Dognosis puts 16 dogs and a helmet of sensors on a 10,000-person cancer trial

Dognosis, a Bengaluru company training dogs to read human breath for cancer, has moved from a published Phase 2 signal to a Phase 3 trial that aims to enrol about 10,000 people across 10 Indian hospitals over 12 months. The trial began in April. It will screen people at higher risk and survivors at risk of recurrence. Anyone the dogs flag is supposed to go on to ordinary imaging and, where needed, biopsy. That follow-up step is the difference between a clever olfactory study and a screening tool.

The Phase 2 paper sat in the Journal of Clinical Oncology. In a test cohort of 1,502 people, drawn from a larger run of 3,275 participants at six Karnataka hospitals, the company reported 90.8 percent sensitivity (95 percent CI 87.2 to 94.5) and 91.3 percent specificity (95 percent CI 89.7 to 92.9), with an AUC of 0.962. Sensitivity for stage I and II disease was 90.6 percent and, the authors said, held across major cancer types. The work covered seven broad groups and more than 20 cancers, including breast.

The method is simple at the patient's end. A person breathes into a cotton mask for about ten minutes. The mask is sealed and taken to a farm lab in Nelamangala, outside Bengaluru. The person never meets the dog. On the other side, 15 or 16 animals, among them beagles, a Labrador, a Labrador-Indie mix and a Dutch Shepherd-Belgian Malinois mix, work platforms with automatic feeders. Chloe, a Labrador mix, wears a 3D-printed helmet and a harness packed with sensors that record brain activity, breathing rate and body language. Co-founder Dayananda Kulgod has called the kit a doggy Iron Man suit. Itamar Bitan, co-founder and product lead, trains the animals.

The scientific claim is that canine olfaction can pick up volatile compounds associated with tumours, and that sensors plus a model can turn a dog's stance into a repeatable score. The company says dogs trained on ten cancer types also flagged an eleventh they had not been taught. That anecdote is not a trial endpoint. It is the sort of observation that makes oncologists ask for Phase 3.

Those oncologists have been cautious in public. Pankaj Chaturvedi, deputy director of the Centre for Cancer Epidemiology at Tata Memorial Centre in Mumbai, said new tools should strengthen existing screening programmes, not replace them. Santosh Kumar Devadas, head of oncology at Ramaiah Memorial Hospital, called the published result a promising early signal that still needs a much larger sample. Nilesh Reddy at Manipal Hospitals said practical benefit for patients remains to be shown.

Dognosis has backing from Accel India and says it is in talks with Punjab's National Health Mission and with officials in Karnataka and Andhra Pradesh about pilots. A commercial at-home kit is pencilled for next year, starting in Bengaluru, if Phase 3 holds. India's screening gap is the market the founders cite: many cancers here are found late, and a cheap breath step that sends the right people to a hospital could, in theory, move diagnosis forward by months.

Theory is the word that still applies. A 1,502-person cohort with a high AUC is not a national programme. Ten thousand people with mandatory conventional follow-up is the first design that can show whether the dogs over-call, under-call, or hold their 90 percent when the sample is no longer a research hospital list. Until those numbers exist, Chloe's helmet is a sensor package on a farm, not a test a family should trust on its own.