Blood Metabolites Linked to Lower Lung Cancer Risk in Massive Two-Cohort Study

Every time your blood is drawn, it carries a chemical diary of your body’s inner workings — sugars, fatty acids, amino acids, albumin, and hundreds of lipoprotein subfractions shuttling cholesterol through the circulation. A new prospective study published in BMC Cancer has now asked whether any of those circulating chemicals, measured years before diagnosis, predict who goes on to develop the four most common cancers in the world: breast, prostate, lung, and colorectal cancer. The answer, drawn from nearly half a million participants across two major population cohorts, is nuanced but intriguing: several lipoprotein subfractions and metabolites were associated with a reduced incidence of lung cancer, while no specific lipid or metabolite profile emerged for the other three cancer types.

The research team, led by Marion Denos and Abhibhav Sharma of the HUNT Center for Molecular and Clinical Epidemiology at the Norwegian University of Science and Technology, together with collaborators at the University of Bristol’s MRC Integrative Epidemiology Unit, quantified 132 lipoprotein subfractions and metabolites using nuclear magnetic resonance spectroscopy. This technique exploits the magnetic properties of hydrogen nuclei to fingerprint the chemical composition of a blood sample without destroying it, allowing researchers to measure cholesterol content within very low-density lipoprotein (VLDL), low-density lipoprotein (LDL), and high-density lipoprotein (HDL) particles of different sizes, along with fatty acids, albumin, and other small molecules. The participants came from the Trøndelag Health Study (HUNT3) in Norway, with 16,877 individuals, and the UK Biobank, with 480,840 individuals — a combined dataset large enough to detect even modest associations between blood chemistry and future cancer diagnoses.

During a mean follow-up of 14 years, 549 women and 741 men in HUNT3 and 12,112 women and 14,749 men in the UK Biobank developed one of the four cancers under study. The researchers used Cox regression, a statistical framework that models the hazard of an event over time while adjusting for confounders such as age, sex, body mass index, and smoking, to estimate the association between each metabolite and each cancer type. The results were expressed as hazard ratios per one-standard-deviation increment in metabolite level — a technical way of asking how much the risk of cancer changes for a typical step up in a given blood chemical.

The most striking findings concerned lung cancer. In the UK Biobank, higher levels of several subfractions of VLDL and LDL were associated with a decreased incidence of lung cancer. Cholesterol in large LDL particles, for example, showed a hazard ratio of 0.84 per 1-SD increment (95% confidence interval 0.81–0.87), meaning that each typical step up in this measure corresponded to roughly a 16 percent lower risk of being diagnosed with lung cancer during follow-up. Similar inverse associations were observed for albumin and some fatty acids. The consistency of the direction of effect across multiple related lipoprotein measures — rather than a single isolated signal — suggests the pattern is not a statistical fluke, although the authors are careful to note that association does not imply causation.

For prostate cancer, the second cancer type to show associations, higher levels of several lipoprotein subfractions were likewise linked to a decreased incidence in the UK Biobank. Breast and colorectal cancer, by contrast, showed no clear associations with any of the 132 metabolites measured. This asymmetry is itself informative: it suggests that the metabolic signatures detectable in routine blood samples may be more relevant to lung and prostate carcinogenesis than to breast or colorectal tumors, or that the associations for the latter are too weak or too heterogeneous to be captured even at this scale.

Perhaps the most sobering result, however, is what did not replicate. In HUNT3, the Norwegian cohort, no clear associations were observed between any metabolite and any of the four cancers. The authors point out that the observed associations showed the same direction of effect across cohorts, despite not reaching statistical significance in HUNT3. This discrepancy has several plausible technical explanations. HUNT3 is roughly 30 times smaller than the UK Biobank, so its statistical power to detect hazard ratios of 0.84 or similar modest effect sizes is limited. Differences in age structure, follow-up duration, smoking prevalence, and background cancer incidence between the Norwegian and British populations may also attenuate or shift the signals. And because the study measured metabolites at a single time point, reverse causation — the possibility that preclinical disease alters blood chemistry rather than the other way around — cannot be excluded, although the prospective design, with metabolites measured years before diagnosis, makes this less likely than in cross-sectional studies.

