Can Dogs Smell Cancer? Yes!

Good Science starts with Good Questions

This topic has been both popular and controversial for several decades already, and it consists of two fundamental questions:

• Does cancer “smell”, at all?
• Can dogs reliably sense that smell?

The publication I will be reading with you is answering positively the first question, confirming that there is a “smell of cancer”.

Filianoti, Alessio, Manuela Costantini, Alfredo Maria Bove, Umberto Anceschi, Aldo Brassetti, Mariaconsiglia Ferriero, Riccardo Mastroianni et al. “Volatilome analysis in prostate cancer by electronic nose: A pilot monocentric study.” Cancers 14, no. 12 (2022): 2927.

At Maecenium, we don’t want to write sensationalistic posts. Instead, we want to gently nudge you towards the original source and let you explore the real, original publication on your own before making final conclusion.

Publication presented today is Open Access, thus you can read it for free. If you stumble upon a publication that you cannot access for free, read our simple “How to Tutorial”, explaining the process of downloading the publications that are behind the pay wall.

Let’s read the publication together, critically, as true Renaissance Minds:

Electronic Nose and PCA

Original Abstract: Urine analysis via an electronic nose provides volatile organic compounds easily usable in the diagnosis of urological diseases. Although challenging and highly expensive for health systems worldwide, no useful markers are available in clinical practice that aim to anticipate prostate cancer (PCa) diagnosis in the early stages in the context of wide population screening.

This is why Maecenium was created. As you can see, there is a chasm between the methods available for medical research and methods available for scientific research. There is no conspiracy in that, nor some sinister reasons, it’s just a matter of division of tasks. Medicine must work with known and well established (just like army for example) while science can operate in the realm of unknown (anyone can buy gadgets that that are better than what army uses).

Some previous works suggested that dogs trained to smell urine could recognize several types of cancers with various success rates. We hypothesized that urinary volatilome profiling may distinguish PCa patients from healthy controls. In this study, 272 individuals, 133 patients, and 139 healthy controls participated. Urine samples were collected, stabilized at 37 °C, and analyzed using a commercially available electronic nose (Cyranose C320). Statistical analysis of the sensor responses was performed off-line using principal component (PCA) analyses, discriminant analysis (CDA), and ROC curves. Principal components best discriminating groups were identified with univariable ANOVA analysis. groups were identified with univariable ANOVA analysis. Here, 110/133 and 123/139 cases were correctly identified in the PCa and healthy control cohorts, respectively (sensitivity 82.7%, specificity 88.5%; positive predictive value 87.3%, negative predictive value 84.2%). The Cross Validated Accuracy (CVA 85.3%, p < 0.001) was calculated. Using ROC analysis, the area under the curve was 0.9. Urine volatilome profiling via an electronic nose seems a promising non-invasive diagnostic tool.

PCA is the staple method for dimensionality reduction. Imagine that dataset consists of 50 variables (50 different proteins for example), and that each of those 50 variables contribute to the final result as the simple  A+B+C+D… PCA will turn those 50 into new set of variables and the first variable (called Principal Component) will explain the majority of variance (effect), the second a bit less and so on… In other words, PCA is used to distinguish between wheat and chaff. If you want to learn more about it, StatQuest made the best possible video explaining it.

Here, 110/133 and 123/139 cases were correctly identified – not bad at all and you can see by pure intuition that there is statistical significance. To develop the feeling for stats, if you see sample size bigger than 50, pretty much everything that is not near 50:50 will be statistically significant.

For this study, we tested a commercially available electronic nose (Cyranose 320, Smith Detections, Pasadena, CA, USA). It is a handheld device with a nano-composite array of 32 organic polymer sensors, and its operation is quite simple: as soon as the sensors are exposed to a mixture of VOCs, the polymers expand, stimulating a variation in their electrical resistance

Brilliant and elegant!

If you want to buy a toy like that, it will cost anywhere between 10.000 and 100.000 dollars. Keep searching on eBay, because you never know X)

Future, and dogs smelling cancer

Can cancer be sniffed? Obviously yes, because the array of sensors was able to pick up relevant volatile components.

Can dogs smell cancer? Obviously yes, I have no doubts that numerous animals can smell better that the device used in this publication.

Can dogs be trained? I guess so… Why not?

Can math be improved? Certainly yes, because PCA is the crudest method for dimensionality reduction and the percentage could go up at least 5% (my intuition after playing with data for 10+ years).

Can the instrument be improved? Yes.

Are we going to see dogs or electric noses smelling cancer? NO! Even if the results were 95%, today, it would take years of clinical studies (and paperwork) to put it on the market and to give it CLIA. For Research Use Only, probably soon.

Actually, I could start replicating this very experiment – today. The instrument is easy to use and I can do PCA in 10 seconds. 

(*cute dog would be a gimmick, a decoy to get more attention from girls. However, my wife would not allow me, probably. Thus no cute dogs)

Could you help the development of similar methods?

Yes, join Maecenium, donate and support scientific projects, because scientists can perform the measurements beyond conventional methods for Research Use Only (not for the Official Diagnostics, but it can help you determine whether you should look for the conventional diagnostics).

Read our blogs, write to us, consider donating, or buy whatever you like after clicking the links below, so we can earn a percentage (it’s seamless donation… You simply buy things that you would buy anyways, for the same price, and you will seamlessly donate).