Abstract
Background
Nonalcoholic fatty liver disease (NAFLD) is a major public health issue in China’s aging population. Insulin resistance (IR) and metabolic dysregulation are key pathogenic factors, yet gold-standard assessment methods are complex and invasive, limiting clinical utility. The triglyceride-high-density lipoprotein cholesterol-glucose-body index (TyHGB index), derived from routine measures including triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), fasting plasma glucose (FPG), and body mass index (BMI), integrates core metabolic disturbances associated with NAFLD. Therefore, the primary aims of this study were to quantify the association between the TyHGB index and prevalent NAFLD, and to characterize its non-linear dose-response pattern and inflection point in middle-aged and older Chinese adults.
Methods
This cross-sectional study included 1,592 Chinese adults aged 40 – 79 years. NAFLD was diagnosed by abdominal ultrasonography. The TyHGB index was analyzed as both a continuous variable and in quartiles. Multivariable logistic regression models were employed to assess its association with NAFLD, adjusting for a comprehensive set of covariates covering demographics, lifestyle, comorbidities, and metabolic parameters. Restricted cubic spline (RCS) models evaluated the dose-response relationship, with piecewise regression identifying potential inflection points. Subgroup analyses and receiver operating characteristic (ROC) curve analysis were conducted to test the robustness and discriminatory ability of the index for prevalent NAFLD.
Results
The prevalence of NAFLD was 61.12%. A significant positive association was found between the TyHGB index and NAFLD prevalence. In the fully adjusted model, each 1-unit increase in the TyHGB index was associated with an 18% higher odds of NAFLD (OR = 1.18, 95% CI: 1.10 – 1.27). Participants in Q4 had a 4.65-fold higher odds of NAFLD compared to those in Q1 (OR = 4.65, 95% CI: 3.04 – 7.13). RCS analysis revealed a significant non-linear dose-response relationship, with an inflection point at a TyHGB index of 9.33. Below this threshold, each 1-unit increase was associated with an 80.9% higher odds of NAFLD (OR = 1.81, 95% CI: 1.53 – 2.14), while no significant association was observed above the threshold. This positive association remained directionally consistent across all pre-specified subgroups, including age, gender, BMI, and metabolic status. Furthermore, the TyHGB index showed acceptable discriminatory ability for prevalent NAFLD, with an area under the curve (AUC) of 0.75 (95% CI: 0.73 – 0.78). Its AUC was higher than those of TG, HDL-C, or FPG alone, but performed only modestly better than BMI. At the optimal cutoff of 7.058, determined by maximizing the Youden index, the accuracy, sensitivity, and specificity were 0.72, 0.59, and 0.81, respectively, indicating a trade-off between relatively high specificity and modest sensitivity.
Conclusion
In middle-aged and older Chinese adults, a higher TyHGB index is independently associated with an increased prevalence of NAFLD, exhibiting a clear non-linear dose-response relationship. This association remains robust across the majority of demographic and metabolic subgroups. Outperforming its individual lipid and glycemic components, the TyHGB index provides acceptable discriminatory ability comparable to BMI, making it a viable complementary screening and clinical stratification tool in health-check settings. Future prospective studies are required to establish causality and evaluate its longitudinal utility for NAFLD.
Clinical trial number
Not applicable.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s12902-026-02471-z.
Keywords: Triglyceride-high-density lipoprotein cholesterol-glucose-body index, Insulin resistance, Metabolic dysregulation, Middle-aged and older adults, Nonalcoholic fatty liver disease, Cross-sectional study, Metabolic dysfunction-associated steatotic liver disease
Background
Nonalcoholic fatty liver disease (NAFLD) is the most prevalent chronic liver disorder worldwide [1], with a global pooled prevalence reaching 32.4% [2]. In China, adult prevalence climbed to 29.6% between 1999 and 2018, paralleling the epidemics of obesity and diabetes [3]. As the hepatic manifestation of metabolic syndrome, NAFLD is intricately linked to insulin resistance (IR), obesity, dyslipidemia, and hypertension, significantly increasing the risks for cardiovascular disease, chronic kidney disease, and extrahepatic malignancies [4]. The NAFLD burden is particularly severe among middle-aged and older adults [5]; for instance, the prevalence reaches 36.9% in Chinese individuals aged ≥ 40 years [6]. Because these patients frequently present with multiple metabolic disturbances that accelerate liver disease progression, identifying early risk factors in this population is crucial for effective stratified management [7]. Recently, a multi-society consensus proposed updating the nomenclature from NAFLD to metabolic dysfunction-associated steatotic liver disease (MASLD) to better emphasize its underlying cardiometabolic pathophysiology [8]. While we retain the term NAFLD in this study to maintain strict consistency with the ultrasound-based diagnostic criteria utilized in the original dataset, the fundamental metabolic principles investigated herein are entirely aligned with the contemporary MASLD framework.
Clinical management of NAFLD faces diagnostic challenges. Liver biopsy, the gold standard, is limited by its invasiveness and cost [9, 10]. Ultrasonography, the primary non-invasive tool, has limited sensitivity for mild steatosis and is operator-dependent [11], while conventional serum biomarkers such as alanine aminotransferase (ALT) and aspartate aminotransferase (AST) often remain within normal ranges in NAFLD patients [8]. Consequently, developing simple, cost-effective, and reliable composite indices reflecting core pathophysiological mechanisms—specifically IR and lipid dysregulation—is a priority for early risk stratification.
While single metrics like body mass index (BMI), fasting plasma glucose (FPG), triglycerides (TG), and high-density lipoprotein cholesterol (HDL-C) are independent NAFLD risk factors [12], they capture limited metabolic dimensions and are susceptible to daily physiological fluctuations [13]. To comprehensively assess IR, composite indices like the triglyceride-glucose (TyG) index have been developed, demonstrating superior discriminatory performance over single indicators [14]. However, the TyG index omits adiposity and HDL-C, a cardioprotective lipid inversely associated with NAFLD risk [15]. Although TyG-related anthropometric indices (e.g., TyG-BMI) incorporate adiposity, they generally still lack HDL-C. Therefore, a novel composite index integrating pro-atherogenic lipids, protective lipids, glycemic status, and adiposity promises a more holistic evaluation of metabolic risk.
The triglyceride-high-density lipoprotein cholesterol-glucose-body (TyHGB) index (TG/HDL-C + 0.7×FPG + 0.1×BMI) [16] theoretically integrates these multiple metabolic abnormalities. Initially proposed for gestational diabetes [16], the TyHGB index has shown significant associations with diabetes [17], diabetic kidney disease [15], hyperuricemia and gout [18], preeclampsia [19], incident stroke [20], and coronary heart disease [21], often outperforming the traditional TyG index [15, 18, 21]. Regarding NAFLD, a 2025 study utilizing Chinese and U.S. cohorts confirmed that higher TyHGB index levels were significantly associated with increased NAFLD risk and exhibited superior discriminatory efficacy compared to the TyG index [22].
Despite these findings, critical gaps remain. Previous TyHGB index–NAFLD evaluations primarily involved younger or mixed cohorts [22], leaving a paucity of data on middle-aged and older Chinese adults who bear the highest disease burden. Furthermore, the precise non-linear dose-response relationship remains unclear. Identifying exploratory thresholds associated with a sharp increase in prevalence could provide an initial epidemiological basis for future risk stratification strategies.
To address these gaps, we performed a secondary analysis of a cross-sectional health-check cohort from China, specifically focusing on middle-aged and older adults. The primary aims were to quantify the association between the TyHGB index and prevalent NAFLD, and to characterize its non-linear dose-response pattern to estimate a data-derived inflection point. Given the cross-sectional design, this study focused on epidemiological association and risk stratification for prevalent disease rather than incident NAFLD prediction, aiming to evaluate the TyHGB index as a convenient, complementary metabolic marker in routine health examinations.
