Overall survival across registry-defined biopsy and registry-defined STR categories in glioblastoma: a population-based matched and weighted SEER study

Abstract

In glioblastoma (GBM), biopsy and subtotal resection are clinically distinct concepts, but registry datasets may not distinguish them with sufficient neurosurgical precision. We therefore evaluated whether SEER registry-defined biopsy and registry-defined STR categories were associated with overall survival (OS), while explicitly recognizing that these codes do not capture radiographic residual tumor volume, operative intent, or validated extent-of-resection definitions. Using Surveillance, Epidemiology, and End Results (SEER) data (2004–2019), GBM cases were identified (ICD-O-3: 9440, 9441, 9442). Surgery was defined by SEER code 20 (biopsy) and code 21 (registry-defined STR). The primary analysis excluded patients with survival of ≤ 1 month; sensitivity analyses included them. Propensity score matching (PSM) and stabilized inverse probability of treatment weighting (sIPTW) were used to reduce measured baseline imbalance. Survival was assessed with Kaplan–Meier analysis and Cox proportional hazards models. Of 10,359 patients, 4,774 were coded as biopsy and 5,585 as registry-defined STR. In the unadjusted cohort, registry-defined STR was associated with better OS in univariable analysis (HR 0.951, 95% CI 0.912–0.991; P = 0.017), but this association was not significant after multivariable adjustment (HR 1.010, 95% CI 0.968–1.054; P = 0.645). Similar non-significant findings were observed in the PSM cohort (HR 1.029, 95% CI 0.978–1.083; P = 0.268) and the sIPTW-weighted cohort (HR 1.008, 95% CI 0.962–1.056; P = 0.726). Sensitivity analyses yielded consistent results. Within SEER registry-coded surgical categories, registry-defined STR was not associated with an adjusted OS difference versus biopsy in patients with GBM. However, because SEER cannot reliably distinguish these procedures as clinically validated surgical entities, the findings should be interpreted primarily as a registry-coding analysis with limited surgical interpretability, rather than as evidence that biopsy and registry-defined STR are equivalent strategies in practice. Future studies with postoperative imaging-based extent-of-resection measures and richer clinical, molecular, radiographic, and functional data are needed.