Triglyceride-glucose index is associated with poor sleep quality in apparently healthy subjects: A cross-sectional study

ABSTRACT Objectives: We aimed to evaluate the association between the triglyceride glucose index (TyG index) and sleep quality and to establish a cut-off value for the TyG index based on the prevalence of subjects with insulin resistance (IR). Materials and methods: This cross-sectional study involved Brazilian health professionals (20-59 years). A total of 138 subjects answered the Pittsburgh Sleep Quality questionnaire to evaluate sleep quality. They were categorized into two groups: good sleep quality (global score ≤ 5 points) and poor sleep quality (global score ≥ 6 points). Also, we classified the subjects as having a high (>8.08 or >4.38) or low TyG index (≤8.08 or ≤4.38). Results: The majority of the subjects (70%) with high TyG index values (>8.08 or >4.38) reported poor sleep quality (p ≤ 0.001). Those with poor sleep quality had a 1.44-fold higher prevalence of IR (TyG index >8.08 or >4.38) compared to those with good sleep quality, regardless of sex, total cholesterol, LDL/HDL ratio, insulin, complement C3, CRP, and adiponectin (p ≤ 0.001). Conclusion: Our data showed a positive and significant association between the TyG index and poor sleep quality. Thus, these findings support the association between poor sleep quality and IR.


INTRODUCTION
I nsulin resistance (IR) is a condition in which the molecular mechanisms of insulin uptake and degradation are impaired, leading to the development of type 2 diabetes (T2D) and cardiovascular diseases in the long term (1)(2)(3). More than 500 million individuals were living with T2D globally in 2018, and it is expected to have a high prevalence in low-income countries (4). Adults with diabetes have a higher risk for all-cause morbidity and mortality because they often present other major comorbidities such as cardiovascular, chronic lower respiratory, and kidney diseases (5). These complications are mediated by several inflammatory markers, including cytokines such as tumor necrosis factor-alpha (TNF-α), interleukin-1 beta (IL-1β), and interleukin-6 (IL-6), that trigger an inflammatory response (6-8). Since IR can contribute to the pathogenesis of diabetes and its related comorbidities (3,9,10), understanding its mechanisms is of great importance.
Poor sleep quality is a common issue in modern society for several reasons, and growing evidence has linked it with . For example, as a direct consequence of the COVID-19 pandemic, sleep problems have affected approximately 40% of people in general and healthcare populations (15)(16)(17)(18). While short sleep duration and metabolic impairments are strongly associated (19), their mechanisms remain largely unknown. There is some support for the roles of the hypothalamic-pituitary-adrenal (HPA) axis and sympathetic activation in glucose impairments and IR due to inadequate sleep quality (20,21).
The triglyceride glucose index (TyG index) has been extensively used as a reliable marker for IR, expressed as the product of triglyceride and glucose levels (22,23). A meta-analysis showed a significant association between higher TyG values and T2D risk (24). Recently, two cross-sectional studies have associated TyG with obstructive sleep apnea (OSA), a sleep breathing disorder that often involves IR (25,26). However, the relationship between TyG and sleep quality has not been previously studied.
Given the limited evidence on the association between TyG index and sleep quality, we aimed to establish a cutoff value for the TyG index based on the prevalence of IR patients under the homeostatic model assessment of IR (HOMA-IR) and evaluate the association between TyG index and sleep quality. We hypothesize that higher TyG index values are positively associated with poor sleep quality.

MATERIALS AND METHODS
This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement. The STROBE checklist used is included in the Supplementary Material.

