Lancet Healthy Longev 2023;
4: e23–33 Published Online December 12, 2022 https://doi.org/10.1016/
S2666-7568(22)00247-1 See Comment page e2 Masira Research Institute, Universidad de Santander, Bucaramanga, Colombia (Prof P Lopez-Jaramillo PhD, D Gomez-Arbelaez PhD, D Martinez-Bello PhD); Division of Adult Medicine, Department of Medicine, Philippine General Hospital, Manila, Philippines (M E M Abat MD); Department of Cardiac Sciences, King Fahad Cardiac Center, College of Medicine, King Saud Medical City, King Saud University, Riyadh, Saudi Arabia (Prof K F Alhabib MBBS);
International Research Center, Hospital Alemão Oswaldo Cruz, São Paulo, Brazil
(Prof Á Avezum PhD); Federal State Budgetary Institution Research Institute for Complex Issues of Cardiovascular Diseases, Kemerovo, Russia (O Barbarash PhD); Physiology Unit, Department of Biomedical Sciences, University of Zimbabwe, Harare, Zimbabwe (J Chifamba DPhil);
Estudios Clínicos Latinoamérica and Instituto Cardiovascular de Rosario, Rosario,
Argentina (M L Diaz MD);
Cardiology Department, Ankara University School of Medicine, Ankara, Türkiye (Prof S Gulec MD); Department of Community Health, Faculty of Medicine, Universiti Kebangsaan Malaysia, Kuala Lumpur, Malaysia (N Ismail MD);
Department of Community Health Sciences, Aga Khan
Association of the triglyceride glucose index as a measure of insulin resistance with mortality and cardiovascular disease in populations from five continents (PURE study):
a prospective cohort study
Patricio Lopez-Jaramillo, Diego Gomez-Arbelaez, Daniel Martinez-Bello, Marc Evans M Abat, Khalid F Alhabib, Álvaro Avezum, Olga Barbarash, Jephat Chifamba, Maria L Diaz, Sadi Gulec, Noorhassim Ismail, Romaina Iqbal, Roya Kelishadi, Rasha Khatib, Fernando Lanas, Naomi S Levitt, Yang Li, Viswanathan Mohan, Prem K Mony, Paul Poirier, Annika Rosengren, Biju Soman, Chuangshi Wang, Yang Wang, Karen Yeates, Rita Yusuf, Afzalhussein Yusufali, Katarzyna Zatonska, Sumathy Rangarajan, Salim Yusuf
Summary
Background The triglyceride glucose (TyG) index is an easily accessible surrogate marker of insulin resistance, an important pathway in the development of type 2 diabetes and cardiovascular diseases. However, the association of the TyG index with cardiovascular diseases and mortality has mainly been investigated in Asia, with few data available from other regions of the world. We assessed the association of insulin resistance (as determined by the TyG index) with mortality and cardiovascular diseases in individuals from five continents at different levels of economic development, living in urban or rural areas. We also examined whether the associations differed according to the country’s economical development.
Methods We used the TyG index as a surrogate measure for insulin resistance. Fasting triglycerides and fasting plasma glucose were measured at the baseline visit in 141 243 individuals aged 35–70 years from 22 countries in the Prospective Urban Rural Epidemiology (PURE) study. The TyG index was calculated as Ln (fasting triglycerides [mg/dL] × fasting plasma glucose [mg/dL]/2). We calculated hazard ratios (HRs) using a multivariable Cox frailty model with random effects to test the associations between the TyG index and risk of cardiovascular diseases and mortality. The primary outcome of this analysis was the composite of mortality or major cardiovascular events (defined as death from cardiovascular causes, and non-fatal myocardial infarction, or stroke). Secondary outcomes were non-cardiovascular mortality, cardiovascular mortality, all myocardial infarctions, stroke, and incident diabetes.
We also did subgroup analyses to examine the magnitude of associations between insulin resistance (ie, the TyG index) and outcome events according to the income level of the countries.
Findings During a median follow-up of 13·2 years (IQR 11·9–14·6), we recorded 6345 composite cardiovascular diseases events, 2030 cardiovascular deaths, 3038 cases of myocardial infarction, 3291 cases of stroke, and 5191 incident cases of type 2 diabetes. After adjusting for all other variables, the risk of developing cardiovascular diseases increased across tertiles of the baseline TyG index. Compared with the lowest tertile of the TyG index, the highest tertile (tertile 3) was associated with a greater incidence of the composite outcome (HR 1·21; 95% CI 1·13–1·30), myocardial infarction (1·24; 1·12–1·38), stroke (1·16; 1·05–1·28), and incident type 2 diabetes (1·99; 1·82–2·16). No significant association of the TyG index was seen with non-cardiovascular mortality. In low-income countries (LICs) and middle-income countries (MICs), the highest tertile of the TyG index was associated with increased hazards for the composite outcome (LICs: HR 1·31; 95% CI 1·12–1·54; MICs: 1·20; 1·11–1·31; pinteraction=0·01), cardiovascular mortality (LICs: 1·44;
1·15–1·80; pinteraction=0·01), myocardial infarction (LICs: 1·29; 1·06–1·56; MICs: 1·26; 1·10–1·45; pinteraction=0·08), stroke (LICs: 1·35; 1·02–1·78; MICs: 1·17; 1·05–1·30; pinteraction=0·19), and incident diabetes (LICs: 1·64; 1·38–1·94;
MICs: 2·68; 2·40–2·99; pinteraction <0·0001). In contrast, in high-income countries, higher TyG index tertiles were only associated with an increased hazard of incident diabetes (2·95; 2·25–3·87; pinteraction <0·0001), but not of cardiovascular diseases or mortality.
Interpretation The TyG index is significantly associated with future cardiovascular mortality, myocardial infarction, stroke, and type 2 diabetes, suggesting that insulin resistance plays a promoting role in the pathogenesis of cardiovascular and metabolic diseases. Potentially, the association between the TyG index and the higher risk of cardiovascular diseases and type 2 diabetes in LICs and MICs might be explained by an increased vulnerability of these populations to the presence of insulin resistance.
