What Does Cancer Smell Like? A Biological Perspective

Cancer cells produce a distinct chemical fingerprint made up of volatile organic compounds, or VOCs, that differ measurably from the compounds healthy cells release. These molecules escape the body through breath, urine, sweat, and stool, creating what researchers sometimes call a “volatilome” unique to cancerous tissue. Most of these signals are far too faint for the human nose to consciously register, but they are real enough to be picked up by trained dogs, tiny worms, and increasingly sophisticated sensor arrays. The biology behind this smell, and what it reveals about the disease, turns out to be surprisingly rich.

Why Cancer Cells Smell Different

The short explanation is that cancer cells run their metabolism differently from normal cells, and the byproducts of that altered metabolism include gases and lightweight molecules that drift into the bloodstream and eventually leave the body. One of the best-understood pathways involves oxidative stress. Tumor cells tend to generate high levels of reactive oxygen species, which attack the fatty acids in cell membranes through a process called lipid peroxidation. When omega-3 and omega-6 fatty acids in those membranes break down, they produce a family of aldehydes, including hexanal, heptanal, nonanal, and decanal, that are reliably elevated in cancer patients compared to healthy people.¹ Tumor cells also tend to have altered membrane lipid compositions, which shifts the specific mix of aldehydes they produce.

But aldehydes are just one slice of the picture. Cancer cells also release abnormal levels of alkanes, ketones, alcohols, sulfur compounds, and even simple gases like methane and hydrogen. A 2024 review described VOCs broadly as products of catalytic peroxidation driven by reactive oxygen species, positioning them as a natural consequence of the heightened metabolic chaos inside tumors.² The result is not a single “cancer smell” but a shifting profile of dozens of compounds whose relative concentrations change depending on the type and stage of the cancer.

Different Cancers, Different Chemical Signatures

Researchers have spent the past two decades cataloging which VOCs appear at abnormal levels in patients with specific cancers. The profiles are surprisingly distinct. In lung cancer, a large prospective study identified 16 breath-borne VOCs, classified as aldehydes, hydrocarbons, ketones, carboxylic acids, and furans, that together could distinguish cancer patients from healthy individuals with roughly 89% accuracy in an external validation group of over 500 people.³ Earlier work had flagged specific compounds like 1-butanol and 3-hydroxy-2-butanone as being at significantly higher concentrations in the breath of lung cancer patients.⁴

Melanoma, the most dangerous skin cancer, produces its own volatile fingerprint. Lab studies have shown that melanoma cells release dimethyl disulfide and dimethyl trisulfide, sulfur-containing compounds not detected from normal skin cells. Melanoma cells also showed higher levels of isoamyl alcohol but lower levels of isovaleric acid compared to healthy melanocytes, both derived from the amino acid leucine, which suggests the cancer rewires how it handles that particular building block.⁵ Those in-vitro findings line up with in-vivo evidence that trained dogs can reliably distinguish invasive melanoma from benign moles, other skin cancers, and healthy skin based on odor alone.⁶

Prostate cancer and bladder cancer have drawn attention for their urinary VOC profiles. One study used an electronic sensor platform to show that methane, iso-butane, hydrogen, and ethanol in urine had significant predictive capability for genitourinary cancers.⁷ Work on prostate cancer specifically found that a model based on four urinary VOCs, including pentanal and two ketones, reached around 63–65% accuracy on its own, but when combined with the standard PSA blood test, accuracy climbed to about 74%.⁸ Separate research distinguished prostate cancer from bladder cancer and from healthy controls through urinary VOC analysis, with area-under-the-curve values reaching 0.89 to 0.97 depending on the comparison.⁹

Colorectal cancer leaves traces in stool, which makes intuitive sense given the location of the tumor. VOCs emitted from stool represent the end products of microbial activity and metabolism, and they shift detectably in disease.¹⁰ Electronic-nose studies on colorectal cancer patients have found consistent VOC elevation across stool, urine, and breath, with particularly strong signals from sensors tuned to nitrogenous compounds, aromatic compounds, methane, and alcohols.¹¹

Animals That Can Smell Cancer

The most dramatic demonstrations of cancer’s chemical signature come from animals with far more sensitive noses than ours. Dogs possess roughly 200 to 300 million olfactory receptors, compared to about 5 to 6 million in humans, and their brains devote a proportionally much larger region to processing smell. This hardware lets trained medical detection dogs pick up VOCs at concentrations down to parts per trillion.¹²

Studies have put this ability to the test in controlled settings with striking results. In one double-blind trial using urine samples from breast cancer patients and healthy controls, a trained dog correctly identified all 40 cancer samples across 40 runs, achieving 100% sensitivity and specificity.¹³ A lung cancer study found that when a dog was given both breath and urine samples, it correctly flagged 40 of 41 cancer cases, for an overall detection rate of about 98%. Using urine alone, accuracy was around 88%; breath alone dropped to about 78%.¹⁴ These numbers are impressive, though it is worth noting that such studies typically involve relatively small sample sizes and highly controlled conditions.

