Personalized Conditional Survival Prediction for Patients with Early-Stage Non-small Cell Lung Cancer.

Abstract

Conditional survival (CS) estimates offer valuable prognostic insights for both clinicians and patients who have already survived a period following diagnosis. This study aimed to evaluate how survival outcomes evolve over time in patients with early-stage nonsmall cell lung cancer (NSCLC) and constructed a personalised CS-nomogram to provide dynamic prognostic predictions. Patients with early-stage NSCLC diagnosed between 2004 and 2015 were identified from the Surveillance Epidemiology End Results (SEER) registry. The Aalen-Johansen estimator was used to estimate cancer-specific survival (CSS). LASSO regression were used to identify key prognostic factors. Multivariable Cox regression was used to construct the CS-nomogram. The model performance was evaluated in terms of discrimination, calibration, and clinical utility. The findings indicated that the 5-year conditional survival rate exhibited an enhancement from an initial 66.1% to 87.6% among patients who had survived for a period of five years. The CS-nomogram demonstrated strong predictive performance, with concordance indices of 0.745 (95% CI: 0.742-0.748) and 0.751 (95% CI: 0.749-0.753) on the training and validation cohorts, respectively. The calibration curves exhibited a high degree of alignment with the ideal reference line, the AUC values for predictions from 1 to 10 years demonstrated stability, and decision curve analysis revealed a high net benefit, indicative of the model's superior performance. The CS-nomogram offered dynamic and individualised prognostic assessments to support the long-term prognosis management.