Construction of a survival prediction model and analysis of influencing factors for brain metastasis in non-small cell lung cancer based on multi‑parameter clinical data.

OBJECTIVE

To identify real-world prognostic indicators affecting patients with non-small cell lung cancer brain metastasis (NSCLC BM) and to develop a pragmatic nomogram for individualized survival prediction.

METHODS

We collected clinical data from 590 patients with NSCLC BM treated in the department of oncology at our hospital from January 2020 to December 2023, and randomly divided them into a training set (n = 413) and a validation set (n = 177) at a 7:3 ratio. Cox regression analysis was used to screen independent risk factors, and a nomogram prediction model was constructed. Model performance was evaluated using receiver operating characteristic(ROC) curves, C-index, calibration curves, decision curve analysis (DCA) and Kaplan-Meier (KM) survival analysis.

RESULTS

Multivariate Cox analysis revealed that Karnofsky Performance Score (KPS) score < 70(HR = 1.95, p < 0.001), number of BMs ≥ 3(HR = 1.40, p = 0.021),and the utilization of whole-brain radiotherapy (WBRT) as a real-world surrogate indicator for high disease burden (HR = 1.65, p = 0.001) were independent predictors of poor overall survival (OS). EGFR mutation status, though with borderline significance in the multivariate analysis (p = 0.052), was retained in the final nomogram due to its established prognostic role. The AUCs for the training set at 6, 12, 18, and 24 months were 0.805, 0.805, 0.846, and 0.836, respectively; the corresponding AUCs for the validation set were 0.745, 0.796, 0.816, and 0.809. Calibration curves showed good consistency between predicted and observed outcomes. DCA demonstrated that our model yields superior net benefit in clinical application. Patients were divided into low-, medium-, and high-risk groups based on nomogram scores. KM analysis showed significant survival differences among the three groups (p < 0.001).

CONCLUSION

The nomogram incorporating KPS score, number of BMs, radiotherapy modality, and epidermal growth factor receptor (EGFR) mutation status demonstrated good discrimination, calibration and clinical utility. Our model distinctively integrates real-world treatment triage (WBRT versus SRS/SRT) as a comprehensive prognostic dimension. This tool enables individualized risk stratification and may aid in early identification of high-risk patients, thereby guiding precise clinical treatment.