Introduction

Hypertension is a major risk factor for cardiovascular diseases (CVD), leading to over 10 million deaths worldwide each year1,2. The number of cardiovascular deaths caused by hypertension exceeds that of any other modifiable risk factor, second only to smoking, making it a leading cause of preventable death3,4,5. According to the World Health Organization, 1.4 billion people suffer from hypertension worldwide, and there are approximately 274 million people in China2,6. Despite the high prevalence of hypertension, treatment and control rates remain low, especially in rural areas7,8,9. The World Health Organization’s first report on the “devastating” impact of hypertension globally indicates that approximately four-fifths of individuals with hypertension are not adequately treated. If treatment coverage were to improve, it could prevent 76 million deaths by 205010. Therefore, the identification and management of the risk factors for hypertension are critical for reducing health risks associated with hypertension, especially in terms of reducing the risk of cardiovascular disease11,12,13,14.

Insulin resistance (IR) is an important risk factor for hypertension15. The triglyceride-glucose (TyG) index is calculated using fasting triglyceride (TG) and fasting blood glucose (FBG), offers an improved and cost-effective marker for identifying IR16[,17. Recent studies have demonstrated that the TyG index is closely associated with hypertension and CVD18,19,20. Given the dynamic nature of individual TyG levels, assessing only a single timepoint TyG level has limitations. Utilizing group-based trajectory modeling (GBTM) enables the tracking of individual trends and recognition of diverse change patterns, facilitating the early identification of high-risk populations for hypertension, which can advance CVD management. However, previous studies have not investigated the association between TyG index trajectories and the risk of hypertension in rural populations.

Consequently, we utilized the data from the Xinjiang Multi-Ethnic Cohort Study, applied group-based trajectory modeling, and explored the relationship between the identified TyG index trajectory and the risk of hypertension in the rural population.

Methods

Study design and participants

Participants were recruited from the Xinjiang Multi-Ethnic Cohort (XMC), a longitudinal, population-based study. Our study is based on two rural areas, Hotan and Yili. Details of the XMC have been presented in a previous paper21.

Our study consisted of two parts: first, to identify longitudinal trajectories of TyG levels from 2017 to 2019 survey in the population; second, to follow up on the incidence of hypertension in each TyG trajectory group until December 2022 (Fig. 1A). In 2017, 20,932 participants received a comprehensive health examination. We excluded participants with hypertension during the trajectory modeling period (n = 8571), incomplete baseline characteristics data (n = 130), Incomplete lifestyle data (n = 820), the records of TyG index less than 3 times (n = 1605), or without follow-up information (n = 463). Ultimately, 9343 participants were included in the present study (Fig. 1B). Approval for this study protocol was granted by the Institutional Review Board of the Traditional Chinese Medical Hospital of Xinjiang Uygur Autonomous Region (2018XE0108). Before enrolling, all participants signed an informed consent document. This study was performed according to relevant guidelines and regulations. Moreover, the research was carried out in strict compliance with the principles stipulated in the Declaration of Helsinki.

Fig. 1
figure 1

A) Study design; B) Flow chart for the selection of study participants.

Full size image

Data collection and definitions

Demographic characteristics (age, sex, ethnicity, marital status, education, and annual income), medical history, family history, medication history, and lifestyle factors (smoking, drinking, physical activity, and diet) were collected by trained clinic staff through a standardized questionnaire. The ethnicity was classified as Han, Hui, Uyghur, Kazakh, and others (Mongolian, and Tibetan). Education status was categorized as primary school or below, middle school, and high school or further education. Marital status was categorized as married, and others (unmarried, widowed, and divorced). annual income was classified as low (< 10,000 RMB), moderate (10,000 RMB–50,000 RMB), high (≥ 50,000 RMB). A smoker was defined as smoking continuously or cumulatively for more than 1 year and at least smoking 1 cigarette one day22. A drinker was defined as drinking alcohol once per week for at least six months23. The physical activity level of the study subjects was assessed using the physical activity questionnaire from the China Kadoorie Biobank (CKB) project. The questionnaire includes the frequency and duration of physical activities related to occupation, commuting, household, and leisure. In this study, work-related physical activities were not differentiated between agricultural and non-agricultural workers; they were classified according to the intensity of their work. The metabolic equivalents (MET) for various physical activities were determined based on the updated Physical Activity Guidelines from 2011 to quantify the intensity of physical activities (Appendix: Table S1)24,25. The MET value for each physical activity was multiplied by its frequency and duration, and then summed to calculate the individual’s overall physical activity level per day (MET-h/d).

