Can dogs ‘smell cancer’?

Posted on December 19, 2011


Dateline: 16 November 2011 // Posted by: Carlos Perez
Source: Daily Mail

Story: “The dogs that can detect cancer: Meet the four-legged ‘bio-detectives’ who are pioneering a health revolution”

Summary: Dogs’ most powerful sense, their sense of smell, is commonly used in drug detection, human rescues, and so on. Now, a group of scientists claim that dogs can smell cancer. According to some trials, it is claimed that they can be 71% accurate in the task.

The evidence is partly anecdotal (such the case of an old woman, whose dog started to sniff her breath and nudge her right breast where the cancer was), and partly empirical (such as the results of experiments conducted by some German researchers).

According to researchers, “It’s believed that cancers produce volatile chemicals that dogs can be trained to smell”. Now dogs are being trained to become cancer detectors using a specialized training programme.

A dog trainer tells that he has long suspected that dogs could do such things, even in early stages of cancer. People involved in that say that there are lots of anecdotal evidence, and lots of cases to suspect of such ability. In 2002 Dr Guest and Dr John Church tried to prove that this was just a coincidence, but the y found a rate of 56% accuracy.

A small demonstration is conducted to show dogs skill. This demonstration consists in detecting among 12 urine samples, the one of a patient with cancer. The dog selects the right one.

Why is this PseudoNews? Dogs are known to have a highly developed sense of smell. However, this does not mean that a dog can smell substances that are linked to cancer. People’s bodies contain a huge number of different smells. It would be very difficult to identify a cancer-smell in particular (if it does really exist). Other strange thing about the reported cases is how dogs could differentiate if a smell is good or bad. When they are trained to look for drugs, dogs do not know if drugs are good or bad, they just search them looking for a reward.

What features of pseudoscience are on show? In general, the narrative of this article lacks parsimony. There are more simple explanations for each case. For one thing, the cases used in the article appear to be of weak reliability. The researchers themselves admit that most of the evidence is quite anecdotal. They explain two cases where dogs supposedly detect a cancer. Those two cases can be also explained by confirmation bias. Millions of dogs around the world have obsessive random behaviors. Only those whose behaviours correspond to actual human cancers will come to light in this way.

The second problem is that the measurement is very vague, if there are measures at all. Quantification is needed, not just the percentage about how many times the dog choses a correct option.

The third problem is the nature of the evidence as discussed. If it is intended to demonstrate that dogs can smell cancer, then researchers will have to organise proper controlled trials designed in ways that attempt to falsify the proposition. Such studies should also be oriented to understand the phenomenon, and not just to show that “something happens”.

For example, if we analyze the description of the demonstration conduct in the training centre, it appears as though the study was not properly blinded: the dog goes to smell the samples with a person that knows what sample is the right one. That reminds me to the case of “Clever Hans”, a German horse that by looking to his owner’s face and hitting the floor with a leg as many times as needed to express a number, ended up convincing the world that it could perform complex arithmetical operations.

The moral of the story: Dogs have certainly better smell sense than humans, but that does not mean that they can smell every molecule in the breath. When more simple explanations can be used to describe why something happens, we have to consider them while studying in a scientific way everything we need to understand it.