Future Cardiovascular Disease in China Markov Model and Risk Factor Scenario Projections From the Coronary Heart Disease Policy Model China

Transcription

1 Future Cardiovascular Disease in China Markov Model and Risk Factor Scenario Projections From the Coronary Heart Disease Policy Model China Andrew Moran, MD, MPH; Dongfeng Gu, MD, MSc; Dong Zhao, MD, PhD; Pamela Coxson, PhD; Y. Claire Wang, MD, MSc; Chung-Shiuan Chen, MS; Jing Liu, MD; Jun Cheng, MD; Kirsten Bibbins-Domingo, MD, PhD; Yu-Ming Shen, PhD; Jiang He, MD, PhD; Lee Goldman, MD, MPH Downloaded from by guest on June 19, 2018 Background The relative effects of individual and combined risk factor trends on future cardiovascular disease in China have not been quantified in detail. Methods and Results Future risk factor trends in China were projected based on prior trends. Cardiovascular disease (coronary heart disease and stroke) in adults ages 35 to 84 years was projected from 2010 to 2030 using the Coronary Heart Disease Policy Model China, a Markov computer simulation model. With risk factor levels held constant, projected annual cardiovascular events increased by 50% between 2010 and 2030 based on population aging and growth alone. Projected trends in blood pressure, total cholesterol, diabetes (increases), and active smoking (decline) would increase annual cardiovascular disease events by an additional 23%, an increase of approximately 21.3 million cardiovascular events and 7.7 million cardiovascular deaths over 2010 to Aggressively reducing active smoking in Chinese men to 20% prevalence in 2020 and 10% prevalence in 2030 or reducing mean systolic blood pressure by 3.8 mm Hg in men and women would counteract adverse trends in other risk factors by preventing cardiovascular events and 2.9 to 5.7 million total deaths over 2 decades. Conclusions Aging and population growth will increase cardiovascular disease by more than a half over the coming 20 years, and projected unfavorable trends in blood pressure, total cholesterol, diabetes, and body mass index may accelerate the epidemic. National policy aimed at controlling blood pressure, smoking, and other risk factors would counteract the expected future cardiovascular disease epidemic in China. (Circ Cardiovasc Qual Outcomes. 2010;3: ) Key Words: China stroke coronary heart disease risk factors computer modeling Since the beginning of economic and social reforms in 1979, China has increased its standard of living and life expectancy. Cardiovascular disease, principally stroke and coronary heart disease (CHD), is the leading cause of death and is expected to increase with further economic development and urbanization, aging of the population, and changes in diet and physical activity 1,2 that will predispose many Chinese to high blood pressure, overweight, dyslipidemia, and diabetes. 3,4 Though male smoking prevalence has declined by more than 10% since the mid-1980s, 62% of Chinese men smoke actively, and at least 49% of nonsmokers (predominantly women) are exposed to passive smoking at home or at work. 5 Others have estimated the impact of risk factors on cardiovascular risk, 6 8 and overall cardiovascular disease in China, 9,10 but prior research has not focused on individual or synergistic effects of risk factors on future cardiovascular disease on a national scale. Building on our predictions of the impact of expected demographic changes on CHD, 11 we forecast the impact of projected future risk factor trends on CHD and stroke in China from 2010 to Editorial see p 226 Received September 21, 2009; accepted March 15, From the Division of General Internal Medicine (A.M., Y.-M.S.), Columbia University Medical Center, New York, NY; Columbia University College of Physicians and Surgeons (A.M., L.G.), New York, NY; the Department of Evidence Based Medicine (D.G.), Cardiovascular Institute and Fu Wai Hospital of the Chinese Academy of Medical Sciences, Beijing, China; National Center for Cardiovascular Diseases (D.G.), Beijing, China; the Department of Epidemiology (D.Z., J.L., J.C.), Beijing Institute of Heart, Lung, and Blood Vessel Diseases, Beijing, China; Capital University of Medical Sciences (D.Z., J.L., J.C.), Beijing, China; School of Medicine (P.C., K.B.-D.), University of California San Francisco, San Francisco, Calif; the Department of Health Policy and Management (Y.C.Y.), Mailman School of Public Health, Columbia University, New York, NY; and the Department of Epidemiology (C.-S.C., J.H.) and the Department of Medicine (J.H.), Tulane University School of Medicine, New Orleans, La. The online Data Supplement is available at Correspondence to Dongfeng Gu, Division of Population Genetics and Prevention, Cardiovascular Institute and Fu Wai Hospital, 167 Beilishi Rd, Beijing , China ( gudongfeng@vip.sina.com) or Dong Zhao, MD, Beijing Institute of Heart, Lung, and Blood Vessel Diseases, Capital University of Medical Sciences affiliated with Beijing Anzhen Hospital, Chaoyang District, Beijing People s Republic of China ( DeeZhao@anzhen.org) American Heart Association, Inc. Circ Cardiovasc Qual Outcomes is available at DOI: /CIRCOUTCOMES

