This randomized, open-label, active-controlled, dose-finding phase IIb study evaluated the efficacy and safety of trabedersen (AP 12009) administered intratumorally by convection-enhanced delivery compared with standard chemotherapy in patients with recurrent/refractory high-grade glioma. One hundred and forty-five patients with central reference histopathology of recurrent/refractory glioblastoma multiforme (GBM) or anaplastic astrocytoma (AA) were randomly assigned to receive trabedersen at doses of 10 or 80 µM or standard chemotherapy (temozolomide or procarbazine/lomustine/vincristine). Primary endpoint was 6-month tumor control rate, and secondary endpoints included response at further timepoints, survival, and safety. Six-month tumor control rates were not significantly different in the entire study population (AA and GBM). Prespecified AA subgroup analysis showed a significant benefit regarding the 14-month tumor control rate for 10 µM trabedersen vs chemotherapy (p= .0032). The 2-year survival rate had a trend for superiority for 10 µM trabedersen vs chemotherapy (p = .10). Median survival for 10 µM trabedersen was 39.1 months compared with 35.2 months for 80 µM trabedersen and 21.7 months for chemotherapy (not significant). In GBM patients, response and survival results were comparable among the 3 arms. Exploratory analysis on GBM patients aged ≤55 years with Karnofsky performance status >80% at baseline indicated a 3-fold survival at 2 and 3 years for 10 µM trabedersen vs chemotherapy. The frequency of patients with related or possibly drug-related adverse events was higher with standard chemotherapy (64%) than with 80 µM trabedersen (43%) and 10 µM trabedersen (27%). Superior efficacy and safety for 10 µM trabedersen over 80 µM trabedersen and chemotherapy and positive risk-benefit assessment suggest it as the optimal dose for further clinical development in high-grade glioma.