Trained physicians or nurses performed anthropometric measurements following standardized protocols. Weight and height were measured using an auto- anthropometer (SK-X80, China) with an accuracy of 0.1 kg and 0.1 cm, respectively. When participants were seated after a 10-min rest, blood pressure was assessed two times with a calibrated sphygmomanometer, and the mean average of the two assessments was applied for subsequent analysis. After an overnight fast, nurses collected participants’ venous blood samples for laboratory tests. Triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), FBG, and creatinine (Cr) levels were performed in the laboratories of the community health centers with automated analyzers (Hitachi, Tokyo, Japan). The TyG index at each visit was calculated as Ln [TG (mg/dL) × FPG (mg/dL)/226. Diabetes was defined as the FBG ≥ 7 mmol/L, taking glucose-lowering agents or previously diagnosed type 2 diabetes by physicians. Hypertension was defined as SBP ≥ 140 mmHg, DBP ≥ 90 mmHg, or taking antihypertensive drugs according to clinical data or self-report27.

Healthy lifestyle score (HLS)

Based on previous studies and the cardiovascular health promotion goals proposed by the American Heart Association, five modifiable lifestyle factors were included in this study: smoking, drinking, physical activity, obesity and diet28,29,30,31. BMI < 28 kg/m² is defined as low-risk; Non-smoking is defined as low-risk; The low-risk group for alcohol consumption includes non-drinkers and those who consume alcohol in light to moderate amounts daily (men with daily alcohol intake < 30 g and women with daily alcohol intake < 15 g) 31; The low-risk group for physical activity includes individuals whose physical activity levels are in the top 25% of their same-sex peer group; For diet, we used a healthy diet score created by the CKB project for calculation (eating fresh vegetables daily, eating fresh fruits daily, eating red meat 1–6 days per week, eating legumes ≥ 4 days per week, and eating fish ≥ 1 day per week). For each criterion met, one point was scored; otherwise, 0. Thus, the diet score ranged from 0 to 5, with a score of 4 to 5 defined as the low-risk group31. Based on the number of low-risk lifestyle factors, a healthy lifestyle score was derived, with a score range of 0 to 5 points, where a higher score indicates a healthier lifestyle for the individual.

Follow‑up and outcomes assessment

Participants were followed up annually through health examinations and on-site follow-up. In addition, the Medical Record Information System, Medical Insurance System, and Chronic Disease Management System were used to confirm health outcomes further. The primary outcome of our study was the time from the baseline to the first occurrence of hypertension, and follow-up was completed in December 2022.

Statistical analysis

Continuous variables with normal distributions were described as mean ± standard deviation (SD), while skewed distributions were presented using medians with interquartile ranges. Differences between multiple groups were analyzed by ANOVA analysis or Kruskal-Wallis rank test according to their distributions. Categorical variables were reported as proportions and compared using the Chi-squared test.

We employed the group-based trajectory modeling (GBTM), using the PROC TRAJ procedure in SAS software, to identify different longitudinal TyG index trajectories during 2017–2019. The optimal number (ranging from 2 to 5) and shape of TyG index trajectories (linear, quadratic, or cubic) were determined according to the following criteria: (1) the lowest Bayesian information criterion (BIC); (2) no less than 5% of the participants within each trajectory group; and (3) higher average posterior probabilities for each trajectory group (> 0.70)32.

The Cox proportional hazards model was applied to assess the associations between different trajectories of the TyG index and the risk of hypertension. Model 1 was unadjusted. Model 2 was adjusted for sex, age, ethnicity, education, marital status, and annual income. Model 3 further adjusted for smoking status, drinking status, physical activity (MET-h/d), diet score, BMI, LDL-C, HDL-C, total cholesterol, and glomerular filtration rate (eGFR). Additionally, the Kaplan–Meier method was used to calculate the cumulative event rate of hypertension for the different TyG trajectory groups, and the log–rank test was used to compare the curves of the cumulative event rates of hypertension by groups. To test the robustness of our findings, we performed several sensitivity analyses. First, we excluded participants with diabetes at baseline. Second, we excluded participants taking hypoglycemic agents or lipid-lowering medications. Additionally, we excluded participants with a family history of hypertension. Finally, we used the random forest method for multiple imputation of missing values in baseline covariates (Appendix: Figure S1). We performed subgroup analyses by sex, age, BMI, and healthy lifestyle score, to elucidate the potential relationship between TyG index trajectories and the risk of hypertension. We further used the restricted cubic spline regression model, with 3 knots, to assess the dose-response relationship between the TyG index (the level at the end of the trajectory modeling period in 2019) and new-onset hypertension. All statistical analyses were conducted with SAS version 9.4 and R version 4.4.2. P value < 0.05 was considered statistical significance.

