Predicting the efficacy of immune checkpoint inhibitors monotherapy in advanced non-small cell lung cancer: a machine learning method based on multidimensional data

Abstract

Immunotherapy has improved the prognosis of patients with advanced non-small cell lung cancer (NSCLC), but only a small subset of patients achieved clinical benefit. The purpose of our study was to integrate multidimensional data using a machine learning method to predict the therapeutic efficacy of immune checkpoint inhibitors (ICIs) monotherapy in patients with advanced NSCLC. We retrospectively enrolled 112 patients with stage IIIB-IV NSCLC receiving ICIs monotherapy. The random forest (RF) algorithm was used to establish efficacy prediction models based on five different input datasets, including precontrast computed tomography (CT) radiomic data, postcontrast CT radiomic data, a combination of the two CT radiomic data, clinical data, and a combination of radiomic and clinical data. The 5-fold cross-validation was used to train and test the random forest classifier. The performance of the models was assessed according to the area under the curve (AUC) in the receiver operating characteristic curve. Survival analysis was performed to determine the difference in progression-free survival (PFS) between the two groups with the prediction label generated by the combined model. The radiomic model based on the combination of precontrast and postcontrast CT radiomic features and the clinical model produced an AUC of 0.92 ± 0.04 and 0.89 ± 0.03, respectively. By integrating radiomic and clinical features together, the combined model had the best performance with an AUC of 0.94 ± 0.02. The survival analysis showed that the two groups had significantly different PFS times (p < 0.0001). The baseline multidimensional data including CT radiomic and multiple clinical features were valuable in predicting the efficacy of ICIs monotherapy in patients with advanced NSCLC.