UTILITY OF TRIGLYCERIDE-GLUCOSE INDEX IN PREDICTING GLYCEMIC CONTROL IN TYPE 2 DIABETES MELLITUS

Santosh Timalsina1*, Shishir Mahato2, Sandesh Nepal2 Timalsina S et al Introduc on Insulin resistance (IR) and glycemic control are two very important aspects to be considered during management of pa ents with Type 2 Diabetes Mellitus (T2DM). The triglyceride-glucose (TyG) index has been proposed as a simple and inexpensive parameter that correlates well with IR and glycemic control. Objec ves To explore the associa on of TyG index (and other TyG derived indices) with glycated hemoglobin (HbA1c) and evaluate their predic ve ability for glycemic control in pa ents with T2DM. Methodology This cross-sec onal study comprised of 160 adult pa ents diagnosed with T2DM visi ng the medical outpa ent department of Chitwan Medical College, Bharatpur, Chitwan between July–December 2019. Socio-demographic data and anthropometric measurements were collected. Glycemic control was assessed by HbA1c. TyG index was calculated by the formula: ln [fas ng TG (mg/dl) x fas ng glucose (mg/dl)/2]. Receiver opera ng characteris c (ROC) curve analysis was performed to analyze the predic ve ability of TyG-index for poor glycemic control. Result One hundred and sixty pa ents (mean age: 53.6 ± 10.7 years, 55.0% males) were included in the study. Eighty (50.0%) had good glycemic control (HbA1c <7.0%). TyG index, along with TyG-BMI and TyG-WC (other TyG derived indices) were significantly increased in the poor glycemic control group. TyG index had a good predic ve ability for poor glycemic control (AUC: 0.803, 95% CI: 0.731 – 0.874). A TyG cutoff ≥ 9.12 was op mal for predic ng poor glycemic control, with 86.1% sensi vity and 61.5% specificity. Conclusion TyG index could be a simple and cost-effec ve screening tool for assessment of glycemic control in pa ents with T2DM.


INTRODUCTION
The global prevalence of type 2 diabetes mellitus (T2DM), a common metabolic disorder characterized by hyperglycemia and insulin resistance is ever increasing; and has been 1 regarded as a major public health issue. In a recent na onwide survey in Nepal, the prevalence of diabetes among adults was 5.8%, and a substan al gap in the 2 diagnosis and treatment was observed. Glycosylated hemoglobin (HbA1c) is the commonly used assessment tool for glycemic control, which reflects average glycemia over approximately 3 months. It has been shown 3,4 to have a strong predic ve value for diabetes complica ons. The frequency of HbA1c tes ng has been recommended to be guided by the clinical situa on, treatment regimen and 5 the clinicians' judgement according to recent guidelines , however frequent tes ng in low-income countries is s ll limited by its high cost and inadequate availability of 6 standardized assays. The triglyceride-glucose (TyG) index is a simple, inexpensive parameter that has been suggested as a surrogate marker for insulin resistance, closely correla ng with Homeostasis 7 Model Assessment of Insulin Resistance (HOMA-IR). Insulin resistance supposedly contributes to poor glycemic control, 8 hypertension, dyslipidemia and accelerated atherosclerosis. Accordingly, TyG index has been iden fied as a predic ve marker for coronary artery atherosclerosis and non- 9,10 alcoholic fa y liver disease (NAFLD). Furthermore, several studies have shown good associa on of TyG index with glycemic control, sugges ng its poten al role in T2DM 11,12 pa ent assessment and management. The aim of this study was to explore the associa on of TyG index (and other TyG derived indices) with HbA1c and evaluate their predic ve ability for glycemic control in pa ents with T2DM. We expect TyG index to have a sa sfactory ability to discriminate pa ents with poor glycemic control, sugges ng its clinical u lity in type 2 DM.

