Can you actually smell disease?

Some diseases, yes. A number of conditions produce a characteristic odour recognised clinically for decades: the sweet acetone of diabetic ketoacidosis, the sulfurous foetor hepaticus of liver failure, the mousy smell of untreated phenylketonuria, maple-sugar urine, and the fishy odour of trimethylaminuria (Messenger et al., 2013). These are real, established scent signals — though most illness does not announce itself so plainly, which is exactly why the field is working to read the fainter chemistry behind it.

Can dogs smell cancer?

Trained dogs can detect some cancers from a urine or breath sample above chance. A 2004 proof-of-principle study trained dogs to pick out bladder cancer from urine odour (Willis et al., 2004), and later work reported high accuracy for prostate cancer (Guest et al., 2021). The dogs are reading volatile organic compounds — the same signal that electronic-nose and breath-VOC research is now trying to read with instruments. It is a real, reproducible signal, not yet a deployed diagnostic.

Is there a medical test that detects disease from breath or scent?

One is in routine use: fractional exhaled nitric oxide (FeNO), a standardized breath test that measures airway inflammation and helps guide asthma treatment (American Thoracic Society, 2021). Beyond it, breath-VOC and electronic-nose panels for cancer and other conditions are under active validation but not yet deployed as clinical screens (de Vries et al., 2025). Distinct volatile signatures have also been identified in the sebum of people with Parkinson’s disease (Trivedi et al., 2019).

Does SHEMIM diagnose or cure disease by smell?

No. SHEMIM studies how to read breath and body scent as an early, honest health signal — prevention a person can own. It does not diagnose disease, and it does not claim to cure anything by smell. Where scent already carries a real signal, SHEMIM points to it and marks how settled the evidence is; the deployed clinical tools here are FeNO and, for now, trained-animal and research-grade breath analysis, cited as others’ work.

SHEMIM / Research / Can You Smell Disease?

Can You Smell Disease?

The Body’s Oldest Alarm

Some illnesses carry a literal smell. Dogs can be trained to catch cancer in a sample. And a breath test is already routine medicine. The honest question is not whether disease can be read by scent — but how far, and how surely.

Yes — and clinics have long known it

Some illness announces itself by smell.

This is not new, and it is not folklore. Clinicians are still taught the sweet acetone of a diabetic crisis, the sulfurous foetor hepaticus of a failing liver, the mousy smell of untreated phenylketonuria, the maple-sugar note of one inherited disorder, and the fishy odour of another (trimethylaminuria, where a single missing enzyme lets an amine escape unbroken). These are established, textbook scent signals — the body, in certain conditions, spelling its state out loud. Most illness is quieter than this. That quieter chemistry is what the research is learning to read.

Noses better than ours

Trained animals catch what we cannot.

Where our own noses fall short, better ones step in. In a 2004 proof-of-principle study, dogs were trained to pick bladder cancer out of urine by odour alone; later work reported high accuracy for prostate cancer. Malaria, for its part, reshapes a person’s body odour so that mosquitoes find them more attractive — the infection editing the host’s scent. The animals are reading volatile organic compounds: the same molecular signal that instruments are now being built to catch.

From nose to instrument

One breath test is medicine; the rest is coming.

This is not only hope — it is already a clinic. Exhaled nitric oxide (FeNO) is measured every day to gauge airway inflammation and steer asthma treatment: disease read straight from the breath. Beyond it, electronic noses and breath-VOC panels paired with machine learning can classify cancers in research settings, and prospective external validation is beginning. And a distinct volatile signature sits in the skin oils of people with Parkinson’s, first noticed by a person who could smell it, then confirmed by mass spectrometry.

Honest limit: only FeNO is in routine clinical use today. The wider breath-VOC and electronic-nose work is emerging and not yet deployed as a validated screen — and a real share of breath molecules come from diet and environment, not disease. We keep the deployed apart from the promising, and say which is which.

The reader is a signal too

And losing the sense itself is a warning.

The signal runs both ways. Not only does illness change our scent — losing our sense of smell is one of the body’s earliest alarms, an early sign of Parkinson’s and Alzheimer’s and among the strongest predictors of five-year mortality. The nose reads the world; and when it goes quiet, that quiet is itself worth reading. What smell loss reveals is a chapter of the same story told from the other side.

Where we draw the line

We read the signal. We do not diagnose by smell.

What SHEMIM does

Studies how to read breath and body scent as an early, honest health signal — prevention a person can own, marking evidence from inference at every step. This is our work.

