Effect of changes in the triglyceride-glucose index on atherogenic lipid profiles in coronary artery disease patients receiving lipid-lowering therapy: a prospective cohort study

Abstract

Background

Insulin resistance (IR) is a key driver of cardiovascular disease. The triglyceride-glucose (TyG) index, derived from fasting triglyceride and glucose levels, has been proposed as a surrogate marker of IR. However, its effect on lipid control in patients with coronary artery disease (CAD) receiving lipid-lowering therapy (LLT) remains unclear.

Methods

In this prospective cohort study, biochemical parameters of 1393 CAD patients were measured and followed over a median of one year. Participants received either low-/moderate-intensity LLT or high-intensity LLT. Linear regression models, logistic regression analyses, and change-to-change analyses were conducted to comprehensively assess the associations between baseline levels, longitudinal changes, and status transitions of the TyG index and lipid parameters.

Results

Lipid levels differed significantly across TyG index tertiles, with the highest tertile showing more adverse profiles at baseline and follow-up. Higher baseline TyG levels were independently associated with increased follow-up atherogenic lipid parameters and failure to achieve targets of non-high-density lipoprotein cholesterol (Non-HDL-C, OR = 1.23, 95% CI 1.01–1.51) and remnant cholesterol (RC, OR = 1.38, 95% CI 1.09–1.76). Participants in the highest tertile had the highest odds of not achieving targets for LDL-C, Non-HDL-C, and RC. Each 1% increase in the index was associated with percent increases in lipids including LDL-C (β = 1.10), Non-HDL-C (β = 1.81), and RC (β = 4.92, all P < 0.001). Patients transitioning from the lowest to the highest TyG tertile showed significant lipid worsening, while improvement from the highest to the lowest tertile was associated with reductions in RC and Non-HDL-C. High-intensity LLT led to greater reductions in TyG and lipid levels, mitigating the adverse lipid effects of TyG elevation. The adverse effects were also more evident in women and in those with obesity or prior revascularization.

Conclusions

Higher TyG levels were positively associated with atherogenic lipid profiles and failure to achieve lipid goals. In addition, upward changes in TyG over time were related to worsened atherogenic lipid status, particularly among patients receiving low-/moderate-intensity LLT. These findings support the routine monitoring of TyG to identify patients at risk of poor lipid control who may require high-intensity LLT.

Graphical abstract

TyG, triglyceride-glucose index; LLT, lipid-lowering therapy; LDL-C, low-density lipoprotein cholesterol; TC, total cholesterol; Non-HDL-C, non-high-density lipoprotein cholesterol; RC, remnant cholesterol.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-025-02910-6.

Research insights

What is currently known about this topic?

• The TyG index reflects insulin resistance (IR) and is linked to dyslipidemia and cardiovascular risk.

• Most studies focus on single-timepoint TyG and cross-sectional lipid associations.

• Strategies to improve TyG levels remain underexplored.

What is the key research question?

• How do levels and changes in IR, as reflected by the TyG index, affect lipid control in CAD patients receiving different intensities of lipid-lowering therapy (LLT)?

What is new?

• This real-world cohort study is the first to assess longitudinal TyG–atherogenic lipid associations in CAD patients under LLT.

• Higher baseline TyG and rising TyG over time are associated with poorer lipid control.

• High-intensity LLT reduces TyG more effectively and attenuates its adverse lipid effects.

How might this study influence clinical practice?

• Incorporating routine TyG monitoring into lipid management may help identify patients at risk of poor lipid control and support high-intensity LLT.

Introduction

Cardiovascular disease (CVD) is the leading cause of morbidity and mortality worldwide, with coronary artery disease (CAD) accounting for a substantial proportion of this burden [1, 2]. Notably, insulin resistance (IR) is now recognized as a critical contributor to adverse outcomes in patients with CAD [3, 4]. The triglyceride-glucose (TyG) index, defined as the natural logarithm (ln) of [fasting triglyceride (TG, mg/dL) × fasting blood glucose (FBG, mg/dL) /2], has emerged as a reliable and convenient clinical marker of IR [5–7]. Numerous studies have demonstrated that elevated TyG levels independently predict adverse cardiovascular outcomes, and correlate closely with dyslipidemia [8–10]. Lipid metabolism plays a central role in CAD development, with traditional markers such as low-density lipoprotein cholesterol (LDL-C) and apolipoprotein B (ApoB) established as key therapeutic targets [11, 12]. Nevertheless, residual cardiovascular risk persists even when LDL-C levels are optimally controlled [11, 13, 14]. Novel parameters such as remnant cholesterol (RC), lipoprotein(a) [Lp(a)], and LDL‑C corrected for the cholesterol content of Lp(a) (LDLLp(a)corr) reflect additional atherogenic burden and predict incident events [15–17]. Elucidating how TyG relates to this broader lipid profile may improve understanding of its impact and support better cardiovascular risk management.

The TyG index has been reported to be associated with various lipid parameters. Specifically, higher TyG levels are linked to elevated TG, decreased high-density lipoprotein cholesterol (HDL-C), and increased atherogenic lipids, including LDL-C, total cholesterol (TC), very-low-density lipoprotein cholesterol (VLDL-C) and small, dense LDL particles [18, 19]. This association reflects not only the index’s mathematical dependence on TG, but also the underlying influence of IR [5]. Beyond impairing glucose regulation, IR plays a central role in lipid metabolism disorders by promoting hypertriglyceridemia, enhancing free fatty acid (FFA) flux from adipose to other tissues, and facilitating the accumulation of cholesterol- and TG-rich ApoB-containing lipoprotein remnants in the arterial wall [20, 21]. This IR-related dyslipidemia is well reflected by elevated TyG index levels, reinforcing its value as a marker of cardiometabolic risk [22]. Additionally, emerging evidence suggests that the association between the TyG index and adverse health outcomes may be modified by clinical conditions and pharmacologic treatments, such as the use of lipid-lowering therapies (LLT), blood pressure or diabetes status, highlighting the importance of accounting for these modifiers when interpreting TyG-related risk [23, 24]. However, most existing evidence is based on single baseline TyG measurements. How longitudinal TyG changes relate to dynamic lipid profiles—particularly novel markers—remains unclear, and these associations may be further influenced by factors such as intensity of LLT.

Beyond examining the cross-sectional associations between TyG and lipid metabolism, recent research has begun to explore whether longitudinal changes in TyG and IR-related markers are linked to health outcomes. Cohort evidence suggests that rising levels of these markers are associated with increased risks of cardiovascular disease (CVD), all-cause and cardiovascular mortality, stroke, and hypertension, highlighting the clinical importance of TyG dynamics [25–28]. However, evidence on TyG variability in relation to lipid metabolism remains limited. To evaluate variable transitions more systematically, the “change-to-change” approach has been introduced. This method assesses the impact of shifting risk states and has been applied in studies on biological aging and lifestyle modification [29]. Applying this framework to TyG could provide new perspectives on its relationship with lipid profiles.

Realworld data on whether changes in TyG associate with lipid improvements, especially under different LLT intensities and across clinical subgroups (e.g., sex, prior interventions, diabetes, hypertension) are limited. Thus, we conducted a longitudinal study of CAD patients to examine how baseline TyG, its longitudinal change, and tertile transitions relate to a comprehensive panel of lipid parameters (including LDL-C, non-HDL-C, RC, ApoB, Lp(a), and LDLLp(a)corr), and to determine whether these associations vary with LLT intensity or across key clinical subgroups.

Methods

Study participants

This was a prospective cohort study conducted at Fuwai hospital (Beijing, China), and enrolled consecutive patients with clinically confirmed CAD at baseline. Eligible patients were those admitted to our center between August 2020 and February 2023 who met at least one of the following CAD related syndrome: acute coronary syndrome, prior myocardial infarction, angina pectoris, previous coronary revascularization, suspected vasospastic or microvascular disease or coronary angiography showing ≥ 50% stenosis in ≥ 1 major coronary artery [30, 31].

