Can a Dog Detect a Human Illness with its Nose?

The Rumor Detector

Agence Science-Presse (

The very idea that a dog’s exceptional sense of smell can be put to use identifying illnesses occasionally arises. The idea doesn’t easily go away; but there are obstacles, says the Rumor Detector and the McGill Office for Science and Society.

The Rumor’s Origin

There is no doubt that a dog is far superior to humans in terms of its sense of smell: its genetics and its anatomy allow for it to detect more odors in lesser concentrations. The Greek Hippocrates mentioned, 2,400 years ago, a change of odor in “fever patients.”  A Chinese medical text from the same period contains the most ancient known observation of the detection by smell of an illness in a human organ. It took until the end of the 1700s for the first attempts to analyze human breath with a machine constructed for that use.

Many molecules contained in our breath are volatile organic compounds (VOCs). These molecules contain carbon atoms, upon which life as we know it is based. They easily evaporate at room temperature, rendering them detectable in the air.

These airborne VOCs exit us in multiple ways. For example, acetone is generated by bacteria in our bodies. Propanol can be found in rubbing alcohol and personal care products.

Breath analyses “without a canine detector” are frequently used. For example, to sample the bacteria responsible for stomach ulcers, we ingest radioactively marked urea, which is metabolized into carbon dioxide, which can in turn be detected in our breath. Hydrogen, nitric oxide and a few other molecules play important roles in diagnoses. And of course, there’s the breathalyzer: your breath can betray how much you’ve had to drink.

Dog Facts

In the medical literature, one of the very oldest observations of dog sniffing out an illness came in the form of a short letter published in The Lancet on April 1st, 1989. And this was no April Fool’s joke. A 44-year-old woman had discovered that she was suffering from skin cancer because her dog would sniff the lesion, and even try to bite it when its master wore shorts.

Researchers then started to train and test small groups of dogs with the goal of detecting specific cancers in urine, sweat and breath samples. It works like this: the dog goes into a room and sniffs several containers. To attract its attention towards the container containing cancerous cells, a bit of dog food is added to it. When the dog sniffs the right container, it eats the dog food as a reward and leaves the room. This is repeated several times. Eventually the food is withdrawn, and we are left with a dog that can theoretically detect a type of cancer in a type of sample.   

The First Issue: Is a Dog Reliable?

The first problem is that the accuracy level of dogs varies a whole lot. Several have noted variations between breeds, and even from one dog to another. That even goes for dogs trained the same way in the same institution.

There’s also a concentration problem. A beagle trained to weed out C. difficile infections in hospitals was easily distracted, we learn by reading a scientific article. A visit to a pediatric wing was useless. The children were too excited by the presence of a dog.

And diabetes is another problem. Some dogs are trained to warn their master if their blood sugar level gets too low or too high. Except that exactly what those dogs are detecting isn’t too clear: it could be elevated levels of isoprene in the breath, or even muscle spasms. Though the presence of these dogs can increase the quality of life and lessen the burden of a diabetic, their reliability in signaling hypoglycemia is debatable. In a study of 14 of these dogs, the performance of only three of them was greater than random chance. The dogs’ owners may not agree, but a study casts doubt on their testimonies: while the aptitude of eight dogs to detect hypoglycemia was judged to be mediocre – an average of 14.5 false alerts per week – the owners gave researchers a “way too optimistic” evaluation of their dogs’ performance.  

The explanation rests in part in what constitutes an “alert” on the part of a dog. It may bring over a medical kit, but it could also bark, lick, jump, growl, scratch… In short, it could act like a dog. That leaves a lot of room for interpretation.

The Second Issue: A Dog is Not a Machine

The question of alerts goes back to the basic difference between a dog and a machine: you can’t “program” a dog with standardized instructions. “Animals are very hard to ‘manufacture’ in large quantities, to train, standardize and regulate,” explains Dr. Madhukar Pai, Director of the McGill International TB Centre. He has written, among many topics, on the laboratory uses of a giant African pouched rat, the Cricetomys of the savannah, in sniffing out tuberculosis in people’s spit.  

Training animals is a long and costly process. It’s not easy to adapt dogs on a large scale, and a certain maintenance of the learning process is necessary so that the dog doesn’t lose its precious ability. And the dog will also have its bad days. A dog with a respiratory infection can’t do this task. It could, moreover, transmit bacteria to humans. And what happens if a dog bites someone while trying to sniff out C. difficile?  

What dogs and other animals do prove is that some diseases can be identified by an odor. It has become possible to identify these molecules and to construct machines capable of detecting them.

This article is an adaptation of a piece by Jonathan Jarry on the McGill Office for Science and Society website

Link to the original article

This article is part of the Rumor Detector series; click here  for more articles in the series.

Be the first to comment

Leave a Reply

Your email address will not be published.