Identification and validation of a serum biomarker for high-specific screening of colorectal cancer and precancerous lesions

This dataset contains de-identified raw clinical data from two independent validation cohorts (local and external) supporting a study on serum EPHB2 as a candidate biomarker for colorectal adenoma and colorectal cancer (CRC). Research hypothesis: We hypothesized that serum EPHB2 protein levels are elevated in patients with colorectal neoplasia (adenoma and CRC) compared to normal controls, and that a fixed diagnostic cutoff derived from a training cohort could independently validate EPHB2's discriminatory performance in geographically distinct populations. Data description: The repository includes two Excel files: "Raw data list_External validation dataset.xlsx" and "Raw data list_Local validation dataset.xlsx". Each file contains serum EPHB2 concentrations (measured by ELISA) for each participant. All direct personal identifiers have been removed to ensure patient confidentiality. Data collection: Both cohorts were recruited from patients undergoing colonoscopy or surgical resection at respective medical centers. Fasting serum samples were collected prior to the procedure, and EPHB2 levels were quantified using a commercially available enzyme-linked immunosorbent assay (ELISA) kit according to the manufacturer's protocol. The local validation cohort was used to assess initial reproducibility, while the external validation cohort served as a fully independent geographic validation. Notable findings: At the training-derived cutoff (EPHB2 ≥ 0.536), the biomarker demonstrated moderate discriminatory ability in the external validation cohort (AUC = 0.794 for adenoma vs. normal; AUC = 0.816 for CRC vs. normal). Sensitivity and specificity were 66.7% and 75.0% for adenoma, and 75.0% and 78.1% for CRC, respectively. The data reveal that EPHB2 exhibits better diagnostic performance for CRC than for adenoma, with a high negative predictive value when adjusted to average-risk screening prevalence. Interpretation and use: These datasets enable independent researchers to (1) validate the diagnostic performance of serum EPHB2 using the prespecified cutoff, (2) compare EPHB2 against established colorectal cancer screening modalities (e.g., FIT, SEPT9), (3) perform decision curve analysis and prevalence-adjusted predictive value calculations, and (4) integrate EPHB2 into multi-marker diagnostic models. Users should note that both cohorts were derived from endoscopy-referred populations with high disease prevalence (spectrum bias), and results should not be directly extrapolated to average-risk asymptomatic screening populations without appropriate prevalence adjustment.