Purpose: Small cell lung cancer (SCLC) is an aggressive malignancy with poor survival. Existing prognostic models provide limited patient-level risk stratification and often overlook accessible clinical and imaging data. There remains a need for refined tools to support individualized prognostication in SCLC. Patients and Methods: We retrospectively analyzed 149 SCLC patients diagnosed between 2010 and 2014, representing a pre-immunotherapy era cohort. Demographics, CT scans, and treatment details were reviewed. Emphysema burden was quantified using an AI-based automated tool that segmented the lungs and calculated the percentage of low-attenuation areas (< − 950 Hounsfield Units). Multivariate Cox regression identified predictors for overall survival (OS), informing nomogram construction. Performance was measured by Harrell’s C-index and calibration plots. Validation included bootstrap resampling and a 3:1 data split. Decision curve analysis (DCA) evaluated clinical utility. Results: Age, emphysema, and treatment modality were independently associated with OS. The nomogram demonstrated excellent discrimination (C-index = 0.807; 95% confidence interval CI, 0.771– 0.843) and good calibration for 1- and 3-year survival. Internal validity was high, with a bootstrap-adjusted C-index of 0.805 (95% CI, 0.779– 0.845), and performance remained robust in the validation subset (C-index = 0.740; 95% CI, 0.668– 0.812). Stratification by nomogram-derived risk quartiles significantly differentiated survival (log-rank p < 0.001). DCA demonstrated superior net clinical benefit compared to treat-all or treat-none strategies across clinically relevant thresholds. Conclusion: This validated nomogram, based on three readily available variables, provides accurate survival predictions for patients with SCLC. It may assist clinicians in refining treatment strategies and enhancing shared decision-making. External validation is warranted. Keywords: cox proportional hazards models, calibration, decision support techniques, risk assessment, survival analysis
Kim et al. (Thu,) studied this question.