A small spot on a CT scan of the lungs can create a very big question. Is it cancer, or something harmless? A paper first published online on Sept. 23 in the journal MedComm asks whether a few breaths into a tube could help answer it.

The work comes from Griffith University in Brisbane, where Professor Chamindie Punyadeera's team tested two ways of analyzing exhaled breath in patients with suspicious lung nodules. The university announced the findings on Sept. 24, and the study was funded by the charity Tour de Cure.

The Nodule Problem

Doctors cannot always tell a cancerous nodule from a benign one on imaging alone. According to the paper, as summarized by Earth.com, up to 95% of nodules found on low-dose CT scans turn out not to be cancer. Infections and other noncancerous conditions can look the same on a scan.

That leaves doctors a hard choice. Ignoring a nodule risks missing an early cancer. Sampling it means a biopsy, which Griffith's Dr. Xi Zhang noted can be technically difficult when a lesion is small or hard to reach, and can cause complications such as a collapsed lung.

What the Griffith Team Tested

The researchers recruited adults whose nodules measured roughly 8 to 30 millimeters, all already referred for further investigation. These were not healthy people taking a screening test. Breath samples were collected before the final diagnosis was known, and patients were treated at the Royal Brisbane and Women's Hospital, according to the paper.

Two technologies were compared. The first is an electronic nose, or eNose, which uses 32 sensors to produce an overall pattern from the chemicals in breath without naming them. It analyzed 89 patients. The second, gas chromatography-mass spectrometry, separates and identifies specific volatile organic compounds, and analyzed 75 patients. That method found 11 chemical features that appeared more often in people with cancer, Earth.com reported.

Punyadeera said both approaches discriminated between malignant and benign nodules substantially better than models built only on standard clinical information.

The Number That Matters Is the Weak One

Here is the detail that deserves more attention than the press statement gave it. According to Earth.com's account of the paper, the breath tests looked very good at first. When the researchers tested the models more strictly, the results became much weaker.

That pattern is familiar in medical prediction. A model fitted to a small group will often fit that group too well, a problem known as overfitting. The stricter result is a better guide to how the test would behave on a new patient. The coverage I reviewed did not give the exact accuracy figures, so I cannot say how large the drop was, and readers should check the paper for the numbers.

My view: the honest headline is not that breath can find lung cancer, but that breath carries a signal worth testing in a bigger study.

Why Breath Is Hard to Read

Breath chemistry is messy. Smoking, other lung diseases and environmental exposures can all change which compounds people exhale, which makes it hard to tie a signal to cancer specifically. The study does not show that lung tumors themselves produce the 11 compounds, Earth.com noted. Some may reflect cell damage or outside exposures.

There are other limits. The patients came from one medical center, and the models were not tested on a separate group from another hospital. For the chromatography analysis, the team simplified results into present or absent, which discards information about how much of each chemical was there. And the two technologies were not combined into a single diagnostic test.

Fewer Biopsies, but Not Yet

Zhang said the breath test is not meant to replace biopsies. The aim is to help clinicians decide who needs invasive intervention most urgently, and perhaps to spare patients with benign nodules unnecessary procedures.

That goal echoes a debate Blogerroom covered in screening. A Danish-led reanalysis argued that mammogram overdiagnosis may be under 5%, shifting attention to false alarms and unneeded procedures as the harms that remain. The same logic applies here. A tool that reduces invasive follow-up could be valuable even if it never finds a single cancer earlier.

It also fits a broader push toward less invasive detection, like the Mayo Clinic model that reads routine records to flag pancreatic cancer risk. Both are early results that need outside validation.

What to Watch Next

The next step is a multicenter study, with breath collected and analyzed identically across several hospitals, and a model locked in before it is tested on fresh patients. Researchers also have to show the test adds information beyond what a doctor already gets from the scan and patient history.

Anyone with a lung nodule should follow the plan their own care team recommends. This breath test is not available in clinics, and this article is general information, not medical advice.