Research on Predicting the Efficacy of Immunotherapy for Non-small Cell Lung Cancer Based on CT Radiomics and...

IntroductionWe developed and validated predictive models based on CT radiomics and peripheral blood inflammatory markers to predict immunotherapy response in non-small cell lung cancer (NSCLC).MethodsThis retrospective study included 128 NSCLC patients receiving anti-PD-1/PD-L1 monotherapy as second-line therapy between January 2020 and January 2024. All eligible patients who met the inclusion criteria during the study period were consecutively included to minimize selection bias. Baseline non-contrast and contrast-enhanced chest CT acquired within 15 days before immunotherapy initiation was used for radiomics. Patients were randomly split into training and testing cohorts (7:3). Radiomics features were extracted and selected using LASSO with cross-validation followed by MRMR. A radiomics model, a clinical model (inflammatory markers), and a combined nomogram model were developed. Model performance was assessed by AUC and calibration; clinical utility was evaluated by decision curve analysis. Reporting followed TRIPOD+AI guidelines.ResultsThe combined model achieved the best performance (AUC 0.8802, 95% CI 0.8106-0.9497 in the training cohort; AUC 0.7329, 95% CI 0.5679-0.8979 in the testing cohort), outperforming the clinical model (AUC 0.6481 and 0.5325) and the radiomics model (AUC 0.7878 and 0.5621). DeLong tests for pairwise AUC comparisons yielded P values of 0.243 and 0.478, respectively.ConclusionsThe combined clinical-radiomics model may help identify NSCLC patients more likely to respond to immunotherapy. External validation is required before clinical implementation.