Prediction of MASLD using different screening indexes in Chinese type 2 diabetes mellitus

Abstract

Background

Formerly known as non-alcoholic fatty liver disease (NAFLD), metabolic dysfunction-associated steatotic liver disease (MASLD) has now become the most widespread chronic liver disease worldwide. The primary goal of this study is to assess the ability of different indexes (including VAI, TyG, HOMA-IR, BMI, LAP, WHtR, TyG-BMI, TyG-WC, and TyG-WHtR) to predict MASLD in individuals diagnosed with type 2 diabetes mellitus (T2DM), particularly within the Chinese population.

Methods

This cross-sectional study involved 1,742 patients with T2DM, recruited from the Metabolic Management Centers (MMC) at Suzhou Municipal Hospital. Abdominal ultrasonography was employed for MASLD diagnosis in patients with T2DM. The predictive accuracy of various screening indexes for MASLD in the Chinese T2DM population was evaluated using logistic regression and receiver operating characteristic (ROC) curve analyses.

Results

Among the 1,742 participants, 996 were diagnosed with MASLD. After adjusting for potential confounding factors, positive associations with the risk of MASLD were found for all the nine indexes. The lipid accumulation product (LAP) exhibited the greatest predictive value for detecting MASLD, with an area under the curve (AUC) of 0.786(95%CI 0.764,0.807), followed by BMI(AUC = 0.785), VAI(AUC = 0.744), TyG(AUC = 0.720), WHtR(AUC = 0.710) and HOMA-IR(AUC = 0.676). The composite Indexes (TyG-BMI, TyG-WC, TyG-WHtR) also showed considerable predictive ability with AUCs of 0.765, 0.752 and 0.748, respectively.

Conclusion

Our results indicated that all nine indexes have favorable correlations with the risk of MASLD, and most of them have a good performance in predicting MASLD. According to our study, LAP was a reliable index for predicting MASLD among Chinese T2DM patients. The exploration of non-invasive screenings will provide significant support for the early detection and diagnosis of MASLD.

Introduction

Nonalcoholic fatty liver disease (NAFLD) ranks among the most prevalent hepatic disorders, with its incidence exhibiting a continuous global increase [1]. NAFLD includes various liver conditions, ranging from mild steatosis and progressing to more severe forms like nonalcoholic steatohepatitis (NASH), fibrosis, and cirrhosis, and it impacts around 25.24% of adults worldwide [2]. NAFLD is closely linked with metabolic dysfunctions and is often regarded as a hepatic manifestation of metabolic syndrome [3, 4].

In recent years, with the gradual updates to the diagnostic criteria for NAFLD, the international academic community has advocated for the use of the term “MASLD” (Metabolic Dysfunction-associated Steatotic Liver Disease) [5]. This term is considered more precise and comprehensive, as it better encompasses the metabolic background and clinical features of the disease [6].

MASLD is a systemic disease affecting multiple organs and demonstrates a bidirectional relationship with insulin resistance (IR) and other clinical manifestations of metabolic dysregulation [7]. There is a particularly strong association between MASLD and type 2 diabetes mellitus (T2DM), with approximately 55.5% of individuals with T2DM presenting with MASLD [8]. This high prevalence is attributed to shared pathophysiological mechanisms, such as insulin resistance, chronic inflammation, and dyslipidemia [9]. Moreover, the presence of MASLD in T2DM patients is associated with an increased risk of developing microvascular complications such as chronic kidney diseases and cardiovascular diseases among others. MASLD can also exacerbate metabolic dysregulations in T2DM, leading to poor glycemic control and complicating T2DM management [3].

Early detection and identification of MASLD is crucial for effective management and treatment. While liver biopsy is widely recognized as a reliable diagnostic method for MASLD, this method is expensive, invasive, and prone to postoperative complications [10]. Additionally, imaging examinations are less invasive, but are constrained by challenges related to accessibility and cost. Due to the multifactorial origins of these conditions, which result from a mix of genetic, metabolic, and environmental factors, accurately predicting the risk of MASLD in individuals with T2DM remains a complex task. Therefore, investigating accessible, reliable, and practical predictors of MASLD holds clinical significance and value. The development of non-invasive biomarkers and affordable, accurate diagnostic tools is essential for early detection of MASLD progression [11].