Study design and subjects
We analyzed data collected by a previous cross-sectional study in Viçosa, Brazil, that involved Brazilian health professionals between 20 and 59 years old (27). All subjects signed a consent form previously approved by the Human Research Ethics Committee of the Federal University of Viçosa (Ref. No. 005/2011;Viçosa, Brazil) under the principles of the Declaration of Helsinki. This study is not registered in "Plataforma Brasil" because it was approved on February 18, 2011, before "Plataforma Brasil" came into effect on January 2, 2012.
To be eligible for this study, health professionals (doctors, nurses, nutritionists, physical trainers, physiotherapists, dentists, pharmacists, biochemists, and psychologists) must work in health facilities or higher education institutions, and students must be in their last two years of courses in a health-related area. The recruitment was performed via phone calls, website disclosures, social networks, local radio, and pamphlets. Individuals who were pregnant, lactating, using corticosteroids, using antibiotics, had a cancer diagnosis within the last three years, or had any serious illness that required hospitalization at the time of this study, were excluded. Individuals who could not follow the measuring protocols such as weighing, blood pressure, or performing blood collection were also excluded. All data were collected between January 2012 and July 2013.
As a baseline, we surveyed 976 healthcare professionals in Viçosa, Brazil. The calculated sample size was 223 subjects, with a 95% confidence interval (CI), 5% sampling error, and an expected metabolic syndrome prevalence of 25%. However, our sample size is contingent on the subset of participants who had completed the Pittsburgh sleep quality index (PSQI) questionnaire (N = 138; Figure 1). Supplementary Table 1 shows the characteristics of participants who had and had not completed the PSQI questionnaire. Subjects who had not completed the questionnaire showed higher TyG index and very-low-density lipoprotein cholesterol (VLDL-c) values than those who completed the PSQI questionnaire. 976 healthcare professionals were identi ed 223 health professionals were required according to sample calculation 100 were excluded: All participants were instructed to answer the PSQI questionnaire, but for personal reasons 100 did not respond 238 were available 138 were analyzed Figure 1. Flowchart of origin of data used in this study. Assuming a prevalence of TyG index > 8.08 or > 4.38 in the exposed (poor sleep quality) and nonexposed (good sleep quality) groups, respectively, the analysis had 89.49% power to detect a difference of this magnitude or larger, determined using the OpenEpi online software (28).

Collected data
Sleep Quality (Exposure) Sleep quality was assessed by the adapted and validated Brazilian version of the PSQI (29). The information refers to the last month and contains nineteen items that cover seven components: subjective sleep quality (contentment at daily sleep), sleep latency (extended sleep onset time), sleep duration, habitual sleep efficiency (proportion of hours slept relative to total hours in bed), sleep disorders (disruption of sleep), use of sleeping medication, and daytime dysfunction (difficulty staying awake during social activities) (30). We assessed the global PSQI score from 0 to 21 points (30). Finally, volunteers were categorized into two groups: good sleep quality (global PSQI score ≤ 5) or poor sleep quality (global PSQI score ≥ 6).

Dietary intake assessment and lifestyle
Dietary intake was assessed by a semi-quantitative food frequency questionnaire, validated for a Spanish population, and adapted for Brazilian citizens, with 136 food items (31). Nutrient intake was estimated using ad hoc computer software specifically developed for this aim. In addition, updated information from Brazilian food composition tables was considered. A trained professional was responsible for administering the questionnaire to minimize potential bias. To evaluate physical activity, we used the international physical activity questionnaire (IPAQ), which is validated for the Brazilian population (32). Smoking habit was determined by asking the participants whether they were smokers, former smokers, or nonsmokers.

Anthropometric, body composition, and blood pressure
Weight and height were measured to calculate body mass index (BMI) by dividing the weight (kg) by height (m) squared. Overweight was classified as BMI ≥ 25 kg/m². Waist circumference (WC) was measured at the midpoint between the last rib and the iliac crest using a flexible and inelastic tape measure. Hip circumference (HC) was measured in the greater protuberance in the gluteal region. The waist-to-height ratio was calculated as the ratio of WC (cm) and height (cm). Body composition was evaluated with a BMI 310 Bioimpedance Analyzer (Biodynamic Research Corporation; San Antonio, TX, USA) according to standardized measurement conditions (27). The systolic and diastolic blood pressures were measured using an Omron HEM-742INT digital sphygmomanometer (Hoffman Estates, IL, USA) according to the protocol recommended by the European Society of Hypertension and the European Society of Cardiology (33). Research team members were suitably trained to obtain these measurements.

Metabolic markers (Outcomes)
Venous blood samples were drawn following a 12hour fast, centrifuged at 3500 rpm at 4 °C for 10 min (Megafuge 11R; Thermo Scientific, Waltham, MA, USA), and stored at -80 °C. A trained health professional was responsible for the blood collection. Triglyceride, total cholesterol, high-density lipoprotein cholesterol (HDL-c), and uric acid levels were determined by the enzymatic colorimetric method.

