A review of the efficacy of a low-cost cancer screening test using cancer sniffing canines

Cancer death rates are significantly higher in socioeconomic disadvantaged areas, which are disproportionately affected by health inequalities and reduced access to healthcare. One of the ways through which better health outcomes can be achieved is through early detection. The Emory University Urban Health Initiative team has been researching the use of inexpensive cancer detection in order to increase detection access both before and after surgery for low-resourced patients. Results from many studies conducted over the past two decades have found extensive evidence that the powerful olfactory system of canines allows them to detect certain odors, in the form of volatile organic compounds (VOCs) and exhaled breath condensate, that are biomarkers of a variety of cancers. In fact, canines have approximately 50 times more olfactory receptors than humans do, in addition to an area of the brain dedicated to scent detection that is 40 times larger than that in humans. This article describes the findings of a literature review that determines cancer sniffing dogs' efficacy in detecting cancer, both generally and for specific carcinomas. Its goal is to determine whether cancer specialists are supported in using breath and other non-invasive samples, tested by dogs, to determine the presence of cancer and the absence of cancer after surgery and other treatments. In this summary of our research, we look at past case studies and experiments to identify the types of canine-detectable cancers, identify the benefits and limits of canine detection, as well as review various canine training and testing methodologies.

There are a few major types of cancer for which the use of the cancer dog test has proved to be remarkably effective. For instance, colorectal cancer, one of the most common in the United States, is found in the lower portion of the digestive tract. Colorectal cancer can be detected by two methods: via watery stool samples and breath tests [1]. Results from a 2010 study conducted at Kyushu University at Fukuoka in Japan indicated that the sensitivity and specificity of scent detection are very high, both found to be above 90% for stool and breath analysis [1]. The canines even detected early stage cancer with surprising accuracy. Another study from Hallym University in Korea found that canines can not only indicate the presence of colorectal cancer, but they can also discriminate between colorectal cancer and other cancer odors, such as breast cancer, with sensitivity and specificity measurements of over 90% [2]. This research concludes that specific chemical compounds indicative of colorectal cancer are indeed present in the body and that canine scent detection of these compounds is an effective means of colorectal cancer diagnosis.

Ovarian cancer, another deadly cancer with a mortality rate of over 50%, often goes undetected for long periods of time, until it reaches a stage where it is difficult or even impossible to treat. With trained dogs, it can usually be detected with a sample of tumor tissue, but it can also be detected via blood samples [3]. In fact, a study conducted by Sahlgrenska University in Sweden found that while sensitivity and specificity for both methods of detection are greater than 95%, canines' accuracy of detecting ovarian cancer in blood samples was found as high as 98%. In addition, it has been established that trained dogs can also distinguish between ovarian cancer and other female reproductive cancers, such as endometrial, cervical, and vulvar carcinomas [3]. Another study from the same institution, which tested canine scent detection of only ovarian tumor tissue samples, yielded similar results, with a sensitivity of 100% and specificity of 97.5% [4]. These findings further emphasize the efficacy of canine scent detection, even more so for ovarian cancer in particular.

Melanoma, a dangerous form of skin cancer, can also be detected by canines via a tissue sample [5]. Melanoma tissue releases a set of chemicals that dogs can be trained to localize and thereby detect cancer. In one study conducted in Florida, canines were able to detect melanoma with perfect accuracy, in other words, with 100% sensitivity and specificity [5]. While not all research supports the efficacy of scent detection of melanoma to this extent, other clinical trials involving canine screening of melanoma have found specificity to be greater than 80% [6]. Therefore, while canines may not provide perfect detection in all cases of melanoma, they generally tend to make the correct diagnosis in most trials.

Prostate cancer, similar to melanoma, releases certain chemicals that are easily recognizable by dogs. However, instead of being released into skin tissue, these chemicals are released into the urine [7]. Canines can determine the presence of cancer by detecting certain odors in the urine that are unique to those with prostate cancer. A study from Italy measured that this process has a sensitivity of 100% and a specificity of almost 99% [7]. Research conducted in France also confirmed the reliability of scent detection for prostate cancer with sensitivity and specificity values of 91% [8].

