[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Figure 1.
Adjusted Hazard Ratios (HRs) for Cardiovascular Disease by Usual Random Plasma Glucose (RPG) Level
Adjusted Hazard Ratios (HRs) for Cardiovascular Disease by Usual Random Plasma Glucose (RPG) Level

Results are stratified by age, sex, and study area and adjusted for educational level, smoking, alcohol intake, systolic blood pressure, and physical activity. Adjusted HRs are plotted against mean usual RPG level per 1 mmol/L (18 mg/dL) in each category. Squares represent the HR, with area inversely proportional to the variance of the log HR; limit lines, the corresponding 95% CIs. To convert RPG levels to milligrams per deciliter, divide by 0.0555.

Figure 2.
Adjusted Hazard Ratios (HRs) for Cardiovascular Death
Adjusted Hazard Ratios (HRs) for Cardiovascular Death

Adjusted HRs are calculated per 1-mmol/L (18-mg/dL) higher usual random plasma glucose levels. Results are stratified by age, sex, and study area and adjusted (except where it is the variable of interest) for educational level, smoking, alcohol intake, physical activity, and systolic blood pressure. Squares represent the adjusted HR with area inversely proportional to the variance of the log HR. Horizontal lines represent the corresponding 99% CI. The dotted line represents the overall adjusted HR. The open diamond represents the overall adjusted HR and its 95% CI. BMI indicates body mass index (calculated as weight in kilograms divided by height in meters squared); MET, metabolic equivalent of task; and SBP, systolic blood pressure.

Figure 3.
Adjusted Hazard Ratios (HRs) for Major Occlusive Vascular Disease
Adjusted Hazard Ratios (HRs) for Major Occlusive Vascular Disease

Adjusted HRs are calculated per 1-mmol/L (18-mg/dL) higher usual random plasma glucose level. Results are stratified by age, sex, and study area and adjusted (except where it is the variable of interest) for educational level, smoking, alcohol intake, physical activity, and systolic blood pressure. Squares represent the adjusted HR with area inversely proportional to the variance of the log adjusted HR. Horizontal lines represent the corresponding 99% CI. The dotted line represents the overall adjusted HR. The open diamond represents the overall adjusted HR and its 95% CI. BMI indicates body mass index (calculated as weight in kilograms divided by height in meters squared); MET, metabolic equivalent of task; and SBP, systolic blood pressure.

