BACKGROUND: Despite available treatment options for patients (pts) with recurrent glioblastoma (GBM), < 5% of pts survive 5 years beyond initial diagnosis, and no single-agent therapy has demonstrated a survival benefit in the second-line setting, including bevacizumab (bev), which is approved for the treatment of recurrent disease. Nivolumab (nivo), a fully human IgG4 monoclonal antibody that inhibits the programmed death 1 receptor, has provided clinical benefit in multiple cancer types. In cohort 2 of the open-label, phase 3 CheckMate 143 study (NCT02017717), the efficacy and safety of nivo was compared with that of bev in pts with GBM experiencing their first recurrence after prior radiotherapy (RT) and temozolomide (TMZ). METHODS: Pts with no prior VEGF therapy were randomized 1:1 to receive nivo 3 mg/kg Q2W or bev 10 mg/kg Q2W until confirmed disease progression; pts were stratified by the presence/absence of measurable disease. The primary endpoint was overall survival (OS); secondary endpoints were 12-mo OS rate and investigator-assessed progression-free survival (PFS) and objective response rate (ORR) per Response Assessment in Neuro-Oncology criteria. RESULTS: At the time of final analyses (Jan 20, 2017), 369 pts were randomized to the nivo (n = 184) or bev (n = 185) treatment arms; of these pts, 182 received nivo and 165 received bev. At baseline, most pts in the nivo (83%) and bev (84%) arms had measurable disease, and 40% (nivo) and 43% (bev) of pts required corticosteroids, with 14% (nivo) and 15% (bev) receiving ≥ 4 mg/day. Deaths were reported in 154 (nivo) and 147 (bev) pts; median OS was 9.8 mo with nivo and 10.0 mo with bev, and the 12-mo OS rate was 42% in both arms. PFS medians were 1.5 mo (nivo) and 3.5 mo (bev). Among evaluable pts treated with nivo (n = 153) or bev (n = 156), ORRs were 8% (nivo) and 23% (bev); duration of response medians were 11.1 mo (nivo) and 5.3 mo (bev). Treatment-related AEs (TRAEs) occurred in 57% (nivo) and 58% (bev) of pts; the most common TRAEs (≥ 10% of pts in either arm; nivo vs bev) were fatigue (21% vs 14%) and hypertension (1% vs 22%). Grade 3–4 TRAEs were reported in 18% (nivo) and 15% (bev) of pts. Serious AEs (all causality) were reported in 46% (nivo) and 35% (bev) of pts; seizure (8% vs 6%) and malignant neoplasm progression (11% vs 7%) were the only serious AEs reported in ≥ 5% of pts in either arm. AEs leading to discontinuation occurred in 10% (nivo) and 15% (bev) of pts. CONCLUSIONS: Nivo did not demonstrate an improved OS compared with bev in pts with recurrent GBM. The ORR was lower with nivo than bev; however, responses with nivo were more durable. The safety profile of nivo was consistent with that observed in other tumor types. Studies of nivo in combination with RT ± TMZ in pts with newly diagnosed GBM are ongoing.