Correlation of radiomics of metastatic lesions in gastroesophageal adenocarcinoma (GEA) with tumoral DKK1 mRNA expression and other immune biomarkers in patients (pts) treated with DKN-01.

Abstract

429 Background: Dickkopf-1 (DKK1) modulates Wnt and PI3K/AKT signaling pathways and contributes to an immune suppressive tumor microenvironment. Recently, anti-DKK1 antibody (DKN-01) [D] plus pembrolizumab (P) in GEA tumors has demonstrated clinical activity. Radiomic Quantitative Texture Analysis (QTA) is a non-invasive method for imaging biomarker discovery that can predict molecular drivers of cancer from CT scans obtained during clinical trials. QTA may be a useful tool for predicting DKK1 expression and other immune biomarkers in metastatic lesions from GEA pts treated with DKN-01. Methods: 13 pts with GEA (12 M; 1F; age 37-81) enrolled in D+P Phase I/II trial with high vs low DKK1 RNA Scope H-Scores (H Score > 35; n = 6 vs H-Score < 35; n = 7) were identified. Metastatic Target lesions (TLs) from baseline CTs with IV contrast underwent additional QTA analysis (TexRAD, CE mark, Essex, UK) after RECIST analysis. ROIs were placed on the TLs and histogram frequency curves (HFC) of the metastatic tumor pixels were generated. First order tumor HFC statistical results (i.e. mean, Standard Deviation-SD, Skewness, Kurtosis, MPP) were correlated with tumoral DKK1 mRNA expression, baseline plasma DKK1 levels and other biomarkers (PD-L1 CPS and MDSCs) along with tumor shrinkage using Pearson’s r correlation (significance p < 0.05). Results: Tumor derived QTA parameter (QTA SD) of the largest metastatic TL from each subject was positively correlated with DKK1 RNAscope H-score (r = 0.661, p value = .014) meaning larger SD texture scores had greater DKK1 expression. Also, tumor derived QTA parameter (QTA MPP) was correlated with baseline plasma DKK1 (r = -0.461, p = 0.005). Finally, QTA SD was correlated with PD-L1 expression (r = 0.630, p = 0.028) and TL shrinkage (r = -0.590, p = 0.034). Conclusions: Radiomics of GEA tumor metastasis using QTA showed significant correlations between tumor texture, tumor DKK1 mRNA expression and baseline plasma DKK1 suggesting that tumor textures may provide a non-invasive tool for assessing DKK1. Although promising as a imaging biomarker, further studies are required to validate these findings. Clinical trial information: NCT02013154.