Table 1.  
Baseline Characteristics of Participants by RPG Level
Baseline Characteristics of Participants by RPG Level
Table 2.  
Adjusted Hazard Ratios (aHRs) for Major Cardiovascular Disease by Baseline RPG Levela
Adjusted Hazard Ratios (aHRs) for Major Cardiovascular Disease by Baseline RPG Levela
1.
Sarwar  N, Gao  P, Seshasai  SR,  et al; Emerging Risk Factors Collaboration.  Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies [published correction appears in Lancet. 2010;376(9745):958].  Lancet. 2010;375(9733):2215-2222.PubMedArticle
2.
Bragg  F, Li  L, Smith  M,  et al; China Kadoorie Biobank Collaborative Group.  Associations of blood glucose and prevalent diabetes with risk of cardiovascular disease in 500 000 adult Chinese: the China Kadoorie Biobank.  Diabet Med. 2014;31(5):540-551.PubMedArticle
3.
World Health Organization, International Diabetes Federation.  Definition and Diagnosis of Diabetes Mellitus and Intermediate Hyperglycaemia: Report of a WHO/IDF Consultation. Geneva, Switzerland: World Health Organization; 2006.
4.
Ford  ES, Zhao  G, Li  C.  Pre-diabetes and the risk for cardiovascular disease: a systematic review of the evidence.  J Am Coll Cardiol. 2010;55(13):1310-1317.PubMedArticle
5.
Gerstein  H, Punthakee  Z. Dysglycemia and the risk of cardiovascular events. In: Yusuf  S, Cairns  JA, Camm  AJ, Fallen  EL, Gersh  BJ, eds.  Evidence-Based Cardiology. 3rd ed. Chichester: Wiley-Blackwell; 2010:179-189.
6.
Lawes  CMM, Parag  V, Bennett  DA,  et al; Asia Pacific Cohort Studies Collaboration.  Blood glucose and risk of cardiovascular disease in the Asia Pacific region.  Diabetes Care. 2004;27(12):2836-2842.PubMedArticle
7.
Park  C, Guallar  E, Linton  JA,  et al.  Fasting glucose level and the risk of incident atherosclerotic cardiovascular diseases.  Diabetes Care. 2013;36(7):1988-1993.PubMedArticle
8.
Singh  GM, Danaei  G, Farzadfar  F,  et al; Global Burden of Metabolic Risk Factors of Chronic Diseases Collaborating Group; Asia-Pacific Cohort Studies Collaboration (APCSC); Diabetes Epidemiology: Collaborative analysis of Diagnostic criteria in Europe (DECODE); Emerging Risk Factor Collaboration (ERFC); Prospective Studies Collaboration (PSC).  The age-specific quantitative effects of metabolic risk factors on cardiovascular diseases and diabetes: a pooled analysis.  PLoS One. 2013;8(7):e65174.PubMedArticle
9.
Benn  M, Tybjaerg-Hansen  A, McCarthy  MI, Jensen  GB, Grande  P, Nordestgaard  BG.  Nonfasting glucose, ischemic heart disease, and myocardial infarction: a Mendelian randomization study.  J Am Coll Cardiol. 2012;59(25):2356-2365.PubMedArticle
10.
Chan  JC, Zhang  Y, Ning  G.  Diabetes in China: a societal solution for a personal challenge.  Lancet Diabetes Endocrinol. 2014;2(12):969-979.PubMedArticle
11.
Xu  Y, Wang  L, He  J,  et al; 2010 China Noncommunicable Disease Surveillance Group.  Prevalence and control of diabetes in Chinese adults.  JAMA. 2013;310(9):948-959.PubMedArticle
12.
Li  H, Ge  J.  Cardiovascular diseases in China: current status and future perspectives.  Int J Cardiol Heart Vasc.2015;6(0):25-31.Article
13.
Chen  Z, Chen  J, Collins  R,  et al; China Kadoorie Biobank (CKB) collaborative group.  China Kadoorie Biobank of 0.5 million people: survey methods, baseline characteristics and long-term follow-up.  Int J Epidemiol. 2011;40(6):1652-1666.PubMedArticle
14.
Chen  Z, Lee  L, Chen  J,  et al.  Cohort profile: the Kadoorie Study of Chronic Disease in China (KSCDC).  Int J Epidemiol. 2005;34(6):1243-1249.PubMedArticle
15.
Allen  N, Sudlow  C, Downey  P,  et al.  UK Biobank: current status and what it means for epidemiology.  Health Policy Technol. 2012;1(3):123-126.Article
16.
Lifescan. SureStep Technology. http://www.cliawaived.com/web/items/pdf/LifeScan_10797_Diabetes_Test~2251file1.pdf. Accessed February 7, 2016.
17.
World Health Organization. International Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10). http://apps.who.int/classifications/icd10/browse/2010/en. Accessed January 15, 2014.
18.
Easton  DF, Peto  J, Babiker  AG.  Floating absolute risk: an alternative to relative risk in survival and case-control analysis avoiding an arbitrary reference group.  Stat Med. 1991;10(7):1025-1035.PubMedArticle
19.
Harrell  FE  Jr.  Regression Modeling Strategies With Applications to Linear Models, Logistic Regression and Survival Analysis. New York: Springer-Verlag; 2001.
20.
Clarke  R, Shipley  M, Lewington  S,  et al.  Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies.  Am J Epidemiol. 1999;150(4):341-353.PubMedArticle
21.
Woodward  M.  Epidemiology Study Design and Data Analysis. 3rd ed. Boca Raton: CRC Press; 2014.
22.
Royston  P, Altman  DG.  Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling.  J R Stat Soc Ser C Appl Stat. 1994;43(3):429-467.
23.
MacMahon  S, Peto  R, Cutler  J,  et al.  Blood pressure, stroke, and coronary heart disease, part 1: prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias.  Lancet. 1990;335(8692):765-774.PubMedArticle
24.
Early Breast Cancer Trialists’ Collaborative Group.  Arithmetic Details of Tests for Trend, for Heterogeneity and for Interaction: Treatment of Early Breast Cancer. Worldwide Evidence 1985-1990: A Systematic Overview of All Available Randomized Trials of Adjuvant Endocrine and Cytotoxic Therapy. Vol 1. Oxford: Oxford University Press; 1990.
25.
Chien  K-L, Hsu  H-C, Su  T-C, Chen  M-F, Lee  Y-T, Hu  FB.  Fasting and postchallenge hyperglycemia and risk of cardiovascular disease in Chinese: the Chin-Shan Community Cardiovascular Cohort study.  Am Heart J. 2008;156(5):996-1002.PubMedArticle
26.
Chien  K-L, Lee  B-C, Lin  H-J, Hsu  H-C, Chen  M-F.  Association of fasting and post-prandial hyperglycemia on the risk of cardiovascular and all-cause death among non-diabetic Chinese.  Diabetes Res Clin Pract. 2009;83(2):e47-e50.PubMedArticle
27.
Einarson  TR, Machado  M, Henk Hemels  ME.  Blood glucose and subsequent cardiovascular disease: update of a meta-analysis.  Curr Med Res Opin. 2011;27(11):2155-2163.PubMedArticle
28.
Kawada  T.  Random blood glucose measurement for epidemiological studies: its significance and limitations.  Curr Med Res Opin. 2012;28(3):447-448.PubMedArticle
29.
Moebus  S, Göres  L, Lösch  C, Jöckel  KH.  Impact of time since last caloric intake on blood glucose levels.  Eur J Epidemiol. 2011;26(9):719-728.PubMedArticle
30.
Wu  B, Lin  S, Hao  Z,  et al.  Proportion, risk factors and outcome of lacunar infarction: a hospital-based study in a Chinese population.  Cerebrovasc Dis. 2010;29(2):181-187.PubMedArticle
31.
Liu  M, Wu  B, Wang  W-Z, Lee  L-M, Zhang  S-H, Kong  L-Z.  Stroke in China: epidemiology, prevention, and management strategies.  Lancet Neurol. 2007;6(5):456-464.PubMedArticle
32.
Liu  J, Hong  Y, D’Agostino  RB  Sr,  et al.  Predictive value for the Chinese population of the Framingham CHD risk assessment tool compared with the Chinese Multi-Provincial Cohort Study.  JAMA. 2004;291(21):2591-2599.PubMedArticle
33.
Chiasson  J-L, Josse  RG, Gomis  R, Hanefeld  M, Karasik  A, Laakso  M; STOP-NIDDM Trial Research Group.  Acarbose treatment and the risk of cardiovascular disease and hypertension in patients with impaired glucose tolerance: the STOP-NIDDM trial.  JAMA. 2003;290(4):486-494.PubMedArticle
34.
Gerstein  HC, Yusuf  S, Bosch  J,  et al; DREAM (Diabetes REduction Assessment with ramipril and rosiglitazone Medication) Trial Investigators.  Effect of rosiglitazone on the frequency of diabetes in patients with impaired glucose tolerance or impaired fasting glucose: a randomised controlled trial.  Lancet. 2006;368(9541):1096-1105.PubMedArticle
35.
Gerstein  HC, Bosch  J, Dagenais  GR,  et al; ORIGIN Trial Investigators.  Basal insulin and cardiovascular and other outcomes in dysglycemia.  N Engl J Med. 2012;367(4):319-328.PubMedArticle
36.
Holman  RR, Haffner  SM, McMurray  JJ,  et al; NAVIGATOR Study Group.  Effect of nateglinide on the incidence of diabetes and cardiovascular events.  N Engl J Med. 2010;362(16):1463-1476.PubMedArticle
37.
Rasmussen-Torvik  LJ, Li  M, Kao  WH,  et al.  Association of a fasting glucose genetic risk score with subclinical atherosclerosis: the Atherosclerosis Risk in Communities (ARIC) study.  Diabetes. 2011;60(1):331-335.PubMedArticle
38.
Ahmad  OS, Morris  JA, Mujammami  M,  et al.  A Mendelian randomization study of the effect of type-2 diabetes on coronary heart disease.  Nat Commun. 2015;6:7060.PubMedArticle
39.
Anderson  KM, Wilson  PW, Odell  PM, Kannel  WB.  An updated coronary risk profile: a statement for health professionals.  Circulation. 1991;83(1):356-362.PubMedArticle
40.
Hippisley-Cox  J, Coupland  C, Vinogradova  Y,  et al.  Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2.  BMJ. 2008;336(7659):1475-1482.PubMedArticle
Views 2,607
Citations 0
Original Investigation
October 2016

Association of Random Plasma Glucose Levels With the Risk for Cardiovascular Disease Among Chinese Adults Without Known Diabetes

Author Affiliations
  • 1Clinical Trial Service Unit and Epidemiological Studies Unit, Nuffield Department of Population Health, University of Oxford, Oxford, England
  • 2Department of Epidemiology and Biostatistics, School of Public Health, Peking University, Beijing, China
  • 3Chinese Academy of Medical Sciences, Beijing, China
  • 4China National Center for Food Safety Risk Assessment, Beijing, China
  • 5Henan Centre for Disease Control and Prevention, Zhengzhou, China
  • 6Heilongjiang Centre for Disease Control and Prevention, Harbin, China
  • 7Hainan Centre for Disease Control and Prevention, Haikou, China
  • 8Liuzhou Centre for Disease Control and Prevention, Liuzhou, China
  • 9Tongxiang Centre for Disease Control and Prevention, Tongxiang, China
JAMA Cardiol. 2016;1(7):813-823. doi:10.1001/jamacardio.2016.1702
Key Points

Question  How are random plasma glucose (RPG) levels associated with the risk for cardiovascular disease (CVD) in Chinese adults without known diabetes?

