The Smell of Disease

Medicine smelled disease before it could measure it. Hippocrates described the breath of liver failure patients. Avicenna used urine odor as a diagnostic indicator. Chinese practitioners were cataloguing breath as a disease marker three thousand years ago. The mechanisms were unknown, and the observations varied in reliability. Some became recognizable clinical signs long before their chemistry was understood.

Modern medicine largely abandoned olfactory diagnosis when laboratory tests became available. More quantifiable tools produced more defensible diagnoses; the nose couldn't be calibrated or documented. What's happened in the past two decades is something interesting: the molecular biology has caught up, and some of those observations have become starting points for molecular investigation. Many diseases have characteristic volatile organic compound (VOC) signatures that emerge reliably from the pathology and are detectable through smell: by trained animals, by sensitive instruments, and in some cases by humans with sufficiently acute perception. A perceptive observation can precede an explanation by centuries.

Why diseases smell

The human body produces hundreds of volatile organic compounds continuously, through metabolism, through bacterial activity on the skin and in the gut, through cellular processes at every level. Most of these are exhaled through the lungs, excreted through sweat, or released in urine and feces in consistent, relatively stable patterns that constitute a person's normal VOC profile.

Disease disrupts this profile in specific ways, depending on the disease. There are four main pathways through which pathology produces a characteristic smell.

The first is metabolic pathway disruption. When normal enzymatic pathways are blocked or overloaded — by an enzyme deficiency, by organ failure, by an infection hijacking cellular metabolism — compounds accumulate that would ordinarily be converted and cleared. These accumulations eventually reach concentrations high enough to volatilize and escape through the lungs and skin. The smell is the byproduct of a metabolic traffic jam: chemicals that aren't supposed to accumulate, accumulating.

The second is filtration failure. The liver and kidneys spend a significant portion of their effort clearing potentially odoriferous compounds from systemic circulation before they reach the lungs or skin surface. When either organ fails, those compounds stay in the blood, reach the lungs, and are exhaled. The disease smell is the smell of the normal filtration system being overwhelmed.

The third is microbial metabolites. Bacteria, fungi, and parasites have their own metabolic processes and produce their own volatile byproducts. An infection by a specific pathogen introduces that pathogen's chemical signature into the body's VOC environment. Different species produce characteristic compounds at identifiable concentrations, which is why some infections can, in principle, be diagnosed by smell without waiting for cultures.

The fourth is altered cellular metabolism in disease states. Cancer cells, degenerating neurons, and other pathological tissue don't metabolize in the same way as healthy cells. They produce different mixes of metabolic byproducts at different concentrations, subtly changing the body's overall VOC profile in ways that don't map to obvious organ failure but are nonetheless detectable at sufficient sensitivity.

Specific diseases and their signatures

Diabetic ketoacidosis is the most widely recognized clinical smell. When the body burns fat for fuel in the absence of sufficient insulin, it produces ketone bodies: acetone, acetoacetate, beta-hydroxybutyrate. Acetone is volatile and exhaled through the lungs in large quantities. The resulting breath has a fruity quality, described as sweet, or like pear drops or nail polish remover. In emergency medicine, this smell on a patient's breath combined with the appropriate clinical picture is considered a diagnostic sign. Its absence is not enough to rule the condition out; odor is one sign within a clinical assessment.

Liver failure produces fetor hepaticus (Latin for "liver stench," though in clinical practice it's sometimes called "the breath of the dead"). The characteristic breath contains the smell of dimethyl sulfide and methyl mercaptan accumulating in blood and exhaled through the lungs. The liver normally processes methionine (an amino acid) and clears the sulfur-containing byproducts of its metabolism. In severe liver failure, this processing fails, thiols pass directly into the lungs via portal-systemic shunting, and the characteristic smell emerges: sweet, musty, faintly fecal, with a sulfurous undertone. The Cleveland Clinic describes it as "rotten eggs and garlic." It's a late sign, appearing at a stage of disease where significant hepatic function is already lost, which is part of why Hippocrates could associate it with a poor prognosis.

Kidney failure produces a different signature: uremic fetor, an ammonia or bleach smell on the breath and sometimes detectable from the skin. When kidneys stop filtering urea from blood, urea accumulates and enters saliva, where bacterial urease can break it down into ammonia. The smell is chemically distinct from fetor hepaticus.

Phenylketonuria (PKU) and maple syrup urine disease (MSUD) are metabolic enzyme deficiencies that produce characteristic smells through the first pathway described above. PKU, caused by a failure to metabolize phenylalanine, produces phenylpyruvic acid and phenylacetic acid: the source of a musty, mousy odor historically strong enough that nurses on pediatric wards would describe being able to identify affected children before testing. MSUD produces an accumulation of branched-chain amino acid metabolites, including sotolone, a lactone with an intensely sweet, caramel, maple syrup character that gives the disorder its name, detectable in the urine and sweat. Both conditions are now screened at birth in most developed countries, meaning the clinical smell is rarely encountered anymore.

