• Mandy

Dogs detecting cancer in humans (from Thesis)

Updated: May 17

The following is one section from the thesis I did for my diploma. This topic reflects my interest in the incredible work Medical Assistance Dogs are doing. I close with a summary.

Picture from:

Can dogs detect cancer in humans

The internet provides anecdotal evidence of dogs helping people to detect their cancers and I have spoken to someone who says her dog alerted her to a lump which turned out to be breast cancer. One story in the news (StaffordshireLive May 2018) was of Thelly Price whose dog, a west highland terrier called Daisy-May, kept sniffing around her throat and if she couldn’t reach Thelly’s neck she would jump up. Thelly was feeling her neck to see if anything was there and then one day felt a lump, she went to the doctors and tests showed she had non-Hodgkin’s lymphoma. Thelly had an operation to remove the cancer and the prognosis was good, far better than if the cancer had been found later. There are many more stories like this one.

Dogs certainly can be trained to help in the detection of some medical issues, Medical Alert Assistance Dogs (undated) provide dogs to help people with the following:

· Type 1 diabetes

· Postural Orthostatic Tachycardia Syndrome (PoTS)

· Addisons disease

· Severe allergies

· Some other Endocrine disorders

· Episodes of sudden health deterioration

They assist by detecting the odour changes that occur when potentially life threatening events are about to occur. For endocrine conditions like diabetes and Addisons disease the dogs are trained to detect the minute odour changes in blood sugar levels and other hormone-related changes. When changes are detected they can warn the owner, get help and fetch medical supplies. With PoTS, a cardiac condition, the dogs detect the odour changes in the owner and can warn them to get to a safe place before they potentially fall unconscious. With Nut allergies the dogs can be trained to detect the air-borne allergens and alert the owner.

Most of the Medical Detection dogs are working with owners who have diabetes.

Research done by Medical Detection Dogs in 2014 showed dogs were capable of detecting tiny traces of odour, roughly one part per trillion, equal to a teaspoon of sugar in two Olympic sized swimming pools.

The Medical Detection Dogs organisation are currently working on the following areas:

Cancer: two NHS approved clinical trials are underway. One working with the Milton Keynes University Hospital NHS Foundation Trust on detecting volatile organic compounds (VOCs), which are associated with disease growth, for Prostate cancer. Prostate cancer is the third biggest cancer killer in the UK and current testing can give false positive results. Another trial is working with Hull and East Yorkshire Hospitals NHS Trust on colorectal cancer using urine samples. Colorectal cancer is the 4th most common cancer in the UK and the current screening process is highly invasive and as such almost half of people offered screening do not take it up.

Providing non-invasive, rapid and accurate testing will allow these cancers to be detected earlier and improve survival rates.

Neurological diseases: work is focused on Parkinson’s disease which affects 1 in every 500 people in the UK. There is no way to detect Parkinson’s and by the time symptoms show and it is diagnosed the disease is already well advanced. Being able to detect the disease earlier would allow for treatment to start earlier which would slow the progression of the disease. Some research suggests the disease could start 20 years before symptoms are obvious. Work with Manchester University and Edinburgh University has determined that dogs can detect Parkinson’s years before symptoms will show.

Bacterial infections: 700,000 people die around the world annually from drug-resistant infections. Working with the Imperial College London dogs are being trained to detect specific bacteria. Dogs trained to detect the ‘super-bugs’ could be used in hospitals to find dangerous bacterial infections like MRSA.

Malaria: People with malaria attract mosquitoes through odour, dogs can be trained to also detect this odour and thus identify people who are asymptomatic and carrying the parasite.

COVID-19: Dogs have been trained to detect the odour of COVID-19, they can screen up to 250 people in 1 hour and tests have shown they correctly identify the odour 94% of the time, showing more accurate detection than Lateral Flow Tests.

Medical Detection Dogs (2018) states that around 30% of a dogs brain is dedicated to analysing odours, percentage-wise an estimated 40 times greater than the area of a human brain devoted to analysis of odour. A human has up to 5 million olfactory receptors, a dog has up to 300 million. Dogs can detect odours direct from source or as a residual odour left in an area, such as on an item of clothing. Odour molecules are chemicals that can be dissolved in water, when small enough they are volatile and vaporise, they then reach the nose and are dissolved in mucus.

When the normal human metabolic functions are altered by a disease this can result in the production of Volatile Organic Compounds (VOCs) which pass into the bloodstream and get excreted through breath or urine, dogs can detect VOCs at a tiny concentration of 0.001 parts per million. The difficulty in the training is teaching the dog to ignore the huge number of other odours it will detect.

Guest et al (2021) performed a study looking at the detection of prostate cancer (the 2nd leading cause of cancer death in men in the developed world) where a more sensitive and accurate detection strategy is needed for population screening, looking at the feasibility of integrating canine olfaction with chemical and microbial profiling and artificial intelligence, a cross discipline approach.