Several anthropometric and metabolic indexes have been proposed as potential predictors of MASLD. It is widely recognized that BMI serves as a standard metric for assessing obesity. WC and WHtR are key measures for evaluating central obesity. The visceral adiposity index (VAI) can indicate fat distribution and function, only with simple anthropometric and functional parameters [12]. Yu et al. found significant correlations between VAI and T2DM [13]. The lipid accumulation product (LAP) is a simple index based on waist circumference and triglyceride levels [14]. Recent research indicates that LAP as a robust indicator for insulin resistance, diabetes, and metabolic syndrome in general populations [15, 16]. According to Dai et al., LAP was significantly correlated to the severity and incidence of NAFLD, establishing its reliability as a predictor of NAFLD risk among Chinese adults [17]. The triglyceride-glucose (TyG) index, calculated with triglycerides and fasting blood glucose, has recently been suggested as an efficient but simple marker of insulin resistance [18]. According to a meta-analysis, the TyG index has good performance in MASLD diagnostic and prediction [19].

While previous studies have investigated the relationship between these indices and MASLD, our study specifically focuses on patients with T2DM, and no direct comparison of their predictive accuracy has been conducted in this population. This research seeks to fill this gap by assessing the predictive accuracy of BMI, WC, VAI, WHtR, LAP, TyG, TyG-WC, TyG-BMI, TyG-WHtR for MASLD risk in T2DM patients. Our results will offer dependable non-invasive indicators for the early diagnosis and recognition of MASLD.

Methods

Study Population

A total of 1,742 patients diagnosed with T2DM were enrolled in this study from the Department of Endocrinology, Suzhou Municipal Hospital of Nanjing Medical University. Since the study was designed to collect data retrospectively from participants’ medical records, the requirement for informed consent was waived. Subjects with hepatic ultrasonography measurements were included in the study. The patients were categorized into two groups based on abdominal ultrasonography results: MASLD and non-MASLD groups. Patients with acute complications of diabetes, other types of diabetes, severe chronic kidney disease (CKD) defined as an estimated glomerular filtration rate (eGFR) ≤ 30 mL/min/1.73 m², or other severe coexisting illnesses were excluded from the study.

Definition of T2DM and MASLD

Diagnosis of type 2 diabetes followed the 1999 World Health Organization guidelines, which define it as fasting glucose levels of ≥ 7.0 mmol/L(126 mg/dL) and postprandial blood glucose levels of ≥ 11.1 mmol/L(200 mg/dL). Every participant was subjected to an abdominal ultrasound procedure, performed by an experienced sonographer from our hospital’s Ultrasound Department. The following ultrasound findings must be met for the imaging diagnosis of fatty liver: ambiguous intrahepatic tube structure, high echo in the liver’s proximal diffusing point, and higher echo intensity in the liver compared to the kidney. The diagnosis of MASLD in this study was based on the presence of type 2 diabetes and fatty liver determined by ultrasonography [20].

Data collection

Trained medical professionals gathered general information, such as age, gender, diabetes duration, history of smoking, hypertension, and diabetic complications. Body composition measurements like weight, height, hip circumference, waist circumference and BMI were also recorded. All participants underwent comprehensive laboratory tests following an 8-hour fasting period, typically conducted in the morning. These tests included measurements of fasting blood glucose, fasting insulin, fasting C-peptide, hemoglobin A1c (HbA1c), triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), low-density lipoprotein cholesterol (LDL-c), and liver function tests including alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma-glutamyl transferase (γ-GT). Other relevant parameters were also measured as per the requirements of the study.

Smokers were classified as individuals who smoked daily or almost daily for a minimum of one year. Drinkers were classified as individuals who drank alcohol weekly or almost weekly for a minimum of one year. Educational attainment was categorized by no high school degree or high school degree and above.