The biological interpretation of the inverse lung cancer associations is equally complex. Lipoproteins are best known for their role in cardiovascular disease, where higher LDL cholesterol is harmful. In cancer, the picture may differ. Lipoprotein particles transport lipids that cells need for membrane synthesis and signaling, and tumor cells are voracious lipid consumers. Some researchers have proposed that the lipid transport capacity of the circulation could influence tumor growth, while others note that lipoproteins also carry antioxidants and modulate immune function. Albumin, which showed similar inverse associations with lung cancer, binds fatty acids, drugs, and countless other molecules, and its concentration reflects nutritional status, liver function, and systemic inflammation. Disentangling which of these mechanisms — if any — drives the observed associations will require studies beyond observational epidemiology.

This is where the study’s design points to the next frontier. Because the same 132 metabolites can be measured cheaply and rapidly by NMR spectroscopy in routine clinical samples, the findings highlight promising directions for future research into metabolic biomarkers and their potential role in cancer prevention. If the inverse associations with lung cancer reflect a genuine protective metabolic state, metabolite profiles could eventually complement existing risk models — which currently rely heavily on age, smoking history, and family history — to identify individuals who might benefit from intensified screening such as low-dose CT. But the authors are explicit that their results do not yet support clinical action: no specific lipid or metabolite profile was associated with the most common cancers overall, and the lung cancer signals, while consistent in direction, require replication and mechanistic validation before they can inform prevention strategies.

The study also illustrates the power and the pitfalls of modern biobank-scale epidemiology. With 480,840 participants in the UK Biobank alone, the statistical machinery can detect hazard ratios far below the threshold of clinical interest, and the sheer number of metabolite-cancer pairs tested — 132 metabolites across four cancer types in two cohorts — raises the specter of multiple testing, where some associations will appear significant by chance alone. The researchers addressed this by examining patterns across related measures rather than isolated signals, and by requiring consistency of direction across cohorts. The fact that the HUNT3 signals pointed the same way, even when not statistically significant, offers a form of directional replication that is more meaningful than a single p-value, but it also underscores how much larger, longer, and more diverse studies will be needed to settle the question.

For now, the take-home message is one of cautious optimism. Blood metabolites, measured by a technique that is fast, inexpensive, and increasingly available, show reproducible patterns of association with lung cancer incidence at biobank scale — patterns that could, with further research, illuminate the metabolic biology of carcinogenesis and point toward new prevention targets. The absence of associations for breast, prostate, and colorectal cancer in this analysis does not rule out metabolic involvement in those diseases; it may simply mean the relevant signals are weaker, later-emerging, or masked by the confounders that observational studies must statistically adjust for. As the authors conclude, their results highlight promising directions for future research into metabolic biomarkers and their potential role in cancer prevention — a reminder that the chemical diary written in our blood may hold clues to cancer risk that we are only beginning to read.

Subject of Research: Cancer

Article Title: Associations between circulating metabolites and incidence of breast, prostate, lung and colorectal cancers in the HUNT Study and UK Biobank

Article References: Denos, M., Sharma, A., Åsvold, B. O., Arham, A. G. A., Smith, G. D., Timpson, N. J., Løset, M., Brumpton, B. M., Debik, J., & Giskeødegård, G. F. (2026). Associations between circulating metabolites and incidence of breast, prostate, lung and colorectal cancers in the HUNT Study and UK Biobank. BMC Cancer. https://doi.org/10.1186/s12885-026-16957-5

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16957-5

Keywords: metabolomics, lipoproteins, lung cancer, cancer prevention, NMR spectroscopy, UK Biobank, HUNT Study, biomarkers, epidemiology, BMC Cancer, Associations, between