Notably, nearly a quarter of patients have no history of smoking, highlighting the need for broader screening approaches beyond traditional risk groups.

Currently, screening programs primarily focus on individuals aged 50–70 with a history of smoking, leaving many at-risk people without access to early detection tools.

Dr Weeden said the findings could pave the way for more inclusive prevention strategies.

“Figuring out who is at risk of developing cancer is the holy grail of cancer prevention medicine,” she said.

“This proof-of-concept study offers new ideas that could fill this critical knowledge gap.”

Machine learning identifies 14-protein signature

The researchers analysed blood plasma samples from more than 48,000 UK Biobank participants, combining the data with cancer registry records. Using machine learning, they identified 14 key proteins that could predict a lung cancer diagnosis within five years.

The model also incorporated age, smoking status, and prior lung disease history.

The signature was validated across eight international datasets and was consistently elevated in individuals who later developed lung cancer, including non-smokers.

Importantly, researchers found that the signature does not originate from tumours themselves, but instead reflects an underlying inflammatory environment in the lungs that precedes cancer development.

Inflammation as a key driver

The study builds on earlier findings that air pollution can trigger inflammation in the lungs, activating dormant mutated cells and increasing cancer risk.

Researchers observed that exposure to pollutants boosts both inflammatory signals and a population of vulnerable “KAC cells,” which can progress toward cancer when mutations are present.

The team also discovered that diverse lung cell types converge into this intermediate state before becoming cancerous, suggesting a shared biological pathway in lung adenocarcinoma.

Dr Weeden said: “Very different lung cell types all seemed to converge on the same intermediate state before becoming cancerous, suggesting a common vulnerability.”

Potential for preventive treatment

The study further found that blocking the inflammatory molecule IL-1β in mice exposed to air pollution reduced the formation of precancerous cells and slowed early tumour development.

This raises the possibility that anti-inflammatory drugs could one day be used to prevent lung cancer in high-risk individuals identified through the blood signature.

Reanalysis of data from the CANTOS clinical trial also showed that patients with high levels of the 14-protein signature experienced a significantly reduced lung cancer risk when treated with the IL-1β inhibitor canakinumab.

In this subgroup, the number of patients needed to treat to prevent one case of lung cancer was just 55—comparable to established preventive therapies such as statins.

A new era of risk-based prevention

Experts say the findings support a shift toward personalised, biology-based screening, rather than relying solely on age or smoking history.

According to researchers, the blood signature could help identify a pre-disease inflammatory state shared across multiple lung conditions, including COPD and idiopathic pulmonary fibrosis.

Professor Charlie Swanton of the Francis Crick Institute said the study reveals a critical window for intervention:

“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.”

He added that future applications may extend beyond lung cancer to other inflammation-driven diseases.

Outlook

While still at an early stage, the findings represent a significant step toward predicting and potentially preventing lung cancer before it develops, marking a shift from reactive treatment to proactive intervention in cancer medicine.