Dog Sniffs Out Lung Cancer, but Something’s Fishy

Pam Harrison

October 02, 2018

TORONTO — A trained dog can distinguish between malignant and benign pulmonary nodules in exhaled breath samples from patients with and those without pulmonary nodules virtually 100% of the time, new research suggests.

"In theory, nearly all dogs can be trained to detect volatile organic compounds [VOCs], which are contained in exhaled breath samples from patients with malignant pulmonary nodules," lead author Angela Guirao, MD, a thoracic surgeon from the Hospital Clinic of Barcelona in Spain, told Medscape Medical News.

"Now the challenge is to identify the VOC pattern detected by the dogs," she added.

The prospective, controlled study was designed to see whether a trained dog could discriminate between patients with and those without malignant pulmonary nodules by sniffing exhaled breath samples from 30 patients with indeterminate pulmonary nodules; 25 of the samples were subsequently identified as stage 1A disease.

The new data were presented here at the 19th World Conference on Lung Cancer (WCLC).

Breath samples were collected prior to the patients' undergoing diagnostic and therapeutic surgery.

Breath samples were also taken from 18 patients with chronic obstructive pulmonary disease and 61 healthy control persons.

Tubes containing the breath samples were placed in wooden boxes that had an opening to enable the dog to smell the sample. The boxes were then randomly placed on the floor.

The dog was then allowed to sniff all the boxes. When a sample from a patient with a malignant pulmonary nodule was identified, the dog would sit before the box, where he would be rewarded with a treat.

In total, the dog was introduced to 90 samples of exhaled gas from patients with indeterminate pulmonary nodules (three per patient) and another 372 samples from patients without pulmonary nodules.

"The dog was confronted 10 times to each sample of pulmonary nodules with different combinations of no lung cancer exhaled gas samples," the investigators explained.

Of a total of 900 attempts, the dog got the right answer 879 times, Guirao reported.

The dog also matched the pathology analysis in detecting 27 malignant nodules and three benign nodules from the whole sample, she added.

Thus, the dog achieved successful results with a 97% sensitivity, a 99% specificity, a 97% positive predictive value, and a 99% negative predictive value.

"These results indicate that trained dogs can accurately identify a malignant solitary pulmonary nodule," Guirao said.

Does It Pass the Smell Test?

An expert not involved with the study believes the results likely reflect the dog's extraordinary knowledge of his trainer.

Alain Tremblay, MD, professor of medicine, University of Calgary, in Alberta, Canada, noted that most, if not all, of these dog-sniffs-out-cancer studies suffer from the same design flaws.

"First, the trainers who are in the room with the dog know which sample is the positive one," he explained. Because dogs learn to read their trainers very closely, "the dog will take any miniscule cue from their trainer — maybe the trainer might just glance at the sample for half a second — but if the dog picks up on that, it will 'guess' which sample is the positive one," Tremblay said.

The other feature about studies that are designed to demonstrate how well dogs can sniff out cancer is the simple fact that dogs don't need to distinguish which sample smells like cancer. Instead, they just have to recognize which sample smells different from the other four samples.

Dogs are typically presented with five samples at a time, one of which is positive for cancer. This is a far cry from a realistic screening situation, Tremblay noted.

Furthermore, he said that people who serve as controls are often very dissimilar to patients with lung cancer.

"When controls are more similar to the population with cancer, the dogs' success rate is usually much lower than it is for studies whose controls are more 'off the street' controls and who are not at risk for the disease," Tremblay observed.

These variables are "pretty strong biases" and likely make the detection of lung cancer "much better than it probably is," Tremblay concluded.

Trainers and dogs spend a lot of time together, lead author Guirao acknowledged.

It takes an experienced dog trainer about 5 months to teach a dog how to distinguish breath samples from a patient with lung cancer from those from a patient without cancer, she noted.

This is not the first study of its kind with this particular dog. In an earlier study, these Spanish investigators reported that the same trained dog was able to identify lung cancer in exhaled gas from lung cancer patients with extremely high accuracy.

Neither Dr Guirao nor Dr Tremblay has disclosed any relevant financial relationships.

19th World Conference on Lung Cancer (WCLC). Abstract MA03.11, presented September 24, 2018.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.