Introduction

Worldwide over 400 million people have diabetes, and the prevalence is increasing rapidly in both developed and developing countries [1]. Previous studies of mostly Western populations have shown that diabetes is typically associated with a 2-fold increased risk of ischaemic heart disease (IHD) [2,3]. Uncertainty remains, however, about whether similar excess risk applies to other populations, and about the strength of the association of diabetes with stroke and, particularly, stroke subtypes [2–4]. Appropriate understanding of these issues is of considerable relevance to China, where stroke rates are high [5].

Since the 1980s, there has been a rapid and substantial increase in the prevalence of diabetes in China [1,6,7], which now affects ~10% of adults [8]. Compared with those in Western populations, individuals with diabetes in China have tended to be leaner and to have worse pancreatic beta-cell function, which may result in greater susceptibility to microvascular complications and cancer than to macrovascular complications [9]. Despite the growing diabetes epidemic, there is limited evidence about the association of diabetes with cardiovascular disease in China [10–14], with previous studies limited by potentially outdated risk estimates [12], highly selected study populations [11], relatively small size [10–12], and lack of proper investigation of the relevance of diabetes duration and other modifiable factors, such as smoking, adiposity, blood pressure, and physical activity, which frequently differ between China and Western populations.

We report relevant findings from a large prospective study, the China Kadoorie Biobank (CKB), of 0.5 million adults, recruited between 25 June 2004 and 15 July 2008 across ten diverse areas of China. The aim of the present study is to examine the association of diabetes—both self-reported and screen-detected—with the risks of major cardiovascular diseases, including IHD and stroke subtypes. By applying population-attributable fractions for cardiovascular mortality from the present study to national cause-specific mortality and representative diabetes prevalence data, we also estimate the number of cardiovascular deaths attributable to diabetes in mainland China.

Methods

Results

Among 488,760 participants without prior cardiovascular disease, the mean baseline age was 51 (standard deviation 11) y, 2.7% (n = 13,284) reported a history of doctor-diagnosed diabetes (self-reported diabetes), and a further 2.7% (n = 13,051) had screen-detected diabetes, of whom 8,554 (65.5%) were identified through RPG measurement (RPG ≥ 7.0 mmol/l and time since last food ≥ 8 h: n = 3,986; RPG ≥ 11.1 mmol/l and time since last food < 8 h: n = 4,568), 1,447 (11.1%) through fasting plasma glucose measurement, and 3,050 (23.4%) through both. For both self-reported and screen-detected diabetes, the prevalence was slightly higher in women than in men (2.9% versus 2.5% for self-reported diabetes and 2.8% versus 2.6% for screen-detected diabetes), mainly at ages > 50 y (S2 Fig), and the prevalence was higher in urban than in rural areas (3.9% versus 1.8% for self-reported diabetes and 3.4% versus 2.1% for screen-detected diabetes). Participants reporting a history of diabetes were older and less likely to be current smokers or alcohol drinkers, but more likely to be ex-smokers or ex-drinkers and to be less physically active than those without diabetes, and had an approximately 4-fold greater prevalence of a family history of diabetes (Table 1). Those with screen-detected diabetes were older than and about twice as likely to have a family history of diabetes as those without diabetes, but had comparable smoking and alcohol consumption patterns. The prevalence of overweight or obesity (BMI ≥ 25.0 kg/m²) and hypertension (SBP ≥ 140 mm Hg) were higher in participants with diabetes, particularly screen-detected diabetes. Among participants with self-reported diabetes, the median age at diagnosis was 53 (range 4 to 77) y, and the median duration of diabetes at baseline was 4 (range 0 to 56) y, with 76% (n = 8,501) reporting use of insulin or oral hypoglycaemic agents.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of participants by diabetes status.

