Beyond traditional survival estimates: conditional survival and individualized prognosis in primary breast lymphoma

Primary breast lymphoma (PBL) is a rare extranodal non-Hodgkin lymphoma with heterogeneous outcomes and limited prognostic guidance. Conventional survival estimates do not account for changes in risk over time. Conditional survival (CS) provides dynamic, time-updated prognostic information. This study aimed to investigate CS patterns in PBL and to develop a validated CS-based nomogram for individualized survival prediction.

Female patients diagnosed with PBL between 2004 and 2021 were identified from the Surveillance, Epidemiology, and End Results (SEER) database. Demographic, clinical, and treatment variables were extracted. CS probabilities and annual hazard rates (AHRs) were calculated to describe survival dynamics. Patients were randomly divided into training (70%) and validation (30%) cohorts. Prognostic factors were selected using least absolute shrinkage and selection operator (LASSO) regression and Cox analysis. A CS-integrated nomogram was constructed, with performance assessed by calibration curves, time-dependent receiver operating characteristic (ROC) analysis, and decision curve analysis (DCA).

A total of 1,552 patients were included. The 3-, 5-, and 10-year overall survival (OS) rates were 85%, 77%, and 58%, respectively. CS analysis showed improved long-term survival with elapsed time, with 10-year CS increasing from 63% for patients surviving one year to 95% for those surviving nine years. The AHR peaked in the first year (8.05%) and declined thereafter. LASSO identified age, histology, radiotherapy (RT), and marital status as significant prognostic factors. Compared with the traditional Cox-based selection, this feature set achieved comparable predictive accuracy while simplifying the model. The CS-nomogram, incorporating these variables, demonstrated good calibration and discrimination, with area under the curves (AUCs) of 0.787-0.814 in the training set and 0.741-0.762 in the validation set, and showed clinical utility by DCA.

This population-based study provided the first CS analysis of PBL and introduced a validated nomogram for dynamic, individualized prognosis. These findings highlighted the early high-risk period after diagnosis and supported tailored follow-up and treatment strategies for this rare malignancy.