To observe clinical features of programmed cell death protein-1 (PD-1) inhibitor-induced liver injury in tumor patients and analyze its influencing factors.

Tumor patients treated with PD-1 inhibitors were enrolled. The clinical data were analyzed. The occurrence of liver injury during treatment was assessed according to CTCAE V5.0 criteria. Clinicopathological features of patients with liver injury were summarized. Univariate and multivariate logistic regression analyses were used to identify independent risk factors for liver injury in tumor patients, and a nomogram prediction model was constructed receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA) were employed to evaluate validity of the model.

The incidence of PD-1 inhibitor-induced liver injury among the 310 patients enrolled was 11.94% (37 cases/310 cases). Grade 1, grade 2, grade 3 and grade 4 liver injuries were observed in 12 cases (32.43%), 15 cases (40.54%), 8 cases (21.62%) and 2 cases (5.41%), respectively. Clinical types included cholestatic type in 23 cases (62.16%), abnormal liver biochemistry type in 12 cases (32.43%), and hepatocellular injury type in 2 cases (5.41%). The median time to the occurrence of liver injury was 28.50 (5.00-217.00) d. After drug withdrawal or symptomatic treatment, liver function recovered, and the mean recovery time was (21.45±6.94) d. Multivariate logistic regression analysis found female gender [odds rotio (OR)=2.36, 95% confidence interval (CI)=1.11-4.99], history of comorbid liver disease (OR=2.62, 95%CI=1.10-6.24), and liver metastasis (OR=3.10, 95%CI=1.28-7.49) as independent risk factors influencing the occurrence of liver injury in tumor patients (all P<0.05). ROC analysis showed that the area under curve (AUC) (95%CI) of the nomogram prediction model for predicting liver injury was 0.86 (0.82-0.90), with sensitivity of 78.38% and specificity of 83.52%. Hosmer-Lemeshow test showed good goodness-of-fit (χ²=12.58, P>0.05). Validation with Bootstrap revealed good agreement between the calibration curve and the actual curve (Brier score=0.09). DCA demonstrated that the nomogram prediction model provided clinical benefit when the threshold was between 0.05 and 0.99.

PD-1 inhibitor-induced liver injury is relatively common in tumor patients. Female gender, history of liver disease and liver metastasis are associated with its occurrence. The nomogram prediction model constructed based on these factors exhibits good predictive performance and can provide reference for early identification of high-risk patients.