Association between rodent xenograft models and phase II outcomes in glioblastoma: a systematic review

Glioblastoma (GBM) has a median overall survival (mOS) under 15 months and few effective treatments. Agents with promising activity in rodent xenograft models rarely succeed in clinical trials. We evaluated the association between preclinical efficacy in rodent models and outcomes in phase II monotherapy trials for GBM. We systematically reviewed phase II trials in recurrent GBM registered on ClinicalTrials.gov from January 1997 through December 2023; the protocol was not preregistered. Monotherapy trials, including those adding an investigational agent to a standard-of-care backbone, that reported mOS, objective response rate (ORR), or 6-month progression-free survival (PFS-6) were matched to in vivo preclinical studies identified via PubMed and trial bibliographies. Combination, vaccine-based, and cell-based trials were excluded. Associations between murine survival benefit or tumor volume reduction and clinical outcomes were assessed using Spearman rank correlation. Of 660 screened trials, 27 met inclusion criteria and had matched preclinical data. Thirteen agents had murine survival benefit paired with mOS, 15 with ORR, and 14 had tumor volume data paired with each outcome. No significant correlations were observed between murine survival benefit and mOS (Spearman ρ = 0.14, 95% CI [-0.44, 0.65], p = 0.64) or ORR (ρ = 0.42, 95% CI [-0.15, 0.77], p = 0.12), or between tumor volume reduction and mOS (ρ = 0.01, 95% CI [-0.52, 0.54], p = 0.98) or ORR (ρ = 0.28, 95% CI [-0.30, 0.71], p = 0.33). Upper confidence limits excluded very strong, but not moderate, correlations. Of 133 phase II trials evaluating distinct agents, 94 (71%) could not be included because rodent GBM efficacy data or trial results were unavailable. This study addresses a critical gap in the neuro-oncology literature by quantitatively comparing preclinical xenograft findings against phase II trial outcomes across GBM. We did not detect a significant association between commonly reported xenograft efficacy endpoints and phase II outcomes. Very strong concordance appears unlikely, although the small number of matched agents limits statistical power and moderate associations cannot be excluded. These findings underscore the need for improved preclinical models, greater reporting transparency, and higher standards for trial initiation in GBM.