Presentation is everything 

A report from Brenner et al. [Am J Roentgenology 176:289-96, 2001] entitled 'Estimated Risks of Radiation-Induced Fatal Cancer from Pediatric CT' has been attracting significant media attention. The authors used data from a variety of sources to estimate the increased frequency of cancer that might occur over a lifetime in children under the age of 15 who receive a CT scan. They used a 1989 British survey to estimate the organ-specific dose of radiation delivered to various sites by CT scan, which was then combined with an estimate of the frequency of CT scan for children in the United States. Many assumptions and different data resources were used to generate their estimates. It is noted that the assumed risk estimates are ultimately derived from analyses of mortality data based on Japanese atomic bomb survivors who were exposed to intermediate radiation doses. This assumes a linear extrapolation without a dose threshold to extrapolate cancer risks to very low doses. The authors generate some quite striking statements including: "Although CT examinations of patients less than 15 years old contribute only approximately 4% by number (i.e, # exposed), based on our calculations they are estimated to contribute approximately 20% of the total potential of cancer mortality from CT examinations." Although they also state, "the estimated projected 500 CT-related deaths represents a small (approximately 0.35%) percentage increase over this background." The authors' major goal is to encourage radiologists to minimize as much as possible the settings used when taking CT images of children, which is a worthy endeavor. This issue is also addressed in a paper in the same journal by Patterson et al. [Am J Roentgenology 176:297-301, 2001]. While worthy, the statements in these papers, and, in particular, the title of the Brenner paper, are prone to being picked up by the media and lay press [e.g. the Burlington Free Press article cited 'CT scans in children linked to cancer in later life'; kindly sent to us by Al Homans, MD). Although the authors state that for an ill child the benefit of a CT scan typically outweighs the risk of receiving one, this is not discussed at all in the media. In addition, this is a population-based estimate of increased risk. It should not be taken to mean that in any individual cancer is directly caused by a CT scan.Care should be taken in the generation of concern in the public about relatively minor risks, and non-inflammatory titles can help in this area. 
- Stella M. Davies

Man vs tick: could it make you sick? 

Several epidemiological studies have investigated the potential association between pesticide exposure and the development of childhood cancer. Daniels et al [Epidemiology 12:20-27, 2001] report the analysis of a case-control study of neuroblastoma. Incident cases of neuroblastoma were identified throughout the US and Canada during the period 1992-1994. One control was individually matched to each case through random digit dialing. A total of 538 case mothers and 405 case fathers, as well as 504 control mothers and 304 control fathers completed the telephone interview. Each parent was asked separately about pesticide use during the period 1 month prior to conception up to the date of diagnosis. Since the pesticide exposure reported by each parent often differed, two exposure indicator variables were created: 1) either the mother or father reported pesticide use, but the other parent reported no use, and 2) both parents reported pesticide use. These groups were compared to a reference group where both parents reported no pesticide use. In addition, the authors analyzed the data with respect to clinical and biological subgroups of neuroblastoma, including MCYN amplification status. Overall, pesticide use in both the home and garden was associated with a modest increased risk (OR=1.6, 95% CI=1.0-2.3; OR=1.7, 95% CI=0.9-2.1, respectively). More notable associations were found with children diagnosed after 1 year (OR=2.2, 95% CI=1.3-3.6), compared to infants (OR=1.0, 95% CI=0.6-2.0). The effects of pesticides were not significantly modified by MCYN amplification status. Regarding agreement between mom and dad's reporting, there was moderately good agreement for extermination (kappa=0.6), but there was considerably less agreement regarding home pesticide use (kappa=0.2). 

COMMENT: Overall, this study provides a detailed evaluation of parental pesticide exposure and the development of neuroblastoma. Daniels et al acknowledge the limitations of parental reporting and emphasize the difficulties for parents to recall exposures up to 5 years ago. In the discussion, the authors report that they had higher confidence that a pesticide was actually used if both parents reported its use. It seems possible, however, that a parent who sprays for ants in the home, or aphids on the dogwood, would not necessarily consider this a reportable event to their partner. The authors acknowledge that multiple compounds can be found in pesticide products, and there is a general lack of specificity when examining such a heterogeneous exposure. It is still unclear, however, if there is any etiologically-relevant association between pesticide exposure and neuroblastoma (or any cancer, for that matter). Nevertheless, given our considerable genetic homology to insects, the recommendation should be to limit the amount of pesticide exposure for everyone. 
- Julie A. Ross

Drinking water and childhood leukemia: A murky mess indeed! 