METHODOLOGY
This cross-sec onal analy cal study comprised of 160 adult pa ents diagnosed with type 2 diabetes mellitus (T2DM) visi ng the medical outpa ent department of Chitwan Medical College, Bharatpur, Chitwan. The sample size was calculated using the sample size formula for es ma ng a sample size necessary to es mate an AUROC curve (Source: h ps://sample-size.net/sample-size-ci-for-auroc/). Based 12 on a previous study, using an es mated AUROC (θ) = 0.83, es mated propor on of diabe cs having poor glycemic control (P) = 0.46, width of confidence interval (0.13) at 95% confidence level, the calculated sample size was 163. Convenience sampling method was used. This study was conducted from July -December 2019. Pa ents under thyroid medica ons or steroid therapy (that could alter blood glucose), pregnant women, pa ents with type 1 diabetes mellitus and other co-exis ng serious illness or inflamma on were excluded from the study. Research par cipants who had highly elevated serum triglyceride (TG) (>500 mg/dl) or had taken medica ons lowering primarily serum TG (e.g. fenofibrate) were also excluded. Ethical approval was obtained from Chitwan Medical A er explaining about the study and receiving wri en consent, socio-demographic data and clinical data including the anthropometric measurements [weight, height, waist circumference (WC)] were collected from the research par cipants, following standard protocol on day 1. On the nextmorning (day 2), about 5 ml of blood was drawn from the pa ents in the fas ng state (a minimum 8 hours of fas ng) by venipuncture using asep c technique in two vacutainers: one for serum analysis (vacutainer with clot ac vator) and the other for HbA1c measurement in whole blood (vacutainer with the antocoagulant EDTA). The biochemical parameters included fas ng blood glucose (FBG), lipid profile [Total cholesterol (TC), Triglyceride (TG), Low-density lipoprotein-cholesterol (LDL-c), High-density lipoprotein-cholesterol (HDL-c)] and glycated hemoglobin (HbA1c). The measurement of glucose and lipid parameters was done by standard colorimetric assays using DIMENSION Clinical Chemistry System, SIEMENS. HbA1c was measured by Ion-exchange High Performance Liquid Chromatography (HPLC) method. The pa ents were categorized into two groups of glycemic control based on HbA1c levels; <7.0% as "good glycemic control" and ≥ 7.0% as "poor glycemic control". The selec on of these 14 cutoff values was based on earlier studies. The categoriza on of pa ents into different obesity statuses was based on 15 WHO classifica on based on Body Mass Index (BMI). TyG indices were calculated according to established 7,16 formulae. TyG index = ln [fas ng TG (mg/dl) x fas ng glucose (mg/dl)/2] TyG-WC = TyG index * WC TyG-BMI = TyG index * BMI Sta s cal Package for the Social Sciences (SPSS) ver. 20 was used to analyze the data. Normality in data distribu on in the variables was assessed by Shapiro Wilk test. Con nuous variables were expressed as mean ± SD or median (Q1 -Q3) depending upon their distribu on, whereas categorical variables were expressed as frequency (%). The comparison of different con nuous variables between good and poor glycemic control groups was done by Independent sample ttest or Mann-Whitney U test as appropriate. Chi-squared test was used to analyze the difference in propor on between groups, for categorical variables. Spearman's rank correla on coefficient (rho) was used to explore correla on between variables. Receiver opera ng characteris c (ROC) curve analysis was performed to analyze the predic ve ability of TyG-index and TyG-WC for poor glycemic control. The op mal cut-off was derived based on the value that had maximal sensi vity and specificity. P<0.05 was considered to be sta s cally significant.

RESULTS
A total of 160 pa ents (mean age: 53.6 ± 10.7 years) were included in the study. Eighty-eight (55.0%) were males.    Table 2 illustrates the differences in different clinical and laboratory parameters between pa ents with poor and good glycemic control. The two groups had similar distribu on of age, gender, BMI and dura on of DM. As expected, the poor glycemic control group had significantly higher fas ng blood glucose (FBG) and glycosylated hemoglobin (HbA1c). TyG index, along with TyG-BMI and TyG-WC were also significantly increased in the poor glycemic control group. [ Table 2] TyG index had the best posi ve correla on with HbA1c, compared to TyG-WC and TyG-BMI (Spearman's rho = 0.65 vs. 0.36 and 0.33 respec vely, P<0.001). No significant associa on was observed between these indices and other lipid profile parameters. On ROC curve analysis, TyG index had a much be er predic ve ability for glycemic control compared to other two indices (AUC: 0.803, 95% CI: 0.731 -0.874). TyG-BMI had the least discrimina ng ability (AUC: 0.644, 95% CI: 0.556 -0.733). Fig 1 depicts the ROC curves for TyG index and TyG-WC. A TyG value ≥ 9.12 was noted to be op mal cutoff for predic ng poor glycemic control, with 86.1% sensi vity and 61.5% specificity. # p<0.05, *p<0.001