These measurements were used to calculate various metabolic indexes:

$$\:\text{B}\text{M}\text{I}=\text{w}\text{e}\text{i}\text{g}\text{h}\text{t}\left(\text{k}\text{g}\right)/{\text{h}\text{e}\text{i}\text{g}\text{h}\text{t}}^{2}\left(\text{m}\right)$$

$$\eqalign{& \:{\text{VAI}} = \frac{{{\text{WC}}\left( {{\text{cm}}} \right)}}{{39.68 + 1.88 \times \:{\text{BMI}}({\text{kg}}/{m^2})}} \times \:\frac{{{\text{TG}}({\text{mmol}}/{\text{L}})}}{{1.03}} \times \: \cr & \frac{{1.31}}{{{\text{HDL}} - {\text{C}}({\text{mmol}}/{\text{L}})}} \cr}$$

for males

$$\eqalign{& \:{\text{VAI}} = \frac{{{\text{WC}}\left( {{\text{cm}}} \right)}}{{36.58 + 1.89 \times \:{\text{BMI}}({\text{kg}}/{m^2})}} \times \:\frac{{{\text{TG}}({\text{mmol}}/{\text{L}})}}{{0.81}} \times \: \cr & \frac{{1.52}}{{{\text{HDL}} - {\text{C}}({\text{mmol}}/{\text{L}})}} \cr}$$

for females

$$\:\text{L}\text{A}\text{P}=\left[\text{W}\text{C}\left(\text{c}\text{m}\right)-65\right]\times\:\text{T}\text{G}(\text{m}\text{m}\text{o}\text{l}/\text{L})\:\:\text{f}\text{o}\text{r}\:\text{m}\text{a}\text{l}\text{e}\text{s}\:\:\:\:\:\:\:\:\:\:\:\:\:\:\:$$

$$\:\text{L}\text{A}\text{P}=\left[\text{W}\text{C}\right(\text{c}\text{m})-58]\times\:\text{T}\text{G}(\text{m}\text{m}\text{o}\text{l}/\text{L})\:\text{f}\text{o}\text{r}\:\text{f}\text{e}\text{m}\text{a}\text{l}\text{e}\text{s}\:\:$$

$$\:\text{W}\text{H}\text{t}\text{R}=\frac{\text{W}\text{C}\left(\text{c}\text{m}\right)}{\text{h}\text{e}\text{i}\text{g}\text{h}\text{t}\left(\text{c}\text{m}\right)}$$

$$\:\text{T}\text{y}\text{G}=\text{ln}\left[\frac{\text{T}\text{G}(\text{m}\text{g}/\text{d}\text{L})\times\:\text{F}\text{P}\text{G}(\text{m}\text{g}/\text{d}\text{L})}{2}\right]$$

$$\:\text{T}\text{y}\text{G}-\text{B}\text{M}\text{I}=\text{T}\text{y}\text{G}\times\:\text{B}\text{M}\text{I}\:\:\:$$

$$\:\text{T}\text{y}\text{G}-\text{W}\text{C}=\text{T}\text{y}\text{G}\times\:\text{W}\text{C}\left(\text{c}\text{m}\right)\:\:$$

$$\:\text{T}\text{y}\text{G}-\text{W}\text{H}\text{t}\text{R}=\text{T}\text{y}\text{G}\times\:\text{W}\text{H}\text{t}\text{R}\:$$

Statistical analysis

The IBM SPSS Statistics 26 software was used for all analyses. The forest plot was drawn with GraphPad Prism 10. Mean ± standard deviation (SD) was used for normally distributed data, while abnormally distributed data were presented as the median and interquartile range and categorical variables were expressed by frequency and proportion.Based on their MASLD status, study participants’ characteristics were examined using the chi-square test for categorical data and the Mann-Whitney U-test for comparing continuous variables. All tests were two-tailed, and a significance level of p < 0.05 was applied. Then, in order to investigate the relationship between these characteristics and MASLD, we converted the variables into quartiles and performed binary multivariable logistic regression. The predictive value of these indicators for MASLD was assessed using ROC curves. Finally, differences between AUCs were tested using DeLong’s test.

Results

Characteristics of the participants

Table 1 provides a summary of the clinical and demographic features of the participants, categorized by the presence of MASLD. The study included 1,742 patients with T2DM, among whom 996 were diagnosed with MASLD, and 746 were not.

Participants with MASLD were generally younger (median age: 54 years) compared to those without MASLD (median age: 60 years) (P < 0.001). The distribution of genders showed no significant differences between the groups (P = 0.704). A higher percentage of participants with MASLD had completed high school or higher education (51.41% vs. 44.37%, P = 0.004). Additionally, more MASLD patients were smokers (35.44% vs. 30.03%, P = 0.018). There was no notable difference in alcohol consumption between the groups(P = 0.112).