Findings  In this prospective cohort study of 467 508 Chinese adults without prior diabetes, a positive log-linear association of RPG levels with CVD risk was found. Each 1-mmol/L (18-mg/dL) higher usual RPG level was associated with an independent 8% to 11% higher risk for cardiovascular death or major ischemic CVD and a 5% greater risk for intracerebral hemorrhage.

Meaning  Higher RPG levels, even within a normal range, are associated with higher CVD risks in adult Chinese without known diabetes.

Abstract

Importance  Diabetes is a known risk factor for cardiovascular disease (CVD). Substantial uncertainty remains, however, about the relevance to CVD risk for blood glucose levels below the diabetes threshold.

Objective  To examine the association of random plasma glucose (RPG) levels with the risk for major CVD in Chinese adults without known diabetes.

Design, Setting, and Participants  This prospective cohort study included 467 508 men and women aged 30 to 79 years with no history of diabetes, ischemic heart disease (IHD), stroke, or transient ischemic attack. Participants were recruited from 5 urban and 5 rural diverse locations across China from June 25, 2004, to July 15, 2008, and followed up to January 1, 2014.

Exposures  Baseline and usual (longer-term average) RPG level.

Main Outcomes and Measures  Cardiovascular deaths, major coronary events (MCE) (including fatal IHD and nonfatal myocardial infarction), ischemic stroke (IS), major occlusive vascular disease (MOVD) (including MCE or IS), and intracerebral hemorrhage. Preliminary validation of stroke and IHD events demonstrated positive predictive values of approximately 90% and 85%, respectively. Cox regression yielded adjusted hazard ratios (aHRs) for CVD associated with RPG levels.

Results  Among the 467 508 participants (41.0% men; 59.0% women; mean [SD] age, 51 [11] years), a significant positive association of baseline RPG levels with CVD risks continued to 4.0 mmol/L (72 mg/dL). After adjusting for regression dilution bias, each 1-mmol/L (18-mg/dL) higher usual RPG level above 5.9 mmol/L (106 mg/dL) was associated with an 11% higher risk for cardiovascular death (6645 deaths; aHR, 1.11; 95% CI, 1.10-1.13). Similarly strong positive associations were seen for MCE (3270 events; aHR, 1.10; 95% CI, 1.08-1.13), IS (19 153 events; aHR, 1.08; 95% CI, 1.07-1.09), and MOVD (22 023 events; aHR, 1.08; 95% CI, 1.07-1.09). For intracerebral hemorrhage, the association was weaker, but also significant (4326 events; aHR, 1.05; 95% CI, 1.02-1.07). These associations persisted after excluding participants who developed diabetes during follow-up.

Conclusions and Relevance  Among adult Chinese without diabetes, lower RPG levels are associated with lower risks for major CVDs, even within a normal range of blood glucose levels.

Introduction

Diabetes is a major risk factor for cardiovascular disease (CVD).1,2 Evidence from Western populations suggests that individuals with prediabetes3 have elevated CVD risks,4 although the magnitude of risk in different populations and population subgroups is less clear. Below this range, uncertainty remains as to whether lower blood glucose levels are associated with a lower CVD risk,5 and, if so, whether the association is continuous6 or a threshold exists.1,7 Furthermore, most studies have focused on fasting blood glucose levels1,7,8 and few on random blood glucose levels, a more practical and arguably more relevant measure.9

Recent decades have seen a marked increase in diabetes prevalence10,11 and high prediabetes prevalence in China.11 Despite this, little reliable prospective evidence exists about the relevance to the risk for CVD of blood glucose levels below the diabetic threshold in China.12 Previous findings from the China Kadoorie Biobank collaborative group (CKB)2 showed a positive association of random plasma glucose (RPG) levels with prevalent CVD, but these findings were limited by their cross-sectional design and self-reported disease outcomes. We herein present 7-year prospective follow-up data from the CKB, examine the associations of RPG levels with risks for incident CVD in individuals without known type 1 or type 2 diabetes, and assess whether factors, such as age, sex, adiposity, and blood pressure, modify these.

Methods
Study Population

Details of the CKB design, survey methods, and population have been described previously.13,14 Briefly, the baseline survey took place from June 25, 2004, to July 15, 2008, in 10 diverse areas (5 urban and 5 rural) of China (eFigure 1 in the Supplement). The areas were selected to provide diversity in exposures and diseases and to take account of population stability, quality of disease and death registries, and capacity, and commitment within the areas. All permanent residents aged 35 to 74 years from 100 to 150 rural villages or urban committees in each area were invited to participate. Overall, approximately 30% responded,13 a proportion comparable with those of other large nationwide prospective studies.15 A total of 512 891 men and women were enrolled, including a small number just outside the target age range (n = 10 168). Ethical approval for this study was obtained from Oxford University, the Chinese Centre for Disease Control and Prevention, and the local Centres for Disease Control and Prevention in the 10 study areas. All participants provided written informed consent.

Data Collection

At local study assessment clinics, participants completed an interviewer-administered questionnaire collating data on demographics, socioeconomic factors, lifestyle measures (including smoking, alcohol consumption, diet, and physical activity), and medical history. Physical measurements included blood pressure, height, weight, and hip and waist circumferences. Measurements were obtained by trained health workers using calibrated instruments and standard protocols. A 10-mL nonfasting (with the exception of one area—Zhejiang—where participants were asked to fast) blood sample was collected from participants, and plasma glucose levels were measured immediately using commercially available meters (SureStep Plus meters; Lifescan, Johnson & Johnson),16 regularly calibrated with manufacturer control solutions. Data were collected on the time since the last food intake. Individuals with a plasma glucose level from 7.8 to less than 11.1 mmol/L (140 to <200 mg/dL) were invited back the following day for a fasting plasma glucose test. A resurvey of a 5% randomly selected sample of surviving participants was undertaken from May 26 to October 10, 2008, using the same procedures as in the baseline survey.

Follow-up for Morbidity and Mortality

Information on the vital status of participants was obtained from local death registries based at China’s Disease Surveillance Points, checked annually against local residential records and health insurance records, and confirmed through street committees or village administrators. Information on cause of death was supplemented by review of available medical records. In deaths without recent medical attention (approximately 5%), verbal autopsies determined probable causes. Information on events resulting in or during hospitalization was collected through linkage to established disease registries (for cancer, ischemic heart disease [IHD], stroke, and diabetes) and, via each individual’s unique national identification number, to the health insurance system, which provides almost universal coverage in the study areas. All events were coded using the International Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10),17 by trained staff who were blinded to baseline information.