Patients were excluded if they met any of the following criteria: [1] missing TG or FBG data; [2] severe hepatic and/or renal dysfunction; [3] cardiomyopathy; [4] serious hematologic disorders; [5] advanced malignancies with limited life expectancy; [6] substantial lack of baseline medical data. In the end, 1393 patients were enrolled, and written informed consent was received from all participants prior to enrollment. This study was approved by the ethics committee of Fuwai hospital and complied with the Declaration of Helsinki.

Data measurement and definitions

Demographic characteristics, clinical data, past medical history, and medication use were extracted from patients’ electronic health records. Self-reported information was cross-validated with clinical documentation. Trained medical personnel assessed height, weight, pulse, systolic and diastolic blood pressure (SBP and DBP), and echocardiographic parameters during hospitalization. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters (kg/m²), and was further categorized according to the China Working Group criteria: underweight (< 18.5 kg/m²), normal weight (18.5–23.9 kg/m²), overweight (24.0–27.9 kg/m²), and obesity (≥ 28.0 kg/m²) [32].

Blood samples were collected after overnight fasting for the measurement of FBG, lipid profiles and biochemical indicators using an automated analyzer at the hospital’s central laboratory. Lipid parameters included TG, TC, LDL-C, HDL-C, ApoB, non-HDL cholesterol (non-HDL-C, calculated as TC − HDL-C), remnant cholesterol (calculated as TC − HDL-C − LDL-C), and Lp(a). Specifically, LDL-C was directly determined using the Low Density Lipoprotein Cholesterol Test Kit (Direct Surfactant Method; CB No. 151857). The TyG index was determined as ln [TG (mg/dL) × FBG (mg/dL) / 2] [33]. LDL-C corrected for Lp(a)-cholesterol content was estimated using two approaches: [1] LDLLp(a)corr 30 = LDL-C − 0.30 × Lp(a) [34]; [2] LDLLp(a)corr 17.3 = LDL-C − 0.173 × Lp(a) [35]. These corrections were based on differing assumptions regarding the proportion of Lp(a) mass attributable to cholesterol content [17].

Comorbidities and treatment definitions

Comorbidities were identified based on clinical history, self-reports, or ongoing medical treatments. Diabetes mellitus (DM) was defined as FBG ≥ 126 mg/dL (7.0 mmol/L), the 2-hour plasma glucose level during an oral glucose tolerance test ≥ 200 mg/dL (11.1 mmol/L), self-reported diabetes, or current use of hypoglycemic agents such as acarbose, metformin, insulin, sodium-glucose cotransporter‑2 (SGLT-2) inhibitors, or glucagonlike peptide-1 (GLP-1) receptor agonists [36, 37]. Hypertension was defined by either SBP ≥ 140 mmHg, DBP ≥ 90 mmHg, self-reported history of hypertension, or the current use of antihypertensive medications [27, 37]. Dyslipidemia was defined according to established clinical criteria and guidelines based on laboratory measurements, specifically the elevation of plasma TC, LDL C, TG, or the presence of low HDL-C levels that contribute to atherosclerosis [11, 38]. The use of LLT was only classified as an associated comorbidity when relevant laboratory abnormalities were present. Stroke and other diseases were identified through a combination of patient self-report, documented clinical diagnosis, and ongoing or historical use of relevant therapies (27). In addition, prior percutaneous coronary intervention (PCI) and other cardiovascular procedures were documented based on patients’ medical records [39].

LLT at discharge was classified based on intensity. High-intensity LLT included combinations of statins with ezetimibe or Proprotein Convertase Subtilisin/Kexin Type 9 (PCSK9) inhibitors, or monotherapy with high-dose rosuvastatin (20 to 40 mg/day) or atorvastatin (40 to 80 mg/day). Low-/moderate-intensity LLT involved lower-dose statins (e.g., atorvastatin 10–20 mg/day, rosuvastatin 10 mg/day) or ezetimibe monotherapy [11, 31].

Endpoints

Follow-up data were collected through scheduled outpatient visits and regular contact via telephone or messaging applications by an independent follow-up team. For this analysis, a median follow-up period of 1 year was selected to evaluate changes in biochemical parameters. Detailed records were maintained regarding biochemical parameters, medication usage, and any adverse events occurring during the follow-up period.

The primary endpoints included follow-up values, longitudinal changes and percent changes in TyG and lipid profiles, specifically LDL-C, TC, non-HDL-C, ApoB, RC, Lp(a), and LDL-C corrected for Lp(a)-cholesterol content. Secondary endpoints encompassed the non-attainment of lipid targets, including LDL-C, non-HDL-C, ApoB, and RC levels. According to the 2019 European Society of Cardiology/European Atherosclerosis Society guidelines for patients at very-high risk of atherosclerotic cardiovascular disease (ASCVD), the target levels are defined as follows: LDL-C < 1.4 mmol/L, non-HDL-C < 2.2 mmol/L, and apoB < 0.65 mmol/L [40]. As current guidelines do not specify a target for RC, a threshold of 0.5 mmol/L was adopted based on findings from previous population studies [31, 41].

Statistical methods

Categorical variables were presented as percentages (%), while continuous variables were described as mean ± standard deviation (SD) for normally distributed data or as median (range) for non-normally distributed data, as determined by the Shapiro-Wilk test. The baseline TyG index was analyzed both as a continuous variable and as categorical variable divided into tertiles (Tertile 1, Tertile 2, and Tertile 3). Between-group comparisons for continuous variables were performed using either Kruskal–Wallis test or one-way Analysis of Variance, depending on the distribution characteristics. Categorical variables were compared using the chi-square test or Fisher’s exact test, as appropriate. Spearman correlation analysis was performed to evaluate the associations between baseline TyG and lipid profiles, with statistical significance assessed by P-values and results visualized using a heatmap.

Univariable and multivariable linear regressions were performed to assess the relationships of baseline TyG (as a continuous variable and in tertiles) with follow-up lipid profiles, as well as the associations between percent changes in TyG and corresponding lipid changes (calculated as [(follow-up − baseline)/ baseline] ×100%). Covariates included age, sex, BMI, LLT at discharge, use of hypoglycemic agents at discharge, the baseline value of the corresponding lipid measure, and histories of PCI, DM, and hypertension. Univariable and multivariable logistic regressions were also conducted to evaluate the impact of baseline TyG on achieving target levels of LDL-C, non-HDL-C, ApoB, and RC at follow-up. Two sensitivity analyses were conducted: [1] Model 1 adjusted for age, sex, and the baseline value of the corresponding lipid measure; [2] Model 2 included all covariates from the final model and additionally adjusted for the pre-admission LLT regimen, in order to mitigate confounding arising from previous changes in the lipid profile before study initiation.

To assess the association between TyG dynamics and lipid profile changes, we applied a change-to-change analytical approach, using TyG status transitions (nine categories based on baseline and follow-up tertiles) as the exposure variable. This method mitigates bias from unmeasured time-invariant confounding by analyzing within-individual changes [42, 43]. A detailed description of the model structure and covariate selection is provided in the Supplementary Material.

As LLT plays a central role in metabolic regulation, we further examined whether its intensity influences TyG by comparing baseline and follow-up values, as well as percentage changes, across LLT intensity groups. Baseline TyG tertiles were compared between the low-/moderate intensity and high-intensity LLT groups using the chi-square test, whereas TyG levels at both time points were compared using the Mann–Whitney U test due to their non-normal distribution. Percentage changes in TyG and lipid parameters (e.g., LDL-C, non-HDL-C, ApoB) were compared between low-/moderate- and high-intensity LLT groups using the Mann–Whitney U test. The distributions were visualized using boxplots with significance annotations.

To investigate whether the relationship between TyG percent change and lipid profile improvements was modified by patient characteristics, subgroup analyses were performed according to LLT intensity, sex, BMI classification, history of PCI, DM, and hypertension. The underweight category in BMI classification was excluded from subgroup analyses due to insufficient sample size. Interaction terms between TyG percent change and each subgroup variable were introduced into generalized linear models, with adjustment for the same covariates used in the main models. Results were visualized using forest plots, with statistical significance for interaction defined as P for interaction < 0.05. All statistical analyses were performed in R software version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria), and P < 0.05 was the threshold for statistical significance.

