Thursday, October 10, 2013

More on the Cancer Sniffers: Dogs vs. Electronic Noses

Dogs have been taught to screen for cancer, alerting to individuals who may have a cancer but whose diagnosis will have to be verified by biopsy and histopathology.  A review article recently appearing in The Netherlands Journal of Medicine (Bijland et al., 2013) finds that electronic noses, at least as to the detection of some cancers, may be better than dogs.  Nevertheless, recent studies on lung and ovarian cancer found dogs uniquely effective, and able to detect smaller concentrations of cancer odors than any other technique.  While most studies continue to suggest that clinical deployment of cancer-smelling canines remains a possibility, researchers continue to emphasize that only more research can make dogs a clinical diagnostic tool.  Also, deploying dogs in medical environments faces hurdles that will not be presented to electronic noses. 

How Cancers Produce Unique Odors

An editorial by three scientists (Leja et al. 2013) regarding using scent as a diagnostic tool explains that volatile organic compounds (VOCs) are released from cancer cells or metabolic processes associated with cancer growth.   “These VOCs are transported with the blood to the alveoli of the lung from where they are exhaled in measurable odorants.”  Thus, cancer has a smell, and “at least in theory, different cancers have different smells.”  The editorial describes an electronic nose as follows:

“The olfactory receptors of the mammalian nose are mimicked by an array of highly sensitive gas sensors.  Like biological receptors, these sensors can each absorb a wide variety of VOCs from the gas phase. The collective sensing signals are statistically analyzed, using pattern recognition algorithms that have previously been trained how to identify a particular smell by controlled exposures in the laboratory. Once established, the electronic patterns of disease allow classifying unknown breath samples from patients or healthy subjects. However, the development of suitable gas sensors for breath testing is technically challenging, because the sensors should be able to detect the delicate smell of diseases such as cancer in the humid atmosphere of exhaled breath albeit strong individual variations of the breath-humidity levels between persons.” (Leja et al.)

Comparing E-Noses and Animals in Disease Identification

Bijland et al. searched various databases for “key studies in scent detection” and found 168 papers, some of which involved electronic noses (e-noses) and other non-canine detection approaches for diseases.  They reviewed studies where scent was used in the detection of lung, ovarian, breast, bladder, colorectal, and melanoma cancers.  This review article compared e-nose studies with studies involving animals, particularly dogs.  As shown in the following table, adapted from the study, in some cancers, such as bladder cancer, e-noses outperformed dogs, but in others, such as breast cancer, dogs were more successful.

Cancer
Type of Nose
Type of Sample
Sensitivity/Specificity
Lung
Dog
Breath
71%/93%
E-nose
Breath
71%/100%
E-nose
Breath
85%/100%
E-nose
Breath
94% success rate
E-nose
Breath
71%/92%
Ovarian
Dog
Tissue and blood
Tissue: 99%/97%
Blood: 100%/98%
E-nose
Tissue
84%/87%
Breast
Dog
Breath
88%/98%
E-nose
Breath
94%/74%
E-nose
Breath
75%/85%
Bladder
Dog
Urine
41% success rate
E-nose
Urine
100%/100%
Colorectal
Dog
Breath and feces
Breath: 91%/96%
Feces: 97%/99%
Melanoma
Dog
Tissue
75-86% success rate
E-nose
Tissue
70%/90%
Tuberculosis
Rats
Sputum
74% accuracy
Rats

68%/87%
E-nose
Sputum
85% accuracy

The tuberculosis study with rats determined that these animals were able to process over 40 times as many sputum samples a day than a lab technician.  Such an ergonomic advantage has not been found in other studies with animals as cancer sniffers, however. 

The review paper noted that dogs have been able to detect certain kinds of intestinal infections.  E-noses have been used to detect diabetes, liver cirrhosis, asthma alone, asthma and COPD, and COPD alone.  E-noses can distinguish different stages of COPD.  This study noted that humans have been able to detect certain diseases with their own noses, though this skill has seldom been the subjected to quantitative study.  Tuberculosis was detected by the ancient Greeks and Chinese by heating the patient’s sputum and smelling the fumes.  As compared to e-noses, the researchers noted the following concerning dogs:

“Dogs … require an average VOC concentration of less than 0.001 part per million.  Enoses on the other hand have a detection threshold of 5 to 0.1 parts per million (ppm), although like animals different types of Enoses have different affinity for different volatiles. In comparison, humans have a detection threshold, on average, ranging from 0 to 80 ppm, again depending on of the type of substance. For example, ammonia cannot be perceived by humans until it reaches 50 ppm. Taken together, many animals smell up to 100 times better than humans and Enoses, and it may well be worth making appropriate use of this superior technology.”

