Dogs can learn to detect chemicals in a man’s urine that may signal a higher risk of having aggressive prostate cancer, according to a new study.
Prostate cancer is the second leading cause of cancer death in men in the developed world. Clinicians have sought accurate and reliable noninvasive diagnostic tools to differentiate early stage, less dangerous, and more treatable stages of the disease from the aggressive, high-grade, and likely-to-spread forms.
Standard blood tests for early detection, such as the prostate specific antigen (PSA) test, often miss cancers in men whose PSA levels are within normal levels or overdiagnose men with clinically insignificant tumors or no cancer at all.
In a small study in PLOS ONE, researchers had two dogs sniff samples of urine from men diagnosed with high-grade prostate cancer and from men without cancer.
The researchers had trained the animals, Florin, a 4-year-old female Labrador, and Midas, a 7-year-old female wirehaired Hungarian vizsla, to respond to cancer-related chemicals—known as volatile organic compounds (VOCs)—added to urine samples and not respond to ones without them.
“Besides PSA, other methods to detect prostate cancer make use of a molecular analyzer called a gas chromatography-mass spectrometer [GC-MS] to find specific VOCs or profiling bacterial population in a urine sample looking for species associated with cancer, but these have limitations,” says Alan Partin, urologist-in-chief at Johns Hopkins Hospital.
“We wondered if having the dogs detect the chemicals, combined with analysis by GC-MS, bacterial profiling, and an artificial intelligence (AI) neural network trained to emulate the canine cancer detection ability, could significantly improve the diagnosis of high-grade prostate cancer.”
Adding the AI analysis, helped the researchers filter the more than 1,000 VOCs present in a typical urine sample down to those most beneficial for cancer diagnosis, Partin says.
The dogs performed their cancer detection roles well, Partin says. Both Florin and Midas identified five of seven urine samples from men with cancer, or 71.4% accuracy. Florin correctly identified 16 of the 21 non-aggressive or no cancer samples (76.2%), and Midas picked out 14 (66.7%).
When researchers combined the canine olfactory (smell) results with GC-MS, bacterial profiling and AI analysis, the multisystem approach proved a more sensitive and more specific means of detecting lethal prostate cancer than any of the methods alone.
A long history of “biobanking” valuable patient samples made the recent study, and other prostate cancer research at Johns Hopkins Medicine, possible, Partin says, adding that “larger sample pools will be the key enabler of statistically powered, multi-institutional future studies seeking to fully integrate VOC and microbiota profiling.”
Additional coauthors are from Medical Detection Dogs in the UK, Massachusetts Institute of Technology, the Prostate Cancer Foundation, and the Cambridge Polymer Group.