Results

Clinical characteristics at baseline

A total of 1,393 participants (75.38% male, mean age 59 ± 10 years) were included in this analysis. Participants were categorized into three groups based on their baseline TyG levels, with the corresponding TyG tertiles being as follows: Tertile 1 (TyG ≤ 8.68, T1), Tertile 2 (8.68 < TyG ≤ 9.26, T2), and Tertile 3 (TyG > 9.26, T3). The BMI values significantly differed across the TyG tertiles (T1: 25.58 ± 3.53, T2: 26.18 ± 3.31, T3: 26.88 ± 3.24, P < 0.001). Additionally, the prevalence of diabetes was higher in the groups with higher TyG levels (T1: 30.11%, T2: 41.16%, T3: 56.47%, P < 0.001), as was the proportion of participants with a history of PCI (T1: 24.89%, T2: 26.36%, T3: 33.41%, P = 0.009).

In terms of medication use at discharge, around 85% of participants in all three TyG groups were treated with high-intensity LLT. However, in the highest TyG tertile (T3), the use of PCSK9 inhibitors and fibrates was significantly more frequent (Table 1). Notably, the proportion of patients on hypoglycemic drugs (e.g., acarbose, metformin, insulin, SGLT-2 inhibitors, GLP-1 agonists) and no background LLT prior to study initiation increased with higher TyG levels (Supplementary Table 1).

Table 1.

Baseline characteristics of the study population according to the TyG index tertiles

Variables	Total (N = 1393)	TyG tertile 1 (N = 465)	TyG tertile 2 (N = 464)	TyG tertile 3 (N = 464)	P value
Male, n (%)	1050 (75.38%)	347 (74.62%)	337 (72.63%)	366 (78.88%)	0.078
Age, years	58.96 ± 9.96	59.33 ± 10.23	59.46 ± 9.71	58.10 ± 9.89	0.073
BMI, kg/m2	26.21 ± 3.40	25.58 ± 3.53	26.18 ± 3.31	26.88 ± 3.24	< 0.001
SBP, mmHg	136.53 ± 16.98	136.19 ± 16.05	136.63 ± 17.35	136.76 ± 17.53	0.868
DBP, mmHg	78.76 ± 10.77	78.92 ± 10.85	78.03 ± 10.29	79.32 ± 11.13	0.179
ACS, n (%)	1232 (88.44%)	416 (89.46%)	412 (88.79%)	404 (87.07%)	0.500
PCI history, n (%)	389 (28.23%)	114 (24.89%)	121 (26.36%)	154 (33.41%)	0.009
CABG history, n (%)	19 (1.38%)	8 (1.74%)	6 (1.31%)	5 (1.09%)	0.688
Stroke, n (%)	103 (7.41%)	36 (7.76%)	33 (7.13%)	34 (7.34%)	0.933
Hypertension, n (%)	955 (68.56%)	299 (64.30%)	326 (70.26%)	330 (71.12%)	0.051
Hyperlipidemia, n (%)	991 (71.14%)	330 (70.97%)	329 (70.91%)	332 (71.55%)	0.972
Diabetes mellitus, n (%)	593 (42.57%)	140 (30.11%)	191 (41.16%)	262 (56.47%)	< 0.001
Smoking history, n (%)	681 (48.92%)	217 (46.67%)	228 (49.24%)	236 (50.86%)	0.435
FBG, mmol/L	6.34 (5.35, 8.11)	5.40 (4.91, 6.20)	6.33 (5.47, 7.66)	8.64 (6.67, 11.41)	< 0.001
Medications at discharge
High-intensity LLT, n (%)	1182 (84.85%)	382 (82.15%)	397 (85.56%)	403 (86.85%)	0.118
Statins, n (%)	1380 (99.07%)	459 (98.71%)	460 (99.14%)	461 (99.35%)	0.693
Ezetimibe, n (%)	1183 (84.92%)	384 (82.58%)	398 (85.78%)	401 (86.42%)	0.215
PCSK9i, n (%)	69 (4.95%)	18 (3.87%)	18 (3.88%)	33 (7.11%)	0.032
Fibrates, n (%)	36 (2.58%)	0 (0.00%)	0 (0.00%)	36 (7.76%)	< 0.001
Hypoglycemic drugs, n (%)	468 (33.60%)	104 (22.37%)	149 (32.11%)	215 (46.34%)	< 0.001

Data are expressed as mean + SD, median (Q1, Q3), or as number (%). Values in bold represent results that reached statistical significance (P < 0.05).

TyG, triglyceride–glucose index; LLT, lipid-lowering therapy; BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; ACS, acute coronary syndrome; PCI, percutaneous coronary intervention; CABG, coronary artery bypass graft; FBG, fasting blood glucose; LLT, lipid-lowering therapy; PCSK9i, proprotein convertase subtilisin/kexin type 9 inhibitors, such as evolocumab and alirocumab

Lipid profile differences across TyG tertiles at baseline and follow-up

Lipid profiles differed significantly across the TyG tertiles, with T3 consistently presenting the highest levels of most atherogenic lipid markers at both baseline and follow-up (Table 2).

Table 2.

Lipid profile differences across TyG index tertiles at baseline and follow-up

Variables	Total (N = 1393)	TyG tertile 1 (N = 465)	TyG tertile 2 (N = 464)	TyG tertile 3 (N = 464)	P value
LDL-C, measured
Baseline, mmol/L	2.14 (1.69, 2.74)	2.01 (1.58, 2.51)	2.17 (1.73, 2.72)	2.29 (1.80, 2.94)	< 0.001
Follow-up, mmol/L	1.62 (1.28, 2.10)	1.58 (1.24, 2.04)	1.59 (1.25, 2.07)	1.71 (1.33, 2.24)	0.010
Median change from baseline, %	− 23.04 (− 42.83, 1.77)	− 21.44 (− 40.44, 2.90)	− 26.19 (− 44.90, 1.30)	− 22.19 (− 43.47, 0.79)	0.200
TC, measured
Baseline, mmol/L	3.85 (3.28, 4.49)	3.61 (3.11, 4.24)	3.78 (3.28, 4.45)	4.16 (3.54, 4.77)	< 0.001
Follow-up, mmol/L	3.25 (2.77, 3.88)	3.17 (2.72, 3.74)	3.20 (2.70, 3.84)	3.46 (2.87, 4.06)	< 0.001
Median change from baseline, %	− 14.53 (− 29.98, 1.89)	− 12.97 (− 26.80, 5.27)	− 14.16 (− 31.09, 2.24)	− 17.23 (− 31.79, − 0.70)	0.004
Non-HDL-C, calculated
Baseline, mmol/L	2.72 (2.16, 3.34)	2.38 (1.92, 2.94)	2.69 (2.19, 3.20)	3.06 (2.50, 3.74)	< 0.001
Follow-up, mmol/L	2.10 (1.68, 2.66)	1.97 (1.60, 2.50)	2.05 (1.63, 2.60)	2.30 (1.84, 2.87)	< 0.001
Median change from baseline, %	− 21.83 (− 39.79, 1.02)	− 18.65 (− 34.68, 5.47)	− 22.70 (− 41.03, 2.58)	− 24.23 (− 42.39, − 4.81)	< 0.001
ApoB, measured
Baseline, g/L	0.74 (0.59, 0.89)	0.65 (0.52, 0.80)	0.73 (0.60, 0.87)	0.81 (0.67, 1.00)	< 0.001
Follow-up, g/L	0.62 (0.50, 0.76)	0.58 (0.49, 0.71)	0.62 (0.49, 0.76)	0.66 (0.55, 0.84)	< 0.001
Median change from baseline, %	− 15.24 (− 32.90, 4.27)	− 14.50 (− 29.13, 7.38)	− 16.42 (− 34.94, 5.27)	− 15.94 (− 33.33, 1.25)	0.294
RC, calculated
Baseline, mmol/L	0.50 (0.37, 0.68)	0.39 (0.30, 0.47)	0.51 (0.40, 0.62)	0.70 (0.54, 0.93)	< 0.001
Follow-up, mmol/L	0.44 (0.29, 0.63)	0.37 (0.23, 0.55)	0.42 (0.30, 0.62)	0.51 (0.35, 0.74)	< 0.001
Median change from baseline, %	− 19.35 (− 47.59, 27.64)	− 6.74 (− 40.37, 55.42)	− 19.51 (− 44.69, 26.48)	− 30.00 (− 55.24, 2.86)	< 0.001
Lp(a), measured
Baseline, mg/L	185.96 (71.76, 406.58)	200.84 (84.94, 451.28)	211.05 (81.25, 441.45)	144.20 (50.02, 344.10)	< 0.001
Follow-up, mg/L	153.61 (53.35, 378.60)	149.64 (56.10, 375.94)	180.00 (52.92, 435.60)	124.90 (49.26, 344.86)	0.229
Median change from baseline, %	− 6.35 (− 40.06, 26.05)	− 4.90 (− 40.25, 22.44)	− 7.16 (− 43.22, 21.30)	− 7.14 (− 33.88, 34.28)	0.414
LDL Lp(a)corr 30, calculated
Baseline, mmol/L	1.91 (1.44, 2.49)	1.79 (1.33, 2.27)	1.90 (1.48, 2.47)	2.07 (1.55, 2.77)	< 0.001
Follow-up, mmol/L	1.52 (1.17, 2.03)	1.49 (1.14, 1.94)	1.50 (1.11, 1.95)	1.59 (1.22, 2.18)	0.006
Median change from baseline, %	− 19.41 (− 42.75, 12.07)	− 16.05 (− 39.89, 18.31)	− 22.48 (− 45.64, 9.40)	− 18.71 (− 42.70, 9.56)	0.127
LDL Lp(a)corr 17.3 , calculated
Baseline, mmol/L	2.00 (1.55, 2.59)	1.86 (1.44, 2.34)	1.99 (1.61, 2.53)	2.16 (1.65, 2.85)	< 0.001
Follow-up, mmol/L	1.56 (1.22, 2.04)	1.52 (1.18, 1.98)	1.53 (1.18, 2.00)	1.66 (1.29, 2.20)	0.008
Median change from baseline, %	− 21.32 (− 42.13, 7.04)	− 18.82 (− 39.52, 12.23)	− 23.54 (− 44.97, 6.68)	− 20.11 (− 42.78, 5.64)	0.148

