Dear Rakesh,
The 'Behind the Headlines' service provided by the NHS Knowledge Service has recently published this critique on the topic:

Dogs that can 'sniff out bowel cancer'

Tuesday February 1 2011

The study only looked at 12 people with early stage cancer

“Dogs can be trained to sniff out bowel cancer, even when the disease is in its early stages,” reported The Guardian. It said researchers have claimed that a specially trained labrador was nearly as good as conventional tests at identifying cancer from sniffing the breath or stool samples of patients.

This study investigated whether a trained dog could differentiate between stool and breath samples from people with and without bowel (colorectal) cancer. In tests, the dog correctly identified cancer in 33 out of 36 tests of breath samples and 37 out of 38 tests of stool samples.

The researchers point out that it is unlikely to be practical to train dogs to do this work. Furthermore, the biggest limitation of this study was its size. It was too small to say whether canine detection is any better or worse than current techniques used to screen for bowel cancer. In particular, there were only 12 people who had early stage bowel cancer so it is not possible to assess how well this method could detect bowel cancer compared to current techniques. However, this research should be followed up, to assess whether chemicals in the breath can be used to screen for cancer.

 

Where did the story come from?

The study was carried out by researchers from The Dental College Hospital at Fukuoka and Fukuoka University in Japan. The research was funded by Fukuoka Dental College. The study was published in the peer-reviewed medical journal Gut.

This research was generally covered accurately by the newspapers, which highlighted that dogs are unlikely to be used to screen for cancer. The focus on successful early diagnosis may be misplaced as the study only looked at 12 people with early stage cancer. Also, the small number of people tested using this method may not be representative of the general population, meaning that the sensitivity and specificity of this method cannot be compared with that of the current screening test of faecal occult blood testing.

 

What kind of research was this?

This laboratory study investigated whether a trained dog would be able to differentiate between stool and breath samples from people with and without bowel cancer. The researchers were interested in whether this was possible due to anecdotal reports that dogs may be able to detect skin cancer. They also cite four other studies that suggested dogs could detect bladder, lung, breast and ovarian cancer.

In this study, the researchers wanted to see how accurate dogs were at detecting colorectal cancer from breath and stool samples. They also wanted to see whether the diagnostic performance of dogs was affected by age, smoking, disease stage, cancer site, inflammation or bleeding in both patients with cancer and control individuals who were cancer free.

 

What did the research involve?

The dog was an eight-year-old female black labrador retriever that had initially been trained for water rescue and had begun training as a cancer dog three years before the start of the study. The training involved presenting the dog with breath samples from a person with cancer and four samples from volunteers without cancer. When the dog correctly identified the cancer sample by sitting in front of the cancer sample, it was rewarded with play with a tennis ball.

The researchers say the dog had detected cancer from breath samples of people with throat, breast, lung, stomach, pancreatic, liver, gall bladder, colorectal, prostate, uterine, ovarian and bladder cancer. Over the course of the training period the dog had been exposed to breath samples collected from several hundred cancer patients and 500 healthy volunteers recruited through the internet.

This study enrolled people who were over 20 years of age. Thirty-seven people with bowel cancer (diagnosed by colonoscopy) and 148 control participants were recruited. It was not clear how the controls were recruited, whether they had no bowel-related problems or whether they were in the clinic awaiting colonscopy.

All of the participants completed a questionnaire about factors that can influence levels of molecules in the breath or cause watery stool samples prior to having their colonoscopy. The questionnaire asked about factors such as age, physical symptoms (e.g. abdominal pain or bloating, blood in their stools, constipation, diarrhoea, body weight loss and abdominal tumour). There were also questions about the particpants history of cancer treatment, present use of anticoagulants and smoking within the previous two weeks.

The participants prepared for the procedure by drinking a balanced electrolyte solution and a chemical called polyethylene glycol (a laxative). During the colonoscopy, a 50ml watery stool sample was collected with a suction tube. The researchers collected 37 samples from people with colorectal cancer and 148 samples from control volunteers.

The researchers collected breath samples from the participants by asking them to exhale into a breath sampling bag. The researchers were not able to collect breath samples from all of the participants and so samples were only collected from 33 people with colorectal cancer and 132 control volunteers.

The dog was tested on whether it could identify cancer from either stool or breath samples. For each experiment, five samples were placed in containers spaced about two feet apart and covered with wire netting to prevent the dog coming into contact with them. One container held the cancer sample and the other four contained a sample from a healthy volunteer. Before going down the line of containers, the dog was exposed to a standard cancer breath sample. The tests were conducted from autumn to spring as the researchers said that the dog’s concentration tended to decrease during the hot summer season.