Funding Full funding sources are listed at the end of the paper (see Acknowledgments).
University, Karachi, Pakistan (R Iqbal PhD); Isfahan Cardiovascular Research Center, Cardiovascular Research Institute, Isfahan University of Medical Sciences, Isfahan, Iran (Prof R Kelishadi MD); Advocate Aurora Research Institute, Advocate Aurora Health, Downers Grove, IL, USA (R Khatib PhD); Institute of Community and Public Health, Birzeit University, Birzeit, Palestine (R Khatib); Universidad de La Frontera, Temuco, Chile (Prof F Lanas PhD); Chronic Disease Initiative for Africa, University of Cape Town, Cape Town, South Africa (N S Levitt MD); Medical Research & Biometrics Center, National Clinical Research Center for Cardiovascular Diseases, Fuwai Hospital, National Center for Cardiovascular Diseases, Peking Union Medical College & Chinese Academy of Medical Sciences, Beijing, China (Y Li PhD, C Wang PhD, Y Wang MSc);
Madras Diabetes Research Foundation, and Dr Mohan’s Diabetes Specialities Centre, Chennai, India (Prof V Mohan DSc); Division of Epidemiology & Population Health, St John’s Medical College
& Research Institute, Bangalore, India (Prof P K Mony MD);
Faculté de Pharmacie, Université Laval, Institut Universitaire de Cardiologie et de Pneumologie de Québec, Québec City, QC, Canada (Prof P Poirier PhD);
Sahlgrenska Academy, University of Gothenburg and Sahlgrenska University Hospital, Gothenburg, Sweden (Prof A Rosengren PhD); Health Action by People, Medical College, and Sree Chitra Tirunal Institute for Medical Sciences and Technology, Kerala, India (B Soman MD);
Department of Medicine, Queen’s University, Kingston, ON, Canada (K Yeates MD);
Independent University, Dhaka, Bangladesh (Prof R Yusuf PhD);
Tamani Foundation, Matemwe, Zanzibar, Tanzania (A Yusufali MD); Department of Population Health, Wroclaw Medical University, Wroclaw, Poland (K Zatonska PhD);
Population Health Research Institute, McMaster University, Hamilton Health Sciences, Hamilton, ON, Canada (S Rangarajan MSc, Prof S Yusuf DPhil)
Copyright © 2022 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC-ND 4.0 license.
Research in context Evidence before this study
The prevalence of type 2 diabetes and cardiovascular diseases has increased in the past five decades, particularly in low- income countries (LICs) and middle-income countries (MICs).
The existing literature has extensively evaluated the associations between risk factors and cardiovascular diseases and mortality. However, the role of insulin resistance has been less studied, particularly in LICs and MICs. Direct measurement of insulin resistance (eg, with hyperinsulinaemic-euglycaemic clamp and the homoeostasis model assessment of insulin resistance [HOMA-IR]) is complex and too expensive to be used in large-scale epidemiological studies and clinical practice.
The triglyceride glucose (TyG) index, a simple, accessible, and reliable clinical surrogate marker of insulin resistance, has shown a good performance in the estimation of insulin resistance compared with the HOMA-IR. Additionally, the TyG index is associated with increased risk of mortality and cardiovascular diseases. However, the association of the TyG index with cardiovascular diseases and deaths has mainly been investigated in Asia, with few data available from other regions of the world.
Added value of this study
In this large prospective cohort study of 141 243 individuals from 22 countries in five continents, we investigated the association of insulin resistance (as determined by the TyG
index) with mortality, cardiovascular diseases, and type 2 diabetes. We also aimed to examine the magnitude of associations between insulin resistance and cardiovascular disease events according to the income level of the countries.
To our knowledge, this is the largest study examining the TyG index as a measure of insulin resistance and associations with mortality, cardiovascular diseases, and type 2 diabetes.
Implications of all the available evidence
The TyG index is associated with cardiovascular mortality, incident cardiovascular diseases (myocardial infarction and stroke), and incident type 2 diabetes, suggesting that insulin resistance is involved in promoting the pathogenesis of cardiovascular and metabolic diseases. The TyG index is suitable for daily clinical practice as it is easy to be measured and calculated from routinely available laboratory variables, leading to early identification of people at high risk of developing a fatal and non-fatal cardiovascular event or type 2 diabetes.
Moreover, the association between the TyG index and risk of cardiometabolic diseases varies in relation to the income level of the countries, with participants from LICs and MICs displaying a higher risk of mortality and cardiovascular diseases with a higher TyG index. Potentially, this association might be explained by an increased vulnerability of these populations from LICs and MICs to the consequences of insulin resistance.
