Epigenetic Age Acceleration Predicts Survival in Early-Stage Lung Cancer

Lung cancer remains the deadliest malignancy in the United States, and even when it is caught early, survival outcomes vary dramatically from patient to patient. Now, a team of researchers at Massachusetts General Hospital and Dana-Farber Cancer Institute reports that a blood-based measure of biological aging can sharpen predictions of who will live and who will die after an early-stage lung cancer diagnosis. The study, published in iScience, found that epigenetic age acceleration, a molecular gauge of how fast a person is aging beyond their chronological years, significantly improved survival prediction in patients with early-stage non-small cell lung cancer, with one particular epigenetic clock, GrimAge2, emerging as the most powerful prognostic tool of all.

The research focused on 75 patients diagnosed with stage I non-small cell lung cancer between 1998 and 2008, all of whom underwent surgical resection with curative intent. DNA methylation data from blood samples collected at diagnosis were used to calculate epigenetic age using four different epigenetic clocks: Horvath, Hannum, PhenoAge, and GrimAge2. Unlike chronological age, which simply counts years since birth, epigenetic age reflects cumulative molecular changes shaped by genetics, environmental exposures, and lifestyle factors. When epigenetic age exceeds chronological age, the deviation is termed epigenetic age acceleration, a biomarker previously linked to cancer incidence, cardiovascular disease, and all-cause mortality.

The results were striking. Short-term survivors, patients who died within roughly four years of diagnosis, showed consistently positive epigenetic age acceleration across all four clocks, while long-term survivors, those who lived well beyond a decade, showed consistently negative values, indicating decelerated biological aging. The largest gap appeared with GrimAge2: short-term survivors had a mean acceleration of 3.4 years, while long-term survivors showed a mean deceleration of 3.7 years, a difference that was highly statistically significant. This pattern suggests that the pace of biological aging, measured at the moment of diagnosis, carries meaningful information about survival that traditional clinical metrics do not capture.

Using Cox proportional hazards models adjusted for chronological age, sex, and smoking status, the researchers found that every year of epigenetic age acceleration increased mortality risk across all clocks. Hannum epigenetic age acceleration showed the strongest hazard ratio at 1.09, meaning a 9 percent increase in mortality risk per year of acceleration, while GrimAge2 delivered the best overall predictive discrimination, achieving a C-index of 0.73 compared to 0.69 for the base model. Notably, all epigenetic clocks except Horvath outperformed pack-years, the standard measure of cumulative smoking exposure, in predicting survival. GrimAge2 improved the C-index by 0.042 over the base model, compared to just 0.008 for pack-years alone.

The team then turned to random survival forests, a machine learning method for time-to-event data, to confirm these findings with non-parametric modeling. Refitting each model 500 times to obtain robust estimates, they found that GrimAge2 epigenetic age acceleration consistently delivered the lowest out-of-bag prediction error at 33.6 percent, compared to 38.4 percent for the base model. More remarkably, in the GrimAge2 model, epigenetic age acceleration ranked as the most important predictor in 97.6 percent of iterations, surpassing even smoking status, which had dominated all other models. This finding underscores that GrimAge2 captures biological signals relevant to lung cancer survival that extend beyond what smoking history alone can explain.

A critical question was whether GrimAge2’s prognostic value simply reflects cumulative tobacco exposure, since the clock includes a methylation-based proxy for pack-years. The sensitivity analyses answered this decisively. Self-reported pack-years explained only 9.6 percent of the variance in GrimAge2 epigenetic age acceleration, and adjustment for pack-years minimally attenuated the association, which remained statistically significant. GrimAge2 also enhanced model discrimination beyond pack-years alone, and remained the most important predictor even after adjustment. The researchers suggest this residual value reflects both superior measurement, avoiding recall bias, and individual biological variability that cumulative exposure alone cannot capture. GrimAge2 is also unique among clocks in including a surrogate for TIMP-1, a protein secreted by lung tumors and identified as a marker of poor prognosis in lung cancer.

Partial dependence plots revealed a non-linear relationship between GrimAge2 acceleration and mortality risk. For positive values, predicted mortality increased approximately linearly from 39 percent at zero years of acceleration to 50 percent at five years, before plateauing. For negative values, the effect was more modest, with risk decreasing to 34 percent at six years of deceleration. When stratified by smoking status, the shape of the relationship remained consistent across groups, but absolute mortality risk differed substantially. Current smokers showed markedly higher predicted mortality than former smokers at equivalent epigenetic age acceleration levels, such as 61 percent versus 45 percent at five years of acceleration. This finding suggests that ongoing tobacco exposure interacts with biological aging pathways in ways not fully captured by epigenetic age alone.

Kaplan-Meier survival curves stratified by GrimAge2 acceleration status revealed dramatic survival differences. Among the 16 patients with high acceleration, defined as more than five years of GrimAge2 epigenetic age acceleration, 94 percent died within five years of diagnosis. Among the 19 patients with high deceleration, defined as more than six years of deceleration, only 26 percent died within the same period. When the researchers combined acceleration status with smoking status, decelerated former smokers had the highest survival probability across all timepoints, while accelerated current smokers had the lowest, with only a 16 percent survival probability at three years post-diagnosis and no survivors beyond 15 years.

The study has important limitations. The extreme phenotype sampling design, which deliberately selected short-term and long-term survivors to maximize statistical power, produces optimistic estimates of model discrimination, and the researchers emphasize relative changes in model performance rather than absolute metrics. The modest sample size of 75 patients limited multivariable adjustment and predictive modeling, and the sample was diagnosed between 1998 and 2008, preceding contemporary advances such as robotic-assisted surgery and perioperative immunotherapy. The population also had limited representation of never smokers, a growing subset of lung cancer patients for whom epigenetic age biomarkers may be particularly valuable.

Despite these constraints, the findings support the potential utility of blood-based GrimAge2 as a readily accessible biomarker that improves survival prediction in early-stage non-small cell lung cancer beyond traditional risk factors. If validated in larger, contemporary cohorts, epigenetic age acceleration could eventually inform personalized risk stratification, helping clinicians identify which patients with seemingly identical early-stage disease face vastly different mortality trajectories, and potentially guiding more informed decisions about adjuvant therapy and follow-up intensity in the quest to close the survival gap that separates lung cancer from other common malignancies.

Subject of Research: Epigenetic age acceleration as a prognostic biomarker for overall survival in early-stage non-small cell lung cancer

Article Title: Epigenetic age acceleration enhances prediction of overall survival in early-stage non-small cell lung cancer

Article References: Gandara, I., Du, M., Li, Y., Su, L., Kothari, J., Lanuti, M., Gainor, J. F., & Christiani, D. C. (2026). Epigenetic age acceleration enhances prediction of overall survival in early-stage non-small cell lung cancer. iScience, 29(11), Article 117843. https://doi.org/10.1016/j.isci.2026.117843

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117843

Keywords: epigenetic age, lung cancer, GrimAge2, DNA methylation, biological aging, survival prediction, Epigenetic, acceleration, enhances, prediction, overall, survival