What we only cite

Diagnosis by scent — the disease odours, the trained dogs, the breath-VOC and FeNO tests above. Real, and others’ work. We point at it; we never claim to diagnose or cure disease by smell.

How sure are we?

Every claim, tagged for how settled it is.

Established

Some illnesses carry a literal smell, taught in clinics for generations: the sweet acetone of a diabetic crisis, the sulfurous foetor hepaticus of liver failure, the mousy note of untreated phenylketonuria, maple-sugar urine, and the fishy odour of trimethylaminuria (Messenger et al., 2013).
One breath test is already routine medicine: exhaled nitric oxide (FeNO) is read directly from breath to gauge airway inflammation and guide treatment (American Thoracic Society, 2021).
Losing the sense of smell is itself among the strongest predictors of five-year mortality in older adults — the reader of the signal behaving like a vital sign (Pinto et al., 2014).

Documented · early

Trained dogs can pick disease out of a urine or breath sample above chance: bladder cancer in a proof-of-principle study (Willis et al., 2004) and prostate cancer with high accuracy in later work (Guest et al., 2021).
Parkinson’s disease leaves a distinct volatile signature in skin sebum — found because one woman could smell it, later confirmed by mass spectrometry (Trivedi et al., 2019).
Malaria infection changes a person’s body odour in ways that make them measurably more attractive to mosquitoes (Robinson et al., 2018).

Emerging

Electronic noses and GC-MS breath panels paired with machine learning can classify cancers from breath-VOC patterns in research settings, with prospective external validation now beginning (de Vries et al., Annals of Oncology, 2025).
Beyond FeNO, no breath-VOC or electronic-nose test is yet deployed as a validated clinical screen — the pattern-reading is promising, not proven.

Common questions

Frequently asked.

Can you actually smell disease?: Some diseases, yes. A number of conditions produce a characteristic odour recognised clinically for decades: the sweet acetone of diabetic ketoacidosis, the sulfurous foetor hepaticus of liver failure, the mousy smell of untreated phenylketonuria, maple-sugar urine, and the fishy odour of trimethylaminuria (Messenger et al., 2013). These are real, established scent signals — though most illness does not announce itself so plainly, which is exactly why the field is working to read the fainter chemistry behind it.
Can dogs smell cancer?: Trained dogs can detect some cancers from a urine or breath sample above chance. A 2004 proof-of-principle study trained dogs to pick out bladder cancer from urine odour (Willis et al., 2004), and later work reported high accuracy for prostate cancer (Guest et al., 2021). The dogs are reading volatile organic compounds — the same signal that electronic-nose and breath-VOC research is now trying to read with instruments. It is a real, reproducible signal, not yet a deployed diagnostic.
Is there a medical test that detects disease from breath or scent?: One is in routine use: fractional exhaled nitric oxide (FeNO), a standardized breath test that measures airway inflammation and helps guide asthma treatment (American Thoracic Society, 2021). Beyond it, breath-VOC and electronic-nose panels for cancer and other conditions are under active validation but not yet deployed as clinical screens (de Vries et al., 2025). Distinct volatile signatures have also been identified in the sebum of people with Parkinson’s disease (Trivedi et al., 2019).
Does SHEMIM diagnose or cure disease by smell?: No. SHEMIM studies how to read breath and body scent as an early, honest health signal — prevention a person can own. It does not diagnose disease, and it does not claim to cure anything by smell. Where scent already carries a real signal, SHEMIM points to it and marks how settled the evidence is; the deployed clinical tools here are FeNO and, for now, trained-animal and research-grade breath analysis, cited as others’ work.

Selected sources

Messenger et al. (2013), J. Clinical and Aesthetic Dermatology — trimethylaminuria (fish-odor syndrome) and FMO3.
American Thoracic Society (2021) — FeNO clinical practice guideline.
Pinto et al. (2014), PLOS One — olfactory loss predicts 5-year mortality.
Willis et al. (2004), BMJ — canine olfactory detection of bladder cancer (proof of principle).
Guest et al. (2021), BMC Cancer — canine VOC detection of prostate cancer.
Trivedi et al. (2019), ACS Central Science — Parkinson’s sebum volatilome.
Robinson et al. (2018), PNAS — malaria alters host odour and mosquito attraction.
de Vries et al. (2025), Annals of Oncology — eNose breath classification, external validation.

A signal worth learning to read.

If disease can be read in the breath and on the skin — sometimes by a dog, sometimes by a machine — the work is to read it early, and honestly. That is what SHEMIM studies, keeping evidence, inference and hypothesis clearly apart.

The Breath as a Signal → The questions behind it