2 244 Circ Cardiovasc Qual Outcomes May 2010 Downloaded from by guest on June 19, 2018 WHAT IS KNOWN It is certain that aging of the Chinese population will result in increased numbers of stroke and coronary heart disease events in future years. Population-based surveys since China s economic reforms in the early 1980s suggest increasing trends in blood pressure, cholesterol, and diabetes, countered by a slight decline in smoking in men. Projecting the cardiovascular disease impact of future risk factor trends can point toward best prevention policies. WHAT THE STUDY ADDS This computer modeling study is the first to use past survey data to project a range of likely scenarios for the effects of future individual and combined risk factor trends on cardiovascular disease in China. Aging of the population is inevitable, but the projected additional impact of unfavorable risk factors is potentially reversible. Methods The CHD Policy Model China The CHD Policy-China is a Markov (state-transition) model of cardiovascular disease in the adult Chinese population. 11 Means and proportions of cardiovascular disease risk factors in Chinese adults in 10-year age categories ages 35 to 84 years in 2000 were estimated from the International Collaborative Study of Cardiovascular Disease in Asia Study (InterASIA). 12 Age trends in risk factor levels were preserved over time. Stroke incidence, 13,14 mortality, 15 and case-fatality 13 estimates were derived from other Chinese studies (supplemental Appendix Table 1). Multivariate CHD and ischemic stroke hazard ratios for age, sex, systolic blood pressure (SBP), total cholesterol (TC), active cigarette smoking, high density lipoprotein (HDL) cholesterol, diabetes, and body mass index (BMI) were estimated from the China Multiprovincial Cohort Study (CMCS, supplemental Appendix Table 2). 6 Cox proportional hazard models for hemorrhagic stroke (same risk factors excepting cholesterol, BMI, and diabetes) and noncardiovascular death (same excepting cholesterol and BMI) were also estimated from CMCS. Significant (P 0.05) age risk factor interactions observed for smoking in CHD proportional hazards models, SBP, smoking, and diabetes in ischemic stroke models, smoking and SBP for hemorrhagic stroke models and smoking and diabetes in noncardiovascular mortality models were incorporated in age-specific risk coefficients. For the main analysis, we assumed BMI effects on cardiovascular risk were mediated through other risk factors: the effect of a 1 kg/m 2 increase in BMI on SBP (males, 1.36 mm Hg; females, 1.40 mm Hg), TC (males and females, 0.05 mmol/l), and HDL (males, 0.03 mmol/l; females, 0.02 mmol/l) were estimated from InterASIA. We assumed 1 kg/m 2 increase in BMI would lead to a 0.21% absolute increase in diabetes prevalence. 16 The main outcomes predicted were CHD events (nonfatal and nonfatal first-ever and repeat episodes of stable and unstable angina, myocardial infarction, or cardiac arrest) and stroke events (nonfatal and fatal ischemic and hemorrhagic strokes). CHD deaths, stroke deaths, and noncardiovascular deaths (total mortality stroke and CHD mortality) are reported in the supplemental Appendix. Cardiovascular disease was defined as combined CHD, ischemic stroke, and hemorrhagic stroke. CHD Risk Factor Trend Analysis (2010 to 2030) Based on secular trends in SBP, TC, active smoking, BMI, and diabetes analyzed from 1980 to 2006 (supplemental Appendix Tables 3 and 4, supplemental Appendix Figure 1) future risk factor trends for the population ages 35 to 84 years were projected forward over 2010 to 2030 (Table 1 and supplemental Appendix Figure 2a 2e). HDL and passive smoking trends were not analyzed due to lack of reliable past data. Unless otherwise noted, the nationally representative InterASIA survey value served as the intercept in To ensure historical consistency and biological plausibility, it was decided a priori that in the main analysis no projected trend exceeded the most extreme adult population value in Japanese or Korean national surveys since Linear main SBP, active smoking, and BMI trends were estimated from six China Health and Nutrition Survey (CHNS) 17 surveys 1991 to 2006 using an ageadjusted random effects model assuming clustering at the level of the individual study participant. Other SBP, smoking, and BMI scenarios were based on alternate trends suggested by CHNS data. Age time interactions observed in trends for SBP, BMI, or active smoking were incorporated into age-specific risk factor trend projections. TC and diabetes projections were based on a number of past surveys, and a logistic trend function reaching a predetermined ceiling value was used for the main and high trends. There were insufficient data for assessing age time interactions for TC or diabetes. Trend analyses were conducted using Stata (Statacorp, Austin, Tex) and Excel (Microsoft, Redmond, Wash). The CHD Policy Model China simulated effects of projected risk factor trends on cardiovascular disease over 2010 to A base case simulated cardiovascular disease events over 2010 to 2030 with risk factors held at year 2000 levels. The base case proportion of CHD, stroke, and cardiovascular disease explained by individual risk factors was determined by running a simulation simultaneously setting all risk factors at optimal levels 28 [zero smoking and diabetes exposures and lowest risk levels of BMI, 29 cholesterol, 30 and blood pressure 31 (supplemental Appendix, Figure 1)]. Subsequently, risk factor trend scenario simulations were run and incremental changes in cardiovascular disease events calculated by comparing the trend cases with the base case. Sensitivity Analyses One-way sensitivity analyses explored uncertainty about main analysis trend projections and potential benefits of controlling smoking or SBP. Risk factor -coefficients estimated from Framingham Heart Study data (supplemental Appendix Table 2) were substituted for CMCS coefficients. Based on evidence from CMCS, a simulation assumed additional BMI effects not mediated by SBP, TC, HDL, or diabetes (supplemental Appendix Table 2). Optimistic sensitivity analyses simulated (1) an extremely aggressive tobacco control policy leading to an exponential decline in active smoking in Chinese men to 20% by 2020 and 10% by 2030 (supplemental Appendix Figure 3), (2) lowering mean SBP by 3.6 mm Hg in 2010 (SBP change associated with lowering mean dietary sodium 6 g/d), 32 and (3) lowering case-fatality to recent US levels for CHD 33 and stroke 34 (supplemental Appendix Table 5). Pessimistic sensitivity analyses (1) repeated the high diabetes trend substituting stronger diabetes relative risk coefficients from Framingham, and (2) simulated a rise in TC by 2030 as high as the mean 6.0 mmol/l measured in 1960s US adults. 18 Results Base Case and Trend Scenarios China s population ages 35 to 84 years is expected to grow from 0.67 billion in 2010 to 0.84 billion in 2030, and the proportion of persons ages 65 years in the total population will double (from 7% to 14%). 35 In the baseline simulation with risk factors held at 2000 levels, 38.6 million CHD events and million strokes were projected from 2010 to 2030 (Table 2). Approximately a quarter of all cardiovascular

3 Moran et al Risk Factor Trends and Future CVD in China 245 Table 1. Future Trends in Selected CHD and Stroke Risk Factors in Chinese Adults 2000 to 2030, Based on Observed Survey Trends During 1980 to 2006: Means or Proportions Age-Standardized for Ages 35 to 84 Directly From the 2000 Chinese Census Downloaded from by guest on June 19, 2018 Scenario Definition Trend Function Start Value 2000 End Value 2030 Base case Risk factors constant at 2000 levels Trend cases SBP, mm Hg Estimated from China Health and Nutrition Survey (CHNS) 17 except low scenario based on US national surveys Main M 0.21 annually Linear M M F 0.17 annually F F High M 0.26 annually Linear M F 0.21 annually F Low 0.14 annually (decline) 18 Linear M F TC, mmol/l Main and high followed Sino-MONICA Beijing trend 19 to prespecified ceiling, low followed Beijing professionals 20 trend Main Rise to ceiling of 5.4* (US surveys peak) 18 Logistic* M 4.7 M 5.3 F 4.8 F 5.4 High Rise to ceiling of 5.7* (Japanese surveys peak) 21 Logistic* M 5.7 F 5.7 Low M annually Linear M 5.0 F annually F 5.0 Smoking, % prevalence Estimated from CHNS, except worst case in females based on Japanese surveys 21 Main M 0.58% annually Linear M 59.8% M 36.6% F 0.13% annually F 7.1% F 1.9% Worst M 0.33% annually Linear M 46.6% F 0.30% annually (increase) F 19.0% Best M 0.83% annually Linear M 26.6% F 0.19% annually F 0.2% Diabetes, % prevalence Main and high assumed steep increase seen between national surveys to prespecified ceiling, low based on past prevalence and urbanization. 25 Higher prevalence based on combined fasting glucose and oral glucose tolerance test 26 Main Rise to ceiling of 15% prevalence (peak in United States) 18 Logistic M 5.3% M 14.8% F 6.1% F 15.6% High Start at higher prevalence 2000, 26 rise to ceiling of 22% (peak in Japan) 21 Logistic M 5.8% F 5.6% M 20.4% F 21.5% Low M 0.08 annually Linear M 5.3% M 8.1% F 0.08 annually F 6.1% F 9.0% BMI, kg/m 2 Main trend estimated from CHNS, effects mediated through downstream changes in SBP, TC, HDL cholesterol, and diabetes Main M 0.10 annually Linear M 23.1 M 26.4 F 0.09 annually F 23.5 F 26.4 High Exponential increase, similar to short-term increases in the United States 27 Exponential M 28.7 F 29.0 Low 0.88 ln (year) Log-linear M 24.9 F 25.6 M indicates male; F, female. Unlabeled values are for both men and women. *Logistic trend functions TC: Main 1.1 Exp(0.1 year-2000)/(1 Exp(0.1 year-2000), High 1.3 Exp(0.22 year-2000)/(1 Exp(0.22 year-2000). Logistic trend functions diabetes: Main 15 Exp(0.19 year-2000)/(1 Exp(0.19 year-2000), High 22 Exp(0.21 year-2000)/(1 Exp(0.21 year-2000). Exponential trend function BMI: High males Exp(0.006 year), females EXP (0.005 year).

4 246 Circ Cardiovasc Qual Outcomes May 2010 Table 2. Incremental Changes in Projected CHD and Stroke Events Attributable to Projected Trends in Cardiovascular Disease Risk Factors, Chinese Adults Ages 35 to 84 Years, 2010 to 2030, the CHD Policy Model China Men Women Downloaded from by guest on June 19, 2018 Simulation CHD Events Stroke Events CHD Events Stroke Events Base case Total Trend cases: incremental events (% change from baseline) SBP Main (4) (9) (3) (8) High (6) (12) (4) (12) Low ( 4) ( 8) ( 2) ( 7) TC Main (16) (3) (8) (3) High (31) (6) (15) (6) Low (6) (1) (2) (1) Active smoking Main ( 8) ( 1) ( 1) ( 1) Worst ( 5) ( 1) (2) ( 1) Best ( 11) ( 1) ( 1) ( 1) Diabetes Main (2) (2) (5) (1) High (3) (3) (10) (2) Low ( 1) ( 1) (1) ( 1) BMI Mediated by SBP, TC, HDL, diabetes Main (6) (5) (7) (3) High (11) (8) (10) (5) Low (4) (3) (5) (2) Simultaneous main trends SBP, TC, diabetes Smoking decline (13) (13) (17) (14) No smoking decline (23) (14) (18) (14) disease events were attributable to SBP 115 mm Hg, and TC 3.8 mmol/l (148 mg/dl) and smoking explained most of the remaining proportion of events that are attributable to the risk factors considered (Figure 1). Assuming constant age-specific event and case-fatality rates, annual CHD and stroke events will increase 50% between 2010 and 2030 and crude event rates will increase steadily due to aging and population growth alone (Figure 2A and 2B, Figure 3). A TC increase of 0.58 mmol/l (22.4 mg/dl) in Chinese men and 0.55 mmol/l (21.6 mg/dl) in Chinese women over Figure 1. Proportions of CHD, stroke, and cardiovascular disease events attributable to selected major risk factors in China, base case scenario 2010 to 2030.