Thus, dogs may be able to detect smaller amounts of cancer-produced chemicals than current e-noses.  The researchers suggested that dogs might be particularly useful for colorectal cancer where other methods, where blood work is uncertain and colonoscopies are much more invasive.  They conclude that “scent detection holds promise for the future and should receive higher priority in terms of research effort and funding.”  

Chemical Means of Analyzing VOCs

The VOCs in the breath can also be analyzed by chemical techniques, but since they appear in the breath in such low concentrations, they must be enriched before analysis.  Buszewski et al. (2012a) note that the “most common method of enrichment of VOCs is solid-phase microextraction (SPME) and sorption on solid sorbents followed by thermal desorption (TD),” after which the enriched VOCs can be subsequently analyzed by gas chromatography (GC) or GC-mass spectrometry (GC-MS).  There are variations on these techniques and other chemical procedures under development. 

This research team from Poland and Austria described the connections between the dog’s nose and its brain, beginning with the mucus lining of the nasal cavity where molecules are bound to odorant-binding proteins.  The axons of olfactory cells reach the olfactory bulb and converge in structures called glomeruli.  In the inner layer of the olfactory bulb, mitral cells form glomeruli with axons of olfactory receptor cells and send signals to the olfactory cortex.  Each individual odor produces a specific spatial map of excitation and, through spatial encoding, the brain distinguishes specific odors.  German shepherds have more than 200 million olfactory cells on an area of about 170 square centimeters, whereas humans have about 5 million cells on about 5 square centimeters of olfactory epithelium.  The proportion of active to inactive genes of the olfactory receptor proteins also enters into the calculation of how much more powerful a dog’s sense of smell is than that of a human.  The team summarizes prior research on canine cancer detection as follows:

“The papers published so far demonstrate that dogs, after appropriate training, are able to discriminate breath, urine, feces or tumor-tissue samples of patients with lung, breast, prostate and ovarian cancers from respective samples taken from healthy humans with sensitivity (the true positive) and specificity (the true negativity) exceeding 80%. The canine indications are easily interpretable in terms of calculating the detection sensitivity and specificity, but no information can be obtained on what chemical compounds dogs are responding to or the quantity of those compounds.”

They also argue that analysis “of odor samples by GC-MS carried out simultaneously with tests using trained dogs” may detect which VOCs are the markers of cancer that dogs respond to.  The team has, in fact, begun to work on this idea (Buszewski et al. 2012b).

Recent Lung and Ovarian Cancer Studies

A recent study (Amundsen et al. 2013) using dogs to detect lung cancer in patients who were suspected of having that cancer, but who had not yet undergone bronchoscopy, found that with “99% sensitivity, the dogs were able to distinguish cancer patients from healthy individuals.”  The authors concluded that “the main challenge is to determine whether the test can sufficiently discriminate between patients at risk, patients with benign disease, and patients with malignant disease.” 

Another research group (Horvath et al. 2013) has been studying the ability of dogs to recognize ovarian cancer in the blood of patients with the disease. Their research has indicated that dogs trained to recognize the odor of ovarian cancer did not recognize odors from other gynecological malignancies.  They argue:

“The fact that the dogs could not recognize cancers other than ovarian cancer strongly suggests that different cancers have different characteristic smells, thus enabling both diagnosis and differential diagnosis. Moreover, the characteristic odor of ovarian carcinoma is likely organ-specific.”

This conclusion appears at odds with research involving a cancer sniffing dog in Japan discussed in a blog here several years ago.  In a paper just published in BMC Cancer, this team sought to analyze how surgery and chemotherapy affected the ability of dogs to detect ovarian cancer in the blood of patients.  They found that dogs were almost flawless when it came to recognizing patients with full-blown ovarian cancer, but also made very few mistakes when asked to detect samples of patients that had received five or six chemotherapy treatments and who had significantly lower cancer antigen levels.  They found that one dog, named Hanna, “was repeatedly able to identify with certainty a piece of fatty abdominal wall containing about 20 microscopically-verified ovarian cancer cells. It is impressive how this very low limit of detection allows dogs to signal probable future recurrences that would not be diagnosed by other methods for another 2–3 years. This is the most important result of the present study.”  They conclude that:

“Detection of odor in the blood, currently only possible with trained dogs, can allow for early and long-term prediction of survival. An early diagnosis of primary or recurrent disease may also significantly improve the patient’s survival.”

