Prediction of immune-related adverse events in non-small cell lung cancer patients treated with immune checkpoint inhibitors based on clinical and hematological markers: Real-world evidence

Abstract

Clinical and hematological parameters can predict immune-related adverse events (irAEs) caused by immune checkpoint inhibitors (ICIs). However, the exact correlation between these parameters and irAEs is unclear. This study aimed to establish a prediction model for irAEs in patients with non-small cell lung cancer (NSCLC) treated with ICIs. This retrospective study included patients with NSCLC treated with a minimum of one dose of ICIs at the Tianjin Medical University Cancer Hospital and Shanxi Bethune Hospital from 2016 to 2020. Baseline characteristics, treatment details, and adverse events were evaluated. The Student's t-test, Chi-square test, and logistic regression were used to identify risk factors for irAEs to establish a prediction model. A total of 667 patients were included; the median age was 62.47 (range, 27–85) years. Most patients were men (74.5%) with stage IV cancer (93.1%). The incidence of any grade and grade 3 or higher irAEs was 21.74% (145/667) and 5.25% (35/667), respectively. A total of 145 patients experienced 220 irAEs; the incidence of endocrinopathies (35.91%, 79/220) was highest in all grade irAEs, while that of pneumonitis (7.73%, 17/220) was the highest in grade 3 or higher irAEs. A prediction model based on treatment lines, aspartate aminotransferase (AST), lactate dehydrogenase (LDH), absolute lymphocyte count (ALC), and systemic immune inflammation index was established. The area under the receiver operator characteristic curve was 0.722 (95% confidence interval: 0.650–0.793), with a cut-off value of 0.247 and a sensitivity and specificity of 62.9% and 74.6%, respectively. The multivariate logistic regression analysis showed that the risk of irAEs was higher in patients undergoing second-line therapy than in those undergoing treatment with adjuvant therapy (odds ratio [OR] = 8.239, p = 0.011). AST (OR = 1.053, p = 0.007) and ALC (OR = 2.556, p = 0.001) showed a positive correlation with the risk of irAEs, while LDH showed a negative correlation with irAEs (OR = 0.994, p = 0.007). The model showed good prediction efficiency, whereas the treatment lines, AST, ALC, and LDH were independent risk factors for the onset of irAEs.