PSIV-3 Canine olfaction as a disease detection technology: A systematic review

Abstract

Abstract Capitalizing on canine olfactory capacity is a promising strategy for detecting and diagnosing human, animal, and plant diseases. The purpose of this review was to assess the extent of current research in canine disease detection. In this systematic review, multiple databases were searched for studies in which dogs were trained to detect diseases or health conditions. Following PRISMA guidelines, 1689 studies were screened and 50 relevant studies identified. The majority of studies (n = 31, 66%) took place in Europe. Lung cancer (n = 11, 22%), prostate cancer (n = 8, 16%), and breast cancer (n = 7, 14%) were the most frequently-studied conditions. Urine (n = 17, 34%) and breath (n = 9, 18%) were the most common sample types. Across all studies, 166 unique detection dogs were tested. The most numerous breed was Labrador Retrievers (n = 24, 14.46%). The median number of dogs per study was 2 (range: 1–20). To analyze experimental design and results, studies including multiple test paradigms were divided into sub-studies (n = 90). In 84.4% of sub-studies (n = 76), dogs were presented with sets of samples and 74.4% (n = 67) reported a constant number of samples per trial. The median number of samples per trial was 7 (range: 2–100). Of the sub-studies reporting a fixed number of positive samples (range: 1–10; n = 55), 87% (n = 48) presented one positive sample per trial. A plurality of sub-studies (n = 44, 49%) presented samples in a lineup. Sensitivity (median: 0.90; range: 0.13 to 1.0; n = 77) and specificity (median: 0.96; range: 0.08 to 1.0; n = 69) were the predominant measures of detection success, although reporting strategies were inconsistent. Dogs appear to have the capacity to detect disease via olfaction; yet the nascent nature of this discipline yields little consistency across studies.