https://doi.org/10.1371/journal.pmed.1002026.t001

Self-Reported Diabetes and Cardiovascular Disease Risk

During ~3.4 million person-years (mean 7 y per person) of follow-up, 7,353 (1.5%) participants died from cardiovascular disease. Individuals with self-reported diabetes had a 2-fold increased hazard of cardiovascular mortality (HR 2.07, 95% CI 1.90–2.26) (Table 2), and the HR for cardiovascular mortality was greater in women than in men (2.37, 95% CI 2.11–2.65, versus 1.73, 95% CI 1.50–2.00) (S3 Fig). Likewise, there were significantly increased risks of MOVD (1.77, 1.69–1.86), MCE (2.44, 2.18–2.73), and IS (1.68, 1.60–1.77) among those with self-reported diabetes (Table 2). For IHD and IS, the HRs were somewhat greater for fatal than non-fatal events. For IS and MOVD (S4 Fig), the associations were stronger at younger ages (both p for trend < 0.001) and in rural areas (p for heterogeneity = 0.003 and < 0.001, respectively). The risk of MCE varied little across population subgroups.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Adjusted hazard ratios for incident cardiovascular diseases by diabetes status.

https://doi.org/10.1371/journal.pmed.1002026.t002

Self-reported diabetes was also associated with significant, but more modest, excess risk of ICH, with an adjusted HR of 1.24 (95% CI 1.07–1.44), more extreme for fatal (1.59, 95% CI 1.30–1.96) than for non-fatal (1.00, 95% CI 0.81–1.23) events (Table 2). The risk varied little across participant subgroups (S5 Fig). For total stroke, self-reported diabetes was associated with an adjusted HR of 1.61 (95% CI 1.54–1.69; Table 2).

Additional adjustment for waist-to-hip ratio modestly, but non-significantly, attenuated the HRs for ischaemic cardiovascular diseases, but not for ICH or cardiovascular mortality (S1 Table). In sensitivity analyses excluding participants who developed incident diabetes during follow-up (n = 8,896), HRs for all cardiovascular diseases remained largely unchanged (S2 Table). Sensitivity analyses adjusting for age, SBP, and physical activity as continuous variables also produced similar findings.

Screen-Detected Diabetes and Cardiovascular Disease Risk

Among participants with screen-detected diabetes, the excess risks were also highly significant, though somewhat less extreme compared with self-reported diabetes. For cardiovascular mortality, the adjusted HR was 1.66 (95% CI 1.51–1.83), while for MOVD, MCE, and IS the adjusted HRs were 1.50 (95% CI 1.41–1.58), 1.62 (95% CI 1.40–1.86), and 1.48 (95% CI 1.40–1.57), respectively. Similarly, for ICH the excess risk was more modest (1.17, 95% CI 1.01–1.36) (Table 2). There was little evidence of heterogeneity in the observed HRs for MCE, ICH (S6 Fig), or cardiovascular mortality (S7 Fig) across population subgroups. For IS and MOVD (S8 Fig), however, the risk was somewhat greater at younger ages (both p for trend < 0.001), in rural residents (p for heterogeneity < 0.001 and 0.003, respectively), and in more physically active individuals (p for trend 0.008 and 0.006, respectively).

Among individuals with diabetes, the risk of major ischaemic cardiovascular events increased progressively with longer duration of diabetes (p for trend < 0.001; Fig 1), but no such trend was seen for ICH (p = 0.3). Similarly, among individuals with self-reported diabetes, the risk of MOVD increased progressively with the number of other presenting cardiovascular risk factors, with adjusted HRs of 1.68 (95% CI 1.53–1.85), 2.08 (95% CI 1.93–2.23), and 2.82 (95% CI 2.59–3.06) in those with one, two, and three or more risk factors, respectively, compared with participants with self-reported diabetes who had no other cardiovascular risk factors (p for trend < 0.001; Fig 2). A similar gradient was seen for screen-detected diabetes (p for trend < 0.001; Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Adjusted hazard ratios for cardiovascular diseases by duration of diabetes diagnosis.