Bladder cancer detection is almost analogous to that of prostate cancer. Trained dogs, when instructed to smell urine, detect the presence of odors that confirm the presence of cancer [9]. However, despite the high success rate of prostate cancer detection, canines are inconsistent when it comes to bladder cancer. In fact, two studies conducted in the United Kingdom found specificity for bladder cancer scent detection to be 64% and 41%, respectively [9, 10]. Therefore, this research indicates that canines are not a reliable means of screening bladder cancer, especially when compared to other carcinomas such as colorectal, ovarian, melanoma, and prostate.

Finally, lung cancer and breast cancer, both of which canines can detect through biopsy (tissue analysis), have been detected accurately by dogs in double-blind tests. Canine lung cancer detection is close to perfect, with both specificity and sensitivity values of 99% for lung cancer in a clinical trial from California's Pine Street Foundation [11]. Similarly, a study in Spain found that as many as 98% of dogs provide the correct diagnosis for lung cancer [12]. The Pine Street Foundation also conducted similar research for breast cancer detection. The results for breast cancer were found to be slightly lower, with a sensitivity of 88% and specificity of 98%. Hallym University in Korea, mentioned earlier in this review, also measured the same results for breast cancer, with specificity and sensitivity values of 88% and 98%, respectively [2]. This is comparable to the accuracy available with traditional cancer detection screening at a fraction of the cost and time required for testing. In addition, for both cancers, the results are fairly similar with respect to the stage of the cancer being tested [11]. Overall, for almost every one of the major cancers studied, canines have consistently made reliable diagnoses; whether it is by analyzing blood, tissue, or urine, these cancer detecting canines have truly made good use of their incredible olfactory capabilities. Table 1 summarizes the efficacy of canine scent detection for the various cancers discussed in this article.

For the past two decades, numerous studies have clearly shown the promising value of canine cancer detection. However, before canines are used in practice, the benefits of canine detection compared to that of conventional methodology must be first elucidated in order to validate their use in a clinical setting.

The most evident benefit of canine detection is that it provides a cheaper alternative to traditional cancer screening [13]. This means that individuals who are interested in their health or lack healthcare coverage have a more accessible option to assess their wellbeing. By making cancer screening more attainable, individuals are more likely to take these tests on a regular basis, increasing the chance of detecting early tumors and therefore improving their survival rate. Notably, this screening would be particularly beneficial to those working in carcinogenic environments, such as firefighters, who stand the most to gain from regular cancer screening. These enormous benefits are reason enough to pursue this novel field of cancer detection. However, the benefits of canine olfactory detection are multifactorial.

Unlike traditional cancer screening, canine detection is usually a non-invasive process. In fact, a 2017 study in Spain revealed that surgical bandages were viable samples to test for cervical cancer VOCs [12]. These surgical bandages are not only less invasive than the collection of fresh samples, but they can also be stored for longer periods of time, allowing greater flexibility in cancer screening. Another study conducted in 2017 tested whether breath collection could be used as a viable means of samples and found promising preliminary results, noting that CancerDogs Inc. a cancer screening organization, has incorporated masks for breath collection and canine examination [14]. In addition, a 2011 study on colorectal cancer screening found that a Labrador Retriever can identify breath samples with a sensitivity (percent of trials where the canine correctly indicated the cancer sample) of 91% and a specificity (percent of trials where the canine identified the control sample) of 99% [1]. Breath collection was shown to be an effective and non-invasive method for presenting test samples to canines.

As previously stated, canines are able to detect cancers based on the VOCs released by tumors. Therefore, in theory, canine detection can be used to identify these olfactive biomarkers, which could then be used to calibrate 'electronic noses' that are then able to detect these same markers at an efficient and reliable scale. This is evidenced by a 2012 study in Germany that found canines were able to distinguish COPD from lung cancer due to a detectable biomarker associated with lung cancer [15]. Though they were not able to specify the identity of this biomarker, they note that the integration of 'sniffer dogs' with electronic nose technologies would allow them to do so and would therefore further the progress and effectiveness of non-invasive lung cancer screening [15]. However, there are certain limitations associated with 'electronic noses.' Namely, they are not able to specify novel odors and can only be used in controlled environments due to their high sensitivity [15]. In contrast, canines have been shown, to a certain extent, to have the ability to filter 'background odors' that may otherwise skew the readings of 'electronic noses' [14]. Though in the future 'electronic noses' may be a non-invasive and efficient strategy for cancer screening at the present technological state, canines show greater promise detecting cancers and identifying biomarkers associated with a variety of cancers, which could serve as valuable information in research and pharmaceutical development.

