Clinical Parameters Combined With Radiomics Features To Predict The Efficacy Of Immunotherapy For Non-small Cell Lung Cancer

Lung cancer is the most common and deadly malignancy with a very poor prognosis.The advent of immunotherapy has changed the treatment landscape of advanced non-small cell lung cancer.As the first immunologic drug approved by the China Food and Drug Administration to enter the Chinese market,nivolumab provided a new treatment option for Chinese patients with advanced non-small cell lung cancer(NSCLC).However,studies have shown that not all patients could benefit from immunotherapy,with less than 1/5 patients experience a durable clinical response.Therefore,it is critical to find predictive biomarkers to screen for those who will benefit from immunotherapy.Part ⅠObjectives:Many scoring models based on clinical parameters have been established to predict the efficacy of immunotherapy,such as the LIPI,mLIPI and EPSILoN scores.We aimed to compare the ability of these three scoring models in predicting the prognosis of Chinese NSCLC patients treated with immune checkpoint inhibitors(ICIs)and to select the most suitable scoring model for Chinese patients.Methods:We retrospectively analyzed 429 patients with NSCLC treated with ICIs at Shandong Cancer Hospital from September 2018 to February 2020.The predictive ability of these models was evaluated using area under the curve(AUC)in receiver operating characteristic curve(ROC)analysis.The predictive accuracy of these models was evaluated by calculating the index of concordance(C-index)of each score.Results:The AUC values of the LIPI,mLIPI and EPSILoN scores for predicting PFS were 0.642(95%CI 0.590-0.694),0.720(95%CI 0.675-0.762)and 0.633(95%CI 0.585-0.679),respectively;the sensitivities were 53.6%,72.6%and 79.5%;specificities were 71.1%,66.3%and 40.4%.The C-index were 0.627(95%CI 0.611-6.643),0.677(95%CI 0.652-0.682)and 0.631(95%CI 0.617-0.645),respectively.Conclusion:The external validations of the LIPI,mLIPI and EPSILoN scores revealed that all these predictive models could be used for prognostic prediction in NSCLC patients receiving ICIs.And mLIPI score had the best predictive efficacy in Chinese NSCLC patients.Part ⅡObjectives:Screen out radiomics features associated with the prognosis of NSCLC patients receiving immunotherapy,establish a radiomics score(Radscore),and combine clinical scoring model with Radscore to further analyze whether the predictive power of the mixed model can be improved.Thus,a mixed scoring model integrating clinical parameters and radiomics features can be established,which in turn could rapidly screen out NSCLC patients who were more likely to benefit from immunotherapy.Methods:We collected contrast-intensive CT images of NSCLC patients received ICIs within 1 month before treatment.We used the semi-automatic segmentation method of the 3D-slicer platform to delineate the region of interest(ROI)of the tumor’s lesion area and extract image features for each patient.The least absolute shrinkage and selection operator(LASSO)was used to screen out the prognostic radiomics features.The selected features and the corresponding weight coefficients were linearly combined to build the Radscore,and then combined with mLIPI score,the best clinical parameter model screened in the first part,to construct Rad-mLIPI,a mixed predictive scoring model integrating clinical parameters and radiomics features.Results:317 patients with advanced NSCLC were included in the study,and 854 radiomics features were extracted.Five prognostic radiomics features were selected,namely,Original.shape.Flatness,Original.GLCM.Joint Entropy,wavelet.HLL.GLMC.Difference Entropy,wavelet.HLL.GLRLM.Gray Level Non Uniformity Normalized,wavelet.LLL.GLRLM.RunVariance.The AUC value and C-index of Radscore for predicting PFS were 0.762(0.697-0.827)and 0.643(0.602-0.684)in the training cohort,and an AUC of 0.682(0.580-0.785)and a C-index of 0.632(0.571-0.693)in the validation group.The AUC value and C-index of Rad-mLIPI model were 0.749(0.655-0.843)and 0.706(0.633-0.778)in the validation group.Conclusion:Radiomics can be used to predict the prognosis of NSCLC patients treated with immunotherapy,and the mixed predictive model Rad-mLIPI had the best predictive efficacy.