5 Moran et al Risk Factor Trends and Future CVD in China 247 Downloaded from by guest on June 19, to 2030 (main assumption) would lead to the highest increase in CHD events of all risk factors modeled (Table 2). The main SBP trend (7.3 mm Hg increase in men, 8.4 mm Hg in women) would lead to the highest increase in stroke and combined cardiovascular disease 6.8 million incremental cardiovascular disease events in men, and 4.2 million in women, but a declining SBP trend ( low scenario) would reduce events in almost equal proportion. Rising BMI was projected to increase CHD and stroke events 6% and 5% in men, and 7% and 3% in women, respectively, mediated through SBP, TC, diabetes, and HDL. Main trends in SBP, TC, diabetes, and smoking combined would lead to an additional 13.2 million more cardiovascular disease events (13% increase) in Chinese men and an additional 9.7 million additional cardiovascular disease events (14% increase) in Chinese women over 2010 to 2030, despite a decreasing smoking trend. Though demographic trends would account for 68% of increases in annual cardiovascular disease 2010 to 2030, unfavorable cardiovascular disease risk factor trends would accelerate crude event rates (Figure 2A and 2B and Figure 3). The magnitude of the impact of the Figure 2. A and B, Number of ischemic strokes, hemorrhagic strokes, and CHD events in Chinese men and women ages 35 to 84 years projected from the CHD Policy Model China for the years 2010, 2015, 2020, 2025, and Dark blue areas represent events projected due to aging and population growth alone; red areas, additional events attributable to projected SBP, TC, diabetes, and smoking trends. combined main trends in risk factors on event rates was greatest for ischemic stroke. Increased CHD in Chinese men from rising TC was blunted by concurrent decline in active smoking. Projection of recent declines in active smoking in Chinese men would not counterbalance the cardiovascular consequences of increasing SBP, TC, diabetes, or BMI. However, a 0.6 percentage point annual decline in active smoking (main assumption) in Chinese men would prevent almost 1 million noncardiovascular disease deaths, such as cancer and chronic obstructive lung disease deaths. An increase in active smoking prevalence in women to 19% by the year 2030 the worst case would lead to a 1% increase in CHD and stroke deaths, and an 1% increase in all-cause mortality over 2010 to 2030 because most of the substantial adverse effects would occur after Sensitivity Analyses When Framingham risk factor coefficients were substituted for the main CMCS coefficients, projected changes in CHD and stroke events varied according to differences in the strengths of the coefficients between the 2 studies (Figure 4A and 4B).

6 248 Circ Cardiovasc Qual Outcomes May 2010 Downloaded from by guest on June 19, 2018 Figure 3. Crude event rates (per ) of ischemic stroke, hemorrhagic stroke, and CHD in Chinese men and women ages 35 to 84 years projected for 2010, 2015, 2020, 2025, and Base case indicates risk factors held at year 2000 levels; risk factor trend case indicates main SBP, TC, diabetes, and smoking trends. Assuming lower US case-fatality rates starting in 2010 would lower cardiovascular mortality in the base case by approximately 25% with a small increase in event rates due to more repeat events (Table 3, supplemental Appendix Table 7). Lower case-fatality would blunt cardiovascular mortality increases from projected trends in SBP, TC, and diabetes. When an aggressive lowering of smoking prevalence in men was simulated, avoided cardiovascular and noncardiovascular deaths were more than twice the main smoking trend simulation, and prevented noncardiovascular deaths would counterbalance cardiovascular death increases from unfavorable SBP, TC, diabetes, and BMI trends, leading to reduced male all-cause mortality. Lowering mean SBP 3.6 mm Hg in 2010 would be even more effective, lowering CHD, stroke, and noncardiovascular mortality in men and women. Blood pressure lowering would reduce cardiovascular and noncardiovascular mortality even if TC and diabetes increased and smoking prevalence stayed at the year 2000 level. Allowing TC mean to peak at 6.0 mmol/l by 2030 or assuming the diabetes high trend coupled with Framingham diabetes coefficients led to much larger projected increases in cardiovascular events. Simulating effects of BMI not mediated by SBP, TC, HDL, or diabetes increased projected cardiovascular disease events an additional 6% in men and 4% in women. Discussion Even if risk factors stay at year 2000 levels, annual cardiovascular disease events in China probably will increase by more than a half between 2010 to 2030 due to aging and population growth. We forecast that projected cardiovascular risk factor trends will increase cardiovascular events by approximately an additional 14% in Chinese adults from 2010 to 2030, above and beyond demographic effects. The recent rate of decline in smoking will not be sufficient to counteract approximately 26 million cardiovascular disease events and nine million cardiovascular deaths added by deleterious trends in SBP, TC, diabetes, and BMI. We projected that an aggressive tobacco control policy lowering active smoking prevalence to 20% by 2020 and 10% by 2030 would produce a reduction in total mortality in Chinese men despite adverse trends in other risk factors. Only lowering SBP across the adult population would reduce cardiovascular and noncardiovascular deaths in men and women. We projected that unfavorable trends in SBP, TC, diabetes, and BMI would substantially augment cardiovascular disease event rates, and especially so for ischemic stroke. Chinese surveys have documented consumption of more dietary fats, 1 overnutrition, 36 and less physical activity 3 over time. Additionally, relatively few Chinese adults with dyslipidemia, 37 high blood pressure, 38 or diabetes 22 are aware of these risk factors. Zhao et al 13 found a

7 Moran et al Risk Factor Trends and Future CVD in China 249 Downloaded from by guest on June 19, 2018 Figure 4. A and B, Sensitivity Analyses with CHD and total stroke risk coefficients estimated from Framingham Heart Study data substituted for China Multi-provincial Cohort Study (CMCS) coefficients, and main trend simulations repeated. Bars represent incremental percent change compared with the base case. transition toward increased ischemic stroke and decreased hemorrhagic strokes in Beijing. 38 In our model, less hemorrhagic stroke coupled with increased ischemic stroke occurred only if we simulated a modest SBP decline and large TC and diabetes increases. Because of the predominance of stroke in China and the strong association between blood pressure and stroke, optimistic blood pressure trend and intervention scenarios reduced cardiovascular and noncardiovascular outcomes most dramatically. If BMI has cardiovascular effects not mediated by SBP, total cholesterol, diabetes, HDL, 39,40 or effects mediated by factors not modeled here, 41 BMI would be on par with SBP and TC as a driver of adverse cardiovascular disease trends. The Chinese government taxes tobacco products and has achieved a steady though slight decline in smoking. Only an extremely aggressive approach to tobacco control would prevent at least 4.5 million deaths from all causes in men from 2010 to 2030, and keep all-cause mortality from rising despite expected increased cardiovascular deaths. A stronger tobacco taxation policy could save millions of lives, and generate government revenues that would eclipse losses to industry and tobacco farmers. 42 We assumed that increasing TC will increase CHD. CHD incidence declined in Japan despite a 0.5 mmol/l (20 mg/dl) mean rise in TC in adults between 1980 and 2000, presumably in part because SBP and smoking decreased, elevated cholesterol requires a long incubation period, 43 or TC does not capture unique dietary influences or subfraction changes. Cardiovascular disease death rates usually decline with economic development, a trend slowed but not reversed by unfavorable cholesterol trends. 44 We simulated 1 driver of decline in deaths with economic development by immediately improving case-fatality lower case-fatality would lead to 25% fewer cardiovascular deaths in the base case and blunt cardiovascular mortality increases from unfavorable risk factor trends. Assuming the higher diabetes prevalence or stronger diabetes coefficients resulted in two thirds to twice more the projected cardiovascular disease events compared with the main assumption fasting glucose-only diabetes definition of diabetes and CMCS diabetes coefficients. CMCS diabetes risk coefficients are weak compared with other studies, 7,45 perhaps due to underdiagnosis or inclusion of predominantly mild cases of diabetes. Prior Markov-style population models of cardiovascular disease in China used risk factor relative risks from Western and Asian cohort studies 10,46 or China-specific risk equations. 47 The accuracy of recalibrated Framingham prediction