The primary outcomes examined were cardiovascular death (ICD-10 codes I00-I25, I27-I88, and I95-I99), myocardial infarction (MI) (ICD-10 codes I21-I23), major coronary event (MCE) (nonfatal MI or fatal IHD [ICD-10 codes I20-I25]), ischemic stroke (IS) (ICD-10 code I63), intracerebral hemorrhage (ICH) (ICD-10 code I61), total stroke (ICD-10 codes I60, I61, I63, and I64), and major occlusive vascular disease (MOVD) (IS, nonfatal MI, or fatal IHD) (eTable 1 in the Supplement). By January 1, 2014, 2411 (0.5%) participants were lost to follow-up.

Statistical Analyses

The present study excluded individuals with self-reported, physician-diagnosed diabetes (n = 16 162), IHD (n = 15 472), or stroke or transient ischemic attack (n = 8884) at baseline and those with missing RPG data (n = 8160) (mainly recruited before formal commencement of blood glucose level testing). Within–study area comparisons of participants with and without RPG data showed no consistent, clinically significant differences. One thousand seventeen participants with missing, implausible, or extreme values for body mass index (BMI), systolic blood pressure (SBP) or diastolic blood pressure, height, waist circumference, hip circumference, or waist to hip ratio were excluded; 467 508 remained for inclusion in the analyses.

The prevalence and mean values of baseline characteristics were calculated across RPG categories, with cut points of 4.3 mmol/L (77 mg/dL), 5.3 mmol/L (95 mg/dL), 5.8 mmol/L (105 mg/dL), 6.8 mmol/L (123 mg/dL), 7.8 mmol/L (140 mg/dL) and 11.1 mmol/L (200 mg/dL), standardized by 5-year age groups, sex, and study area. The RPG cut points were chosen to include oral glucose tolerance test 2-hour postload thresholds for diabetes and impaired glucose tolerance3 and to ensure reasonable participant numbers in all groups.

Cox proportional hazards models were used to estimate hazard ratios (HRs) for the associations of baseline RPG levels with incident CVD, stratified by age at risk, sex (where appropriate), and study area and adjusted for educational level (no formal education, primary school, middle school, high school, or college/university), smoking (never, occasional, previously regular, or current regular), alcohol intake (never, occasional, previously regular, reduced, or weekly), SBP (<100, 100-109, 110-119,120-129, 130-139, 140-149, 150-159, 160-169, or ≥170 mm Hg), and physical activity (<10.0, 10.0-19.9, 20.0-29.9, 30.0-39.9, or ≥40.0 metabolic equivalent of task hours per day). Confounding variables were selected based on a priori knowledge of underlying biological mechanisms and demonstrated associations with RPG levels and CVD outcomes. The floating absolute risk method was used; this method does not alter the value of the HRs, but provides 95% CIs for all RPG measurement categories, thus enabling comparisons between any 2 categories and not only with the reference group.18 We examined discrimination of the models using the Harrell C statistic.19

Single RPG measurements may not accurately reflect an individual’s usual or longer-term average, RPG level owing to random measurement error and more systemic changes over time, resulting in regression dilution bias when assessing the associations with disease risks.20 To correct for this, data on repeated RPG levels measured at resurvey (mean of 2.6 years after the baseline survey) in 17 863 participants were used to estimate usual (mean resurvey) RPG levels for individuals in each baseline RPG category. Usual RPG levels for the lowest 3 RPG measurement categories were similar; these categories were therefore combined when investigating associations of usual RPG levels. Departure from linearity was assessed using the likelihood ratio test.21 If the shape was log-linear, baseline RPG value was also investigated as a continuous variable. As sensitivity analyses, we conducted fractional polynomial analyses of baseline RPG value that allow a continuous variable to be modeled using a nonlinear relationship.22 Examination of HRs for the first 4 and subsequent years of follow-up showed no strong evidence of departure from the proportional hazards assumption. The overall regression dilution ratio was calculated as the ratio of the range of the mean resurvey RPG levels, between top and bottom RPG measurement categories, to the range of the mean baseline RPG levels.23 Log HR estimates for baseline RPG level examined as a continuous variable were multiplied by the reciprocal of the regression dilution ratio to obtain regression dilution bias–corrected estimates.23 Adjusted HRs (aHRs) were compared across strata of other CVD risk factors and fasting time, and χ2 tests for trend and heterogeneity (ie, effect modification or statistical interaction) were applied to the log HRs and their SEs.24

Separate analyses were performed excluding individuals with a baseline plasma glucose level suggestive of diabetes,2 individuals from Zhejiang (where 72.5% reported not having consumed food for ≥8 hours before testing), or individuals diagnosed as having diabetes during follow-up as identified from diagnoses in mortality, disease surveillance, or health insurance data. Sensitivity analyses examined the association of RPG levels with all-cause mortality. All analyses used SAS (version 9.3; SAS Institute, Inc). Figures were produced using R software (version 2.13.1; https://www.r-project.org/).

Results

Among the 467 508 participants (191 555 men [41.0%] and 275 953 women [59.0%]) without known diabetes or CVD at baseline, the mean (SD) age was 50.9 (11) years (Table 1). Mean (SD) baseline RPG level was 5.9 (1.9) mmol/L (106 [34] mg/dL), slightly higher in women than men (6.0 [1.9] vs 5.8 [1.9] mmol/L [108 [34] vs 105 [34] mg/dL]). Baseline RPG level was associated positively with age, educational level, SBP, and adiposity and inversely with physical activity. We found no clear trend in fasting time across baseline RPG measurement categories.

During approximately 3.3 million person-years of follow-up (mean, 7 years), 19 214 deaths, 6645 cardiovascular deaths, and 3270 MCE, 19 153 IS, 22 023 MOVD, and 4326 ICH events were recorded. For all CVDs, the risk increased progressively with higher baseline RPG levels, with no evidence of a threshold in the association (Table 2). Multivariable-adjusted fractional polynomial models examining the association of RPG levels with cardiovascular death, MCE, IS, and MOVD consistently indicated that models using the linear form of RPG level best fitted the data (eFigure 2 in the Supplement). We found a strongly significant, positive association of baseline RPG level with cardiovascular death, MCE, IS, and MOVD (P for trend <.001), and a weaker association with ICH (P for trend = .10). The incremental changes in the Harrell C statistic for comparing the base model (ie, a Cox proportional hazards model including educational level, smoking, alcohol intake, SBP, and physical activity stratified by age at risk, sex, and study area) with the model that additionally included baseline RPG level were very modest (incremental changes of 0.0050, 0.0053, 0.0036, and 0.0033 for cardiovascular death, MCE, IS, and MOVD, respectively). The overall values of the Harrell C statistic for the multivariable-adjusted Cox proportional hazards models for cardiovascular death, MCE, IS and MOVD were 0.68, 0.63, 0.61, and 0.61, respectively.

