Application of Cardiovascular disease (CVD) risk assessment equations to the Thai population

This document is the author deposited version. You are advised to consult the publisher's version if you wish to cite from it. Published version KHATAB, Khaled, INTHAWONG, Rungkarn, WHITFIELD, Malcolm, COLLINS, Karen, RAHEEM, Maruf A and ISMAIL, Mubarak (2019). Application of Cardiovascular disease (CVD) risk assessment equations to the Thai population. Biostatistics and Epidemiology International Journal.


Introduction
Cardiovascular disease (CVD), a set of disease affecting either the heart or blood vessels, is a global health problem, the prevalence of which has led to an increased epidemiological transition of the disease from high-income countries to low-middle income countries. CVD risk prediction models have been developed in many studies, as a tool to predict the levels of CVD in the future. The most influential prediction models, using statistical multivariate analysis equations, were derived from the Framingham Study. [1][2][3][4] Many studies found that the Framingham's function was suitable for predicting future CVD in the middle-aged, US white population and others with similar profiles of CVD risks. 5, 6 However, some studies showed an overestimated prediction for groups which have different risk profiles and ethnicity such as Asian and some EU populations. [7][8][9][10] Nevertheless, the Framingham equation gives a generalised prediction, concerning the biomarker risk factors, such as blood pressure and cholesterol level. Several studies have developed adjustments to the Framingham CVD risk prediction equation, to adapt it for use in specific populations. For example, the QRISK(cardiovascular risk calculator) score in the United Kingdom and the Asia Pacific Collaborative Cohort Studies (APCCS). 11,12 In Thailand, to date, there is only one CVD risk score, which has been derived from the EGAT cohort study. 13 This study has followed-up 3,499 participants aged 35 to 54 years since 1985. Several studies found that RAMA-EGAT risk score provided a better prediction of CVD than the Framingham equation. 14 Although the CVD risk score was derived specifically for the Thai population, this cohort study was well representative only for Thai middle-class men aged 35 years and over, but it was not well represented in women. 1 To estimate the number of 8-10 years CVD events; 3) To identify and validate the most suitable CVD risk equations for the Thai population.
This research presents the results of the application of the CVD risk assessment equations to the Thai populations and the probability of CVD events using the three different equations, the estimated number of non-fatal CVD events in the Thai people and the validation of the mathematic model. The details of the analysis and CVD risk assessment equations have been described in the previous study.

Materials and Methods
The 4 th National Health Examination Survey IV(NHESIV), 2009dataset has been used as the baseline population and provided the risk factors profile of the Thai population. This survey represents the prevalence of chronic diseases and the potential health risk factors to the Thai population. This survey is a national representative of the Thai health status by age, gender, administrative area (urban/rural) and geographic regions. TheNHESIV dataset has been used as the study population.

Calculation of the CVD probability
The individual data and risk factors profile has been entered to an Excel spreadsheet. The APCS, Framingham-Asia and the original-Framingham equations have been applied to the data to calculate the probability of getting CVD in each. Then, the mean of 8-10 years probability has been calculated by age groups and gender. The primary outcomes are the probability of IHD, stroke and all CVD events.  15 Third, the Framingham's equation (Framingham original), using the concept of the Weibull accelerated failure-time models, which have been applied to the Thai population. 16 The main outcome prediction is the of probabilities of CVD related events, such as stroke, heart attacks and heart failure. Details of the equations and the process of calculations are described in the appendix.

Outcome of interests
8 to 10 years probability of getting CVD, IHD or Stroke of the individuals and the 8 to 10 years probability of getting CVD, IHD or Stroke events by age groups and gender.

Estimated number of CVD events
The number of CVD events at the national level is estimated by multiplying the CVD probability to the number of the mid-year population in 2009, by age groups and gender, to obtain the number of all CVD events over the next 8 to 10 year period. The probability of CVD events includes both fatal and non-fatal CVD. The analysis excluded those who died based on the statistical information. The model deducted the number of people who are likely to die from CVD, which will estimate the number of people who are alive with the CVD conditions. The probability of dying from a CVD condition is calculated from the national CVD mortality rate in Thailand, from the national statistical report over the past decade.

Validation of the CVD estimated events
The estimated number of CVD events are validated with the actual number of CVD hospital admissions in Thailand in 2009. The number of CVD cases are classified by the ICD-10 code in which I0-I99 refers to all CVD events, I20-I25 refers to IHD and I60-I69 relates to stroke. The actual number of CVD admissions comes from the National Health Security Office, Thailand (NSO) which covers 75% of all hospital admissions in Thailand and includes the patients who were in the universal coverage health care scheme.

