Methods

The methods used in the global INTERHEART study have been described previously.15 16 Patients admitted to hospital with first incident AMI were enrolled within 24 h of onset of symptoms. At least one age- (±5 years) and sex-matched control (without a history of CVD) was recruited for each case from non-cardiac wards or unrelated visitors of cardiac patients, or patients at the same centres with illnesses not obviously related to CVD or its risk factors.

Trained personnel administered the structured questionnaires and physical examinations in a standardised manner. Information about demographic factors, income and education, lifestyle, personal and family history of CVD and risk factors, psychosocial factors, physical activities, nutritional and smoking history, alcohol use were obtained. Blood pressure, height, weight, waist and hip circumferences were measured. Non-fasting blood samples (20 ml) were drawn and centrifuged within 2 h of admission, frozen immediately after processing. Samples were shipped in nitrogen vapour tanks from every site to a blood storage site and stored at −70°C in freezers or −170°C (Chinese samples were stored in Beijing but, samples from Hong Kong in Hamilton, Canada) in liquid nitrogen. Samples were analysed for total cholesterol, high-density lipoprotein (HDL) cholesterol, and apolipoproteins B (ApoB) and A-1 (ApoA-1). Analytic techniques, described in detail previously,17 were closely coordinated, standardised and calibrated between the laboratories in Beijing and Hamilton.

The Beijing Hypertension League Institute and Fu Wai Hospital, Beijing, coordinated the study. The Population Health Research Institute, Hamilton, Canada, received, managed and statistically analysed all the data, and coordinated the overall study.

Statistical analysis

As described in previously published reports,16 univariate associations were explored with frequency tables and the Pearson χ² test for independent proportions. Continuous variables, summarised by means or medians, were compared with t tests or appropriate non-parametric tests. For comparison of means across subgroups, values were adjusted for age and sex with analysis of covariance models. Sex-specific quintile values in controls were used to categorise continuous variables. Unconditional logistic regression with adjustment for matching factors was used to control for confounding by other risk factors.17

Population-attributable risks (PARs) and their 95% confidence intervals (CIs) were calculated based on unconditional logistic regression using the methods of Benichou and Gail, and with the Interactive Risk Attributable Program (US National Cancer Institute, 2002),18 19 and adjusted for confounders as the logistic regression models for odds ratios. PAR  =  (Pr(E) (R−1))/(Pr(E) (R−1) +1), where Pr(E) is probability of exposure to the risk factors and R is the relative risk of the disease in exposed versus unexposed subjects. The PAR of any given risk factor for MI is an estimate of the portion of MI that is explained by the risk factor in the specified population.16

Results

Participants, 3030 cases and 3056 controls, were recruited from 26 regional hospitals situated in urban areas and equipped to admit patients with MI in the region, including Hong Kong. A total of 5135 were recruited from 19 northern Chinese centres and 951 from seven southern Chinese centres (fig 1). Table 1 shows risk factor profiles of cases and controls in China versus other countries. Among controls, fewer participants from China had diabetes, were smokers (current and former), were depressed or experienced permanent global stress (all p<0.001). Similarly, lower levels of risk factors were seen among Chinese participants for body mass index (BMI), waist to hip ratios, total cholesterol, low-density lipoprotein (LDL) cholesterol, ApoB levels and lower ApoB/ApoA-1 ratios, and higher ApoA-1 and HDL cholesterol levels. However, Chinese participants had similar rates of self-reported hypertension and fruit and vegetable intake compared with those seen in the other countries and fewer Chinese reported regular physical activity (table 1). The standardised distributions for BMI, waist to hip ratio, total cholesterol, LDL cholesterol and ApoB/ApoA-1 ratios were all lower in the Chinese population, while HDL cholesterol levels were higher, than corresponding distributions in other countries (fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Recruiting centres from the INTEHEART study by region in China.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Standardised distribution of body mass index (BMI), waist to hip ratios, total, high-density lipoprotein (HDL) and low-density lipoprotein (LDL) levels and ApoB to ApoA-1 ratios in the populations (both cases and controls) from China and other countries. Mean values differ significantly between the two populations.

