No bones about it: we are going to the dogs!

Little is known about the etiology of osteosarcoma, a bone tumor that is typically diagnosed in adolescents and young adults. Rapid skeletal growth and large body size have been hypothesized to be etiologically relevant, given the age-specific incidence peak observed during the pubertal years. It is speculated that gonadal sex hormones may be im-portant, although it is unclear how these hormones may play a role in development. Cooley DM et al [Cancer Epi Biom Prev, 2002; 11:1434-1440] evaluated potential risk factors for bone tumors in an animal model, specifically Rottweiler dogs. Spontaneous osteosarcoma in pet dogs behaves similarly to osteosarcoma in humans. Further, like humans, larger size dogs may be at greater risk, although this is controversial. The investigators were particularly interested in whether gonadal hormones play a role since Ru G et al [Vet J 1998; 156:31-9] showed a 2-fold increased risk in neutered dogs. In this historical cohort study, questionnaires were mailed to 1500 dog owners identified through membership in national Rottweiler breed specialty clubs in North America; the questionnaire was also published in a national breed magazine. A total of 730 questionnaires were returned within a four month period. Data, completed with the assistance of a veterinarian, included general dog information such as date of birth, gender, date of neuter, body condition, adult body weight, adult height, growth rate, bone structure, diet, etc. Questions regarding bone sarcoma included age at diagnosis, location, family history, vaccination history, etc. Bone tumors were reported in 133 dogs and medical records and radiographs were requested by the authors for independent review. Forty-seven dogs were excluded from analysis because medical records were incomplete, radio-graphs were inconsistent with long bone sarcoma, or histopathology was inconsistent with osteosarcoma. Thus, data from 86 confirmed cases were compared to 597 dogs that were free of bone cancer.  Interestingly, the authors found that the risk of bone sarcoma was significantly influenced by the age of gonadectomy (neutering or spaying). Male and female dogs who underwent gonadectomy before the age of 1 year were significantly more likely to develop bone sarcoma compared to dogs who were sexually intact (Relative Risk (RR)=3.8 (95%CI=1.5-9.2) for males; RR=3.1, 95% CI=1.1-8.3) for females). Moreover, there was a significant inverse dose-response relationship between life-time gonadal hormone exposure and risk of bone sarcoma. All of these associations were independent of adult weight or height. The authors conclude that, at least in Rottweiler dogs, endogenous hormone exposure reduces the risk of bone sarcoma.

COMMENT: As noted previously [C3, vol 10, no 1], Dr. Kelsey has written a review [Epidemiol Rev 20:204-217, 1998] that remarked on the potential benefit of studying the behavior of tumors in dogs to learn about the etiology of cancer in humans. The study by Cooley et al suggests that castration leads to an increased risk of bone sarcoma in Rottweiler dogs as did the study by Ru et al. While humans do not routinely undergo gonadectomy, particularly at a young age, there are situations where reduced gonadal function does occur (e.g., childhood cancer survivors) and it may be of interest to investigate whether there is an increased risk of osteosarcoma  associated with hypogonadism in these popu-lations. Dogs may also be a potential animal model for therapy. On a lighter note, one can see several benefits in working in the area of comparative oncology. For example, this study had a relatively high participation rate. In contrast, mailing out questionnaires to the general population to assess human health characteristics typically reaches (at most) a 30% response rate, and this is after several repeated attempts. Obviously, people are much more willing to talk about their pets than themselves. 

 

Viral load?

The existence of oncogenic viruses has inspired many researchers to search for one that might cause childhood ALL. Leukemic cells have been tested for viral DNA with little result, though the absence of such genetic sequences cannot rule out past infection.  Others have tested for antibodies to candidate viruses, which may persist longer than viral DNA or RNA. Salonen MJH et al [Leukemia 2002; 16: 716-719] compared the antibodies to human herpesvirus-6 (HHV-6) in 40 children recently diagnosed with leukemia (33 with ALL, 7 with AML) with 39 children “with various neurological symptoms of suspected viral origin,” matched on age, sex, and season. Subjects were tested for both HHV-6 specific IgM antibodies, which indicate a recent infection, and HHV-6 specific IgG antibodies, which persist long after initial infection.  IgG levels were comparable between the two groups, with 39 (97.5%) cases and 36 (92.3%) reference subjects being positive.  However, sixteen (40%) children with leukemia were positive for IgM while only three (7.7%) reference children were, which was a significant difference (p=0.005).  Whereas 22/40 cases were ages two to five years, the majority of IgM positive cases were ages six to fourteen years old.  The authors concluded that evidence of recent infection in the leukemia group supported a role for HHV-6 in initiating leukemia.