Values in bold represent results that reached statistical significance (P < 0.05). LLT, lipid-lowering therapy; LDL-C, low-density lipoprotein cholesterol; TC, total cholesterol; Non-HDL-C, non-high-density lipoprotein cholesterol; ApoB, apolipoprotein B; RC, remnant cholesterol; Lp(a), lipoprotein(a); LDL _{Lp(a)corr 30}, LDL-C corrected for Lp(a)-cholesterol content as 30% of total Lp(a) mass; LDL _{Lp(a)corr 17.3}, LDL-C corrected for Lp(a)-cholesterol content as 17.3% of total Lp(a) mass

At baseline, T3 showed the highest median levels of LDL-C (2.29 mmol/L vs. 2.17 mmol/L in T2 and 2.01 mmol/L in T1, P < 0.001), TC (4.16 mmol/L vs. 3.78 mmol/L in T2 and 3.61 mmol/L in T1, P < 0.001), and non-HDL-C (3.06 mmol/L vs. 2.69 mmol/L in T2 and 2.38 mmol/L in T1, P < 0.001). Similar trends were observed for ApoB, RC, LDLLp(a)corr 30, and LDLLp(a)corr 17.3 (all P < 0.01).

After LLT, these markers decreased across all groups, yet T3 maintained the highest follow-up levels for LDL-C (1.71 mmol/L), TC (3.46 mmol/L), non-HDL-C (2.30 mmol/L), RC (0.51 mmol/L), ApoB (0.66 g/L), LDLLp(a)corr 30 (1.59 mmol/L), and LDLLp(a)corr 17.3 (1.66 mmol/L) (all P ≤ 0.01). In terms of percentage change, greater reductions were observed in T2 and T3 compared with T1. For example, TC decreased by − 17.23% in T3, compared to − 14.16% in T2 and − 12.97% in T1 (P = 0.004); non-HDL-C showed a median reduction of − 24.23% in T3 vs. − 22.70% in T2 and − 18.65% in T1 (P < 0.001). Although the reduction in LDL-C was not statistically significant (P = 0.200), greater decreases were still observed in T2 (–26.19%) and T3 (–22.19%) compared to T1 (–21.44%).

Association of baseline TyG with Follow-up lipid goal attainment and lipid levels

As shown in Supplementary Fig. 1 and Table S2, baseline TyG index was associated with a more atherogenic lipid profile.

We next investigated the impact of baseline TyG on the likelihood of failing to achieve target lipid levels at follow-up, treating TyG both as a continuous variable and as a categorical variable based on tertiles (low, middle, and high). Logistic regression analysis demonstrated that higher baseline TyG was significantly associated with increased odds of failing to achieve target levels for LDL-C, non-HDL-C, ApoB, and RC (Table 3). After adjusting for potential confounders, baseline TyG remained independently associated with a higher risk of non-attainment of non-HDL-C target (odds ratio [OR] = 1.23, 95% confidence interval [CI] 1.01–1.51, P = 0.040) and RC target (OR = 1.38, 95% CI 1.09–1.76, P = 0.008). Although not statistically significant, similar negative trends were observed for LDL-C and ApoB.

Table 3.

Logistic regression analysis of baseline TyG index (as a continuous variable and categorical variable) and non-attainment of lipid profile targets

Variables	Baseline TyG	Univariable logistic regression			Multivariable logistic regression
		OR	95% CI	P value	OR	95% CI	P value
LDL-C ≥ 1.4mmol/L	Continuous	1.27	1.06, 1.53	0.009	1.21	1.00, 1.47	0.052
	Tertile 1	–	–	–	–	–	–
	Tertile 2	1.01	0.76, 1.34	0.968	0.98	0.72, 1.32	0.873
	Tertile 3	1.46	1.08, 1.97	0.014	1.40	1.01, 1.94	0.044
Non-HDL-C ≥ 2.2mmol/L	Continuous	1.56	1.31, 1.86	< 0.001	1.23	1.01, 1.51	0.040
	Tertile 1	–	–	–	–	–	–
	Tertile 2	1.18	0.89, 1.57	0.254	1.03	0.76, 1.38	0.870
	Tertile 3	2.07	1.56, 2.76	< 0.001	1.54	1.12, 2.13	0.008
ApoB ≥ 0.65 g/L	Continuous	1.39	1.11, 1.76	0.005	1.15	0.88, 1.49	0.301
	Tertile 1	–	–	–	–	–	–
	Tertile 2	1.48	1.01, 2.17	0.047	1.28	0.85, 1.92	0.236
	Tertile 3	1.95	1.33, 2.87	0.001	1.48	0.96, 2.27	0.075
RC ≥ 0.5mmol/L	Continuous	1.85	1.55, 2.22	< 0.001	1.38	1.09, 1.76	0.008
	Tertile 1	–	–	–	–	–	–
	Tertile 2	1.38	1.03, 1.85	0.030	1.18	0.87, 1.60	0.287
	Tertile 3	2.47	1.85, 3.30	< 0.001	1.63	1.15, 2.33	0.007

Multivariable logistic regression was adjusted for age, sex, body mass index, the baseline value of the corresponding lipid measure, discharge LLT, hypoglycemic agents at discharge, and the history of percutaneous coronary intervention, diabetes and hypertension. Values in bold represent results that reached statistical significance (P < 0.05).

