TABLE 1.

Summarizing comments and recommendations for medical canine scent detection of samples from SARS-CoV-2-infected individuals.

Disease- or metabolism-derived VOCs	Canine detection of SARS-CoV-2-infection is thought to be mainly based on the detection of volatile organic compounds (VOCs). Canine detection of VOCs can occur in real-time with high level of accuracy. However, VOC-detection is susceptible to environmental factors, which can be difficult to standardize. The success of medical detection dogs’ VOC-detection depends largely on training with the right variety of target odors.
Ethical considerations	Dogs have different personalities and experiences. They are sentient beings. The learning method should only include positive reinforcement. Dogs can fatigue and get frustrated, which should be considered in the training procedure and when they are deployed in the field. Thus, dogs require adequate work/break cycles and regular positive rewards for their work.
Dog selection	Not only anatomical but also the dog’s behavior and personality significantly impacts suitability as a detection dog. Physical and mental fitness as well as high levels of motivation are of crucial importance. Dogs should have a solid willingness to work with humans. Prior detection experience can be helpful.
Dog training	Appropriate training is the key for success in detection. Defining the correct target scent in advance is challenging, especially when the VOC-profile of interest remains unknown. The right grade of olfactory generalization vs. discrimination has to be achieved during training. Sufficient variety of new samples of symptomatic and asymptomatic patients at different stages of the disease process are here required. Duration of training can be variable and should be tailored to the individual dog’s success rate. Few days of “retraining” dogs after a longer break are sufficient to reach initial levels of detection accuracy. Line-up, scent-wheel, and detection dog training system (DDTS) have been used for training successfully. Apart from imprinting the specific scent to be recognized, also the search context needs to be trained for. While automated approaches such as DDTS might offer a more randomized and rapid training by providing higher repetition rates, line-up settings are closer to the search context in the field. Blank trials are important in order to test for forced choice decisions and to understand the individual dog’s frustration threshold. Dogs should not only be trained with negative samples, but also ideally with samples from other viral respiratory infections to reduce false-positive rates. Further work is needed to standardize and certify training procedures.
Susceptibility of dogs for SARS-CoV-2	Dogs can be infected with SARS-CoV-2, but have a low susceptibility to the virus. Clinical signs are, if at all present, mild. However, biosecurity measures for safe sample presentation, such as virus inactivation and/or safety sample containers during training and/or deployment are recommended, not only for the dogs but also for the handlers.
Sample types	Saliva, sweat, urine, and breath but also respiratory secretions and immediate body odor of SARS-CoV-2-infected individuals express specific COVID-19-associated VOC-profiles, which can be used for training and testing. Sweat collected with cotton pads is not thought to be infectious. Other sample types can be infectious and should be inactivated or presented in a container ensuring biosecurity. Only inactivation procedures should be used, which have shown not to alter the target scent and could bias canine scent detection (see below). Most studies have used sweat samples for practicality reasons. However, it is not clear if cotton-bound VOCs have a similar storage resilience than fluid-bound VOCs such as saliva or urine, which may impact training. Further work is required to provide standardized sample materials.
Virus inactivation	Beta-propiolactone (BPL), heat, ultraviolet radiation (UV), and detergent/solvent are possible measures for virus inactivation. While BPL does not appear to alter canine VOC-detection, heat and detergents might have a greater impact on altering VOC-profiles, which remains ambiguous for UV. However, the use of BPL-inactivation is more time-consuming, requiring laboratories with high safety standards. The least VOC-altering method is to omit inactivation, which works especially well for sweat samples, providing a neglectable risk for infection. In general, biosecurity aspects should never be disregarded and be approved by authorities.
Training sample alternatives	Currently, well-established sample alternatives for a more standardized training for COVID-19-detection do not exist. Artificial “VOC-cocktails,” samples from animal models, cell cultures, or pure virus protein are currently being tested and the tests are not yet conclusive. It is likely that proteins can only be used in parts of the training and that the certification procedure will require samples from SARS-CoV-2-infected individuals.
Target population and operational applicability	Studies showed high accuracies for canine COVID-19-detection within seconds with similar or better detection performances than with antigen tests. Depending on disease prevalence and characteristics of the population to be screened, the performance can alter. To ensure certainty in defining the infection/disease status of an individual, multiple back-up dogs can be involved. Changing or distracting environmental factors in the operational setting should be reduced or avoided.