MAFLD earlier denoted as non-alcoholic fatty liver disease (NAFLD), has become a significant worldwide health concern, and is a medical condition identified by a high level of fat within the liver, combined with at least one cardiometabolic risk factor, while excluding excessive alcohol intake and other coexisting liver conditions (1, 2). This drives a further understanding and categorizing NAFLD. MAFLD diagnosis requires imaging evidence, blood biomarkers/scoring, or histological evidence of hepatic steatosis, along with one of the following three conditions: overweight/obesity, T2DM, or abnormal metabolic disease risk (2). It has now become the leading chronic liver disease, impacting approximately one-third of adults worldwide. The prevalence of MAFLD among adults in the Asia-Pacific region ranges from 28 to 40%, with a yearly increase, it not only increases the risk of liver-related events, but also leads to extrahepatic complications--overweight/obesity, T2DM, atherosclerotic dyslipidemia, and hypertension, approximately one-third of patients may eventually develop to advanced stages, such as progressive fatty hepatitis, cirrhosis, and hepatocellular carcinoma, significantly increasing risks of liver-related and adverse cardiovascular disease (CVD) outcomes (3–6). MAFLD has increasingly become a public health concern in China. Characterized by insidious onset and progression, its severity and prevalence are frequently underestimated (7). Therefore, developing effective non-invasive screening methods and diagnostic indicators remains crucial for MAFLD prevention and control.

The exact mechanism of MAFLD remains unclear. It is widely believed that when intrahepatic adipose accumulation exceeds 5% of liver mass, metabolic disorders may occur, including abnormal fatty acid oxidation, oxidative stress, mitochondrial dysfunction, protein homeostasis disruption, and gut microbiota imbalance (8). Patients with MAFLD often exhibit dyslipidemia as high as 70%. Among patients with dyslipidemia, the prevalence of MAFLD ranges from 27 to 92%, while 69.2% of MAFLD patients also have dyslipidemia, characterized by elevated triglycerides (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C) and reduced high-density lipoprotein cholesterol (HDL-C) in traditional lipid parameters, which are hallmark features of atherogenic dyslipidemia (9, 10).

However, it cannot reliably diagnose MAFLD caused by fat accumulation by relying on above these single biomarkers. Therefore, we developed and evaluated nine non-conventional lipid parameters for identifying MAFLD: (1) body mass index (BMI), (2) non-HDL-to-HDL cholesterol ratio (NHHR), (3) atherogenic index of plasma (AIP), (4) cholesterol, high-density lipoprotein, and glucose index (CHG), (5) FBG to HDL-C (GHR), (6) remnant cholesterol (RC), (7) lipoprotein combine index (LCI), (8) triglyceride-glucose index (TyG), and (9) TyG derivative indicators (TyG-BMI). These indicators are closely associated with insulin resistance (IR), obesity, T2DM, and CVD. Previous studies indicated that NHHR and AIP were more strongly associated with MAFLD risk than other traditional lipid indicators, and exhibited a significant linear positive relationship (11–13). Elevated RC and CHG levels significantly increased the incidence and mortality rates associated with CVD (14, 15). GHR and the prevalence of NAFLD was an inverted L-shaped curve relationship and saturation effect, with an inflection point of 7.443 (16). Significant nonlinear associations were observed between LCI and prediabetes risk in a Chinese population (17). TyG and TyG-BMI showed significant correlations with all-cause mortality, cardiovascular mortality, and diabetes mortality in NAFLD/Metabolic dysfunction-associated steatotic liver disease (MASLD) patients, with the association demonstrating a clear nonlinear trend (18).

Despite existing research on the association of certain non-conventional lipid parameters—including BMI, NHHR, AIP, TyG, and TyG-BMI—with MAFLD, determining the optimal predictor remains a challenge (19). In the present study, we systematically compared nine non-conventional lipid parameters against conventional parameters and conducted a comprehensive assessment of their respective predictive capacities for MAFLD risk. Our analysis aimed to identify the parameter with the highest discriminative performance, thereby offering potential clinical utility for enhancing MAFLD screening and diagnostic strategies.