Conclusion

The recent research and analysis indicates that e-noses may become a valid methodology for detecting some cancers, where samples can be sufficiently large. It is difficult to imagine dogs wandering the corridors of hospitals smelling the rear ends of patients even if this were proven to be a useful diagnostic tool, so clinical deployment would likely involve the creation of separate testing facilities.  Samples would have to be carried to the such facilities, and specialized dog handlers and technicians would have to be employed.  This might be practical for a large centralized medical provider treating a great number of patients, or for a specialized facility serving an array of hospitals and medical centers in a large, perhaps multistate area.  Still, this might for a time remain cheaper than buying electronic noses or developing laboratories with sophisticated extraction equipment.  Almost all papers published in this area emphasize the need for further research along all these lines, but the number of scientists focusing on these approaches is increasing, and we can expect advances to continue.  

Thanks to Tadeusz Jezierski for directing me to some of the research discussed here.  Thanks to Kingsbury Parker, particularly for noting that "the dog's ability to detect .001 ppm is truly amazing.  Since I have done similar quantitative analysis in the lab at the accuracy of 1 ppm I know what sort of effort is required." Thanks to L.E. Papet for additional sources and, as he often does, restraining my more fanciful arguments.  Thanks to Richard Hawkins, also as usual, for catching mistakes.    

Sources:

Altomore, D.F., Di Lena, M., Porcelli, F., Trizio, L., Travaglio, E., Tutino, M. Dragonieri, S., Memeo, V., and d Gennaro, G. (2013). Exhaled Volatile Organic Compounds Identify Patients with Colorectal Cancer.  British Journal of Surgery, 100(1), 144-150. 

Amundsen, T., Sundstrom, S., Buvik, T., Gederaas, O.A., and Haaverrstad, R. (2013). Can Dogs Smell Lung Cancer? First Study Using Exhaled Breath and Urine Screening in Unselected Patients with Suspected Lung Cancer.  Acta Oncologica (posted online ahead of publication August 19, 2013).

Bijland, L.R., Bomers, M.K., and Smulders, Y.M. (July/August 2013). Smelling the Diagnosis: A Review on the Use of Scent in Diagnosing Disease.  The Netherlands Journal of Medicine, 71(6), 300-307.

Buszewski, B., Rudnicka, J., Ligor, T., Walczak, M., Jezierski, T., and Amann, A. (2012a). Analytical and Unconvential Methods of Cancer Detection Using Odor.  Trends in Analytical Chemistry, 38, 1-12.

Buszewski, B., Ligor, T., Jezierski, T., Wenda-Piesik, A., Walczak, M., and Rudnicka, J. (2012b).  Identification of Volatile Lung Cancer Markers by Gas Chromatography-Mass Spectrometry: Comparison with Discrimination by Canines.  Analytical and Bioanalytical Chemistry, 404(1),141-146.

Horvath, G., Andersson, H., and Nemes, S.(2013) Cancer Odor in the Blood of Ovarian Cancer Patients: A Retrospective Study of Detection by Dogs During Treatment, 3 and 6 Months Afterward. BMC Cancer, 13, 396.

Leja, M., Liu, H., and Haick, H. (2013). Breath Testing: The Future for Digestive Cancer Detection.  Expert Review of Gastroenterology and Hepatology, 7(5), 389-391. 

1 comment:

  1. Research about dogs and disease detection seems to have reached the point at which calculations concerning the return on investment should be made. Nationally there would need to be several thousands of dogs in kennels located in the countryside
    probably no more than 100 dogs per location. The cost has to be less than $100 per test depending on the expected accuracy and the dogs need to make at least 25 tests per day in order to be competitive with e-noses. Naturally, all costs of training, exercise, maintenance and human supervision must be included in these calculations as ultimately the ROI will determine how all this moves into the future.

    ReplyDelete