Introduction
The prevalence of type 2 diabetes and cardiovascular diseases has increased in the past five decades, particularly in low-income countries (LICs) and middle- income countries (MICs).1 This epidemic is most likely being driven by fast epidemiological transition shift, socioeconomic inequalities, and a greater vulnerability of these populations to the pathogenic consequences of obesity compared with individuals with similar characteristics in high-income countries (HICs).1–3 In fact, previous data from the Prospective Urban Rural Epidemiology (PURE) study showed a higher incidence of mortality and cardio vascular diseases in LICs and MICs, despite lower rates of obesity in those regions.4
Insulin resistance has been widely shown to be a critical pathophysiology pathway for the development of type 2 diabetes and cardiovascular disease.5–7 In populations in LICs and MICs, a mismatch between poverty and undernutrition in early life and exposure to an obesogenic environment in adulthood is thought to lead to an increased susceptibility to insulin resistance, low-grade inflammation, and chronic diseases, such as type 2 diabetes, and cardiovascular diseases.2,8,9
The reference method for the assessment of insulin resistance is the hyperinsulinaemic-euglycaemic clamp
(HIEC), but this technique is not feasible for use in large population-based studies or clinical settings because of its complexity and costs.10 The homoeostasis model assessment of insulin resistance (HOMA-IR) is the most often used surrogate of insulin resistance, but it has little value in people receiving insulin treatment or those who do not have functioning β cells.11 Furthermore, circulating insulin concentration is not routinely measured, which also limits the use of the HOMA-IR.10 The triglyceride glucose (TyG) index is a simple, accessible, and reliable clinical surrogate marker of insulin resistance.12,13 The diagnostic accuracy of the TyG index in identifying insulin resistance using the HIEC and HOMA-IR as reference methods has been tested in several studies. The highest achieved sensitivity was 96% with HIEC, and the highest specificity was of 99% with HOMA-IR.14 The TyG index has also shown good performance in the estimation of insulin resistance compared with the HOMA-IR in individuals with and without diabetes.14 Additionally, the TyG index does not require insulin quantification and might be used in all people, regardless of their insulin treatment status.11 The TyG index is associated with increased risk of mortality and cardiovascular diseases, including carotid atherosclerosis, coronary artery
Correspondence to:
Prof Patricio Lopez-Jaramillo, Masira Research Institute, Universidad de Santander, 680003 Bucaramanga, Colombia jplopezj@gmail.com
disease, metabolic syndrome, and type 2 diabetes.15–18 However, the association of the TyG index, as a surrogate marker of insulin resistance, with cardiovascular diseases and deaths, has principally been investigated in Asia, with few data available from other regions of the world or from LICs and MICs.
In the current analysis from the PURE study, we aimed to assess the association of the TyG index with mortality and cardiovascular disease events in populations who are at different levels of economic development, from five continents. We also aimed to examine whether the magnitude of associations between insulin resistance (as determined by the TyG index) and cardiovascular disease events differ according to the income level of the countries.
Methods
Study design and participants
The design and methods of the PURE study has been previously described.4,19,20 Briefly, PURE is a large-scale cohort study of participants aged between 35 and 70 years with an unbiased approach of sampling individuals from over 660 communities in 22 countries. Participating countries were selected to represent significant socioeconomic heterogeneity. The countries included four HICs (Canada, Saudi Arabia, Sweden, and the United Arab Emirates), 13 MICs (Argentina, Brazil,
Chile, China, Colombia, Iran, Malaysia, occupied Palestinian territory, Philippines, Poland, Russia, South Africa, and Türkiye), and five LICs (Bangladesh, India, Pakistan, Tanzania, and Zimbabwe; figure 1). Countries were classified on the basis of the World Bank classification for 2006. Recruitment began on Jan 1, 2003, and was completed by Feb 26, 2019. Within each country, rural and urban communities were selected on the basis of the national definition for each country. Individuals within the prespecified age range, who were expected to remain in their community for at least 4 years, were eligible for inclusion. Guidelines for the selection of countries, communities, households, and individuals to participate are in the appendix (pp 1–5). The study complies with the Declaration of Helsinki, the protocol was approved by the ethics committee at each participating site, and all participants provided written informed consent.
Procedures
At the baseline visit, a fasting blood sample was collected and frozen at temperatures between –20°C and –70°C.
Serum samples were shipped in nitrogen vapour tanks by courier from every site to a blood storage site, where they were stored at −160°C in nitrogen vapour (in Hamilton, ON, Canada) or at −70°C (in China, India, and Türkiye). Fasting blood samples were analysed for
Figure 1: Map of the countries participating in this analysis from the PURE study according to income level High-income countries
Middle-income countries Low-income countries
Canada
Colombia
Brazil
Chile
Argentina
Sweden
Poland
Occupied Palestinian territory
Saudi Arabia
United Arab Emirates Iran
Türkiye
Pakistan India
Bangladesh China
Malaysia Philippines Russia
Tanzania
South AfricaZimbabwe
See Online for appendix
glucose, total cholesterol, LDL cholesterol, HDL cholesterol, and triglycerides at the Population Health Research Institute, Hamilton, ON, Canada. Blood samples from China, India, and Türkiye were analysed in a central laboratory in each country after standardisation with the laboratory in Hamilton.21 The TyG index was calculated as Ln (fasting triglycerides [mg/dL] x fasting plasma glucose [mg/dL]/2).12,13
Trained study personnel administered standardised questionnaires and measured participants’ anthro po- metrics.20,21 Information on sociodemographic charac- teristics, education status, personal medical history,
tobacco and alcohol use, and cardiovascular disease risk factors was recorded. Physical measurements included weight, height, and waist and hip circumferences.
Handgrip strength was measured with a Jamar dynamometer (Sammons Preston, Bolingbrook, IL, USA) according to a standardised protocol,22 and two recordings of resting blood pressure with the use of an Omron HEM- 757 automatic digital monitor (Omron, Tokyo, Japan) were recorded in all participants.
