Breath tests for disease: what the evidence shows so far

By the Vrulum team

Why breath is being studied

Exhaled breath carries gases and very small amounts of volatile organic compounds (VOCs). Researchers are asking whether particular patterns in breath could help screen for or monitor illness. Because a breath sample needs no needle, the idea is attractive. The evidence is still early, so this article sorts it into three groups to show how much weight each claim can bear.

How the evidence is sorted

Established means a guideline or consensus document. Demonstrated means published studies, named here with their year. Reported or speculative covers company claims and press coverage.

Established

The most relevant consensus document comes from a European Respiratory Society (ERS) Task Force on exhaled biomarkers in lung disease, published in the European Respiratory Journal. An ERS news item about it states that the document is "not intended to provide clinical guidance on disease diagnosis and management" (ERS news item). The same item says there were no official recommendations for exhaled VOCs or exhaled particles at the time. The full document was not read for this article, so its contents are not described here.

Fractional exhaled nitric oxide (FeNO) is the only breath test with a NICE diagnostics guidance listing in the sources opened for this article. A bibliographic database record (INAHTA) lists the National Institute for Health and Care Excellence (NICE) guidance DG12, dated 2014, titled "Measuring fractional exhaled nitric oxide concentration in asthma: NIOX MINO, NIOX VERO and NObreath" (INAHTA record). The record does not state the recommendation, and the NICE guidance text was not read, so this article does not describe what it advises.

Demonstrated

Lourenço and Turner (2014), in Metabolites, reviewed breath analysis for disease. They wrote that there are no accepted standardised methods for collecting and analysing breath VOCs. They also noted that several studies used small control groups or small groups of people with the disease (review).

Jia and colleagues (2019), in Metabolites, reviewed 25 clinical breath studies of lung cancer. They concluded that no consistent, robust and validated VOC signature for lung cancer had been identified. The choice of control groups differed between studies and contributed to the inconsistent results (review). The review also notes that Tedlar sample bags are known to emit VOCs from the bag material, which poses a threat of contamination, and that lung cancer patients were often significantly older than the control subjects.

Erdrich and colleagues (2021), in Scientific Reports, studied hydrogen and methane breath tests, measured by gas chromatography. The team reviewed 388 consecutive tests, and 96 (24.7%) showed baseline gas of at least 10 ppm. In the analysed group, 64 people gave 69 breath tests. Participants followed a one-day low-residue diet and fasted overnight. Those with raised baseline gas received a chlorhexidine mouthwash and gave a follow-up sample. The authors concluded that a single mouthwash at baseline has a high probability of returning a false positive. The study had no control arm and took a single baseline sample from each participant. The first author discloses ownership of a company that runs hydrogen-methane breath testing (study).

Reported or speculative

Owlstone Medical's own news page, published on 11 December 2016, describes a feature in the trade publication Medtech Insight about its ReCIVA breath test (company page). The page refers to a 3,000-patient LuCID study at 21 sites in the UK, funded by the NHS at a cost of £1.1m. The study was collecting data to determine the test's sensitivity and specificity. It gave no accuracy figures. The company described ReCIVA as CE-marked, but also said it was still working on regulatory approval, including in the UK and Europe. These are company statements from 2016, not peer-reviewed results, and this article could not confirm the current regulatory status.

Limits and open problems

Confounding is a practical problem. Erdrich and colleagues found that a mouthwash could change a hydrogen-methane breath result (study). Lourenço and Turner report earlier work suggesting that when ambient levels of a compound exceed about 5% of exhaled levels, exhaled readings cannot be correlated with blood levels with confidence (review). Age is another factor, since lung cancer patients in the Jia review were often older than the control groups (review).

Oral sources add a further problem. Lourenço and Turner note that ammonia and ethanol are largely generated in the mouth, so mouth-exhaled air can carry signals that are not from the lungs or blood. They suggest nasal sampling for this reason. Sampling equipment can also add contamination, as the Jia review notes for sample bags.

Standardisation is still the central barrier. The ERS document is a step towards common methods, but it is a consensus document rather than proof that any single test works. Small study groups make it hard to judge accuracy. A small study does not show how a test would perform in routine use.

The gap between research and routine NHS use is wide. The sources opened for this article do not show any breath test for a named disease that is validated for routine diagnosis. Breath tests should not be read as a substitute for established clinical tests. The ERS document and the NICE guidance text were not read for this article, so their detailed contents are not described here.

What to watch next

Whether new studies use the ERS document, so results from different laboratories can be compared.
Whether breath tests report accuracy from independent test groups, with sample sizes given, rather than from the same group used to build the test.
Whether the LuCID study publishes results and whether the ReCIVA test gains regulatory approval (company page).
Whether new studies control for mouthwash, fasting and diet.