A retrospective analysis of survival prognostic factors and risk stratification in recurrent glioblastoma

Abstract

Background: Traditional Cox analysis identifies independent prognostic factors in recurrent glioblastoma (rGBM) but often overlooks their interrelationships. We aimed to develop a predictive nomogram integrating these multifaceted factors to establish a clinically applicable risk stratification model. Materials and methods: In a retrospective analysis of IDH-wildtype rGBM, we used Cox regression to evaluate prognostic factors including age, sex, Ki-67 index, Karnofsky Performance Status (KPS), time to first progression, number of recurrent lesions, tumor location, O6-methylguanine-DNA methyltransferase (MGMT) methylation status, and post-recurrence treatment. Significant predictors were used to construct a nomogram in R software, generating a risk stratification model by converting risk scores into categorical levels. The model underwent bootstrap validation. Results: Our cohort included 206 patients with a median overall survival (OS) of 8.3 (95% CI, 7.2–9.4) months. Multivariate analysis revealed KPS > 50 (p = 0.009; HR 0.61, 95% CI: 0.42–0.88), MGMT methylation (p = 0.033; HR 0.68, 95% CI: 0.48–0.97), time to first recurrence >12 months (p = 0.048; HR 0.69, 95% CI: 0.47–1), single lesion (p = 0.005; HR 0.63, 95% CI: 0.46–0.87), and post-recurrence therapy (surgery: HR 0.35, 95% CI: 0.21–0.59; targeted therapy/Tumor-treating fields (TTF)/re-irradiation: HR 0.5, 95% CI: 0.35–0.71; both p < 0.001) were favorable independent prognostic factors for OS. The nomogram-based risk stratification model successfully stratified patients into low-, medium-, and high-risk groups, yielding distinct OS outcomes (13.9 vs. 6.5 vs. 3.98 months; p < 0.0001). Its predictive performance was confirmed with area under curve (AUC)s of 0.76, 0.72, and 0.73 at 6, 12, and 24 months, respectively. Conclusions: We developed and internally validated a nomogram-based risk stratification model for rGBM. By integrating key clinical and molecular factors, this tool accurately predicts patient survival.