Abstract BACKGROUND Although monoclonal antibodies targeting PD-1 and its ligand PD-L1 have improved overall survival (OS) for many cancers, recent trials of anti-PD-(L)1 in glioblastoma revealed no OS benefit. However, a small subset of patients appears to benefit, warranting evaluation of predictive markers of improved survival. METHODS We evaluated 108 patients managed at Dana-Farber Cancer Institute for IDH-wildtype glioblastoma, who received anti-PD-(L)1 therapy and had complete clinical, pathologic, and radiographic data. OS was measured from anti-PD-(L)1 initiation to death or censored at last follow-up. Predictors of OS were evaluated by multivariable Cox regression, adjusted for age at diagnosis, sex, disease setting (newly-diagnosed vs. recurrent), MGMT status, chemotherapy and bevacizumab before/during anti-PD-(L)1, extent of resection prior to anti-PD-(L)1, and, at anti-PD-(L)1 initiation: KPS, dexamethasone use, absolute lymphocyte count (ALC), ADC mean value, FLAIR volume, T1 post-contrast volume. RESULTS Of 108 patients, 28 (25.9%) were newly-diagnosed, 80 (74.1%) were treated in the recurrent setting, and 100 died (92.3%). In multivariable analysis, unmethylated MGMT (HR 2.31, 95%CI: 1.38–3.88, p=0.002), and, at the time of anti-PD-(L)1 initiation, baseline dexamethasone use (HR 1.79, 95%CI: 1.07–3.00, p=0.03), lower ADC mean values (reference 2nd quartile [1,130-1,262 10-6mm2/s]; 3rd quartile [1,268–1,388] HR 0.32, 95%CI: 0.15–0.67, p=0.002; 4th quartile [1,402–11,966] HR 0.30, 95%CI: 0.14–0.66, p=0.003), and higher T1 post-contrast volumes (reference 2nd quartile [4,121–8,157mm3]; 3rd quartile [8,701–14,663mm3] HR 2.08, 95%CI: 1.03–4.19, p=0.04; 4th quartile [15,566-64,362mm3] HR 3.15, 95%CI: 1.49–6.69, p=0.003) independently predicted worse OS. Older age (HR 1.02/year, 95%CI: 1.00–1.05, p=0.07) and lower ALC (HR 0.66/unit, 0.40–1.10, p=0.11) trended towards worse OS. CONCLUSION Unmethylated MGMT, and, at the time of starting anti-PD-(L)1, baseline dexamethasone, low ADC mean value, and high T1 post-contrast volume may be predictive of worse OS in IDH-wildtype glioblastoma patients treated with anti-PD-(L)1. These findings may help better select patients who could benefit from PD-(L)1 inhibitors.