In recent years, Asian populations exhibit higher MAFLD prevalence and shown a global upward trend. A national research report from China showed that the overall prevalence of MAFLD was 37%, with males (46%) higher than females (24%), and the prevalence gradually increases with age, which may be attributed to excessive carbohydrate intake, central obesity, and genetic susceptibility (22). The MAFLD prevalence of 58.86% observed in our cohort exceeds estimates from general population surveys in China. We attribute this discrepancy primarily to selection bias inherent to hospital-based health examination populations, who predominantly comprise white-collar workers with high socioeconomic status, sedentary occupations, and higher health-seeking behaviors—characteristics clustering MAFLD risk factors. This phenomenon has been consistently reported in tertiary center-based studies. Ma X et al. found that the proportion of individuals diagnosed with MAFLD was 50.16% among the health examination population in the Kailuan Group from July 2006 to December 2007 (n = 49,518) (23). Charatcharoenwitthaya P et al. also found that the proportion of individuals diagnosed with MAFLD was 45.32% among the National Health Examination Survey in Thailand from 2008 to 2009 (n = 18,323) (24). Compared to other chronic liver diseases such as hepatitis B and C, MAFLD typically progresses more slowly. However, given the large number of affected individuals, the resulting health damage and disease burden cannot be underestimated.

MAFLD not only affects the liver but also increases cardiovascular disease risk. Exploring novel biomarkers can aid early intervention and diagnosis. The hallmark of MAFLD is lipid metabolism disorder, primarily manifested as elevated TG, and reduced HDL-C levels. The mechanisms underlying MAFLD caused by lipid abnormalities are multifaceted and complex, with IR playing a pivotal role (25, 26). Under normal physiological conditions, plasma free fatty acid (FFA) levels are higher when fasting. After eating, the increased insulin secretion inhibits the activity of hormone-sensitive lipase, resulting in an anti-lipolytic effect that leads to a decrease in plasma FFA. However, patients with T2DM exhibit peripheral IR, which impairs insulin’s ability to inhibit lipid breakdown in adipose tissue. Hepatic de novo synthesis of fat, and elevated serum TG levels collectively drive excessive production and secretion of LDL-C in the liver. This leads to elevated fatty acid levels in the liver, triggering hepatocyte dysfunction, inflammation, and oxidative stress, thereby accelerating the progression of MAFLD (27). Furthermore, IR-induced reduction in TG hydrolysis and impaired HDL-C synthesis result in both increased serum TG levels and decreased HDL-C concentrations (28). These atherosclerotic lipid abnormalities heighten the risk of MAFLD development. Thus, monitoring and screening for atherosclerotic lipid biomarkers proves beneficial for primary prevention of MAFLD.

Previous research has established that conventional lipid parameters—HDL-C, LDL-C, TG, and TC—are directly or indirectly associated with MAFLD (29–32). In this study, TG and HDL-C were identified as the most valuable conventional lipid parameters for MAFLD screening (Supplementary Figure 2). These results not only corroborate prior findings but also offer robust evidence supporting the clinical utility of TG and HDL-C as conventional lipid parameters in both the screening and management of MAFLD. In recent years, novel lipid parameters have garnered considerable scientific attention, as multiple studies confirm their substantial association with MAFLD. These emerging markers demonstrate meaningful utility in evaluating the risk of MAFLD. Nevertheless, determining which of these non-conventional lipid parameters holds the greatest clinical relevance for MAFLD detection continues to be an unresolved question.

In this retrospective single-center study, analysis of health examination data from Chinese adults revealed that eight non-conventional lipid parameters—namely NHHR, BMI, AIP, RC, GHR, LCI, TyG, and TyG-BMI—demonstrated effectiveness in distinguishing MAFLD patients, with the exception of CHG. Based on ROC curve analysis, TyG-BMI exhibited the strongest predictive performance, achieving an AUC of 0.81. TyG-BMI integrates TG levels, FBG, and BMI into a single composite parameter, offering a novel, simple, and cost-effective indicator for clinical use. IR is a well-established pathogenic driver of MAFLD (33). Although the HOMA-IR remains the gold standard for diagnosing IR and demonstrates high accuracy in MAFLD detection, its reliance on complex and costly assays limits its feasibility in large-scale epidemiological studies (34, 35). By contrast, the TyG index—calculated from fasting TG and glucose levels—serves as a practical, noninvasive surrogate marker for IR and has been validated in clinical settings (36). The TyG index has proven effective in identifying individuals at risk for T2DM, CVD, and metabolic syndrome (37). Recent studies further support its utility in detecting MAFLD, including among lean individuals (38–40). Incorporating BMI into the TyG index enhances its clinical relevance, as TyG-BMI better captures the influence of adiposity and body composition on metabolic dysregulation. Evidence suggests that TyG-BMI surpasses TyG alone as a marker of IR, offering improved stratification in populations with diverse metabolic phenotypes (41–43).