Statistical analysis
Statistical analyzes were performed using MedCalc (v.9.3; Ostend, Belgium) and the R statistical software (v.4.1.0). Subjects were classified into two groups according to the global PSQI score: good sleep quality (≤5) and poor sleep quality (≥6). Subjects with missing PSQI data were excluded from the analyses. A cutoff value for the TyG index was estimated, taking the presence of IR (HOMA-IR>2.71) as a reference (36). Then, the area under the curve (AUC) for receiving operating characteristic (ROC) curves was calculated to obtain sensitivity and specificity estimates. Variable normality was assessed by the Shapiro-Wilk test. The Student's t-test or Pearson Chi-square test was used to compare the subject characteristics according to the TyG index cutoff values. The data are presented as mean with standard deviation (SD) and frequencies. The prevalence of subjects with TyG index values above the cutoff was greater than 10% in our data. For this reason, Poisson regression with robust variance was used to assess the association between the TyG index (categorical and dependent variable) and the sleep quality (categorical and independent variable). The variables associated with the TyG index by the hypothesis test were used to adjust the regression analysis. The variable VLDL-c was calculated using the triglyceride values. Therefore, we do not include it as an adjustment variable. The data are presented as prevalence ratio (95% CI). A 5% significance level was used for all tests performed.

RESULTS
Of the 138 subjects included in the study, 103 (75.7 %) were female with a mean age of 29.17 (SD = 7.23) years, 39.9 % presented poor sleep quality evaluated through PSQI, and 8.9% presented IR evaluated by HOMA-IR. We estimated cutoffs for the TyG index, taking IR presence as a reference. Therefore, subjects were classified as having a TyG index of ≤8.08 or >8.08, or ≤4.38 or >4.38 ( Figure 2). 100% sensitivity and 51.2% specificity were found for the optimal TyG index cutoff (Supplementary Tables 3 and 4).
Higher levels of total cholesterol, VLDL-c, LDL-c, LDL/HDL ratio, insulin, complement C3, CRP, and lower adiponectin levels were found in subjects with TyG index >8.08 or >4.38 ( Table 1). Characteristics of the participants according to sleep quality are presented in Supplementary Table 2. Briefly, subjects with poor sleep quality were mostly male and presented higher values of muscle mass, VLDL-c, insulin, and TyG index but lower percentages of body fat and carbohydrate intake than those with good sleep quality. However, while most subjects (70%) with high TyG index values (>8.08 or >4.38) were likely to have poor sleep quality, 41% had good sleep quality ( Figure 3A). We found that subjects with poor sleep quality had a 1.44-fold higher prevalence of IR (TyG index > 8.08 or > 4.38) compared to those with good sleep quality, regardless of sex and total cholesterol, LDL/HDL ratio, insulin, complement C3, CRP, and adiponectin levels ( Figure 3B).
Finally, we constructed a correlation matrix to explore the correlations of variables related to TyG index. TyG index was positively correlated with the PSQI score, waist circumference, waist-to-height ratio, body fat, diastolic blood pressure, cardiac frequency, total cholesterol, insulin, complement C3, and CRP. Conversely, higher TyG values were negatively associated with adiponectin levels ( Figure 4).