During follow-up, participants were contacted at least once every 3 years either by telephone or in-person visits by the local research team. Information on events of
Overall (n=141 243) High-income countries
(n=16 294) Middle-income
countries (n=104 194)* Low-income countries
(n=20 755)† p value‡
Triglyceride glucose index 8·57 (0·69) 8·49 (0·64) 8·60 (0·70) 8·51 (0·70) <0·0001
Age, years 51·31 (9·58) 52·22 (9·44) 51·68 (9·38) 48·76 (10·26) <0·0001
Sex
Male 57 868 (41·0%) 7642 (46·9%) 41 494 (39·8%) 8732 (42·1%) <0·0001
Female 83 375 (59·0%) 8652 (53·1%) 62 700 (60·2%) 12 023 (57·9%) <0·0001
Location
Urban 79 602 (56·4%) 12 856 (78·9%) 56 159 (53·9%) 10 587 (51·0%) <0·0001
Rural 61 641 (43·6%) 3438 (21·1%) 48 035 (46·1%) 10 168 (49·0%) <0·0001
Education
Presecondary school 49 757/127 958 (38·9%) 1814/16 248 (11·2%) 38 011/91 056 (41·7%) 9932/20 654 (48·1%) <0·0001 Secondary school 48 913/127 958 (38·2%) 4696/16 248 (28·9%) 36 255/91 056 (39·8%) 7962/20 654 (38·5%) <0·0001 Postsecondary school 29 288/127 958 (22·9%) 9738/16 248 (59·9%) 16 790/91 056 (18·4%) 2760/20 654 (13·4%) <0·0001 Smoking status
Current 28 058/139 903 (20·1%) 2151/16 242 (13·2%) 21 523/103 022 (20·9%) 4384/20 639 (21·2%) <0·0001 Former 20 919/139 903 (15·0%) 5309/16 242 (32·7%) 14 639/103 022 (14·2%) 971/20 639 (4·7%) <0·0001 Never 90 926/139 903 (64·9%) 8782/16 242 (54·1%) 66 860/103 022 (64·9%) 15 284/20 639 (74·1%) <0·0001 Drinking status
Current 48 055/138 744 (34·6%) 10 827/16 233 (66·7%) 35 048/101 832 (34·4%) 2180/20 679 (10·5%) <0·0001
Former 7220/138 744 (5·2%) 994/16 233 (6·1%) 5608/101 832 (5·5%) 618/20 679 (3·0%) <0·0001
Never 83 469/138 744 (60·2%) 4412/16 233 (27·2%) 61 176/101 832 (60·1%) 17 881/20 679 (86·5%) <0·0001 Hypertension 61 248/140 709 (43·5%) 6489/16 254 (39·9%) 47 891/103 732 (46·2%) 6868/20 723 (33·1%) <0·0001 Diabetes 11 073/140 714 (7·9%) 1468/16 252 (9·0%) 7407/103 746 (7·1%) 2198/20 716 (10·6%) <0·0001 Coronary artery disease 6744/140 675 (4·8%) 648/16 239 (4·0%) 5568/103 717 (5·4%) 528/20 719 (2·5%) <0·0001
Stroke 2685/140 725 (1·9%) 213/16 249 (1·3%) 2259/103 761 (2·2%) 213/20 715 (1·0%) <0·0001
BMI, kg/m2 26·20 (5·20) 27·79 (5·45) 26·48 (5·01) 23·51 (4·98) <0·0001
Waist-to-hip ratio
Male 0·92 (0·08) 0·94 (0·07) 0·92 (0·07) 0·91 (0·08) <0·0001
Female 0·85 (0·08) 0·83 (0·08) 0·85 (0·08) 0·84 (0·09) ··
Handgrip strength/weight 0·43 (0·15) 0·43 (0·14) 0·43 (0·15) 0·41 (0·15) <0·0001
Systolic blood pressure, mm Hg 132·83 (60·08) 129·20 (19·56) 134·47 (68·68) 127·36 (22·20) <0·0001
Diastolic blood pressure, mm Hg 82·32 (58·07) 81·24 (11·72) 82·89 (67·13) 80·28 (13·19) <0·0001
Fasting glucose, mmol/L 5·34 (1·89) 5·54 (1·61) 5·29 (1·72) 5·43 (2·73) <0·0001
Total cholesterol, mmol/L 4·99 (1·19) 5·23 (1·15) 5·04 (1·20) 4·59 (1·06) <0·0001
LDL cholesterol, mmol/L 3·14 (0·98) 3·25 (0·90) 3·12 (0·93) 3·20 (1·23) <0·0001
HDL cholesterol, mmol/L 1·24 (0·36) 1·38 (0·40) 1·23 (0·34) 1·18 (0·40) <0·0001
Triglycerides, mmol/L 1·55 (1·04) 1·34 (0·87) 1·60 (1·08) 1·45 (0·91) <0·0001
Data are n (%), n/N (%), or mean (SD). Pairwise comparison adjusted with Bonferroni correction and Tukey’s adjustment were used for multiple comparisons. *p<0·05 for all values in this column (except for sex) are in comparison with high-income countries. †p<0·05 for all values in this column (except for sex) are in comparison with middle- income and high-income countries. ‡Pearson’s χ² test (categorical variables) and one-factor ANOVA (continuous variables).
Table 1: Characteristics of the study participants at enrolment by country income level
interest (including available documentation) was obtained from participants, family, and from hospital or clinic records whenever possible. Standardised case- report forms, death certificates, medical records, and verbal autopsies were used to record data on major cardiovascular events and mortality (classified by cause), which were adjudicated centrally in each country by trained physicians using common definitions. Verbal autopsy reports were obtained when the cause of death could not be ascertained from available medical records (appendix pp 4–5).1,4 For the current analysis, we included all outcome events known to us until Dec 31, 2020.
Outcomes
The primary outcome of this analysis was the composite of mortality or major cardiovascular events (defined as death from cardiovascular causes, and non-fatal myocardial infarction, or stroke). Secondary outcomes were non- cardiovascular mortality, cardiovascular mortality, all myocardial infarctions, stroke, and incident diabetes.
Statistical analysis
Continuous variables were expressed as means (SD) and categorical variables as percentages. We used comparisons with ANOVA followed by post-hoc analysis with Tukey’s adjustment for multiple comparisons, and Pearson’s χ² test with pairwise comparison as post-hoc test (with Bonferroni corrected p values), as appropriate.
Education was categorised as presecondary school (first 6 years of schooling), secondary school (7–11 years of schooling), and postsecondary school (>11 years of schooling). Smoking and alcohol status were categorised as never, former, or current. Participants were categorised according to their country’s income level (ie, HICs, MICs, and LICs) and into TyG index tertiles.
The Kaplan-Meier method was done to describe the incidence rate of clinical outcomes. Moreover, we calculated hazard ratios (HRs) using a multivariable Cox frailty model with random effects to account for centre clustering to test the associations between the TyG index and risk of cardiovascular diseases and mortality (appendix pp 5–6).23 The proportional hazards assumption was checked by visual inspection of log to log plots, and by scaled Schoenfeld residuals. We present HRs and 95% CIs for TyG index tertiles, adjusted for multiple confounding variables (age, sex, location [urban or rural], education, smoking status, alcohol use, marital status, history of hypertension, diabetes, coronary artery disease, stroke, cancer, BMI, waist-to-hip ratio, weight-adjusted handgrip strength, LDL and HDL cholesterol, and study centre as a random effect). Adjustment variables were selected on the basis of clinical knowledge, previous literature, and following the stepwise selection method.