Given the established links between MAFLD, IR, and dyslipidemia, the TyG-BMI index is regarded as a promising predictive marker for MAFLD (38). However, findings across epidemiological investigations have been inconsistent (44). For instance, a cross-sectional study conducted in the Suzhong region of China reported that BMI, ALT, TG, TyG, TyG-BMI, and AIP all demonstrated strong predictive performance for MAFLD, with AUC values of 0.834, 0.803, 0.809, 0.822, 0.880, and 0.824, respectively (45). Among these, TyG-BMI exhibited the highest clinical diagnostic utility and decision-making benefit. Similarly, other studies have shown that TyG-BMI achieved the highest diagnostic accuracy in U. S. adolescents (AUC = 0.917) and in Japanese adults undergoing health examinations (AUC = 0.837) (46, 47). Additionally, a cross-sectional study of U. S. adults identified a significant nonlinear relationship between TyG-BMI and MAFLD, with a critical threshold near 176.78 (48).

According to the ROC curve results in this study, TyG-BMI has the highest AUC is 0.81. Consistent with previous research, RCS revealed a significant nonlinear relationship between TyG-BMI and MAFLD (P for nonlinearity<0.001). Subsequent threshold analysis indicated an inflection point value of about 222.426 for TyG-BMI. On the left side of the threshold value, each unit increase in TyG-BMI was associated with a 5.2% increase in the odds of developing MAFLD (OR = 1.052, 95% CI: 1.04–1.064, P < 0.001). The effect size (OR) for the right side of the threshold value was 1.02 (95% CI: 1.009–1.031, P < 0.001). Clinically, this threshold can assist physicians in risk stratification for MAFLD, to determine whether more aggressive management measures or closer monitoring are necessary. Therefore, incorporating the TyG-BMI threshold as part of a comprehensive risk assessment tool can aid in the development of public health policies and preventive measures.

Subgroup analyses revealed that TyG-BMI demonstrated stronger predictive performance for MAFLD in females compared to males. In general, females exhibit a lower overall risk of MAFLD than males. The peak incidence of MAFLD in males occurs between 40 and 49 years of age, whereas in females it typically presents around 60–69 years, a delay of approximately two decades that is largely attributed to the protective effects of estrogen. However, following menopause, the prevalence of MAFLD in women approaches that of men, likely due to factors such as weight gain, altered fat distribution, dyslipidemia, and declining estrogen levels—all of which elevate MAFLD risk. Furthermore, studies such as that by Boushehri et al. highlight sex-based differences in how TyG-BMI relates to MAFLD, suggesting that identical increases in TyG-BMI may confer differing risk profiles between sexes (49). The precise biological mechanisms underlying these sex differences in MAFLD pathogenesis and presentation remain incompletely understood but may involve variations in carbohydrate and lipid metabolism, as well as menopause-related changes in adiposity and metabolic susceptibility (50). Subgroup analyses also indicated that TyG-BMI was a more accurate predictor in non-diabetic individuals compared to those with diabetes. A systematic review and meta-analysis confirmed that TyG-BMI maintains strong diagnostic accuracy for MAFLD, with the highest AUC observed in the general population and lower values in patients with T2DM (50). This finding appears contradictory given that individuals with MAFLD are at elevated risk of developing diabetes; for instance, a biopsy-confirmed retrospective cohort study reported diabetes incidence rates of 51% in MAFLD patients with advanced fibrosis (F3–F4) and 31% in those with mild-to-moderate fibrosis (F0–F2) (51). One plausible explanation is that the presence of diabetes may attenuate the discriminative performance of TyG-BMI, possibly due to more pronounced and complex metabolic dysregulation in this population. Nevertheless, these observations warrant further validation in larger prospective cohort studies.

Moreover, studies indicate that CVD represents the most common and clinically significant extrahepatic manifestation in individuals with MAFLD, serving as the primary cause of mortality and accounting for approximately 40–45% of all deaths in this population. Extrahepatic malignancies—such as endometrial, breast, prostate, colorectal, and lung cancers—along with liver-related complications constitute the subsequent leading causes (52, 53). Consequently, we extended our analysis to examine the relationship between TyG-BMI and ASCVD risk. Our findings demonstrate a significant association, with MAFLD patients in the highest TyG-BMI tertile exhibiting a markedly elevated ASCVD risk compared to those in the lowest tertile.