TyG, triglyceride-glucose index; LLT, lipid-lowering therapy; LDL-C, low-density lipoprotein cholesterol; TC, total cholesterol; Non-HDL-C, non-high-density lipoprotein cholesterol; ApoB, apolipoprotein B; RC, remnant cholesterol; OR, odds ratio; CI, confidence interval

When TyG was analyzed by tertiles, participants in the highest group (T3) had significantly higher odds of failing to achieve lipid targets compared to those in the lowest group (T1). Specifically, the odds of non-attainment of LDL-C targets were 1.40 times higher in T3 compared with T1 (adjusted OR: 1.40; 95% CI 1.01–1.94; P = 0.044), and for non-HDL-C targets, the odds increased by 54% (adjusted OR: 1.54; 95% CI 1.12–2.13; P = 0.008). Similarly, participants in T3 had a 63% higher risk of failing to achieve RC targets (adjusted OR: 1.63; 95% CI 1.15–2.33; P = 0.007). For ApoB targets, the association in T3 approached statistical significance (adjusted OR: 1.48; 95% CI 0.96–2.27; P = 0.075). These findings were consistently observed across two models with different covariate adjustments, further supporting that high baseline TyG remained independently associated with the risk of lipid target non-attainment (Supplementary Table 3).

To further validate the observed associations, linear regression analyses showed that higher baseline TyG was independently associated with higher follow-up levels of LDL-C, TC, non-HDL-C, RC, and Lp(a)-corrected LDL-C, consistent with the direction observed in the target attainment analysis (Supplementary Table 4).

Association between longitudinal change in TyG and lipid profiles

To better characterize the longitudinal impact of TyG dynamics on lipid changes, we investigated both the association of percent changes in TyG with lipid profile changes and the effect of TyG status transitions over time on these lipid changes.

Multiple linear regression analysis showed that percent increases in TyG were significantly and positively associated with percent changes in several atherogenic lipid parameters. Specifically, each 1% increase in TyG was associated with greater percent increases in LDL-C (β = 1.10; 95% CI 0.67–1.53; P < 0.001), TC (β = 1.02; 95% CI 0.75–1.29; P < 0.001), non-HDL-C (β = 1.81; 95% CI 1.44–2.17; P < 0.001), RC (β = 4.92; 95% CI 3.73–6.11; P < 0.001), and Lp(a)-corrected LDL-C levels (Table 4). These associations remained robust in sensitivity analyses, where effect estimates were highly consistent after further adjustment for prior LLT, and also in models that adjusted solely for age, sex, and baseline levels of corresponding lipid and TyG index (Supplementary Table 5).

Table 4.

Linear regression between the percentage change in TyG index and the percentage change in lipid profiles

Variables	Simple linear regression			Multiple linear regression
	Beta	95% CI	P value	Beta	95% CI	P value
LDL-C percent change	1.13	0.69, 1.57	< 0.001	1.10	0.67, 1.53	< 0.001
TC percent change	1.03	0.75, 1.30	< 0.001	1.02	0.75, 1.29	< 0.001
Non-HDL-C percent change	1.81	1.44, 2.19	< 0.001	1.81	1.44, 2.17	< 0.001
ApoB percent change	0.75	− 14.09, 15.59	0.921	− 1.08	− 16.27, 14.12	0.890
RC percent change	4.89	3.71, 6.07	< 0.001	4.92	3.73, 6.11	< 0.001
Lp(a) percent change	− 0.37	− 2.67, 1.93	0.753	− 0.22	− 2.38, 1.93	0.841
LDL Lp(a)corr 30 percent change	1.71	0.80, 2.62	< 0.001	1.72	0.82, 2.63	< 0.001
LDL Lp(a)corr17.3 percent change	1.34	0.79, 1.88	< 0.001	1.32	0.78, 1.86	< 0.001

Multiple linear regression results were adjusted for age, sex, body mass index, the baseline values of the TyG index and the corresponding lipid measure, lipid-lowering therapy at discharge, hypoglycemic agents at discharge, and the history of percutaneous coronary intervention, diabetes, and hypertension. Values in bold represent results that reached statistical significance (P < 0.05). TyG, triglyceride-glucose index; LDL-C, low-density lipoprotein cholesterol; TC, total cholesterol; Non-HDL-C, non-high-density lipoprotein cholesterol; ApoB, apolipoprotein B; RC, remnant cholesterol; Lp(a), lipoprotein(a); LDL _{Lp(a)corr 30}, LDL-C corrected for Lp(a)-cholesterol content as 30% of total Lp(a) mass; LDL _{Lp(a)corr 17.3}, LDL-C corrected for Lp(a)-cholesterol content as 17.3% of total Lp(a) mass

In the TyG status transition analysis, patients whose TyG levels transitioned from the lowest at baseline to the highest tertile during follow-up showed significant increases in LDL-C (β = 0.23, 95% CI 0.03 to 0.43, P = 0.027), TC (β = 0.33, 95% CI 0.09 to 0.57, P = 0.007) and non-HDL-C (β = 0.33, 95% CI 0.09 to 0.56, P = 0.006), compared to those who consistently maintained a low TyG status. Conversely, transitioning from the highest to the lowest tertile was significantly associated with reductions in RC (β = − 0.27, 95% CI − 0.41 to − 0.13, P < 0.001) and non-HDL-C (β = − 0.42, 95% CI − 0.74 to − 0.11, P = 0.009). Additionally, a transition from the highest to the middle tertile was also negatively associated with RC (β = − 0.23, 95% CI − 0.37 to − 0.10, P < 0.001) (Table 5).

Table 5.

Change-to-change analysis of TyG index tertile transitions and changes in lipid profiles from baseline to follow-up

Characteristic		LDL-C		TC			Non-HDL-C					RC
		Beta (95% CI)	P value	Beta (95% CI)	P value		Beta (95% CI)			P value		Beta (95% CI)			P value
TyG status transition
Low → Low		–	–	–	–		–			–		–			–
Low → Middle		0.25 (− 0.04, 0.54)	0.093	0.30 (− 0.05, 0.64)	0.098		0.35 (0.01, 0.69)			0.045		0.08 (− 0.07, 0.23)			0.310
Low → High		0.23 (0.03, 0.43)	0.027	0.33 (0.09, 0.57)	0.007		0.33 (0.09, 0.56)			0.006		0.10 (− 0.01, 0.20)			0.065
Middle → Low		− 0.12 (− 0.35, 0.12)	0.341	− 0.14 (− 0.43, 0.14)	0.322		− 0.25 (− 0.52, 0.03)			0.080		− 0.14 (− 0.26, − 0.02)			0.027
Middle → Middle		0.05 (− 0.19, 0.30)	0.657	0.13 (− 0.16, 0.42)	0.368		0.11 (− 0.17, 0.39)			0.436		0.04 (− 0.08, 0.17)			0.518
Middle → High		0.06 (− 0.14, 0.26)	0.543	0.02 (− 0.23, 0.26)	0.902		0.00 (− 0.23, 0.24)			0.986		− 0.04 (− 0.14, 0.07)			0.457
High → Low		− 0.20 (− 0.47, 0.07)	0.148	− 0.29 (− 0.61, 0.04)	0.083		− 0.42 (− 0.74, − 0.11)			0.009		− 0.27 (− 0.41, − 0.13)			< 0.001
High → Middle		− 0.08 (− 0.34, 0.19)	0.574	− 0.21 (− 0.52, 0.10)	0.190		− 0.29 (− 0.59, 0.02)			0.067		− 0.23 (− 0.37, − 0.10)			< 0.001
High → High		0.00 (− 0.20, 0.20)	0.982	− 0.16 (− 0.39, 0.08)	0.195		− 0.20 (− 0.43, 0.03)			0.096		− 0.21 (− 0.31, − 0.11)			< 0.001

Multivariable regression result adjusted for age, sex, body mass index, lipid-lowering therapy at discharge, hypoglycemic agents at discharge, the baseline value of the corresponding lipid measure, and the history of percutaneous coronary intervention, diabetes, and hypertension. Values in bold represent results that reached statistical significance (P < 0.05). LDL-C, low-density lipoprotein cholesterol; TC, total cholesterol; Non-HDL-C, non-high-density lipoprotein cholesterol; ApoB, apolipoprotein B; RC, remnant cholesterol; CI, confidence interval

Impact of LLT intensity on TyG index and lipid profiles

Given the well-recognized lipid-lowering effects of LLT on lipid profiles and the observed changes in TyG, we further investigated whether LLT intensity also influenced TyG dynamics. As shown in Fig. 1, high-intensity LLT led to greater percentage reductions across multiple lipid parameters, including LDL-C, TC, non-HDL-C, ApoB, RC, and Lp(a)-corrected LDL-C, as well as a modest but significantly larger decline in TyG index compared with the low-/moderate-intensity group.