Method
Data source
Data for this study were obtained from Dryad, a well-established open-access scientific data repository that curates publicly available research datasets across diverse disciplines. The specific dataset utilized is linked to a cross-sectional study published in BMJ Open in 2023, entitled “Association of fat-to-muscle ratio with non-alcoholic fatty liver disease: a single-centre retrospective study” [23]. Researchers can access the raw data via the digital object identifier (DOI): https://doi.org/10.5061/dryad.7d7wm3809. The dataset includes information on individuals who underwent routine health examinations at Wuhan Union Hospital between January 2020 and November 2021. Participants completed standardized questionnaires detailing gender, age, tobacco use, alcohol use, and medical and medication history. In accordance with Dryad’s policy, re-analysis of publicly shared data for new scientific inquiries is permitted within ethical and legal frameworks. As the original study received ethical approval from the Ethics Committee of Tongji Medical College, Huazhong University of Science and Technology, this secondary analysis did not involve new participants and was exempt from additional ethical review and informed consent procedures. All study procedures strictly adhered to the ethical principles outlined in the Declaration of Helsinki and complied fully with national regulations and institutional policies.
Study population
This study conducted a secondary analysis based on a previously established cross-sectional dataset from mainland China. Therefore, it involved no new data collection and no a priori sample size calculation was performed. The original study by Yan et al. prospectively enrolled 1,830 participants aged 40–79 years who voluntarily underwent body composition analysis and liver ultrasound examination during routine health check-ups. Demographic and clinical information, including gender, age, tobacco use, alcohol use, and medical and medication history, were collected through standardized questionnaires. Participants were excluded if they met any of the following criteria: excessive alcohol consumption (defined as consumption exceeding 210 g per week for males and 140 g per week for females); known history of liver diseases (including viral, autoimmune, or drug-induced hepatitis); acute illness, renal insufficiency (estimated glomerular filtration rate < 60 mL/min/1.73 m2), or active cancer diagnosed or treated within the past six months; use of oral or injectable steroids; or missing key biochemical data or interview records. After applying these inclusion and exclusion criteria, 1,592 participants constituted the preliminary analytic cohort.
This study aimed to investigate the association between the TyHGB index and NAFLD in middle-aged and older Chinese adults. As data for FPG, BMI, HDL-C, and TG were complete for all participants (Supplementary Table 1), the entire cohort of 1,592 individuals was included in the final analysis.
Data collection and variable definitions
Trained personnel performed all anthropometric measurements. Body composition parameters, including body weight, fat mass, and muscle mass, were assessed using a multi-frequency bioelectrical impedance analyzer (BIA; Tsinghua Tongfan BCA-2A, China). Participants were evaluated after an overnight fast, wearing light clothing and standing barefoot. The BIA device underwent regular manufacturer calibration. The fat-to-muscle ratio (FMR) was calculated by dividing fat mass by muscle mass. BMI was computed as body weight in kilograms divided by the square of height in meters (kg/m2). After resting for at least 10 min, seated blood pressure (BP) was measured twice at 5-minute intervals using an electronic sphygmomanometer (Panasonic EW3106, China), and the average value was used for analysis.
Fasting venous blood samples (≥ 8 h) were processed at the hospital’s central laboratory. Platelet count (PLT) was determined using a Beckman-Coulter hematology analyzer. Biochemical parameters, including total cholesterol (TC), TG, HDL-C, low-density lipoprotein cholesterol (LDL-C), ALT, AST, uric acid (UA), and FPG, were analyzed on a Beckman AU5800 automated chemistry system.
Diabetes was defined by self-reported physician-diagnosed diabetes or current use of glucose-lowering medications. Hypertension was defined by self-reported physician-diagnosed hypertension or current use of antihypertensive agents. The diagnosis of fatty liver was established through conventional abdominal ultrasonography performed by trained technicians (Philips IU22, Philips Healthcare). Participants exhibiting fatty liver without evidence of other concomitant liver diseases were classified as having NAFLD. The exposure variable for this study was the TyHGB index, treated as a continuous variable. It was calculated using the formula proposed in the literature [16]: TyHGB index = TG (mmol/L) / HDL-C (mmol/L) + 0.7 × FPG (mmol/L) + 0.1 × BMI (kg/m2). All components of the TyHGB index formula were expressed in SI units, including TG, HDL-C, and FPG in mmol/L and BMI in kg/m2. Previous studies have validated the association between the TyHGB index and NAFLD in the general Chinese population [22].
Statistical method
The distribution of the TyHGB index among the study population is illustrated in Fig. 1. To examine the dose-response relationship between the TyHGB index and NAFLD, all participants were categorized into quartiles based on the index values: Q1 (≤ 6.74), Q2 (≤ 7.66), Q3 (≤ 9.04), and Q4 (≤ 30.91).
Fig. 1.
Distribution of the TyHGB index
The Kolmogorov-Smirnov test was used to assess the normality of continuous variables. Data with a normal distribution are presented as mean ± standard deviation, while non-normally distributed variables are expressed as median (interquartile range) [M (IQR)]. Categorical variables are summarized as frequency (percentage) [n (%)]. For between-group comparisons, one-way analysis of variance (ANOVA) was applied to normally distributed continuous variables with homogeneity of variance; otherwise, the Kruskal-Wallis H test was used. Comparisons of categorical variables were performed using the chi-square test.
Logistic regression models were constructed to evaluate the association between the TyHGB index and NAFLD prevalence, with results quantified as odds ratios (ORs) and 95% confidence intervals (CIs). The crude model was unadjusted. Model I was adjusted for age, gender, tobacco use, alcohol use, hypertension, and diabetes. Model II was further adjusted for TC, LDL-C, FMR, PLT, ALT, AST, UA, systolic blood pressure (SBP), and diastolic blood pressure (DBP) based on Model I. Covariates were selected based on prior clinical knowledge, previously reported NAFLD-related factors, and variables available in the original dataset. As a rationale for the covariate selection, Model II was considered a conservative adjustment model. Because some Model II covariates, such as TC, LDL-C, FMR, BP, and UA, may be closely related to both TyHGB index components and NAFLD-related metabolic pathways, adjusting for them may partially attenuate the TyHGB index signal. However, this comprehensive adjustment was intentionally designed as a highly conservative approach to rigorously test whether the TyHGB index provides incremental epidemiological value beyond traditional metabolic markers. While we acknowledge that this strategy risks introducing partial overadjustment, it ensures that any persisting independent association is strictly robust. Multicollinearity was assessed using the variance inflation factor (VIF), and all covariates in Model II had VIF values below 5.0, indicating no severe multicollinearity (Supplementary Table 2). Additionally, as a sensitivity analysis to evaluate the association without the potential physiological overlap of related metabolic parameters, we constructed a parsimonious model (Model Ⅲ). Model Ⅲ was adjusted for age, gender, tobacco use, alcohol use, hypertension, diabetes, PLT, AST, and ALT. Restricted cubic splines (RCS) with four knots placed at the 5th, 35th, 65th, and 95th percentiles of the TyHGB index were used to assess the dose-response relationship. If statistical tests indicated a significant non-linear trend (non-linear P-value < 0.05), a piecewise logistic regression model was further employed to estimate a data-derived inflection point. This inflection point was exploratory and should not be interpreted as a clinically established cutoff. The strength and direction of the association between the TyHGB index and NAFLD prevalence were then estimated separately on either side of the identified threshold. Subgroup analyses were conducted to evaluate potential effect modifications by age, gender, BMI (< 28 kg/m2 vs. ≥ 28 kg/m2), tobacco use, alcohol use, hypertension, diabetes, TC (< 6.2 mmol/L vs. ≥ 6.2 mmol/L), TG (< 2.3 mmol/L vs. ≥ 2.3 mmol/L), HDL-C (< 1 mmol/L vs. ≥ 1 mmol/L), and LDL-C (< 4.1 mmol/L vs. ≥ 4.1 mmol/L). These subgroup analyses were exploratory. No formal correction for multiple testing was applied; therefore, the subgroup results were interpreted cautiously, with emphasis on consistency of direction rather than statistical significance in individual strata, and these analyses should be considered strictly hypothesis-generating. Except for the stratifying variable itself, all logistic regression models within each subgroup were uniformly adjusted for the covariates in Model II to assess the stability and consistency of the main association across different population characteristics. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the discriminatory ability of the TyHGB index and its individual components (FPG, TG, HDL-C, and BMI) for prevalent NAFLD. DeLong’s non-parametric test was used to conduct pairwise comparisons of the areas under the ROC curves (AUCs) between different variables. The ROC cutoff was determined by maximizing the Youden index and was distinct from the RCS reference value and the piecewise regression inflection point.
