Conditional survival patterns and individualized prognostic prediction in malignant peritoneal mesothelioma

Abstract

Malignant peritoneal mesothelioma (MPM) is a rare, aggressive cancer with limited treatment options and extremely poor survival outcomes. Due to the disease’s low incidence, large-scale cohort studies to clarify survival outcomes and the impact of treatments like chemotherapy are limited. This study aimed to use conditional survival (CS) analysis to assess survival trends and explore the survival benefits of chemotherapy in MPM patients. Using SEER data (2000–2019), CS analysis was applied to capture survival probabilities conditional on having survived specific durations after diagnosis. Additionally, we used machine learning algorithms, including Random Survival Forest and the least absolute shrinkage and selection operator regression, combined with Cox proportional hazards models, to develop and validate a nomogram model for CS prediction. Propensity score matching (PSM) was conducted to evaluate the effect of chemotherapy on survival. Among 1,549 MPM patients, CS rates improved with time survived post-diagnosis but plateaued later, highlighting the need for early intervention and follow-up. The CS-based nomogram model demonstrated strong predictive performance, with C-index values of 0.701 and 0.717 in the training and validation cohorts, respectively. The model was well-calibrated, with AUC values indicating consistent predictive accuracy. DCA analysis validated the clinical applicability of the model. PSM analysis showed that chemotherapy was associated with longer overall survival. CS analysis provides a time-updated complement to traditional survival estimates and may inform follow-up intensity and patient counseling. The CS-based nomogram offers a useful adjunct to current staging for individualized risk assessment in MPM. Moreover, our study highlighted the potential survival benefit of chemotherapy, though further research is needed to confirm these findings.