Fig. 1.

Comparison of biomarker percent changes between low-/moderate-intensity and high-intensity lipid-lowering therapy groups. Note: TyGPC, triglyceride-glucose index percentage change; LDLPC, low-density lipoprotein cholesterol percentage change; TCPC, total cholesterol percentage change; HDLPC, high-density lipoprotein cholesterol percentage change; NonHDLPC, non-high-density lipoprotein cholesterol percentage change; ApoBPC, apolipoprotein B percentage change; RCPC, remnant cholesterol percentage change; LDL30PC, LDL-C corrected for 30% Lp(a)-cholesterol content percentage change; LDL17.3PC, LDL-C corrected for 17.3% Lp(a)-cholesterol content percentage change; Lp(a)PC, lipoprotein(a) percentage change

At baseline, TyG levels were comparable between low-/moderate- and high-intensity LLT groups, with a median of 8.82 and 8.96, respectively (P = 0.053). The distribution across TyG tertiles was relatively balanced (P = 0.118), although a higher proportion of patients in the low-/moderate-intensity group were in tertile 1 (39.34% vs. 32.32%). Following LLT, both groups showed reductions in TyG, with median follow-up levels of 8.74 and 8.71, respectively (P = 0.514). Notably, the high-intensity group exhibited a significantly greater median percentage reduction in TyG compared with the low-/moderate-intensity group (− 2.40% vs. −1.04%; P = 0.030) (Table 6).

Table 6.

Baseline and follow-up TyG index levels, distribution, and percent changes stratified by lipid-lowering therapy intensity

Variables	Total (N = 1393)	Low-/moderate-intensity LLT (N = 211)	High-intensity LLT (N = 1182)	P value
TyG index
Baseline	8.95 (8.53, 9.44)	8.82 (8.48, 9.40)	8.96 (8.55, 9.44)	0.053
Follow-up	8.71 (8.35, 9.10)	8.74 (8.37, 9.13)	8.71 (8.35, 9.09)	0.514
Median change from baseline, %	− 2.17 (− 7.57, 1.33)	− 1.04 (− 5.80, 2.85)	− 2.40 (− 7.84, 1.15)	0.030
Baseline TyG group				0.118
Tertile 1	465 (33.38%)	83 (39.34%)	382 (32.32%)
Tertile 2	464 (33.31%)	67 (31.75%)	397 (33.59%)
Tertile 3	464 (33.31%)	61 (28.91%)	403 (34.09%)

Values in bold represent results that reached statistical significance (P < 0.05). TyG, triglyceride–glucose index; LLT, lipid-lowering therapy

Interaction between TyG percent change and lipid profile changes across subgroups

To further clarify the relationship between TyG reduction and lipid profile improvements, we conducted subgroup analyses. Significant interactions were observed between TyG percent change and several covariates (Fig. 2, Supplementary Fig. 3). Notably, the associations with percent changes in LDL-C, TC, and non-HDL-C were modified by LLT intensity and sex (all P for interaction < 0.05). Additionally, the association between TyG and RC was more pronounced in patients with prior PCI (β = 5.15 vs. β = 3.16; P for interaction = 0.025) and those with obesity (normal weight: β = 4.50; overweight: β = 2.87; obesity: β = 4.89; P for interaction = 0.022).

Fig. 2.

Subgroup analysis of estimated associations between percent change in TyG index and percent change in lipid profiles Note: All models were adjusted for age, sex, BMI category, LLT at discharge, the baseline value of the corresponding lipid measure, history of PCI, diabetes and hypertension, with exclusion of the stratified variable as appropriate. The boxes represent point estimations. Horizontal lines represent 95% CI. TyG, triglyceride-glucose index; LDL-C, low-density lipoprotein cholesterol; TC, total cholesterol; Non-HDL-C, non-high-density lipoprotein cholesterol; RC, remnant cholesterol; LLT, lipid-lowering therapy; BMI, body mass index; PCI, percutaneous coronary intervention; CI, confidence interval.

Further subgroup analyses revealed that the associations were stronger in the low-/moderate-intensity group, with higher effect estimates of TyG observed for LDL-C (β = 1.74, P = 0.021 vs. β = 0.79, P < 0.001), TC (β = 1.32, P = 0.001 vs. β = 0.73, P < 0.001), and non-HDL-C (β = 2.07, P = 0.001 vs. β = 1.40, P < 0.001) (Fig. 2). Similar trends were observed for LDL Lp(a)corr 17.3 (β = 1.83 vs. β = 0.99) and LDL Lp(a)corr 30 (β = 2.05 vs. β = 1.37), although these associations did not reach significance in the low-/moderate-intensity LLT group (Supplementary Fig. 3). Sex-specific analyses revealed that the effect of TyG percent change on LDL-C (β = 1.86, P < 0.001 vs. β = 0.60, P = 0.010), TC (β = 1.38, P < 0.001 vs. β = 0.63, P < 0.001) and non-HDL-C (β = 2.17, P < 0.001 vs. β = 1.29, P < 0.001) was significantly stronger in females than in males.

Discussion

In this prospective cohort of 1393 CAD patients receiving LLT and followed for a median of one year, a higher baseline TyG index was significantly associated with suboptimal lipid control at follow-up, reflected by lower attainment of LDL-C, non-HDL-C, and RC targets and higher levels of atherogenic lipid profiles. Moreover, longitudinal reductions in TyG were positively correlated with improvements in these lipid parameters, as demonstrated by both percent-change and change-to-change analyses based on TyG status transitions. Notably, high-intensity LLT not only reduced TyG levels but also attenuated the adverse impact of TyG changes on lipid profile improvements. These findings extend current knowledge by emphasizing the prognostic value of both baseline TyG and its longitudinal changes in predicting lipid outcomes and, for the first time, demonstrating that LLT intensity can modify this relationship in CAD patients.

The TyG index, derived from fasting TG and FBG concentrations, has been widely recognized as a simple and reliable surrogate marker for IR, which plays a central role in the pathogenesis of ASCVD and dyslipidemia [3, 5, 10]. Numerous studies have demonstrated significant associations between higher TyG levels and adverse lipid profiles [22]. In a cohort study of acute coronary syndrome patients undergoing PCI, TyG was positively associated with TG, TC, LDL-C, and negatively associated with HDL-C [18]. Similarly, in a cross-sectional investigation of obese individuals, those in the highest TyG tertile exhibited markedly elevated levels of TCand TG, along with reduced HDL-C [44]. In another study of Korean obese adults, TyG index showed a robust independent inverse association with mean LDL particle size (β = − 0.038, P < 0.001), and effectively identified the predominance of small dense LDL particles (cut-off 8.72; AUC = 0.897), supporting TyG as a sensitive marker of adverse lipid subfractions in this high-risk group [19]. Moreover, UK Biobank data (N = 403,335) revealed that participants in the highest TyG quartile had a hazard ratio (HR) of 1.19 (95% CI 1.12–1.26) for incident CVD, with dyslipidemia mediating ~ 46% of this association [45].

