Zhejiang University index predicts metabolic dysfunction-associated steatotic liver disease in type 2 diabetes mellitus patients

Abstract

The Zhejiang University (ZJU) index, which combines body mass index, fasting blood glucose, triglyceride level and alanine aminotransferase/aspartate aminotransferase ratio, can be used to predict metabolic dysfunction-associated steatotic liver disease (MASLD) in patients with type 2 diabetes mellitus (T2DM). The ZJU index of 38.87 has been identified as the key threshold for diagnosing MASLD. The new model for predicting MASLD in T2DM based on ZJU index shows high diagnostic value. While the study is methodologically robust and offers a valuable clinical tool, it is limited by its cross-sectional design, inpatient cohort bias, unadjusted pharmacotherapy effects, and reliance on ultrasound for MASLD diagnosis. Future validation in outpatient settings, incorporating medication data and advanced fibrosis assessment, is crucial to translate this cost-effective biomarker into wide practice.

Key Words: Zhejiang University index; Metabolic dysfunction-associated steatotic liver disease; Type 2 diabetes mellitus; Predictive biomarker; Cross-sectional study; Pharmacotherapy confounding

Core Tip: This correspondence critically evaluates the study by Tao et al on the ability of the Zhejiang University index to predict metabolic dysfunction-associated steatotic liver disease (MASLD) in type 2 diabetes mellitus patients. It commends the index as it is a minimally invasive biomarker panel and methodologically sound but highlights several limitations, such as the reliance on a hospitalized patient sample, lack of information on pharmacological treatments, the inherent constraints of ultrasound imaging, and the cross-sectional nature of the study. The authors suggest that future research should aim at prospective validation in outpatient settings and exploring the underlying mechanisms of MASLD.

TO THE EDITOR

We have read with great interest the pivotal study by Tao et al[1], entitled “Association between Zhejiang University index and metabolic dysfunction-associated steatotic liver disease in patients with type 2 diabetes mellitus”, recently published in the World Journal of Diabetes. This rigorously designed cross-sectional investigation explores the efficacy of the Zhejiang University (ZJU) index[2] as a predictive tool for metabolic dysfunction-associated steatotic liver disease (MASLD) in a sample of 688 individuals diagnosed with type 2 diabetes mellitus (T2DM). The research is both timely and significant, considering the rising global incidence of T2DM and MASLD and the complex, bidirectional pathophysiological interplay between these conditions. The authors report that the ZJU index exhibits an independent and positive correlation with MASLD risk, establish a critical cutoff value of 38.87, and develop a novel multivariate predictive model that achieves enhanced diagnostic performance [area under the curve (AUC) = 0.76] relative to individual metabolic indicators.

We extend our heartfelt commendations to the authors for their significant contribution to the domain of metabolic medicine. The development of simple, minimally invasive, and cost-efficient methodologies for the early detection and risk stratification of MASLD in high-risk groups, such as individuals with T2DM, remains a critical clinical priority. The ZJU index, which integrates commonly measured clinical parameters, including body mass index (BMI), fasting blood glucose (FBG), triglycerides, and the alanine aminotransferase (ALT) to aspartate aminotransferase (AST) ratio—into a composite score, offering a practical solution to this challenge. The research effectively adapts and validates this index within the T2DM population, where the clinical demand for such tools is particularly pronounced.

The methodological rigor demonstrated in this study is noteworthy. The application of restricted cubic spline analysis to characterize the non-linear association between the ZJU index and the prevalence of MASLD represents a significant strength. This technique transcends simplistic linear assumptions, facilitating a more refined understanding of the risk continuum. This threshold aligns with findings from prior research conducted within general populations, thereby underscoring the robustness and potential generalizability of the metric[3,4]. Additionally, the construction of a composite predictive model that integrates the ZJU index with other pertinent covariates—including alcohol consumption, age, white blood cell (WBC) count, lipid parameters (total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), uric acid, and blood urea nitrogen—substantially improves predictive accuracy. The employment of advanced statistical validation methods, such as bootstrapping (resulting in an AUC of 0.73), calibration plots, and decision curve analysis, provides strong evidence for the model’s internal validity and clinical applicability. Moreover, the development of a nomogram offers a practical and accessible tool for clinicians to estimate individual patient risk in clinical settings.

Notwithstanding these significant strengths, we seek to engage in a scholarly discourse concerning several elements of the study that merit additional examination. This dialogue aims to enhance the interpretation of the results and to outline directions for future research that can extend this robust groundwork.

