Predicting response to neoadjuvant therapy in oesophageal adenocarcinoma pre-treatment biopsies

Abstract Aims We currently cannot predict which patients with locally advanced oesophageal adenocarcinoma will be amongst the 15-20% to gain a clinically important response to neoadjuvant therapy (NAT). This pilot study aimed to identify differentially expressed genes from oesophageal adenocarcinoma pre-treatment biopsies between responders and non-responders to NAT and develop methodology for predicting response. Method Response to NAT was assessed pathologically using Tumour Regression Grading (TRG). Pre-treatment formalin-fixed paraffin embedded samples were analysed with two nuclease protection assays (EdgeSeq, HTG = Oncology Biomarker Panel (OBP) and Precision Immuno-Oncology Panel (PIP)). Sequencing was performed on the NextSeq500 (Illumina). Result Whilst there was no difference in pre-treatment characteristics, responders (TRG1-2, n=26) had significantly better post-treatment pathology and overall survival than non-responders (TRG4-5, n=30). Genes up-regulated in responders were involved in regulating cell cycling, whereas genes up-regulated in non-responders were involved in cytokine signalling and the immune response. Neuronal artificial network models could predict response to NAT with overall accuracy of 73% and 68% for the OBP and PIP, respectively, which is promising considering the small sample size. As no model will be 100% accurate, we developed a model that could take patient's views into consideration with an adjustable probability threshold for classification. Conclusion This pilot study informs a biologically sound hypothesis for the basis of response to NAT and suggests prediction from pre-treatment biopsies may be possible using EdgeSeq. We now aim to validate these results in a larger study to inform a bespoke classifier of response to enable delivery of precision therapy. Take-home message In oesophageal adenocarcinoma, responders and non-responders to neoadjuvant therapy have different expression profiles. Through using EdgeSeq in larger studies, we may be able to predict which patients will respond to treatment, allowing for delivery of precision therapy.