Abstract
BACKGROUND: The real-time prognosis of patients with primary ocular adnexal lymphoma (POAL) who have survived for several years remains uncertain. Our objective was to assess survival dynamics over time in POAL using conditional survival (CS) analysis and annual hazard functions. METHODS: We utilized data from the SEER database (2004–2019) to conduct CS analysis on POAL patients. A total of 1901 patients were included. CS was estimated using the Kaplan–Meier method, and the annual hazard rate (AHR) was calculated to examine changes in mortality risk over time. A predictive nomogram incorporating CS analysis was developed using a random survival forest (RSF) approach, with model performance evaluated through calibration, concordance index (C-index), receiver operating characteristic (ROC) curves, and decision curve analysis (DCA). RESULTS: CS analysis revealed a significant improvement in survival over time, with 10-year CS rates increasing from 69% at diagnosis to over 96% for patients surviving 1–9 years. The AHR peaked at 4.1% in the first year, declining progressively to 1.7% by year 10. The RSF algorithm identified five key prognostic factors—age, tumor site, histology, tumor stage, and radiotherapy—as important predictors. A CS-based nomogram was developed, and its reliability was further confirmed through calibration curves, C-index values, and ROC analysis, with AUC values exceeding 0.80 for 3-, 5-, and 10-year survival predictions. CONCLUSIONS: This study outlined the changes in CS prognosis and AHR for patients with POAL. Furthermore, we developed a CS-based nomogram using SEER data, which effectively integrated dynamic survival improvements and time-dependent risk factors. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s12672-026-04800-z.