which tend to increase the likelihood of a young child being exposed to infection. Jourdan-Da Silva et al. [Br J Cancer 2004; 90: 139-145] report on a French study that is the latest to address Greaves’ hypothesis.

The investigators enrolled 473 cases of leukemia less than 15 years of age and 567 controls frequency-matched on age, sex, and region. Cases were identified through France’s population-based childhood leukemia registry and controls were selected, “from a sample of 30,000 phone numbers representative of the French population with respect to area of residence and municipality size categories.” Data on history of ear, nose, or throat (ENT), gastrointestinal (GI) and other infections, as well as data on infection proxies and general perinatal characteristics, were collected by a self-administered questionnaire. Children with Down syndrome, a known risk factor for leukemia, were excluded from analysis.

Results were copious; besides having several exposures of interest they analyzed data separately for cases by diagnosis (ALL or AML), by age (2-6 or 7-15 years), and by immunophenotype (cALL or other ALL). There were significant or near significant inverse relationships of having had > 4 ENT infections (OR = 0.8; 95% CI: 0.6-1.1), > 4 GI infections (OR = 0.1; 95% CI: 0.03-0.6), or > 4 infections of any type (OR = 0.8; 95% CI: 0.6-1.0) with ALL. No significant association was apparent with ALL and duration of breast-feeding. There was a weakly significant inverse association of attendance at day care and ALL (OR = 0.7; 95% CI: 0.6-1.0) as well as a significant decreasing trend in ALL risk (p < 0.05) with decreasing age at day care enrollment; the latter results did not hold when analysis was limited to age at enrollment in full-time day care. Lastly, there was a significantly elevated risk of ALL among children with birth order of 4 or higher compared to those who were first-born (OR = 2.0; 95% CI: 1.1-3.7; p for trend = 0.07). There were no significant results for the above exposures and AML. Having > 4 ENT, GI, or other infections, age at start of day care, and birth order were weakly significantly associated with leukemia at ages 2-6 years, in the same directions as above, but not with leukemia at older ages. Similar results also obtained among cALL but not among other ALL case subanalyses.

COMMENT: Epidemiologists evaluate a hypothesis in part by looking for consistency of results in different populations. This study covers familiar territory and adds geographical diversity to the investigation of Greaves’ hypothesis. Most previous studies have been in the U.S. or the U.K. Unfortunately, investigators often must rely on maternal recall and this will affect data to varying degrees. Infant illness is common and may be difficult for mothers to remember accurately; the high proportion of missing responses to these questions in the current study may reflect this reality. History of day care and birth order are probably recalled with a higher level accuracy that is comparable between cases and controls. However, the inverse association of day care was only weakly significant and the results for birth order are the opposite of those expected. Moreover, other research suggests that there is a complex interplay between breastfeeding, birth order, day care, and the incidence of infections in infancy [Pettigrew M et al. Annals of Epi 2003; 13: 431-435], and it may be inappropriate to analyze these factors separately. This study therefore offers only very qualified support for Greaves’ hypothesis.

Logan G. Spector  

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