(A) Cardiovascular disease mortality; (B) MOVD; (C) MCE; (D) IS. Stratified by age, sex, and study area and adjusted for education, smoking, alcohol consumption, physical activity, and SBP. Diabetes duration defined as 0 y in individuals with screen-detected diabetes. Diabetes duration data missing or implausible for 17 participants. Squares represent the HR, with area inversely proportional to the variance of the log HR. Vertical lines represent the corresponding 95% confidence intervals. Numbers above the CIs are HRs, and numbers below the CIs are the number of participants with cardiovascular events.

https://doi.org/10.1371/journal.pmed.1002026.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Adjusted hazard ratios for major occlusive vascular disease by number of cardiovascular risk factors at baseline among individuals with diabetes.

Stratified by age, sex, and study area and adjusted for education and alcohol consumption. Cardiovascular risk factors: hypertension (self-reported hypertension, mean SBP ≥ 140 mm Hg or diastolic blood pressure ≥ 90 mm Hg), overweight or obesity (BMI ≥ 25 kg/m²), ever regular smoking, and physical inactivity (<10 metabolic equivalent of task hours/day). Squares represent the HR, with area inversely proportional to the variance of the log HR. Horizontal lines represent the corresponding 95% confidence intervals.

https://doi.org/10.1371/journal.pmed.1002026.g002

Discussion

Our study provides, to our knowledge, the first large-scale prospective evidence of the cardiovascular consequences of diabetes among adults in China. It showed that self-reported doctor-diagnosed diabetes was associated with 1.5- to 2.5-fold increased risks of cardiovascular mortality and incident IHD and IS. Moreover, it provided strong prospective evidence of a significant, though more modest, adverse association of diabetes with ICH risk. Individuals with screen-detected diabetes also had significantly increased cardiovascular disease risks. If these associations are causal, then almost 0.5 million cardiovascular deaths a year in China can now be attributed to diabetes.

Previous studies of predominantly Western populations have shown approximate doubling of IHD risk with diabetes [2,3,10,11,23]. The comparability of our risk estimates with previous studies suggests that differences between East Asian and Western populations in the relative importance of insulin resistance and beta-cell dysfunction in the aetiology of diabetes [9] have little, if any, impact on associated IHD risk. The method of ascertainment of IHD in the CKB may misclassify a proportion of individuals with “silent” IHD, which is more prevalent among individuals with diabetes [24], leading to underestimation of diabetes-associated IHD risk. The more modest effects of screen-detected—compared to self-reported—diabetes may reflect the shorter duration of, or less severe glycaemic aberration in, screen-detected diabetes, and greater potential for misclassification.

The CKB provides reliable evidence about the association of diabetes with stroke and, in particular, stroke subtypes. The association of screen-detected diabetes with IS in our study is largely consistent with data from studies of mostly Western populations [3]; the association with self-reported diabetes was moderately weaker than in those studies [2,3] and also weaker than in a small prospective Chinese study of 30,000 men and women [10] and a study—based on routine data and adjusted only for age—in Taiwan [25]. In mainland China, computed tomography or MRI is widely used, often leading to detection of a relatively high proportion of lacunar infarcts without major, or any, apparent focal neurological deficit [26]. The relatively low IS case fatality rate in our study might explain, at least in part, the more modest association with diabetes.

Existing evidence relating ICH to diabetes is more limited [2,3,10,25,27], partly due to lower ICH rates in Western populations [2,3] and the lack of widespread imaging in earlier studies. Where available, findings have been inconsistent [2–4,10,25]. Two of the largest studies in Western populations, including 1,000 [3] and 2,500 [4] ICH events, found modest elevated risks associated with known diabetes (HR 1.56 [3] and 1.28 [4]). Our study included almost 5,000 well-characterised ICH cases and provides the most reliable evidence to date of a positive, but modest, association between self-reported diabetes and ICH, and the first opportunity, to our knowledge, to investigate effect modification beyond age and sex. The large number of well-characterised stroke events (~90% of validated stroke events were confirmed with computed tomography/MRI) is a strength of the CKB. Outcome adjudication, through review of medical records for all IHD and stroke cases, is currently underway; findings to date have shown high positive predictive values for IHD events (~85%) and for stroke (~90%). The apparent difference in excess risk between fatal and non-fatal cardiovascular disease events could reflect poorer survival following such events in individuals with diabetes [28,29] or more severe disease in fatal cases.

