Accurate Non-endoscopic Detection of Early Esophageal Squamous Malignant Lesions Using Sponge-based Methylated DNA Biomarkers.

Background: Early detection of esophageal squamous cell carcinoma (ESCC) significantly improves patient quality of life and outcomes. We aimed to train and validate a predictive algorithm using methylated DNA markers (MDMs) based on encapsulated sponge cell collection device (CCD) samples. Methods: CCD samples in the training set (n=120) were taken from two medical centers, and in the validation set (n=70) were prospectively collected from another medical center. The case group contained a diverse range of esophageal high-grade lesions, including high-grade intraepithelial neoplasia (HGIN), early ESCC, and advanced ESCC. The algorithms were trained on the training set and tested to evaluate diagnostic performance in a prospective validation set. Results: The MDM model was constructed using logistic regression with two MDMs (OTOP2 and KCNA3). The areas under the receiver operating characteristic curves (AUCs) were 0.933 (95%CI 0.881 to 0.986) and 0.911 (95%CI 0.841 to 0.981) in the training and validation sets, respectively. The overall sensitivity was 90.0% at a specificity of 91.4% in the training set. The sensitivity was 90.0%, 95.0%, and 90.0% for HGIN, early ESCC, and advanced ESCC in the training set, respectively. The sensitivity and specificity were 87.5% and 86.7% in the validation set, respectively. The AUCs for the complex model that included the two MDMs and Age were 0.961(95%CI, 0.920 to 1.000) and 0.940 (95%CI, 0.889 to 0.990) in the training and validation sets, respectively. Conclusion: The assay of MDMs in CCD samples offers a highly accurate method of predicting HGIN/ESCC, thereby providing the potential for early diagnosis and screening for HGIN/ESCC.