The study was designed as a prospective controlled study. Samples of exhaled air were collected from 30 patients with solitary pulmonary nodules (SPN) (25 with stage 1A disease) before the diagnostic and therapeutic surgery, and from 77 controls — 18 with chronic obstructive pulmonary disease and 61 healthy subjects.
Patients with other neoplasms and those who had received chemotherapy were excluded from the study. Participants had to refrain from eating, drinking, or smoking 30 minutes before they exhaled in a crystal tube filled with hydrophilic and hydrophobic wool and closed with silicon taps. Each tube was introduced in a wooden box with an open side to enable Blat to smell.
Blat was trained on a progressive prize-dependent learning method. The dog was confronted with samples with and without the pulmonary nodules in a proportion of 1:4.The lung cancer sample was also randomly assigned a different physical location each time Blat was presented with a sample to smell.
Results obtained with Blat were compared with the final pathology reports.
Blat was confronted with 90 breath samples of patients with SPN (3 per person) and 382 control samples, 10 times to each sample of SPN with different combinations of the control exhaled gas samples — for a total of 900 attempts. Blat sat beside the correct malignant samples 879 times. Of 30 patients with SPN, Blat was able to recognize 27 as positive for lung cancer as per the pathology report. Blat’s responses provided a predictive positive value of 0.97 and a predictive negative value of 0.99.
“Trained dogs can identify accurately the malignant origin of SPN, so they can be helpful for their diagnosis management,” Dr Guirao said in an interview with Cancer Therapy Advisor. “These results prove that the exhaled gas of patients with lung cancer has a distinct odor (likely in relation to the presence of a specific VOCs signature). The challenge is now to identify such signature using appropriate technology (e-nose),” she added.
When asked whether such an approach can affect patient management, Dr Guirao indicated that dogs can be very useful in the development of lung cancer screening programs. “We completely believe that dogs’ help would be mandatory in the detection of specific VOCs, which is key for the development of new noninvasive lung cancer diagnosis methods,” she said.
Diagnosis of early-stage lung cancer poses a challenge, because more than 75% of patients are diagnosed when disease has advanced beyond the point of being eligible for a cure, Dr Guirao explained. “It is mandatory to develop lung cancer screening programs, and we think that VOCs and lung cancer exhaled breath detection could be used in combination with computerized tomography screening recommendations,” she said.
How Blat Was Trained
Blat was trained based on positive reinforcement through a reward-based method aimed at the association of smell patterns. Dr Guirao explained that Blat was first exposed to the odor of interest (ie, the exhaled gas of a patient with lung cancer), followed by a reward, which enabled him to associate the smell with the reward, memorize it, learn to recognize it, and then sit upon recognition. “This training phase was concluded when the dog sat automatically when confronted [with] a known lung cancer sample,” Dr Guirao said. In the second phase, the dog trainer exposed the dog to 5 different exhaled gas samples each in its own wooden container; only one of those samples came from a patient with lung cancer.
Blat was then trained to indicate which samples came from patients with lung cancer by sitting next to the boxes that contained these samples. “It took us 4 to 5 months to train Blat. With the exception of plane snout dogs, nearly all dogs can be trained,” Dr Guirao said.
The dog is selected by a specialist dog trainer who looks for skills, including playfulness and gluttony — if the prize is something the dog likes to eat, he will do his best to get it.
- Guirao Á, Molins L, Ramón I, et al. Trained dogs can identify malignant pulmonary nodules in exhaled gas. Presented at: International Association for the Study of Lung Cancer World Conference on Lung Cancer 18th Annual Meeting; Toronto, Canada: September 23-26, 2018. Abstract MA03.11.
- International Association for the Study of Lung Cancer. Study demonstrates reliability of trained dogs in the identification of malignant pulmonary nodules [press release]. Published September 24, 2018. Accessed September 27, 2018.
- Krilaviciute A, Heiss JA, Leja M, Kupcinskas J, Haick H, Brenner H. Detection of cancer through exhaled breath: a systematic review. Oncotarget. 2015;6(36):38643-38657.
- Guirao Montes Á, Molins López-Rodó L, Ramón Rodríguez I, et al. Lung cancer diagnosed by trained dogs. Eur J Cardiothorac Surg. 2017;52(6):1206-1210.