Background: Intravenous (IV) administration of ipilimumab (IPI) and nivolumab (NIVO) has low activity in recurrent glioblastoma (rGB). Intratumoral (IT) and intracavitary (IC) administration of IPI and NIVO is under evaluation in the GlITIpNi phase I clinical trial. Methods: Patients (pts) with resectable rGB were recruited to cohorts C1, C2 and C4; pts with non-resectable rGB were recruited in C3 (biopsy only). IT administration (brain tissue lining the resection cavity during surgery) of IPI (10 mg)(C1), or IPI (5 mg) plus NIVO (10 mg)(C2, C3 and C4), was followed by IC administration of NIVO at escalating doses of 1, 5 or 10 mg Q2w in both C3 and C4 (via an Ommaya reservoir). In all cohorts, pts received 10 mg NIVO IV Q2w (6x in C1/C2, and 12x in C3/C4). Corticosteroids were contraindicated. Results: Forty-six pts (31 male; median age 56y (38-74); IDH1 R132H mutation in 2 pts in C1/C2; NGS somatic mutation analysis for C3/C4 ongoing) with rGB following resection, RT and temozolomide were enrolled (3, 24, 13 and 6 pts in C1, C2, C3 and C4, respectively). All pts received IT administrations. Pts in C1/C2 received a median of 5 IV NIVO administrations. Study treatment has been completed in all pts in C1/C2, in 9 pts in C3, and in 3 pts in C4; pts received a median of 4 (0-10) and 3 (0-7) postoperative IC/IV administrations in C3 and C4, respectively. Two pts in C2 and 1 pt in C3 had an increased perilesional cerebral edema (G3) with neurological deterioration after surgery/IT-injection, that was reversible with steroids. Most frequent AE were fatigue (32 pts, 64%), fever (20 pts, 44%), and headache (25 pts, 50%). In 4 pts from C3, the Ommaya was removed because of bacterial colonization (asymptomatic). There were no G5 AE. There was no dose/AE correlation with increasing IC NIVO doses in C3/C4. Repetitive CSV analysis during therapy (C3/C4) revealed increased lymphocyte counts in 4 pts; scRNA- and TCR-sequencing is ongoing. Gene expression profiling for C1/C2, and pharmacokinetic analysis of NIVO and IPI in CSV for C3/C4 are ongoing. After a median FU of 62w (16-165) for pts in C1/C2, 16 pts have died; median OS is 71w (95% CI 8-134), 1- and 2y-OS% are respectively 51% (95% CI 31-71), and 34% (95% CI 10-59). OS compares favorably to a historical cohort of Belgian rGB pts (n = 469; Log-Rank p .001). After a median FU of 10w (1-37) for pts in C3/C4, 2 pts have died; median OS has not been reached. One pt in C3 achieved a PR that is ongoing at 12m. Conclusions: IT/IC administration of NIVO and IPI is feasible and sufficiently safe to warrant further investigation in pts with rGB. Clinical trial information: NCT03233152.