Drinking water is known to contain many contaminants (both natural and man-made) including trihalomethanes (such as chloroform, bromoform, etc), and metals (cadmium, lead, zinc, arsenic). Some of these contam-inants have been associated with cancer development in both animal and epidemiologic studies. Infante-Rivard et al [Epidemiology 12:13-19, 2001] examined the possible association between water contaminants and childhood ALL in a matched case-control study conducted in Quebec. The authors compared 491 cases diagnosed at less than 10 years of age, between 1980-1993, and 491 controls. Briefly, cases and controls were assigned to a municipality-exposure matrix for specific water contaminants that was constructed from 3 sources: a) information from parental interviews covering the period from pregnancy to the reference date (diagnosis date for cases), b) historic data provided by municipalities and c) a tapwater survey that was carried out in a subsample of 227 homes. For the historic data, 305 municipalities were identified where cases and controls had lived from birth until the reference date. For the tap water samples, 227 homes of study subjects were visited between 1995 and 1996, and the tap water was run for five minutes prior to sample collection. All case and control addresses were mapped and the authors constructed an exposure matrix based on the three sources of data. Overall, the authors report no increased risk associated with average levels of exposure during the prenatal period. However, they did find some positive associations including postnatal exposure to chloroform (OR=1.63, 95% CI=0.84-3.19) and zinc (OR=2.48, 95% CI=0.99-6.24). While risks were also increased for cadmium and arsenic, no positive associations were found for nitrates. The authors recognize several of the limitations to interpreting these imputed values. 

COMMENT: This was an admirable attempt to explore the possible association between drinking water contaminants and childhood ALL. However, factors that weaken any possible interpretation of this study include a lack of biological support for an association between the specific contaminants examined here and leukemogenesis, as well as the considerable limitations of exposure assessment through imputed data. Moreover, for the 227 houses that had their tap water directly measured, the water was run for five minutes before a sample was taken. While this is how municipalities perform routine testing, it is unlikely that most individuals run tap water in their home for 5 minutes before drinking. This study emphasizes the need for epidemiologists to find accurate methods to measure exposures. Recent studies that directly measured electromagnetic field exposures in homes a child occupied prior to leukemia [See C3, Vol 8, No 8] provide a good example. In the case of water contaminants, a better method might be to obtain a water sample (drawn after an interval a parent usually uses) from all homes occupied prior to the child's development of leukemia. A major limitation of measuring water contaminants or other exposures (including EMFs) after diagnosis is the assumption that these measures reflect the exposure level during the pertinent time period, which may not be the case. Nevertheless, direct measure of exposure (rather than extrapolation from matrices) offers the best possible method of investigating potential associations. 
- Julie A. Ross 

Why big isn't always better 

As reported previously [see C3, Vol 10, No 3, Vol 9, No 2], there is considerable interest in the relationship between high birthweight and childhood cancer. Several studies have reported that birthweight greater than 4000 grams (g) is associated with about a 2-fold increased risk of leukemia in younger children. Moreover, Wilms' tumor, neuroblastoma, and hepatoblastoma have also been associated with high birthweight. The interest in the potential link between high birthweight and malignancy extends to adults, where several studies have suggested a link between high birthweight and cancer of the breast and prostate. In the most recent study to date, Kaijser et al [JNCI 93:60-61, 2001] examined the association between birthweight and breast cancer in the female twin of opposite sex pairs. It was previously reported that a female twin of a male-female twin pair has an increased risk of breast cancer, however, there was no ability to examine birth weight as an independent risk factor. In this study, records from 13,280 male-female twin pairs identified through the Swedish Twin Registry were linked the Swedish Cancer Registry. Cases included females who developed breast cancer between 1972 and 1995. Controls were selected from the same population, age matched, alive and without breast cancer at the time of the case diagnosis. Birth records were obtained for all cases (n=104) and their matched controls. The analysis was restricted to twins with a gestational age of 33 weeks or more, given that prematurity has been identified as an independent risk factor for breast cancer, leaving 90 case-control pairs. Notably, the risk of breast cancer increased in a stepwise fashion with increasing birth-weight (p trend, 0.007). Although based on small numbers, women with a birthweight of more than 3500g had a 12-fold increased risk of breast cancer compared to women who had a birthweight less than 2500g. No interaction was found with the male co-twin's birthweight or gestational age, suggesting that high birthweight is a strong independent risk factor for breast cancer. 

COMMENT: The positive link between high birthweight and cancer both in children and adults needs to be more thoroughly investigated. With the exception of gestational diabetes (which has been controlled for in several of the childhood cancer studies), we know very little about what causes high birthweight. While high birthweight is associated with high levels of growth factors (notably insulin-like growth factor-1), it is unclear how these levels could play a role in carcinogenesis, particularly forty or fifty years later. Biological studies are necessary to identify genetic and environmental predictors of high birthweight, which may lead to an understanding of the mechanism of increased cancer risk. 
- 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