Anybody out there? Anybody home?

 

In the 1980s and early 1990s, random digit dialing (RDD) was the method of choice for selecting controls in case-control studies of childhood cancer. Most people answered their telephones and the majority were willing to participate in studies that may help children. However, with the advent of answering machines, cellular telephones and caller identification (along with the increasing annoyance of telemarketers), RDD appears no longer feasible on a national basis. Further, it is difficult to characterize who among the population actually answers their telephone and agrees to complete a telephone interview compared to those who do not. Our recent experience in the Children's Oncology Group (COG) suggests that of those individuals screened, approximately 60% are agreeable to interview. Although not a bad participation rate, there is difficulty in determining who among those unreachable might have been eligible. There are conservative methods to calculate the true response rate that take these unreachable households into account [Slattery M et al, Ann Epidemiol 1995; 5:245-9] . Applying these calculations can decrease the perceived overall response rate to often less than 50%. While it is still not clear whether an RDD group is substantially biased, control mothers in our studies do tend to be more often white and better educated than our case mothers. The COG Epidemiology Steering Committee has made it a priority to identify alternatives to random digit dial controls for future studies of childhood cancer. This is especially important in light of the developing national pediatric cancer registry.

 

Ma X et al [Am J Epidemiology 2004; 159:915-921] recently reported their experience in using birth certificate controls in an on-going case-control study of childhood leukemia in Northern California . Overall, the investigators successfully traced and enrolled 49% of their selected birth certificate controls. In this evaluation study of the representativeness of their controls, two different control groups were compared: one from birth certificates, the other from a roster of friends of the case mother. A total of 64 birth certificate and 64 friend control pairs were compared. Information on the birth certificates (including maternal and paternal age, maternal and paternal education, birth weight, number of live births, previous pregnancy losses and time since live birth) was compared between control pairs enrolled in the study and 192 ‘ideal' birth certificate controls selected without tracing. With the exception of birth weight, the differences between the 64 birth certificate controls and the 192 ideal controls were smaller than those between ideal and friend controls. Importantly, the ideal controls were comparable to the enrolled birth certificate controls with the exception of birth weight: the enrolled controls were slightly heavier than the ideal controls (3498 grams versus 3407 grams). In an accompanying editorial, COG epidemiologists noted that this comparison study suggests that birth certificates may represent a useful alternative to RDD for control selection [Ross JA et al Am J Epidemiol 2004; 159:922-924].  

 

As a follow-up, the COG Epidemiology Committee recently completed a survey of 52 birth registrars in the United States (including the District of Columbia and New York City , which has a health department separate from that of the state). Thirty-seven states along with New York City (73%) indicated that they could provide us randomly selected birth certificates for our studies. Twelve states (23%) indicated that they could not provide identifying information and two states (3%) did not reply. In a follow-up pilot test of 6 states (including 2 states that initially said they could not comply), we obtained birth certificates from three states, assurances from two states that information could be provided pending IRB approval, and indication from one state (which initially indicated they could not provide them) that they are working on a mechanism to comply to these types of requests. Based on the experience of Ma et al, and our survey and pilot work, we believe that birth certificates could provide a feasible alternative to RDD in the identification of controls for national pediatric case-control studies.

Julie A. Ross

 

I can SEER for miles and miles and miles and miles..

 

The National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program has gathered data since 1973 on virtually all malignancies occurring within nine U.S. states and metropolitan areas including Connecticut, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Utah, Seattle-Puget Sound (since 1974) and Detroit (since 1974). This population-based registry has been an invaluable resource for the study of cancer patterns in general.  

 

Two recent analyses have made use of SEER data to study childhood leukemia. In the first, Wartenberg D et al. [Br J Cancer 2004; 90: 1771-1776] addressed the population mixing hypothesis of childhood leukemia etiology proposed by Leo Kinlen [see C3, vol 13, no 6].   Briefly, the hypothesis maintains that leukemia incidence increases following the rapid influx of migrants into previously isolated areas because of the concomitant increased likelihood that susceptible children would be exposed to a putative leukemogenic virus. Thus, Wartenberg et al analyzed those cases of acute lymphoblastic leukemia (ALL) that were diagnosed in rural counties (<100 people/mile 2 ) in SEER, which were identified in Iowa , New Mexico , and Utah .   Cases were also limited to Caucasians and to those who were born in the state in which the county of diagnosis was located. The exposure of interest was the percent of population increase between decennial census years, which was classified into < 0%, > 0-10%, > 10-20%, and > 20%. Odds ratios (ORs) and 95% confidence intervals (CIs) were obtained using logistic regression and parallel analyses of central nervous system (CNS) malignancies were conducted for comparison.   In the main analysis of interest, limited to children aged 0-4 years and the years 1980-1989, there was a significant association of ALL, but not of CNS tumors, with population increase in 1979-1988. The ORs comparing > 0-10%, > 10-20%, and > 20% to < 0% population increase were 1.9 (95% CI: 1.0-3.6), 2.1 (95% CI: 1.1-3.8), and 2.6 (95% CI: 1.5-4.6) for ALL and 1.2 (95% CI:   0.6-2.4), 1.3 (95% CI: 0.7-2.4), and 0.7 (95% CI: 0.3-1.7) for CNS, respectively.   A more conservative analysis excluded cases from counties adjacent to a non-rural county; the corresponding ORs were 2.6 (95% CI: 1.1-6.4), 2.4 (95% CI: 1.0-5.5), and 4.3 (95% CI: 2.0-9.7) for ALL and 1.0 (95% CI: 0.3-3.0), 1.6 (95% CI: 0.7-3.6), and 1.1 (95% CI: 0.4-3.2) for CNS.    