8 250 Circ Cardiovasc Qual Outcomes May 2010 Table 3. Incremental Changes in Projected CHD and Stroke Events Projected With Optimistic and Pessimistic Scenarios for TC, SBP, Smoking, and Diabetes Assumptions, Chinese Adults Ages 35 to 84 Years, 2010 to 2030, the CHD Policy Model China Men Women Downloaded from by guest on June 19, 2018 Simulation CHD Events Stroke Events CHD Events Stroke Events Base case (totals) Trend cases: incremental events absolute and (%) change from main base case; except low case-fatality risk factor trend scenario compared with low case-fatality base case Optimistic scenarios Base case with lower case-fatality starting (totals)* Main trends SBP, TC, diabetes lower (13) (13) (19) (14) case-fatality starting 2010 Smoking Aggressive anti-tobacco policy ( 23) ( 2) SBP, TC, diabetes main trends ( 6) (10) aggressive anti-tobacco SBP Lower SBP 3.6 mm Hg ( 5) ( 10) ( 3) ( 9) TC, diabetes, smoking trends, lower (3) ( 6) (10) 2 437,000 ( 5) SBP 3.6 mm Hg TC, diabetes, lower SBP 3.6 mm Hg, (12) ( 5) (11) ( 5) no smoking change Pessimistic scenarios TC Increase ceiling to 6.0 mmol/l (48) (9) (24) (9) Diabetes Diabetes high trend and Framingham Heart (14) (3) (13) (6) Study diabetes coefficients BMI Main (mediated by SBP, TC, HDL, diabetes) (6) (5) (7) (3) Additional nonmediated effects (10) (5) (3) (5) *Main effect of lower case-fatality was approximately 25% decrease in CHD and stroke mortality (see Appendix Tables 7a and 7b). Incremental change compared with the lower case-fatality base case. equations for Chinese populations remains controversial. 6 8 Our simulations substituting Framingham coefficients for the CMCS coefficients yielded CHD and stroke projections that varied from the main projections by up to 16 percentage points. Stroke projections varied mostly because TC was not a significant predictor of total stroke in Framingham. 48 Stroke predictions were more detailed and probably more accurate using China-specific stroke equations, but there was uncertainty regarding whether CMCS or Framingham CHD diabetes and cholesterol coefficients should be used. limited past survey data. Artificial ceilings limiting highest future risk factor levels may be overly conservative: on Nauru, diabetes prevalence already exceeds 30%, 25 and total cholesterol was as high as 7.0 mmol/l in 1970s Finland. 49 On the other hand, generalizing the rapid rise in total cholesterol observed in the urban Beijing population 19 to all of China may have led to overestimation. For this analysis, for the sake of simplicity, uncertainty about trend projections was tested using only 1-way sensitivity analyses, which are limited compared with multiway analyses. Limitations Aging and growth of the Chinese population are certain, but the trends projected here were based on limited survey data gathered since China s economic reforms after 1979 and remain uncertain. Much hinges on future rates of economic development and urbanization. HDL was not modeled (except as an indirect product of BMI), nor was widespread passive smoking exposure in Chinese women, both due to Implications In this computer modeling study, unfavorable trends in SBP, TC, and diabetes from 2010 to 2030 were projected to increase cardiovascular disease events by approximately 14% above and beyond the increase expected due to aging and population growth, even if active cigarette smoking continues the recent rate of decline. Population-wide risk reduction policies, screening for cardiovascular disease

9 Moran et al Risk Factor Trends and Future CVD in China 251 Downloaded from by guest on June 19, 2018 risk factors, and scaling up of successful local risk factor prevention and treatment programs should be included in China s health system reform. Even if other adverse risk factor trends continue unabated, national policy targeted toward aggressive tobacco control policy or blood pressure lowering could save 2.9 to 5.7 million lives during the next 20 years. Acknowledgments The authors thank the many investigators and participants who contributed to the surveys of cardiovascular risk factors in China over the years 1980 to 2008 reviewed. We particularly thank investigators and participants from the Chinese Multiprovincial Cohort Study for contributing the risk factor relative risks and the International Collaborative Study of Cardiovascular Disease in Asia Study and its participants for providing risk factor means and prevalence in China. We thank the China Health and Nutrition Survey and its participants, funded by NIH (R01- HD30880, DK056350, and R01-HD38700), and the Carolina Population Center and the Chinese Centers for Disease Control for providing the primary data for trends in blood pressure, BMI, and smoking. The Framingham Heart Study (FHS) and Framingham Offspring Study (FOS) are conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with the FHS and FOS Investigators. This article was prepared using a limited access dataset obtained by the NHLBI and does not necessarily reflect the opinions or views of the FHS, the FOS, or the NHLBI. Sources of Funding This study was supported by a grant from the Flight Attendants Medical Research Institute and a grant from the Swanson Family Fund to the University of California, San Francisco (to Dr Goldman), Mentored Career Development Award number K08HL from the United States National Heart, Lung, and Blood Institute of the NIH, and a grant from the William J. Matheson Foundation to Columbia University (to Dr Moran). None. Disclosures References 1. Drewnowski A, Popkin BM. The nutrition transition: new trends in the global diet. Nutr Rev. 1997;55: Wang LD. Comprehensive report, Chinese nutrition and health survey in Beijing: People s Medical Publishing House; Paeratakul S, Popkin BM, Keyou G, Adair LS, Stevens J. Changes in diet and physical activity affect the body mass index of Chinese adults. Int J Obes Relat Metab Disord. 1998;22: Wildman RP, Gu D, Muntner P, Wu X, Reynolds K, Duan X, Chen CS, Huang G, Bazzano LA, He J. Trends in overweight and obesity in Chinese adults: between 1991 and Obesity (Silver Spring, Md). 2008;16: Gu D, Wu X, Reynolds K, Duan X, Xin X, Reynolds RF, Whelton PK, He J. Cigarette smoking and exposure to environmental tobacco smoke in China: the international collaborative study of cardiovascular disease in Asia. Am J Public Health. 2004;94: Liu J, Hong Y, D Agostino RB Sr, Wu Z, Wang W, Sun J, Wilson PW, Kannel WB, Zhao D. Predictive value for the Chinese population of the Framingham CHD risk assessment tool compared with the Chinese Multi- Provincial Cohort Study. JAMA. 2004;291: Wu Y, Liu X, Li X, Li Y, Zhao L, Chen Z, Li Y, Rao X, Zhou B, Detrano R, Liu K. Estimation of 10-year risk of fatal and nonfatal ischemic cardiovascular diseases in Chinese adults. Circulation. 2006;114: Barzi F, Patel A, Gu D, Sritara P, Lam TH, Rodgers A, Woodward M. Cardiovascular risk prediction tools for populations in Asia. J Epidemiol Commun Health. 2007;61: Leeder S, Raymond S, Greenberg H, Liu H, Esson K. Race Against Time: The Challenge of Cardiovascular Disease in Developing Countries. New York, NY: Columbia University Center for Global Health and Economic Development; Cheng J, Zhao D, Zeng Z, Critchley JA, Liu J, Wang W, Sun J, Capewell S. The impact of demographic and risk factor changes on coronary heart disease deaths in Beijing, BMC Public Health. 2009;9: Moran A, Zhao D, Gu D, Coxson P, Chen CS, Cheng J, Liu J, He J, Goldman L. The future impact of population growth and aging on coronary heart disease in China: projections from the coronary heart disease policy model-china. BMC Public Health. 2008;8: He J, Neal B, Gu D, Suriyawongpaisal P, Xin X, Reynolds R, MacMahon S, Whelton PK. International collaborative study of cardiovascular disease in Asia: design, rationale, and preliminary results. Ethn Dis. 2004;14: Zhao D, Liu J, Wang W, Zeng Z, Cheng J, Liu J, Sun J, Wu Z. Epidemiological transition of stroke in China: twenty-one-year observational study from the Sino-MONICA-Beijing Project. Stroke. 2008;39: Gu D, Kelly TN, Wu X, Chen J, Duan X, Huang JF, Chen JC, Whelton PK, He J. Blood pressure and risk of cardiovascular disease in Chinese men and women. Am J Hypertens. 2008;21: He J, Gu D, Wu X, Reynolds K, Duan X, Yao C, Wang J, Chen CS, Chen J, Wildman RP, Klag MJ, Whelton PK. Major causes of death among men and women in China. N Engl J Med. 2005;353: Stevens J, Truesdale KP, Katz EG, Cai J. Impact of body mass index on incident hypertension and diabetes in Chinese Asians, American Whites, and American Blacks: the People s Republic of China Study and the Atherosclerosis Risk in Communities Study. Am J Epidemiol. 2008;167: China Health and Nutrition Survey. [cited 2007 October 31]; Available at: Health, United States, [cited 2009 November 12]; Available at: Liu S, Zhao D, Wang W, Liu J, Qin L, Zeng Z, Wu Z. The trends of cardiovascular risk factors in urban and rural areas of Beijing during J Cardiovasc Pulmon Dis. 2006;25: Li Z, Yang R, Xu G, Xia T. Serum lipid concentrations and prevalence of dyslipidemia in a large professional population in Beijing. Clin Chem. 2005;51: World Health Organization. Global Infobase. [cited 2008 November]. Available at: Gu D, Reynolds K, Duan X, Xin X, Chen J, Wu X, Mo J, Whelton PK, He J. Prevalence of diabetes and impaired fasting glucose in the Chinese adult population: International Collaborative Study of Cardiovascular Disease in Asia (InterASIA). Diabetologia. 2003;46: Pan XR, Yang WY, Li GW, Liu J. Prevalence of diabetes and its risk factors in China, National Diabetes Prevention and Control Cooperative Group. Diabetes Care. 1997;20: Wang KLT, Xiang H. Study on the epidemiology characteristics of diabetes mellitus and IGT in China. Chin J Epidemiol. 1998;19: International Diabetes Federation. Diabetes Atlas. 2nd edition. [cited 2009 February]. Available at: Atlas% Summary.pdf. 26. Qiao Q, Hu G, Tuomilehto J, Nakagami T, Balkau B, Borch-Johnsen K, Ramachandran A, Mohan V, Iyer SR, Tominaga M, Kiyohara Y, Kato I, Okubo K, Nagai M, Shibazaki S, Yang Z, Tong Z, Fan Q, Wang B, Chew SK, Tan BY, Heng D, Emmanuel S, Tajima N, Iwamoto Y, Snehalatha C, Vijay V, Kapur A, Dong Y, Nan H, Gao W, Shi H, Fu F. Age- and sex-specific prevalence of diabetes and impaired glucose regulation in 11 Asian cohorts. Diabetes Care. 2003;26: Wang Y, Beydoun MA. The obesity epidemic in the United States: gender, age, socioeconomic, racial/ethnic, and geographic characteristics: a systematic review and meta-regression analysis. Epidemiol Rev. 2007; 29: Ezzati M, Hoorn SV, Rodgers A, Lopez AD, Mathers CD, Murray CJ. Estimates of global and regional potential health gains from reducing multiple major risk factors. Lancet. 2003;362: Gu D, He J, Duan X, Reynolds K, Wu X, Chen J, Huang G, Chen CS, Whelton PK. Body weight and mortality among men and women in China. JAMA. 2006;295:

