Materials and Methods

Study participants—This study was approved by the institutional review boards of the CDC and Colorado State University. A case-control study of dogs evaluated at Colorado State University Veterinary Teaching Hospital was conducted. The case group included 100 dogs with a histologically confirmed bladder neoplasm diagnosed between 1996 and 1998. The control group included 100 dogs that had nonneoplastic diseases or injuries and that were admitted to the veterinary hospital during the same time period. Controls were frequency matched with cases on age (within 2 years) and sex.

Questionnaire—Owners of dogs enrolled in the study provided verbal informed consent and were interviewed via telephone. The telephone questionnaire included a complete residence history for each dog. For the relevant time period for each residence, it was determined which utility company provided household tap water and how the tap water was disinfected. Information about potential confounders was collected by use of the following components of the questionnaire: smoking history for each household member to examine exposure to environmental tobacco smoke; medication history (including aspirin, phenacetin, reproductive hormones, parasiticides, and other medications associated with bladder cancer in humans); history of exposure to specific foods and dietary supplements; information about other dogs in the household; and information describing pesticide use in the home and on the pet.¹⁵

Drinking water utility company data—Concentrations of TTHMs are considered reasonable surrogates for other classes of DBPs³ and were used as an exposure variable in this case-control study. Each dog's TTHMs exposure history was reconstructed from its residence history and utility company data. For dogs with bladder cancer and control dogs from 1996 to 1998, assuming a maximum longevity of 15 years for a dog with bladder cancer diagnosed in 1996, the relevant period of exposure was from 1981 onward. Water-quality data were collected from 61 drinking water companies. In addition to providing the TTHMs data, a representative from each utility company completed a short questionnaire about the source of water for the utility company, the number of humans served by the utility company, disinfection processes, and any changes in disinfection processes that occurred during the study period.

Statistical analysis—A software program was used for statistical analyses.^a A χ² analysis was used in univariate analyses of demographic data and exposure data. Crude ORs and their 95% CIs were calculated for each categoric variable. Continuous variables were tested for difference between cases and controls, by use of a 2-sample t test.

For some utility companies, many of the relevant quarterly TTHMs values were missing. Complete quarterly data for the most recent household for 90 control dogs and 94 case dogs were available. Complete data for the last 3 households in which the dog lived for 47 control dogs and 39 case dogs were available. Imputing the missing TTHMs values on the basis of the drinking water utility company data, seasonal fluctuation, yearly fluctuation, water source (surface vs groundwater), and population served by the drinking water utility company were examined. However, in general linear models, it was found that only utility company and year contributed importantly to the variability in log TTHMs concentrations. Therefore, missing values were not imputed. The exposure variable for each dog was the mean TTHMs value over the dog's lifetime. To calculate the mean TTHMs value, mean quarterly data were calculated from utility companies serving the 3 most recent households where each dog lived during the period from birth to date of diagnosis. For TTHMs concentrations reported as below the limit of detection, the value (0.5/√2) was substituted.¹⁶ If the drinking water utility company did not provide TTHMs concentrations for any part of the relevant exposure period, the exposure was considered unknown during that period.

Odds ratios were calculated for varying concentrations of exposure and duration of exposure to chlorinated water. Because some exposure data were missing for some dogs, the analyses were repeated to include only dogs with TTHMs concentration data for at least 90% of their exposure history. The analyses were repeated by use of TTHMs concentration data for an assumed 5-year latency period, ending 2 years prior to diagnosis.

To reflect the frequency matching, conditional logistic regression was used to obtain the maximum likelihood point estimate of the OR and the 95% CI and to adjust for potential confounding effects.^17,18 The data were stratified on age by 2-year categories and sex; therefore, there were strata for every 2-year and sex combination. Age was controlled for in age categories (stratifying by age and sex) in conditional logistic regression. In sensitivity analyses, a model using conditional logistic regression controlling for age in age categories (stratifying by age and sex) was also used and also included a continuous age variable to control for the residual difference in age. In further sensitivity analyses, unconditional logistic regression with age as a variable (age categories in 2-year increments) was performed and included additional model terms for exposure, weight, whether the dog was purebred or mixed breed (to partially control for socioeconomic status and genetic predisposition to develop certain forms of cancer), and history of smoking in the dog's household to estimate the ORs.

Results

A working telephone number was available for 123 case and 132 control households. We were unable to contact 18 case households and 16 control households after a minimum of 3 attempts; thus, contact response rate was 81.3% for case households and 75.8% for control households.

Among successful contacts, 5 owners of case dogs and 16 owners of control dogs refused to participate, yielding participation rates of 95.2% and 86.2%, respectively. The proportions of purebred dogs among the refusals were 80% and 81.25%, respectively.

Results of univariate analyses were determined (Table 1). When comparing cases with controls, no differences were detected in purebred status, source of the dog's drinking water, or exposures to environmental contaminants. Case dogs, however, were slightly younger, weighed less, and drank less water than control dogs. Univariate analyses using medication history (including reproductive hormones, parasiticides, and other medications associated with bladder cancer in humans) and exposure to specific foods, dietary supplements, and flea and tick treatments did not reveal any significant differences between the exposed and control dogs.