Based on resurvey data from 17 863 randomly selected participants, we estimated the usual RPG levels in each baseline RPG measurement category. Figure 1A shows the association between usual RPG level and the risk for cardiovascular death. We found a positive, log-linear association between usual RPG measurement and cardiovascular death continuing to at least 5.9 mmol/L (106 mg/dL); each 1-mmol/L (18-mg/dL) higher usual RPG level was associated with an aHR of 1.11 (95% CI, 1.10-1.13) with application of the calculated regression dilution ratio of 0.56. The positive association appeared stronger in men than women (P = .005) and in individuals with lower SBP (P for trend = .002) or higher levels of education (P = .009) (Figure 2).

We also found a positive, log-linear association between usual RPG levels and the risk for ischemic CVD, with no evidence of a threshold (Figure 1). For MCE, each 1-mmol/L (18-mg/dL) higher usual RPG level was associated with an aHR of 1.10 (1.08-1.13), while for IS it was 1.08 (1.07-1.09). For MI and IS, the aHRs were somewhat greater for fatal than nonfatal events (aHRs for MI, 1.13 [95% CI, 1.09-1.16] vs 1.05 [95% CI, 1.01-1.10]; aHRs for IS, 1.15 [95% CI, 1.10-1.21] vs 1.08 [95% CI, 1.07-1.09]) (eTable 2 in the Supplement). For MOVD, each 1-mmol/L (18-mg/dL) higher usual RPG level was associated with an 8% (aHR, 1.08; 95% CI, 1.07-1.09) greater risk, with some suggestion of a stronger association at younger ages (P for trend = .003) (Figure 3).

The associations with ischemic CVD and cardiovascular death did not appear to differ across fasting periods (eFigure 3 in the Supplement). We found no clear difference in the strength of association per 1-SD higher nonfasting (fasting period, <8 hours; 2.0 mmol/L [36 mg/dL]) and fasting (fasting period ≥8 hours; 1.1 mmol/L [20 mg/dL]) baseline plasma glucose levels with cardiovascular death or ischemic CVDs (eFigure 4 in the Supplement).

The association of usual RPG levels with ICH was more modest, with each 1-mmol/L (18-mg/dL) higher usual RPG level associated with an aHR of 1.05 (95% CI, 1.02-1.07) (eFigure 5 in the Supplement), driven mainly by fatal ICH (aHR, 1.10; 95% CI, 1.07-1.13) rather than nonfatal ICH (aHR, 0.98; 95% CI, 0.95-1.02) (eTable 2 in the Supplement). For ICH, the association was apparently stronger with nonfasting than with fasting baseline plasma glucose levels (P for heterogeneity = .004) (eFigure 4 in the Supplement).

Additional adjustment for waist to hip ratio did not materially alter the associations of usual RPG level with CVD risk (eTable 2 in the Supplement). The associations also persisted after excluding participants diagnosed with diabetes during follow-up (n = 12 048) (eTable 3 in the Supplement) or those with a baseline plasma glucose level suggestive of diabetes (n = 13 050) (eTable 4 in the Supplement). Exclusion of individuals from Zhejiang (n = 51 656) did not materially alter risk estimates. In sensitivity analyses, the association of usual RPG level with all-cause mortality was similar to the association with cardiovascular death (aHR per 1-mmol/L [18-mg/dL] higher usual RPG, 1.11; 95% CI, 1.10-1.12).

Discussion

The present study is, to our knowledge, the largest prospective investigation in China of the association of plasma glucose levels with the risk for CVD in individuals without known diabetes and the only study to date with power to investigate the associations of RPG levels with CVD. We showed positive, log-linear associations between usual RPG levels and the risk for cardiovascular death and major ischemic CVD that continued to at least a usual RPG level of 5.9 mmol/L (106 mg/dL) and no evidence of a threshold. Each 1-mmol/L (18-mg/dL) higher usual RPG level was associated with an approximately 10% higher CVD risk.

Prospective studies of mostly Western populations have investigated the association of blood glucose levels—mainly fasting blood glucose—with CVD risks and shown a relatively consistent greater CVD risk in the prediabetes range when compared with lower blood glucose levels.1,6,7 Below this range, however, evidence is conflicting. In the Asia Pacific Cohort Studies Collaboration6 of approximately 240 000 participants from 13 cohorts, a positive, log-linear association was found between usual fasting blood glucose levels and incident IHD (n = 816) and cardiovascular death (n = 1661), continuing to at least 4.9 mmol/L (88 mg/dL). In contrast, in a study of approximately 1.2 million Koreans with approximately 60 000 IHD and more than 45 000 IS events,7 a J-shaped association with baseline fasting plasma glucose levels was found, with the lowest risks at approximately 5.0 mmol/L (90 mg/dL). In the Emerging Risk Factors Collaboration,1 including approximately 260 000 participants from 51 studies with approximately 11 000 IHD and approximately 1500 IS events, no significant association of fasting plasma glucose level with IS was found, but a J-shaped association was found with IHD, with the lowest risk at 3.9 to 5.6 mmol/L (70-101 mg/dL). No prospective studies in mainland China have reported on the association, and 2 small Taiwanese studies25,26 have produced conflicting findings.

Limited evidence is available about the association of random blood glucose levels with CVD. A published data meta-analysis,27 including 7 cohort studies, found no convincing evidence of an association of random blood glucose levels with cardiac (314 events; HR per 1-mmol/L [18-mg/dL] higher random blood glucose level, 1.02; 95% CI, 0.98-1.07), stroke (544 events; HR per 1-mmol/L [18-mg/dL] higher random blood glucose level, 1.11; 95% CI, 0.95-1.31) or cardiovascular (1782 events; HR per 1-mmol/L [18-mg/dL] higher random blood glucose level, 1.11; 95% CI, 1.00-1.24) mortality, and only weak evidence of a positive association with total CVD (2087 events; HR per per 1-mmol/L [18-mg/dL] higher random blood glucose level, 1.12; 95% CI, 1.01-1.25). Our study provides, to our knowledge, the first convincing evidence of a positive association of RPG levels with CVDs.