Results

Baseline characteristics

In total, 9343 participants were available for analysis, 3868 (41.40%) were male, and the mean age at baseline was 48.38 ± 9.24 years. At a median follow-up period of 35 months, 1687 participants experienced hypertension, the incidence of hypertension was 18.06%. We identified three distinct trajectories of the TyG index. Trajectories were labeled as low stable (n = 4239, 45.37%, mean TyG index range: 8.00–8.10), moderate stable (n = 4561, 48.82%, mean TyG index range: 8.64–8.70), and high stable (n = 543, 5.81%, mean TyG index range: 9.43–9.55) (Fig. 2). The range of TyG index at baseline was 6.19–11.09, and this baseline data corresponds to the TyG level at the end of the trajectory modeling period in 2019. The average posterior probabilities for each trajectory group are 0.78, 0.75, and 0.81, respectively (Appendix: Table S2). The baseline characteristics of each trajectory were shown in Table 1. Compared with other groups, participants in the high stable group were more likely to be older, male, smokers, and drinkers; to have diabetes mellitus; to take hypoglycemic or lipid-lowering medications; and to have higher levels of BMI, FBG, TG, TC, LDL-C, systolic blood pressure (SBP), and diastolic blood pressure (DBP); They were also more likely to have lower levels of physical activity, HDL-C, and eGFR.

Fig. 2
figure 2

The trajectory of TyG index derived from group-based trajectory modeling.

Full size image
Table 1 Baseline characteristics of the different TyG index trajectory groups of the participants.
Full size table

Association between TyG index trajectory and risk of hypertension

The cumulative incidence of hypertension increases with the elevation of TyG trajectory levels. At a median follow-up period of 35 months, the cumulative event rate of hypertension in the low stable group, moderate stable group, and high stable group was 12.90%, 21.25%, and 31.49%, respectively. Elevated TyG trajectory levels are associated with an increased risk of hypertension (log-rank P < 0.001; Fig. 3). The results of the hazard proportionality test are shown in Figure S2, showing no evidence of non- proportionality.

Fig. 3
figure 3

Kaplan-Meier estimation of cumulative incidence curves of hypertension in different TyG index trajectories.

Full size image

The Cox regression models were applied to investigate the association of TyG index trajectory with the risk of hypertension (Table 2). In the fully adjusted model, compared with the low stable group, the risk of hypertension was 1.41 (95% CI, 1.26–1.58), and 1.82 (95% CI, 1.50–2.21) for the participants in the moderate stable and high stable group, respectively.

Table 2 Association between TyG index trajectory and the risk of hypertension.
Full size table

Sensitivity analysis

Sensitivity analyses all generated similar findings to those of the primary analysis (Table 3). Excluding participants with diabetes at baseline, the associations between TyG index trajectories and hypertension remained unchanged. Similarly, after excluding participants taking hypoglycemic agents or lipid-lowering medications, the TyG index trajectories were still significantly associated with the incidence of hypertension. After excluding participants who used hypoglycemic or lipid-lowering medications, we analyzed the associations between different TyG index trajectory groups and the risks of three outcomes separately: hypertension only, diabetes only, and hypertension-diabetes comorbidity. The results showed that compared with the low stable group, the risk of hypertension only remained increased in the moderate stable group and high stable group (Table S3). Moreover, after excluding participants with a family history of hypertension, we found that the association between hypertension incidence and TyG index trajectories remained significant. Additionally, imputing the missing data using multiple imputations yielded similar results.

Table 3 Sensitivity analysis of the association between TyG index trajectory and the risk of hypertension.
Full size table

Subgroup analysis

In subgroup analyses, the association between TyG index trajectory and the risk of hypertension remained consistent across sex (male or female), age (< 50 years or ≥ 50 years), BMI (< 25 kg/m² or ≥ 25 kg/m²), and healthy lifestyle score (HLS) (0–3 or 4–5). Notably, we found a significant interaction between the TyG index trajectory and HLS on hypertension (P for interaction = 0.033). The association between a high TyG longitudinal trajectory and incident hypertension was stronger in the high HLS group compared to the low HLS group (Fig. 4).

Fig. 4
figure 4

Association between TyG index trajectory and the risk of hypertension in different subgroups. Each subgroup was adjusted for sex, age, ethnicity, education, marital status, annual income, smoking status, drinking status, physical activity (MET-h/d), diet score, BMI, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, total cholesterol, and glomerular filtration rate, except for the stratification variables. HLS, healthy lifestyle score.

