The Bladder Cancer Study
Experimental design
Despite individual claims that dogs have drawn attention to cancers, to date there has been no scientific study, validated by peer review, establishing that dogs are capable of recognising a distinctive smell (odour signature) for cancer.
Our primary objective was, therefore, to conduct a rigorous experiment to address the question, ‘Can dogs be trained to detect individuals with bladder cancer on the basis of urine odour more successfully than would be expected by chance alone’.
We chose to test the dogs against bladder cancer because we reasoned that urine, which is easy to collect, store and present to the dogs, would act as a reservoir for any distinctive smelling compounds that may be produced by the abnormal metabolism of bladder cancer cells. The challenge was to train the dogs to recognise an, as yet unknown, odour signature for cancer from amongst the many smells present in urine.
We trained six dogs of various breeds, ages and of both sexes over a period of seven months to discriminate between urine from bladder cancer patients and urine from both healthy people and patients with diseases other than cancer. Recognition of the bladder cancer scent was achieved by rewarding the dogs for correctly selecting the cancer urine from non-cancer controls, a process known as positive motivational training. The dogs were taught at the outset to signal their choice by lying down next to the sample.
 |
 |
 |
|
'Tangle' during training
One of our cancer detection dogs being taught to recognise the smell of bladder cancer within urine.
|
|
 |
|
Initially, the controls were water or diluted urine from healthy individuals, but progressively, the dogs were presented with discrimination tasks of increasing complexity, until they learned to preferentially choose urine from bladder cancer patients over that from patients with a variety of other diseases, containing blood or other factors that might be expected to confuse them.
We assessed the dogs’ abilities in a blinded trial, in which they were presented with a panel of samples consisting of one bladder cancer urine placed among six control samples. Each dog repeated this process nine times, and, on each occasion, they were faced with a new set of seven samples which they had not encountered before, with the cancer sample positioned randomly throughout. All dogs had the same nine test sets, but the samples were positioned differently each time.
The control urine specimens set against each bladder cancer sample were carefully chosen so that we could demonstrate unambiguously that the dogs had learnt to recognise a characteristic odour signature for bladder cancer. Hence, all samples within each test set were from people of the same sex, at least two controls were of a similar age to the patient with bladder cancer, half of the control samples came from patients with non-cancerous bladder problems, and the abnormalities present in the cancer urine sample (blood, protein, white blood cells etc.), were also present in at least some of the controls.
Blinding was achieved by coding the samples in two stages so that neither the dog trainers nor the research scientists conducting the assessments knew the position of the cancer urine in the line-up of seven samples. This ensured an objective interpretation of the dogs’ signalling and prevented human behaviour from influencing the dogs’ choice. The double blinding codes were not broken until all the testing was complete.
Results
Taken together as a group, the dogs were presented with a total of 54 test runs (each consisting of one cancer urine sample and six non-cancer samples). They selected the bladder cancer urine sample correctly on 22 out of the 54 occasions, equating to an accuracy of 41%, compared to the 14% (1in 7) which we would have expected if the dogs had chosen a sample at random each time. Statistical analysis confirmed that the dogs’ success rate when compared to chance alone was statistically significant. Further analysis confirmed that the dogs were recognising bladder cancer and not just the abnormalities present in the urine.
Concluding remarks
Our study provides the first piece of experimental evidence that dogs do have the ability to detect a distinctive odour in urine associated with bladder cancer. At this stage, the exact nature of the odour signature is unknown.
The combination of a successful protocol for training the dogs and stringent controls in the testing phase has produced an experimental model suitable for extending our current work. Our future aims are to optimise this model further, study the potential of dogs to detect other types of cancers, particularly skin cancer, and investigate the nature of the odiferous compounds associated with cancer.
|
|
Related Links
See the encouraging results of our latest study published in the BJM:
» BMJ.com
Find out more about the multi-disciplinary team and their work:
» cancerdogs.org
|