Fasting and postload blood glucose levels are arguably more robust glycemic measures than random blood glucose levels, which may be subject to greater interindividual and intraindividual variation. However, nonfasting glucose levels may be more relevant to CVD risks because people spend most time in a nonfasting state.9 Furthermore, we found fasting time explained only a small proportion of variation in plasma glucose levels in the CKB (<8 hours, r2 = 0.01; ≥8 hours, r2 = 0.001), with no consistent difference in associations with CVD risks across fasting time strata. In addition, use of fasting time–adjusted plasma glucose levels (eFigure 6 in the Supplement) or additional adjustment for fasting time did not materially alter risk estimates. Thus, despite recognized limitations,28 in large-scale population-based epidemiologic studies, RPG level appears to be a reliable and practical glycemic indicator.29

The large number of well-characterized stroke events (approximately 90% of validated stroke events had been confirmed on computed tomography or magnetic resonance imaging) is a strength of this study and partly reflects frequent use of computed tomography or magnetic resonance imaging scans in China. Medical record review for all stroke cases is under way; findings to date have shown a positive predictive value of approximately 90% for stroke (approximately 85% for IHD). Frequent use of scans detects a relatively high proportion of lacunar infarcts without major, or any, apparent focal neurologic deficit,30 likely contributing to the relatively low rate of IS case fatality in the CKB. Rates of stroke, particularly hemorrhagic stroke,31 are characteristically high in Chinese populations, as reflected in the CKB. Owing to lower hemorrhagic stroke rates in Western populations and limited availability of scanning technology in earlier studies, evidence of the association of plasma glucose levels with hemorrhagic stroke has been limited. The Korean Cancer Prevention Study7 included approximately 19 000 cases of hemorrhagic stroke and showed a more modest association than was seen for other CVDs, with clearly elevated risks only in the highest fasting plasma glucose measurement categories. The stronger association for fatal than nonfatal ICH, and to a lesser extent MI and IS, in the CKB may reflect a survival effect or more severe disease in fatal cases, although this has not been reported previously.6,7 The models using baseline RPG levels showed moderate ability to discriminate between participants developing and not developing CVDs. The discriminatory ability of our models appears to be somewhat lower than that reported in previous Chinese studies.32 This difference may reflect our exclusion of individuals with known diabetes and the current lack of data on lipid levels. However, our study included much larger numbers of well-characterized CVD end points, so our results are more statistically robust. This discriminatory ability is, however, of limited relevance to disease etiology, which is the main focus of our analyses.

Higher CVD risks at higher glucose levels could reflect undiagnosed or future diabetes.1,2 However, persistence of the associations after excluding individuals with plasma glucose levels suggestive of diabetes or who developed new diabetes during follow-up supports the existence of independent log-linear associations of RPG with CVD risks. Loss to follow-up in the study was low, and any resulting bias would be negligible. Residual confounding could not be excluded, especially given our current inability to adjust for lipid levels (known CVD risk factors associated with plasma glucose levels1). Lack of renal function data prevented investigation of its influence on RPG level–associated CVD risks, but would not bias risk estimates. Randomized trials of agents to lower glucose levels in prediabetes have, so far, been inconclusive in their effects on CVD risk.3336 However, evidence from mendelian randomization studies is generally compatible with a causal association between higher blood glucose levels and CVD throughout the glycemic range.9,37,38

Conclusions

The present analyses provide clear evidence of an independent, continuous relationship of RPG levels with the risk for CVD in Chinese adults without known diabetes. They support consideration of blood glucose levels as a continuous variable (rather than simply the presence or absence of diabetes39,40) in cardiovascular risk prediction models and suggest the need to consider CVD primary prevention at glucose levels below the diabetes threshold. Our findings, supported by mendelian randomization9,37,38 and some trial33 evidence, suggest that interventions to reduce plasma glucose levels may reduce CVD risk in individuals without diabetes, but further data are required.

Back to top
Article Information

Corresponding Author: Zhengming Chen, DPhil, Clinical Trial Service Unit and Epidemiological Studies Unit, Nuffield Department of Population Health, University of Oxford, Richard Doll Building, Old Road Campus, Oxford OX3 7LF, England (zhengming.chen@ctsu.ox.ac.uk).

Accepted for Publication: April 29, 2016.

Published Online: July 20, 2016. doi:10.1001/jamacardio.2016.1702

Open Access: This article is published under the JAMA Cardiology open access model and is free to read on the day of publication.

Author Contributions: Drs Bragg and Z. Chen had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Bragg, Li, Peto, Z. Chen.

Acquisition, analysis, or interpretation of data: Bragg, Li, Bennett, Lewington, Bian, Yang, J. Chen, Y. Chen, Collins, Peto, Zhu, Yin, Hu, Zhou, Pan, Z. Chen.

Drafting of the manuscript: Bragg, Z. Chen.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Bragg, Bennett, Lewington, Yang, Peto, Z. Chen.

Obtained funding: Li, Collins, Peto, Z. Chen.

Administrative, technical, or material support: Li, Guo, Bian, Yang, J. Chen, Collins, Zhu, Yin, Hu, Zhou, Pan, Y. Chen, Z. Chen.

Study supervision: Li, Bennett, Guo, Peto, Z. Chen.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: The baseline survey and the second survey were supported by the Kadoorie Charitable Foundation, Hong Kong. The long-term follow-up was supported by grants 088158/Z/09/Z and 104085/Z/14/Z from the UK Wellcome Trust, grant 2011BAI09B01 (2012-2014) from the Chinese Ministry of Science and Technology, and grant 81390541 from the Chinese National Natural Science Foundation. The British Heart Foundation, UK Medical Research Council, and Cancer Research UK provide core funding to the Clinical Trial Service Unit and Epidemiological Studies Unit, University of Oxford. This study was also supported by the British Heart Foundation Centre of Research Excellence, Oxford (Dr Bragg).

