Dogs Detect Prostate Cancer: Give Hope for Better Diagnostic Test

German shepherds, Liu and Zoe, were trained to detect
prostate cancer in urine samples.

The diagnosis of prostate cancer is most often prompted by an elevated PSA (prostate specific antigen) level, or less commonly an abnormal digital rectal examination. An abnormality in PSA or digital rectal examination often prompts a prostate biopsy – almost X performed per year in the United States. However, PSA testing and digital rectal examination are far from perfect (See our prior blog on PSA Screening Controversies). Part of this stems from the fact that both normal and cancerous prostate cells make PSA. Distinguishing a normal PSA from a "cancerous" PSA can be very difficult. Therefore, the accuracy of detecting prostate cancer following an abnormal PSA test is only 30% - with a lot of room for improvement!!

An average dog nose has 200 million olfactory or smell-
sensing cells!  The average human has 5 million.

PSA is not the only molecule that prostate and prostate cancer cells release. For many years, researchers have been looking at the urine of men with and without prostate cancer in search of a better diagnostic test. Recently, researchers in Italy used two highly trained German shepherds to evaluate urine samples of men with and without prostate cancer – and the dogs were correct 98% of the time. Dogs have a dramatically stronger sense of smell than humans. The average human nose has approximately 5 million olfactory cells that detect volatile (airborne) compounds (our sense of smell) – dogs have 200 million. We have known this for years, making use of our canine friends in law enforcement, to detect drugs and bombs; and in medicine. In several studies, dogs have been able to detect the onset of epileptic seizures and to aid in the diagnosis of breast, lung and bladder cancers.

This study was completed in Italy with two German shepherds, a team of urologists, veterinarians, dog handlers, and over 900 urine samples from patients with and without prostate cancer. Three-hundred and sixty two men had prostate cancer in various stages, from very early, low-risk cancer to metastatic disease; the 540 men in the control group had neither benign prostatic hyperplasia nor prostate cancer. The dogs were exposed to half a dozen specimens at a time and were trained to sit down when they detected a cancer. The dogs were accurate in detecting either cancer or no cancer in 98% of cases. In fact, the sensitivity (or ability of the dogs to rule out cancer) was >99%; and if they identified cancer, they were right 97% of the time (specificity).

One of the biggest problems in using dogs in "sniffing" tests is that they can pick up on queues from their trainers. In this experiment, the trainers were completely blinded to the samples and the dogs were left to determine the status of the urine on their own. To ensure well-trained dogs, the process of teaching the dogs to smell cancer took over six months.

One of the biggest questions that remains is what were the dogs actually smelling? The dogs could be smelling a single molecule present in the cancer or a combination of chemicals produced by the cancer cells. In fact, it may not even be a cancer molecule, but a product of the microenvironment of that man's prostate that allowed the cancer to grow in the first place. Additional laboratory studies using complex gas chromatographers or "electronic noses" are needed to determine the exact molecule or chemicals present in the urine of men with prostate cancer. Whether dogs are in the future of prostate cancer detection remains to be seen, but this study proves that there is something detectable in the urine of men with prostate cancer that can be detected with incredible precision!

This study was presented by Dr. Gian Luigi Taverna, MD, at the American Urological Association (AUA) Annual Meeting in Orlando, Florida in May.  Read the abstract below:

PD19-01: Prostate Cancer Urine Detection Through Highly-Trained Dogs’ Olfactory System: A Real Clinical Opportunity 
Gianluigi Taverna*, Milan, Italy, Lorenzo Tidu, Grosseto, Italy, Fabio Grizzi, Guido Giusti, Mauro Seveso, Alessio Benetti, Rodolfo Hurle, Silvia Zandegiacomo, Luisa Pasini, Alberto Mandressi, Pierpaolo Graziotti, Milan, Italy
Abstract: PD19-01 
Introduction and Objectives: The analysis of volatile organic compounds (VOCs) in urines is a promising approach to cancer detection. Here, we establish the level of accuracy at which a rigorously trained canine olfactory system can recognize specific prostate cancer-VOCs in urine samples, thus reducing unnecessary prostate biopsies, and pinpointing patients at high-risk for prostate cancer. 
Methods: A total of 677 subjects were investigated. All the subjects included in the study were placed in one of two main Groups: Prostate Cancer Group (n = 320) and Control Group (n = 357). Prostate Cancer Group includes patients with PC ranging from those at very-low risk to metastatic PC. Control Group includes a large and heterogeneous cohort of healthy subjects or patients affected by non-neoplastic diseases or non-prostatic tumors. Each patient's clinical status and therapeutic regiment was known. We took into consideration all predictive values currently used in the clinical practice for prostate cancer management. Two dogs and a full-time, highly specialized, multidisciplinary team was dedicated. We have standardized a “work training” procedure and eliminated any possible olfactory interference. Each test was carried out blindfold. 
Results: The dogs achieved the following performances: Dog 1: sensitivity 100%, specificity 97.8%, accuracy 98.9%. Dog 2: sensitivity 98.6%, specificity 95.9%; accuracy 97.3%. When considered together, Dogs 1 + 2, we found an accuracy of 98.1% with a sensitivity of 99.2% and a specificity of 97.1%. Whether excluded the female control sub-groups we found that Dog 1 achieved a sensitivity of 100% and a specificity of 97.9%, while Dog 2 a sensitivity of 98.6% and a specificity of 94.5%. When evaluating the wrongly detected cases no differences were found between epidemiological, clinical or histopathological characteristics. 
Conclusions: The present study first demonstrates that a trained canine olfactory system detects specific PC-VOCs in urine samples from a large number of patients with PC at different stages and risks versus a heterogeneous control group, which is unthinkable in current clinical urological practice. Thanks to the early intuition, we have definitely turned what used to seem a myth into a real clinical opportunity.