Study population and generalizability: The spectrum of T2DM

The study sample consisted exclusively of hospitalized patients diagnosed with T2DM. Although this approach enables comprehensive data acquisition, it inherently restricts the cohort to a subgroup characterized by potentially more advanced disease severity, suboptimal glycemic control, and a higher prevalence of comorbid conditions (e.g., hypertension at 52.33%, coronary artery disease at 10.32%) compared to the broader T2DM population typically managed in primary care or outpatient environments. This selection bias may result in an overestimation of both the prevalence of MASLD and the predictive accuracy of the ZJU index. The observed MASLD prevalence of 49.85%, while credible within a hospitalized cohort, may not accurately reflect the prevalence in the wider T2DM community. For instance, the latest research shows that among individuals with T2DM the prevalence of MASLD was approximately 6.5% in the Valencian Community region of Spain[5], which was significantly lower than the 49.85% prevalence in this study. Accordingly, the diagnostic performance metrics of the ZJU index and the proposed predictive model—including sensitivity, specificity, and AUC, may vary when applied to healthier, community-based T2DM populations or individuals with newly diagnosed diabetes. Therefore, future validation studies involving outpatient cohorts are imperative to ascertain the external validity of these findings and to delineate the clinical utility of the ZJU index across diverse healthcare settings, potentially extending its application to population-level screening initiatives.

Unmeasured confounder: Influence of pharmacotherapy

A significant limitation identified by this research pertains to the absence of data regarding concomitant medication use. The management of T2DM is complex, as the patients are frequently prescribed pharmacological agents that substantially affect the parameters underlying the ZJU index as well as the pathophysiology of MASLD. The frequently prescribed pharmacological agents are as follows: (1) Metformin: As the first-line treatment for T2DM, metformin is recognized for its capacity to reduce hepatic fat accumulation[6] and enhance insulin sensitivity[7]; (2) Pioglitazone: This potent insulin sensitizer is specifically recommended for T2DM patients with biopsy-confirmed MASLD and has a marked impact on ALT levels and hepatic steatosis[8]; (3) GLP-1 receptor agonists (e.g., liraglutide, semaglutide): These agents facilitate weight reduction, improve glycemic control, and have demonstrated direct effects in decreasing liver fat content and inflammation[9]; (4) SGLT2 inhibitors: Similarly, these medications promote weight loss and glycemic improvement and are associated with favorable changes in liver enzyme profiles and steatosis[10]; and (5) Statins and fibrates: Commonly prescribed for lipid regulation, these drugs influence triglyceride and LDL-C levels and may exert modest effects on liver histology.

The failure to account for these pharmacotherapies constitutes a considerable confounding factor. It is plausible that the observed correlation between elevated ZJU index values and MASLD is mitigated among patients receiving effective MASLD-targeted treatments. For instance, GLP-1 receptor agonists and SGLT2 inhibitors significantly reduce body weight and FBG, directly lowering the BMI and FBG components of the ZJU index. Pioglitazone can markedly normalize ALT levels, directly affecting the ALT/AST component. Therefore, in a treated population, the ZJU index might be artificially lowered, potentially underestimating the true MASLD risk, whereas the association might appear stronger in untreated patients. This differential effect constitutes a significant unmeasured confounder. Therefore, future research should rigorously document and adjust for medication use to accurately determine the independent predictive validity of the ZJU index.

Diagnostic criteria for MASLD: Limitations of ultrasonography

While ultrasonography is a pragmatic and widely available tool for large-scale studies like this one, it is important to acknowledge its limitations. Ultrasonography is subject to well-recognized limitations. Its sensitivity markedly decreases when hepatic steatosis is below 20%-30% fat infiltration, which may result in the misclassification of individuals with mild MASLD as non-MASLD cases. Additionally, ultrasonography does not provide quantitative assessment of steatosis severity and, critically, is inadequate for evaluating liver fibrosis—the principal factor influencing long-term liver-related morbidity and mortality in MASLD.

The integration of more advanced, yet still minimally or non-invasive, diagnostic modalities could have enriched the analysis substantially, for instance: (1) Controlled attenuation parameter, utilized in conjunction with vibration-controlled transient elastography (VCTE), offers a quantitative evaluation of hepatic steatosis[11]; and (2) Serum-based fibrosis indices, such as the fibrosis-4 index (FIB-4) index or the non-alcoholic fatty liver disease fibrosis sibrosis score, derived from routine clinical data, could facilitate examination of the relationship between the ZJU index and not only steatosis but also fibrosis risk within the study population[12].

While acknowledging the impracticality of liver biopsy in a study of this magnitude, future investigations might consider stratifying patients by both MASLD presence and fibrosis stage using these techniques. Such an approach would enable assessment of whether the ZJU index correlates with disease severity in addition to disease presence. Even with the ultrasound-defined MASLD diagnosis, the authors could potentially explore the correlation between the ZJU index and established minimally invasive fibrosis scores like the FIB-4 index, which can be calculated from age, ALT, AST, and platelet count, the parameters already available in their dataset.