Results
The probability of CVD events Table 1 shows the comparison of the mean likelihood of CVD, which has been calculated by using the APCS equation, the Framingham-Asia equation and the original Framingham equation in the NHESIV dataset. The mean probability of CVD has been estimated by age groups and gender. When applying the APCCS equation, the overall mean of 8-year CVD probability is 8.3% in men and 7.8% in women. The trend of 8-year probability increased according to age groups in both men and women. The likelihood of CVD is lowest at age 15-24 years, and the possibility of CVD starts to grow from the 45-54 years age group The CVD probability in men is higher than in women from the ages 15 to 74 years; whereas, women have CVD probability higher than men at age 75 years and over.
The mean 8-year probabilities of CVD when using the Framingham-Asia equation are 7.2% in men and 8.1% in women. The minimum CVD probability is at age 15-24 year. At age 45-54 years, the CVD probability in men is 3.2% and 2.6% in women. The CVD probability is equal to 6.7% in both men and women at age 55-64 years. The CVD probability increases with an increase of age groups. Although the trends of CVD probability of the Framingham-Asia equation are similar to the CVD probability that is calculated with APCS equation, the Framingham-Asia calculated the CVD probability in women as higher than in men from age 65 years and over.The 8-year probability of CVD is highest at age 75 years and over.
Regarding the original Framingham equation, the overall mean of 10year CVD probability is 18.8% in men and 11.1% in women. The CVD probability in men is higher than in women in all age groups. The lowest CVD probabilities are 1.9% in men and 0.8% in women at age 15-24 years. The CVD probability continuously increases when age increases, in both men and women. The CVD probability is two times higher in men than in women at the age 35-44 and 45-54 years.
In old age groups, the 10-year CVD probability is higher than in the younger age groups, and the highest likelihood of CVD is at age 75 years and over, which are 47.6% in men and 30.8% in women. Table 2 shows the 10-year probabilities of CVD, which are calculated by the Framingham original equation in the NHESIV data set. This equation is capable of calculating three events, which are IHD, stroke, and all CVD events. Table 2 presents the mean 10-year probability, by CVD condition, age groups and gender.
The overall mean of 10-year IHD probability is 14.5% in men and 8.0% in women. The trend of IHD probability rises with an increase in age groups. Men have more likelihood of getting IHD than women in all age groups. The IHD probability is the lowest at age 15-24, which are 1.9% in men and 0.7% in women. The likelihood of IHD at age 25-34 is 3.9% in men and 1.4% in women. The IHD probability in men continually increases in the middle period and elderly age groups. At age 35-44 years, the IHD probability in men is nearly triple that in women, which is 6.3% in men and 2.4% in women. The IHD probability is twice as high in men as in women at age 45-54 years, which is 10.2% in men and 4.8% in women. At age 55-64, the IHD probability in men is 15.9% and 9.1% in women. The IHD probability in elderly age groups, 65-74 years and 75 years and above is higher than the IHD probability in younger age groups. At age 65-74 years, there are 22.8% of men and 14.0% of women who have a chance of getting IHD. The IHD probability reaches the highest at age 75 years and over, which is 32.4% in men and 20.15 in women. The average 10-years probability of stroke is 6.2% in men and 3.8% in women. The graph shows that the stroke probability is below 1% from age 15 to 44 years in both men and women. At age 45 -54 years, there is 1.5% probability in men and 1.2% of women who suffer a stroke. The trend of stroke probability increase from the age of 55 years and men are more likely to have a chance of suffering a stroke than women. The stroke probability is 4.5% in men and 3.1% in women at age 55-64 years. Then, the stroke probability increases in the elderly age groups, which is 10.6% in men and 6.8% in women at age 65-74 years. The stroke probability reaches 24.7% in men and 14.6% in women at age 75 years and above. Compared to figure 5.3, the mean of the 10-year likelihood of stroke is lower than the probability of IHD, in both men and women in all age groups. The overall mean of stroke probability is twice as high as the mean of IHD probability. In men, the likelihood of IHD is 14.5% and stroke 6.2%. In women, the possibility of IHD is 8.0%, while the chance of stroke is 3.8%.   the mean CVD probability in Thai men, when applying the original Framingham equation is 18.8% within a 10-period. While the APCS and the Framingham-Asia equations estimated the 8-year likelihood of getting CVD in men as 8.3% and 7.2% respectively; comparing across age groups, the trend of CVD probability increases with an increase in age. However, the original Framingham equation presents a higher likelihood of CVD than APCS or the Framingham-Asia equations in all age groups. Additionally, the trend of 8-year CVD probability is similar between the APCS equation and the Framingham-Asia equation at age 15 to 54. However, the 8-year CVD probability using the APCS is slightly higher than the Framingham-Asia equation at ages more than 55 years.