The researchers also used the standard method of screening for bowel cancer from stool samples: the faecal occult blood test (FOBT).

 

What were the basic results?

The people with colorectal cancer tended to be older than the control group with an average age of between 70 and 71, compared to the controls who were on average between 64 and 65 years of age.

Approximately half of the control group had colorectal polyps and a small proportion (6.1% of those that had given breath samples and 10.5% of those that had given watery stool samples) had bleeding or inflammatory bowel diseases.

The researchers compared the results from the dog’s diagnosis to the diagnostic results of a colonoscopy (looking for tumours inside the bowel using a camera). They found that the dog’s sensitivity (the number of people with cancer that were correctly identified) was 91% accurate for breath samples and 97% for stool samples. The dog’s specificity meanwhile (the number of people without cancer that were correctly identified) was 99% for both the stool and breath samples.

The researchers then compared the dog’s accuracy from the watery stools test with how well the conventional FOBT test performed at identifying people who had been demonstrated to have cancer by colonoscopy.

 

How did the researchers interpret the results?

The researchers say that this study represents ‘the first step in the development of an early detection system using odour materials from patients with colorectal cancer’. They say, ‘it may be difficult to introduce canine scent judgement into clinical practice owing to the expense and time required for the dog trainer and for dog education’. However, they do say that breath chemicals (volatile organic compounds) have been identified as candidate substances for early detection of cancer and that these may be possible to measure using chemical analysis techniques in the future.

 

Conclusion

This small study showed that a trained dog could accurately differentiate between people with bowel cancer and healthy volunteers from stool and breath samples.

There are various practical and methodological limitations to this technique that indicate it may be unfeasible to use dogs to screen for cancer. The researchers point out that it is unlikely to be practical to train dogs to do this work, citing the cost and highlighting that the dog did not concentrate as well in the summer months.

Other limitations of the study include:

• That this was a small study in only in only 37 people with bowel cancer of whom only 12 had early stage. When testing potential screening tools it is important to test sensitivity and specificity on a large number of samples to be sure the results are representative of the wider population. In particular, this study sample was too small to compare the accuracy of colorectal cancer screening using dog scent detection compared with the currently used screening method of faecal occult blood testing.
• On average, the sample of cancer patients was older than the control group. This raises a potential problem as the age of the person may affect the mixture of chemicals found in their stool or breath samples. Further studies will need to address this limitation.

Despite these limitations, this preliminary research warrants follow-up studies as it appears that the dog was able to be trained to detect cancer in this small sample. These studies would need to assess whether there are detectable chemicals in breath or stool samples that could lead to the development of diagnostic tools for bowel cancer.

 

Ash

Dr Ash Paul
Medical Director
NHS Bedfordshire

21 Kimbolton Road

Bedford

MK40 2AW

Tel no: 01234897224

Email: [log in to unmask]

 

---

From: Rakesh Biswas <[log in to unmask]>

To: [log in to unmask]

Sent: Sun, 3 April, 2011 11:53:51

Subject: Re: Clinical Decision Making and Diagnostic Error

Is this an interesting example of system 1 decision making where researchers utilized a certain breed of our 'best-friends' to detect a malignancy with a specificity and sensitivity of 91% ?

In the Feb 2011 issue of European Urology, Cornu et al. report an experiment where after a learning phase and a training period ( ?analogous to learning to drive a car-Neal? system 2?), a Belgian Shepherd dog's ability to discriminate Prostate carcinoma and control urine was tested in a double-blind procedure. Urine was obtained from 66 patients referred to a urologist for elevated prostate-specific
antigen or abnormal digital rectal examination.

All patients underwent prostate biopsy and two groups were considered:
33 patients with cancer and 33 controls presenting negative biopsies.

During each “run,” the dog was asked to signal a cancer urine among six samples containing only one cancer urine and five randomly selected controls. Sensitivity and specificity of the test were assessed.

The dog completed all the runs and correctly designated the cancer samples in 30 of 33 cases. Of the three cases wrongly classified as cancer, one patient was rebiopsied and a PCa was diagnosed. The sensitivity and specificity were both 91%.

This was postulated to be due to recognition ( ? pattern recognition) of VOCs ( volatile organic compounds) and for most humans ( and the market ) this naturally means looking forward to a 'mass spectroscopic'  standardized detection methodology for mass consumption.

For more light Sunday reading on the topic: http://www.europeanurology.com/article/S0302-2838%2810%2900944-9/abstract

http://www.europeanurology.com/article/S0302-2838%2810%2900998-X/fulltext

regards,

rakesh