Development and validation of a nomogram for predicting overall survival in small cell lung cancer: a multicenter analysis integrating multimodal therapy and molecular pathology.

BACKGROUND

Small cell lung cancer (SCLC) is an aggressive malignancy with heterogeneous survival outcomes. Conventional staging systems do not fully capture the influence of multimodal treatment strategies or molecular pathological characteristics on prognosis. However, prognostic models integrating treatment modalities with molecular pathology remain limited in SCLC. This study aimed to evaluate the survival impact of multimodal treatments and develop a nomogram incorporating clinical and molecular features for individualized prognostic prediction.

METHODS

This retrospective multicenter study included patients diagnosed with SCLC between 2020 and 2024 at three medical centers. Patients from two centers were randomly divided into a derivation cohort (70%) and an internal validation cohort (30%), while patients from the third center served as an independent external validation cohort. Overall survival (OS) was the primary endpoint. Candidate predictors were selected using the Boruta algorithm and the least absolute shrinkage and selection operator (LASSO) regression, and a multivariable Cox proportional hazards model was used to construct a nomogram for predicting 1-year OS. Model performance was evaluated using time-dependent receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA), and was further compared with the tumor, node, metastasis (TNM) staging system and Veterans Administration Lung Study Group (VALG) staging system.

RESULTS

A total of 728 patients were included and assigned to the derivation cohort (n=468), internal validation cohort (n=200), and external validation cohort (n=60). A nomogram incorporating pathological score, therapy, clinical T stage, clinical N stage, and TNM clinical stage was developed. The nomogram showed better discrimination than the TNM and VALG staging systems in the derivation cohort, with an area under the ROC curve (AUC) of 0.801 and a 95% confidence interval (CI) of 0.758-0.844, versus 0.664 (95% CI: 0.609-0.719) and 0.603 (95% CI: 0.051-0.655), respectively; similar results were observed in the internal validation cohort [0.800 (95% CI: 0.734-0.866) vs. 0.649 (95% CI: 0.560-0.738) and 0.569 (95% CI: 0.488-0.650)] and the external validation cohort [0.776 (95% CI: 0.646-0.906) vs. 0.611 (95% CI: 0.460-0.762) and 0.593 (95% CI: 0.462-0.725)]. Calibration curves showed good agreement between predicted and observed 1-year OS, and DCA demonstrated favorable clinical utility across all cohorts.

CONCLUSIONS

Multimodal treatment strategies were associated with survival in advanced SCLC. The proposed nomogram improved individualized 1-year survival prediction beyond conventional staging systems in SCLC. By integrating treatment-related and pathological information with clinical staging variables, this approach may support risk stratification, treatment intensity, and inform follow-up planning in clinical practice.