Objectives: To determine whether canines could be trained to identify patients with cancer by sniffing the urine obtained from a patient with breast or prostate cancer from among samples obtained from healthy volunteers.
Design: Dogs of different breeds were trained by their owners to detect the urine sample from a patient with cancer from among 6 other age- and sex-matched healthy volunteers. After the training was completed, using new samples, 2 test runs were used for each patient with breast cancer and three runs for the patients with prostate cancer against the same matched samples. The configuration of the samples was different for each run. A total of 18 and 33 runs were carried out, respectively.
Results: For each cohort, specificity and sensitivity were measured. In the breast cancer tests, of 6 dogs, only 2 performed better than chance in specificity and none were more sensitive than chance. For the prostate sample testing, 4 dogs were used. Two performed significantly better than chance in specificity and none in sensitivity.
Conclusions: Although this study did not produce the outcomes desired, the literature supports a potential to use canines for human cancer detection. Better management of urine samples and a more stringent training protocol during our study may have provided new evidence as to the feasibility of using canines for cancer detection. A comparison of the 3 dog cancer scenting studies is also presented.