 

The second report was prompted by an apparent cluster of childhood leukemias in Churchill County , Nevada , home of the Fallon Naval Air Station [Steinmaus C et al., Environ Health Persp 2004; 112: 766-771] . Eleven cases of leukemia (10 of which were ALL) were diagnosed in children < 19 years of age in 1999-2001 in this county of 24,000.   The investigators' first goal was to estimate the probability of this occurrence by combining data from SEER with that of the California Cancer Registry (CCR), calculating the rate of childhood leukemia in all counties with comparable military aviation facilities (MAFs), and comparing this rate to that in Churchill County. Using these data, only 0.92 cases would have been expected, whereas eleven were observed, resulting in a rate ratio (RR) of 12.0 (95% CI: 6.0-21.4).   The RR was still significantly elevated (11.2; 95% CI: 3.6-26.3) when limited to cases diagnosed after the Nevada state health department investigation had begun (July 2000) and who were therefore not susceptible to post hoc ascertainment bias.   The authors calculated that a cluster such as this would occur by chance once every 22,000 years, compared to once every 120 years in the well known cluster in Woburn, Massachusetts (chronicled in the book and movie “A Civil Action”).   The investigators' second goal was to address the theory that exposure to jet fuel and other exposures presumably common to MAFs contributed to the Nevada cluster. They compared incidence of childhood leukemia in counties with MAFs (excluding Churchill County ) to those without, and found RRs of 1.04 (95% CI: 0.97-1.12) and 1.01 (95% CI: 0.93-1.10) among children ages 0-19 years in SEER and the CCR, respectively.   Similar results were obtained when analysis was limited to counties with low population density.

 

COMMENT: These studies both made good use of registry data.   Many caveats apply to analyses such as these, since the investigators typically must make broad assumptions in the absence of individual-level data.   For instance, in the former study the authors had to assume that children born and diagnosed within a state did not change county of residence. In the latter study, the authors assumed that children in counties with MAFs were equally exposed to jet fuel and that children in counties without MAFs were not exposed (i.e. through civilian airports). Nevertheless, these studies are important contributions to the literature. The findings of Wartenburg et al. offer support for the population mixing hypothesis, especially since the association was specific to ALL. Steinmaus et al. formally confirmed that the rate of childhood leukemia was drastically elevated in Churchill County and that the cluster was very unlikely to have occurred by chance alone.   On the other hand, their results suggest that exposures related to military aviation are not risk factors for childhood leukemia elsewhere. Interestingly, population mixing may be an explanation for the Nevada cluster since Fallon Naval Air Station is host to 55,000 short-term trainees each year. Lastly, both studies were beset by small numbers. The establishment of a nationwide childhood cancer registry would thus be a welcome advance in assisting in studies of potential clusters.

Logan G. Spector

 

Briefly Noted:

Phototherapy- Given the blue light

Phototherapy is the standard of care for neonatal jaundice. There has been some concern that blue lamps, which are most effective in reducing hyperbilirubinemia, may increase nevus formation in neonates.   In this cross-sectional study from Germany to investigate predictors of nevi in young children, parents of 1812 white children (ages 2 to 7 years) completed a questionnaire regarding their child's sun exposure, along with presence of neonatal jaundice and phototherapy at birth [Bauer J, et al. Arch Dematol 2004; 140:493-494] . A total of 554 children experienced jaundice at birth, and 333 received therapy with blue lamps. There was no statistically significant difference (p=0.56) in the number of melanocytic nevi between the treated and untreated children. Importantly, the major predictors of nevi number were number of days on vacation and time spent outdoors. This study provides reassurance that phototherapy is unlikely a major contributor to the development of melanocytic nevi.

Julie A. Ross

 

AML following organ transplant

It is well known that the immunosupression needed after solid organ transplantation is associated with an increased risk of skin cancer and lymphoma. A recent review of 170,000 organ transplants shows an increased risk of AML also, which is more frequent after heart or lung transplant than kidney, and is associated with use of azathioprine [Offman J, et al Blood 2004; April 15 th , Epub ahead of print] . Study of AML cells from 7 such cases shows microsatellite instability typical   of defective mismatch repair, as expected after thiopurine exposure.   Azathioprine has been largely replaced by newer immune suppressants; a repeat study is needed in five years to determine if risk is reduced.

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