Serum markers modestly improve fecal SDC2 detection of colorectal cancer

Combining fecal measurements of SDC2 gene methylation with four serum tumor markers modestly improved the discrimination of colorectal cancer from other examination records in a retrospective hospital-based study, researchers reported. The full model achieved an area under the receiver operating characteristic curve, or AUC, of 0.872, compared with 0.814 for fecal SDC2 methylation alone. At the threshold selected by the Youden index, the combined approach had a sensitivity of 87.3%, a specificity of 78.6%, a positive predictive value of 42.9% and a negative predictive value of 97.1%. The results suggest that serum CEA and CA19-9 supplied most of the added information, while CA242 and CA50 contributed little further discrimination.

The analysis included 354 examination records from 325 people treated at the First People’s Hospital of Jiangxia District in Wuhan, China, between July 2021 and November 2025. The records comprised 55 colorectal cancer cases, 188 records involving adenomas or polyps, 106 non-tumor or inflammatory examinations and five with other or unclear lesions. The median age was 60 years, and 76.6% of records came from male patients. Because some individuals contributed examinations at different clinical time points, the investigators described the results primarily at the examination-record level rather than treating all 354 records as independent people.

Participants had no previous gastrointestinal surgery, anticancer treatment or history of inflammatory bowel disease. Fecal SDC2 methylation was measured with a commercial real-time fluorescence PCR workflow that included stool DNA extraction, bisulfite conversion and methylation-specific amplification. ACTB served as the internal control, and samples were considered analytically valid when its cycle threshold, or Ct, value was no greater than 36. The assay manufacturer’s interpretation classified SDC2 Ct values below 38 as methylation-positive and values of 38 or higher, or no detectable amplification, as negative. The investigators used continuous Ct values, rather than only these binary results, when constructing diagnostic models.

Blood samples were collected before colonoscopy or surgery, and serum CEA, CA19-9, CA242 and CA50 concentrations were measured with magnetic microparticle chemiluminescence immunoassays. The reference limits were 5.093 ng/mL for CEA, 37.0 U/mL for CA19-9, 20.0 U/mL for CA242 and 25.0 U/mL for CA50. Prespecified logistic regression models were built in nested fashion: fecal SDC2 alone; SDC2 plus CEA; SDC2 plus CEA and CA19-9; and a full model adding CA242 and CA50. No data-driven variable selection was applied, and the analyses included patient-grouped cross-validation, bootstrap optimism correction, calibration assessments, decision-curve analysis and comparison of correlated ROC curves with DeLong’s test.

Median fecal SDC2 Ct was lowest in the colorectal cancer group, indicating relatively higher detected methylation, at 34.95 with an interquartile range of 33.25 to 35.91. Corresponding values were 36.76 for adenoma or polyp records, 36.98 for non-tumor or inflammatory records and 35.77 for the five other or unclear cases. Serum CEA, CA19-9 and CA50 were also relatively higher in colorectal cancer. Correlations between SDC2 Ct and the serum markers were weak: Spearman coefficients were -0.1331 for CEA and -0.1119 for CA19-9, with no statistically significant correlations for CA242 or CA50. The investigators proposed that this limited correlation was consistent with the markers reflecting partly different biological processes, although the study did not directly test a mechanism.

In ROC analysis, fecal SDC2 alone produced an AUC of 0.814 with a 95% confidence interval of 0.751 to 0.876. The individual serum markers performed less well: AUCs were 0.709 for CEA, 0.666 for CA19-9, 0.495 for CA242 and 0.567 for CA50. Adding CEA to SDC2 increased the AUC to 0.866, with a 95% confidence interval of 0.814 to 0.918. Adding CA19-9 yielded an AUC of 0.872, and the complete five-predictor model also reached 0.872, with a 95% confidence interval of 0.818 to 0.926. CA242 and CA50 were not independently associated with colorectal cancer status in the full model, with P values of 0.784 and 0.335, respectively.

The absolute increase in AUC from 0.814 to 0.872 was 0.058, but its statistical significance was borderline by DeLong’s test, at P = 0.055. The study included only 55 colorectal cancer records for five predictors, approximately 11 events per predictor, and estimated post-hoc power for the observed AUC increment at 48%. Additional analyses nevertheless favored the full model. Its patient-grouped 10-fold cross-validated AUC was 0.854, while bootstrap optimism correction produced an estimated AUC of about 0.858. The cross-validated Brier score was 0.095 and the calibration slope was 0.864. Integrated discrimination improvement was 0.098, and continuous net reclassification improvement was 0.657, while decision-curve analysis showed generally greater net benefit than SDC2 alone across the examined threshold probabilities.

These findings should not be interpreted as validation of a population-screening test or as evidence of improved early-stage detection. The hospital-based cohort had a colorectal cancer prevalence of approximately 17%, far above that expected in a general screening population, which can inflate positive predictive values and limit direct generalization. Complete tumor-node-metastasis staging was unavailable, preventing stage-specific analysis, particularly for stage I and II disease. Lesion-level classification was also incomplete, so the investigators could not distinguish advanced adenomas, non-advanced adenomas and sessile serrated lesions. The results therefore establish exploratory discrimination among available examination records rather than performance for early colorectal cancer or advanced precancerous lesions.

Further limitations included the retrospective design, single-center setting, restricted geographic and ethnic diversity, absence of an independent external validation cohort and the possibility of overfitting from the small number of cancer cases. Information on tumor location, histological subtype, smoking, family history and comorbidity was not systematically available, and the adenoma or polyp group could not be classified consistently at the lesion level. Adjustment for age and sex left discrimination essentially unchanged, increasing the AUC from 0.872 to 0.879, but internal cross-validation and bootstrap adjustment cannot substitute for testing in an independent prospective population. The authors also noted that adding four serum assays would increase laboratory costs and workload, and patient-level cost, reimbursement, downstream diagnostic use and health-outcome data were unavailable for a formal economic analysis.

Established stool-based molecular screening platforms use different biomarkers, positivity algorithms, intended-use populations and regulatory frameworks, making direct numerical comparisons with this retrospective model inappropriate. The study therefore does not show that the combined panel is superior or equivalent to licensed multitarget stool DNA tests, nor does it establish a clinical role for blood-based screening. Prospective multicenter validation with prespecified sample-size planning, stage-specific analyses, advanced-precancerous-lesion endpoints, complete clinicopathological characterization and health-economic evaluation would be needed before considering implementation. As tested, fecal SDC2 methylation did most of the work in separating colorectal cancer from other records, while CEA and CA19-9 provided a modest incremental signal and CA242 and CA50 added essentially no further discrimination.

Subject of Research: Diagnostic performance of fecal SDC2 methylation combined with serum tumor markers for colorectal cancer discrimination

Article Title: Integrated fecal SDC2 methylation and serum tumor markers improve noninvasive detection of colorectal cancer

Article References: Tong, S., Rao, K., Sun, M., Wu, Y., Dong, Y., & Xiang, C. (2026). Integrated fecal SDC2 methylation and serum tumor markers improve noninvasive detection of colorectal cancer. Frontiers in Oncology, 16, Article 1919347. https://doi.org/10.3389/fonc.2026.1919347

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1919347

Keywords: colorectal cancer, SDC2 methylation, fecal biomarkers, serum tumor markers, CEA, CA19-9, noninvasive diagnosis, retrospective cohort, diagnostic accuracy, Integrated, fecal, SDC2