Optimized response prediction in esophagogastric junction adenocarcinomas (EGJAc) with combination of molecular biomarkers and FDG-PET

Background: Predictive biomarkers (BMs) for EGJAc would optimise treatment selection and avoid ineffective therapy. Metabolic response (MR) defined as >35% decrease in tumour FDG Standardized Uptake Value (SUV) between day 0 & 14 after starting chemotherapy has a high negative predictive value (95%) for response, but limited positive predictive value (50%). Combining molecular BMs with FDG-PET may optimise response prediction. We used global gene expression profiling (GEP) to identify molecular BMs that when combined with FDG-PET would improve predictive accuracy. Methods: 28 patients with locally advanced or metastatic EGJAc received platinum based chemotherapy (PBC). FDG PET CT scans were at day 0 and day 14 and GEP (Affymetrix ST1.0 Exon Genechips) on day 0 tumour biopsies. A tissue microarray comprising an independent set of 154 OGJAc who underwent surgery +/- neoadjuvant PBC was used with immunohistochemistry (IHC) for qualification of GEP results. Radiological response was assessed after 3/ 4 cycles of PBC by RECISTv1.1. Results: We identified a gene expression signature (86 genes) that separated FDG PET MR patients(>35% fall SUV day 0 to14) into those that do and do not go on to have a RECIST response. In cross validation, this signature correctly predicted response in 28/28. Pathway analysis on GEP data identified potential novel mechanisms of response, including the Leptin pathway. Leptin mRNA was higher in FDG metabolic responders who did not have a RECIST response compared to those that did. In the independent set, high Leptin protein by IHC was strongly associated with lack of histopathologic response to neoadjuvant PBC (n=64, p=0.002). High Leptin expression also had a therapy independent prognostic effect with longer survival in the absence of histopathologic response or with no neoadjuvant PBC and in low Leptin patients poor survival was mitigated to a certain extent by neoadjuvant PBC (n=154, Kaplan-Mieier, log rank p=0.041 & Cox MVA p=0.040). Conclusions: Molecular biomarkers (Leptin in particular) combine with FDG PET to optimise response prediction in EGJAc. Further investigation of this combined molecular and imaging approach is warranted.