Comment on “Sequential FIB-4 Index and M2BPGi Combination Improve Detection of Advanced Fibrosis in MASLD: A MultiCenter Study”

Dear Editor,

We read with great interest the study by Morishita et al evaluating Mac-2 binding protein glycosylation isomer in biopsy-confirmed metabolic dysfunction–associated steatotic liver disease and proposing a sequential strategy with the Fibrosis-4 index to refine detection of advanced fibrosis.¹ The effort to optimize noninvasive triage pathways in a large multicenter cohort is timely and clinically relevant.

First, although Mac-2 binding protein glycosylation isomer achieved an area under the curve of 0.774 (95% confidence interval: 0.742–0.803) for advanced fibrosis, its sensitivity of 76.1% and specificity of 65.3% at the 0.99 threshold translate into a meaningful proportion of misclassified patients. Given the 25.6% prevalence of stage F3–F4 disease in this cohort, the reported negative predictive value of 88.3% still implies that approximately 1 in 8 patients with advanced fibrosis may remain undetected. In a condition where stage F3 already confers increased risk of hepatocellular carcinoma and liver-related events, even modest false-negative rates can delay surveillance enrollment and metabolic risk intensification.² Future analyses reporting absolute risk differences and number needed to misclassify would better contextualize whether the incremental gain over the Fibrosis-4 index (area under the curve: 0.719) is clinically sufficient to modify referral thresholds.

Second, Mac-2 binding protein glycosylation isomer demonstrated significant associations not only with fibrosis stage but also with steatosis, lobular inflammation, and ballooning. While this pleiotropic behavior reflects its biological link to hepatic stellate cell activation and inflammatory signaling, it raises concerns regarding construct specificity for fibrosis. The modest correlation between Mac-2 binding protein glycosylation isomer and the Fibrosis-4 index (r = 0.42) suggests partial overlap but also residual confounding by inflammatory activity.³ The multivariable model retained alanine aminotransferase and platelet count as independent predictors alongside Mac-2 binding protein glycosylation isomer, indicating that fibrosis discrimination partly relies on composite inflammatory signals.⁴ Stratified calibration analyses across alanine aminotransferase quartiles and diabetes status would help determine whether uniform cutoffs are appropriate or whether dynamic thresholds anchored to metabolic phenotype are required to preserve clinical reliability.

Third, the proposed 2-step algorithm reduced referrals by 36.5% while maintaining a false-negative rate of 13.9%. Although operational efficiency is appealing, the decision tree approach was derived and tested within the same dataset. Without internal resampling techniques such as bootstrapping to estimate optimism-corrected performance, decision thresholds may appear more stable than they would be in real-world deployment.⁵ Reporting calibration metrics, such as the Hosmer–Lemeshow statistic or calibration slope, would clarify whether predicted probabilities align with observed event rates across risk strata. From a health systems perspective, integration of cost-effectiveness modeling comparing this strategy with elastography-based pathways would further strengthen translational applicability.

Finally, advanced fibrosis is a surrogate endpoint for clinically meaningful outcomes. While histology remains the reference standard, prognostic validation against incident hepatic decompensation or hepatocellular carcinoma would provide stronger justification for embedding Mac-2 binding protein glycosylation isomer into population-level screening algorithms.⁶ Longitudinal assessment of biomarker dynamics and their association with fibrosis regression or progression could determine whether this marker is suitable not only for detection but also for disease monitoring in therapeutic trials.

These considerations may help refine risk stratification frameworks and guide the integration of serum-based biomarkers into scalable care pathways for metabolic dysfunction–associated steatotic liver disease.

Declaration of Generative AI and AI-Assisted Technologies in the Writing Process

During the preparation of this work, the authors used Grammarly and ChatGPT in order to refine language, grammar, and style. After using this tool/service, the authors reviewed and edited the content as needed and takes full responsibility for the content of the publication.