DISCUSSION
To the best of our knowledge, this is the first study to evaluate the association between the TyG index and sleep quality in ostensibly healthy adults. The two previous cross-sectional studies have associated TyG with OSA (25,26) but did not consider sleep quality. We found that subjects with poor sleep quality had a 1.44-fold higher risk of having a TyG index above the cutoff than those with good sleep quality, regardless of sex and total cholesterol, LDL/HDL ratio, insulin, complement C3, CRP, and adiponectin levels. A number of established mechanisms have addressed the link between metabolic disorder and IR, defined as a state that stimulates impairments in glucose uptake, particularly glycogen synthesis (2). This metabolic disorder causes hyperglycemia and leads to oxidative stress and inflammatory responses (2). IR also leads to dyslipidemia because adipocytes increase their release of free fatty acids, which are absorbed by the liver to form triglyceride-rich and VLDL-c particles in large circulating amounts (2,36).  Evidence has shown that sleep fragmentation can change glucose metabolism by reducing insulin sensitivity (19). Endocrine mechanisms underlie the influence of sleep on IR through inflammatory pathways and persistent activation of the sympathetic and HPA axis (37). Poor sleep quality has been associated with high inflammatory marker (38) and cortisol (39) levels. Sympathetic and HPA axis activation has been reported to increase catecholamine and cortisol secretion (37). Combined with a proinflammatory state, these factors could contribute to IR development. Additionally, poor sleep quality seems to have epigenetic effects and share genetic architecture with metabolic syndrome (40). A study found a genetic correlation between insomnia symptoms and HOMA-IR, suggesting the involvement of genetic variants (41).
Current data have indicated a close relationship between HOMA-IR and TyG index (22,42). In Brazil, a validated study concluded that the TyG index had a better performance than the HOMA-IR index for measuring IR in clinical practice (22). Furthermore, a population-based cross-sectional study found a correlation between TyG and other markers of IR, such as HOMA-IR and the hyperinsulinemic-euglycemic clamp (HIEC), in healthy subjects (42). A systematic review has found that the highest achieved sensitivity was 96% using HIEC (43). The highest specificity was 99% using HOMA-IR, with a cutoff value of 4.68 (43) (36). In this study, the TyG index > 8.08 or > 4.38 was the optimal value to identify IR in our samples, with 100% sensitivity and 51.2% specificity. These values suggest that subjects with a TyG index > 8.08 or > 4.38 have IR, with 0% false-positive cases. However, the TyG index is not a good measure to detect subjects without IR, with a high false-negative rate.
Previous studies have reported a positive association between the TyG index, IR, and related conditions such as T2D (24) and cardiovascular events (44). Subjects in our study with IR determined by a TyG index > 8.08 or > 4.38 had worse metabolic profiles, with higher total cholesterol, VLDL-c, LDL-c, LDL/ HDL ratio, insulin, complement C3, and CRP values and lower adiponectin levels than those with a TyG index lower ≤ 8.08 or ≤ 4.38. Moreover, we found simultaneous and positive correlations between the TyG index and cardiometabolic risk variables such as waist circumference, waist-to-height ratio, body fat, diastolic blood pressure, cardiac frequency, total cholesterol, and fractions (including VLDL-c, LDL-c and LDL/ HDL ratio), insulin, complement C3, and CRP. These observations are consistent with previous studies that found a worse metabolic profile in subjects classified in the highest quartiles of the TyG index (24,44).
Poor sleep quality and IR could contribute to chronic inflammation, but it remains challenging to manipulate factors such as diet and sleep that may affect inflammation experimentally (45). Complement C3 and CRP are prominent biomarkers for IR (46,47). As mentioned above, subjects with a TyG index > 8.08 or > 4.38 had higher values for these inflammation markers. Uemura and cols. have reported that higher serum CRP was associated with IR in a dose-dependent manner (47). In another study, complement C3 was strongly associated with IR, independent of the other components of metabolic syndrome (46). Moreover, our results have shown a simultaneous correlation between the TyG index and adiponectin, a crucial modulator of insulin sensitivity and chronic inflammation (48).
This study used a cutoff value for the TyG index > 8.08 or > 4.38 as a surrogate marker to estimate IR. It was associated with poor sleep quality among apparently healthy adults, and its predictive significance also correlates with other important independent risk factors. Although the prevalence in women and men was not statistically different between the sleep quality and TyG index categories, selection bias potentially limits our study because of the high frequency of females in our data. In addition, since it is a crosssectional study, it cannot establish a causal relationship. While the HOMA-IR test is not the gold standard for diagnosing IR, the euglycemic-hyperinsulinemic clamp is the gold standard, but it is impractical for use in large cohort studies. The PSQI questionnaire was available for all subjects in the study. However, 100 subjects did not respond to the questionnaire, creating additional selection biases in our study.
This study's greatest benefit is its use of ostensibly healthy adults before the onset of chronic diseases to highlight the potential involvement of poor sleep quality in IR, even before it manifests clinically. Our findings reinforce the need for further research into using the TyG index as a surrogate marker of IR and its relationship with sleep. In addition, understanding modifiable risk factors for IR in adults may offer more effective primary prevention efforts in an at-risk population and expansion of interventions to improve sleep quality.