Dynamic conditional survival nomogram for non-early-stage infiltrating ductal carcinoma based on SEER database

The prognosis of breast cancer varies by histologic subtype, with infiltrating ductal carcinoma (IDC) being the most prevalent, comprising 70% of infiltrating breast cancers. Long-term survival outcomes for IDC remain unclear, and some non-early-stage IDC patients exhibit better survival than initially expected. Dynamic survival probability offers a more accurate approach to life expectancy estimation compared to traditional methods such as Kaplan-Meier analysis and the tumor-node-metastasis (TNM) staging system–based nomogram (TNM nomogram) .This study aims to develop a conditional survival prediction model for non-early-stage infiltrating ductal carcinoma patients, with a particular focus on long-term survival assessment of female IDC patients at stage IIB and beyond (according to the AJCC classification), to support personalized treatment planning. Female IDC patients from the SEER database (2000–2022) were analyzed. Overall survival probabilities for additional y years after surviving x years were calculated using the Kaplan-Meier method. Lasso, subset regression, and multivariate Cox proportional hazards analysis were applied to identify significant predictors for constructing a conditional survival (CS) nomogram, where CS is defined as the probability of surviving y additional years given that a patient has already survived x years. The CS nomogram was compared to the TNM-nomogram using the concordance index, ROC curves, calibration plots and decision curve analysis (DCA). The conditional survival nomogram, incorporating 12 variables, predicted 3-, 5-, and 10-year survival, as well as 10-year conditional survival. It demonstrated higher predictive accuracy compared to the TNM-nomogram, reflected by a superior concordance index and improved ROC curves. Calibration plots further confirmed its greater precision, and decision curve analysis (DCA) highlighted better net benefits across varying risk thresholds. Unlike the traditional nomogram, which only provides unconditional survival probabilities at diagnosis, the CS-nomogram dynamically updates survival estimates according to the time already survived (CS(y|x) = OS(x + y)/OS(x)). The conditional survival nomogram provides more accurate and personalized survival predictions for non-early-stage IDC patients, surpassing traditional nomogram. This tool offers valuable guidance for clinical decision-making and improves the precision of survival estimations.