Dogs’ Detection of Symptomatic and Asymptomatic SARS-CoV-2 by Non-Working Dogs: Feasibility and Limits Under Controlled Laboratory Conditions

TL;DR

This study explores whether dogs can detect symptomatic and asymptomatic SARS-CoV-2 infections, finding that they can generalize for Delta variant but perform worse with Omicron samples from vaccinated individuals.

Contribution

Demonstrates the feasibility of using non-working dogs to detect asymptomatic SARS-CoV-2 cases and highlights limitations with Omicron variant and vaccinated individuals.

Findings

Dogs trained on Delta samples could detect both symptomatic and asymptomatic cases with similar performance.

Performance dropped significantly when testing asymptomatic Omicron samples from vaccinated individuals.

A line-up protocol improved detection compared to a strict yes/no protocol.

Abstract

Dogs can detect odors associated with infectious diseases, but their ability to distinguish between symptomatic and asymptomatic SARS-CoV-2 infections has not been fully established. In this exploratory proof-of-concept study with a limited number of dogs, eight non-working dogs began training with either symptomatic or asymptomatic samples of the Delta variant. We next tested whether dogs trained with asymptomatic samples could detect symptomatic samples, and vice versa. Poor performance in a task where odors had to be strictly identified (yes/no protocol) led us to use a simpler task in which dogs had to discriminate the targeted odor among five other samples. When presented with novel Delta samples, each group of dogs significantly discriminated them from controls, and the two groups performed equally well. In contrast, performance dropped when dogs were tested with a small number of asymptomatic Omicron-positive samples from vaccinated individuals. These results show that dogs can generalize across symptomatic and asymptomatic COVID-19 cases for the Delta variant but showed reduced performance with Omicron samples, likely due to altered odor profiles in vaccinated individuals. While proof-of-concept feasibility was demonstrated in a small group of dogs using a categorization task for detecting asymptomatic scents, detection dogs should not currently be recommended for large-scale screening.

Dogs can detect volatile organic compounds (VOCs) associated with infectious diseases, but their ability to distinguish between symptomatic and asymptomatic SARS-CoV-2 infections has not been fully established. In this exploratory proof-of-concept study with a limited number of dogs, eight non-working dogs (without experience in scent detection) began training with either symptomatic or asymptomatic Delta samples; seven reached the testing phase and were then evaluated on the alternate group to assess cross-generalization. Training initially used a yes/no protocol but was adapted to a line-up design after poor performance. When presented with novel Delta samples from the other category, dogs significantly discriminated them from controls, achieving a mean sensitivity of 71% and a specificity of 51%, with no difference between training groups. In contrast, performance dropped when dogs were tested with a small number of asymptomatic Omicron-positive samples from vaccinated individuals, with a mean sensitivity of 35% and a specificity of 55%. Dogs can thus generalize across symptomatic and asymptomatic COVID-19 cases for the Delta variant; performance decreased for Omicron samples, likely due to altered VOC profiles in vaccinated individuals. While proof-of-concept feasibility was demonstrated with a line-up protocol for detecting asymptomatic scent, detection dogs should not currently be recommended for large-scale screening; the findings underscore the need for standardized protocols and variant-specific retraining.

Figures