Sensitivity and subgroup analyses
To evaluate the robustness of the results, sensitivity analysis was done by excluding those participants with a
previous history of cardiovascular diseases, or diabetes (as appropriate), and those participants who developed the clinical outcome within the first 2 years of follow-up.
To evaluate whether the TyG index could provide additional information for the risk assessment of adverse cardiovascular events compared with glucose and triglycerides alone, the HRs and 95% CIs for these biochemical parameters were also estimated.
We also did subgroup analyses to examine the magnitude of associations between insulin resistance (ie, the TyG index) and outcome events according to the income level of the countries based on the assumption that the populations from LICs and MICs might be more susceptible to adverse outcomes than participants from HICs. To determine whether associations between the TyG index and outcomes were uniform across countries of different income, we first modelled the TyG index (divided by tertiles) as part of an interaction term with country
Overall
(n=141 243) Triglyceride glucose index p value*
Tertile 1
(n=47 099) Tertile 2
(n=47 067)† Tertile 3 (n=47 077)‡
Triglyceride
glucose index 8·58 (0·70;
3·42–14·50) 7·86 (0·38;
3·42–8·27) 8·54 (0·15;
8·27–8·81) 9·33 (0·48;
8·81–14·50) <0·0001 Hypertension,
n (%) 61 248/140 709
(43·5%) 15 082/46 922
(32·1%) 20 012/46 875
(42·7%) 26 154/46 912
(55·8%) <0·0001 Diabetes, n (%) 11 073/140 714
(7·9%) 1170/46 927
(2·5%) 2287/46 884
(4·9%) 7616/46 903
(16·2%) <0·0001 Coronary artery
disease, n (%) 6744/140 675
(4·8%) 1665/46 914
(3·5%) 2178/46 865
(4·6%) 2901/46 896
(6·2%) <0·0001
Stroke, n (%) 2685/140 725
(1·9%) 706/46 931
(1·5%) 877/46 880
(1·9%) 1102/46 914
(2·3%) <0·0001
BMI, kg/m2 26·2 (5·2) 24·4 (4·7) 26·2 (5·1) 27·9 (5·2) <0·0001
Waist-to-hip ratio
Male 0·92 (0·08) 0·89 (0·07) 0·91 (0·07) 0·94 (0·07) <0·0001
Female 0·85 (0·08) 0·82 (0·07) 0·85 (0·07) 0·88 (0·07) <0·0001
Handgrip
strength/weight 0·43 (0·15) 0·45 (0·15) 0·43 (0·15) 0·40 (0·14) <0·0001 Systolic blood
pressure, mm Hg 132·8 (60·1) 128·1 (75·3) 132·5 (50·4) 137·9 (50·7) <0·0001 Diastolic blood
pressure, mm Hg 82·3 (58·1) 79·8 (74·0) 82·1 (48·1) 85·0 (48·1) <0·0001 Fasting glucose,
mmol/L 5·3 (1·9) 4·6 (0·8) 5·0 (0·9) 6·3 (2·8) <0·0001
Total cholesterol,
mmol/L 5·0 (1·2) 4·6 (1·0) 5·0 (1·1) 5·4 (1·3) <0·0001
LDL cholesterol,
mmol/L 3·1 (1·0) 2·9 (0·9) 3·2 (1·0) 3·3 (1·1) <0·0001
HDL cholesterol,
mmol/ 1·2 (0·4) 1·3 (0·4) 1·2 (0·3) 1·1 (0·3) <0·0001
Triglycerides,
mmol/L 1·6 (1·4) 0·8 (0·2) 1·3 (0·3) 2·6 (2·0) <0·0001
Data are n/N (%), mean (SD), or mean (SD; range). Pairwise comparison adjusted with Bonferroni correction and Tukey’s adjustment were used for multiple comparisons.*Pearson’s χ² test (categorical variables) and one-factor ANOVA (continuous variables). †p<0·05 in comparison with tertile 1. ‡p<0·05 in comparison with tertile 1 and tertile 2.
Table 2: Metabolic characteristics of the study participants at enrolment by tertiles of triglyceride glucose index
income (stratified as HICs, MICs, or LICs). A significant interaction indicates heterogeneity of the association between the TyG index and outcome across country- income strata.
All analyses were done with the R statistical software (version 3.6.3), and the R packages survival (version 3.2-13), survminer (version 0.4.9), finalfit (version 1.0.4), emmeans (version 1.7.2), and forestplot (version 2.0.1). A two-sided p<0·05 was considered statistically significant.
Role of the funding source
The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report.
Results
This study includes the 141 243 participants in whom baseline values of triglycerides and fasting plasma glucose, and follow-up data, were available (appendix p 7).
Of these, 16 294 (11·5%) individuals are from HICs,
104 194 (73·8%) from MICs, and 20 755 (14·7%) from LICs. Baseline participant characteristics stratified by country income level are in table 1. The mean age of the overall population was 51·3 years (SD 9·5); participants from HICs were older than those from MICs or LICs.
Among participants from HICs, the majority were residents from urban areas, compared with about half from urban areas in LICs. In HICs, a high proportion had a post-secondary education compared with less than 20% in MICs and LICs. Participants from HICs weighed more, compared with those from LICs, and adjusted handgrip strength was lower in those from LICs compared to those in HICs and MICs. Additional details on participant characteristics are presented in table 1.