DISCUSSION
Our study showed that TyG index and other TyG-derived indices (such as TyG-WC and TyG-BMI) are significantly associated with glycemic control in adult pa ents with type 2 diabetes mellitus. TyG index, in par cular showeda good discriminant ability for good glycemic control. The indices had poor correla on with other lipid profile parameters, however. The product of TG and FBG (TyG index) has been shown to be a surrogate marker for es ma ng insulin resistance (IR) in healthy subjects, with a high degree of correla on with the commonly used HOMA-IR index and the gold standard 7,17 euglycemichyperinsulinemic clamp test. IR basically involves two phenomena: reduced sensi vity of the muscle and adipose ssue towards insulin and reduced ability of the 18 liver to suppress hepa c glucose produc on and output. Because there is impairment in the oxida on and u liza on of fa y acids (FAs) in IR, the resultant increase in the flux of free FAs from adipose to non-adipose ssue amplifies the 19 fundamental metabolic derangements, characteris c of IR. TyG index has been reported to be a be er and efficient marker than other indices such as Visceral Adiposity Index (VAI), lipid ra os and Lipid Accumula on Product (LAP) in 20 early iden fica on of IR. Similarly, BMI and WC are simple, non-invasive anthropometric parameters that are adopted as indicators of obesity. As obesity and IR are closely related, Timalsina S et al combina on of these with TyG index could provide addi onal value in iden fying IR, as has been shown in one 16 of the studies. Moreover, TyG index has been proposed as a simple biochemical marker that could iden fy individuals 21 at high risk of developing diabetes in a large cohort study. In our study, TyG index was significantly and posi vely associated with HbA1c in pa ents with T2DM, also 11,12 observed in studies elsewhere. A study by Babic et al. in 113 pa ents showed elevated TyG index in pa ents with poor glycemic control and suggested the u lity of the index in the overweight and obese group among the diabe c 11 pa ents. Similarly, Hameed et al. in 293 pa ents found that TyG and TyG-derived indices (TyG-WC and TyG-BMI) were significantly increased in the "poor glycemic control" group. Moreover, the study also reported that TyG index had the largest AUC (0.833) for predic on of poor glycemic 12 control, as observed in our study (AUC = 0.803). The dyslipidemia present in T2DM pa ents is characterized by elevated triglycerides, low HDL-c and predominance of small-dense LDL par cles. It is increasingly being recognized that these lipid changes are not only the consequences of 22 the impaired glucose metabolism, but also cause them. It is hypothesized that elevated levels of free fa y acids (FFAs) in these pa ents, a consequence of elevated TG, induce insulin resistance and beta-cell dysfunc on by a) disrup ng or modula ng the cascade linking insulin receptors with glucose transporters (GLUTs) and b) inducing subclinical inflamma on ac ng synergis cally with pro-inflammatory 23,24 adipokines released by adipose ssue. It has been shown that comprehensive biomarker tes ng of IR (and beta cell func on) has greater sensi vity to detect diabetes risk and is associated with improved glycemic 25 control in clinical prac ce. HbA1c is the most commonly used and established method for assessment of glycemic control, however the test is expensive, needs to be highly standardized and might be unavailable in many hospital laboratories of low-income countries such as ours. TyG index, a reliable surrogate marker for IR, could be a simple and low-cost alterna ve for screening for insulin resistance, glycemic control status, and effec veness of the treatment regimen during pa ent follow up.

CONCLUSION
TyG index could be a simple and cost-effec ve screening tool for assessment of insulin resistance and glycemic control in type 2 diabe c pa ents, par cularly in primary health care se ng or when HbA1c measurement might not be financially/logis cally feasible.

RECOMMENDATIONS
Further prospec ve researches are needed exploring the poten al clinical u lity of TyG index in terms of disease progression and effec veness of diabe c treatment regimen. The reliability of the TyG index has to be confirmed by studies involving a large group of type 2 DM pa ents.

LIMITATIONS OF THE STUDY
The major limita ons of our study are small sample size and single-center study design. Because of the cross-sec onal nature of the study, cause-effect rela onships could not be deduced (such as between TyG index and glycemic control). Some of the clinical informa on, such as use of insulin/oral hypoglycemic agents, family history of diabetes were missing.