The large urban—rural differences in excess risk associated with self-reported diabetes in the CKB, especially for IS, may reflect the higher proportion of undiagnosed diabetes cases in rural areas, such that self-reported diabetes in rural areas may represent more severe disease. More frequent use of diabetes medications in rural areas supports this hypothesis. Several studies have previously reported greater diabetes-associated IHD risk [3,23] and, less consistently, stroke risk [3,4,30] in women. We found a sex difference for cardiovascular mortality only, and this may have contributed to the higher diabetes-associated risks in never or occasional smokers and never or occasional alcohol drinkers. Diabetes, particularly type 2 diabetes, frequently coexists with other cardiovascular risk factors [31]. In our study population, the cardiovascular disease risk among those with diabetes was also dependent on the presence of other common and modifiable cardiovascular risk factors, which highlights the need for multi-factorial approaches to managing cardiovascular disease risk, even after development of diabetes. The small proportion of participants with self-reported diabetes who reported use of medications for lowering cardiovascular disease risk suggests that there is considerable scope for improvement in this regard in China. The impact of diabetes diagnosis duration on disease risk is largely consistent with the almost 2-fold greater odds of prevalent cardiovascular disease in early-onset (<40 y of age), compared to late-onset (≥40 y), diabetes in China, with the majority of the association explained by diabetes duration [32].

The CKB was not designed to be nationally representative, but given the size and diversity of the CKB study population and the CKB’s minimal loss to follow-up, this would not be expected to bias risk estimates or reduce their generalisability to the Chinese adult population [33]. Furthermore, the prevalence of diabetes in the CKB was similar to that reported in a reasonably contemporaneous representative Chinese survey in 2000–2001 [7]. More recent surveys have reported a higher rate of diabetes in China [7], with, for example, a prevalence rate of 11.6% in the 2010 survey [8]. As well as reflecting secular trends, the lower prevalence of diabetes in the CKB may reflect different study settings, sampling methods, and approaches used to identify undiagnosed diabetes. The prevalence of self-reported diabetes was, however, similar (3.5% in the 2010 survey versus 2.7% in the present CKB population and 3.2% in the CKB population including individuals with prior cardiovascular disease). Although self-reported diabetes is widely used [33], it may be subject to misclassification. However, amongst almost 30,000 participants resurveyed in the CKB, over 90% of those who reported a history of diabetes at baseline gave the same answer at resurvey. Furthermore, the mean plasma glucose level among those with self-reported diabetes was consistent with the diagnosis. We did not collect specific information on diabetes type, but, based on age at diagnosis and medication use, <1% of diabetes in the CKB would likely be type 1.

Separate examination of screen-detected diabetes should reduce bias from lifestyle changes and treatment following diagnosis, although screen-detected diabetes may be subject to greater misclassification resulting from use of RPG [34] measurement on a glucometer [35] for diagnosis. Such misclassification likely contributed to the lower screen-detected diabetes prevalence in the CKB compared to the most recent national survey [8] (2.7% in CKB versus 8.1% in the 2010 survey), which used multiple glycaemic indicators (oral glucose tolerance testing, fasting plasma glucose, and HbA1c), and would result in underestimates of diabetes-associated risks. However, excluding participants who developed new diabetes during follow-up (identified from mortality, disease surveillance, and national health insurance data) did not materially alter risk estimates. The high proportion of undiagnosed diabetes [8] and the elevated cardiovascular risk associated with it lend support to population screening for diabetes to enable more effective primary prevention of cardiovascular diseases. Data on lipids are not currently available; this may have resulted in residual confounding, although progressive adjustment for confounders in one published study suggests that this confounding would be minimal [3]. Similarly, renal function data are not currently available in the CKB; this precludes investigation of effect modification of the association of diabetes with cardiovascular diseases but would not bias our risk estimates. Additional adjustment of presented risk estimates for other potential confounding variables, including dietary intake, had minimal effect.

