Multimodal Assessment of Biological Age Following Radiation Therapy Among Patients With Early-Stage NSCLC

Accurately estimating physiological fitness and life expectancy is challenging in older patients with early-stage non-small cell lung cancer (NSCLC) considered for stereotactic body radiotherapy (SBRT). Noninvasive biological age metrics may improve assessment beyond chronological age.

To evaluate whether a photography-based age estimating algorithm (face age) and spirometry-based estimated age (lung age) are associated with overall survival and early mortality.

This retrospective cohort study included patients aged 60 years and older with early-stage NSCLC treated with definitive SBRT from June 2009 to March 2023. Participants were recruited from 6 radiation oncology clinics affiliated with a single academic cancer center and had a median follow-up of 44 months. Pretreatment identification photographs were available for face age estimation in all patients; baseline pulmonary function tests were available for a subset and used to compute lung age.

Face age used a validated deep learning model to estimate age from pre-SBRT photographs; lung age was derived from spirometry using validated equations.

Primary outcomes were overall survival and 2-year mortality. Multivariate Cox models adjusted for clinical covariates.

Among 670 patients (median [range] age, 77 [60-98] years; 406 [61%] female), the median (range) face age was 79 (range, 47-98) years. Overall, 113 participants (17%) vs 149 (22%) were 85 years or older by chronological age vs face age, respectively. The median overall survival was 47 (95% CI, 42-54) months, and 161 (24%) died within 2 years. On multivariate analysis adjusting for sex, Eastern Cooperative Oncology Group performance status, cancer stage, smoking pack-years, and histology, older face age was associated with worse overall survival (adjusted hazard ratio [HR] per decade, 1.39; 95% CI, 1.13-1.71; P = .002), whereas chronological age was not. Older face age and face age of 85 years or older were associated with increased 2-year mortality (adjusted HR per decade, 1.35; 95% CI, 1.03-1.78; P = .03; age ≥85 years: adjusted HR, 1.59; 95% CI, 1.06-2.40; P = .03), whereas chronological age per decade and age of 85 years or older were not. In the subset of 477 participants with spirometry data, median lung age was 98 (range, 22-139) years and showed minimal correlation with face age (r = 0.07). Face age remained independently associated with survival in sensitivity analyses adjusting for lung age.

In this cohort study of older patients with early-stage NSCLC receiving SBRT, face age was an independent biomarker associated with survival and early mortality and complemented lung age. These noninvasive measures, obtainable from routine photographs and spirometry, may aid individualized risk stratification and treatment planning.