Statistical software
All results are reported in accordance with the STROBE statement. Statistical analyses were performed in the R environment (version 4.2.2; https://www.R-project.org/) and the Free Statistics Analysis Platform (version 2.1.1; https://www.clinicalscientists.cn/freestatistics). A two-tailed P-value of less than 0.05 was considered statistically significant. As the present study constitutes a secondary analysis of pre-existing data, no formal a priori sample size calculation was performed.
Results
Characteristics of the study population
A total of 1,592 middle-aged and older Chinese adults were included in this analysis. The mean TyHGB index for the entire cohort was 8.43 ± 2.84. Among the participants, 1,148 (72.11%) were male and 444 (27.89%) were female. The age range was 40–79 years, with the largest proportion falling within the 50–59 years age group (709 participants, 44.54%). The overall prevalence of NAFLD was 61.12% (973/1,592).
Participants were categorized into quartiles (Q1 – Q4) based on the TyHGB index, with 398 individuals in each group. The baseline characteristics of the total population and across TyHGB index quartiles are presented in Table 1. Significant differences in gender and age distributions were observed across the TyHGB index quartiles (both P < 0.001). The proportion of females decreased progressively from 50% in Q1 to 13.32% in Q4. Age distribution also varied significantly across groups. Participants aged 40–49 years accounted for the highest proportion in Q1 (26.38%) and the lowest in Q2 (17.84%), with a slight rebound in Q3 (19.60%) and Q4 (25.13%). Conversely, the proportion of participants aged 50–59 years increased gradually with rising TyHGB index quartiles, from 39.20% in Q1 to 47.99% in Q4. Participants aged 60–69 years and 70–79 years comprised relatively higher proportions in Q2 and Q3, but decreased to the lowest levels in Q4 (18.59% and 8.29%, respectively). Levels of BMI, UA, SBP, DBP, FPG, TG, ALT, and AST all showed an upward trend with increasing TyHGB index quartiles (all P < 0.001). In contrast, HDL-C levels exhibited a gradual decline across ascending quartiles (P < 0.001). The FMR increased from Q1 to Q3 and slightly decreased in Q4 (P = 0.002). TC levels remained relatively stable from Q1 to Q3 and were highest in Q4 (P < 0.001). The proportions of tobacco users, alcohol users, individuals with hypertension, those with diabetes, and NAFLD cases all increased progressively with higher TyHGB index quartiles (all P < 0.001). No statistically significant differences were observed in LDL-C levels or PLT across the TyHGB index quartiles.
Table 1.
Characteristics of the study population by TyHGB index quartiles
| Variables | Total (n = 1592) |
TyHGB index | ||||
|---|---|---|---|---|---|---|
| Q1 (n = 398) | Q2 (n = 398) | Q3 (n = 398) | Q4 (n = 398) | P-value | ||
| TyHGB index, Mean ± SD | 8.43 ± 2.84 | 6.24 ± 0.37 | 7.19 ± 0.26 | 8.25 ± 0.40 | 12.02 ± 3.56 | < 0.001 |
| Age (years), n (%) | 0.013 | |||||
| 40–49 | 354 (22.24) | 105 (26.38) | 71 (17.84) | 78 (19.6) | 100 (25.13) | |
| 50–59 | 709 (44.54) | 156 (39.2) | 178 (44.72) | 184 (46.23) | 191 (47.99) | |
| 60–69 | 360 (22.61) | 90 (22.61) | 102 (25.63) | 94 (23.62) | 74 (18.59) | |
| 70–79 | 169 (10.62) | 47 (11.81) | 47 (11.81) | 42 (10.55) | 33 (8.29) | |
| Gender, n (%) | < 0.001 | |||||
| Female | 444 (27.89) | 199 (50) | 105 (26.38) | 87 (21.86) | 53 (13.32) | |
| Male | 1148 (72.11) | 199 (50) | 293 (73.62) | 311 (78.14) | 345 (86.68) | |
| BMI (kg/m2), Mean ± SD | 25.41 ± 2.92 | 23.28 ± 2.07 | 24.97 ± 2.11 | 26.41 ± 2.72 | 26.97 ± 3.14 | < 0.001 |
| FMR, Mean ± SD | 0.39 ± 0.12 | 0.38 ± 0.11 | 0.39 ± 0.12 | 0.41 ± 0.12 | 0.40 ± 0.11 | 0.002 |
| UA (µmol/L), Mean ± SD | 366.75 ± 95.83 | 319.19 ± 81.35 | 362.57 ± 84.57 | 389.30 ± 90.22 | 395.95 ± 106.02 | < 0.001 |
| SBP (mmHg), Mean ± SD | 130.94 ± 15.82 | 126.03 ± 15.12 | 130.59 ± 15.52 | 132.55 ± 15.18 | 134.57 ± 16.20 | < 0.001 |
| DBP (mmHg), Mean ± SD | 81.60 ± 11.19 | 77.97 ± 10.93 | 80.93 ± 10.89 | 82.86 ± 10.65 | 84.64 ± 11.23 | < 0.001 |
| FPG (mmol/L), Mean ± SD | 5.54 ± 1.65 | 4.61 ± 0.40 | 5.03 ± 0.52 | 5.48 ± 0.82 | 7.05 ± 2.54 | < 0.001 |
| TC (mmol/L), Mean ± SD | 4.48 ± 1.09 | 4.38 ± 1.03 | 4.35 ± 1.18 | 4.40 ± 0.99 | 4.78 ± 1.10 | < 0.001 |
| TG (mmol/L), Median (IQR) | 1.42 (0.98, 2.18) | 0.86 (0.69, 1.06) | 1.23 (1.00, 1.51) | 1.73 (1.33, 2.15) | 2.92 (2.16, 4.19) | < 0.001 |
| HDL-C (mmol/L), Mean ± SD | 1.11 ± 0.32 | 1.38 ± 0.31 | 1.15 ± 0.29 | 1.03 ± 0.22 | 0.90 ± 0.22 | < 0.001 |
| LDL-C (mmol/L), Mean ± SD | 2.67 ± 0.89 | 2.61 ± 0.88 | 2.69 ± 0.92 | 2.71 ± 0.86 | 2.64 ± 0.90 | 0.393 |
| ALT (U/L), Median (IQR) | 21.00 (15.00, 30.00) | 16.00 (12.00, 22.00) | 21.00 (15.00, 27.00) | 24.00 (17.00, 34.75) | 27.00 (20.00, 39.00) | < 0.001 |
| AST (U/L), Median (IQR) | 21.00 (17.00, 26.00) | 20.00 (17.00, 23.75) | 20.00 (17.00, 24.00) | 21.00 (18.00, 27.00) | 23.00 (18.00, 29.00) | < 0.001 |
| PLT (109/L), Mean ± SD | 211.15 ± 53.31 | 209.68 ± 53.69 | 211.93 ± 50.17 | 209.03 ± 54.64 | 213.96 ± 54.65 | 0.547 |
| Tobacco use, n (%) | < 0.001 | |||||
| No | 1054 (66.21) | 314 (78.89) | 268 (67.34) | 246 (61.81) | 226 (56.78) | |
| Yes | 538 (33.79) | 84 (21.11) | 130 (32.66) | 152 (38.19) | 172 (43.22) | |
| Alcohol use, n (%) | < 0.001 | |||||
| No | 1072 (67.34) | 312 (78.39) | 281 (70.6) | 239 (60.05) | 240 (60.3) | |
| Yes | 520 (32.66) | 86 (21.61) | 117 (29.4) | 159 (39.95) | 158 (39.7) | |
| Hypertension, n (%) | < 0.001 | |||||
| No | 649 (40.77) | 232 (58.29) | 153 (38.44) | 147 (36.93) | 117 (29.4) | |
| Yes | 943 (59.23) | 166 (41.71) | 245 (61.56) | 251 (63.07) | 281 (70.6) | |
| Diabetes, n (%) | < 0.001 | |||||
| No | 1094 (68.72) | 353 (88.69) | 317 (79.65) | 270 (67.84) | 154 (38.69) | |
| Yes | 498 (31.28) | 45 (11.31) | 81 (20.35) | 128 (32.16) | 244 (61.31) | |
| NAFLD, n (%) | < 0.001 | |||||
| No | 619 (38.88) | 285 (71.61) | 165 (41.46) | 106 (26.63) | 63 (15.83) | |
| Yes | 973 (61.12) | 113 (28.39) | 233 (58.54) | 292 (73.37) | 335 (84.17) | |
As shown in Table 2, distinct characteristics were observed between the 973 participants diagnosed with NAFLD and the 619 participants without NAFLD. Compared to the non-NAFLD group, individuals with NAFLD exhibited a significantly higher TyHGB index (P < 0.001). Additionally, the NAFLD group demonstrated significantly elevated levels of BMI, FMR, UA, SBP, DBP, FPG, TC, and TG (all P < 0.001). Levels of ALT and AST were also markedly higher in the NAFLD group (P < 0.001), whereas HDL-C levels were significantly lower (P = 0.001). The proportions of males, current tobacco users, current alcohol users, individuals with hypertension, and those with diabetes were all substantially greater in the NAFLD group (all P < 0.001). However, no significant differences were detected between the two groups regarding age distribution, LDL-C levels, or PLT.
