International Case-Control Study of DNA Methylation Biomarkers for the Minimally Invasive Diagnosis of Esophageal Squamous Cell Carcinoma.

Abstract

Introduction: Esophageal squamous cell carcinoma (ESCC), the most common type of esophageal cancer worldwide, is usually diagnosed by esophagogastroduodenoscopy, an expensive procedure having limited availability in low- and middle-income countries. We therefore developed a diagnostic strategy that applies ESCC-specific DNA methylation biomarkers to nonendoscopically-obtained esophageal mucosal samples.

Methods: We used a novel approach based on a search of 5 publicly available GEO and TCGA datasets, which contained 196 organ-specific normal controls. We selected 6 new methylation markers, combining these with 5 markers identified in a previous study. We then performed a case-control analysis of 174 esophageal cytologic samples collected using a swallowable sponge-capsule device, one sample each from 93 patients with ESCC and 81 noncancer control patients. Based on these molecular data, a classification algorithm was built in a training set of 122 patients using the least absolute shrinkage and selection operator (LASSO) method, which was then tested in a separate cohort of 52 patients. Coefficients of the LASSO regression model were used to select markers that contributed most strongly to the classification.

Results: All 11 markers tested exhibited significantly higher methylation levels in all 93 ESCC sponge samples than in 81 normal control sponges, with P -values below 0.001. Furthermore, individual areas under the receiver operating characteristic curve (AUCs) for the 11 markers, C1ORF70, PPFIA3, SKOR1, ZNF132, TAC1, JPH4, ZNF542, KLF16, CG20655070, SLC35, and PAX9, were 0.91, 0.90, 0.89, 0.89, 0.89, 0.89, 0.87, 0.87, 0.83, 0.82, 0.81, and 0.65, respectively. Based on LASSO regression applied to all 11 markers, 4 (C1ORF70, CG20655070, SKOR1, and PPFIA3) were retained in our final model. A composite score based on this final model achieved an area under the receiver operating characteristic curve of 0.93 in the 122-patient training cohort and 0.92 in the 52-patient test set.

Discussion: When used in combination with a minimally invasive sponge-capsule device, ESCC methylation biomarkers may improve the diagnosis of ESCC. This strategy harbors potential to mitigate the deadly worldwide public health impact of ESCC.