Association of long-term triglyceride-glucose index patterns with the incidence of chronic kidney disease among non-diabetic population: evidence from a functional community cohort

The present study was performed based on a prospective cohort design and the aim was to evaluate the association of TyG index variability or cumulative TyG index with the incidence of CKD among non-diabetic population. Participants who underwent annual health examination from 2015 to 2017 (baseline and index year) were included and long-term TyG index patterns from three health examination were calculated as predictors of future CKD. All the participants were annually followed under same identical conditions for the development of CKD until December 31, 2021.

The population data used in this study were obtained from the Beijing Functional Community Cohort, which was initiated in 2010 and 8671 individuals with annual examination were recruited at the Health Management Center of Xuanwu Hospital, Capital Medical University [18]. These participants were from various occupations in Xicheng District, Beijing, including medical workers, teaching staff, government workers, workers, and service industries, representing most of the occupational population in Beijing. This cohort excluded participants with a history of medication for elevated blood glucose and dyslipidaemia. Those with acute inflammatory disease, hepatic or renal failure, heart failure, autoimmune disease or cancer were also excluded. The study protocol was approved by the ethics committee of Capital Medical University and Xuanwu Hospital, and it was conducted according to the principles of the Declaration of Helsinki. All study participants were informed at enrollment.

In this analysis, as the collection of information on the cohort population was more limited prior to 2015, we included participants who underwent one health examination per year between January 2015 and December 2017. Of 6759 participants, we excluded those with with missing data on FBG or lipid in 2015, 2016 or 2017 wave (n = 286) and those who were treated with antidiabetic or lipid-lowering agents (n = 371). We further excluded those with a history of type 2 diabetes (T2D) or CKD in 2017. Finally, 5484 participants were included in our analysis and were annually followed under same identical conditions for the development of CKD until December 31, 2021. Fig. S1 in Additional file 1 showed the flow chart of this prospective cohort study and Table S1 in Additional file 1 presented the baseline characteristics of participants included and excluded from this study.

Venous blood samples were collected between 7:30 and 8:30 am after an overnight fast. In a calm state, all samples were immediately centrifuged for laboratory measurements. Low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), TG, alanine aminotransferase (ALT), aspartate aminotransferase (AST), Urea, creatinine (Cr) and uric acid (UA) were measured using standard laboratory methods (Hitachi Autoanalyzer 7060; Hitachi, Tokyo, Japan). FBG was determined by the glucose oxidase method. Hemoglobin A1c (HbA1c) levels were measured by Ion-exchange chromatography. T2D was defined as FBG ≥ 126 mg/dL or HbA1c ≥ 6.5% or self-reported history of T2D.

The TyG index was measured ln (fasting TG [mg/dL]×FPG [mg/dL]/2). TyG index variability was defined as intraindividual variability recorded in the health examinations from 2015 to 2017 and coefficient of variation (CV) was first used to determine the variability of TyG [17, 19]. Standard deviation (SD), average real variability (ARV) and variability independent of the mean (VIM) were also used to determine the variability of TyG in sensitivity analysis and detailed information was presented in Supplementary Materials.

As previous studies described [20, 21], cumulative TyG index was defined as the summation of the average TyG index for each pair of consecutive examinations multiplied by the time between these two consecutive visits, in years: ([TyG index_{visit 1}+TyG index_{visit 2}]/2×time_1 − 2) + ([TyG index_{visit 2}+TyG index_{visit 3}]/2× time_2 − 3).

TyG index_{visit 1}, TyG index_{visit 2}, TyG index_{visit 3} indicate the TyG index at the first, second and third examinations, respectively, and time_1–2, time_2–3 indicate the participant-specific time intervals between consecutive visits in years.

In this study, CKD was defined as estimate glomerular fltration rate (eGFR) < 60 ml/min/1.73 m² or onset of macroalbuminuria (UACR > 300 mg/g) according to Kidney Disease: Improving Global Outcomes (KDIGO) guidelines [22]. Besides, self-reported diagnosis of CKD was also included in the diagnostic criteria for CKD in this study.

eGFR was calculated based on Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) [23]. The sex specific formula is as follows:

Women: if Cr ≤ 0.7 mg/dL, eGFR = 144*(Cr/0.7)^−0.329*(0.993)^Age.

if Cr > 0.7 mg/dL, eGFR = 144*(Cr/0.7)^−1.209*(0.993)^Age.