In China, about 10% of the adult population is estimated to have diabetes [8]. The present nationwide prospective study provides large-scale evidence from mainland China that individuals with diabetes are at significantly increased risk of major cardiovascular diseases, similar in magnitude to that observed in Western populations [2,3]. Much of the diabetes-associated excess risk is likely to be causal, accounting for almost 0.5 million cardiovascular deaths annually in China. With further adverse changes in lifestyle, the prevalence of diabetes will likely increase further in China, especially among young adults [1,7], foreshadowing an even greater diabetes-attributable burden of cardiovascular (and other) diseases.

Supporting Information

Acknowledgments

The chief acknowledgment is to the participants, the project staff, and the China CDC and its regional offices for access to death and disease registries. The Chinese national health insurance system provided electronic linkage to all hospital treatment.

The members of the China Kadoorie Biobank collaborative group are as follows. International Steering Committee: Junshi Chen, Zhengming Chen (Principal Investigator), Rory Collins, Liming Li (Principal Investigator), Richard Peto. International Co-ordinating Centre, Oxford: Daniel Avery, Derrick Bennett, Yumei Chang, Yiping Chen, Zhengming Chen, Robert Clarke, Huaidong Du, Xuejuan Fan, Haiyan Gao, 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, Paul Ryder, Sam Sansome, Dan Schmidt, Paul Sherliker, Rajani Sohoni, Robin Walters, Jenny Wang, Lin Wang, Ling Yang, Xiaoming Yang. National Co-ordinating Centre, Beijing: Zheng Bian, Ge Chen, Lei Guo, Yu Guo, Bingyang Han, Can Hou, Peng Liu, Jun Lv, Pei Pei, Shuzhen Qu, Yunlong Tan, Canqing Yu, Huiyan Zhou. Regional Co-ordinating Centres: Qingdao CDC: Zengchang Pang, Shaojie Wang, Yun Zhang, Kui Zhang. Licang CDC: Silu Liu, Wei Hou. Heilongjiang Provincial CDC: Zhonghou Zhao, Shumei Liu, Zhigang Pang. Nangang CDC: Weijia Feng, Shuling Wu, Liqiu Yang, Huili Han, Hui He, Bo Yu. Hainan Provincial CDC: Xianhai Pan, Shanqing Wang, Hongmei Wang. Meilan CDC: Xinhua Hao, Chunxing Chen, Shuxiong Lin, Xiangyang Zheng. Jiangsu Provincial CDC: Xiaoshu Hu, Minghao Zhou, Ming Wu, Ran Tao. Suzhou CDC: Yeyuan Wang, Yihe Hu, Liangcai Ma, Renxian Zhou, Guanqun Xu, Yan Lu. Guangxi Provincial CDC: Baiqing Dong, Naying Chen, Ying Huang. Liuzhou CDC: Mingqiang Li, Jinhuai Meng, Zhigao Gan, Jiujiu Xu, Yun Liu, Jingxin Qing. Sichuan Provincial CDC: Xianping Wu, Yali Gao, Ningmei Zhang. Pengzhou CDC: Guojin Luo, Xiangsan Que, Xiaofang Chen. Gansu Provincial CDC: Pengfei Ge, Jian He, Xiaolan Ren. Maiji CDC: Hui Zhang, Enke Mao, Guanzhong Li, Zhongxiao Li, Jun He, Yulong Lei, Xiaoping Wang. Henan Provincial CDC: Guohua Liu, Baoyu Zhu, Gang Zhou, Shixian Feng. Huixian CDC: Yulian Gao, Tianyou He, Li Jiang, Jianhua Qin, Huarong Sun. Zhejiang Provincial CDC: Liqun Liu, Min Yu, Yaping Chen, Ruying Hu. Tongxiang CDC: Zhixiang Hu, Jianjin Hu, Yijian Qian, Zhiying Wu, Chunmei Wang, Lingli Chen. Hunan Provincial CDC: Wen Liu, Guangchun Li, Huilin Liu. Liuyang CDC: Xiangquan Long, Xin Xu, Youping Xiong, Zhongwen Tan, Xuqiu Xie, Yunfang Peng, Weifang Jia.