Full size image

Dose-response relationship of TyG index and risk of hypertension

To further explore the relationship between the TyG index and risk of hypertension, Cox model with restricted cubic splines (with three knots) was performed, demonstrating a positive linear dose-response relationship between TyG index and the log hazard of hypertension (Pnon-linearity = 0.400; Appendix: Figure S3).

Discussion

In this multiethnic cohort study, we investigated the relationship between longitudinal trajectories of the TyG index and hypertension. Our results indicated that compared with the low stable group, elevated TyG index trajectories were associated with an increased risk of hypertension. Further subgroup analysis revealed that these associations remained significant in different subgroups, including sex, age, BMI, and HLS.

Limited research has explored the relationship between longitudinal trajectories of the TyG index and hypertension risk over time. A study conducted from 2011 to 2022 involving 15,056 participants indicated that those in the moderate increasing group and the high stable group of the TyG index trajectories were associated with the risk of hypertension22. However, this study was a retrospective study, and the subjects were recruited from the hospital’s physical examination center. Our study adds to the existing evidence by confirming the association between high TyG index longitudinal trajectories and hypertension in rural population. Although a limited number of studies have investigated the longitudinal trajectories of the TyG index and the risk of hypertension in large prospective cohort, several studies have demonstrated a close association between TyG index trajectories and cardiovascular events. A large-scale prospective cohort study demonstrated that participants with medium stable or high gradual increase trajectories of TyG index were associated with an elevated risk of developing CVDs in elderly populations33. The CARDIA study confirmed the association between TyG index trajectories and cardiovascular disease in young adults34. Moreover, another study demonstrated that distinct TyG index trajectories were significantly associated with subsequent risk of CVD in normal-weight individuals35. Meanwhile, the restrictive cubic spline model showed a positive dose-response relationship between the TyG index and the risk of hypertension. This is consistent with the findings of the previous studies18,36,37. However, some studies indicate that the TyG index has a non-linear dose-response relationship with the risk of hypertension22, and this difference may be attributed to variations in study populations and sample sizes.

In our study, the healthy lifestyle score includes five factors: smoking, drinking, physical activity, obesity and diet. We found that the association between the high TyG longitudinal trajectory and the risk of hypertension appears to be stronger in the group with a high healthy lifestyle score. Individuals with a low healthy lifestyle score, other risk factors, such as abnormal lipid concentrations38,39,40,41,42,43,44, may obscure the effect of the TyG index on hypertension risk. In summary, among individuals with a high healthy lifestyle score, exposure to a high TyG longitudinal trajectory is associated with a higher risk of hypertension. This study highlights the importance of paying attention to TyG levels, even for people with a healthy lifestyle.

Although the three TyG index trajectory groups in our study maintained a stable trend, this result still exhibits practical significance. The stable trajectory reflects the basic characteristics and trends of the TyG index, indicating that TyG levels in this study population remain relatively stable, especially in the high stable group, where fasting blood glucose and/or fasting triglyceride levels consistently remain elevated. This suggests the need for enhanced health management for this population.

The mechanisms underlying the association between the TyG index and hypertension remain unclear. IR is a central pathophysiological mechanism of hypertension, and induces hypertension through various pathways, such as oxidative stress, systemic inflammation, endothelial dysfunction, formation of advanced glycation end products, decreased nitro oxide synthesis, increased renal sodium reabsorption, as well as the inappropriate activation of the sympathetic nervous system and the renin–angiotensin–aldosterone system45,46,47,48,49. These mechanisms lead to increased peripheral vascular resistance and preload, ultimately resulting in the development of hypertension.

Some strengths in this study should be acknowledged. First, our study confirms that TyG index trajectories remain stable over time in large populations. Second, our study makes full use of longitudinal data from health examinations in rural areas to better integrate health policies into the residents’ health management. Third, based on a multiethnic prospective cohort study, the association between TyG longitudinal trajectories and hypertension was indicated in a rural population. Finally, we studied the association between TyG longitudinal trajectories and hypertension under different lifestyle subgroups. However, the current study also had several limitations. First, although potential risk factors were adjusted for in the model, there may still be some residual or unassessed confounding variables. Second, the absence of insulin level measurements limits direct validation of TyG as a marker of insulin resistance. Third, relevant exposure factors and confounding factors may change during the follow-up process. Fourth, data of this study were from in a rural population in Xinjiang, China, limiting the generalizability of its results to other populations.

Conclusions

In rural residents, elevated TyG index trajectories were associated with the risk of hypertension. The TyG index is valuable for assessing insulin resistance among residents with hypertension in remote rural areas. In public health practice, we should focus on the longitudinal changes in the TyG index, even among individuals with a healthy lifestyle. Large studies with long follow-up time are needed to confirm this observation.