Summary

The Framingham equation is likely to overestimate the risk of coronary heart disease (CHD) for the Hong Kong population. The 10-year CHD incidence reported in the Chinese Multi-provincial Cohort Study (CMCS) was only one fifth that of the original Framingham cohort. The Framingham equation was calibrated by the CMCS to predict the risk of cardiovascular events (iCVS) for Chinese. We appraised the CMCS equation on its validity and applicability in local Chinese. CMCS achieved good/acceptable internal and external validity. >From limited evidence, the prevalence of risk factors in Hong Kong lies between the two cohorts. However, available local data was neither well described nor consistent, precluding the calibration of a local equation. We have developed a computer programme / charts directly from the CMCS equation. If we have good data for the equivalent risks/ event rates, it is possible to calibrate an equation for local use.

摘要

採用佛萊明翰（Framingham）方程式預測，可能會高估香港人患冠心病之風險。據中國多省市人口組別研究（CMCS）報告，華人在十年間的心血管病發生率只及原來Framingham人口組別研究所得的五分之一。Framingham方程式經CMCS標定後可用來預測華人患上心血管病變的風險。

在評估這方程式在本地華人的正確性和可適用性時，我們得知經CMCS標定的預測等式有良好/可接受的內在及外在正確性。從少數證據顯示，港人的風險因素普遍性介乎上述兩個研究人口組別之間。然而，由於數據不足及缺乏一致性，未能標定出適用於本地的預測等式。但我們以CMCS的等式直接研製了一些電腦程式和圖表。當香港在危險因素及心血管病變方面有更完整的數據時，我們可標定一列適合本地使用的預測等式。

---

Introduction

The Framingham risk equation or derived risk assessment tables/charts¹ are commonly used to assess patients' cardiovascular risks. Since most of the Framingham participants were Caucasians, it may overestimate the risk for other ethnic populations.² Even in Europe, the equation overestimates coronary heart disease (CHD) risk in some European populations.^3,4 In Hong Kong both the prevalence and death incidence of CHD were much lower than those in most western countries.⁵ It is likely that the Framingham equation also overestimates the risk for the Hong Kong population. However, evidence for or against this probability is lacking. Local doctors may be unsure of what to do; 50% of family physicians in a local survey did not use any risk assessment tool for patients with dyslipidaemia.⁶

There are recent reports on a prospective cohort of more than 30,000 Chinese adults, the Chinese Multi-provincial Cohort Study (CMCS) for coronary heart disease and ischaemic stroke (IS). The term ischaemic Cardiovascular Disease (iCVD) was used by the CMCS investigators to report the combination of CHD and IS events.^7,8 The 10-year CHD incidence of this cohort was only one fifth that of the original Framingham cohort (Table 1). Accordingly the Framingham equation was calibrated and modified by the CMCS investigators to predict the risk of CHD and iCVD for Chinese.

This discussion paper consists of two parts. In Part I, we appraise the CMCS study in order to assist local colleagues to make the decision: given the latest evidence from the CMCS, can we directly apply the CMCS calibrated Framingham equations to guide clinical decisions in Hong Kong? In Part II, we shall discuss in more details the coloured charts and the computer programme that have been derived from the re-calibrated equation. (Readers are strongly advised to read the original English CMCS paper⁷ if they wish to understand the methodology of how the calibrated equations were constructed.)

The CMCS

The CMCS cohort included 30121 Chinese adults aged 35 to 64 years at baseline. Participants were recruited from 11 provinces in China and were followed for new "hard" CHD events (coronary deaths and myocardial infarction) from 1992 to 2002. In 1992 and 1993, 27003 were recruited from 16 centres that had taken part in the Sino-Monica Project. The MONICA (Multinational MONItoring of trends and determinants in CArdiovascular disease) Project was established in the early 1980s in many centres around the world to monitor trends in cardiovascular diseases, and to relate these to risk factor changes in the population over a ten-year period. The total population (aged 25-64 years) monitored was ten million men and women. Additionally, 3118 participants from Beijing were added to the CMCS cohort in 1996 and 1999.

The Framingham investigators used the Cox regression model to predict the absolute risk of CHD of a patient. The mathematical model involved regression coefficients and the mean values of the risk factors in the cohort. In both the Framingham and the CMCS cohorts, all the major risk factors, namely, age, blood pressure, smoking, diabetes, total cholesterol and high-density lipoprotein cholesterol (HDL), showed a similar relation to CHD. Body mass index (BMI) was not significantly associated with CHD risk; neither was physical exercise after adjustment for other risk factors. The magnitude of the relative risk of both cohorts did not differ significantly. However, if the original Framingham risk equation was applied to the CMCS cohort, a systematic overestimation of the 10-year absolute risk of CHD was observed, especially in those patients with more risk factors. Using their own observed data, the CMCS group calibrated the Framingham equation for the prediction of CHD risk for the Chinese cohort.

