This study aimed to identify potential biomarkers for adverse reactions (ADRs) and efficacy of osimertinib, and to construct prediction models in patients with non-small cell lung cancer(NSCLC). NSCLC patients treated with osimertinib were retrospectively enrolled from January 2021 to July 2022. Their clinical characteristics, ADRs and prognostic outcomes (e.g., progression-free survival, objective response rate) were extracted from medical records. Univariate and multivariate analysis were performed on baseline data to identify potential biomarkers associated with ADRs and efficacy. Prediction models were constructed and validated using logistic regression. All analysis were conducted with SPSS 19.0 and Graphpad Prism 8.0. A total of 90 patients in training cohort and 74 patients in the validation cohort were enrolled. Univariate analysis revealed that drinking history, total bilirubin(TB), indirect bilirubin(I-Bil), and aspartate aminotransferase(AST) were potential biomarkers for ADRs. Multivariate analysis identified drinking history, TB, and AST as key predictors of severe ADRs. No variate was associated with efficacy. A prediction model for severe ADRs was established using logistic regression: y = -2.714×(drinking history) − 0.4×(TB) − 0.099×(AST) + 6.196 Area Under the Curve (AUC) = 0.87; 95% confidence interval (CI): 0.752–0.995; P < 0.001; sensitivity: 90%; specificity: 82.5%; optimal Youden index: 0.725. Validation of this model in 74 NSCLC patients showed a sensitivity of 83.3% and specificity of 98.2%. This study developed and validated a prediction model for severe osimertinib-related ADRs in NSCLC patients based on baseline characteristics. The model exhibits good predictive performance for severe osimertinib-related ADRs and may facilitate individualized osimertinib therapy in clinical practice.
Cui et al. (Sun,) studied this question.