Canine Olfaction Combined With Bayesian Modeling for Multicancer Detection From Breath Samples: A Phase II Study in India

Abstract

PURPOSE Low-cost, acceptable, high-sensitivity triage tests are needed to address low cancer prevalence in population screening, particularly in low- and middle-income countries. Breath-based canine olfaction may serve this role, but evidence, to date, has been limited to small, single-cancer studies, largely from high-income settings. We evaluated the analytical validity of a multicancer breath detection system using trained dogs integrated with Bayesian fusion modeling. METHODS We conducted an assessor-masked, multicenter case-control study across six hospitals in Karnataka, India (CTRI/2024/10/075938). A total of 3,275 participants were enrolled (1,773 training; 1,502 testing). The test cohort included 283 treatment-naïve, biopsy-confirmed cancer cases spanning seven major cancer groups and 1,219 controls (healthy volunteers, nononcologic chronic disease, or benign biopsy). Breath was collected using cotton surgical masks, stored under –20°C cold-chain conditions, and evaluated by trained detection dogs. Individual dog indications were integrated using a Bayesian fusion framework incorporating historical dog performance and participant-level covariates. RESULTS The fusion system achieved 90.8% sensitivity (95% CI, 87.2 to 94.5) and 91.3% specificity (95% CI, 89.7 to 92.9), with a receiver operating characteristic AUC of 0.962 (95% CI, 0.952 to 0.969). The sensitivity for early-stage disease (stage I to II) was 90.6% and remained consistent across major cancer types. CONCLUSION In a 1,502-participant test cohort, canine olfaction combined with Bayesian fusion demonstrated high analytical accuracy for multicancer detection from breath. These findings establish analytical validity and support prospective evaluation in true screening populations.