As with the global INTERHEART study, the risk factors with the strongest relationship to AMI were smoking (current and former), hypertension, diabetes, highest versus lowest tertile of ApoB/ApoA-1 ratio, depression and global stress.16 Table 2 shows the strengths of association of risk factors and AMI between Chinese patients and those of other countries. Regular exercise, moderate alcohol consumption and regular consumption of both fruits and vegetables were similarly associated with reduced risk of AMI in China versus other countries and the risks associated with current/former smoking, lower education and lower income were similarly associated with increased risk of AMI in China versus other countries (table 2). While China had less adverse lipid profiles (total and LDL cholesterol levels and ApoB/ApoA-1 ratio) than other countries, the risks associated with these lipid fractions and AMI were similar between China and other countries. The odds ratios for increasing quintiles of ApoB/ApoA-1 ratio for the two populations showed similar patterns of significantly graded increasing risks of AMI with quintiles of increasing ratio, with no evidence of heterogeneity (p = 0.298) (fig 3). Consistent findings were found when the relationship between changes in standard deviations of the ApoB/ApoA-1 ratio was compared between China and other countries. However, significant heterogeneity in the associated risk of some risk factors and AMI were seen between China and other countries. The association of diabetes (5.10 vs 2.84), depression (2.27 vs 1.37) and permanent stress (2.67 vs 2.06) in China were associated with a higher risk of AMI compared with other countries. Conversely, the risk of AMI associated with a high waist to hip ratio (1.33 vs 2.62) was lower for China than for other countries. This heterogeneity was also displayed when the odds ratio associated with an increase of 1 standard deviation of waist to hip ratio was compared between China and other countries.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Graded increases in risks for acute myocardial infarction associated with increasing quintiles (Q) of waist to hip ratio and ApoB to ApoA-1 ratio in the populations from China and other countries. There were significant differences in the graded relationship between the two populations for waist to hip ratios (heterogeneity p<0.001) but no significant differences for ApoB and ApoA-1 ratio quintiles.

Population attributable risks (PARs)

The PAR associated with the nine risk factors in China accounted for 89.4% of AMI risk. Education had a PAR for AMI of 16.1% and income 9.3%. Inclusion of these variables with the nine INTERHEART risk factors raised the total PAR to 90.4% (table 3). The PAR for self-reported diabetes in the Chinese population was relatively low at 10.0%, despite the high risk associated with this risk factor, owing to the relatively low prevalence of diabetes among the study subjects.

Northern versus southern China

Risk factor levels

Table 4 shows a comparison of risk factor levels between northern and southern Chinese centres. Northern Chinese controls had more adverse levels of current smoking (23.2% vs 15.5%, p = 0.001), hypertension (16.7% vs 11.8%, p<0.001), mean BMI (24.6 vs 24.0, p<0.001), waist to hip ratio (0.88 vs 0.86, p<0.001), HDL cholesterol (1.15 vs 1.26, p<0.001). Southern Chinese controls reported lower levels of daily fruit and vegetable consumption (40.2% vs 47.3%, p = 0.005), higher prevalence of depression (17.3% vs 8.1%, p<0.001) and of stress (1.91 vs 0.60, p = 0.003).

Risk factor effects

Differences in levels of associated risk were found between southern and northern centres in China (table 5). The odds ratios associated with hypertension (OR 3.80 vs 2.09), ApoB/ApoA-1 ratios (5.63 vs 2.48) and waist to hip ratios (2.27 vs 1.14) were significantly higher in the southern centres than in those in the north. Conversely, associated risk due to depression was higher in the north (1.09 vs 2.70) (heterogeneity for all four, p⩽0.001). These differences were seen despite controlling for age and sex.

Discussion

This study showed that the AMI risks associated with the nine modifiable risk factors studied by the INTERHEART study were consistent in the populations studied in China and in the other countries in the INTERHEART study.17 The Chinese participants generally had better risk factor profiles than other countries. In particular, Chinese had lower levels of lipids, diabetes, psychosocial factors and BMI/waist to hip ratio. Smoking rates and self-reported hypertension rates were, however, similar to those of other countries. These findings are consistent with the results of previous surveys of CVRFs in China.3 4 5 6 7 8 9 10 11 12 13 14 In this study we also found that the AMI risk associated with adverse levels of ApoB/ApoA-1 ratios were similar between China and other countries. Smoking was similarly associated with AMI compared with other countries, supporting findings from other studies that the Chinese are not relatively protected from smoking, and that the high rates of tobacco consumption in this population are a continuing and significant public health concern.4 5 Similarly, the relationship between hypertension and AMI was consistent compared with other countries. However, the strength of association of diabetes and psychosocial factors with AMI was higher for the Chinese than for other countries, and the strength of association of waist to hip ratio and AMI was lower in Chinese than in other countries.

Levels of total and LDL cholesterol, ApoB levels and the ApoB/ApoA-1 ratios were lower and HDL cholesterol and ApoA levels higher in the Chinese population than in other, mainly Western, populations.6 7 Despite these potentially beneficial levels, the AMI risk associated with increasing lipid fractions (ApoB/ApoA-1 ratios, increasing LDL cholesterol in China) and decreasing HDL cholesterol levels, was almost identical to that found in other countries (fig 3). Thus, the risks associated with dyslipidaemia, despite the lower profiles in the Chinese population, are substantial and warrant similar preventive measures. If the levels of lipids increase in China, especially if a rise is coupled with high rates of smoking, may be a harbinger of future increase in coronary heart disease.

