Blood proteins flag lung cancer risk years before diagnosis

Clinical takeaway: A blood-based signature may eventually help identify high-risk patients for lung cancer prevention that current screening misses. This is not yet a validated clinical test.

Lung cancer screening still hinges on age and smoking history, which means people who have never smoked and who breathe polluted air fall outside it even when their risk is real. If we could catch that risk earlier and in a wider pool of patients, the window for prevention would open further. Researchers at the Francis Crick Institute and University College London went looking for that early signal in blood proteins.

Applying machine learning to plasma proteomics from more than 48,000 UK Biobank participants, the team landed on 14 proteins that, alongside age, smoking status, pack-years, and prior COPD, predicted lung cancer a median of more than five years out. The signature beat established risk models with the biggest edge two to four years before diagnosis. It was replicated across eight cohorts from the UK, US, Iceland, and China, including one made up mostly of never-smokers.

Notably, the signal does not seem to come from the tumor. It did not track with stage and did not drop after surgical resection, which points away from a tumor-shed marker. Instead, the proteins trace back to lung epithelial and myeloid cells and capture an inflamed, pre-cancer state: the same signature ran high in people who later developed idiopathic pulmonary fibrosis or COPD. It also climbed with current smoking and particulate exposure.

The researchers also took a fresh look at CANTOS, a trial that showed the IL-1β blocker canakinumab cut lung cancer incidence but at a treated population figure that was too high to justify. Among 4,651 participants, canakinumab nearly halved lung cancer in those with a high baseline signature and did nothing for those with a low one. Selecting on the signature pulled the number needed to treat from roughly 1,500 down to 55, in the same range as statins for cardiovascular prevention.

The mechanism behind the signal holds together. In mouse models, different lung epithelial lineages carrying EGFR mutations all funneled into a single transitional alveolar cell state primed to seed tumors. Air pollution expanded that cell pool and pushed up the signature proteins through IL-1β, and blocking IL-1β reined both in, supporting a real window where intervention could work before cancer establishes itself.

The CANTOS analysis was retrospective, and because CANTOS was a cardiovascular trial rather than a lung cancer one, the formal test of whether the signature predicted canakinumab response fell short of significance, even though response clearly differed between the high- and low-signature groups. The UK Biobank work had no baseline imaging, and the assays measure relative, rather than absolute, protein levels so there is no actionable cutoff yet.

"Finding a signal for an inflammatory state in the lungs has given us insight into this window of opportunity, when preventative treatment could work best. This work supports a relatively new idea in the field, that some common age-related diseases, causing a high burden of disease in the community, share a common, presymptomatic state of inflammation,” said Charlie Swanton, Clinical Research Director and Principal Group Leader at the Crick and Professor in Cancer at UCL. “We think the signature could in the future help to predict and help prevent lung cancer and other lung diseases."

Source: Pandya T, et al. Cell. 2026 Jun 25. Plasma signals of lung tumor promotion for molecular cancer prevention