10 252 Circ Cardiovasc Qual Outcomes May 2010 Downloaded from by guest on June 19, Chen Z, Peto R, Collins R, MacMahon S, Lu J, Li W. Serum cholesterol concentration and coronary heart disease in population with low cholesterol concentrations. BMJ (Clin Res Ed). 1991;303: Eastern Stroke and Coronary Heart Disease Collaborative Research Group. Blood pressure, cholesterol, and stroke in eastern Asia. Lancet. 1998;352: He FJ, MacGregor GA. How far should salt intake be reduced? Hypertension. 2003;42: Vaccarino V, Krumholz HM, Yarzebski J, Gore JM, Goldberg RJ. Sex differences in 2-year mortality after hospital discharge for myocardial infarction. Ann Intern Med. 2001;134: Rosamond WD, Folsom AR, Chambless LE, Wang CH, McGovern PG, Howard G, Copper LS, Shahar E. Stroke incidence and survival among middle-aged adults: 9-year follow-up of the Atherosclerosis Risk in Communities (ARIC) cohort. Stroke. 1999;30: International Database, US Census Bureau [cited 2007 May 17]. Available at: Zhai F, Wang H, Wang Z, Popkin BM, Chen C. Closing the energy gap to prevent weight gain in China. Obes Rev. 2008;(Suppl 1): He J, Gu D, Reynolds K, Wu X, Muntner P, Zhao J, Chen J, Liu D, Mo J, Whelton PK. Serum total and lipoprotein cholesterol levels and awareness, treatment, and control of hypercholesterolemia in China. Circulation. 2004; 110: Gu D, Reynolds K, Wu X, Chen J, Duan X, Muntner P, Huang G, Reynolds RF, Su S, Whelton PK, He J. Prevalence, awareness, treatment, and control of hypertension in china. Hypertension. 2002; 40: Davi G, Guagnano MT, Ciabattoni G, Basili S, Falco A, Marinopiccoli M, Nutini M, Sensi S, Patrono C. Platelet activation in obese women: role of inflammation and oxidant stress. JAMA. 2002;288: Rosito GA, D Agostino RB, Massaro J, Lipinska I, Mittleman MA, Sutherland P, Wilson PW, Levy D, Muller JE, Tofler GH. Association between obesity and a prothrombotic state: the Framingham Offspring Study. Thromb Haemost. 2004;91: Koch R, Sharma AM. Obesity and cardiovascular hemodynamic function. Curr Hypertens Rep. 1999;1: Hu TW, Mao Z. Effects of cigarette tax on cigarette consumption and the Chinese economy. Tob Control. 2002;11: Iso H. Changes in coronary heart disease risk among Japanese. Circulation. 2008;118: Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to PLoS Med. 2006;3:e Lawes CM, Parag V, Bennett DA, Suh I, Lam TH, Whitlock G, Barzi F, Woodward M. Blood glucose and risk of cardiovascular disease in the Asia Pacific region. Diabetes Care. 2004;27: Critchley J, Liu J, Zhao D, Wei W, Capewell S. Explaining the increase in coronary heart disease mortality in Beijing between 1984 and Circulation. 2004;110: Qin X, Jackson R, Marshall R, Lee L, Cao W, Zhan S, Hu Y. Modelling the potential impact of population-wide and targeted high-risk blood pressure-lowering strategies on cardiovascular disease in China. Eur J Cardiovasc Prev Rehabil. 2009;16: Wolf PA, D Agostino RB, Belanger AJ, Kannel WB. Probability of stroke: a risk profile from the Framingham Study. Stroke. 1991;22: Puska P, Tuomilehto J, Salonen J, Neittaanmaki L, Maki J, Virtamo J, Nissinen A, Koskela K, Takalo T. Changes in coronary risk factors during comprehensive five-year community programme to control cardiovascular diseases (North Karelia project). BMJ. 1979;2:

11 Downloaded from by guest on June 19, 2018 Future Cardiovascular Disease in China: Markov Model and Risk Factor Scenario Projections From the Coronary Heart Disease Policy Model China Andrew Moran, Dongfeng Gu, Dong Zhao, Pamela Coxson, Y. Claire Wang, Chung-Shiuan Chen, Jing Liu, Jun Cheng, Kirsten Bibbins-Domingo, Yu-Ming Shen, Jiang He and Lee Goldman Circ Cardiovasc Qual Outcomes. published online May 4, 2010; Circulation: Cardiovascular Quality and Outcomes is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX Copyright 2010 American Heart Association, Inc. All rights reserved. Print ISSN: Online ISSN: The online version of this article, along with updated information and services, is located on the World Wide Web at: Data Supplement (unedited) at: Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published in Circulation: Cardiovascular Quality and Outcomes can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office. Once the online version of the published article for which permission is being requested is located, click Request Permissions in the middle column of the Web page under Services. Further information about this process is available in the Permissions and Rights Question and Answer document. Reprints: Information about reprints can be found online at: Subscriptions: Information about subscribing to Circulation: Cardiovascular Quality and Outcomes is online at:

12 SUPPLEMENTAL MATERIAL (Technical Appendix) Supplemental Methods Age Assumptions for the analysis Risk factor levels at baseline are age-specific estimates from the International Collaborative Study of Cardiovascular Disease in Asia (InterASIA). The age-trend for each risk factor (e.g. higher SBP with higher age, lower cholesterol in oldest age category) was preserved by a risk factor transition function in the Model software: all base year (year 2000) risk factor levels are age- and sex-specific and taken from the nationally-representative InterASIA study (male and female and age categories 35-44, 45-54, 55-64, 65-74, and years). The CHD Policy Model software includes a function that ensures that the age-trend in risk factor levels is preserved as age cohorts age upward and new waves of 35 year olds enter with each successive year of the model simulation. Transfers from one risk factor level to another are included to ensure that the InterASIA proportions of the population with each risk factor level are maintained. Transfers from one risk factor level to another were included to preserve the InterASIA Study proportions of the population with each risk factor level. For example, the proportion of year old men with low (<100 mg/dl) LDL cholesterol is For year old men the proportion is The shift toward higher LDL cholesterol levels is most likely caused by increasing LDL levels as people age. In higher age ranges, this trend reverses, so that by age 75-84,

13 the proportion is The change in the upper age ranges is most likely due to a more complex array of factors, including the fact that people with higher risk are more likely to die. Annual transfer rates between risk factor levels were calculated to reduce the low risk population from to over 10 years, without regard to the reason for the change, but taking into account the effect of the Model s CHD incidence and non-chd death rates. As the age composition of the population changes (specifically, in China the population shifts to a relatively larger proportion of the population in older age groups), risk factor exposures change due to a larger proportion of the population exposed to the risk factor levels characteristic of the older age categories and a smaller proportion exposed to risk factor levels characteristic of the younger population. Significant (P < 0.05) age*risk factor risk coefficient interactions (higher risk at higher ages) were observed for smoking in China Multiprovincial Cohort Study (CMCS) multivariate CHD models, SBP, smoking, and diabetes in ischemic stroke models, smoking and SBP for hemorrhagic stroke models and smoking and diabetes in non-cardiovascular mortality models, so these were incorporated in age-specific risk coefficients. Analysis of risk factor time trends in the China Health and Nutrition Survey (CHNS) period trends in SBP and BMI showed time*age interactions (smaller increases over time in older persons), so we assumed smaller increases in SBP and BMI with each higher 10-year age category. No age*time interaction was observed in the CHNS data for proportional decline in smoking prevalence (perhaps because of relatively

14 lower prevalence in the oldest age categories). For total cholesterol and diabetes we lacked sufficient longitudinal data in order to assess for age*time interactions, so the relative increases in total cholesterol and diabetes were assumed uniform across age categories (though added to age-specific means or proportions). Modeling of stroke in the CHD Policy Model-China The objective of the CHD Policy Model-China stroke analysis was to predict hemorrhagic and ischemic strokes in Chinese men and women over the years Base case stroke projections from the CHD Policy Model-China are dependent on the incidence assumptions, and risk factor trends are additionally dependent on risk factor relative risk assumptions. Incident stroke rates were identified from three Chinese studies the China Multiprovincial Cohort Study (CMCS), the Beijing Sino-MONICA Study 1, and the China Hypertension Epidemiology Follow Up Study (CHEFS). 2 All studies defined strokes by ICD-9 codes Age- and sex-specific total stroke incidences for Chinese adults aged years old were estimated from the Sino-MONICA study, the CMCS, and the CHEFS. Twenty-eight day stroke case fatality rates were based on Sino-MONICA case-fatality rates averaged over the years We assumed a total mortality envelope of all-cause mortality in China for the year 2002 reported by the Global Burden of Disease Study, 3 and the proportion of total deaths attributed to stroke reported by the CHEFS. 4 Keeping

15 case-fatality constant, stroke incidence was calibrated to reproduce the number of stroke deaths and the proportion of stroke of total deaths expected for the year 2002 (Appendix Table 1). The CMCS was a cohort study of 30,121 male and female participants aged years and with no cardiovascular disease at baseline in (more details about the CMCS provided in a manuscript by Liu et al). 5 These participants were recruited from 16 centers in 11 Chinese provinces using a multistage sampling method. Twelve centers (80.3% of participants) were in urban areas and the remainder in rural areas. Overall baseline participation rate was 82%. Baseline measurement of risk factors using a standard protocol (WHO-MONICA protocol) 6 and blood samples were processed at a central laboratory. Case-finding of new CHD and stroke events and noncardiovascular deaths was first done by face-to-face interview. Events were ascertained by 1) detailed interview of participants or family members, 2) review of hospital records. These events were later adjudicated by investigators at the Beijing Institute for Heart, Lung, and Blood Vessel Diseases. After 1996, six centers ceased follow up because of completion of that national research project, but the remaining 10 centers (16,552 participants) were followed up through the end of Follow up rate was 86% for the centers followed all of , and 65% of the original 16 center cohort. Multivariate Cox proportional hazard ratios for systolic blood pressure (SBP), diabetes mellitus, total cholesterol, BMI, and active cigarette smoking were estimated from

16 baseline measurements and ischemic and hemorrhagic events occurring over 159,400 person-years of observation in CMCS participants aged years (Appendix Table 2). 5 Statistically significant age*risk factor interactions were found for SBP, smoking, and diabetes for ischemic stroke and smoking and SBP for hemorrhagic stroke, so these interactions were modeled in the Policy Model risk coefficients. For simplicity, this manuscript reports the sum of ischemic stroke and hemorrhagic stroke events together as total strokes in many instances. The CHD Policy Model-China stroke-specific mortality estimate for 2002 for ages years (1,573,000) was close to the 2002 Global Burden of Disease estimate for the same age range (1,485,000). 7 The similarity between the Model s results and the GBD results for stroke does not hold for CHD in our past analysis, 8 assuming the proportion of total deaths attributable to CHD observed in CHEFS, the Policy Model-China predicted fewer CHD deaths for China in 2002 (486,000) compared with the Global Burden of Disease estimate for the same year (590,000). It is may be that the difficulty involved in assigning sudden and/or unwitnessed deaths to an ischemic heart disease ICD code leads to variability in classifying CHD deaths across surveys, while stroke deaths are more easily recognized and categorized. The epidemiology of stroke in China is complex, with regional variations, 9 and recent evidence from Beijing Sino-MONICA that there has been a shift toward an overall lower stroke case-fatality rate. 1 Therefore our current model of stroke is very simplified: we assume uniform national incidence and assumed

17 a constant case-fatality rate into the future. As is evidenced in the attributable proportion analysis, a substantial proportion of stroke events predicted by the Policy Model-China were unexplained by SBP, total cholesterol, HDL cholesterol, active and passive smoking, BMI or diabetes. This may be explained by the absence of information in the current Policy Model on atrial fibrillation, alcohol consumption, and other important stroke risk factors. Estimation of risk factor beta coefficients from the China Multiprovincial Cohort Study (CMCS) and the Framingham Heart Study Estimation of multivariate Cox proportional hazard functions with CHD, ischemic stroke, hemorrhagic stroke, and non-cardiovascular death as outcomes from the CMCS is described above on Appendix page A4 and A5 above and in a publication by Liu et al. (CHD). 5 Age*risk factor interactions modeled in Policy Model coefficients are described on Appendix page A2. Beta coefficients from the Framingham Heart Study were determined for a 60-yearold, the average age of the first onset of CHD events, stroke events, or noncardiovascular disease deaths in individuals in examinations 9 to 13, 24, and 25 from the original Framingham cohort and 1-6 from the Framingham offspring cohort, for whom adequate data were available for a time-dependent logistic regression analysis. 10 Main stroke risk factor coefficient estimates are from Wolf et al. 11 Standard errors for the stroke risk factor coefficients were calculated

