Long-term responders have been observed with anti-programmed death 1 and anti-programmed death ligand 1 (anti-PD(L)1). Optimal duration of therapy in responding and stable disease (SD) patients is unclear with various attitudes encompassing treatment until progression disease, stopping therapy after a defined timeframe.We report the experience of 13 patients who discontinued immune checkpoint inhibitor in phase I trials as per protocol while experiencing a tumour-controlled disease. According to protocols, patients could restart the same immunotherapy if radiological or clinical progression occurred.Patients were treated for colorectal microsatellite instability-high genotype (n = 5), urothelial carcinoma (n = 3), melanoma (n = 2), non-small-cell lung cancer (n = 2) and triple-negative breast cancer (n = 1) for a median time of 12 months (range 10.6-12). Patients achieved 1 (8%) complete response, 10 (77%) partial response (PR) and 2 (15%) SD. The median progression-free survival 1 (PFS1) defined as the time from the first infusion until progression was 24.4 months (range 15.8-49). The median time free-treatment after discontinuation was 12.6 months (range 4-39.7). Eight patients experienced disease progression and were retreated. Best responses observed after rechallenging were 2 PR (25%) and 6 SD (75%). Median PFS2 defined from the first day of retreatment until disease progression or the last news was 12.9 months (range 5-35.4). No grade 3/4 events occurred during the study period.Our data suggest that anti-PD(L)1 therapy should be resumed if progression occurs after a planned anti-PD(L)1 interruption. Further prospective studies are needed to confirm these results.