Combination Assay of Methylated HOXA1 with Tumor Markers for Detection of Early-Stage Hepatocellular Carcinoma

Dear Editor,

We read with great interest the article by Kunimune et al. [1] reporting a novel diagnostic index (ASDAm-H1) combining methylated HOXA1 (m-HOXA1) with conventional tumor markers for hepatocellular carcinoma (HCC) detection. While this innovative approach deserves attention, several methodological concerns require careful consideration.

First, the study’s control group composition demonstrates significant selection bias. The control group predominantly comprised healthy subjects (62.3%) rather than high-risk populations requiring HCC surveillance [2]. This proportion is substantially higher than in previous validation studies of HCC biomarkers, where control groups typically consist predominantly of patients with chronic liver disease. Notably, patients with cirrhosis, who represent the most needed target population for HCC screening, constituted only 12.6% of controls. While we acknowledge that the changing epidemiology of viral hepatitis may influence control group composition, recent validation studies have consistently maintained chronic liver disease patients as the primary control population to ensure clinically relevant specificity assessment. Additionally, no matching or adjustment for important confounding factors such as liver function parameters was performed.

Second, the diagnostic golden standard for HCC is questionable. While we acknowledge that an imaging-based diagnosis is widely accepted and recommended by current guidelines [2, 3], our concern focuses primarily on the lack of standardization in the study’s imaging criteria. Only 50.6% of HCC cases were pathologically confirmed, with the remainder relying on imaging criteria. The authors neither provided standardization details for imaging diagnosis nor analyzed potential differences in ASDAm-H1 performance between pathologically and radiologically diagnosed groups. This lack of detailed imaging criteria standardization could substantially impact the reported diagnostic accuracy.

Third, the remarkably high diagnostic accuracy of the ASDAm-H1 index (AUC 0.96) is questionable, given the modest performance of its individual components (AUC: AFP 0.77, DCP 0.82, m-HOXA1 0.82). For early-stage HCC, while the index achieved 76.3% sensitivity, the individual markers showed poor sensitivity (AFP 22.9%, DCP 50.9%, m-HOXA1 56.8%). Such dramatic improvement through combination lacks biological plausibility and requires more robust statistical validation [3, 4]. The perfect separation observed between early-stage HCC and controls (ASDAm-H1 index medians: 0.05 vs. 0.89) strongly suggests overfitting.

Fourth, the methodological validation of the 1-step CORD assay is insufficient. Validation using only HCT116 DNA serial dilutions inadequately represents clinical complexity. No analytical validation studies were performed to establish the assay’s limit of detection, precision, or analytical specificity. The model development lacks rigorous statistical validation methodology that is essential for biomarker studies of this nature [4]. Furthermore, the 17-h processing time for the 1-step CORD assay remains clinically impractical for routine screening. The authors failed to evaluate several crucial aspects: operational complexity, technical expertise requirements, batch-to-batch variation, inter-laboratory reproducibility, and most importantly, cost-effectiveness analyses [5]. The estimated reagent cost and labor intensity of this complex molecular testing would significantly impact its clinical implementation.

Fifth, the sample size limitations affect result reliability, particularly for early-stage HCC and non-viral HCC subgroups. The small numbers in non-viral HCC etiological subgroups (alcoholic n = 56, NAFLD n = 34) preclude meaningful conclusions about the C-index’s performance in these increasingly important populations. In addition, no formal sample size calculation was reported by the authors.

Sixth, a major concern is the analysis of BCLC stage D patients (n = 7). Given that BCLC stage D primarily reflects liver function and performance status rather than tumor burden, including this small subgroup in tumor marker analysis is inappropriate. We suggest either excluding this group or combining BCLC stages B-D (n = 151) as intermediate/advanced-stage HCC for subgroup analysis.

Last but not least, the heterogeneous non-viral HCC group (including alcoholic liver disease, NAFLD, and autoimmune hepatitis) was analyzed as a single entity, despite potentially different effects of various etiologies on tumor marker expression [6]. The small sample sizes in these non-viral subgroups severely limit result reliability and generalizability. In conclusion, while we acknowledge the innovative nature of the ASDAm-H1 index and its potential utility in HCC detection, addressing these methodological concerns through rigorous validation studies is essential before considering clinical implementation.