Table 2.
Characteristics of the study population according to NAFLD status
| Variables | Non-NAFLD (n = 619) | NAFLD (n = 973) | P-value |
|---|---|---|---|
| TyHGB index, Mean ± SD | 7.39 ± 2.12 | 9.09 ± 3.04 | < 0.001 |
| Age (years), n (%) | 0.085 | ||
| 40–49 | 127 (20.52) | 227 (23.33) | |
| 50–59 | 267 (43.13) | 442 (45.43) | |
| 60–69 | 146 (23.59) | 214 (21.99) | |
| 70–79 | 79 (12.76) | 90 (9.25) | |
| Gender, n (%) | < 0.001 | ||
| Female | 229 (37) | 215 (22.1) | |
| Male | 390 (63) | 758 (77.9) | |
| BMI (kg/m2), Mean ± SD | 23.99 ± 2.41 | 26.31 ± 2.86 | < 0.001 |
| FMR, Mean ± SD | 0.38 ± 0.11 | 0.40 ± 0.12 | < 0.001 |
| UA (µmol/L), Mean ± SD | 337.81 ± 86.51 | 385.17 ± 96.95 | < 0.001 |
| SBP (mmHg), Mean ± SD | 128.93 ± 15.51 | 132.21 ± 15.88 | < 0.001 |
| DBP (mmHg), Mean ± SD | 79.73 ± 11.17 | 82.79 ± 11.05 | < 0.001 |
| FPG (mmol/L), Mean ± SD | 5.19 ± 1.45 | 5.77 ± 1.73 | < 0.001 |
| TC (mmol/L), Mean ± SD | 4.37 ± 1.01 | 4.55 ± 1.13 | 0.001 |
| TG (mmol/L), Median (IQR) | 1.10 (0.82, 1.55) | 1.74 (1.17, 2.56) | < 0.001 |
| HDL-C (mmol/L), Mean ± SD | 1.21 ± 0.36 | 1.05 ± 0.27 | 0.001 |
| LDL-C (mmol/L), Mean ± SD | 2.63 ± 0.86 | 2.69 ± 0.91 | 0.193 |
| ALT (U/L), Median (IQR) | 17.00 (13.00, 24.00) | 25.00 (17.00, 35.00) | < 0.001 |
| AST (U/L), Median (IQR) | 20.00 (17.00, 24.00) | 21.00 (18.00, 27.00) | < 0.001 |
| PLT (109/L), Mean ± SD | 209.69 ± 53.28 | 212.08 ± 53.33 | 0.385 |
| Tobacco use, n (%) | < 0.001 | ||
| No | 458 (73.99) | 596 (61.25) | |
| Yes | 161 (26.01) | 377 (38.75) | |
| Alcohol use, n (%) | < 0.001 | ||
| No | 456 (73.67) | 616 (63.31) | |
| Yes | 163 (26.33) | 357 (36.69) | |
| Hypertension, n (%) | < 0.001 | ||
| No | 316 (51.05) | 333 (34.22) | |
| Yes | 303 (48.95) | 640 (65.78) | |
| Diabetes, n (%) | < 0.001 | ||
| No | 493 (79.64) | 601 (61.77) | |
| Yes | 126 (20.36) | 372 (38.23) |
Figure 2 visually illustrates the trend of increasing NAFLD prevalence with ascending TyHGB index quartiles. Specifically, the prevalence rates in Q1 through Q4 were 28.39%, 58.54%, 73.37%, and 84.17%, respectively. Compared with the Q1 group, participants in the Q4 group exhibited a 55.78% absolute increase in NAFLD prevalence.
Fig. 2.
Prevalence of NAFLD across TyHGB index quartiles. The horizontal axis represents the TyHGB index quartile groups, and the vertical axis represents the prevalence of NAFLD (%). Blue indicates participants without NAFLD, and red indicates participants with NAFLD
Association between the TyHGB index and NAFLD prevalence
Multivariable logistic regression was performed to assess the association between the TyHGB index and the prevalence of NAFLD (Table 3). In the fully adjusted Model II (adjusted for age, gender, tobacco use, alcohol use, hypertension, diabetes, TC, LDL-C, FMR, PLT, AST, ALT, UA, SBP, and DBP), each one-unit increase in the TyHGB index was associated with an 18% higher odds of prevalent NAFLD (OR = 1.18, 95% CI: 1.10–1.27; P for trend < 0.001). Compared to the unadjusted crude model (OR = 1.51), the effect estimate in Model II was attenuated by approximately 60% (based on the change in log-odds), reflecting the expected impact of adjusting for shared metabolic covariates.
Table 3.
Association between TyHGB index and prevalence of NAFLD, assessed using logistic regression in different models
| Variables | Crude model | Model Ⅰ | Model Ⅱ | |||
|---|---|---|---|---|---|---|
| OR (95% CI) | P-value | OR (95% CI) | P-value | OR (95% CI) | P-value | |
| TyHGB index | 1.51 (1.40–1.62) | < 0.001 | 1.39 (1.29–1.50) | < 0.001 | 1.18 (1.10–1.27) | < 0.001 |
| (TyHGB index quartiles) | ||||||
| Q1 | 1.00 (Reference) | 1.00 (Reference) | 1.00 (Reference) | |||
| Q2 | 3.56 (2.65–4.79) | < 0.001 | 3.28 (2.42–4.46) | < 0.001 | 2.47 (1.79–3.41) | < 0.001 |
| Q3 | 6.95 (5.09–9.48) | < 0.001 | 6.07 (4.38–8.41) | < 0.001 | 3.37 (2.37–4.81) | < 0.001 |
| Q4 | 13.41 (9.49–18.96) | < 0.001 | 10.24 (7.0–14.98) | < 0.001 | 4.65 (3.04–7.13) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | |||
Crude model: we did not adjust other covariates
Model I: adjusted for age, gender, tobacco use, alcohol use, hypertension, and diabetes
Model II: further adjusted for TC, LDL-C, FMR, PLT, AST, ALT, UA, SBP, and DBP based on Model I
When analyzed by TyHGB index quartiles in Model II, participants in the higher quartiles exhibited significantly increased odds of NAFLD compared to the Q1. The ORs (95% CIs) for Q2, Q3, and Q4 were 2.47 (1.79–3.41), 3.37 (2.37–4.81), and 4.65 (3.04–7.13), respectively, demonstrating a robust dose-response relationship.
This positive association and dose-response trend were also consistently observed in the less adjusted models. In the unadjusted crude model, the ORs (95% CIs) per one-unit increment and for Q2, Q3, and Q4 vs. Q1 were 1.51 (1.40–1.62), 3.56 (2.65–4.79), 6.95 (5.09–9.48), and 13.41 (9.49–18.96), respectively. In Model I (adjusted for age, gender, tobacco use, alcohol use, hypertension, and diabetes), the corresponding ORs (95% CIs) were 1.39 (1.29–1.50), 3.28 (2.42–4.46), 6.07 (4.38–8.41), and 10.24 (7.00–14.98).