Men: if Cr ≤ 0.9 mg/dL, eGFR = 141*(Cr/0.9)^−0.411*(0.993)^Age.

if Cr > 0.9 mg/dL, eGFR = 141*(Cr/0.9)^−1.209*(0.993)^Age.

Factors associated with outcomes were considered as covariates. Based on previous studies, covariates in this study included age, sex, current smoking, current drinking, educational level, physical activity, hypertension, BMI, LDL-C, HDL-C, ALT, AST, Cr, UA, Urea, eGFR and baseline TyG index.

Structured standard questionnaires were used to collect basic information about the study participants, including sex, age, current smoking, current drinking, education level, and physical activity. Current smoking was defined as those who smoked ≥ 1 cigarette per day. Current drinking was defined as the intake of wine/beer/cider/spirits ≥ 1 time per week. Physical activity was defined as walking or cycling ≥ 15 min per day, exercising or physical activity > 2 h per week or lifting or carrying heavy loads at work per day.

Height and weight were measured while the participants wearing light clothing without shoes, and body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters (kg/m²). Blood pressure (BP) was averaged after three consecutive measurements using a standard mercury sphygmomanometer on the right arm of the study subject after 5 min of sitting still. Hypertension was defined as systolic blood pressure (SBP) ≥ 140 mmHg or diastolic blood pressure (DBP) ≥ 90 mmHg or self-reported history of hypertension.

Categorical variables were presented as number (proportions), continuous variables were presented as mean ± SD or median (quartiles). Difference between groups were compared using Chi-square test for categorical variables and t test, One-way ANOVA or Kruskal-Wallis rank sum test for continuous variables, as appropriate.

Intraclass correlation coefficient (ICC) was first calculated to assess the stability and consistency of repeated TyG index measurements and if the ICC is greater than 0.75, it indicates a high stability of the repeated measurements. Multivariable-adjusted Cox proportional hazards regression models were used to estimate the association of quartiles of baseline TyG index, TyG index variability or cumulative TyG index with incident CKD and hazard ratios (HR) and 95% confidence intervals (CI) were listed. Model 1 was adjusted for age and sex. Model 2 was further adjusted for current smoking, current drinking, educational level, physical activity, hypertension, BMI, LDL-C, HDL-C, ALT, AST, Cr, UA, Urea and eGFR based on Model (1) Model 3 was further adjusted for baseline TyG index based on Model (2) Time to first CKD event was examined using Kaplan-Meier survival curves and compared using Log-rank test. Restricted cubic spline (RCS) was conducted to exploit the dose-response relationship between TyG index variability or cumulative TyG index and the incidence of CKD. We further divided participants into four groups according to median of TyG index variability and median of cumulative TyG index and compared the risks of CKD among different groups. The relative excess risk due to interaction (RERI) and the attributable proportion due to interaction (AP) were also calculated to assess the synergy. If 0 is outside the 95% CIs of RERI and AP, there are additive interactions of TyG index variability and cumulative TyG index on the incidence of CKD. C statistics, net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were used to estimate the improvement in discrimination and reclassification after adding baseline TyG index, TyG index variability and cumulative TyG index to the clinical risk model for CKD prediction [24].

Multiple sensitivity analyses were performed to validate the robustness and reliability of our results. First, participants experiencing CKD within one year were excluded to minimize potential reverse causality. Second, to account for the possible changes in TyG index levels before index year, the mean values of TyG from 2015 to 2017 instead of baseline values was adjusted in Model 3 of the Cox proportional-hazards model. Third, considering the possible influence of abnormal TG concentrations on TyG index levels or outcomes, restricted analysis was performed by excluding participants with baseline TG concentrations ≥ 1.7 mmol/L. Besides, SD, VIM and ARV were further used to determine the variability of TyG index and the primary analysis was repeated. The potential effect modification by age, sex, BMI and hypertension categories was also evaluated using stratified analysis and interaction testing using a likelihood ratio test.

Statistical analyses were performed using R software (version: 4.2.1; R Foundation for Statistical Computing) and MedCalc® Statistical Software (version 20.100; MedCalc Software Ltd, Ostend, Belgium; https://​www.​medcalc.​org; 2022). The difference was considered statistically significant at two-side significance level of P < 0.05.