Pseudomonas aeruginosa wound infections smell fruity, sometimes specifically of grape. The compound responsible is 2-aminoacetophenone, a metabolic byproduct produced by Pseudomonas specifically. It's distinctive enough that experienced wound care nurses report being able to identify Pseudomonas infections by smell before cultures return. Clostridium difficile produces one of the most recognizable infectious disease smells: intensely foul, barnyard-adjacent, with a specific pungency attributed to p-cresol and isovaleric acid. Staff have reported recognizing it, but a blinded study found that nurses could not identify positive stool samples better than chance. Recognizing a ward situation and identifying a pathogen from odor alone are different tasks. Bacterial vaginosis has its characteristic fishy smell from trimethylamine produced by anaerobic bacteria, the same compound responsible for trimethylaminuria (fish odor syndrome), a metabolic disorder where the enzyme that normally oxidizes trimethylamine to the odorless N-oxide is absent or deficient.

Historically documented disease smells include typhoid fever, described as freshly baked bread; yellow fever as a butcher's shop; and measles as freshly plucked feathers. The specific compounds behind these historical descriptions have not been fully characterized, and repeated descriptions are not enough to establish diagnostic reliability. They remain an interesting part of medical history rather than validated tests.

Animals and the extended sensitivity problem

The reason dogs are useful here isn't that disease smells are detectable only by extraordinary means. It's that the concentrations at which disease VOCs appear, especially in early-stage disease, often fall below the human olfactory threshold while sitting well above a dog's. Humans have around 6 million olfactory receptors; dogs can have up to 300 million. The dog isn't detecting something chemically different from what we could detect in principle. It's detecting the same compounds at concentrations we can't register.

Trained scent dogs have been studied for detecting hypoglycemia, seizures, infections, and several cancers. The COVID-19 studies provide a particularly concrete example. In one controlled evaluation using sweat samples from a hundred cases and more than four hundred non-cases, fifteen trained dogs achieved overall sensitivity of about 95% and specificity of about 97%. Those are impressive results for that task and setting. The next question is how well the detection holds up with new people, different collection conditions, and routine use.

Research programs have reported dogs, rats, ants, and other animals detecting cancers, diabetes, epilepsy, tuberculosis, malaria, urinary tract infections, and SARS-CoV-2. Rats (specifically African giant pouched rats trained by the APOPO organization) have been used in tuberculosis screening in sub-Saharan Africa for years. They work on sputum samples faster than microscopy and with higher sensitivity. A rat can screen 100 samples in twenty minutes; a laboratory technician takes up to four days. TB follows the same pattern: the bacteria have a specific VOC profile, the rats are trained on it, and detection remains reliable below the human threshold.

Joy Milne and Parkinson's disease

The most remarkable human case in this literature is Joy Milne, a Scottish nurse whose husband Les was diagnosed with Parkinson's disease in 1986. For many years, she had noticed that he emitted a musky odor, but assumed it was unique to him. In 2012 she attended a Parkinson's support group and smelled the same distinct odor on other members with the disease, which led her to question whether it was a wider phenomenon.

She approached Dr. Tilo Kunath at the University of Edinburgh, who designed a test where Joy was asked to smell T-shirts worn by Parkinson's patients and controls. She correctly identified all six patient T-shirts. She also flagged one control group T-shirt, which turned out to have been worn by someone who was subsequently diagnosed with Parkinson's eight months later. Her result raised the possibility that the odor change could precede diagnosis.

Preliminary tests indicated the odor was present in areas of high sebum production (the upper back and forehead) and not in the armpits. The disease was altering sebum chemistry, not axillary odor. Researchers found a sebum signature associated with Parkinson’s, including altered signals tentatively assigned to eicosane and perillic aldehyde. Other candidate compounds appeared too, with less consistent statistical evidence. Joy was later listed as co-author on the published paper in ACS Central Science. The work has led toward the development of a diagnostic skin-swab test.

What Parkinson's disease does to sebum chemistry, and why it does it, isn't entirely understood. But the observation was real, the detection was real, and what enabled it was one person with an unusually acute sense of smell in close contact with someone over a long enough period to notice a change that she couldn't initially explain. She had noticed the change years before the diagnosis.

The diagnostic direction

The scientific response to these observations has been the development of breathomics and volatolomics: fields dedicated to characterizing the complete VOC profile of breath, sweat, urine, and other biofluids as a disease detection and monitoring tool. Electronic nose devices and mass spectrometry of breath are being developed as potential clinical instruments. The underlying hypothesis is straightforward: if a disease produces a sufficiently distinctive VOC pattern, and if that signature can be measured with sufficient sensitivity and specificity, then non-invasive early disease detection becomes possible through smell, or rather, through instruments that do what noses do, at the sensitivity levels dogs achieve rather than the ones humans can manage.

The difficult step is moving from an intriguing observation to a reproducible test. Joy Milne’s experience gave researchers a specific place to look: the chemistry of sebum. The C. difficile example shows why the next step matters. A smell that seems unmistakable in a familiar clinical setting may not identify the disease when the surrounding clues are removed.

Breath and skin chemistry offer something we have barely begun to use systematically. What interests me is the possibility of catching a change while it is still a chemical disturbance, before it has become an unmistakable illness.