Furthermore, in the parsimonious sensitivity model (Model Ⅲ, Supplementary Table 3), which omitted potentially overlapping metabolic covariates to evaluate the unattenuated independent effect, the association remained robust and highly significant (OR = 1.31, 95% CI: 1.22–1.41; P for trend < 0.001). Participants in Q4 of Model Ⅲ exhibited an 8.06-fold higher odds of NAFLD compared to those in Q1 (OR = 8.06, 95% CI: 5.45–11.92).
Non-linear association between the TyHGB index and NAFLD prevalence
Figure 3 illustrates the non-linear relationship between the TyHGB index and the prevalence of NAFLD using an RCS curve, with adjustment for all covariates in Model II. The analysis revealed a significant non-linear association (P for non-linearity < 0.001). In the gender-stratified analysis (Supplementary Fig. 1), this non-linear relationship remained significant in both males and females (P < 0.001). Specifically, NAFLD prevalence increased rapidly with rising TyHGB index values at lower-to-intermediate levels, while the association at very high TyHGB index values became statistically non-significant, strictly aligning with the wide CIs crossing 1.0 in this upper range. Similarly, the age-stratified analysis (Supplementary Fig. 2) demonstrated a significant non-linear association across all age groups (40−49, 50–59, 60–69, and 70–79 years; Supplementary Fig. 3; P < 0.001 for all). The pattern of change was consistent across strata: NAFLD prevalence increased initially with rising TyHGB index levels and subsequently decreased after reaching a threshold, confirming a non-linear dose-response relationship.
Fig. 3.
Non-linear dose-response relationship between the TyHGB index and prevalent NAFLD. The model was adjusted for age, gender, tobacco use, alcohol use, hypertension, diabetes, TC, LDL-C, FMR, PLT, AST, ALT, UA, SBP, and DBP. The solid curve represents the adjusted OR, and the shaded area represents the 95% CI
Piecewise logistic regression analysis identified a significant inflection point in the association between the TyHGB index and NAFLD prevalence (P for likelihood ratio test < 0.001) (Table 4). The inflection point was located at a TyHGB index value of 9.33. The reference value in the RCS plot was used only for graphical comparison, whereas the inflection point of 9.33 was derived from piecewise logistic regression. After adjusting for all covariates in Model II, each one-unit increase in the TyHGB index was associated with an 80.9% higher odds of NAFLD (OR = 1.81, 95% CI: 1.53–2.14, P < 0.001) when the index was below 9.33. However, when the TyHGB index was at or above 9.33, no statistically significant association with NAFLD prevalence was observed (OR = 0.94, 95% CI: 0.80–1.11, P = 0.486).
Table 4.
Data-derived inflection point for the association between the TyHGB index and prevalent NAFLD identified by piecewise logistic regression. The model was adjusted for age, gender, tobacco use, alcohol use, hypertension, diabetes, TC, LDL-C, FMR, PLT, AST, ALT, UA, SBP, and DBP
| TyHGB index (per 1-unit) | OR (95% CI) | P-Value |
|---|---|---|
| < 9.33 | 1.81 (1.53–2.14) | < 0.001 |
| ≥ 9.33 | 0.94 (0.80–1.11) | 0.486 |
| P for likelihood ratio test | < 0.001 |
Subgroup analysis
As illustrated by the forest plot in Fig. 4, the positive association between elevated TyHGB index and increased prevalence of NAFLD was observed in the majority of prespecified subgroups. These subgroups were stratified by age, gender, BMI, tobacco use, alcohol use, hypertension, diabetes, and lipid profiles (TG, TC, HDL-C, and LDL-C). All subgroup analyses were adjusted for the same set of covariates specified in Model II. While the vast majority of adjusted ORs were greater than 1.0, the association was not completely consistent across all strata. Notably, the point estimates reversed (OR < 1.0) among participants with a BMI ≥ 28 kg/m2 and those with TG ≥ 2.3 mmol/L. Furthermore, statistical significance was not achieved in several strata where the positive trend remained but 95% CIs crossed 1.0, specifically among participants aged 60–69 years or 70–79 years, current tobacco users, current alcohol users, individuals without hypertension, those with diabetes, and participants with TC ≥ 6.2 mmol/L, HDL-C < 1 mmol/L, or LDL-C ≥ 4.1 mmol/L.
Fig. 4.
Subgroup analysis of the association between the TyHGB index and prevalent NAFLD
Comparative discriminatory performance of the TyHGB index and its individual components for NAFLD
We further evaluated the discriminatory performance of the TyHGB index and its individual components for identifying NAFLD using ROC curve analysis. As shown in Fig. 5; Table 5, and Table 6, the TyHGB index demonstrated acceptable discriminatory ability for identifying prevalent NAFLD, with an AUC of 0.75 (95% CI: 0.73–0.78). The AUC of the TyHGB index was significantly higher than that of TG (AUC = 0.72, 95% CI: 0.69–0.74), HDL-C (AUC = 0.64, 95% CI: 0.61–0.67), and FPG (AUC = 0.64, 95% CI: 0.61–0.67). However, the TyHGB index performed only modestly better than BMI, with no statistically significant difference observed (AUC = 0.74, 95% CI: 0.72–0.77; △AUC = 0.0111, P = 0.4278). Based on the Youden index, the optimal cutoff value for the TyHGB index to distinguish NAFLD was 7.058, yielding an accuracy of 0.72 (95% CI: 0.70–0.75), a relatively low sensitivity of 0.59 (95% CI: 0.55–0.63), and a specificity of 0.81 (95% CI: 0.79–0.84).
Fig. 5.
Comparison of the discriminatory ability of TyHGB index, TG, HDL-C, FPG, and BMI for NAFLD. This figure presents ROC curves evaluating the associative and discriminatory ability of the TyHGB index and its individual components (TG, HDL-C, FPG, and BMI) to identify NAFLD. The results demonstrate that the TyHGB index (dark blue curve) achieves the highest AUC, indicating that its associative performance significantly outperforms that of TG (orange curve), FPG (light blue curve), or HDL-C (green curve) alone, but its AUC was comparable to BMI (purple curve). The black dashed line represents the random reference line
Table 5.
Performance of individual TyHGB index components and the TyHGB index for identifying NAFLD. This table summarizes the AUC, accuracy, sensitivity, specificity, and optimal cut-off values for the TyHGB index and its constituent variables—TG, HDL-C, FPG, and BMI—in identifying NAFLD. The composite TyHGB index achieved the highest AUC (0.75, 95% CI: 0.73–0.78), followed by BMI (0.74, 95% CI: 0.72–0.77), TG (0.72, 95% CI: 0.69–0.74), HDL-C (0.64, 95% CI: 0.61–0.67), and FPG (0.64, 95% CI: 0.61–0.67), reinforcing the acceptable discriminatory ability of TyHGB index demonstrated in Fig. 5. However, pairwise comparison using DeLong’s test showed no statistically significant difference in AUC between TyHGB index and BMI (△AUC = 0.0111, 95% CI: −0.0163–0.0385; P = 0.4278)
| Variables | AUC (95% CI) | Accuracy (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | Cut off |
|---|---|---|---|---|---|
| TyHGB index | 0.75 (0.73–0.78) | 0.72 (0.70–0.75) | 0.59 (0.55–0.63) | 0.81 (0.79–0.84) | 7.058 |
| BMI | 0.74 (0.72–0.77) | 0.70 (0.68–0.72) | 0.59 (0.56–0.63) | 0.77 (0.74–0.79) | 24.2 |
| TG | 0.72 (0.69–0.74) | 0.63 (0.61–0.66) | 0.82 (0.79–0.85) | 0.51 (0.48–0.54) | 1.715 |
| HDL-C | 0.64 (0.61–0.67) | 0.62 (0.59–0.64) | 0.65 (0.61–0.68) | 0.57 (0.53–0.61) | 1.112 |
| FPG | 0.64 (0.61–0.67) | 0.58 (0.56–0.61) | 0.71 (0.68–0.75) | 0.50 (0.47–0.53) | 5.285 |
Table 6.