Metabolic characteristics of the study participants at enrolment by tertiles of the TyG index were recorded (table 2). Participants with a higher TyG index had a higher prevalence of hypertension, diabetes, coronary artery disease, and stroke; a higher BMI, waist-to-hip ratio, systolic blood pressure, diastolic blood pressure,
Figure 2: Kaplan-Meier curves to estimate cumulative hazard for the development of clinical outcomes by baseline triglyceride glucose index tertiles The primary composite outcome was a major cardiovascular event (non-fatal myocardial infarction, or stroke) or death from cardiovascular causes.
Number at risk (number censored) Tertile 1 Tertile 2 Tertile 3
0
39 500 39 927(0) (0) 39 279
(0) 5
37 994 (1035) 37 990 (1257) 37 243 (1045)
10
35 230 (3125) 34 718 (3529) 33 618 (3359)
15
9027 (29 037)
7323 (30 546) (30 363)6122
20
83 (37 961)
73 (37 782) (36 387)81 0
0·025 0·050 0·075 0·100
0
39 500 39 927(0) (0) 39 279
(0) 5
37 689 (1276) 37 823 (1572) 37 200 (1479)
10
34 770 (3589) 34 511 (4182) 33 585 (4254)
15
8679 (29 314)
7171 (31 145) (31 330)6061
20
71 (37 848)
70 (38 177) (37 231)78 0
0·05 0·10 0·15
0
39 500 39 927(0) (0) 39 279
(0) 5
38 296 (1036) 38 445 (1258) 37 911 (1051)
10
35 967 (3141) 35 817 (3557) 35 153 (3399)
15
9281 (29 753)
7608 (31 657) (31 938)6465
20
85 (38 939)
76 (39 187) (38 312)83 0
0·01 0·02 0·03 0·04
Cumulative hazard
Composite Non-cardiovascular mortality Cardiovascular mortality
Number at risk (number censored) Tertile 1 Tertile 2 Tertile 3
0
39 500 (0) 39 927
(0) 39 279
(0) 5
38 240 (1036) 38 323 (1258) 37 675 (1050)
10
35 846 (3140) 35 595 (3557) 34 689 (3395)
15
(29 595)9272 7579 (31 414) (31 444)6397
20
(38 774)83 75 (38 909) (37 750)83 Years
0 0·01 0·02 0·04 0·03 0·05
0
39 500 (0) 39 927
(0) 39 279
(0) 5
38 212 (1036) 38 312 (1258) 37 763 (1046)
10
35 725 (3147) 35 458 (3548) 34 773 (3388)
15
(29 482)9225 7576 (31 231) (31 470)6449
20
(38 615)87 77 (38 726) (37 830)86 Years
0 0·01 0·02 0·04 0·03 0·05
0
39 500 (0) 39 927
(0) 39 279
(0) 5
38 218 (1036) 38 133 (1259) 37 240 (1048)
10
35 715 (3150) 35 067 (3559) 33 725 (3383)
15
(29 482)9183 7417 (30 870) (30 296)6329
20
(38 561)81 78 (38 189) (36 535)77 Years
0 0·025 0·050 0·075 0·100
Cumulative hazard
Myocardial infarction Stroke Incident diabetes
Tertile 1 Tertile 2 Tertile 3
fasting glucose, total cholesterol, LDL cholesterol, and triglycerides levels; but lower adjusted handgrip strength and HDL cholesterol levels compared with participants in the tertile 1 group.
During a median follow-up of 13·2 years (IQR 11·9–14·6), there were 6345 composite events, 2030 cardiovascular deaths, 3038 incident cases of myocardial infarction, 3291 incident cases of stroke, and 5191 cases of incident type 2 diabetes. The proportional hazards assumption was checked visually and found to be met. After adjusting for all other variables, participants in tertile 3 of the baseline TyG index had a higher risk for cardiovascular and metabolic events compared with participants in the other two groups during follow-up (figure 2). Estimates of the associations of the TyG index with risks of various clinical outcomes were recorded (figure 3). The risk of cardiovascular and metabolic events increased by higher baseline TyG index tertiles and remained significant even after adjustment for potential confounding factors. Compared with the lowest
tertile (tertile 1), the highest TyG index group (tertile 3) was associated with a greater incidence of the composite outcome (HR 1·21; 95% CI 1·13–1·30), myocardial infarction (1·24; 1·12–1·38), stroke (1·16; 1·05–1·28), and incident diabetes (1·99; 1·82–2·16). As expected, the TyG index was not associated with non-cardiovascular mortality.
Associations of the TyG index with outcomes were not modified when participants with a previous history of cardiovascular diseases, or diabetes (as appropriate), and those who developed an outcome of interest within the first 2 years of follow-up were excluded (appendix pp 8–9).
Moreover, the TyG index showed a greater association with cardiovascular events and type 2 diabetes than glucose and triglyceride levels alone (appendix pp 10–11).
The associations between the TyG index and outcomes varied considerably by country income level (figure 4, appendix pp 12–13). In HICs, the TyG index was only significantly associated with increased hazard of incident diabetes. Those in the highest tertile of the TyG index
Composite Tertile 1 Tertile 2 Tertile 3 Overall
Non-cardiovascular mortality Tertile 1
Tertile 2 Tertile 3 Overall
Cardiovascular mortality Tertile 1
Tertile 2 Tertile 3 Overall
Myocardial infarction Tertile 1
Tertile 2 Tertile 3 Overall Stroke Tertile 1 Tertile 2 Tertile 3 Overall Incident diabetes Tertile 1 Tertile 2 Tertile 3 Overall
39 500 (1457) 39 927 (2074) 39 279 (2814) 118 706 (6345) 39 500 (1582) 39 927 (1682) 39 279 (1975) 118 706 (5239) 39 500 (477) 39 927 (666) 39 279 (887) 118 706 (2030)
39 500 (644) 39 927 (945) 39 279 (1449) 118 706 (3038) 39 500 (799) 39 927 (1126) 39 279 (1366) 118 706 (3291) 39 500 (859) 39 927 (1662) 39 279 (2670) 118 706 (5191) At risk (events)
1
1·10 (1·02−1·18) 1·21 (1·13−1·30)
1
1·00 (0·93−1·08) 1·04 (0·96−1·12)
1
1·06 (0·94−1·20) 1·09 (0·96−1·24)
1
1·04 (0·94−1·15) 1·24 (1·12−1·38)
1
1·15 (1·05−1·27) 1·16 (1·05−1·28)
1
1·54 (1·41−1·67) 1·99 (1·82−2·16) HR (95% CI)
0·9 1·2 1·5 1·8 2·1 2·4
Figure 3: Association of triglyceride glucose index with clinical outcomes by tertiles
HRs were calculated with the multivariable Cox frailty model with random effects to account for centre clustering. Models were adjusted for date of entrance to the cohort; age at recruitment; sex; centre (as a random effect); location (urban or rural); education level; tobacco and alcohol use; marital status; history of hypertension, diabetes, coronary artery disease, stroke, and cancer; BMI; waist-to-hip ratio; weight-adjusted handgrip strength; and LDL and HDL cholesterol. HR=hazard ratio.