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Group Information: The China Kadoorie Biobank Collaborative Group included the following: International Steering Committee: Junshi Chen, Zhengming Chen (principal investigator [PI]), Rory Collins, Liming Li (PI), and Richard Peto; International Coordinating Centre, Oxford: Daniel Avery, Derrick Bennett, Yumei Chang, Yiping Chen, Zhengming Chen, Robert Clarke, Huaidong Du, Xuejuan Fan, Simon Gilbert, Alex Hacker, Michael Holmes, Andri Iona, Christiana Kartsonaki, Rene Kerosi, Ling Kong, Om Kurmi, Garry Lancaster, Sarah Lewington, John McDonnell, Winnie Mei, Iona Millwood, Qunhua Nie, Jayakrishnan Radhakrishnan, Sajjad Rafiq, Paul Ryder, Sam Sansome, Dan Schmidt, Paul Sherliker, Rajani Sohoni, Iain Turnbull, Robin Walters, Jenny Wang, Lin Wang, Ling Yang, and Xiaoming Yang; National Coordinating Centre, Beijing: Zheng Bian, Ge Chen, Yu Guo, Bingyang Han, Can Hou, Jun Lv, Pei Pei, Shuzhen Qu, Yunlong Tan, Canqing Yu, and Huiyan Zhou; Qingdao Regional Coordinating Centre: Zengchang Pang, Ruqin Gao, Shaojie Wang, Yongmei Liu, Ranran Du, Yajing Zang, Liang Cheng, Xiaocao Tian, and Hua Zhang (Qingdao Centre for Disease Control and Prevention [CDC]); Silu Lv, Junzheng Wang, and Wei Hou (Licang CDC); Heilongjiang Regional Coordinating Centre: Jiyuan Yin, Ge Jiang, Shumei Liu, Zhigang Pang, and Xue Zhou (Provincial CDC); Liqiu Yang, Hui He, Bo Yu, Yanjie Li, Huaiyi Mu, Qinai Xu, Meiling Dou, and Jiaojiao Ren (Nangang CDC); Hainan Regional Coordinating Centre: Jianwei Du, Shanqing Wang, Ximin Hu, Hongmei Wang, Jinyan Chen, Yan Fu, Zhenwang Fu, Xiaohuan Wang, and Hua Dong (Provincial CDC); Min Weng, Xiangyang Zheng, Yijun Li, Huimei Li, and Chenglong Li (Meilan CDC); Jiangsu Regional Coordinating Centre: Ming Wu, Jinyi Zhou, Ran Tao, and Jie Yang (Provincial CDC); Jie Shen, Yihe Hu, Yan Lu, Yan Gao, Liangcai Ma, Renxian Zhou, Aiyu Tang, Shuo Zhang, and Jianrong Jin (Suzhou CDC); Guangxi Regional Coordinating Centre: Zhenzhu Tang, Naying Chen, and Ying Huang (Provincial CDC); Mingqiang Li, Jinhuai Meng, Rong Pan, Qilian Jiang, Jingxin Qing, Weiyuan Zhang, Yun Liu, Liuping Wei, Liyuan Zhou, Ningyu Chen, Jun Yang, and Hairong Guan (Liuzhou CDC); Sichuan Regional Coordinating Centre: Xianping Wu, Ningmei Zhang, Xiaofang Chen, and Xuefeng Tang (Provincial CDC); Guojin Luo, Jianguo Li, Xiaofang Chen, Jian Wang, Jiaqiu Liu, and Qiang Sun (Pengzhou CDC); Gansu Regional Coordinating Centre: Pengfei Ge, Xiaolan Ren, and Caixia Dong (Provincial CDC); Hui Zhang, Enke Mao, Xiaoping Wang, and Tao Wang (Maiji CDC); Henan Regional Coordinating Centre: Guohua Liu, Baoyu Zhu, Gang Zhou, Shixian Feng, Liang Chang, and Lei Fan (Provincial CDC); Yulian Gao, Tianyou He, Li Jiang, Huarong Sun, Pan He, Chen Hu, Qiannan Lv, and Xukui Zhang (Huixian CDC); Zhejiang Regional Coordinating Centre: Min Yu, Ruying Hu, Le Fang, and Hao Wang (Provincial CDC); Yijian Qian, Chunmei Wang, Kaixue Xie, Lingli Chen, Yaxing Pan, and Dongxia Pan (Tongxiang CDC); Hunan Regional Coordinating Centre: Yuelong Huang, Biyun Chen, Donghui Jin, Huilin Liu, Zhongxi Fu, and Qiaohua Xu (Provincial CDC); Xin Xu, Youping Xiong, Weifang Jia, Xianzhi Li, Libo Zhang, and Zhe Qiu (Liuyang CDC).

Additional Contributions: We thank the participants, the project staff, and the China National CDC and its regional offices for access to death and disease registries. The Chinese National Health Insurance scheme provided electronic linkage to all hospital treatment.