Evaluation included 2,029 years of exposure to household drinking water, 70% of which was from surface water sources. Potential risk for bladder cancer from exposure to TTHMs was evaluated by comparing dogs that lived in households with chlorinated drinking water with those that lived in households without chlorinated drinking water (Table 2). Exposure was defined in several ways on the basis of the literature and the etiology of canine bladder cancer. No evidence of an association between bladder cancer and exposure to chlorinated water or with the number of years exposed to chlorinated water in the household was detected. The results did not change when analyses were restricted to dogs for which 90% of their lifetime exposure history was known or to dogs with exposure during an assumed 5-year latency period that ended 2 years prior to diagnosis.

Mean years of exposure to chlorinated surface water was less for case dogs (8.03 exposure-years) than for control dogs (8.88 exposure-years) but not significantly. Results did not change when analyses were restricted to dogs with at least 90% of their exposure history available or to dogs with exposure data for a 5-year latency period that ended 2 years prior to diagnosis.

Mean ± SD TTHMs concentrations for case dogs was 32.6 ± 13.5 μg/L and for control dogs was 31.7 ± 13.5 μg/L (P = 0.7). There were no significant associations between exposure and disease when TTHMs concentrations were treated as a continuous variable or categorized into tertiles or quartiles. When exposure was defined as having a mean TTHMs concentration in the 97th percentile (> 39.97 μg/L), the crude OR was 1.2 (95% CI, 0.58 to 2.7). When exposure was defined as having a mean quarterly TTHMs concentration > 39.97 μg/L and being exposed to chlorinated surface water at some time during the 5-year latency period that ended 2 years prior to diagnosis, the OR was 1.3 (95% CI, 0.54 to 3.1). For conditional logistic regression analyses performed to examine how other variables contributed to the association between TTHMs exposure (TTHMs concentration > 39.97 μg/L) and case status, weight and purebred status were significant in the model (Table 3); however, the exposure variable was not significant in any of the models. In all the sensitivity analyses, the results for the exposure variable were approximately the same; that is, slightly negative and less than half an SD from the null.

Discussion

Canine bladder cancer resembles human bladder cancer with respect to histopathologic characteristics, molecular features, biological behavior including metastasis, response to medical treatment, and prognosis.¹⁹ A genetic predisposition for bladder cancer exists among Scottish Terriers and other terrier breeds.¹¹ Recently, a series of case-control studies in dogs revealed that environmental exposures to herbicides,²⁰ topically administered flea and tick pesticides,²¹ and diet²² are associated with bladder cancer risk in dogs. Obesity and topically administered insecticide exposure were risk factors for bladder cancer in an earlier study²³ of multiple breeds. Thus, because some environmental exposures increase risk for bladder cancer in dogs and chlorination by-products have been associated with an increased risk of bladder cancer in humans, assessment of the role of these compounds in canine bladder cancer is relevant and of potential clinical importance. Furthermore, because exposure to TTHMs can be reduced by allowing water to stand, heating, or substituting bottled water for tap water, preventive strategies could be considered if a substantial risk is determined to be present.

Canine bladder cancer is associated with some of the same exposures that increase human bladder cancer risk.²⁴ Hayes et al¹² reported a positive correlation between the proportional mortality ratios for canine bladder cancer and the level of industrial activity in the county where the dog was hospitalized. Human deaths from bladder cancer in the same US counties were also correlated with industrial activity. In addition, Claude et al²⁵ reported increased risk for bladder cancer in humans employed in rubber, plastics, and textiles industries and those exposed to coal pitch, chromium, and mining activities. Although dogs are not likely to have those occupational exposures directly, bladder cancer has been experimentally induced in dogs administered aromatic hydrocarbons such as those used in the textile industry.¹³ These similarities in risk factors for human and canine bladder cancers suggest that dogs may be good sentinels for environmentally induced cancers.¹²

This study was conducted to determine whether the reported association between exposure to DBPs and bladder cancer in humans could be detected in dogs. Previous studies of human illnesses and exposures to DBPs used various methods to define exposure. Those methods included asking study participants to recall their personal tap water drinking history, creating TTHMs concentration categories by imputing or estimating historical data from currently available data, or both. Defining exposure via recall over a lifetime or via modeled drinking water TTHMs concentrations could easily result in exposure misclassification and biased study results.

In the present study, several methods were used to reduce exposure misclassification and the associated bias. First, it was verified that the participating dogs drank tap water and exposure was then defined by use of historical drinking water utility company TTHMs concentration data for the relevant time period for each home where the dog had lived. In comparison to human studies, recall bias would be less for dogs because the retrospective exposure period was shorter and actual drinking water TTHMs concentration data were available, rather than data generated from modeling.

