Dogs outperform blood tests for cancer

Cancer screening is the domain here — and the claim getting attention is unusually bold. A startup called Dognosis says trained dogs, smelling volatile compounds in breath samples(ascopubs.org)tests. The gap is obvious: early cancer is hard to catch, especially across many tumor types at once. What changed is that Dognosis presented a 1,000-sample mul(ascopubs.org)rs strong enough to force a real look. (ascopubs.org) ### What exactly did the dogs do? They d(jamanetwork.com)h samples collected under standardized conditions from 1,000 people across three sites in Hubli, India — 105 with cancer and 895 healthy controls. Trained biomedical detection dogs then analyzed those samples, and the team combined the dogs’ responses in a consensus framework instead of trusting one animal alone. (ascopubs.org) ### What were the numbers? The headline figure was 96% sensitivity and 100% specificity in the test set across multiple cancers, including oral, breast, (ascopubs.org)are striking numbers because early-stage disease is the hard version of the problem — there is usually less tumor material in blood and a weaker chemical signature overall. (ascopubs.org) ### Why would breath work at all? Cancer changes metabolism, and metabolism changes the mix of volatile organic compounds — VOCs — that leave the body in breath. Dogs are basically(ascopubs.org)is not a brand-new idea; reviews of the field have been arguing for years that trained dogs can detect cancer-linked scent signatures, but reproducibility and standardization have been the bottleneck. (ascopubs.org) ### So did dogs really beat blood tests? In a narrow sense, yes — at least versus some current blood-screening benchmarks pe(ascopubs.org)d test for every cancer.” It is that this breath-plus-dog system reported higher sensitivity than many blood-based screening approaches have shown, especially for early disease. For context, one large 2025 colorectal blood-test study showed 79.2% sensitivity, and multicancer blood tests discussed in oncology circles still show wide variation by stage and cancer type. (jamanetwork.com) ### What’s th(ascopubs.org)n screening trial. That matters a lot. Case-control designs can make a test look cleaner than it will in ordinary screening, where cancer is rare and false positives become a huge practical problem. That is exactly why blood-based multicancer screening still faces demands for randomized trials showing actual outcome benefits, not just good-looking accuracy metrics. (ascopubs.org) ### Why use dogs if machines exist? Because dogs may be the discovery engine, not the final product. Dognosis recorded motion, (jamanetwork.com) training machine-learning models on those signals. The idea is to use canine olfaction to help identify the chemical patterns that matter, then turn that into a scalable diagnostic system. (ascopubs.org) ### What would have to happen next? Independent validation. Bigger prospective studies. Direct head-to-head comparisons with blood tests and standard screening tools. And proof that the metho(ascopubs.org)out that, this is promising science — not a screening program. (ascopubs.org) ### Bottom line? The interesting part is not “dogs replace labs.” It is that dogs may be revealing a cancer signal in breath that labs have not fully captured yet — and that could end up improving the next generation of both breath tests and blood tests. (ascopubs.org)