had the highest hazards of incident type 2 diabetes (HR 2·95; 95% CI 2·25–3·87) compared with the lowest tertile, even after adjusting for all confounders. In contrast, in MICs and LICs the hazards for the composite outcome, cardiovascular mortality, myocardial infarction, stroke, and incident diabetes, were substantially higher with increasing TyG index tertiles. Within MICs, there was an increased risk of the composite outcome (HR 1·20; 95% CI 1·11–1·31), myocardial infarction (1·26; 1·10–1·45), stroke (1·17; 1·05–1·30), and incident type 2 diabetes (2·68; 2·40–2·99) among those in the highest tertile of the TyG index, compared with the lowest tertile. Similar results were seen in LICs, with a 1·44-times higher risk of cardiovascular mortality in the highest TyG group compared with those in the lowest TyG group (1·44; 1·15–1·80).
Discussion
In this large prospective cohort study from 22 countries across five continents, we found that the TyG index was
significantly associated with incident cardiovascular diseases (eg, myocardial infarction and stroke), cardiovascular mortality, and incident type 2 diabetes. As expected, the TyG index was not associated with non- cardiovascular mortality and, therefore, there was no association with total mortality. These results were consistently maintained even after sensitivity analyses were done. Subgroup analyses showed that the association between the TyG index and risk of cardiovascular diseases and type 2 diabetes varied according to the income level of the countries; participants from MICs and LICs displayed a higher risk of cardiovascular mortality, myocardial infarction, stroke, and incident diabetes with increases in the TyG index, whereas in HICs, there was only an increased risk of incident type 2 diabetes, but not of other outcomes. These findings show that insulin resistance is a risk factor for incident metabolic and cardiovascular diseases, and suggest that individuals in MICs and LICs are potentially more susceptible to this metabolic disturbance.
Figure 4: Association between triglyceride glucose index tertiles and clinical outcomes by the income level of the countries
HRs were calculated with the multivariable Cox frailty model with random effects to account for centre clustering. pinteraction are pfor testing the interaction between triglyceride glucose index by tertiles and country income. Models were adjusted for date of entrance to the cohort; age at recruitment; sex; centre (as a random effect); location (urban or rural); education level; tobacco and alcohol use;
marital status; history of hypertension, diabetes, coronary artery disease, stroke, and cancer; BMI; waist-to-hip ratio; weight-adjusted handgrip strength; and LDL and HDL cholesterol. HR=hazard ratio.
Composite Tertile 1 Tertile 2 Tertile 3 Overall
Non-cardiovascular mortality Tertile 1
Tertile 2 Tertile 3 Overall
Cardiovascular mortality Tertile 1
Tertile 2 Tertile 3 Overall
Myocardial infarction Tertile 1
Tertile 2 Tertile 3 Overall Stroke Tertile 1 Tertile 2 Tertile 3 Overall Incident diabetes Tertile 1 Tertile 2 Tertile 3 Overall
5951 (150) 5029 (194) 4307 (247) 15 287 (591) 5951 (143) 5029 (188) 4307 (172) 15 287 (503) 5951 (15) 5029 (28) 4307 (45) 15 287 (88) 5951 (88) 5029 (115) 4307 (155) 15 287 (358) 5951 (59) 5029 (71) 4307 (80) 15 287 (210)
5951 (84) 5029 (198) 4307 (338) 15 287 (620) At risk (events)
1
0·94 (0·75−1·18) 0·97 (0·76−1·24)
1
1·16 (0·92−1·47) 0·98 (0·75−1·28)
1
1·41 (0·73−2·73) 1·72 (0·86−3·43)
1
0·90 (0·67−1·20) 0·92 (0·67−1·26)
1
0·90 (0·62−1·30) 0·88 (0·59−1·31)
1
1·76 (1·35−2·30) 2·95 (2·25−3·87) HR (95% CI)
1
1·14 (1·05−1·24) 1·20 (1·11−1·31)
1
1·00 (0·91−1·09) 1·04 (0·94−1·15)
1
0·99 (0·85−1·15) 0·91 (0·78−1·07)
1
1·09 (0·95−1·26) 1·26 (1·10−1·45)
1
1·18 (1·06−1·31) 1·17 (1·05−1·30)
1
1·65 (1·47−1·84) 2·68 (2·40−2·99) HR (95% CI)
26 511 (950) 28 804 (1504) 29 598 (2018) 84 913 (4472) 26 511 (833) 28 804 (1044) 29 598 (1387) 84 913 (3264) 26 511 (297) 28 804 (441) 29 598 (564) 84 913 (1302) 26 511 (329) 28 804 (575) 29 598 (906) 84 913 (1810) 26 511 (614) 28 804 (930) 29 598 (1118) 84 913 (2662) 26 511 (454) 28 804 (1048) 29 598 (1968) 84 913 (3470) At risk (events)
0·01
0·02
0·01
0·08
0·19
<0·0001 pinteraction
1
1·01 (0·87−1·17) 1·31 (1·12−1·54)
1
0·92 (0·81−1·04) 0·98 (0·85−1·13)
1
1·16 (0·93−1·43) 1·44 (1·15−1·80)
1
0·98 (0·81−1·18) 1·29 (1·06−1·56)
1
1·12 (0·87−1·46) 1·35 (1·02−1·78)
1
1·49 (1·27−1·73) 1·64 (1·38−1·94) HR (95% CI)
7038 (357) 6094 (376) 5374 (549) 18 506 (1282)
7038 (606) 6094 (450) 5374 (416) 18 506 (1472)
7038 (165) 6094 (197) 5374 (278) 18 506 (640) 7038 (227) 6094 (255) 5374 (388) 18 506 (870) 7038 (126) 6094 (125) 5374 (168) 18 506 (419) 7038 (321) 6094 (416) 5374 (364) 18 506 (1101) At risk (events)
0·8 1·2 1·6 2·4 0·8 1·2 1·6 2·4 0·8 1·2 1·6 2·4
High-income countries Middle-income countries Low-income countries
Our findings are consistent with previous studies done mostly in Asia.15,17,24–29 In a cohort study of 6078 Chinese patients older than 60 years, quartile 3 of the TyG index was associated with a 1·33-times higher risk of cardiovascular diseases, and quartile 4 with at 1·72-times risk, compared with the lowest TyG index quartile.27 Another prospective cohort study, comprising 5014 White participants, showed that people in the highest quintile group had a 2·32-times higher risk of developing cardiovascular diseases than those in the lowest quintile group.28 Data from the National Health Information Database of South Korea also showed that, independent of other traditional cardiovascular risk factors, individuals in the highest TyG index quartile were at higher risk of stroke and myocardial infarction.29 A long-term follow-up of a Korean cohort reported that individuals in the highest baseline quartile of the TyG index had a four-times higher risk of developing diabetes,30 and that changes in the TyG index over time altered the incidence and risk of diabetes.31 To our knowledge, however, the present study is the first to report the association between the TyG index and cardiometabolic diseases across countries at different stages of economic development.