The study acknowledged that dogs could be trained to detect the cancer through smell with high levels of accuracy but the level of training required meant it was not a scalable solution, and therefore a machine equivalent, a ‘machine nose’, with artificial intelligence based on the canine capabilities should be looked at. For the study 12 urine samples were used from men with prostate cancer and 38 negative samples, two Medical Detection trained dogs were used. Out of the samples some were used for training leaving 7 positive and 21 negative for testing, both dogs accurately identified 5 out of the 7 positive samples, and for the negatives one of the dogs correctly identified 16 and one dog identified 14. The results gave 71% sensitivity (identifying positive samples) and 70-76% specificity (identifying negative samples). The training given was limited, and the study group believed that with further training the results for sensitivity/specificity could be improved to over 99%.

The volatile compounds from the samples were analyzed using gas chromatography – mass spectrometry (GC-MS) which found some volatiles elevated or reduced. The team used the dogs diagnoses to train an artificial intelligence: an artificial neural network (ANN) to evaluate the volatile chemicals found by the GC-MS and the ANN detected the cancer with high accuracy. The next step would be to apply the canine-trained machine algorithm to an electronic nose that contains synthetic versions of the dogs olfactory receptors. This study was based on a small set of samples and the patients with cancer used were very advanced with the disease, however the results showed that combining approaches from the different areas of science could lead to an overall improved solution for more accurate and scalable population screening. Key to the success of this approach are the trained Medical Detection dogs.

Worthington (2021), on behalf of the Prostate Cancer Foundation reported on the use of the Medical Detection dogs detecting prostate cancer and explained how they were looked after. The dogs live in normal homes, not kennels, and go to work in the morning. Following a walk and breakfast they go to the training area and work for about half an hour. After a break they will do another working session, typically they do 3 sessions of training per day. When the dogs are successful they get rewarded with treats and praise, they are successful for correctly identifying a sample with or without cancer, as the dogs are rewarded when they don’t find cancer a false-positive is less likely as the dogs aren’t feeling pressure to find ‘something’. The dogs chosen for training are dogs that enjoy the scent work, often hunting or working dogs like Labradors and Spaniels. Some of the dogs used are rescue dogs taken out of shelters. The chosen dogs have to love their work and never tire of wanting to get to the next sample they can sniff.

There are numerous other studies reporting success in dogs detecting cancer such as the following:

The University of Pennsylvania ‘PennVet’ (2018) reported on training dogs to detect malignant ovarian cancer using plasma samples. As the dogs identify that the cancer has an odour signature the team can look to isolate and identify the signature components of the odour. The ultimate aim being to create an electronic system, an electronic nose of nanosensors to mimic the canine nose. Ovarian cancer is hard to detect at an early stage and difficult to treat later on, PennVet state that detecting it earlier can increase the 5 year survival rate from 17% to 90%.

In 2012 McCulloch et al reported on dogs identifying lung cancer. The study involving 60 patients with lung cancer and 110 healthy patients resulted in the trained dogs correctly identifying the cancer patients with a sensitivity of 90% and those without cancer with specificity of 72%. They reported that the 5 year survival rate could be increased from <5% to >70% with earlier detection. The dogs were detecting the cancer from breath samples.

Kure et al (2021) performed a study using a trained Labrador Retriever to identify patients with breast cancer from urine samples. 200 samples were used, 40 coming from patients with breast cancer and the dog correctly identified all 40 breast cancer patients, sensitivity and specificity were both 100%.

Jendrny et al (2021) gathered details from a number of studies, they concluded that dogs could be better at detecting infectious diseases than non-infectious ones like cancer where results have been quite varied, which is highly influenced by the level of training the dog has had. The dogs performance can also be impacted by how it is kept, dogs in good physical condition do better, and dogs with lower stress levels enjoying social contact and an enriched environment had enhanced cognitive performance. Other items which could affect the dogs performance include hydration, food, general health and medications. Standards for training medical detection dogs are needed, such as exists for explosive detection dogs.

There is ample evidence that dogs can be trained to detect a number of different cancers, however the training has to be sufficiently done to give a high level of sensitivity and specificity and the number of dogs which can be made available is small compared to the millions of samples that need testing to successfully screen a population. The ability to scale up is leading to the development of electronic noses which can replicate what the dog does.


Dogs have been shown to have an amazing capacity for sniffing out Cancers, offering promise for a future where some cancers can be detected earlier offering much improved survival rates for the patients, and in some cases replacing invasive screening options which are not widely available or widely taken up. In some parts of the world the screening equipment currently in use is not available as the costs can be high and using dogs could offer much needed screening in these areas, even if not able to screen large numbers. The training of medical detection dogs needs to be standardised and of a high standard to deliver good levels of detection without lots of false-positives, which does limit their use as a primary medical option currently. I believe it is likely an electronic nose with artificial intelligence can be created, using the dog olfaction capabilities as a base to build on, which would offer a more standardised and scalable solution.

References - see under article on emotional support dogs