Nomogram of conditional survival probability in oldest old patients with colorectal cancer after surgical resection: a multicenter retrospective cohort study

Abstract Background Prognostic models for oldest old patients with colorectal cancer (CRC) are needed to inform treatment decision-making. We aimed to develop a conditional survival (CS) nomogram tailored specifically to this patient population. Methods We examined 594 patients from the Keio Surveillance, Epidemiology, and End Results database who were aged over 80 years and underwent curative surgery for CRC. Overall survival (OS) was analyzed using the Kaplan–Meier method. CS was calculated using the following formula: CS (y|x) = OS (y + x)/OS (x). Least absolute shrinkage and selection operator regression and multivariate Cox regression were used to identify risk factors. A CS nomogram was constructed based on the prognostic factors identified. The performance of the nomogram was assessed using the concordance index, calibration curves, and time-dependent area under curve (AUC). Internal validation was performed using the bootstrap method. Results CS analysis showed gradual improvement in real-time survival over time after surgery. Age, pT stage, pN stage, pM stage, R status, and CEA value were predictors of CS. The CS nomogram demonstrated a concordance index of 0.718 (95% confidence interval (CI), 0.625–0.810). Calibration curves and time-dependent AUCs provided evidence of the model’s stability and reliability. After internal validation via bootstrapping, the model still had a high discriminative ability (a concordance index of 0.716 [95% CI, 0.695–0.729]), and its predicted calibration curve and time-dependent AUC also demonstrated good performance. Conclusions The nomogram developed in this study accurately predicted postoperative CS in oldest old patients with CRC.