Pairwise comparison of the AUC between the TyHGB index and its individual components for identifying NAFLD
| Categories | △AUC (95% CI) | SD | Z-value | P-value |
|---|---|---|---|---|
| TyHGB index vs. BMI | 0.0111 (− 0.0163–0.0385) | 0.0140 | 0.793 | 0.4278 |
| TyHGB index vs. TG | 0.0378 (0.0202–0.0554) | 0.00898 | 4.206 | < 0.0001 |
| TyHGB index vs. HDL-C | 0.117 (0.0922–0.141) | 0.0125 | 9.310 | < 0.0001 |
| TyHGB index vs. FPG | 0.114 (0.0910–0.138) | 0.0119 | 9.588 | < 0.0001 |
Discussion
This cross-sectional analysis extends prior evidence on the TyHGB index–NAFLD association by focusing specifically on middle-aged and older Chinese adults from a health-check cohort. Its main added value lies in characterizing the non-linear dose-response pattern and estimating a data-derived inflection point in this age-specific population. Our findings demonstrate a significant positive association between higher TyHGB index levels and NAFLD prevalence, which remained robust even after extensive multivariable adjustment. The attenuation of effect estimates after adjustment suggests part of the association may be shared with related metabolic factors; moreover, because TyHGB index already includes BMI, TG, HDL-C, and FPG, further adjustment for overlapping variables (TC, LDL-C, FMR) may have introduced partial over-adjustment. Nevertheless, the persistence of a highly significant association even in this conservatively adjusted model robustly underscores the independent epidemiological value of the TyHGB index. Notably, a significant non-linear relationship was identified: below the 9.33 inflection point, each one-unit increase correlated with an 80.9% higher odds of prevalent NAFLD, whereas associations above this data-derived threshold lost statistical significance. The direction of this positive association was consistent across all prespecified demographic and metabolic subgroups, suggesting generalizability. However, because no correction for multiple comparisons was applied, these subgroup findings remain exploratory and are primarily hypothesis-generating. Furthermore, the TyHGB index showed acceptable discriminatory ability for prevalent NAFLD (AUC = 0.75), outperforming TG, HDL-C, and FPG alone; however, it is important to note that it performed only modestly better than BMI without statistical significance. While the index integrates lipid imbalance, glycemic status, and adiposity to provide broader metabolic context, the Youden cutoff of 7.058 yielded a relatively low sensitivity (0.59) despite high specificity (0.81). Given the cross-sectional nature of our data, it is important to emphasize that this cutoff is strictly exploratory and not yet ready for direct clinical implementation. Therefore, rather than serving as a standalone screening or diagnostic tool, the TyHGB index might eventually be explored as a potential complementary marker in health-check settings. This combination of high specificity and modest sensitivity supports its use for rule-in risk stratification to prioritize asymptomatic individuals for further structural hepatic evaluation (e.g., ultrasound or elastography) rather than triggering immediate therapeutic intervention.
The observed association between the TyHGB index and NAFLD aligns with the index’s integration of core pathophysiological mechanisms—specifically, atherogenic lipid profiles, IR, and generalized adiposity [24–26]. By synergistically capturing these metabolic disturbances, TyHGB index provides a more holistic evaluation of metabolic disequilibrium than its individual components. Notably, this study identified a distinct non-linear dose-response relationship and pinpointed an inflection point at a TyHGB index value of 9.33. Below this threshold, each one-unit increment in the index was associated with a rapid surge in NAFLD prevalence (OR = 1.81). This suggests that within this range, the cumulative effect of metabolic aberrations represented by the TyHGB index exerts its most potent “driving force” on hepatic steatosis. Conversely, once the index exceeded this critical value, the escalating trend in prevalence was no longer statistically evident. The identification of this inflection point carries potential epidemiological implications, though we acknowledge it may reflect specific statistical modeling choices rather than a universally true biological threshold. This specific value can serve as a purely data-derived, exploratory reference point for identifying high-risk individuals in whom metabolic disturbances exert a particularly pronounced effect on liver health. This assists clinicians in prioritizing patients for further specialized hepatic evaluation, such as ultrasound or elastography, once this threshold is surpassed, thereby facilitating early risk stratification and enabling individualized risk assessment [27]. The biological explanation for the high-value segment of the curve remains uncertain and requires further validation.
Our findings align with and extend the growing body of evidence supporting the utility of the TyHGB index as a marker of metabolic dysfunction. Prior studies have established independent associations between the TyHGB index and a spectrum of metabolic disorders, including diabetes [16, 17], hyperuricemia [18], gout [18], and diabetic kidney disease [15]. These investigations underscore its value as a comprehensive indicator of systemic metabolic disturbance. A 2025 study utilizing a Chinese hospital cohort and the U.S. NHANES dataset first highlighted the association between the TyHGB index and NAFLD, suggesting a non-linear trend [22]. Our results corroborate this core conclusion, reinforcing the TyHGB index as a robust correlate of NAFLD. Nevertheless, our study advances this field in specific ways. First, by focusing exclusively on middle-aged and older Chinese adults—a demographic bearing the highest burden of NAFLD—we provide targeted evidence directly applicable to this high-risk population. Second, beyond confirming the non-linear relationship, we precisely quantified the inflection point at a TyHGB index of 9.33 using RCS modeling. This provides a data-derived exploratory threshold for risk stratification, whereas previous research only suggested a non-linear trend without identifying a specific threshold in comparable populations. Furthermore, our extensive subgroup analyses demonstrated that the positive association remained directionally consistent across diverse demographic and metabolic strata. This enhances the robustness and generalizability of the TyHGB index within this specific population. Thus, while supporting its role as an effective tool for identifying NAFLD, our study supports the utility of the TyHGB index as a complementary tool for risk stratification.
It is important to acknowledge the recent evolution in disease terminology from NAFLD toward metabolic dysfunction-associated fatty liver disease (MAFLD) and MASLD. The present study retained the term NAFLD to remain consistent with the original dataset definition and the ultrasound-based diagnostic criteria used in the source cohort. Nevertheless, because the newer MASLD diagnostic frameworks, such as those outlined in the 2024 Chinese and 2023 American Association for the Study of Liver Diseases (AASLD) guidance [8], explicitly emphasize cardiometabolic risk factors including adiposity, fasting glucose, and lipid profiles, the TyHGB index aligns perfectly with these updated risk-stratification paradigms. Therefore, while our data analysis strictly relies on the NAFLD definition, the clinical utility of TyHGB index remains highly relevant to the contemporary MASLD landscape. At the clinical and public health levels, the TyHGB index provides an accessible tool for identifying middle-aged and older adults at higher risk for NAFLD. Because its constituent parameters are routinely measured, it facilitates straightforward calculation and implementation across various healthcare settings [28]. Utilizing this data-derived exploratory threshold may assist in prioritizing resources for hepatic imaging or lifestyle interventions, aiding in long-term, personalized metabolic health management [29, 30].
