Breathomics

Blood will surrender its dissolved secrets only through needles; its volatile fraction leaks out through the lungs all day for free. Exhaled breath carries thousands of distinct organic compounds — metabolic products evaporating from circulation, lipid-peroxidation fragments marking oxidative stress, gases fermented by gut microbes reabsorbed then respired — each typically present below a part per billion. Breathomics proposes that this free exhaust stream can be read as diagnostics, and half its technology already exists; the contested half is what the readings mean.

How volatiles reach the alveolar air

Gas exchange obeys simple partitioning: anything volatile dissolved in blood equilibrates with alveolar air according to its solubility, so compounds produced anywhere in the body — the liver’s chemistry, tumours’ aberrant redox, adipose ketogenesis — ride the bloodstream to the lungs and exit within minutes of formation. That speed is the field’s genuine promise: unlike biopsy or blood draw, breath samples current physiology, capturing metabolic state minutes old. It also imports ambient reality directly into the sample: room air pollutants, cosmetics, last night’s garlic and the clinic’s disinfectant all appear alongside endogenous signal, making collection protocol — inhaled-air controls, mouth versus nose routes, end-tidal fraction rather than mixed volume — part of the analytical method itself.

Instruments for the invisible

Concentrations near parts per trillion sit far below ordinary detectors. Proton-transfer reaction mass spectrometry answers in real time by soft-ionising analytes with hydronium ions fast enough to follow breath cycle by cycle. Gas-chromatography mass-spectrometry separates first and identifies slower but with authoritative specificity. Both must wage war on water — exhaled breath is fully saturated vapour, and humidity murders adsorbents, ionisation efficiency and chromatography alike — which is why preconcentration traps, drying stages and inlet engineering occupy as much patent space as detection itself.

Pattern recognition and its recurring embarrassment

No single compound flags cancer; the hope lives in multivariate breathprints fed to statistical classifiers. Here the field carries scar tissue. Multiple celebrated studies taught algorithms to distinguish patients from healthy volunteers using signatures that later proved to encode recruitment site, dietary habits, collection hardware — anything correlated with group membership, exactly as detection dogs famously keyed on handlers’ cues. Reproduction in independent centres remains the graveyard of candidate biomarkers, and rigorous designs now mandate blind multi-centre validation against boring covariates before any accuracy claim earns belief. This is the same prevalence-and-confounding arithmetic that governs every screening dream, sharpened by breath’s unusually promiscuous background.

Meanwhile three humble tests have quietly held clinics for years, and their common shape explains the contrast. The urea breath test feeds isotopically labelled urea; only Helicobacter pylori’s urease cleaves it, exhaling labelled carbon dioxide — mechanism so specific the result needs no statistics. Exhaled nitric oxide tracks eosinophilic airway inflammation well enough to steer asthma therapy as a routine measurement. And breath acetone reliably flags ketosis, one compound carrying one physiological story. Single molecules with accountable chemistry pass; thousand-compound symphonies still audition.