18 from original Framingham data, then used to calculate 95% intervals for the estimates in Appendix Table 2. The Asian cohorts from the Asia Pacific Cohort Studies collaboration (APCSC) have been the source of the largest pooled estimates of risk factor relative risks for cardiovascular disease in Asian popultions. 12 Similar to the APCSC, the CMCS coefficients generally showed a stronger association between SBP and events in men and women, and a lower association between total cholesterol and events in men compared with Framingham Heart Study estimates. 13 On the other hand, CMCS diabetes coefficients were weaker than both the Framingham coefficients (Table 1), and APCSC coefficients. 14 The weaker diabetes coefficients estimated from the CMCS may be due to a preponderance of participants with less severe or more recently diagnosed diabetes, the limitations of a single one off glucose measurement for diagnosis, 14 or other unknown measurement error. Comparing the cardiovascular disease risk factor coefficients between three major Asian cohort studies (CMCS, the PRC-USA Study, 15 and APCSC 13 ) and the Framingham Heart Study led us to only two conclusions: the effect size for systolic blood pressure was generally larger for all of the Asian cohorts compared with Framingham, and the CMCS diabetes coefficients were generally weaker compared with all of the other studies. Attributable Risk Analysis

19 Annual risk for CHD, ischemic stroke or hemorrhagic stroke events and non-cvd deaths is calculated for each model cell by multivariate logistic regression equations. Therefore annual risk for events is determined by the age, sex, and risk factor relative levels assigned to that cell, and the combined multiplicative effect of the risk factor coefficients (Appendix Table 2). Simulations of individual risk factor reductions were used to determine the relative magnitude of the risk factors studied, but these individual population attributable fractions (PAFs) are not reported because the objective of the analysis was not to assess the hypothetical effect of removing single exposures but to give the reader a sense of the relative contributions of the major cardiovascular risk factors in China. Individual risk factor simulations were necessary for gauging relative magnitude because we are not able to decompose the contribution of individual risk factors from Model output resulting from a simultaneous multiple risk factor reduction simulation. The overall proportion of cardiovascular events explained by the selected major risk factors was obtained by setting all of the risk factors to the minimum risk exposure level for the entirety of the 10-year simulation and comparing with a base case in which risk factor levels stayed constant at year 2000 levels. The method for calculating the change in risk is quite similar to the multiple risk factor PAF methods developed by Yusuf et al. in INTERHEART. Risk for events in each model cell predicted by logistic function Formula 1:

20 Rate of CHD or stroke in cell = exp(α + Σ β RF *rskave RF )/[1+exp(α + Σ β RF *rskave RF )] where α = age-specific rate of disease in the overall population (intercept), β = age-specific risk coefficient, and RF = age-specific risk factor mean. Thus, in this method the effect of combined and simultaneous risk factor reductions is multiplicative of the relative risks, after adjusting for correlation between risk factors. As in INTERHEART, the multiple risk factor PAF is constrained in a logistic model approach because percent of risk for events explained can approach but not reach 100%. The proportion of incident CHD and total stroke attributable to SBP, TC, HDL cholesterol, combined active and passive smoking, and diabetes (and separately for mediated BMI effects) were estimated for the base case simulation for the years A baseline with risk factors at 2000 levels was compared with a simulation in which all risk factors were simultaneously set at optimal levels 16 (zero smoking and diabetes exposures and levels of BMI, 17 cholesterol, 18 and blood pressure, 19 with lowest risk for events found in large epidemiologic studies, Figure 1). SBP > 115 mm Hg explained 16% of CHD in men, 10% of CHD in women, 33% of stroke in men, 25% of stroke in women, 29% of CVD in men, and 23% of CVD in women. Active smoking explained 28% of CHD in men, 7% of CHD in women, 7% of stroke in men, 3% of stroke in women, 11% of CVD in men, 3% of CVD in women. Passive smoking explained 1% of CHD in men, 6% of CHD in women, 3% of stroke in men, 16% of stroke in women, 2% of CVD in men, and 14% of CVD in women. Diabetes (diagnosis or fasting glucose

21 7.0 mmol/l) explained 1% of CHD in men, 6% of CHD in women, 3% of stroke in men, 2% of stroke in women, 2% of CVD in men, and 2% of CVD in women.. Total cholesterol > 3.8 mmol/l explained 17% of CHD in men, 14% of CHD in women, 9% of stroke in men, 11% of stroke in women, 11% of CVD in men, and 12% of CVD in women.. HDL cholesterol < 1.03 mmol/l explained 4% of CHD in men, 19% of CHD in women, 6% of stroke in men, 3% of stroke in women, 5% of CVD in men, and 5% of CVD in women. BMI >21 kg/m² (mediated through SBP, total cholesterol, HDL, and diabetes) explained 10% of CHD in men, 20% of CHD in women, 6% of stroke in men, 7% of stroke in women, 7% of CVD in men, and 9% of CVD in women. For CHD, 34% in men, 39% in women was unexplained by these risk factors at these levels, though some of CHD that is unexplained may be accounted for by other risk factors, and even lower levels of exposure for some risk factors may contribute to at lower levels. For stroke 40% was unexplained for men and women. Systematic Review of past CVD risk factor surveys in China (Appendix Figure 1) In order to capture changes that have occurred following China s period of economic development, past cardiovascular disease (CVD) risk factor trends were estimated from published surveys of CVD risk factors from China over the years Before reviewing past Chinese surveys, it was decided a priori that studies included would meet the following criteria: population-based

22 observational CVD risk factor surveys that conducted multiple samples of a population or cohort over time, plus single sample nationally representative or multi-provincial studies which sampled urban and rural areas of China. Longitudinal studies were selected for estimating time trends; nationallyrepresentative or multi-provincial single samples were used as references in order to examine the generalizability of means and proportions from smaller or regional studies. Selected surveys included adult men and women aged 20 years or above, and the minimum age range was years. Manuscripts published in either English or Chinese were eligible for inclusion. The Methods section of each study considered for inclusion was reviewed to ensure that risk factors were measured using internationally standard methods. The literature search for past epidemiologic surveys was conducted using MEDLINE via PubMed (U. S. National Library of Medicine, November, 2008, search terms listed in Appendix Table 3). Two search strategies were employed: first, a search was conducted using the broad Medical Subject Heading (MeSH) terms cardiovascular disease/epidemiology/etiology and restricting the search to the publication date range and to China using the keywords China and Chinese. Second, searches of the specific risk factors of interest were performed by entering the MeSH terms hypertension/epidemiology, hyperlipidemia/epidemiology OR cholesterol, smoking/epidemiology, diabetes/epidemiology, and obesity/epidemiology and restricting to China or Chinese as described above.

23 After eliminating duplicate titles, the resulting lists of manuscripts were combined into a single list of 1,699 titles. The initial list of titles was checked against the World Health Organization (WHO) Global Infobase ( list of studies pertaining to chronic disease risk factors in China. One investigator (AM) screened the resulting 1,699 abstracts in English or English translation for eligibility for review. The titles and abstracts of 98 manuscripts potentially fulfilled the eligibility criteria described above and were selected for review of the complete manuscripts. Of these 98 included manuscripts, 85 complete English language publications were reviewed and abstracted by one investigator (AM) and 13 complete Chinese language publications were reviewed and abstracted by a second investigator (JC). Review of the WHO Infobase added nine titles, predominantly data from multinational studies including Chinese cohorts, Chinese government publications, unpublished WHO World Health Survey data, 23 and some data from the 2002 Chinese Survey of Health and Nutrition 24 that were not captured by the PubMed search terms. Twenty-seven published peer-reviewed studies met the inclusion criteria described above, and were chosen for abstraction and consideration in trend estimates (Appendix Table 4). Additional data were found from on the WHO Infobase website (World Health Survey data, 23 and the