The clinical applicability of CMCS to predict CHD risk alone is likely to be limited by underestimating other cardiovascular outcomes in Chinese. Stroke is the most common manifestation of cardiovascular disease among Chinese men, its occurrence being higher in China than in Scandinavian and European countries.^9,10

In another paper reporting the same CMCS cohort (published in Chinese language⁸), ischaemic stroke (IS) events were reported on top of hard CHD events. There were some minor differences in the numbers reported in the English and Chinese articles (Table 2). CHD and IS were combined as one endpoint, "ischaemic cardiovascular events" (iCVD). It can be seen that for Chinese, IS event rates were higher than CHD event rates. Yet, the combined iCVD event rate was still lower than pure CHD event rates in the Framingham cohort. The Framingham equation was also modified by the CMCS investigators to predict iCVD risk in Chinese.

Appraisal of the CMCS
A.	Internal validity
	The School of Population Health in the University of Auckland developed the Generic Appraisal Tool for Epidemiology, GATE,11 in 2003. It is based on the framework of PECOT (Participants-Exposure-Comparison-Outcome-Time) and is a comprehensive appraisal tool. Using GATE, we considered that CMCS achieved good/acceptable internal validity. A brief summary is given below to point out the area of potential bias in the CMCS. The full appraisal can be found in Appendix 1.
	1.	Participants - There were sufficient details on the inclusion and exclusion criteria on the recruitment of eligibles, which were all without prior CHD. While the selection process was appropriate, the randomization process was not described.
	2.	Exposure and Comparison - For risk factor exposure, considerable details were given according to the WHO-MONICA protocol. Efforts were made to ensure objective and valid measurements of the risk factors, e.g. blood pressure; but it was unclear who did the measurements.
	3.	Outcome - There was clear definition of outcome in terms of "hard" CHD events and outcome was recorded by trained physicians, using standardized forms and verified by independent investigators. However, both the physicians and the investigators were not blinded to the risk factors.
	4.	Time - The follow-up was long enough for the original 27003 participants (10 years) but not for the additional 3118 participants who were followed up for only 3 years. There were different rates of loss to follow-up at different stages of the study. The total loss to follow-up was 35%. The addition of new participants near the end of the study could lead to bias, but a separate model was created to assess the effect of drop-out and the effect was found to be minimal.
B.	External validity
	Important issues that may affect the applicability of the CMCS equations to our patients include the following:
	1.	CMCS stratified the risk factors according to the original Framingham study, for example, stages of hypertension were classified according to the Fifth Joint National Committee (JNC V)12 rather than the most updated JNC VII.13
	2.	The cut-point for diabetes mellitus was taken as fasting blood glucose >7.8 mmol/l. That is, some patients who would be diagnosed as diabetic in Hong Kong were classified as non-diabetic by CMCS.
	3.	The validity of the CMCS results is limited to patients within the age-range 35 to 64; applicability in the older populations who are at higher risk is unknown.
	4.	The calibrated equations have not been validated in another similar population.
C.	Other points of interest
	1.	The apparent protective effect of smoking in Chinese women could be a random error due to small numbers (only one hard CHD event was reported among the 4% who were smokers).
	2.	Using the same data and sample equation for CHD as published,7 we constructed the prediction equation of CHD in females. We found that the CMCS calibrated equation gave unexpectedly high risk for females aged above 52 years (higher than that for males of the same age). Our speculation is: the assumption that the event rates vary with the risk factors in a linear relationship does not hold for females whose risk of CHD rises disproportionally after menopause.

Applicability of the calibrated Framingham equation for the Hong Kong population

Direct application of prediction models will be most appropriate for individuals who resemble the cohort from which the result is derived. Prevalence rates of cardiovascular risk factors in Hong Kong¹⁴ are given in Table 3, comparing those in the Framingham cohort and the CMCS cohort. Salient differences between the local population and the CMCS cohort include a lower smoking rate amongst Hong Kong male, higher prevalence of raised total cholesterol level and a lower prevalence of hypertension (the cut-offs are chosen because of the limited information with respect to local data). The cutoff point of fasting plasma glucose lower than 7.0 mmol/l (instead of 7.8 mmol/l in CMCS) is nowadays taken as diagnostic criterion for diabetes mellitus for the Hong Kong population.¹⁵ This difference can affect the potential applicability of the equation in two ways. Firstly, the difference in prevalence of diabetes mellitus by the two diagnostic criteria may influence its weighting among other risk factors. Secondly, since the factor of diabetes, unlike blood pressure, is entered in a dichotomous manner (either "yes" or "no"), the older cut-off of 7.8 mmol has to be employed if the equation were to be applied in the local population.

