Background: Neoplasia has been linked to exercise levels in humans for some time. Due to the similarities in the behavior of canine and human tumors this has often been extrapolated to dogs as well. The purpose of this study is to explore how exercise levels in canines correlate with the development of neoplasia.Hypothesis/
Objectives: We hypothesize machine learning models will be able to correctly classify if a golden retriever develops cancer, and important predictors will be related to the level of aerobic exercise over the patients’ lifetime.Animals: The first 7 years of available medical records and owner surveys from 3,044 golden retrievers enrolled in the Morris Animal Foundation (MAF) Golden Retriever Lifetime Study (GRLS) were included.
Methods: All golden retriever records related to neoplastic diagnosis, activity and exercise engagement underwent data pre-processing and feature selection prior to the development of binary mixed model (BiMM) forest prediction models.
Results: Preliminary BiMM models incorporating 14 variables performed well when classifying if a golden retriever would develop cancer (average out-of-bag error of 4.77%) . The top 5 most important predictors were the pace; average yearly activity level, the total time engaged in activity, the grade of the surface, and frequency of exercise.Conclusions and Clinical Importance:Our results suggest that overall amount of time exercising and the pace of the activity is more influential than type of exercise. These results can help clinicians provide better recommendations related to exercise when discussing with owners how to reduce the risk of cancer.
