Development and validation of a high-confidence diagnostic model integrating ctDNA methylation and serum biomarkers for early-stage hepatocellular carcinoma detection

TL;DR

A new diagnostic model called GAMAD improves early detection of liver cancer by combining blood tests and DNA methylation analysis.

Contribution

GAMAD integrates ctDNA methylation with serum biomarkers for enhanced early HCC detection in high-risk populations.

Findings

GAMAD achieved 86.5% sensitivity for early-stage HCC with an AUC of 0.952.

GAMAD outperformed existing models like GALAD in both validation and test cohorts.

HepaAiQ alone showed 74.6% sensitivity and 88.1% specificity for HCC detection.

Abstract

Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related mortality, with significantly improved prognosis when detected early in high-risk populations. Current serum biomarkers show limited sensitivity for early-stage disease. We developed and validated GAMAD, a multimodal diagnostic model integrating circulating tumor DNA (ctDNA) methylation with established serum markers to enhance early HCC detection in China. In this multicenter prospective trial, a total of 1,692 patients were enrolled: 476 with HCC, 645 with hepatitis, 443 with cirrhosis, and 128 with no detectable liver abnormalities. Blood tests including AFP, AFP-L3, DCP, and ctDNA methylation (HepaAiQ) were performed. Using training, validation, and independent test cohorts, we developed the GAMAD model by integrating HepaAiQ with gender, age, AFP, and DCP. HepaAiQ demonstrated a superior performance compared with AFP, DCP, and AFP-L3 with a sensitivity of 74.6%, specificity of 88.1%, and an area under the curve (AUC) of 0.862 (95% CI, 0.842–0.883) in all samples. The GAMAD model, optimized specifically for early-stage HCC, achieved sensitivity of 80.5%, specificity of 90.4%, and AUC of 0.934 (95% CI, 0.911–0.957) in the validation cohort. In the independent test cohort, GAMAD showed superior performance with sensitivity of 86.5% for stage 0/A and 91.7% for stage B-C HCC (AUC 0.952, 95% CI, 0.931–0.973), substantially outperforming the GALAD model across all cohorts. The GAMAD model represents a significant clinical advancement for early HCC detection. Its intended clinical use is to serve as an adjunct to standard imaging for high-risk populations, substantially increasing early diagnosis rates.

The online version contains supplementary material available at 10.1186/s43556-026-00426-3.

Figures