Participants with MASLD had higher HbA1c (9.7% vs. 9.3%, P = 0.031), fasting blood glucose (7.455 mmol/L vs. 6.755 mmol/L, P < 0.001), and fasting insulin levels (6.695 µIU/mL vs. 3.915 µIU/mL, P < 0.001). Although the values of liver enzymes were statistically different in the two groups, the values were low and within normal limits: ALT (28 IU/L vs. 17 IU/L, P < 0.001), AST (21 IU/L vs. 16 IU/L, P < 0.001), and γ-GT (34 IU/L vs. 21 IU/L, P < 0.001). There was no significant different of ALP levels between participants with or without MASLD. (P = 0.227). It means that liver enzymes are not satisfactory indicators for assessing the progression of MASLD [11].

MASLD patients had significantly higher triglycerides (1.945 mmol/L vs. 1.150 mmol/L, P < 0.001), total cholesterol (4.73 mmol/L vs. 4.51 mmol/L, P < 0.001), and LDL-C (3.07 mmol/L vs. 2.85 mmol/L, P < 0.001), and lower HDL-C levels (1.04 mmol/L vs. 1.18 mmol/L, P < 0.001). MASLD patients also had significantly higher weight (73.2 kg vs. 63.7 kg, P < 0.001), waist circumference (95 cm vs. 88 cm, P < 0.001), and hip circumference (99 cm vs. 95 cm, P < 0.001).

Indexes of obesity and metabolic dysfunction were markedly elevated in the MASLD group: VAI (3.05 vs. 1.62, P < 0.001), HOMA-IR (2.27 vs. 1.20, P < 0.001), BMI (26.71 vs. 23.60, P < 0.001), LAP (65.4 vs. 30.0, P < 0.001), WHtR (0.57 vs. 0.53, P < 0.001), and TyG and its variants (P < 0.001 for all).

Table 1 Demographic and clinical characteristics of the cohort by MASLD

Relationship between MASLD occurrence and various parameters

Table 2; Fig. 1 display significant associations between various metabolic indexes and MASLD presence, as revealed by logistic regression models. After adjusting for potential confounders, elevated levels across the nine quartile-stratified parameters were associated with an increased risk of MASLD (S1 Table).

In model 1, individuals in the top quartile of all indexes had a significantly higher probability of developing MASLD compared to those in the lower quartile, after adjusted for gender and age. Notably, the Triglyceride-Glucose Body Mass Index (TyG-BMI) and Lipid Accumulation Product (LAP) exhibited the highest odds ratios (OR). Specifically, those in the top quartile of TyG-BMI and LAP exhibited a 16.981-fold (95% CI: 11.863, 24.308) and 16.609-fold (95% CI: 11.693, 23.592) increased risk of MASLD, respectively. Although the Homeostatic Model Assessment for Insulin Resistance (HOMA-IR) had the lowest odds ratio, it showed a 5.83-fold (95% CI: 4.297, 7.911) increased risk of MASLD between those in the upper and lower quartiles of HOMA-IR.

In model 2, which was adjusted for traditional risk factors, including gender, age, smoking status, drinking status and education levels, these significant associations persisted and were even slightly enhanced. In this model, LAP remained the most robust predictor, with individuals in the higher quartile exhibiting a 16.64-fold (95% CI: 11.702, 23.663) higher likelihood of developing MASLD compared to those in the lower quartile (P < 0.001).

These relationships are depicted in the forest plot (Fig. 1), illustrating how increased risk of MASLD is associated with upper quartile of these metabolic indexes.

Table 2 The adjusted logistic regression model of nine parameters with MASLD in type 2 diabetes mellitus

Fig. 1

Odds Ratios (95% CI) for MASLD in Type 2 Diabetes According to the Upper Quartile of Nine Predictive Indexes. Model 1: adjusted for age and gender; Model 2: adjusted for traditional risk factors including gender, age, smoking status, drinking status and education levels

Diagnostic predictive value of the nine indexes for MASLD

The predictive ability of all the indexes was analyzed using ROC curves (S1 Figure). For each metric, the Area Under the Curve (AUC), 95% confidence interval (CI), threshold value, sensitivity, and specificity were calculated (Table 3). To enhance the clarity of the results, we re-conducted the ROC curve analysis using LAP, VAI, HOMA-IR, and TyG-BMI. Additionally, given that LAP and TyG-BMI exhibited the highest AUC values, we further evaluated the predictive power of their combination (Fig. 2).

