Multimodal modeling of CCI, CCI, FaceAge, thymic health, and OS among older adults with early-stage NSCLC undergoing SBRT.

Abstract

e20006 Background: Lung cancer disproportionately affects older adults, who have diverse care needs and inferior clinical outcomes. While pre-treatment geriatric assessment can address these challenges, existing tools remain limited. AI-based metrics of biological aging and thymic health have shown prognostic utility for toxicity, locoregional recurrence, and survival, but have not been explored together or in geriatric populations. We thus sought to assess whether a multimodal model incorporating Charlson Comorbidity Index (CCI), FaceAge, and thymic health was associated with all-cause mortality in older adults undergoing SBRT for early-stage NSCLC. Methods: We retrospectively reviewed the records of 708 patients ≥65 years old with stage I-II NSCLC treated with SBRT at our institution between June 1, 2009, to March 31, 2023. We abstracted demographics, functional status (Eastern Cooperative Oncology Group [ECOG] score), CCI, oncologic history, SBRT details, and time-to-death. Using validated deep learning systems, biological FaceAge was calculated from SBRT simulation photographs, and thymic health scores (range 0-1, with higher scores indicating better health) were calculated from pre-treatment CT scans. We examined associations between a model incorporating CCI, FaceAge, and thymic health and all-cause mortality, using Cox models adjusted for age, sex, stage, ECOG, and covariates significant to p<.10 in univariate analyses. Results: Overall, patients (median age 76.2 years; 60.7% female; median stage IA) had high multimorbidity (median CCI 7; IQR 6-8), elevated biological FaceAge (median 2.6 years above chronological age), and low thymic health (median 0.14; IQR 0.07-0.23). In the adjusted Cox regression model, worse performance on all 3 geriatric vulnerability measures was independently associated with higher all-cause mortality (hazard ratio [HR] CCI =1.09, 95% confidence interval [CI] CCI 1.01-1.19, P<.044; HR FaceAge =1.04, 95% CI FaceAge 1.02-1.06, P=.003; HR thymic =0.21, CI thymic 0.06-0.79, P=0.021). Conclusions: The use of an AI-based multimodal tool using CCI, FaceAge, and thymic health may offer prognostic information in older adults with NSCLC. Associations between geriatric measures and all-cause mortality. Characteristic Univariate AnalysisHR/OR (95% CI) Univariate AnalysisP-value Multivariate AnalysisHR/OR (95% CI) Multivariate AnalysisP-value Female sex 0.63 (0.50-0.78) <0.001 0.61 (0.46-0.80) <0.001 Age 1.03 (1.01-1.04) <0.001 0.99 (0.97-1.01) 0.564 Stage IB 0.76 (0.47-1.22) 0.259 0.67 (0.39-1.14) 0.141 Stage II 2.41 (1.38-4.20) 0.002 1.97 (1.09-3.57) 0.024 ECOG 1 1.90 (1.32-2.74) 0.001 1.87 (1.18-2.95) 0.007 ECOG 2+ 4.38 (3.04-6.31) <0.001 4.39 (2.76-6.97) <0.001 CCI 1.19 (1.13-1.26) <0.001 1.08 (1.02-1.17) <0.001 FaceAge 1.03 (1.02-1.05) <0.001 1.04 (1.02-1.06) 0.002 Thymic Health 0.19 (0.06-0.59) 0.004 0.30 (0.10-0.92) 0.036