COMMENT: Though it was interesting that leukemia cases should display evidence of recent infections, mainly at ages older than 3 years, when other surveys have demonstrated that 90% of two-year-olds have antibodies to HHV-6 [Parker CA & Weber HM. J Virol Meth 1993; 41:265-276], two aspects of the study give pause. First, the choice of neurological patients as controls may have been inappro-priate. For instance, if there was a long latency between acquiring HHV-6 and appearance of neurological symptoms, assuming the latter can cause the former, then HHV-6 infections in the reference group would be older by artifact. The finding of low anti-HHV-6 IgM positivity among popu-lation controls would have strengthened the comparison. Second, the study had problematic temporality since antibody prevalence was measured at diagnosis.  Compromised immunity, such as occurs in leukemia, can allow the reactivation of HHV-6 infection.  Thus, even if the choice of reference group did not bias the results, it may be that high anti-HHV-6 IgM levels among cases was a result of leukemia rather than vice-versa. This study highlights the need for great care in reference group selection and interpretation in case-control studies.  

Go play in traffic

Motor vehicle exhaust contains many carcinogens including benzene. Previous ecologic studies have suggested an association between surrogate measures of motor vehicle exhaust (including traffic density, car density, and estimates of the concentration of nitrogen in the air) and risk of childhood cancer, particularly leukemia, although the data are equivocal. In this study byReynolds P et al [Cancer Causes & Control, 2002; 665-672], geographical information system (GIS) technology in combination with three separate traffic measures was used to evaluate whether childhood cancers in California occur more frequently in high traffic neighbor-hoods. Addresses at diagnosis were used to assign census blocks. Using 1990 census data, these blocks in turn were assigned vehicle density (dividing the number of vehicles owned by a block group by the land area in square miles), road density (miles of road within a block group per square mile of area), and traffic density (sum of the vehicle miles traveled divided by the block group area in square miles). Further, average vehicle, road and traffic density values were compared to available ambient air monitoring data from 1990 and correlations were calculated. During the period 1988-1994, encompassing 46 million-person years of observation for children in California, 7143 newly diagnosed childhood cancers occurred. Nearly all (97.8%) cases were assigned to a census block group. None of these measures was associated with an increased risk of childhood cancer nor with leukemias or gliomas. For example, for traffic density, the rate ratios for geographical areas in the 90th percentile of density (over 320,700 vehicle miles traveled per day per square mile) were 1.08 (95%CI=0.98-1.20) for all cancers combined, 1.15 (95%CI=0.97-1.37) for leukemias, and 1.14 (95%CI=0.90-1.45) for gliomas. Compared to air monitoring data, traffic density values were strongly correlated with benzene, carbon monoxide, and 1,3-butadiene, therefore, this measure may be a useful surrogate exposure marker. 

COMMENT: This study demonstrates the powerful nature of GIS methods in generating hypotheses, and to some extent, testing them. The fairly strong correlations observed between air monitoring data and traffic density values are encourag-ing. It is important to note, however, as the authors acknow-ledge, that these types of studies are ecologic in nature. They contain no individual exposure level data. Moreover, in these types of studies, there is a lack of information on potential confounders (such as other sources of air pollution and benzene exposure, etc). Finally, it is possible that the time of diagnosis is not the most relevant period. The authors are planning on performing a similar study using the mother’s address from birth certificates to evaluate the potential importance of this exposure during pregnancy, which will be of interest. 

Interesting Leeds

Kinlen’s observations of increased incidence of childhood leukemia during periods of immigration to previously isolated areas (which he termed “population mixing (PM)) have evinced great interest, since it suggests an infectious origin for the disease. Whereas his original studies examined extreme instances of PM, such as military encampments, later studies have applied various quantitative measures of PM to wide geographic areas. In the latest study Parslow RC, et al [Eur J Cancer 2002; 38:2033-40] used Poisson regression to relate the rate of childhood leukemia, CNS tumors, and other cancers in 532 electoral wards in Yorkshire, UK, to population density, deprivation, and two measures of PM (percent of residents who migrated to a ward in the year prior and diversity of migrants’ origins). Each of these PM measures was calculated for all ages and for children only. The rate of ALL was significantly lower in districts with child migrants from the most diverse origins, controlling for the other factors, with the rate ratio (RR) comparing the 90th percentile of diversity to the 10th-90th percentiles being 0.67 (95% CI=0.47-0.94). Neither the percentage of new migrants among residents nor population density were associated with ALL.  There was no evidence that PM was associated with other childhood cancers.

COMMENT: The findings of this well-constructed study contrast with previous studies. The researchers constructed their diversity of origins variable to compare their results with those of Stiller and Boyle [BMJ 1996; 313:1297-1300], who found an increased rate of leukemia with this measure.  By way of explanation, Parslow et al. offer that their unit of observation, the electoral ward, is much smaller than the county districts employed by Stiller and Boyle.  However, the authors did not elaborate how a finer unit of observation could produce opposite results. Moreover, yet another study [Dickinson and Parker BJC 1999; 81:144-151], which they did not cite, also used electoral ward as the unit of observation and found results similar to Stiller and Boyle’s. The present study continues the question of the relationship between PM and childhood leukemia, but will no doubt stoke more interest in research along these lines. 

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