Cross-sectional design and the issue of causality

The cross-sectional design employed in this study constitutes a fundamental limitation, as it permits the identification of associations but precludes the determination of causal relationships or directionality. A critical question remains unresolved: Does an elevated ZJU index contribute to the onset of MASLD, or does the presence of advanced MASLD—characterized by hepatic insulin resistance and inflammation—result in an increased ZJU index? It is plausible that a complex, bidirectional feedback loop exists between these factors. To elucidate this relationship, longitudinal prospective cohort studies are necessary to ascertain whether baseline or increasing ZJU index values can serve as predictors for the future development of MASLD in individuals with T2DM who are initially free of the condition. Furthermore, such studies would be instrumental in evaluating whether interventions aimed at reducing the ZJU index, such as weight reduction and improved glycemic control, effectively reduce the risk of MASLD onset or progression.

External validation and model parsimony

Although the use of bootstrapping for internal validation represents a methodological strength, it does not obviate the need for external validation. To rigorously evaluate the robustness and generalizability of the proposed multivariate model, its predictive performance should be assessed in an independent cohort of patients with T2DM, preferably drawn from a distinct geographic region or healthcare system. Additionally, while the model demonstrates a satisfactory performance, its complexity—incorporating up to ten variables in the full specification—may hinder its feasibility in time-constrained clinical settings. A comparative analysis quantifying the incremental predictive value contributed by each variable beyond the core ZJU index would provide valuable insights. It is conceivable that a more parsimonious model, comprising the ZJU index alongside one or two of the most influential predictors (such as age and HDL-C), could achieve comparable predictive accuracy, thereby facilitating broader clinical implementation. For example, the receiver operating characteristic curve AUC of the MASLD prediction model incorporating 10 variables such as the ZJU index, age and HDL-C was 0.76. If the AUC value of the prediction model incorporating three variables such as the ZJU index, age and HDL-C can reach 0.7. This raises the question if this model with less data will be more convenient for clinical use.

In-depth examination of mechanisms and specificity

The current correspondence appropriately identifies the ZJU index as an indicator of insulin resistance. Nonetheless, a more comprehensive analysis of the underlying pathophysiological mechanisms connecting the index’s components to MASLD in T2DM would enhance the depth of the discourse. For example, the ALT/AST ratio serves not only as a biomarker of hepatic cellular injury but may also signify liver fibrosis[13]. The incorporation of the WBC count in the final analytical model further underscores the contributory role of inflammatory processes. It would be valuable to ascertain whether additional inflammatory biomarkers, such as high-sensitivity C-reactive protein or pro-inflammatory cytokines, were assessed. Investigating their associations with the ZJU index could provide meaningful insights and represent a promising direction for future research.

Furthermore, although the study primarily focuses on MASLD, it would be valuable to evaluate the performance of the ZJU index in differentiating MASLD from other types of liver diseases in individuals with diabetes. Despite the implementation of exclusion criteria intended to reduce confounding, conducting such analyses could enhance the specificity of the index for MASLD in the context of T2DM.

Conclusion and future directions

In summary, the study conducted by Tao et al[1] merits commendation for its insightful and clinically pertinent contributions. The authors have presented robust evidence supporting the ZJU index as a reliable and effective marker for assessing the risk of MASLD in individuals with T2DM. Their research successfully validates the utility of this index within a high-risk patient population and introduces a practical multivariate tool that demonstrates superior predictive performance compared to individual metabolic parameters.

The observations outlined herein are not intended to diminish the significance of these findings but rather to identify avenues for further refinement and broader clinical application. We advocate for the following future research directions: (1) Prospective validation: Implementation of longitudinal cohort studies to confirm the predictive capacity of the ZJU index for MASLD onset and disease progression; (2) External validation: Evaluation of the model’s generalizability through testing in independent external cohorts, including those in outpatient clinical settings; (3) Medication adjustment: Integration of comprehensive pharmacotherapy data in subsequent analyses to account for potential confounding effects of medication use; (4) Advanced disease staging: Incorporation of non-invasive fibrosis assessment techniques, such as FIB-4 scoring and VCTE, to examine correlations between the ZJU index and disease severity; (5) Mechanistic investigations: Exploration of the associations between the ZJU index and biological mediators including adipokines, inflammatory cytokines, and gut microbiota, to elucidate underlying pathophysiological mechanisms; (6) Dynamic monitoring: Beyond a single baseline measurement, future studies could investigate whether longitudinal changes in the ZJU index track with the progression or regression of MASLD in response to therapeutic interventions, thereby enhancing its utility as a dynamic monitoring tool; and (7) Mechanistic exploration: To move beyond correlation towards mechanism, integrating the ZJU index with omics data (e.g., metabolomic or proteomic profiles from the included patients) could uncover novel biological pathways linking insulin resistance, hepatic inflammation, and steatosis in T2DM.

In conclusion, we commend Tao et al[1] for their valuable contribution, which robustly validates the ZJU index in a high-risk T2DM inpatient cohort. The critical discussions in this correspondence regarding generalizability, confounders, and diagnostic depth are not meant to diminish these findings but to chart a path forward. Addressing these aspects in future work will be crucial to translate this promising, cost-effective tool from a research setting into widespread clinical practice, ultimately enabling personalized risk stratification and early intervention for T2DM patients at risk of MASLD.