The ROC analysis demonstrated that LAP is the most reliable predictor with an AUC of 0.786, an optimal threshold of 42.061, a sensitivity of 74.3%, and a specificity of 70.0%. TyG-BMI ranked second with an AUC of 0.765. However, the AUC of the combined index did not increase as expected and remained below that of LAP, with a value of 0.771. Furthermore, BMI, TyG-WC, TyG-WhtP and VAI exhibited notable predictive capabilities, with AUCs of 0.75, 0.752, 0.748 and 0.744, respectively.

Fig. 2

Receiver operating characteristic (ROC) curves and the area under the ROC curve (AUC) values of the several parameters in diagnosing metabolic-associated fatty liver disease (MASLD)

Table 3 Evaluation of nine indexes for MASLD Prediction: AUC(95%CI), cut-off value, sensitivity and specificity

Comparison in AUC values between LAP and other indexes

To compare the diagnostic predictive value of different metabolic indexes for MASLD, DeLong’s test was performed to evaluate the AUC for each index (Table 4). The results indicated that LAP exhibited a significantly higher AUC in comparison to all other indexes, including VAI, HOMA-IR, BMI, WHtR, TyG, TyG-BMI, TyG-WC, and TyG-WHtR. The differences in AUC ranged from 0.020 to 0.110, all of which were statistically significant with p-values less than 0.05. The DeLong analysis underscores the LAP index’s reliability in diagnosing MASLD among T2DM patients, emphasizing its potential as a reliable and accurate marker for early detection and management of MASLD.

Table 4 DeLong’s test for two correlated ROC curves: LAP Versus other indexes

Discussion

In the present study, we observed strong and positive associations between VAI, TyG, HOMA-IR, BMI, LAP, WHtR, TyG-BMI, TyG-WC, TyG-WHtR and risk of MASLD, after adjustment for potential confounders. We demonstrated that LAP was better than other indexes for identifying MASLD in the Chinese T2DM patients. Triglyceride glucose index–related parameters also showed good predictive ability for MASLD. Then because of the highest AUC values of LAP and TyG-BMI, we further evaluated the predictive power of their combination. The predictive ability of this combined index was not as strong as that of LAP. In conclusion, our study suggests that LAP holds strong predictive value for determining the presence of MASLD in patients with T2DM, providing support for the early detection and diagnosis of MASLD.

The Lipid Accumulation Product (LAP) is a calculated index derived from the combination of waist circumference (WC) and fasting triglyceride (TG). Studies have shown that LAP is associated with liver fat content and metabolic disorders (such as hyperglycemia and dyslipidemia) [21]. Recent research further supports the role of LAP as an effective predictor of MASLD progression. For example, in a cross-sectional study involving 40,459 participants from southern China, LAP was significantly associated with a higher prevalence and severity of NAFLD in both men and women [17]. Similarly, a study by Liu et al. indicated that LAP is positively correlated with the incidence of MASLD and is a convenient index for the screening and quantification of NAFLD [22]. Our research conclusions are consistent with this. LAP, which can effectively identify visceral adiposity [23], is the most reliable predictor with an AUC of 0.786, an optimal threshold of 42.061. This also reflects the important role of dyslipidemia in the progression of MASLD [24]. Delong’s analysis confirmed that the AUC of LAP significantly differed from that of other indices. Meanwhile, our study emphasize its application in populations with type 2 diabetes (T2DM). The correlation between LAP and T2DM has long been established [25]. T2DM patients are five times more likely to develop MASLD compared to those without T2DM. This may be due to the shared pathophysiological mechanisms between NAFLD and T2DM, such as insulin resistance, hepatic lipidomic abnormalities, and triglyceride metabolism dysfunction [26,27,28]. The American Diabetes Association (ADA) updated its 2023《Standards of Care in Diabetes》guidelines, recommending routine MASLD screening for all adults with T2DM or prediabetes [29]. Therefore, we emphasize the application of LAP in T2DM populations, which will facilitate the early detection and diagnosis of MASLD.

