Comparison of the Diagnostic Value Between Triglyceride-Glucose Index and Triglyceride-to-High Density Lipoprotein Cholesterol Ratio in MAFLD Patients: A Retrospective Cross-Sectional Study

Background The triglyceride and glucose index (TyG) and triglyceride-to-high density lipoprotein cholesterol ratio (TG/HDL-C) were found to be substitute markers of insulin resistance (IR). We aimed to compare the ecacy of the two indicators in the diagnosis of Metabolic-Associated Fatty Liver Disease (MAFLD), which was rarely covered in the literature, and to construct a novel disease diagnosis model. Methods A retrospective cross-sectional study was carried out in West China Hospital of Sichuan University and 229 people (97 MAFLD and 132 Non-MAFLD) were included. Biochemical indexes were collected and analyzed by logistic regression to screen out indicators that expressed differently in MAFLD patients and healthy controls and incorporate them into a diagnostic model. MAFLD was diagnosed by Ultrasound. Results After adjusting for age, gender and BMI, Serum ALT, AST, AST/ALT (A/A), FPG, Cys-C, URIC, TG, HDL-C, ALP, GGT, nonHDL-C, LDL-C/HDL-C, nonHDL-C/HDL-C, TG/HDL-C, TC/HDL-C, TyG and TyG-BMI were risk factors of MAFLD through binary logistics regression analysis. The odds ratio of TG/HDL-C and TyG were 5.387 (95%CI: 2.986-9,718) and 107.945 (95% CI: 25.824-451.222). In identifying MAFLD, TyG, TG/HDL-C and TG were found to be the most vital indexes by the random forest method and the area under the curve (AUC) of them are all greater than 0.9 respectively. In addition, the combination of gender, BMI, ALT, TG, HDL-C, TyG and TyG-BMI had a great diagnostic eciency for MAFLD. Conclusions TyG and TG/HDL-C were potential risk factors for MAFLD and the former made a better performance in The combination of gender, BMI, ALT, HDL-C, TyG and TyG-BMI improved the diagnostic capability of MAFLD.


Introduction
Renaming "Nonalcoholic Fatty Liver Disease (NAFLD)" to "MAFLD" takes into account the current understanding of the pathogenesis of MAFLD and its rising prevalence (1). Affecting more than onequarter of adults in the world, MAFLD is becoming one of the most vital causes of liver diseases (2). The incidence of MAFLD will continue to increase in developed countries such as the US and the European countries (Germany, France, Italy, and United Kingdom), causing a tremendous economic burden (3).
However, the occurrence rate of MAFLD in China will be the fastest, which is expected to rise from 243.66 million in 2016 to 314.58 million in 2030 4 . The challenges of MAFLD in China are as follows: a large number of cases, high genetic susceptibility, the appearance of patients young and with a low body mass index (BMI), absence of attention and recognition, lack of adequate diagnostic methods and treatments (5).
Although the pathogenesis theory of MAFLD has changed from two-stroke theory to the multi-stroke theory, IR still plays a signi cant role in the development of MAFLD(6). Regarded as a gold standard method to measure IR, hyperinsulinemic-euglycemic clamp (HIEC) is complicated, time-consuming and expensive. TG/HDL-C and TyG are found to be useful biomarkers in identifying IR individuals from a large group of Chinese (7,8). Early in 2005, TG/HDL-C was considered to be an important factor in predicting IR and increased risk of cardiovascular diseases in patients (9). As a result, TG/HDL-C is correlated with the occurrence of diabetes mellitus with related vascular diseases and fatty liver (10)(11)(12). The TyG, calculated by FPG and TG, has a close connection with IR which is related to glucolipid metabolism (13). Being used to detect IR, TyG seems to be cheaper and more convenient than the homeostasis model assessment of insulin resistance (HOMA-IR) index (14). TyG is also associated with cardiometabolic diseasesand is a risk prognostic factor for stroke, diabetes mellitus, acute myocardial infarction and acute coronary syndrome (15)(16)(17). Besides, TyG may be a useful indicator of MAFLD not only in adults but also in children and the elderly(18-20).
Therefore, the purpose of this study was to compare the ability of TG/HDL-C and TyG to distinguish MAFLD from healthy people and establish a better prediction model for MAFLD. Pearson's chi-square test for categorical variables was performed. Logistic regression analysis was used to nd independent risk factors for MAFLD, and the OR value was expressed with a 95% con dence interval (CI). Receivers operating characteristic curve (ROC curve) was used for diagnostic value analysis, and the maximum value of the Youden index (sensitivity+speci city-1) was taken as the optimal cut-off value. The difference was statistically signi cant when P<0.05.

Results
Baseline characteristics of the study participants 229 patients were enrolled in the present study. Table 1 listed the baseline characteristics of the subjects with and without MAFLD. Compared to non-MAFLD, individuals in MAFLD had higher levels ofBMI, ALT, AST, FPG, UREA, CREA, Cys-C, URIC, TG, TC, ALP, GGT, LDH, HBDH, nonHDL-C, TyG, TyG-BMI and lowerA/A, TP, ALB, GLB and HDL-C (all P < 0.005).As a result, ratios of the above indicators such as LDL-C/HDL-C, nonHDL-C/HDL-C, TG/HDL-C and TC/HDL-C were greater in MAFLD patients than in the other group.  (Table 2). Eventually, a predictive model consisted of gender, BMI, ALT, TG, HDL-C, TyG and TyG-BMI by binary logistic regression.As shown in Figure 1  Diagnostic performance ofvital indexesand the predictive model in MAFLD Through the random forest method, patients were randomly divided into a training set and a test set at a ratio of 2:1. When adopted all the variables in the model, a graph provided an overview of the importance score of each variable in predicting MAFLD. As can be seen in Figure2, TyG, TG/HDL-C and TG were the most important indicators for identifying MAFLD.
ROC curve analyses were conducted to identify the diagnostic value of TG, TG/HDL-C, TyG and the predictive model (Figure3). As a result, the area under receivers operating characteristic curve (AUROC) of the predictive model was 0.989 (95% CI 0.980-0.998) with 0.979 sensitivity and 0.947 speci city when the cut-off value was 0.323, showing the best capacity in assessing MAFLD. The ability in detecting MAFLD of TG, TG/HDL-C and TyG were also good, which were concluded in Table 3.