From limited evidence, in Hong Kong, the prevalence of risk factors lies between the corresponding rates of the Framingham and CMCS cohorts. If the Framingham equation likely over-estimates the risk of CHD for Hong Kong population, the CMCS calibrated equation likely under-estimates it. Which one would be closer to the "true" risk? Could these estimates be taken as equivalent to the upper and lower limits of the predicted risk?

Discussion

This discussion paper introduces the CMCS and tries to appraise its internal and external validity, with special reference to Hong Kong Chinese. The authors deliberately avoid taking sides on which risk prediction equation would give a more accurate estimation of cardiovascular risk. Local family physicians could use the presented evidence, together with their own expertise and their patients' preference, to discuss the management plan.

The CMCS illustrates how the Framingham equation can be calibrated for a particular population that shows prevalence of CHD, CVD and risk factors different from the original Framingham cohort. Among the more striking findings is the observation that the association of BMI and physical exercise to CHD risk is not significant; probably they act through other factors such as the development of diabetes mellitus. This finding is helpful in interpreting the associations of cardiovascular events with various risk factors as reported in local studies.^16-18

While it is quite tempting to construct similar calibrated equations for the Hong Kong population, it is disappointing to find that the basic information on the prevalence of various risk factors in Hong Kong is not well described or is inconsistent. There are so far only two epidemiological studies on the cardiovascular risk factors in Hong Kong, by Janus¹⁴ and Chan WK et al.¹⁶ The former was a population-based cross-sectional study. Random telephone interviews were performed but the details of recruitment of the 2900 participants for laboratory tests from the 7730 eligibles were not described in detail; so were the result data. Chan et al was a retrospective study of 17716 participants for purposes of health screening, among whom 54.2% had total cholesterol levels >5.2 mmol/l, 18.5% were hypertensive, 15.1% were smokers, and 3.7% had diabetes mellitus. These latter figures were quite different from those cited by Janus (see Table 3). We could not find any other local cohort studies.

From a clinician's perspective, the combination of heart and stroke endpoints are more useful in guiding clinical decisions with respect to management of the modifiable risk factors. In this aspect, the CMCS calibrated equation for iCVD is more helpful than that for CHD alone. Moreover, the former does not give an unexpectedly higher predicted risk in females aged above 52 years in comparison to males of the same age.

From the CMCS articles and basing on the New Zealand risk charts,¹⁷ one of the authors (WKL) has developed coloured risk charts for the Chinese cohort. Another author (YTW) has developed a computer programme for a more precise calculation of the absolute risks. These tools will be described in more detail in Part (II) of this paper.

Conclusion

The predicted risks derived from the CMCS calibrated Framingham equations were closer to the observed cardiovascular events in China than the original Framingham equation. Further adjustment is required for the CHD risk for females after menopause. If good local data for the equivalent risks and event rates were available, one may develop an equivalent prediction equation for use in the Hong Kong population.

Acknowledgement

The content of this paper was presented in the Third Asian Pacific Conference on Evidence-Based Medicine at the Chinese University of Hong Kong on 26-28th November, 2004. We thank the Librarian, Professional Development and Quality Assurance, Department of Health, for searching the original articles of CMCS. We would also thank the JASPA Group (Drs Ching Kam Wing, Wendy Lo, Ngan Po Lun, Kinson Lau, Yanny Yu) and the EBM Group, PD&QA, DH (Drs Roger Wong, Douglas Lai, Lai Wing Yiu, Fung Kin Chor) for their input in the initial discussions.

Key messages

1. The Framingham risk equation is likely to over-estimate risk of coronary heart disease (CHD) in Chinese.
2. A Chinese Multi-provincial Cohort Study (CMCS) involving more than 30,000 Chinese was recently published. The 10-year incidence rate of coronary heart disease was only one-fifth that of the original Framingham cohort. The Framingham equation was calibrated to predict absolute CHD event rates in Chinese.
3. The ischaemic stroke (IS) event rate of the CMCS cohort was reported separately and the Framingham equation was modified to predict combined CHD and IS events (iCVS).
4. CMCS achieved good/acceptable internal and external validity. It is uncertain whether the new evidence can be applied directly in Hong Kong. Data on incidence of iCVS and prevalence of risk factors are insufficient for the calibration of a local equation.

---

---

References