Breath diagnostics

In plain language

Every breath carries tiny amounts of chemicals called volatile organic compounds. Some of them come from the blood and tissues, and the mix can change when the body is unwell. The idea behind breath diagnostics is to collect a breath sample, measure its chemical profile, and look for a pattern that suggests a particular disease.

The tools range from electronic noses, which use arrays of sensors that respond to many gases at once, to laboratory instruments that identify individual molecules. Studies on lung cancer report high accuracy, but most are small and the results vary widely between devices. Few breath tests have reached routine clinical use, mainly because the evidence is not yet strong enough to convince clinicians and regulators.

Technical explanation

Two main approaches are used. Gas chromatography mass spectrometry (GC-MS) separates and identifies individual VOCs, giving a detailed chemical fingerprint. Electronic noses (e-noses) use arrays of partly selective sensors and pattern recognition to classify a sample without identifying each compound. A 2024 meta-analysis of 25 lung cancer studies by Fan and colleagues reports pooled figures of 85% sensitivity and 86% specificity for exhaled VOC detection, with an AUC of 0.93. The same authors judged the risk of bias to be significant, and found that only 37 of 190 detected VOCs appeared in three or more studies, which suggests low reproducibility across studies.

Accuracy is usually reported as sensitivity (the share of people with cancer who test positive), specificity (the share without cancer who test negative) and the area under the receiver operating characteristic curve (AUC). These figures can hide trade-offs. In a 2023 multicentre study of an e-nose (aeoNose) for non-small cell lung cancer, Kort and colleagues reported 95% sensitivity but only 49% specificity in a validation set of 199 people, which means about half of the people without cancer would be flagged. An earlier study by Capuano and colleagues (2015) reported that an e-nose classified cancer and non-cancer samples with over 90% correct classification in 30 people (20 with lung cancer, 10 with other lung disease). The authors noted the models had not been validated on an independent dataset.

Breath sampling is also harder to standardise than blood sampling. Results depend on the device, sampling method and classification model. Owlstone Medical's Breath Biopsy platform is used in the PAN Cancer trial with Cancer Research UK Cambridge Centre, which aims to find breath signatures across several cancer types. As of the trial information reviewed here, its stated target cancers did not include lung cancer, and no results had been reported.

What the evidence says

Each statement is typed by the strength of its evidence and linked to the source that supports it.

Limitations and open problems

• Many lung cancer accuracy studies are small. The 2015 Capuano study had 30 participants and its authors state the models had not been validated on an independent dataset.
• The 2024 meta-analysis judged the risk of bias to be significant, which the authors say warrants cautious interpretation.
• High sensitivity has come with low specificity in some studies (Kort and colleagues, 2023: 49% specificity), so many people without cancer could need follow-up tests.
• Studies use different devices, sampling methods and cut-off values, and only 37 of 190 detected VOCs appeared in three or more studies, so results are hard to compare or reproduce.
• Few breath tests have been approved by the US Food and Drug Administration. Modak (2013) reported three approvals in the preceding 15 years.
• The PAN Cancer trial had not reported results in the sources reviewed, so there is no published evidence yet from a large trial that breath tests can find cancer early in routine care.

Possible future uses

Product concepts: Clinic breath screening unit

Timeline

1. Jun 2013Modak's review in the Journal of Breath Research reports that only three breath tests had been approved by the FDA in the previous 15 years. Sources: [4]
2. 12 Nov 2015Capuano and colleagues publish in Scientific Reports that an e-nose classifies cancer and non-cancer samples with over 90% correct classification in 30 people, without independent validation. Sources: [3]
3. Jan 2019The PAN Cancer trial of Owlstone Medical's Breath Biopsy technology starts recruiting, according to the NIHR Cambridge Biomedical Research Centre. Sources: [6]
4. Mar 2023Kort and colleagues publish a multicentre e-nose validation study for non-small cell lung cancer in Chest, which states that clinical implementation has not yet been realised. Sources: [2]
5. 27 Jun 2024A meta-analysis of 25 exhaled VOC lung cancer studies is published in BMC Cancer, reporting pooled sensitivity of 85% and specificity of 86%. Sources: [1]
6. 30 Sept 2024Owlstone Medical signs a five-year research collaboration with the FDA's Center for Devices and Radiological Health to improve identification of breath VOCs. Sources: [5]

Futures this could help bring about

How far away?

Checking blood sugar without needles

Engineering prototypes · 2 of 5 steps done

Glossary

Volatile organic compound (VOC)

A carbon-based chemical that easily turns into a gas at normal temperatures. VOCs in breath come from the body and from the environment.

Electronic nose (e-nose)

A device with an array of gas sensors and pattern-recognition software that classifies a smell or breath sample without identifying each chemical.

Sensitivity and specificity

Sensitivity is the share of people with a disease who test positive. Specificity is the share of people without the disease who test negative.

Area under the curve (AUC)

A number from 0.5 (no better than chance) to 1.0 (perfect) that summarises how well a test separates people with and without a disease across all cut-off values.

Sources

1. review article
   Exhaled VOC detection in lung cancer screening: a comprehensive meta-analysis
   Xianzhe Fan, Ran Zhong, Hengrui Liang et al.. BMC Cancer, 27 Jun 2024. DOI: 10.1186/s12885-024-12537-7. Accessed 21 Oct 2026.
2. peer reviewed paper
   Diagnosing Non-Small Cell Lung Cancer by Exhaled Breath Profiling Using an Electronic Nose: A Multicenter Validation Study
   Sharina Kort, Marjolein Brusse-Keizer, Hugo Schouwink et al.. Chest, Mar 2023. DOI: 10.1016/j.chest.2022.09.042. Accessed 21 Oct 2026.
3. peer reviewed paper
   The lung cancer breath signature: a comparative analysis of exhaled breath and air sampled from inside the lungs
   Rosamaria Capuano, Marco Santonico, Giorgio Pennazza et al.. Scientific Reports, 12 Nov 2015. DOI: 10.1038/srep16491. Accessed 21 Oct 2026.
4. review article
   Regulatory issues on breath tests and updates of recent advances on [13C]-breath tests
   Anil S. Modak. Journal of Breath Research, 18 Jun 2013. DOI: 10.1088/1752-7155/7/3/037103. Accessed 21 Oct 2026.
5. company announcement
   Owlstone Medical Enters into a Research Collaboration Agreement with the FDA to Support the Confident Identification of Volatile Organic Compounds on Breath and Expansion of its Breath Biopsy VOC Atlas
   Owlstone Medical, 30 Sept 2024. Accessed 21 Oct 2026.
6. news
   Can a simple breath test detect cancer?
   NIHR Cambridge Biomedical Research Centre, 16 Oct 2024. Accessed 21 Oct 2026.
7. peer reviewed paper
   Lung cancer detection by electronic nose analysis of exhaled breath: a multicentre prospective external validation study
   A I G Buma, M B Muntinghe-Wagenaar, V van der Noort et al.. Annals of Oncology, 31 Mar 2025. DOI: 10.1016/j.annonc.2025.03.013. Accessed 21 Oct 2026.

Revision history

3 changes since this profile was created.