CD8 radiomics signature to assess inter-lesion spatial heterogeneity and cold liver lesions in advanced non-small cell lung cancers treated with durvalumab.

Abstract

2511 Background: The objective of this study was to assess whether a validated CD8 radiomics signature may help to evaluate patient inter-lesion heterogeneity and to predict the clinical outcome of advanced non-small cell lung cancers (NSCLC) patients treated with durvalumab in Study 1108 phase I/II trial (NCT01693562). Methods: Clinical data and imaging data from patients with naïve and pretreated advanced NSCLC who received durvalumab monotherapy were used. Radiomic features were extracted on contrast-enhanced CT scans and a validated CD8 radiomics signature was applied. A progressive lesion was defined by an increase in lesion longest diameter of 20% at 8 weeks. Dispersion metrics of the radiomics signature were estimated to evaluate the impact of inter-lesion heterogeneity on patient’s response. Results: A total of 188 patients were included in this study, accounting for a total of 1137 lesions (median [IQR] = 4 [3 - 9] lesions per patient) evaluated at baseline using a radiomics approach. A low CD8 radiomics score at baseline was associated with a significantly higher risk of progression at the lesion-level (AUC=0.59, P-value<0.0001), and was especially performant for liver lesions (AUC=0.66, P-value=0.0002). At the patient level, the least infiltrated lesion of a patient according to the radiomics score of CD8 T-cells was positively associated with OS (HR=0.70, P-value=0.029) and PFS (HR=0.68, P-value=0.014), the highest values being associated with the best outcomes. 55 patients had liver lesions, with worse prognosis than patients without liver lesion (HR=2 for OS and PFS, P-value = 0.00012 and 0.00022 respectively). In these patients, the CD8 radiomic score enabled the stratification of patients according to hot and cold liver metastasis (HR=0.6, P-value=0.072 and HR=0.54, P-value=0.038 for OS and PFS respectively). A 4-class stratification of the whole cohort based on the least infiltrated (cold/hot) and liver or non-liver lesion was independently associated with clinical outcomes. These radiomics approaches have shown independent prognostic values when adjusting for PD-L1 status in multivariate analyses (Table). Conclusions: These results confirm the predictive value at a lesion level and the patient level of the biologically inspired CD8 radiomics score for advanced NSCLC patients treated with durvalumab. It has shown interesting results in discriminating outcomes of patients with liver lesions by identifying hot and cold lesions. [Table: see text]