Discussion
This retrospective cross-sectional study assessed the ability of TG, TG/HDL-C and TyG to predict MAFLD and nally constructed a novel model in diagnosing MAFLD. TyG performed the best among the three indexes. The sensitivity and speci city of the new diagnostic model made up by gender, BMI, ALT, TG, HDL-C, TyG and TyG-BMI were improved to more than 90%. It was apparent from the Hosmer-Lemeshow test and the pictures of the calibration plot and DCA curve that the model ts well and could improve the net bene t.
MAFLD is not only the leading cause of chronic liver disease but also an increasing reason for HCC. With MAFLD-HCC, patients were older and had shorter survival times and more advanced tumor (22). Therefore, identifying MAFLD is of vital importance. Liver biopsy is still the gold standard for MAFLD although it has a variety of safety risks such as bleeding, pain, death and so on (23,24). In addition, there exists sampling errors and sampling bias (25). Noninvasive biomarkers which are cheap and effective for diagnosing MAFLD have been under study.
Triglycerides accumulate in the liver and blood of MAFLD. Interestingly, both TG/HDL-C and TyG which are surrogate IR markers derive from triglycerides. Independent of age, BMI and waist circumference, it was observed that whether the fatty liver patients with a normal or even higher level of ALT, the expression of TG/HDL-C was higher than that of the control group(26). Subsequently, several studies found that TG/HDL-C can be considered as a risk factor and even a predictor of MAFLD with a lower cutoff value and a greater AUROC in women rather than in men (27,28). What's more, in a randomized controlled study, decreases of TG and TG/HDL-C were associated with the resolution ofNonalcoholic steatohepatitis (NASH) (29).Another lipid ratio parameter, TC/HDL-C, was exerted to diagnose MAFLD with an AUROC of 0.645 in Ren's study (30). Both higher levels of TC/HDL-C and TG/HDL-C indicated more severe liver steatosis while the former showed a higher speci city (31). However, in our study, TG/HDL With the increment of TyG, the incidence of MAFLD signi cantly increased (33,34). The AUCs of the TyG index demonstrated by some studies in detecting adult MAFLD were 0.714-0.782 and their MAFLD was diagnosed by ultrasound (19,(35)(36)(37)(38)(39). Two researches among them mentioned that the AUC of TyG was larger than ALT.Fedchuk L et al. discovered that the AUC of TyG was up to 0.9 (95%CI 0.84-0.94) in biopsy-proven MAFLD with 80% sensitivity and 92% speci city when the cut-off value was 8.38 (40).
Similar to the research, the AUC of TyG in our study in diagnosing MAFLD was 0.943 (95%CI 0.913-0.973) with 87.6% sensitivity and 93.9% speci city when the cut-off value was 8.804. Among 50 asymptomatic women who underwent a liver biopsy, the TyG displayed a high sensitivity in screening simple steatosis and NASH (41). Based on a large amount of MAFLD participants ( n=11424) in a follow-up study, it was found that patients with a high TyG index were more likely to have MAFLD progression (42). TyG-BMI, one of Modi ed TyG Indices, performed a better ability to identify MAFLD than TyG itself not only in males but also in females. The AUROC of TyG-BMI in diagnosing MAFLD ranged from 0.774-0.9084 (43)(44)(45). We discovered that the AUROC of TyG-BMI was 0.949 (95% CI 0.922-0.976), which was higher than other researches. As for non-obese subjects, TyG-BMI could identify MAFLD and predict its occurrence effectively (46,47). In addition to liver steatosis, TyG and its related marker TyG-BMI were discovered to be related to liver brosis (39,48).
The limitations of our study were as follows. First of all, lack of a gold standard for diagnosis of MAFLD by liver biopsy. As we all know, ultrasound detection of steatosis less than 20% has limited sensitivity.
What also worries us are its poor differentiating effect between steatosis and brosis and poor consistency of repeated operations. The detection capability mostly depends on the examiners (49). Secondly, be short of reliable markers of IR such as HOMA-IR, which is a commonly used substitute for hyperinsulinemic-euglycemic glucose clamp technique to detect insulin resistance and verify the relationship between TG/HDL-C, TyG and IR. Last of all, to measure the diagnostic signi cance of the indicators for MAFLD more accurately, the suspected population of MAFLD should be included, otherwise the value of these indicators will be exaggerated.

Conclusion
This study compared the ability of two indices of surrogate for IR, TG/HDL-C and TyG, in predicting MAFLD for the rst time. TG/HDL-C and TyG can be easily measured with high accuracy and low cost due to the convenience of detecting TG, FPG and HDL-C. In conclusion, TyG performed better than TG/HDL-C in predicting MAFLD. In addition, the combination ofgender, BMI, ALT, TG, HDL-C, TyG and TyG-BMI could promote the diagnosis of MAFLD.   Calibration plot of the model Figure 5 DCA curves of the model