Background: Canine cancer is a leading cause of canine deaths, often resulting from complex interactions between germline-risk genetics, somatic mutations, and environmental exposures. To help identify major dietary, genetic, and environmental exposure risk factors for canine cancer, Morris Animal Foundation launched the Golden Retriever Lifetime Study, the first prospective longitudinal study in veterinary medicine. We hypothesized that responses from the physical activity section of the GRLS annual questionnaire could be used to develop a BiMM forest model that accurately classifies which Golden Retrievers develop cancer within the first seven years of the study. Furthermore, we expected that the most important predictors of cancer development would be the frequency and duration of the physical activity, with more rigorous activities—such as swimming—would be the most important predictors of cancer development.
Methods: Activity and lifestyle questionnaire data for 3,044 purebred Golden Retrievers enrolled in the Golden Retriever Lifetime Study were obtained from Morris Animal Foundation. Two BiMM forest models were developed to predict the development of cancer: the “Years 0–7” model using consistently asked questions over the seven years, and the “Years 3–7” model, which incorporated additional questions about the pace and duration of physical activity starting in study year 3.
Results: Of the enrolled dogs, 277 were diagnosed with cancer. The “Years 3–7” model achieved the best performance, with overall accuracy of 80.7%, a F1 score of 74.9% and a fair ROC AUC of 0.763. Key predictors of cancer development included year in study, frequency, pace, duration, and the frequency of warm and cold weather swimming. After Golden Retrievers were diagnosed with cancer, owners reported an 8–10% increase in exercise frequency and a 15.6% to 68.88% increase in cold weather swimming whereas warm weather swimming decreased by 2.0% to 13.9%. Similar declines in the pace and duration were also observed. The surface type where the exercise took place and the specific types of physical activity were lower in importance.
Conclusions: Including pace and duration of the physical activity improved model performance, highlighting these predictors as high importance alongside the frequency of the physical activity. Future prospective studies should seek to determine specific physical activity guidelines for dogs, focusing on frequency, duration, and pace to potentially reduce cancer risk.