This study possesses several notable strengths. First, conducting a secondary analysis using an openly accessible dataset with standardized protocols enhances the transparency and reproducibility of our findings. Second, our comprehensive analytical strategy (multivariable logistic regression, RCS, and piecewise regression) allowed for a systematic evaluation of the dose-response relationship and inflection points. Third, focusing on middle-aged and older Chinese adults provides direct evidence for this high-risk demographic. Finally, the TyHGB index relies on inexpensive, routine clinical parameters, facilitating its translation into practical risk stratification tools. However, several limitations must be acknowledged. First, the cross-sectional design precludes causal inference between the TyHGB index and NAFLD. Second, the data were derived from a single tertiary hospital health-check cohort, which may introduce selection bias compared to community populations. Individuals undergoing routine health checks may exhibit different socioeconomic statuses and health awareness compared to general community populations, which could influence the external validity of our findings. Third, the sex distribution of the study population was imbalanced, with men accounting for 72.11% of participants and only 444 women included. This imbalance may limit the statistical power and precision of sex-specific analyses, particularly among women. Although a significant gender interaction was observed, the relatively small female sample may have affected the stability of the female-specific estimates. Nevertheless, our gender-stratified subgroup and RCS analyses demonstrated that the robust positive association persisted in both women and men, supporting internal validity across genders. Fourth, NAFLD ascertainment relied exclusively on conventional abdominal ultrasonography. While acceptable and standard for large-scale epidemiologic screening, ultrasonography has limited sensitivity for detecting mild steatosis and cannot distinguish simple steatosis from nonalcoholic steatohepatitis (NASH) or accurately stage liver fibrosis. Because the TyHGB index integrates multiple core metabolic disturbances, its elevation might theoretically correlate with more severe liver pathology (i.e., NASH or advanced fibrosis). However, given that this study is a secondary analysis of a pre-existing dataset, we were strictly constrained by the diagnostic modalities employed in the original cohort. Consequently, our analysis is restricted to the mere detection of steatosis. The inability to assess disease severity means the full clinical implications of a high TyHGB index cannot be entirely captured in this study. Future research employing advanced quantitative imaging or liver biopsy is required to elucidate the index’s association with NASH and fibrosis staging. Fifth, residual confounding from unmeasured factors (e.g., diet, physical activity) may remain. Additionally, because some Model II covariates (e.g., TC, LDL-C, and UA) are closely related to TyHGB index components, potential overadjustment cannot be fully excluded, and the fully adjusted estimates should be interpreted conservatively. Sixth, in our exploratory subgroup analyses, we observed attenuated point estimates and statistical interactions in strata defined by BMI and specific lipid profiles (such as TG). Because these parameters are mathematically coupled to the TyHGB calculation, stratifying the cohort based on the index’s own components inherently restricts variance, limiting the clinical interpretation of these specific strata. Finally, because this study was restricted to a Chinese population, the applicability of these results and exploratory cutoffs to broader non-Chinese populations warrants future cross-ethnic validation.
While this study establishes a cross-sectional association, it delineates critical avenues for future investigation. First, large-scale prospective cohort studies are required to establish causality and evaluate TyHGB index as a predictor of incident NAFLD. Second, external validation across multi-center and multi-ethnic populations is necessary to confirm generalizability, alongside investigations into disease severity. Third, future research should employ precise hepatic quantification techniques and systematically control for broader lifestyle and genetic confounders. Finally, interventional trials assessing whether reducing the TyHGB index through lifestyle or pharmacological means correspondingly reduces NAFLD progression will validate its efficacy as a target in prevention strategies.
Conclusion
In middle-aged and older Chinese adults, a higher TyHGB index is independently associated with an increased prevalence of NAFLD, exhibiting a clear non-linear dose-response relationship. This association remains robust across the majority of demographic and metabolic subgroups. Outperforming its individual lipid and glycemic components, the TyHGB index provides acceptable discriminatory ability comparable to BMI, making it a viable complementary screening and clinical stratification tool in health-check settings. Future prospective studies are required to establish causality and evaluate its longitudinal utility for NAFLD.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
None.
Abbreviations
- TG
Triglycerides
- HDL-C
High-density lipoprotein cholesterol
- FPG
Fasting plasma glucose
- BMI
Body mass index
- RCS
Restricted cubic spline
- ROC
Receiver operating characteristic
- AUC
Area under the curve
- OR
Odds ratio
- CI
Confidence interval
- VIF
Variance inflation factor
- BIA
Bioelectrical impedance analyzer
- FMR
Fat-to-muscle ratio
- BP
Blood pressure
- SBP
Systolic blood pressure
- DBP
Diastolic blood pressure
- PLT
Platelet count
- TC
Total cholesterol
- LDL-C
Low-density lipoprotein cholesterol
- ALT
Alanine aminotransferase
- AST
Aspartate aminotransferase
- UA
Uric acid
- DOI
Digital object identifier
- NAFLD
Nonalcoholic fatty liver disease
- MAFLD
Metabolic dysfunction-associated fatty liver disease
- MASLD
Metabolic dysfunction-associated steatotic liver disease
- IR
Insulin resistance
- TyHGB index
Triglyceride-high-density lipoprotein cholesterol-glucose-body index
- NASH
Nonalcoholic steatohepatitis
- TyG
Triglyceride-glucose index
- NHANES
National Health and Nutrition Examination Survey
- ANOVA
Analysis of variance
- IQR
Interquartile range
- SD
Standard deviation
- STROBE
Strengthening the reporting of observational studies in epidemiology
- AASLD
American Association for the Study of Liver Diseases
Author contributions
YTZ: Data curation, Methodology, Writing – original draft, Writing – review & editing.HBL: Data curation, Methodology, Writing – original draft, Writing – review & editing.YL: Data curation, Methodology, Writing – original draft, Writing – review & editing.RZL: Visualization, Writing – original draft.ZTL: Visualization, Writing – original draft.SL: Visualization, Writing – original draft.HKZ: Writing – review & editing.DY: Conceptualization, Data curation, Methodology, Writing – original draft, Writing – review & editing. QXH: Conceptualization, Data curation, Methodology, Writing – original draft, Writing – review & editing. LSZ: Conceptualization, Data curation, Methodology, Writing – original draft, Writing – review & editing.
Funding
None.
Data availability
The dataset presented in this study is available in the online repository (https://doi.org/10.5061/dryad.7d7wm3809).
Declarations
Ethical approval and consent to participate
Ethics review and approval for this research involving human subjects were provided by the Ethics. Committee of Tongji Medical College, Huazhong University of Science and Technology. This retrospective study, which utilized fully anonymized data, was performed in accordance with local legislative and institutional requirements. Given the anonymized nature of the data, the study qualified for exemptions from obtaining further approval from an institutional ethics committee and from procuring written informed consent from participants or their legal guardians/next of kin.
Consent for publication
Not applicable.
Generative AI statement
The authors declare that no generative artificial intelligence technologies were employed in the.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Yiting Zhang, Huibao Lyu, and Yan Lu contributed equally to this work and should be considered as co-first authors.
Duo Yang, Qixia He, and Longsheng Zhang contributed equally to this work and should be considered as co-corresponding authors.
Contributor Information
Duo Yang, Email: duoyang15775096614@126.com.
Qixia He, Email: gdmufy2011@163.com.
Longsheng Zhang, Email: zhangls@gdmu.edu.cn.
