Childhood cancers and the environment: Is there anything to worry about?

The considerable glare of publicity that surrounds the regular reports in the general press defining yet another potential environmental ‘cause’ of childhood cancer mandates frequent reassessment of the true state of the art. As health advisers to the parents of our patients and to public policy makers, it behooves us to have a working knowledge of the relationship between environmental events and childhood cancer.

In considering this relationship, the criteria bona fide epidemiologists use to determine the existence of a causal relationship must be applied. These criteria include biological plausibility, temporality, strength of the association, dose-response relationship and consistency across studies. For environmental exposure studies, two fundamental approaches are possible: case control studies and cohort studies. While the low incidence of childhood cancer makes it more difficult to demonstrate the possible role of a given environmental agent by either methodology (case control or cohort), the evaluation of any given study ought to include rigorous consideration of these criteria.

The key to properly planned case control studies is the selection of appropriate control subjects. The issue of recall bias (differences in the accuracy of subject recall across compared groups) is often cited as a confounding factor in case control studies. Concern about recall bias has recently been addressed in an elegant study, framed within a leukemia epidemiology study. Canadian investigators demonstrated that the sensitivity of recall did not differ significantly between parents of patients, hospital controls and population-based controls (1), unless the particular exposure under consideration was a current media issue.

Chemical agents that have been evaluated as potential environmental hazards include pesticides, petroleum products and N-nitroso compounds. With regard to pesticides, there is not consistent evidence of an association with acute lymphoblastic leukemia (ALL) (2); however, several small studies suggest a relationship with acute nonlymphoblastic leukemia, consistent with studies in adult acute nonlymphoblastic leukemia (3). Two small studies have suggested an association with no-pest pesticide strips and brain tumours, but the weight of evidence does not yet support this association (4,5).

Petroleum products have not demonstrated a consistent relationship to any childhood malignancy, despite innumerable studies addressing the issue. N-nitroso compounds are commonly found in cured foods of various types, notably meats. Several early studies suggested an association between the intake of cured meats, in general, and hot dogs, in particular, and brain tumours (6). Biological plausibility is lent by the known carcinogenic effect of N-nitroso compounds on the developing brain of mouse embryos. A recent review of publications on this subject considered 14 studies and concluded that the number of variables that bore on the consumption of cured meats, including socioeconomic status, intake of vegetables and other variables, did not permit a clear-cut conclusion of a relationship nor did it allow ruling one out (7).

Physical agents for which exposure has been studied include radiation and electromagnetic fields. Environmental exposure to nuclear power plants and nuclear reprocessing plants, either in the child or in the father preconceptionally, have been the subject of intense investigation. The early studies in this regard were generated by a television program reporting a perceived increase in acute lymphoblastic leukemia incidence in the area surrounding the Sellafield nuclear installation in the United Kingdom (8). Extensive investigation concluded that there was indeed an increased incidence of leukemia, but that it could not clearly be attributed to nuclear exposure of the child or preconceptual paternal exposure (9,10). A large study in Ontario (11) specifically designed to replicate the Sellafield study did not demonstrate an increase in leukemia incidence or mortality in areas surrounding nuclear installations.

The extensive study of risks of childhood leukemia and brain tumours in relation to electromagnetic field (EMF) exposure has yielded contradictory results. A large American study (12) that included EMF measurements in the index child’s bedroom, selected other rooms and on the exterior perimeter in all homes in which the child had resided for at least 70% of his or her lifetime, demonstrated no statistically significant increase in the risk of ALL with increasing exposure (OR 1.24, CI 0.86 to 1.79). A smaller Canadian study (13) that used similar measurement techniques found an increased risk for the development of ALL in children with higher EMF exposure who were less than two years of age at the time of exposure. The risk, if it exists, is small.

The infectious theory of etiology of childhood ALL is derived from observations of age distribution in white populations, with peaks of incidence between ages two to five years. Greaves (14) first suggested the two-step infectious induction of leukemic change – the first step being an aberrant mitotic event in utero in B lineage lymphoblasts as they expand rapidly to generate multiple clones capable of responding to a huge array of antigens. The second step is postulated to be induced by exposure to environmental infectious agents, which may vary in timing according to socioeconomic status, thus accounting for the absence of an early age peak in African children and a peak incidence in the two to five age group in white children reflecting later exposure (14). Kinlen et al (15) postulated an influence of high population mixing as a result of migration to ‘new towns’ occurring as a result of industrial expansion. The migrant population, Kinlen et al (15) suggested, might have different infectious exposures from the initial inhabitants of the town, resulting in a mixing of immune and nonimmune children. This may result in aberrant immune responses in the previously unexposed pre-existing inhabitants. Some evidence supports this theory in several contexts of population migration, including areas of new offshore oil exploration (16), mass tourism and urban migration, as in Hong Kong (17). While intriguing, neither of these hypotheses has been definitively proven.

Is there anything to fear from environmental exposure? Given the limitations of the methodologies and the risk of bias, and the historical grouping together of disease entities that contemporary molecular biology allows to be clearly separated on the basis of nonrandom genetic molecular events, it is hard to identify clear-cut population exposures for childhood cancers that are causal, although some suggestive associations are present. For those exposures that are suggestive, little opportunity for modification of the exposure as a matter of public policy exists.

The true answer may begin to emerge as science moves from the era of descriptive epidemiology to that of molecular epidemiology (the science of defining inherited characteristics at their molecular expression level and correlating these with environmental exposures). It is intuitively comfortable to postulate that, for any given potentially oncogenic environmental exposure, the effect of that exposure may be both mediated and modulated by biological characteristics that are specific to the exposed host. This formulation is a rational explanation for why only a small proportion of a cohort exposed to a particular environmental risk goes on to develop malignant change. Elegant studies of genetic susceptibility in childhood leukemia have defined a combination of heritable genetic polymorphisms in the phase I, chemical and/or drug metabolizing enzyme cytochromes P-450, and in phase II, conjugating enzymes glutathione-S-transferase and N-acetyltransferase, all involved in the metabolism and excretion of known carcinogens, which defines a group at increased risk of developing acute lymphoblastic leukemia with significant odds ratios in the 1.5 to 1.8 range (95% CI 1.0 to 2.2) (18,19).

If there was a strong direct relationship between any of the environmental exposures and disease, the studies that have been performed should have shown this. By the opposite token, if uncommon or multiple mutations are required to make an individual child susceptible to a given carcinogen, only a small subset of the exposed childhood population would be susceptible. Watch this space – society is at the beginning of an exciting ride that combines classic epidemiological techniques and an ever expanding range of molecular techniques that may allow us to answer, in a much more specific manner, the questions of whether we have anything to worry about, and who should worry about it.