Exploration Of A Predictive Model For Immunotherapy Efficacy In Advanced Non-small Cell Lung Cancer Based On Radiomics And Clinical Features

Objective:The aim of our study was to explore the predictive factors of immunotherapy efficacy in real-world advanced non-small cell lung cancer(NSCLC)and to construct an efficacy prediction model based on radiomics and clinical features.Methods:Clinical data and computed tomography(CT)images of the chest were collected from advanced NSCLC patients treated with immune checkpoint inhibitors(ICIs).The study assessed the correlation between clinical features and progression-free survival(PFS)using Kaplan-Meier survival analysis and multivariate COX regression analysis.The region of interest(ROI)in the CT images was outlined using ITK-SNAP software,and the radiomics signature(RS)was extracted from the ROI using the pyradiomics toolkit,using the least absolute shrinkage and selection operator(LASSO)to filter highdimensional features associated with immune efficacy.The study used logistic regression to analyze the extracted radiomics signature and to create RS scores.Univariate COX regression assessed the correlation between RS scores and PFS.A predictive model of efficacy based on clinical characteristics and radiomics signature was developed using multivariate COX regression and represented by plotting an alignment diagram.The receiver operating characteristic(ROC)curve and the area under curve(AUC)were applied to evaluate the predictive validity of the integrated predictive model for 1-year PFS.Calibration curve,Hosmer-Lemeshow test and decision curve were used to evaluate the calibration and clinical applicability of the integrated predictive model.Results:1.A total of 104 patients with advanced NSCLC treated with first-line ICIs were included in this study,with a median follow-up time of 21.0 months(range: 1.0-60.5).As of the follow-up date,a total of 68 patients(65.4%)experienced disease progression with a median PFS of 11.1 months(95% CI: 8.5-13.7).1-year PFS rate was 53.8%.2.The results of the univariate and multivariate COX regression analyses of clinical characteristics suggested that clinical stage III,absence of liver metastases and prognostic nutritional index(PNI)≥47.8 were independent influences on longer PFS in patients with advanced NSCLC treated with first-line ICIs.A predictive model of clinical features was constructed based on 3 indicators(clinical stage,liver metastasis and PNI),which had a C-index of 0.68 and an AUC of 0.749 for predicting the 1-year PFS rate.3.By performing a logistic regression analysis of the radiomics signature(RS),our study established the RS scores.Univariate COX analysis suggested that high radiomics signature(High-RS)score was a predictive factor of longer PFS(HR: 0.44;95% CI: 0.250-0.772;P=0.004).4.Based on the predictive role of clinical features(clinical stage,liver metastases,PNI)and RS scores on PFS,the study constructed a comprehensive predictive model combining clinical features and radiomics signature.This model had a C-index of 0.71 and an AUC of 0.780 for predicting the 1-year PFS rate.The study produced an alignment diagram for the integrated model.Both the calibration curve analysis and the Hosmer-lemeshow test suggested a strong agreement between the disease progression rates predicted by the integrated model and the actual disease progression rates.Decision curve analysis(DCA)suggested good clinical applicability of the integrated model.Conclusions:1.Clinical stage III,absence of liver metastases and PNI≧47.8 are predictive factors of longer PFS in patients with advanced NSCLC treated with first-line ICIs.2.Radiomics signature can be used as a predictor of PFS in patients with advanced NSCLC treated with first-line ICIs.3.The integrated model combining clinical features(clinical stage,liver metastases and PNI)and radiomics signature can accurately predict the efficacy of ICIs in patients with NSCLC.