References
- 1.Younossi ZM, Golabi P, Paik JM, Henry A, Van Dongen C, Henry L. The global epidemiology of nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis (NASH): a systematic review. Hepatology. 2023;77(4):1335–47. 10.1097/HEP.0000000000000004. Epub 2023 Jan 3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Riazi K, Azhari H, Charette JH, Underwood FE, King JA, Afshar EE, Swain MG, Congly SE, Kaplan GG, Shaheen AA. The prevalence and incidence of NAFLD worldwide: a systematic review and meta-analysis. Lancet Gastroenterol Hepatol. 2022;7(9):851–61. 10.1016/S2468-1253(22)00165-0. Epub 2022 Jul 5. [DOI] [PubMed] [Google Scholar]
- 3.Zhou J, Zhou F, Wang W, Zhang XJ, Ji YX, Zhang P, She ZG, Zhu L, Cai J, Li H. Epidemiological Features of NAFLD From 1999 to 2018 in China. Hepatology. 2020;71(5):1851–64. 10.1002/hep.31150. [DOI] [PubMed] [Google Scholar]
- 4.Fan JG, Xu XY, Yang RX, Nan YM, Wei L, Jia JD, Zhuang H, Shi JP, Li XY, Sun C, Li J, Wong VW, Duan ZP. Chinese Society of Hepatology, Chinese Medical Association. Guideline for the Prevention and Treatment of Metabolic Dysfunction-associated Fatty Liver Disease (Version 2024). J Clin Transl Hepatol. 2024;12(11):955–74. 10.14218/JCTH.2024.00311. Epub 2024 Nov 4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Wang Q, Guo J, Liu S, Cao X, Guo Z, Rui L, Zheng L, Wang C. Global, regional, and national burden of nonalcoholic fatty liver disease among adults aged ≥ 45 years: A comprehensive analysis of epidemiological trends and projections to 2035. PLoS ONE. 2026;21(3):e0342697. 10.1371/journal.pone.0342697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Zeng J, Qin L, Jin Q, Yang RX, Ning G, Su Q, Yang Z, Fan JG. Prevalence and characteristics of MAFLD in Chinese adults aged 40 years or older: A community-based study. Hepatobiliary Pancreat Dis Int. 2022;21(2):154–61. 10.1016/j.hbpd.2022.01.006. Epub 2022 Jan 31. [DOI] [PubMed] [Google Scholar]
- 7.Lou TW, Yang RX, Fan JG. The global burden of fatty liver disease: the major impact of China. Hepatobiliary Surg Nutr. 2024;13(1):119–23. 10.21037/hbsn-23-556. Epub 2024 Jan 12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Rinella ME, Neuschwander-Tetri BA, Siddiqui MS, Abdelmalek MF, Caldwell S, Barb D, Kleiner DE, Loomba R. AASLD Practice Guidance on the clinical assessment and management of nonalcoholic fatty liver disease. Hepatology. 2023;77(5):1797–835. 10.1097/HEP.0000000000000323. Epub 2023 Mar 17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Martinou E, Pericleous M, Stefanova I, Kaur V, Angelidi AM. Diagnostic Modalities of Non-Alcoholic Fatty Liver Disease: From Biochemical Biomarkers to Multi-Omics Non-Invasive Approaches. Diagnostics (Basel). 2022;12(2):407. 10.3390/diagnostics12020407. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Nowak K, Nowak A, Jabłońska A, Potaczek A, Salacha J, Dardzińska N, Janczura J. Implementation of noninvasive liver disease screening tools in primary care. Korean J Fam Med. 2025;46(6):381–90. 10.4082/kjfm.25.0144. Epub 2025 Nov 20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Wentworth BJ, Caldwell SH. Pearls and pitfalls in nonalcoholic fatty liver disease: tricky results are common. Metab Target Organ Damage. 2021;1(1):2. 10.20517/mtod.2021.02. Epub 2021 Jul 2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Li N, Liu C, Lu Z, Wu W, Zhang F, Qiu L, Shen C, Sheng D, Liu Z. Establishing of a risk prediction model for metabolic dysfunction-associated steatotic liver disease: a retrospective cohort study. BMC Gastroenterol. 2025;25(1):39. 10.1186/s12876-025-03598-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Lin H, Wang Z, Tang J, Liu X. Metabolic characteristics and risk profile of non-obese fatty liver disease in an elderly Chinese community cohort. BMC Gastroenterol. 2026;26(1):244. 10.1186/s12876-026-04730-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Ling Q, Chen J, Liu X, Xu Y, Ma J, Yu P, Zheng K, Liu F, Luo J. The triglyceride and glucose index and risk of nonalcoholic fatty liver disease: A dose-response meta-analysis. Front Endocrinol (Lausanne). 2023;13:1043169. 10.3389/fendo.2022.1043169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yang J, Xie B, Deng Z, Zhang Z, Zhou H. The triglyceride-high-density lipoprotein-glucose-body index: a superior novel biomarker for diabetic kidney disease in type 2 diabetes. Front Endocrinol (Lausanne). 2026;16:1749826. 10.3389/fendo.2025.1749826. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Song S, Zhang Y, Qiao X, Duo Y, Xu J, Peng Z, Zhang J, Chen Y, Nie X, Sun Q, Yang X, Lu Z, Liu S, Zhao T, Yuan T, Fu Y, Dong Y, Zhao W, Sun W, Wang A. HOMA-IR as a risk factor of gestational diabetes mellitus and a novel simple surrogate index in early pregnancy. Int J Gynaecol Obstet. 2022;157(3):694–701. 10.1002/ijgo.13905. Epub 2021 Sep 24. [DOI] [PubMed] [Google Scholar]
- 17.He XY, Ma DH, Liu Z, Guo YF, Zhao RC, Sun PP, Zhong XL. Association between novel obesity- and lipid-related indices and diabetes across different FBG among elderly: a prospective cohort study. Front Endocrinol (Lausanne). 2026;17:1798687. 10.3389/fendo.2026.1798687. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Zhang Q, Xu J, Qian Y. Association between TyHGB and hyperuricemia and gout in patients with type-2 diabetes mellitus. BMC Endocr Disord. 2026 Apr 30. 10.1186/s12902-026-02294-y. Epub ahead of print. [DOI] [PMC free article] [PubMed]
- 19.Bai Y, Xi Y, Huang J, Wu K, Yin B, Zhang J. TyHGB as a predictive marker for preeclampsia: association with increased risk in pregnant women with normal glucose tolerance. Lipids Health Dis. 2026 Apr 27. 10.1186/s12944-026-02962-x. Epub ahead of print. [DOI] [PMC free article] [PubMed]
- 20.Lu A, Xu Q. Association between the triglyceride-high-density lipoprotein-glucose-body index and new-onset stroke risk: a prospective cohort study. Lipids Health Dis. 2026 Apr;20. 10.1186/s12944-026-02953-y. Epub ahead of print. [DOI] [PMC free article] [PubMed]
- 21.Sun M, Lin W, Gong R, Guo R, Zhang M, Yin Y, Wang Y, Zhang S, Dong T, Luo D. TyHGB combined with diagonal earlobe crease and traditional risk factors for constructing a diagnostic model of coronary heart disease. Lipids Health Dis. 2026;25(1):66. 10.1186/s12944-026-02880-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Zhao E, Wang R, Chen Y, Xie H, Gao Y, Dong B, Zhang C. Evaluation of a new TyG indicator, TyHGB, in predicting non-alcoholic fatty liver disease: evidence from two independent populations. Lipids Health Dis. 2025;24(1):342. 10.1186/s12944-025-02742-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Yan F, Nie G, Zhou N, Zhang M, Peng W. Association of fat-to-muscle ratio with non-alcoholic fatty liver disease: a single-centre retrospective study. BMJ Open. 2023;13(10):e072489. 10.1136/bmjopen-2023-072489. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Kim AH, Son DH, Lee YJ. Modified triglyceride-glucose index indices are reliable markers for predicting risk of metabolic dysfunction-associated fatty liver disease: a cross-sectional study. Front Endocrinol (Lausanne). 2024;14:1308265. 10.3389/fendo.2023.1308265. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Heeren J, Scheja L. Metabolic-associated fatty liver disease and lipoprotein metabolism. Mol Metab. 2021;50:101238. 10.1016/j.molmet.2021.101238. Epub 2021 Apr 20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lin L, Pan G, Liu L, Huang J. The association between triglyceride glucose-waist-to-height ratio and non-alcoholic fatty liver disease in adults aged over 60 in the United States: a cross-sectional study. Front Public Health. 2026;13:1569324. 10.3389/fpubh.2025.1569324. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Wang X, Liu J, Yu K, Huang Z, Liu H, Li X. Association between TyG-related parameters and NAFLD risk in Japanese non-obese population. Sci Rep. 2025;15(1):7119. 10.1038/s41598-025-88478-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Zhu G, Qu Y, Chen K, He D, Wang J, Tang Z, Wang X, Zhang M, Jiang P, Zhang R, Cai K. Comparison of the triglyceride-glucose index and triglyceride-glucose-body mass index for predicting non-alcoholic fatty liver disease in elderly diabetic patients. PLoS ONE. 2026;21(2):e0341109. 10.1371/journal.pone.0341109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Zhu X, Wang Z, Yang X, Ning Z. Diagnostic performance of TyG-related indices for MAFLD screening in older adults at moderate altitude: TyG-BMI showed the highest discriminatory ability. BMC Gastroenterol. 2026;26(1):172. 10.1186/s12876-026-04604-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Chen Y, Lin X, Chen Y, Wang B, Xu C, Wang Y. Triglyceride-glucose index and related parameters in the diagnosis and predictive efficacy for older adults with NAFLD/MASLD: a prospective cohort study. Front Med (Lausanne). 2025;12:1626672. 10.3389/fmed.2025.1626672. Erratum in: Front Med (Lausanne). 2026;13:1804927. doi: 10.3389/fmed.2026.1804927. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The dataset presented in this study is available in the online repository (https://doi.org/10.5061/dryad.7d7wm3809).