A major criticism of canine cancer screening is its lack of efficiency in terms of resources, time, and output. Canines can only work for a certain number of hours per day before losing motivation or becoming fatigued. This limits the number of samples that can be analyzed and therefore also limits the scale and impact of canine screening. The accuracy of detection may also vary based on the conditioning of the canines, which will in turn require a systematic process for checking canine conditioning with known controls. In addition, a canine's accuracy is unknown until further testing in a clinical setting. Therefore, these canines are unable to immediately receive a reward, reducing their motivation or even sensitivity to cancer samples. For that reason, canines would have to undergo ongoing training regimens to improve motivation and their detection skills [16].

Another limitation of canine cancer screening are confounding variables that may skew a canine's analysis. These confounding variables include both intrinsic properties of the sample, such as age, and extrinsic properties, such as extraneous odors [14, 17, 18]. In fact, a 2011 study conducted in the United Kingdom found that there was a decrease in the canine performance in patients with abnormal urine, and specificity tended to be lower for older patients [18]. Another study explains the issue of using factory made masks to collect breath samples as there is no guarantee that these masks would not harbor confounding odors [14]. They also argue that different environments surrounding the subject could foster odors that may influence the results. However, a different study showed that canine scent detection through breath samples and watery stools were not influenced by current smoking, colorectal disease, or inflammatory disease [1]. Therefore, there is some evidence that canines can ignore certain extraneous odors. In addition, it is worth nothing that both the Willis and Sonoda studies trained and tested canines with a limited number of samples, usually between 30 and 60 [1, 10]. Therefore, there is an argument that with a large enough library of scents (hundreds to thousands), canines would be able to ignore these variables and focus on the odors associated with cancerous cells.

It is unclear whether canines can distinguish between types of tumors, unlike traditional cancer screening. In fact, a 2013 study found that their trained dogs were not able to distinguish between benign and malignant lung tissue [17]. However, another study showed that its dogs could distinguish between malignant and benign ovarian tissues [4]. The cause of these discrepancies is unknown, but it is fair to attribute some of it to different training regimens and experiences. As mentioned before, canines who have undergone more training and have a larger library of scents to draw from may be more likely to correctly distinguish tumors. Whether canines can distinguish between cancer stages is also unknown and requires further study, particularly the ability to detect preclinical stages of disease with the same level of accuracy as cases that have already been diagnosed. There may be limitations due to other substances that are not specific for cancer that may be found in breath samples. This might mislead the canines' ability for detection due to unknown characteristics. For example, a 2015 study from Poland claimed that tobacco smoking reduced canine accuracy when analyzing breath samples [19]. In addition, 'hospital odor' was found to be in the breath samples of lung cancer patients and was proven to be a confounding variable. Specifically, the study found that samples taken in a hospital setting tended to have more false alerts than those collected in ambient air. However, this disparity was only found when canines were given a line of samples and instructed to find the one that had cancer. Conversely, when canines were given one sample at a time and instructed to indicate whether this sample had cancer, there were no significant differences in canine accuracy [19]. These findings indicate that the structure of canine cancer training and screening can have a significant impact on a study's results. Therefore, it is important to consider these factors in order to minimize confounding odors when procuring samples and thereby improve the generalizability of canine cancer screening by providing canines an abundant sample library. However, this periodic process of ensuring canines remain properly trained is both expensive and time-consuming when compared to traditional methodology, highlighting a serious limitation of canine screening. It is also important to note that given that cancer cells can emit unique odors, the specific compounds responsible for those unique scents have not been completely identified by scientists. In fact, a combination of several VOCs can be displayed from a single cancer. Thus, there may be a qualitative or quantitative difference from a combination of VOCs that might limit canine detection ability.

The canine training protocols utilized amongst all studies varied slightly in their methodology though they had key parallels. First, there was no indication in any of the studies that one breed exhibited greater olfactory competence than others, though German Shepherds and Retrievers seemed to be popular. All canines used in these studies had undergone basic obedience training and therefore recognized cues from trainers. Some canines may have had experience in being rescue dogs or drug testers, though none had been trained to detect cancer samples prior to their respective studies [1, 6, 13].