References
1.
Sarwar  N, Gao  P, Seshasai  SR,  et al; Emerging Risk Factors Collaboration.  Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies [published correction appears in Lancet. 2010;376(9745):958].  Lancet. 2010;375(9733):2215-2222.PubMedArticle
2.
Bragg  F, Li  L, Smith  M,  et al; China Kadoorie Biobank Collaborative Group.  Associations of blood glucose and prevalent diabetes with risk of cardiovascular disease in 500 000 adult Chinese: the China Kadoorie Biobank.  Diabet Med. 2014;31(5):540-551.PubMedArticle
3.
World Health Organization, International Diabetes Federation.  Definition and Diagnosis of Diabetes Mellitus and Intermediate Hyperglycaemia: Report of a WHO/IDF Consultation. Geneva, Switzerland: World Health Organization; 2006.
4.
Ford  ES, Zhao  G, Li  C.  Pre-diabetes and the risk for cardiovascular disease: a systematic review of the evidence.  J Am Coll Cardiol. 2010;55(13):1310-1317.PubMedArticle
5.
Gerstein  H, Punthakee  Z. Dysglycemia and the risk of cardiovascular events. In: Yusuf  S, Cairns  JA, Camm  AJ, Fallen  EL, Gersh  BJ, eds.  Evidence-Based Cardiology. 3rd ed. Chichester: Wiley-Blackwell; 2010:179-189.
6.
Lawes  CMM, Parag  V, Bennett  DA,  et al; Asia Pacific Cohort Studies Collaboration.  Blood glucose and risk of cardiovascular disease in the Asia Pacific region.  Diabetes Care. 2004;27(12):2836-2842.PubMedArticle
7.
Park  C, Guallar  E, Linton  JA,  et al.  Fasting glucose level and the risk of incident atherosclerotic cardiovascular diseases.  Diabetes Care. 2013;36(7):1988-1993.PubMedArticle
8.
Singh  GM, Danaei  G, Farzadfar  F,  et al; Global Burden of Metabolic Risk Factors of Chronic Diseases Collaborating Group; Asia-Pacific Cohort Studies Collaboration (APCSC); Diabetes Epidemiology: Collaborative analysis of Diagnostic criteria in Europe (DECODE); Emerging Risk Factor Collaboration (ERFC); Prospective Studies Collaboration (PSC).  The age-specific quantitative effects of metabolic risk factors on cardiovascular diseases and diabetes: a pooled analysis.  PLoS One. 2013;8(7):e65174.PubMedArticle
9.
Benn  M, Tybjaerg-Hansen  A, McCarthy  MI, Jensen  GB, Grande  P, Nordestgaard  BG.  Nonfasting glucose, ischemic heart disease, and myocardial infarction: a Mendelian randomization study.  J Am Coll Cardiol. 2012;59(25):2356-2365.PubMedArticle
10.
Chan  JC, Zhang  Y, Ning  G.  Diabetes in China: a societal solution for a personal challenge.  Lancet Diabetes Endocrinol. 2014;2(12):969-979.PubMedArticle
11.
Xu  Y, Wang  L, He  J,  et al; 2010 China Noncommunicable Disease Surveillance Group.  Prevalence and control of diabetes in Chinese adults.  JAMA. 2013;310(9):948-959.PubMedArticle
12.
Li  H, Ge  J.  Cardiovascular diseases in China: current status and future perspectives.  Int J Cardiol Heart Vasc.2015;6(0):25-31.Article
13.
Chen  Z, Chen  J, Collins  R,  et al; China Kadoorie Biobank (CKB) collaborative group.  China Kadoorie Biobank of 0.5 million people: survey methods, baseline characteristics and long-term follow-up.  Int J Epidemiol. 2011;40(6):1652-1666.PubMedArticle
14.
Chen  Z, Lee  L, Chen  J,  et al.  Cohort profile: the Kadoorie Study of Chronic Disease in China (KSCDC).  Int J Epidemiol. 2005;34(6):1243-1249.PubMedArticle
15.
Allen  N, Sudlow  C, Downey  P,  et al.  UK Biobank: current status and what it means for epidemiology.  Health Policy Technol. 2012;1(3):123-126.Article
16.
Lifescan. SureStep Technology. http://www.cliawaived.com/web/items/pdf/LifeScan_10797_Diabetes_Test~2251file1.pdf. Accessed February 7, 2016.
17.
World Health Organization. International Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10). http://apps.who.int/classifications/icd10/browse/2010/en. Accessed January 15, 2014.
18.
Easton  DF, Peto  J, Babiker  AG.  Floating absolute risk: an alternative to relative risk in survival and case-control analysis avoiding an arbitrary reference group.  Stat Med. 1991;10(7):1025-1035.PubMedArticle
19.
Harrell  FE  Jr.  Regression Modeling Strategies With Applications to Linear Models, Logistic Regression and Survival Analysis. New York: Springer-Verlag; 2001.
20.
Clarke  R, Shipley  M, Lewington  S,  et al.  Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies.  Am J Epidemiol. 1999;150(4):341-353.PubMedArticle
21.
Woodward  M.  Epidemiology Study Design and Data Analysis. 3rd ed. Boca Raton: CRC Press; 2014.
22.
Royston  P, Altman  DG.  Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling.  J R Stat Soc Ser C Appl Stat. 1994;43(3):429-467.
23.
MacMahon  S, Peto  R, Cutler  J,  et al.  Blood pressure, stroke, and coronary heart disease, part 1: prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias.  Lancet. 1990;335(8692):765-774.PubMedArticle
24.
Early Breast Cancer Trialists’ Collaborative Group.  Arithmetic Details of Tests for Trend, for Heterogeneity and for Interaction: Treatment of Early Breast Cancer. Worldwide Evidence 1985-1990: A Systematic Overview of All Available Randomized Trials of Adjuvant Endocrine and Cytotoxic Therapy. Vol 1. Oxford: Oxford University Press; 1990.
25.
Chien  K-L, Hsu  H-C, Su  T-C, Chen  M-F, Lee  Y-T, Hu  FB.  Fasting and postchallenge hyperglycemia and risk of cardiovascular disease in Chinese: the Chin-Shan Community Cardiovascular Cohort study.  Am Heart J. 2008;156(5):996-1002.PubMedArticle
26.
Chien  K-L, Lee  B-C, Lin  H-J, Hsu  H-C, Chen  M-F.  Association of fasting and post-prandial hyperglycemia on the risk of cardiovascular and all-cause death among non-diabetic Chinese.  Diabetes Res Clin Pract. 2009;83(2):e47-e50.PubMedArticle
27.
Einarson  TR, Machado  M, Henk Hemels  ME.  Blood glucose and subsequent cardiovascular disease: update of a meta-analysis.  Curr Med Res Opin. 2011;27(11):2155-2163.PubMedArticle
28.
Kawada  T.  Random blood glucose measurement for epidemiological studies: its significance and limitations.  Curr Med Res Opin. 2012;28(3):447-448.PubMedArticle
29.
Moebus  S, Göres  L, Lösch  C, Jöckel  KH.  Impact of time since last caloric intake on blood glucose levels.  Eur J Epidemiol. 2011;26(9):719-728.PubMedArticle
30.
Wu  B, Lin  S, Hao  Z,  et al.  Proportion, risk factors and outcome of lacunar infarction: a hospital-based study in a Chinese population.  Cerebrovasc Dis. 2010;29(2):181-187.PubMedArticle
31.
Liu  M, Wu  B, Wang  W-Z, Lee  L-M, Zhang  S-H, Kong  L-Z.  Stroke in China: epidemiology, prevention, and management strategies.  Lancet Neurol. 2007;6(5):456-464.PubMedArticle
32.
Liu  J, Hong  Y, D’Agostino  RB  Sr,  et al.  Predictive value for the Chinese population of the Framingham CHD risk assessment tool compared with the Chinese Multi-Provincial Cohort Study.  JAMA. 2004;291(21):2591-2599.PubMedArticle
33.
Chiasson  J-L, Josse  RG, Gomis  R, Hanefeld  M, Karasik  A, Laakso  M; STOP-NIDDM Trial Research Group.  Acarbose treatment and the risk of cardiovascular disease and hypertension in patients with impaired glucose tolerance: the STOP-NIDDM trial.  JAMA. 2003;290(4):486-494.PubMedArticle
34.
Gerstein  HC, Yusuf  S, Bosch  J,  et al; DREAM (Diabetes REduction Assessment with ramipril and rosiglitazone Medication) Trial Investigators.  Effect of rosiglitazone on the frequency of diabetes in patients with impaired glucose tolerance or impaired fasting glucose: a randomised controlled trial.  Lancet. 2006;368(9541):1096-1105.PubMedArticle
35.
Gerstein  HC, Bosch  J, Dagenais  GR,  et al; ORIGIN Trial Investigators.  Basal insulin and cardiovascular and other outcomes in dysglycemia.  N Engl J Med. 2012;367(4):319-328.PubMedArticle
36.
Holman  RR, Haffner  SM, McMurray  JJ,  et al; NAVIGATOR Study Group.  Effect of nateglinide on the incidence of diabetes and cardiovascular events.  N Engl J Med. 2010;362(16):1463-1476.PubMedArticle
37.
Rasmussen-Torvik  LJ, Li  M, Kao  WH,  et al.  Association of a fasting glucose genetic risk score with subclinical atherosclerosis: the Atherosclerosis Risk in Communities (ARIC) study.  Diabetes. 2011;60(1):331-335.PubMedArticle
38.
Ahmad  OS, Morris  JA, Mujammami  M,  et al.  A Mendelian randomization study of the effect of type-2 diabetes on coronary heart disease.  Nat Commun. 2015;6:7060.PubMedArticle
39.
Anderson  KM, Wilson  PW, Odell  PM, Kannel  WB.  An updated coronary risk profile: a statement for health professionals.  Circulation. 1991;83(1):356-362.PubMedArticle
40.
Hippisley-Cox  J, Coupland  C, Vinogradova  Y,  et al.  Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2.  BMJ. 2008;336(7659):1475-1482.PubMedArticle
×