Introduction: Sniffer dogs can diagnose lung cancer. However, the diagnostic yields of different samples and training methods for lung cancer remain undetermined.
Objective: Six dogs were trained in three stages with the aim of improving the diagnostic yield of lung cancer by comparing training methods and specimens.
Methods: The pathological tissues of 53 lung cancer patients and 6 non-lung cancer patients in the Department of Thoracic Surgery of Kaohsiung Chang Gung Hospital were collected, and the exhaled breath samples and urine samples were collected. Urine and exhaled breath samples were also collected from 20 healthy individuals. The specimens were sent to the Veterinary Department of Pingtung University of Science and Technology.
Results: The dogs had a very low response rate to urine target samples in the first and second stages of training. The experimental results at the second stage of training found that after lung cancer tissue training, dogs were less likely to recognize lung cancer and healthy controls than through breath target training: the response rate to exhaled breathing target samples was about 8-55%; for urine target samples, it was only about 5-30%. When using exhaled air samples for training, the diagnosis rate of these dogs in lung cancer patients was 71.3% to 97.6% (mean 83.9%), while the false positive rate of lung cancer in the healthy group was 0.5% to 27.6% (mean 7.6%). Compared with using breathing target samples for training, the diagnosis rate of dogs trained with lung cancer tissue lung cancer was significantly lower (p < 0.05). The sensitivity and specificity of lung cancer tissue training (50.4% and 50.1%) were lower than the exhaled breath target training (91.7% and 85.1%). There is no difference in lung cancer diagnostic rate by sniff dogs among lung cancer histological types, location, and staging.
Conclusion: Training dogs using breathing target samples to train dogs then to recognize exhaled samples had a higher diagnostic rate than training using lung cancer tissue samples or urine samples. Dogs had a very low response rate to urine samples in our study. Six canines were trained on lung cancer tissues and breathing target samples of lung cancer patients, then the diagnostic rate of the recognition of exhaled breath of lung cancer and non-lung cancer patients were compared. When using exhaled air samples for training, the diagnosis rate of these dogs in lung cancer patients was 71.3% to 97.6% (mean 83.9%), while the false positive rate of lung cancer in the healthy group was 0.5% to 27.6% (mean 7.6%). There was a significant difference in the average diagnosis rate of individual dog and overall dogs between the lung cancer group and the healthy group (p < 0.05). When using lung cancer tissue samples for training, lung cancer diagnosis rate of these dogs among lung cancer patients was only 15.5% to 40.9% (mean 27.7%). Compared with using breathing target samples for training, the diagnosis rate of dogs trained with lung cancer tissue lung cancer was significantly lower (p < 0.05). The sensitivity and specificity of lung cancer tissue training (50.4% and 50.1%) were lower than the exhaled breath target training (91.7% and 85.1%). The diagnostic rate of lung cancer by sniffer dogs has nothing to do with the current stage of lung cancer, pathologic type, and the location of tumor mass. Even in stage IA lung cancer, well-trained dogs can have a diagnostic rate of 100%. Using sniffer dogs to screen early lung cancer may have good clinical and economic benefits.
Keywords: breathing target samples; detection dog; lung cancer.