In line with existing studies, our study confirmed that higher baseline TyG was significantly associated with higher levels of atherogenic lipids at both baseline and follow-up, extending prior cross-sectional evidence into a longitudinal CAD cohort under LLT. Our analysis also incorporated a broader range of lipid markers, including non‑HDL‑C, ApoB, RC, and Lp(a)‑corrected LDL‑C, which recent guidelines and studies recognize as important risk indicators and therapeutic targets [40, 46, 47]. Individuals in the highest TyG tertile (T3) consistently exhibited elevated LDL-C, TC, non-HDL-C, ApoB, RC, and Lp(a)-corrected LDL-C throughout the treatment course. Although lipid levels declined in all tertiles after LLT, the T3 group maintained the highest residual concentrations, reflecting a persistently adverse lipid profile. Furthermore, baseline TyG index showed independent associations with follow-up lipid levels, suggesting that individuals with higher TyG were more likely to exhibit a less favorable lipid response. Interestingly, we observed an inverse association between baseline TyG and baseline Lp(a), whereas no significant relationships were found with follow-up Lp(a) or its changes over time. Although Lp(a) levels are predominantly determined by genetic variation in the LPA gene, emerging evidence suggests that metabolic factors may exert modulatory effects [48]. Insulin, for instance, can suppress apo(a) synthesis in hepatocytes, providing a plausible biological link between IR (as reflected by TyG) and lower Lp(a) concentrations [49]. This finding aligns with prior studies reporting negative correlations between Lp(a) and IR markers in various populations, including Chinese adults [50] and middle-aged hypertensive patients [51], as well as with observations that lower Lp(a) is more common in dyslipidemic individuals with increased markers of IR or metabolic syndrome [52]. The absence of significant associations longitudinally likely reflects the biological stability of Lp(a) and its minimal intra-individual variability, indicating that short-term metabolic fluctuations may not translate into measurable long-term changes.

Current guidelines recommend LDL-C as the primary target, with non-HDL-C and ApoB recognized as important secondary targets in cardiovascular risk management. In our CAD cohort of very-high-risk patients, targets were set at LDL-C < 1.4 mmol/L, non-HDL-C < 2.2 mmol/L, and ApoB < 65 mg/dL, consistent with expert recommendations [11, 40, 53, 54]. However, real-world data showed that many patients, even under high-intensity LLT, failed to achieve LDL-C and non-HDL-C targets, exposing residual cardiovascular risk. For example, the international DA VINCI observational study found that only about one-third of patients achieved 2019 guideline-based LDL-C goals, and specifically just 18% of very-high-risk patients met those targets [55]. Even when a stepwise LLT optimization was modeled in ASCVD patients, approximately 58% remained above LDL-C targets after ezetimibe addition [56]. Our previous prospective study also found over 40% of CVD patients failed to meet non-HDL-C (< 2.2 mmol/L) and RC (< 0.5 mmol/L) goals despite high-intensity LLT, reflecting persistent residual dyslipidemia.

However, the underlying factors contributing to this poor goal attainment remain insufficiently explored. To address this, the present study assessed whether baseline TyG index was associated with the likelihood of failing to achieve recommended lipid targets, and found a significant relationship between higher TyG and poorer goal attainment. As a continuous variable, baseline TyG increased the risk of non-HDL-C and RC non-attainment by approximately 23% and 38%, respectively. Compared to the lowest TyG tertile (T1), participants in the highest TyG tertile (T3) were 40–63% more likely to miss LDL-C, non-HDL-C, and RC targets (Table 3). These findings suggested that TyG might serve as a practical marker for identifying patients prone to persistent dyslipidemia and residual risk despite standardized LLT, supporting its potential role in guiding more intensive or individualized lipid-lowering strategies.

While the association between a single TyG index and CVD is well-established, the clinical importance of its longitudinal fluctuations is an area of growing interest [57–59]. Moving beyond static measurements, recent research has begun to explore the prognostic significance of changes in the TyG index in cardiovascular health. In a large prospective cohort study of over 62,000 Chinese adults without CVD at baseline, individuals in the highest quartile of TyG index change had a significantly increased risk of developing CVD (HR: 1.37; 95% CI 1.21–1.54), stroke (HR: 1.38; 95% CI 1.19–1.60), and myocardial infarction (HR: 1.36; 95% CI 1.05–1.76) compared to those in the lowest quartile, over a median follow-up of 7.01 years. Critically, substantial changes in the TyG index independently predict CVD risk, and incorporating this change into baseline risk models significantly enhanced their predictive accuracy (all P < 0.01) [25]. Similarly, in a prospective cohort of 20,185 Chinese older adults followed for a mean of 4.25 years, Wang et al. identified five distinct TyG index trajectories, with the medium stable and high gradual increase groups showing approximately 17% and 25% higher risks of CVD onset, respectively, compared with the low gradual increase group, highlighting the prognostic value of longitudinal TyG patterns in this population [59]. In a large Korean cohort of over 230,000 adults, individuals with increasing TyG index trajectories over four years had significantly higher risks of all-cause mortality (HR: 1.09; 95% CI 1.03–1.15) and CVD mortality (HR: 1.23; 95% CI 1.01–1.50) during an 8.13-year follow-up, compared to those with stable TyG levels. Notably, the association was particularly pronounced in subgroups such as younger adults (< 50 years), men, and individuals with obesity, diabetes, or hypertension [26]. This concept has been extended to related metrics, such as the TyG-body mass index (TyG-BMI). For example, Huo et al. investigated the relationship between changes in TyG-BMI and stroke incidence using data from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative prospective cohort. Compared to those with consistently low TyG-BMI, individuals with a high TyG-BMI and a slow rising trend had a 62% higher risk of stroke (OR: 1.62), while those with the highest TyG-BMI and a slow declining trend had a 71% higher risk (OR: 1.71) [28]. Similarly, increasing TyG-BMI has also been linked to a higher risk of hypertension. Zhang et al. examined the association between changes in TyG-BMI and the incidence of hypertension using data from the CHARLS. Participants were categorized into four groups based on the dynamic patterns of TyG-BMI. A clear gradient in hypertension risk was observed: compared to those with persistently low TyG-BMI, individuals with moderate (OR: 1.60), higher (OR: 1.93), and the highest TyG-BMI levels (OR: 2.33) had progressively greater odds of developing hypertension [27].

Building upon previous findings, our study provided new evidence that temporal changes in the TyG index are significantly associated with alterations in lipid metabolism. Specifically, each 1% increase in TyG was associated with proportional increases in several atherogenic lipid parameters, such as LDL-C, TC, non-HDL-C, RC, and Lp(a)-corrected LDL-C. Notably, participants whose TyG levels shifted from the lowest to the highest tertile over time exhibited significant elevations in LDL-C (β = 0.23), TC (β = 0.33), and non-HDL-C (β = 0.33). These findings suggested that worsening IR, indicated by a deterioration in TyG status, was associated with a broad spectrum of lipid abnormalities, including impacting LDL-C—the principal therapeutic target in CVD prevention. The concomitant rises in TC and non-HDL-C further underscored the atherogenic lipid shift accompanying TyG aggravation. Our clinical observation was consistent with mechanistic studies demonstrating that IR modulates lipid metabolism [20]. Impaired insulin action promoted hepatic de novo lipogenesis and TG production, while dysregulated adipose lipolysis increased FFA flux to the liver, collectively contributing to elevated LDL-C and atherogenic dyslipidemia [60]. Conversely, we found that individuals whose TyG index improved from the highest to the lowest tertile showed significant reductions in RC (β = − 0.27) and non-HDL-C (β = − 0.42). This inverse relationship highlighted the potential of TyG amelioration to induce comprehensive benefits in lipid metabolism, particularly in residual risk factors beyond LDL-C. Mechanistically, enhanced insulin sensitivity might suppress hepatic lipogenesis, reduce VLDL-C production, and attenuate adipose-derived FFA influx to the liver, thereby favorably altering lipid profiles [61].

Clinically, these findings underscored the value of TyG as both a surrogate marker of metabolic health and a modifiable indicator for lipid risk stratification and individualized therapeutic interventions. Therefore, longitudinal monitoring and targeted modulation of the TyG index might offer a practical strategy to reduce traditional and residual lipid-related cardiovascular risks.

Expanding on this clinical implication, our study provided evidence that high-intensity LLT not only led to greater reductions in traditional lipid parameters, but also resulted in a modest yet statistically significant decrease in the TyG index compared to low-/moderate-intensity LLT (− 2.40% vs. − 1.04%). We hypothesized that more aggressive lipid-lowering strategies might have a favorable impact on IR, as reflected by the TyG index. These observations aligned with previous researches. For instance, a study on familial hypercholesterolemia patients found that combination therapies involving statins with ezetimibe or PCSK9 inhibitors led to slightly greater reductions in the TyG index compared to statin monotherapy, although the differences were not statistically significant after adjusting for baseline levels [62]. Additionally, a prospective cohort study based on the U.S. National Health and Nutrition Examination Survey (NHANES) revealed that the use of lipid-lowering or antidiabetic medications modified the association between the TyG index and cardiovascular mortality. Specifically, a U-shaped relationship was observed in medication users, whereas a linear positive association was found in non-users [23]. These findings suggested that pharmacological interventions may influence the relationship between the TyG index and cardiovascular risk.

