Delta radiomic parameters for the evaluation of liver metastases in patients with colorectal cancer treated with cetuximab rechallenge plus avelumab in the CAVE mCRC study.

Abstract

241 Background: We performed a retrospective analysis of metastatic colorectal cancer (mCRC) patients enrolled in CAVE trial. Cave mCRC is a single arm, academic, multicenter, phase II trial that evaluated the efficacy of cetuximab rechallenge plus avelumab in mCRC patients. Methods: CT imaging of patients were collected and retrospectively analysed. Inclusion criteria included evidence of liver metastases that could be measured and segmented, availability of an enhanced contrast phase at baseline and restaging CT imaging (within 3 months from baseline). The gross tumor volume (GTV) of liver metastases was evaluated at both baseline and first restaging CT. Multiple texture parameters were extracted with LifeX Software, and delta-texture analysis (D-TA) was calculated as percentage variations in the two time points. Patients were divided in a training cohort (coordinating center, University of Campania “L. Vanvitelli”) and validation cohort (other centers involved in CAVE mCRC trial). Patients were further divided in responders/non-responders according to median progression free survival (mPFS). ROC curve were calculated to obtain a cut-off. Survival analysis of progression free survival (PFS) and overall survival (OS) with Kaplan-Meier method was used in the two subgroups to test the cut-off found with the methods previously described. Results: The original dataset of CAVE mCRC protocol included 77 patients. Liver metastasis were present in 55 patients (71%); after application of inclusion and exclusion criteria, a total of 42 patients were included in the current analysis. The training cohort consisted of 22 patients (for a total of 32 liver metastasis) and the validation cohort of 20 patients (for a total of 28 liver metastasis). After Bonferroni correction, the only D-TA parameters that significantly correlated with mPFS in the training cohort were EntropyHistogram (p:0,021), HomogeneityGLCM (p<0,001) and Dissimilarity GLCM (p:0,002). At multivariate analysis only HomogeneityGLCM resulted significant both in training (p:0,001, OR 0,1) and validation cohort (p:0,021, OR 0,3). ROC curves were generated and the cut-off of HomogeneityGLCM was equal to 0. Patients who developed a reduction of HomogeneityGLCM at first restaging CT scan showed a better PFS and OS (in both training and validation cohorts). Conclusions: Our results suggest that D-TA parameter such as HomogeneityGLCM is able to discriminate better survival outcomes in patient treated with cetuximab rechallenge plus avelumab, further prospective investigations are needed. [Table: see text]