SummaryBackgroundThere are few data on corticosteroids (CS)‐sparing strategies for checkpoint inhibitor (ICI)‐induced liver injury (ChILI).AimWe aimed to assess the performance of a 2‐step algorithm for severe ChILI, based on ICI temporary discontinuation (step‐1) and, if lack of biochemical improvement, CS based on the degree of necroinflammation at biopsy (step‐2).MethodsProspective study that included all subjects with grade 3/4 ChILI. Peripheral extended immunophenotyping was performed. Indication for CS: severe necroinflammation; mild or moderate necroinflammation with later biochemical worsening.ResultsFrom 111 subjects with increased transaminases (January 2020 to August 2023), 44 were diagnosed with grade 3 (N = 35) or grade 4 (N = 9) ChILI. Main reason for exclusion was alternative diagnosis. Lung cancer (13) and melanoma (12) were the most common malignancies. ICI: 23(52.3%) anti‐PD1, 8(18.2%) anti‐PD‐L1, 3(6.8%) anti‐CTLA‐4, 10(22.7%) combined ICI. Liver injury pattern: hepatocellular (23,52.3%) mixed (12,27.3%) and cholestatic (9,20.5%). 14(32%) presented bilirubin >1.2 mg/dL. Overall, 30(68.2%) patients did not require CS: 22(50.0%) due to ICI discontinuation (step‐1) and 8/22 (36.4%) based on the degree of necroinflammation (step‐2). Biopsy mainly impacted on grade 3 ChILI, sparing CS in 8 out of 15 (53.3%) non‐improvement patients after ICI discontinuation. CD8+ HLA‐DR expression (p = 0.028), central memory (p = 0.046) were lower in CS‐free managed subjects, but effector‐memory cells (p = 0.002) were higher. Time to transaminases normalisation was shorter in those CS‐free managed (overall: p < 0.001, grade 3: p < 0.001). Considering our results, a strategy based on ICI discontinuation and biopsy for grade 3 ChILI is proposed.ConclusionsAn algorithm based on temporary immunotherapy discontinuation and biopsy allows CS avoidance in two thirds of cases of severe ChILI.