Furthermore, we found that the relationship between percent change in TyG index and concurrent changes in several lipid parameters was significantly modified by several patient characteristics, including the intensity of LLT, sex, obesity status, and history of PCI.

Subgroup analyses revealed that the association between percent changes in the TyG index and alterations in lipid parameters, specifically LDL-C, TC, and non-HDL-C, was more pronounced in patients receiving low-/moderate-intensity LLT and in female participants. This suggested that in patients undergoing less intensive lipid-lowering strategies (primarily low-/moderate-intensity statin monotherapy), fluctuations in IR, as reflected by the TyG index, had a more substantial impact on lipid metabolism.​ High-intensity LLT (combination of statins and ezetimibe/ PCSK9 inhibitors/ fibrates) might attenuate the influence of IR on lipid metabolism by more effectively and directly reducing lipid parameters. Our previous studies demonstrated that high-intensity LLT independently improves follow-up levels and goal attainment rates of LDL-C, non-HDL-C, and RC [31]. While statins were potent in lowering atherogenic lipids and reducing cardiovascular risk, evidence indicated that statin monotherapy, especially at high-intensity doses, would impair pancreatic β-cell function and exacerbate IR, increasing the risk of new-onset diabetes [63]. In contrast, combination LLT (e.g., statin plus ezetimibe) modulated lipid metabolism via complementary mechanisms and appeared to mitigate IR-related metabolic disturbances [64]. Therefore, this combination strategy might offset the adverse glycemic effects of statin monotherapy, offering a more favorable metabolic profile.

Additionally, the stronger association between TyG index reductions and lipid profile improvements in females might be attributed to sex-specific differences in IR, lipid metabolism, and fat distribution [65]. Estrogen modulated lipid metabolism by promoting subcutaneous fat storage and enhancing insulin sensitivity, potentially augmenting the lipid-lowering effects of TyG index reduction in women. Lu et al. also indicated that the TyG index is more strongly associated with subclinical atherosclerosis in non-diabetic women compared to men, suggesting a sex-specific metabolic response to IR [66].​ In individuals with obesity or prior PCI, TyG percent change correlated more strongly with reductions in RC, consistent with NHANES findings showing that the TyG index and its combination with obesity indicators were significantly associated with CVD risk [67]. Collectively, these subgroup-specific associations underscored the heterogeneous metabolic effects of TyG change, and supported a more personalized approach to lipid management and cardiometabolic risk control.

Study limitations and strengths

This study has several limitations. First, as a real-world observational study, causality cannot be inferred. Although we adjusted for confounders and stratified by key variables, unmeasured factors such as diet, physical activity, genetics, or detailed data on medication changes during follow-up may have influenced both IR and lipid outcomes. In particular, most patients (~ 90%) were already receiving LLT prior to study enrollment, and a large statin-naive cohort was not available. To address this, we defined baseline lipid levels at the time of hospitalization, when LLT regimens were reassessed and adjusted, and calculated lipid changes from that point onward. Sensitivity analyses further adjusting for prior LLT confirmed the robustness of our findings, supporting the independence of the observed associations. Second, the relatively short follow-up period (median 1 year) and the limited number of clinical events recorded, resulting in an absence of hard cardiovascular endpoints such as major adverse cardiovascular events, cardiovascular death, or myocardial infarction, restrict the evaluation of long-term outcomes and the prognostic value of TyG dynamics. Third, the single center design may limit the generalizability of the findings to broader CAD populations, particularly given potential ethnic and regional differences in lifestyle and metabolic characteristics that could also affect IR and lipid profiles. Lastly, although TyG is a convenient surrogate of IR, we did not directly measure insulin levels or use reference methods such as the clamp test. Moreover, as TyG was not a predefined therapeutic target and no interventions aimed at modifying IR were implemented, neither causal inference nor its modifiability can be reliably assessed.

Despite these limitations, the study has important strengths. To our knowledge, it is the first to assess how dynamic changes in IR (reflected by TyG) relate to lipid outcomes under varying LLT intensities. The prospective cohort design, featuring a relatively large sample size and serial biochemical measurements, strengthens the study’s validity through multiple analytical approaches. These include stratification by LLT intensity and clinical characteristics, along with multidimensional analyses of TyG index dynamics (baseline levels, temporal changes, and transition status). Using change-to-change analysis, we provided additional insights into how shifts in TyG status over time affect lipid changes. Furthermore, the study explored associations between TyG and a comprehensive panel of atherogenic lipid parameters, including LDL-C, non-HDL-C, RC, Lp(a), and the novel LDLLp(a)corr, with appropriate adjustment for potential confounders. By including multiple lipid targets, we offer a more comprehensive view of lipid control in relation to IR.

Conclusion

In this study, we identified a longitudinal association between the TyG index and atherogenic lipid profiles in patients with CAD. Notably, both elevated baseline TyG levels and upward transitions over time were associated with more adverse lipid profiles. Importantly, intensified LLT appeared to attenuate the adverse impact of TyG elevation on lipid profiles, suggesting that improving TyG through more aggressive treatment may contribute to better lipid control. Maintaining lower TyG levels may help stabilize lipid profiles and improve cardiovascular risk management. Collectively, these results highlight the importance of incorporating TyG and its temporal changes into routine cardiovascular risk assessment to guide personalized lipid management and therapeutic decisions.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CVD

Cardiovascular disease

CAD

Coronary artery disease

IR

Insulin resistance

TyG

Triglyceride-glucose index

TG

Triglyceride

FBG

Fasting blood glucose

LDL-C

Low-density lipoprotein cholesterol

ApoB

Apolipoprotein B

RC

Remnant cholesterol

Lp(a)

Lipoprotein(a)

LDLLp(a)corr

LDL‑C corrected for the cholesterol content of Lp(a)

HDL-C

High-density lipoprotein cholesterol

TC

Total cholesterol

VLDL-C

Very-low-density lipoprotein cholesterol

FFA

Free fatty acid

LLT

Lipid-lowering therapy

CVD

Cardiovascular disease

SBP

Systolic blood pressure

DBP

Diastolic blood pressure

BMI

Body mass index

Non-HDL-C

Non-high-density lipoprotein cholesterol

DM

Diabetes mellitus

SGLT-2

Sodium-glucose cotransporter‑2

GLP-1

Glucagonlike peptide-1

PCI

Percutaneous coronary intervention

PCSK9

Proprotein Convertase Subtilisin/Kexin Type 9

ASCVD

Atherosclerotic cardiovascular disease

SD

Standard deviation

OR

Odds ratio

CI

Confidence interval

HR

Hazard ratio

NHANES

U.S. National Health and Nutrition Examination Survey

CHARLS

China Health and Retirement Longitudinal Study

TyG-BMI

Triglyceride-glucose-body mass index

Author contributions

Na-Qiong Wu and Zhi-Fan Li designed the study. Zhi-Fan Li, Zheng Yin, Xi Li, and Meng-Ying Lu performed the study and collected the data. Zhi-Fan Li, Na-Qiong Wu and Hong Qiu analyzed the data and wrote the manuscript. Zhi-Fan Li and Zheng Yin prepared the figures. All the other authors (Wen-Jia Zhang, Fang Luo, Yan-Lu Xu) edited the manuscript and provided comments. Ke-Fei Dou, Xiao Wang, Jian-Jun Li and Hong Qiu provided consultations regarding the study design and statistical analysis.

Funding

This work is funded by the Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences (CIFMS) (2021-I2M-1-008).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This work was approved by the Fuwai Hospital Ethics Review Committee and strictly adhered to the Declaration of Helsinki.

Consent for publication

All participants provided written informed consent for publication.

Competing interests

The authors declare no competing interests.