Second, dog owners were asked to quantify the amount of water their dog drank each day. If a drinking water contaminant adequately represented by the TTHMs concentration is an important risk factor for canine bladder cancer, these 3 variables—drinking household tap water, TTHMs concentration data for the households in which the dogs lived, and an owner's estimate of how much water the dog drank each day—should reduce misclassification and improve exposure assessment. Case and control dogs were equally likely to drink household tap water, indicating potential exposure to drinking water contaminants for both groups. Case dogs weighed less and drank less water than control dogs. Thus, it was not clear whether case dogs were more likely to be exposed to drinking water contaminants than control dogs.

When lifetime exposure was defined as the mean quarterly TTHMs concentration determined from historical quarterly utility company data, results indicated that case dogs were exposed to the same mean TTHMs concentrations over the relevant exposure period as were control dogs. An exposure index similar to the one constructed for this study was used by Villanueva et al²⁶ in their analysis of pooled data from 6 human studies. Historical records of TTHMs concentrations were not available for the relevant time periods; thus, the mean TTHMs exposure was modeled by use of current TTHMs concentrations, information about water sources and treatment methods, or both. On the basis of the mean TTHMs exposure index for each human study, cases were exposed to higher mean TTHMs con-centrations than were controls. The differences between the mean TTHMs exposure for case and controls ranged from 0.2 μg/L to 4.5 μg/L in the 6 studies.²⁶ Although we found a similar magnitude of difference between the mean household TTHMs concentrations for case and control dogs (0.8 μg/L), the 95% CIs included the possibility of no difference between the 2 groups. Including the covariates of weight and history of bladder infection or blood in the urine in conditional logistic regression analyses did not change the results.

We assumed that dogs in the present study would drink the same household water as their owners. However, a dog's exposure to drinking water disinfection by-products may be different from that of the owners. For example, THMs are volatile and concentrations in water left out in a bowl are expected to decrease over time. Thus, dogs whose drinking water has been standing for several hours are likely exposed to lower THMs concentrations than are humans who drink water immediately after obtaining it from the tap.

These results suggest 2 issues about the association between TTHMs exposure and canine bladder cancer. First, TTHMs may not be the appropriate measure of exposure for examination of this association because > 50% of this class of DBPs, total organic halides, formed in chlorinated water remains unidentified.²⁷ Second, even if TTHMs concentrations are good surrogates of exposure to DBPs and the exposure is a potential risk factor, a dog's total exposure to TTHMs is likely lower than a human's living in the same household. Dogs are primarily exposed to tap water by drinking; only infrequently exposed through bathing; and not likely exposed through showering, hand washing, or other household activities. Showering and bathing are important routes of exposure to TTHMs for humans and result in higher increases in TTHMs concentrations in blood and exhaled breath²⁸ than other household exposures such as washing clothes or dishes.²⁶ Villanueva et al²⁹ found that the risk of human bladder cancer was associated with the duration of showering and bathing and TTHMs concentrations, and the risk estimate was higher than that for ingestion of TTHMs alone. Thus, the risk of bladder cancer in dogs from household exposure to TTHMs may be limited because their exposure is primarily through drinking water, rather than through inhaling household aerosols.

There were a number of strengths and weaknesses in the present study. The major strength was limiting exposure misclassification bias by use of historical TTHMs data, rather than extrapolating or estimating TTHMs concentrations from current measurements. Nearly all owners reported giving their dog tap water to drink. Only 1 (0.5%) respondent reported giving their dog bottled water. In addition, 193 (97%) of the dogs were not allowed to roam free, thus limiting their exposure to other sources of drinking water and improving the ability of the tap water measurements to reflect exposure. The use of an animal model should be largely free of potential confounding factors, such as occupational exposures to bladder carcinogens and cigarette smoke. Secondhand smoke was not associated with risk in a case-control study²³ of canine bladder cancer.

The major limitations were associated with the paucity of dogs that were not exposed to chlorinated water or that had only a few years of exposure and missing information on quarterly TTHMs concentrations. This was addressed by evaluating dogs with extremely high TTHMs concentrations (> 97th percentile) in some analyses, but misclassification could still have affected results. Furthermore, volatilization of THMs in a dog's water bowl before the dog drank it could have contributed to exposure misclassification in both the control and exposed groups and reduced the risk estimates as well as yielded lower exposures to chlorination by-products.

On the basis of the etiologic and pathologic similarities between human and canine bladder cancer, we anticipated that exposure to disinfection by-products might also be a risk factor for dogs. However, defining exposure by use of historical TTHMs data from water utility companies did not provide strong evidence for an association. Consistent with reports from studies in humans, dogs with bladder cancer were exposed to higher TTHMs concentrations than were control dogs; however, the difference was not significant. Volatization of THMs in water allowed to stand likely reduced the dose obtained from ingestion. Similarly, the fact that dogs are seldom bathed resulted in lower exposures to TTHMs than would occur in humans. However, other classes of DBPs that are not volatile, such as the haloacetonitriles, would not be affected by those factors, so evaluation of the role of exposure of dogs to DBPs remains a valid approach.