Across all studies, there seemed to be two common phases of training. During the first phase canines learn the odor signature of cancer samples, usually displayed to the canines in low concentrations. This was done by placing the sample, such as in the form of rag, and repeatedly exposing this to the canine until it became interested in the odor [3]. Once the canines expressed interest and recognized the cancer samples consecutively over numerous trials, they would move on to the second phase: odor discrimination. In this phase canines were trained to distinguish cancer samples from controls that usually came from healthy patients or those afflicted with non-cancerous diseases. The samples were in ventilated containers that allowed canines to smell the sample without any physical contact [1, 3, 11, 17]. These samples were placed in identical containers or boxes and usually given a randomized number to keep double blind conditions [20, 21]. Furthermore, samples were not reused to prevent the canines from recognizing past scents [1, 11]. Canines were trained to give cues to communicate the identity of the sample by lying down or sitting next to samples they believed were cancerous and walking past those they believed were not [1, 4–6, 9, 11, 18, 20]. All training regimens used some form of operant conditioning, usually in the form a clicker, food, or tennis ball, where canines would be rewarded for correctly identifying the sample to maintain their motivation across multiple trials [1, 7, 9, 11, 13, 20, 21]. Additionally, equipment, containers, and rooms were sanitized or replaced between trials of testing to avoid cross contamination [1, 9, 13, 20]. This protocol was also followed during testing periods. Understandably, canine testing protocols had similar methodology to that of the training protocols. However, testing usually had stricter guidelines on what cues were acceptable to ensure only clear and consistent cues were recorded. Therefore, hesitation or an incomplete response was marked as a failure even if the canine later made the correct identification [1, 11, 13, 18]. These results, often double checked via video recordings, were then used to calculate sensitivity (true positive rate) and specificity (true negative rate) to assess the canine's ability to distinguish cancer and control samples.

Overall, for almost all the types of cancers tested, specially trained dogs have reliably made accurate predictions regarding the presence of cancer. Measured by sensitivity and specificity, the accuracy of the results produced by cancer sniffing dogs are quite promising. By simply analyzing a blood, tissue, or urine samples, dogs have proved to detect multiple carcinomas with only their unique olfactory system. In fact, dogs have 300 million olfactory receptors, which is about 50 times that of humans. In addition, like all organisms, canines have a portion of their brain dedicated to scent detection. However, this area of the brain is 40 times as big in dogs as it is in humans. Correspondingly, it is not surprising that canines are able to complete what is an impossible task for humans. Cancer sniffing dogs can assess the presence of a variety of cancers with a high sensitivity and specificity and should therefore be considered as a serious option for non-invasive cancer screening.

This systematic review of the various canine training and testing protocols utilized by scientists, canine trainers, and testing institutions across the world serves to identify the critical methodologies that either were common place amongst most studies or were devised as unique strategies to optimize training and testing. These same strategies may be used as a guideline for designing future canine cancer detection studies. In addition, this section serves to better elucidate assessment of the olfactory aptitude of canines and its related terminology.

Canine cancer detection is an exciting and relatively novel avenue of cancer screening, though, due to its various limits, it should not be the only source in making an informed medical decision. Instead, canine testing should be used in concert with traditional cancer screening. In fact, we believe that despite the many limits of canine cancer detection, it still may prove useful for preliminary screening as it would provide a cheaper and more accessible option for cancer screening, which would be especially beneficial to those in underserved areas. Cancer screening using canines can also be an efficient and cost-effective means of surveillance with the development of specific guidelines. After a patient with cancer is treated, he or she is usually checked at regular intervals of time to determine if the cancer has returned through various scans and other tests. If a protocol is developed that only requires the patient's blood or urine to be evaluated instead of expensive, time-consuming scans, it would save a significant amount of money and resources, not only for the patient being tested, but also for the organization conducting the test. More frequent and lower-cost cancer assessments would ideally encourage those who have 'positive' results to seek further medical attention to validate the health concerns they might have. In addition, canine detection would prove very useful in specialized areas such as detecting cancers in higher-risk populations, such as firefighters, where frequent cancer screening is critical and necessary.

Further research is needed to examine combinations of potential biomarkers for various cancers. This analysis will help detect specific cancers, as well as assist with increasing the specificity and sensitivity values of screening performed by cancer sniffing canines. Overall, trained detection dogs are a useful and simultaneously cost-efficient screening tool for finding the best possible biomarker for an effective cancer diagnostic system.

Funding: None

Conflicts of Interest: None

Availability of Data and Material: Requests for data and material will be honored.

Code Availability: None