Convection-enhanced delivery (CED) of cintredekin besudotox (CB) was compared with Gliadel wafers (GW) in adult patients with glioblastoma multiforme (GBM) at first recurrence. Patients were randomized 2:1 to receive CB or GW. CB (0.5 microg/mL; total flow rate 0.75 mL/h) was administered over 96 hours via 2-4 intraparenchymal catheters placed after tumor resection. GW (3.85%/7.7 mg carmustine per wafer; maximum 8 wafers) were placed immediately after tumor resection. The primary endpoint was overall survival from the time of randomization. Prestated interim analyses were built into the study design. Secondary and tertiary endpoints were safety and health-related quality-of-life assessments. From March 2004 to December 2005, 296 patients were enrolled at 52 centers. Demographic and baseline characteristics were balanced between the 2 treatment arms. Median survival was 36.4 weeks (9.1 months) for CB and 35.3 weeks (8.8 months) for GW (P = .476). For the efficacy evaluable population, the median survival was 45.3 weeks (11.3 months) for CB and 39.8 weeks (10 months) for GW (P = .310). The adverse-events profile was similar in both arms, except that pulmonary embolism was higher in the CB arm (8% vs 1%, P = .014). This is the first randomized phase III evaluation of an agent administered via CED and the first with an active comparator in GBM patients. There was no survival difference between CB administered via CED and GW. Drug distribution was not assessed and may be crucial for evaluating future CED-based therapeutics.