A study of predictive and early response radiomics signatures to atezolizumab in molecularly selected populations with advanced solid tumours
Despite the excellent results of immunotherapy trials, some caveats remain concerning patient selection and response assessment. While some patients achieve excellent responses, others do not respond to immunotherapies at all. Moreover, distinct imaging patterns in response to immunotherapies have been described (e.g. pseudoprogression, hyperprogression). It is therefore critical to better predict and identify patients responding and not responding to therapy in order to maximize benefit and minimize patients’ risk.
Medical imaging enables non-invasive cancer diagnosis and guides decision making in clinical practice. Radiomics use imaging data to quantitatively assess tumour tissue features using automatically extracted data-characterization algorithms. Radiomics can thus be used to facilitate a deeper understanding of tumour biology, capture tumour heterogeneity, and monitor tumour evolution and response to therapy. Different research groups within CEE have reported promising data on the value of radiomics signatures for predicting response to immune checkpoint inhibitors (Trebeschi S. et al. Annals of Oncology 10.1093/annonc/mdz108; Ligero M et al. Radiology 10.1093/annonc/mdz108). Several challenges, however, still need to be addressed for their implementation in clinical practice. The CCE-Imaging-Task-Force has joined forces to address this issue using data from the Basket-of-Basket clinical study.
We hypothesize that CT-radiomics signatures are robust indicators of response to the immune-checkpoint-inhibitor atezolizumab (PD-L1 monoclonal antibody), in patients with advanced solid tumours. Within the CCE consortium we aim to develop and validate a CT-radiomics signature of solid tumours that predicts response to atezolizumab. As secondary objectives we will evaluate: i) the early changes in CT-radiomics signatures as response to atezolizumab, ii) the correlation of radiomics features with tumour mutational status to develop radiogenomics phenotypes and, iii) the correlation of CT-radiomics features with histological ‘measurements’ from tumour samples.