Triglyceride glucose index–related parameters also have good performance in MASLD prediction, especially the composite indexes. Studies have confirmed that TyG index is closely related to MASLD and T2DM [30, 31]. A meta-analysis revealed that TyG-BMI (AUC = 0.84) and TyG-WC (AUC = 0.81) offer better predictive accuracy than the standalone TyG index (AUC = 0.75) [19]. This suggests that combining triglycerides with body composition metrics substantially improves the predictive performance of the TyG index.In our study, TyG-BMI, TyG-WC, and TyG-WHtR also demonstrated high predictive value for MASLD. Each composite indexes surpassed the single TyG index in the ROC analysis, highlighting their high predictive value. These results suggest that combining triglycerides with BMI, waist circumference, and waist-to-height ratio provides a more comprehensive assessment, thereby enhancing early detection and risk stratification of MASLD. In contrast, HOMA-IR didn’t have a good performance, with a relatively lower AUC value of 0.676 in our analysis. It indicates that while insulin resistance is a fundamental mechanism of MASLD development, other metabolic and body composition indicators are equally crucial for effective and accurate prediction.

Considering that LAP and TyG-BMI have the largest areas under the curve (AUC), we conducted a joint analysis of the two indices to assess whether combining them would enhance the predictive ability for MASLD in T2DM populations. However, the results did not align with our expectations. The ROC curve for the combined index showed an AUC of 0.771, which did not exceed that of LAP alone. Several factors could explain this phenomenon. LAP and TyG-BMI are both reflective of similar metabolic processes related to insulin resistance, visceral fat accumulation, and metabolic dysfunction [18, 32], which are key drivers of MASLD in T2DM patients [33, 34]. LAP incorporates waist circumference and triglycerides, while TyG-BMI is based on the triglyceride-glucose index adjusted for BMI. Given these overlapping components, the combined index may not add significant predictive value over and above what is already captured by LAP alone. Meanwhile, insufficient sample size is another important potential reason for this outcome. Nonetheless, this result may suggest that LAP, as a standalone indicator, better integrates key factors influencing T2DM and MASLD, thereby offering stronger predictive capability.

Present findings have important public health implications. A graphic abstract has been created to highlight this point(Fig. 3). In clinical practice, the advantages of LAP are not limited to its high predictive performance. With significant advantages such as simplicity, low cost and ease of use, it’s an ideal tool for large-scale screening of T2DM patients with MASLD, especially in resource-limited settings. When LAP is elevated, it provides a reasonable basis to recommend further abdominal ultrasound screening for MASLD. The widespread use of these metabolic indicators in T2DM patients can effectively improve the early detection rate of MASLD, thereby helping to reduce the risk of complications associated with fatty liver. Furthermore, for the T2DM population, the simplicity of calculating metabolic indicators can encourage patients to pay more attention to changes in lipid levels, body weight, and waist circumference, as well as to prioritize early MASLD screening, ultimately contributing to comprehensive health management.

There are some undeniable limitations in this study that deserved mention. Above all, the diagnosis of MASLD was determined using ultrasound imaging instead of liver biopsy pathology. Although ultrasound is a commonly used non-invasive method, simple grayscale ultrasound has reduced sensitivity in detecting mild hepatic steatosis [35]. In addition, US is significantly limited by both inter- and intraobserver reliability [36]. Therefore, future research could consider incorporating more precise imaging techniques, such as magnetic resonance imaging (MRI), or liver biopsy when feasible, to enhance the accuracy and early detection of MASLD [37]. Since our sample size primarily consisted of a population from Suzhou, China, this may lead to regional bias and may constrain the applicability of our results to the broader Chinese T2DM population. In addition, due to the cross-sectional design of our study, establishing a definitive relationship between MASLD risk and the nine evaluated indexes is challenging.

Fig. 3

A graphic abstract. Non-invasive examinations have more clinical advantages compared to invasive procedures in the diagnosis of MASLD

Conclusion

Our results indicated that all nine indexes have favorable correlations with the risk of MASLD, and most of them have a good performance in predicting MASLD. According to our study, LAP was a reliable index for predicting MASLD among Chinese T2DM patients. The exploration of non-invasive screenings will provide significant support for the early detection and diagnosis of MASLD.