There is no conventional treatment for patients with severe drug-induced liver injury (DILI) except for discontinuation of liver injury drugs and symptomatic supportive therapy. Opinions on whether corticosteroids can be used to treat severe DILI are conflicting, and most of the relevant clinical studies are case reports or retrospective studies, which still need to be supported by high-level evidence-based medical studies. This study aimed to evaluate the effect and tolerance of corticosteroids in patients with severe DILI. Risk factors associated with patient failure to cure were also explored.

Propensity score matching based on nearest-neighbor 1:1 matching was used to screen severe DILI patients in the corticosteroids and control groups. Severe DILI was defined as elevated serum ALT and/or ALP with TBIL≥5 ULN (5 mg/dL or 85.5 μmol/L) with or without INR ≥1.5. Patients were treated with conventional therapy combined with corticosteroids in the corticosteroids group and only conventional therapy in the control group.

A total of 146 patients, 73 each in the corticosteroids and control groups, were included in this study. By analyzing the entire cohort, we found no significant difference in cure rates between patients in the corticosteroid group and control group (34.2% vs. 20.5 %, p = 0.095), and there was no significant difference in the incidence of adverse effects between the two groups (20.5% vs. 20.5 %, p = 1.000). However, TBIL decreased more in the corticosteroids group on day 7 (89.2 ± 107.6 μmol/L vs. 58.8 ± 70.7 μmol/L, p = 0.046). In subgroup analyses, patients whose TBIL remained elevated despite conventional treatment had a higher TBIL decline on day 7,14 after use of corticosteroid (99.2 ± 98.5μmol/L vs. -23.3 ± 50.4μmol/L, p < 0.001; 120 ± 119.1μmol/L vs. 61.2 ± 98.5μmol/L, p = 0.047). The cure rate of patients in the corticosteroid group was significantly higher than that of the control group (36.1 % versus 4.5 %, p = 0.016). The proportion of patients with TBIL <85.5 μmol/L was also significantly higher in the corticosteroid group than in the control group at day 7 (p = 0.016) and day 14 (p = 0.004) after treatment. In the subgroup analysis of patients with different clinical phenotypes, the causative agent was herbal, autoimmune antibody-positive and 40 % < PTA ≤ 50 % of patients, corticosteroid use did not increase the cure rate of the patients. Univariate and multifactorial analyses found corticosteroid use to be a protective factor for failure to cure in patients with severe DILI (p < 0.001, OR:0.191,95 % CI:0.072–0.470), and peak TBIL to be a risk factor (p = 0.003, OR:1.016,95 % CI:1.007–1.028).

The addition of corticosteroids could not increase the cure rate in patients with severe DILI, but it could rapidly reduce the patient's TBIL at an earlier stage. Corticosteroids could also promote curing in patients with elevated TBIL after conventional treatment. Corticosteroid use was a protective factor for failure to cure in patients with severe DILI and peak TBIL was a risk factor.