Two-step investigation of lung cancer detection by sniffing dogs

Early detection of lung cancer (LC) is a priority since LC is characterized by symptoms mimicking other respiratory conditions, but remains the leading cause of oncological disease death. Properly trained dogs can perceive the volatile organic compounds (VOCs) related to cancer, thanks to their acute sense of smell. Dogs' use for LC detection could be advantageous: reliable trained dogs would represent a valuable, cost-effective, non-invasive method of screening, which gives a clear-cut yes/no response. However, whether sniffer dogs are able to maintain their discriminative capacity at long-term control, and in different types of environments, needs further investigation. In this study, we sought to test two hypotheses: firstly, if dogs can be trained to perceive LC-related VOCs in human urine, a substrate which may be a good candidate for large-numbers screening, and secondly, whether trained dogs retain their performance stable over time, even if the environment in which the tests are carried out varies. We have selected three family dogs that underwent a one-year training period (two weekly training sessions), by clicker training method. At the end of the training, dogs underwent two separate test phases, in two different locations, one year apart. All other procedures had been maintained unchanged. The donors of the samples submitted to the dogs were recruited by the European Institute of Oncology (IEO), Milan, Italy. Results show that dogs had different sensitivity (range: 45%-73%) and specificity rate (range: 89%-91%), and were deceived neither by lung conditions (that the dogs did not consider) nor by the existence of tumors in the beginning stage, that was correctly reported by dogs. The one-year interruption of the research work and the changes in the test environment did not induce statistically significant differences in the dogs' perceptive capacity. To our knowledge, so far, these issues have never been highlighted.