The estimated number of CVD events in the Thai population over 8 to 10 year periods
This section presents the estimated number of CVD events when multiplying the probability of CVD in an 8-year and 10-year period, to the number of the national mid-year population in Thailand in 2009. The number of CVD events, which have been calculated using three different equations, is the number of total CVD events which included both fatal and non-fatal events of IHD and stroke. Therefore, the number of estimated CVD events will deduct the number of people who are dead from CVD, to obtain the number of people who are alive with CVD. The number of deaths from CVD is estimated by applying the average of 10-year CVD mortality probability, which has been calculated from the national CVD mortality rate in Thailand over a 10-year period (1996-2006).   The national hospitalisation data in 2009 from NSO, Thailand, have been used to validate the model which covers the in-patient's hospital admissions for the whole country.Thein-patient's admissions from the universal coverage health care scheme (UC), which accounts for 75% coverage of the all hospitalisation data in Thailand. The average number of the estimated CVD patients per year has been calculated to make it comparable with the 1-year hospital admissions in Thailand and is compared by age groups and gender. The CVD conditions have been identified by using the ICD-10 code, which I00-I99 refers to all CVD, I20-I25 refers to IHD and I60-I69 refers to stroke.   both men and women. However, the original Framingham equation projects a higher CVD probability than either the APCS equation or the Framingham-Asia equation, in both men and women at all age groups. This might be because the original Framingham equation was derived from the Framingham cohort, who have different characteristics of CVD risk factors and ethnicity, compare to the Asian population. Many studies have found that the Framingham's function was suitable for predicting future CVD in middle-aged US white populations, and others with similar profiles of CVD risks. 5,6 However, some studies showed an overestimated prediction for groups which have different risk profiles and ethnicity, such as Asian and some EU populations. [7][8][9]17 The APCS equation was derived from the pooled data of the cohort studies around Asia and included the EGAT cohort study in Thailand. Hence, the Framingham-Asia equation was also recalibrated with the APCS cohort. Therefore, the APCS equation and the Framingham-Asia equation are more suitable than the original Framingham equation, when applied to the Thai population.
When comparing genders, the original Framingham equation calculated the higher probability of CVD in men than in women. Conversely, the likelihood of CVD in women is higher than in men in the Framingham Asia equation. This might because the Framingham Asia equation was reported to overestimate the risk of CVD by an average 4% in women and underestimate the CVD risk by an average 2% in men, in the non-Chinese cohorts. 16 Also, the prevalence of regular smoking in women is lower than in men, which might have an impact and affect the calculations. There are some metabolic risk factors in women which are higher than men. For example, the mean of BMI in men and women is 23 kg/m2 and 24.36 kg/m 2 respectively. The mean of total cholesterol in women and men are respectively, 5.58 mmol, and 5.27 mmol. Hence, the prevalence of diabetes in women is 10.9% but in men is 9.3%.  Furthermore, there are some limitations on the availability of data in Thailand for undertaking the validation. As mention above, The Chronic disease surveillance system did not capture all CVD events for the whole country. The data only included out-patients admissions in public hospitals in 43 provinces, which was not segregated by gender. Hence, the number of CVD patients is not available to compare by gender. The number of in-patients hospitalisation data covered 75% of the hospital admissions in Thailand. This data, however, did not take account of approximately 8% of the Thai population who are in the social security health care scheme, 8% who are in the civil servant health care scheme and the other 9% who are in the private health insurance system, for which data was not available to access.
Regarding the mortality data, the number of deaths from CVD may be underestimated, because the average 10-year CVD mortality probability has been calculated from the national CVD mortality rate in Thailand during 1996 to 2006, but the estimation is based on the year 2009. Hence, the death registry in Thailand might be underreporting or miss classifying the causes of death in its data. Moreover, the CVD mortality rate in 2009 was presented as a total mortality rate, but not given by age groups and gender. Therefore, the 10-years mortality from 1996 to 2006 was used instead because it was the best available data during the period of this study.

Conclusions
The When: