COVID‐19 sniffer dog experimental training: Which protocol and which implications for reliable sidentification?

Abstract The introduction of trained sniffer dogs for COVID‐19 detection could be an opportunity, as previously described for other diseases. Dogs could be trained to detect volatile organic compounds (VOCs), the whiff of COVID‐19. Dogs involved in the study were three, one male and two females from different breeds, Black German Shepherd, German Shepherd, and Dutch Shepherd. The training was performed using sweat samples from SARS‐CoV2 positive patients and from SARS‐Cov2 free patients admitted at the University Hospital Campus Bio‐medico of Rome. Gauze with sweat was collected in a glass jar with a metal top and put in metal boxes used for dog training. The dog training protocol was performed in two phases: the olfactory conditioning and the olfactory discrimination research. The training planning was focused on the switch moment for the sniffer dog, the moment when the dog was able to identify VOCs specific for COVID‐19. At this time, the dog was able to identify VOCs specific for COVID‐19 with significant reliability, in terms of the number of correct versus incorrect (p < 0.0001) reporting. In conclusion, this protocol could provide a useful tool for sniffer dogs' training and their introduction in a mass screening context. It could be cheaper and faster than a conventional testing method.

such as breast and lung cancers with a percentage of detection rate ranging from 88% to 99% 4 , a malarial disease with a detection rate of about 82% 5 , and viral or bacterial infections with a detection rate of 77%-92.6%. [6][7][8] Several volatile organic compounds (VOCs) account for the odor released by the expiration phase of breathing, skin emanation, urine and breath vapors, saliva, and pathological conditions. These odors depend on biochemical modification occurring in the body with the consequent release of these specific compounds that are volatile. 9 The metabolic changes occurring in the body in presence of specific conditions, such as inflammation, infections, or neoplastic disease can be recognized by the dogs that are provided by a powerful olfactory apparatus if adequately trained for detection of the VOCs. 10 The same approach could be used for COVID-19 detection, as described in previous studies. [11][12][13][14] The VOCs could be useful in clinical diagnosis of different disease including bacterial and viral infections as SARS-CoV-2 causing COVID-19interstizial bilateral pneumonia. [11][12][13][14] In a recent study, dogs professionally trained were evaluated for glucose level detection in patients with diabetes. This study suggested that dogs, after adequate training, have the ability to detect hypo and hyperglycemic conditions. 15 VOCs' detection-trained dogs could provide early detection of SARS Cov-2 infected patients at low cost. The trained dog has the ability to screen more than 200 individuals per hour, enough to allow mass screening at airports, stadiums, or in case of crowded events where the virus transmission control by asymptomatic individuals is fundamental. This is in agreement with World Health Organization recommendation about mass screening and its application also in low-income countries where the use of sophisticated and expensive screening tools could be limiting.
The study aims to evaluate the sniffer dogs' ability to discriminate VOCs emanated by the skin in course of COVID-19, demonstrating that this disease is characterized by a specific odor and that dogs are able to identify it efficiently and quickly.

| Experimental design
The training planning was developed involving dogs from different breeds. Dogs involved in the study were from different working dog breeds since their features are useful to standardize the characteristics, the management and the training coherence, to the advantage of more homogeneity in results recording. The intention was to concentrate the experiment on high quality rather than on the number of dogs. In fact, dogs were selected for their specific talents suitable to this kind of experimental design, such as temperament, docility, and resistance.
COVID-19 conditioned dogs, once involved in the study, will be recoverted to other activities of safety and security, if necessary or at the end of the pandemic, to guarantee the service continuity and mental and physical dogs health in the future. The study was approved by the Local Ethics Committee of University campus Bio-Medico of Rome (PAR 17.21 OSS).
The dogs' training plan was divided into two steps: the first step was "specific conditioning" to COVID-19 VOCs, consisting of the association of the odor research and consequent reporting. This critical and fundamental step is developed using several sweat samples from patients admitted to the COVID Center of the University Hospital Campus Bio-Medico of Rome, for COVID-19. The second step of "olfactory discrimination research" consisted of the discrimination between the COVID-19 odor of interest and everything else that has to be discarded. Here, the discrimination was

| Sniffer dogs' characteristics
The dogs involved in this study belonged to the SecurityDogs, a brand of NGS srl Security company (Italy), one male and two females from three different breeds: Black German Shepherd, German Shepherd, and Dutch Shepherd. The demographic characteristics of dogs are reported in Table 1.

| Materials used for sweat samples collection and training
The gauze was the elective support chosen for sweat collection, for its common distribution and consequent easy availability. The gauze used belongs to Class IIa surgical device for its specific characteristics to be sterile, 100% cotton, latex, and phtalate free ( Figure 1A).
Gauze with sweat was collected in a glass jar with the metal top

| Statistical analysis
The percentage difference between correct and incorrect identification registered in the verification of the protocol procedure for each dog was evaluated by χ 2 test for proportions. A p value of less than 0.05 was considered statistically significant.

| De-briefing session
Each training session has been video-recorded (as in the supplementary material) for further analysis during the de-briefing section performed at the end of each training session. Moreover, a specific report has been made for the training sessions tracing.

| Sanitizing
Training site and training equipment were daily sanitized at the end of each training session, while wastes were disposed of in special waste containers.

| Troubleshooting
Sweat self-collection by inserting gauze in the underarms of the patients has to be carefully performed, as previously described, to avoid any influence on training.

| RESULTS
In

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

AUTHORS CONTRIBUTIONS
Silvia Angeletti and Massimo Ciccozzi: study design, data analysis, and study supervision. Francesco Travaglino, Silvia Spoto, Maria