In contrast, the relationship between a raised waist to hip ratio and risk for AMI appeared less pronounced in the Chinese population. Both BMI and waist to hip ratio were much lower than those seen in other countries in INTERHEART and in other studies.20 The reason for the heterogeneity in the odds ratio associated with an increase of one standard deviation of the ratio when compared between China and other countries is unclear, although it might be partly explained by the relatively narrow distribution of waist to hip ratio and the lower average ratios in the Chinese. Other studies have shown that being overweight remains an independent risk factor for CVD in the Chinese population, even when overweight and obesity are less prevalent than in Western populations.8 9 Unfortunately, the prevalence of obesity is increasing rapidly in China and it is likely that the PARs will similarly rise.9

The observation that risks associated with diabetes were higher in the Chinese population than in other countries has been seen in previous studies.10 11 It may be that, as diabetes was self-reported and there was a lower awareness of this disease in China owing to lower rates of screening/diagnosis, those with known diabetes might have had worse disease as a consequence of late diagnosis or undertreatment. In contrast, awareness of hypertension is likely to be better among the Chinese as the risk of hypertension associated with stroke is better understood in the Chinese population. Consequently, this risk factor would have been better recognised and managed, and may be the reason why AMI hypertension risk did not differ between the Chinese and other populations.

The association of stress and other psychosocial factors with AMI risk is very large in this study.21 22 Although the prevalence of depression and permanent stress was lower in China than in other countries, the associated AMI risk was significantly higher in the Chinese. While rates of depression have been reported to be lower in China and other Asian countries, the physical disabilities associated with depression have been shown to be consistent with those of other countries. Those reporting depression in China might have had severe disease owing to late presentation or treatment and this may explain the stronger association with AMI.23 24 25

We did note differences between southern and northern regions of China. However, it is to be noted that there were no agreed criteria on classifying cities into northern and southern regions, and it is likely that the differences are graduated.12 Still, the differences we have noted in this analysis can help in emphasising the differential importance of these risk factors in regions of the country. The higher levels of certain risk factors in controls from northern regions are consistent with findings from previous surveys of CVRFs in China. The Sino-Monica project studied the trends and determinants of CVD in different parts of China two decades previously and found that CVD mortality and risk factor levels were higher in the north and lower in the south; however, it only examined smoking habits, blood pressure, serum total cholesterol, weight and height.12 13 26 Odds ratios associated with hypertension, high ApoB/ApoA-1 ratios and high waist to hip ratio were all higher in southern Chinese, perhaps owing to a longer time exposed to these risk factors; the odds ratio associated with depression was higher in the north, despite depression being more prevalent in the south and this may be because northern Chinese only reported depression if they had severe forms. Depression is more common in Westernised countries and the higher rates of depression in southern centres may be because these are more Westernised and developed than northern centres.23 24 25 Risk associated with higher income was also stronger in the south, but risk associated with education, which may be a better marker of socioeconomic status in developing countries, was similar. It is likely that the incidence of CVD in both northern and southern regions of China will rise rapidly with rapid economic development.

The strengths of this study are its large size, the simultaneous and contemporaneous nature of data collection for the common risk factors. We have used the ApoB/ApoA-1 ratio as a powerful marker of dyslipidaemia, the psychosocial factors of depression and stress at work or home, diet and exercise, all of which have not been simultaneously assessed in the Chinese population. This common protocol allowed us to compare directly the various risk factors and associated AMI risk between the Chinese populations and populations in other countries.

In order to obtain a large number of cases from each region of the world, a case–control approach, which is open to some potential biases in selection of participants, was used. We minimised the susceptibility to bias inherent in a case-control study design by careful matching of cases and controls. By enrolling incident cases of first AMI, we reduced selection bias and minimised any biases resulting from lifestyle changes in people with previous coronary artery disease, and by recruiting cases and controls from the same source population and from both the hospital and community. The use of uniform standardised methods of data collection by trained research assistants for both the cases and controls also reduced measurement biases. The absence of major selection bias was shown by the finding that with a separate analysis of the different types of control, similar results were obtained.16

This study shows that the relative importance of known and modifiable risk factors for AMI in the Chinese were comparable to those seen in other countries. Indeed the risks associated with emerging risk factors such as diabetes and psychosocial factors may be greater than for other countries. Lifestyle changes and social changes in a rapidly developing economy like China usually include decreasing levels of exercise, increasing obesity, changes in diets and increasing cigarette smoking, resulting in worsening of risk factors such as lipids, glucose or blood pressure. The risk of AMI is likely to increase substantially in China in the years to come. Programmes that prevent the development of risk factors, the early identification and treatment of undiagnosed risk factors (such as diabetes and abnormal lipids) and the effective control of already prevalent risk factors (particularly smoking and hypertension) through individual effort and policy intervention will mitigate much of this potential future rise in AMI in China and must be addressed urgently for this nation.