The scent detection prowess of dogs has prompted interest in their ability to detect cancer. The purpose of this study was to determine whether dogs could use olfactory cues to discriminate urine samples collected from dogs that did or did not have urinary tract transitional cell carcinoma (TCC), at a rate greater than chance. Dogs with previous scent training (n=4) were initially trained to distinguish between a single control and a single TCC-positive urine sample. All dogs acquired this task (mean =15±7.9 sessions; 20 trials/session). The next training phase used four additional control urine samples (n=5) while maintaining the one original TCC-positive urine sample. All dogs quickly acquired this task (mean =5.3±1.5 sessions). The last training phase used multiple control (n=4) and TCC-positive (n=6) urine samples to pro-mote categorical training by the dogs. Only one dog was able to correctly distinguish multiple combinations of TCC-positive and control urine samples suggesting that it mastered categorical learning. The final study phase evaluated whether this dog would generalize this behavior to novel urine samples. However, during double-blind tests using two novel TCC-positive and six novel TCC-negative urine samples, this dog did not indicate canine TCC-positive cancer samples more frequently than expected by chance. Our study illustrates the need to consider canine olfactory memory and the use of double-blind methods to avoid erroneous conclusions regarding the ability of dogs to alert on specimens from canine cancer patients. Our results also suggest that sample storage, confounding odors, and other factors need to be considered in the design of future studies that evaluate the detection of canine cancers by scent detection dogs.
Keywords: cancer detection dogs; cancer odor; multiple sample learning; olfactory memory; urinary tract cancer.