Dogs are not the only animals in this story. Researchers have found that the tiny roundworm C. elegans, a creature barely a millimeter long with just 302 neurons, is attracted to the urine of cancer patients and repelled by the urine of healthy people. One study reported 88% sensitivity and 93% specificity for cancer detection using worm behavior.¹⁵ A separate team working with breast cancer urine samples found 75% sensitivity and over 97% specificity.¹⁶ Perhaps most telling, the worms’ attraction to a patient’s urine decreased after the tumor was surgically removed, suggesting their response tracks the tumor’s presence in real time.¹⁷

Even ants have gotten into the act. Researchers showed that individual ants could be trained in just a few trials to associate the odor of human ovarian cancer cells with a sugar reward. In subsequent memory tests without any reward, the ants spent significantly more time near the cancer-cell odor than near the control, demonstrating they had learned the VOC signature and could recall it.¹⁸ The advantage of ants, the authors noted, is that they are cheap, abundant, and fast to train compared to dogs. The obvious disadvantage is that ant-based diagnostics remain far from clinical reality.

Electronic Noses and Machine Breath Analysis

The animal studies, compelling as they are, have pushed researchers toward a practical question: can we build a device that does what a dog’s nose does? The answer is the electronic nose, or e-nose, which uses arrays of chemical sensors that each respond differently to different gas molecules. When a breath or urine sample passes over the sensor array, the pattern of responses creates a “smell print” that machine-learning algorithms can classify as cancerous or not.

Results from e-nose studies have been encouraging. One lung cancer screening study using six metal-oxide gas sensors and logistic regression achieved 95% sensitivity and 100% specificity across a group of 118 people, correctly classifying cancer in roughly 97% of cases.¹⁹ Another team using a similar sensor array with an artificial neural network reported about 86% sensitivity, 100% specificity, and 94% overall accuracy.²⁰ These numbers rival or exceed the performance of some conventional screening methods, and the sensors themselves are relatively cheap and stable.

For colorectal cancer, researchers have used a technique called selected-ion-flow-tube mass spectrometry to measure VOCs in exhaled breath.²¹ This kind of analytical chemistry does not rely on pattern matching the way an e-nose does; instead, it identifies and quantifies individual molecules. Both approaches have their strengths. E-noses are cheaper and faster but cannot tell you which specific compound triggered the alarm. Analytical instruments give you a precise chemical readout but are expensive and complex to operate.

What Humans Can Actually Smell

If you are wondering whether you yourself could ever smell cancer on someone, the honest answer is: probably not at an early stage, but sometimes at an advanced one. The VOCs produced by internal tumors are diluted enormously by the time they reach your breath or skin, making them invisible to the roughly 5 to 6 million olfactory receptors in a human nose. There are anecdotal reports of people or their partners noticing a change in body odor before a diagnosis, and some palliative-care nurses describe a particular scent around patients with advanced disease, but these observations have never been rigorously validated as diagnostic tools.

Where smell does become undeniable is with fungating wounds, a grim complication of advanced cancer in which the tumor breaks through the skin. These wounds arise from primary, secondary, or recurrent malignant disease and produce copious exudate, often accompanied by strong and distressing odor caused by tissue necrosis and bacterial colonization.²² The smell is frequently described as sweet-rotten or sickeningly putrid, difficult to mask, and deeply distressing to patients and caregivers alike. Topically applied metronidazole, an antibiotic, remains the most robustly supported treatment for controlling the odor, with emerging evidence for cadexomer iodine and manuka honey as complementary options that help disrupt the bacterial biofilms responsible for much of the stench.²³ Palliative radiotherapy can also reduce odor and exudate by shrinking the tumor mass. The management of fungating wound odor is fundamentally a quality-of-life issue: healing is rarely possible, but controlling the smell can make an enormous difference to a patient’s dignity and comfort in their remaining time.

Why This Has Not Reached Your Doctor’s Office Yet

Given how promising the numbers look in many of these studies, it is reasonable to ask why your doctor is not already having you breathe into a sensor at your annual check-up. The gap between laboratory performance and clinical deployment is wide, and several stubborn problems sit in it.