Although our study further supports the association between the TyG index and cardiovascular mortality, incident cardiovascular diseases, and incident type 2 diabetes, the underlying mechanisms have not yet been fully elucidated. The TyG index is the calculated transformation of fasting triglyceride and glucose levels.12,13 Substantial evidence has shown the relationship between the TyG index and insulin resistance.10,12,14 In fact, it was documented that the TyG index was the best index to identify individuals with insulin resistance, even superior to visceral adiposity indicators and other lipid parameters.32 Hence, the TyG index has been proposed as a simple and reliable clinical surrogate marker of insulin resistance.12,13 Consequently, we speculate that insulin resistance might be important in linking the TyG index with cardiovascular diseases and type 2 diabetes.
Insulin resistance causes altered glucose metabolism, chronic hyperglycaemia, and abnormal lipid metabolism, resulting in chronic inflammation, oxidative stress, and endothelial dysfunction that might lead to cellular damage and atherosclerosis.5–7,33 Moreover, an inverse relationship between insulin resistance and insulin secretion has been confirmed, so the decrease in insulin sensitivity is compensated by an increase of β-cell secretion. Thus, insulin production rises proportionally to insulin resistance to try to maintain the glucose metabolism homoeostasis.34 Persistent insulin resistance might induce the development of chronic diseases, such as type 2 diabetes, and cardiovascular diseases,33,35 and a high TyG index might promote the development of cardiovascular disease events through these pathways.
Moreover, it has been previously described that fasting plasma glucose mainly reflects insulin resistance in the liver, whereas fasting triglycerides mainly reflects insulin
resistance in adipocytes.36 This is consistent with the fact that the TyG index, as a product of triglycerides and plasma glucose, had a higher association with outcomes than glucose and triglyceride values separately (appendix pp 10–11).
Previous analyses from the PURE study showed a higher incidence of mortality and cardiovascular diseases in MICs and LICs, despite lower BMI levels in those regions.4 Moreover, cardiovascular diseases rates and mortality were markedly higher among those with type 2 diabetes in LICs.37 These contrasts between countries at different levels of economic development might be related, in part, to differences in access to health- care systems. However, the current study suggests a plausible pathophysiological mechanism, which might also contribute to these findings. Potentially, the association between the TyG index and the higher risk of cardiovascular diseases and type 2 diabetes in MICs and LICs might be explained by an increased vulnerability of these populations to the presence of insulin resistance.
This might be due to the discrepancy between maternal undernutrition and its consequence (ie, low birthweight infants) and subsequent exposure to modern lifestyles and an obesogenic environment in adulthood.2,9 The rapid changes resulting from the urbanisation process are associated with socioeconomic inequalities and have produced an environment that discourages physical activity and encourages the consumption of energy dense diets.2,8,9,38 This mismatch between conditions experienced during fetal program ming, and current environmental conditions, makes adaptation difficult for people in LICs and MICs, and will increase their susceptibility to insulin resistance, obesity, and cardiovascular diseases. Future studies are needed to confirm the variable relationship between insulin resistance and cardiometabolic outcomes among countries at different levels of socioeconomic development.
The major strength of our study is that, to our knowledge, it is the only study that has prospectively recorded information on individual-level risk factors, characteristics, treatments, diseases, hospital admissions, and deaths from several HICs, MICs, and LICs using a standardised methodology. Therefore, the PURE study is uniquely positioned to inform comparisons of risk factors, morbidity, and mortality across countries at different income levels. Nonetheless, the present study also has some potential limitations. First, the level of fasting insulin was not measured, and thus we could not compare the association level of the TyG index with those of HOMA-IR and the HIEC test, the reference method of insulin resistance assessment. However, the TyG index has previously been reported to perform well in the estimation of insulin resistance compared with HIEC and HOMA-IR;14,17 and direct measurement of insulin resistance (HIEC and HOMA-IR) is complex and too expensive to be used in large-scale epidemiological studies, or in clinical practice. Second, for the current