EXTH-38. ORTHOGONAL IN VIVO MODELS WERE THE SOLE PRECLINICAL PREDICTOR OF CLINICAL EFFICACY IN PHASE 1 TRIALS OF TARGETED THERAPIES FOR GLIOBLASTOMA: RESULTS OF A SYSTEMATIC REVIEW

Abstract

Abstract INTRODUCTION No drug has improved survival in recurrent glioblastoma despite encouraging activity preclinically. We undertook a systematic review of matched preclinical and Phase 1 trials (P1Ts) of targeted agents to investigate potential preclinical predictors of clinical efficacy. METHODS We identified all adult glioblastoma monotherapy P1Ts of targeted agents & preceding preclinical data published between 2006–2019 via structured searches of EMBASE/MEDLINE/PUBMED. For preclinical studies, data regarding in vitro models, in vivo models (species, implantation site, cell-line type) and efficacy (growth inhibition, regression rate, survival) were extracted. For P1T, response rate (RR) data were collected as absolute (%) and categorical (RR< 5% vs. RR≥ 5%) variables. Associations were compared by chi-square/Fisher’s exact test, Kruskal-Wallis or Mann-Whitney U testing as appropriate with 2-sided p-values. RESULTS We found 28 P1Ts with median RR 2.9% (range 0.0–33.3%) and mOS 8.0mo (range 4.6–13.0mo). Seven (25%) had ‘minimal’ published pre-clinical data (5 missing entirely; 2 in vitro only); 12 (43%) utilised one cell line in vivo (‘single model’ group); and 9 (32%) used 2+ biologically distinct in vivo models (‘orthogonal’ group). There was strong reliance on U87-based cell lines (14/21 (71%)) in the latter groups. None of the variables tested were associated with RR except for use of ‘orthogonal models’. Compared to the ‘orthogonal’ group, the P1T RR rate was lower in ‘single’ and ‘minimal’ groups (6.8% vs 1.2%, p= 0.043 and 6.8% vs 0.0%, p= 0.026 respectively). The frequency of P1T with a RR > 5% was also higher in the ‘orthogonal’ compared to the same two groups (78% vs 20%, p= 0.042 and 78% vs 17%, p= 0.041). CONCLUSION The availability of good quality pre-clinical data, especially the use of orthogonal models in vivo, was significantly associated with P1T response rates and warrants further investigation as a minimal threshold of evidence in future drug development.