Standardization of breath collection is one of the biggest. Factors as seemingly minor as how fast you exhale, whether you hold your breath beforehand, how much air you push out, and even the composition of the room air you breathe in before the test can all significantly change the VOC concentrations measured.²⁴ If different research groups collect samples differently, their results cannot be meaningfully compared, and no consensus emerges about which compounds truly matter. Diet, medications, smoking, and recent meals add further noise. A garlic-heavy dinner can spike certain sulfur compounds in your breath for hours, and distinguishing that from a melanoma-related sulfur signal is not trivial.

The other major bottleneck is validation at scale. A review in Molecular Diagnosis & Therapy put it bluntly: the success of VOC-based disease identification has so far been limited to laboratory settings, and large-scale clinical data are still needed to establish diagnostic robustness.²⁵ Most of the impressive sensitivity and specificity numbers come from studies with dozens to a few hundred participants, not the thousands typically required before a regulatory body would approve a new screening test. Cancer screening also demands extremely low false-positive rates, because telling a healthy person they might have cancer has real psychological and financial costs. A 90% specificity rate sounds good until you realize it means roughly one in ten healthy people would get a false alarm.

Engineering Cells to Broadcast a Detectable Scent

One creative approach sidesteps the problem of detecting naturally occurring VOCs altogether. Instead of listening for cancer’s faint chemical whisper, researchers have tried giving cancer cells a loudspeaker. A team at Columbia University engineered human cervical cancer cells to express limonene, the compound responsible for the citrus smell of lemons and oranges, by inserting a single plant gene encoding limonene synthase. In a mouse model, tumors made from these engineered cells produced enough limonene to be detected by mass spectrometry, and the signal was sensitive enough to flag tumors as small as 5 millimeters. The researchers improved the signal further by co-expressing a modified enzyme from the cholesterol synthesis pathway, roughly doubling the limonene output.²⁶

This is still a proof of concept, far from anything you would encounter in a clinic. You cannot retroactively insert a limonene gene into a tumor that has already formed on its own. But the work illustrates a broader idea: if cancer’s natural VOC signal is too faint or too variable to be a reliable diagnostic, perhaps the signal can be amplified or replaced with something cleaner. Future versions of this concept might use gene therapies or engineered probiotics that home in on tumors and produce a detectable volatile marker once they arrive. It is speculative, but the mouse data suggest the underlying biology works.

Bladder Cancer and the Urine Connection

Among cancers detectable through body-fluid odor, bladder cancer has a particularly direct relationship with urine. The tumor sits in the organ that produces and stores urine, so its metabolic byproducts have an unusually short path to an easily collected sample. A metabolomics study comparing urine odor from bladder cancer patients with healthy controls identified 12 distinct metabolites, and the chemical profiles of most preoperative patients separated cleanly from controls using principal component analysis. The researchers noted that this VOC-based approach may actually have higher sensitivity than urinary cytology, the standard method of examining urine under a microscope for abnormal cells.²⁷ Cytology is highly specific but misses a lot of early-stage disease, so a smell-based supplement could fill a genuine gap.

The nematode studies described earlier add another layer here. In one experiment, the worms’ behavioral response shifted back toward the “healthy” pattern after patients had their bladder tumors removed, essentially showing that the cancer-related scent in the urine disappeared along with the tumor. If that finding holds up in larger studies, it suggests urine-odor testing could serve not just as a screening tool but as a way to monitor whether a treatment is working. That kind of noninvasive, repeatable follow-up test would be genuinely useful in a cancer where recurrence after surgery is common.

How Smell Research Fits with Other Liquid Biopsies

VOC analysis is not the only approach trying to detect cancer through body fluids rather than tissue biopsies. Circulating tumor DNA, exosomes, and protein biomarkers in blood are all under active investigation. What sets VOC-based methods apart is their potential simplicity and cost. A blood-based liquid biopsy requires sophisticated molecular biology techniques and expensive sequencing. An e-nose, by contrast, is a small device with sensors that cost a few dollars each and an algorithm running on an ordinary computer. If the standardization and validation problems can be solved, the barrier to deploying breath or urine VOC testing in low-resource settings or routine primary care would be substantially lower than for genomic liquid biopsies.

That said, smell-based detection is unlikely to replace other methods entirely. The chemical signal varies with diet, medications, comorbidities, and even the time of day, introducing noise that more targeted molecular tests can avoid. The most plausible near-term role for VOC analysis is as a first-pass triage tool: cheap, fast, noninvasive, and good enough to flag people who should get further workup, rather than serving as a definitive diagnostic on its own. Whether it reaches even that modest goal depends on the large-scale clinical trials that have not yet been completed.