MAFLD and CVD are closely linked in their pathophysiological mechanisms, with shared risk factors including IR, inflammatory responses, endothelial cell damage, and lipid accumulation. At the metabolic level, chronic hyperinsulinemia caused by IR accelerates the progression of atherosclerosis by activating vascular smooth muscle cell proliferation and promoting vascular wall inflammation. At the inflammatory level, systemic low-grade inflammation induced by hepatic steatosis triggers endothelial dysfunction and increased plaque instability through the TNF-α/NF-κB signaling pathway. At the lipid metabolism level, abnormal secretion of hepatic very low-density lipoprotein (VLDL) leads to the formation of atherosclerotic lipoprotein phenotype, characterized by increased small and dense LDL-C and impaired HDL-C functionality. In individuals diagnosed with T2DM and a BMI greater than 23 kg/m², the presence of MAFLD is positively correlated with an increased risk of CVD. Clinical studies have demonstrated that patients with elevated liver triglyceride levels exhibit reduced myocardial perfusion, decreased glucose uptake in the myocardium, and impaired myocardial energy metabolism, regardless of myocardial ischemia presence. Notably, no significant changes were observed in cardiac mass or functional parameters. These findings highlight the substantial predictive value of liver triglyceride levels for myocardial metabolic impairment and coronary artery dysfunction (54). Therefore, TyG-BMI which combines TG levels, glucose levels, and BMI can effectively reflect the CVD risk in patients with MAFLD. Liver fibrosis is a significant component of the MAFLD. Regardless of BMI, the presence of two or more metabolic disorders is significantly associated with an increased risk of liver fibrosis in MAFLD patients. However, there was no significant association between TyG-BMI and FIB-4 in participants with MAFLD in our study. Nevertheless, the OR in the highest tertile was 1.8, which was greater than 1, potentially indicating a risk, and further relevant studies were still required. Some degree of population selection bias may led to this results. FIB-4 is a non-invasive index and cannot fully meet the criteria for pathological diagnosis. TyG-BMI primarily reflects early metabolic disturbances, whereas FIB-4 assesses structural fibrosis damage. This discovery suggests that risk stratification for MAFLD requires the integration of multidimensional biomarkers, as reliance solely on metabolic parameters may underestimate fibrosis risk.

This study represents a comprehensive assessment comparing the strength of association and diagnostic accuracy of nine non-conventional lipid parameters for MAFLD. Our findings indicate that these novel lipid parameters, particularly TyG-BMI, demonstrate superior identification capability for MAFLD compared to conventional lipid parameters. Furthermore, we established optimal cutoff values and corresponding AUCs for TyG-BMI in MAFLD detection. Through multiple sensitivity analyses, the robustness of these associations was confirmed. Exploratory stratified analyses also yielded insights supporting precision medicine approaches. Finally, TyG-BMI was identified as an independent predictor of ASCVD risk in MAFLD patients, highlighting its potential clinical utility in early cardiovascular risk stratification within this population.

Despite these findings, our study has several limitations. First, given its cross-sectional design, no causal inference can be drawn regarding the relationship between TyG-BMI and MAFLD. Although TyG-BMI demonstrated superior predictive ability for MAFLD risk compared with other parameters, its clinical utility requires further validation in prospective studies. Second, as this is a retrospective study, the analysis was limited by incomplete collection of metabolic risk-related clinical data—such as hs-CRP, WC, and HOMA-IR—which may have contributed to under-diagnosis of MAFLD in some individuals. This have led to some degree of selection bias into the study, as less healthy older adults—who often carry higher CVD risk—are generally less likely to participate in medical health examination. Consequently, our study may represent a healthier subset of the target population, potentially resulting in an underestimation of the true association strength. Third, MAFLD diagnosis was based on ultrasonography rather than liver biopsy, which, while non-invasive and suitable for large-scale screening, has lower accuracy than histopathological examination. Fourth, although multiple covariates were adjusted for, residual confounding from factors such as family history of MAFLD, genetic susceptibility to cholesterol metabolism abnormalities, and other heritable influences cannot be ruled out. Fifth, this study utilized MAFLD criteria reflecting the diagnostic framework prevalent during our data collection period (2020–2021). While MASLD nomenclature has since been adopted, the core diagnostic requirements—hepatic steatosis with metabolic risk factors—remain substantially unchanged. Therefore, we expect the observed performance of lipid parameters to remain broadly informative for MASLD-related hepatic steatosis; however, the change in criteria may introduce some degree of misclassification and could modestly affect cross-study comparability and the direct generalizability of specific estimates. Finally, as a single-center retrospective study conducted in Chinese adults, the generalizability of our results to other populations requires further investigation from external validation cohorts in other centers and contexts.

This study demonstrates that among the nine evaluated non-conventional lipid parameters, all except CHG—namely BMI, NHHR, AIP, RC, GHR, LCI, TyG, and TyG-BMI—showed positive associations with MAFLD risk. Among these, TyG-BMI exhibited the strongest predictive performance for MAFLD in the Chinese adult population and also served as an independent predictor of ASCVD risk in patients with MAFLD. As a non-invasive and easily accessible parameter, TyG-BMI holds promise as a practical and reliable tool in clinical practice for MAFLD assessment, offering a feasible means for early identification and risk stratification.