Construction and validation of a prognostic risk score model for malignant mesothelioma.

PURPOSE: Malignant mesothelioma is a rare, aggressive cancer. Although the World Health Organization (WHO) histologic subtypes–epithelioid (best prognosis), biphasic, and sarcomatoid (worst prognosis)—provide baseline stratification, architectural and molecular features (e.g., aneuploidy, immune infiltration) introduce heterogeneity not fully captured by histotype alone. METHODS: Using The Cancer Genome Atlas (TCGA) data and an independent validation cohort, we found that TNM stage was not independently prognostic after multivariable adjustment. We therefore developed a clinicomolecular model integrating transcriptomic and clinicopathological variables. Differential expression analysis of TCGA tumors identified 1,055 genes associated with poor prognosis (FDR < 0.05, |log2FC| > 1). Least absolute shrinkage and selection operator (LASSO) followed by multivariate Cox regression yielded five independent predictors: pathological subtype, receipt of pharmaceutical (systemic) therapy, and tumor expression of SPAG5, CORO1C, and SGCE. The final score was: RiskScore = 1.590 × (pathological type) − 2.258 × (pharmaceutical therapy) + 3.048 × SPAG5 + 1.652 × CORO1C + 0.908 × SGCE. RESULTS: In the TCGA cohort, high-risk patients had significantly shorter overall survival (log-rank p < 0.001), and time-dependent ROC AUCs exceeded 0.70 at 1–5 years. In the validation cohort, the RiskScore significantly stratified overall survival (log-rank p < 0.001) and demonstrated superior predictive accuracy over both TNM staging and the established European Organization for Research and Treatment of Cancer (EORTC) prognostic score. CONCLUSION: This clinicomolecular model provides prognostic discrimination superior to TNM stage and EORTC score, offering a robust tool for risk-stratified management of mesothelioma. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s12672-026-04597-x.