24 International Diabetes Federation website ( 25 Comparison with United States and Japanese National Survey Risk Factor Trends The shift away from a predominance of infectious and maternal/fetal illnesses and deaths toward non-communicable diseases and deaths as nations develop economically is termed the epidemiologic transition. 26, 27 Because the United States represents an example of a Western nation that has already gone through the epidemiologic transition, the same age- and sex-specific risk factor data for adults aged years were retrieved from the United States National Health and Nutrition Survey (NHANES, for total cholesterol, SBP, diabetes, and active smoking status). For the NHANES, diabetes was defined as physician diagnosis, taking antihyperglycemic medications, or a fasting glucose 7.0 mmol/l. National estimates of risk factor means or proportions were estimated from NHANES data by applying survey weights. Weighted population estimates from past NHANES were also available from Health, United States 2008 data tables. 28 Japan and the Republic of Korea (South Korea) represent nations with dietary and smoking patterns and a cardiovascular disease epidemiology (i.e. stroke more common than CHD) more similar to China, but which are further along in the epidemiologic transition than China. Past national risk factor means

25 and proportions for Japan and South Korea were obtained from the WHO Infobase. 23 Age-specific estimates from the US, Japan and Korea were used to calculate age-standardized estimates using the 2000 Chinese population as the standard. Age-standardized risk factor levels were then plotted in time trend figures for comparison and reference. Past risk factor surveys in China (Appendix Table 4, Appendix Figures 2a-2e) Systolic blood pressure, body mass index, and active smoking in China: estimates from the China Health and Nutrition Survey The CHNS is an ongoing multi-provincial household survey of urban and rural China conducted by the Carolina Population Research Center at the University of North Carolina, the Chinese Center for Disease Control and Prevention, and China s National Institute of Nutrition and Food Safety. The survey is a multi-stage, random cluster sample of nine Chinese provinces that sampled households in low and high income, urban and rural areas. Since the study s inception, some households have dropped out, and others have been added, with the result that SBP, BMI and current smoking status were recorded in some individuals once, and in others up to six times. Fasting serum cholesterol and fasting glucose were not measured in CHNS. In the most recent (2006) survey 4,400 households were sampled. Data are publicly available for download at

26 SBP trends in Chinese adults aged years were analyzed from the CHNS surveys. SBP was measured on the right arm by trained health workers using appropriately sized cuffs, with the participant in a seated position, and the blood pressure cuff at the level of the heart. Blood pressure was obtained three times at each survey visit by trained observers using a standard mercury sphygmomanometer. SBP was measured at the first appearance of a pulse sound (Korotkoff phase 1). For this analysis, SBP was the defined as the mean of the second two measurements. BMI (kg/m²) was calculated from CHNS weight and height data for each survey year. The methods for measuring weight and height in the CHNS are described in detail elsewhere. 29 In brief, weight was measured to the nearest 0.01 kg with a balance-beam scale, and height to the nearest 0.10 cm using a stadiometer, both following a WHO reference protocol. The prevalence of active cigarette smoking in the CHNS was defined as a positive response to two survey questions: Have you ever smoked cigarettes (including hand-rolled or device-rolled and Do you still smoke cigarettes now? Main linear trends in mean SBP and BMI and smoking prevalence over time were estimated using a random effects model that adjusted for age and assumed clustering at the level of the individual study participant (Appendix Figures 2a, 2b, and 2c). Analysis of risk factor time trends in the CHNS surveys period trends in SBP and BMI showed time*age interactions (smaller increases over time in older persons), so we assumed smaller increases in SBP and BMI with

27 each higher 10-year age category. No age*time interaction was observed in the CHNS data for proportional decline in smoking prevalence (perhaps because of relatively lower prevalence in the oldest age categories). The high SBP trend scenario used the upper limit of the 95% confidence interval of the estimated linear trend. Because CHNS surveys in 2004 and 2006 suggest a decline in mean SBP, the low SBP trend projected an annual mm Hg decline, similar to the decline in mean blood pressure observed in the U.S Based on evidence of an exponential increase in BMI in U.S. adults from , 30 an exponentially increasing trend was estimated for the high trend scenario. A log-linear trend was used for the low BMI trend. A linear declining trend of 0.58% annually in smoking prevalence for men and 0.13% in women estimated from the CHNS fit well with prevalence reported in past Chinese national smoking surveys , The lower bound of the 95% confidence interval for the main linear estimate was used to for the smoking trend best case and the upper bound for the worst case (still a decline) under current tobacco policies. For the women s worst case, active smoking was assumed to increase 0.30% annually to reach the smoking prevalence currently observed in Japanese women (approximately 15% in ages 35-64) by the year , 35 Studies of total cholesterol in China

28 Few past population-based surveys in China measured serum cholesterol. The Sino-MONICA study randomly sampled total cholesterol in Beijing area adults in 1984, 1988, 1993, 1996, and 1999, 36 with a slope of the Sino-MONICA total cholesterol trend largely dependent on the 1999 mean (Appendix Figure 2d). Even ignoring the 1999 measurement, a linear projection would rapidly result in mean values exceeding highest historical values observed in the U.S.( >5.6 mmol/l or >220 mg/dl in ages years in ). 37 Therefore logistic functions were fit to reproduce a steep early rise in cholesterol but attenuation approaching a ceiling mean of 5.7 mmol/l (220 mg/dl), the highest mean total cholesterol observed in Japanese 38 or South Korean 23 national surveys. The bestcase trend perpetuated the more conservative linear trend in mean total cholesterol between 1983 and 2002 measured in a Beijing professional population. 39 For total cholesterol we lacked sufficient longitudinal data in order to assess for age*time interactions, so the relative increases in total cholesterol and diabetes were assumed uniform across age categories. Other surveys of total cholesterol were reviewed for this analysis. Mean total cholesterol appeared to increase between the first and second China Studies (Appendix Figure 2d, China Study I, 1983 and China Study II, 1989), but this increase should be interpreted with caution as the method of pooling serum in that study prevents age-standardization, which was done with all the other data presented for this analysis [The China Study study design selected subjects to give approximately equal numbers in three year age groups (35-44, 45-54, 55-

29 64), personal communication, Jill Boreham, Oxford University August 5, 2008]. Changes in total cholesterol observed between to in the PRC-USA Study are likewise difficult to interpret given the different patterns observed in the two study populations. 40 Other single sample surveys from China were plotted for reference. Mean total cholesterol from the 2002 national survey were not plotted because results were reported in different age ranges, 41 preventing age standardization. The reported means for 15-44, 45-64, and >65 year olds suggest a mean lower than that reported by InterASIA. Studies of diabetes in China The WHO currently defines diabetes as a plasma glucose 11.1 mmol/l (200 mg/dl) two hours after an oral 75g glucose load (the Oral Glucose Tolerance Test or OGTT), or a fasting plasma glucose 7.0 mmol/l (126 mg/dl). 42 The WHO favors screening with the 2-hour oral glucose tolerance test (OGTT, defined as the level of plasma glucose 2 hours after a 75g glucose ingestion), as past studies have suggested that the fasting glucose criteria is insufficiently sensitive: results from the European DECODE study suggested that using the fasting glucose criteria alone would fail to diagnose 30-35% of persons with diabetes, especially in leaner individuals, 43 and may be a superior predictor of mortality risk 44 and the DECODA study suggested that OGTT may be the preferable to fasting glucose for screening for diabetes in Asian populations. 22 Most surveys in China screened for diabetes using OGTT. However, because the

30 both the Model s base year diabetes prevalence estimate from InterASIA 45 and because fasting glucose was the measure used for risk coefficients in the CMCS and Framingham studies (both of the cohort studies used for risk coefficients in this analysis) our main assumption was that diabetes was defined as a having a past diagnosis of diabetes, taking antidiabetes medications, or having a new diagnosis based on a fasting glucose 7.0 mmol/l. 45 Diabetes prevalence in China was assessed in three large, multi-provincial national surveys in 1980, , 47 and , 23, , 45 and repeated surveys in Da Qing City 47, 49 (Appendix Figure 2e). Even with accounting for the fact that surveys earlier than 2000 screened with OGTT and not fasting glucose, diabetes prevalence appears to have increased over time in China, mirroring the Japanese trend Because of the steep rise in diabetes prevalence and the a priori assumption of an upper limit of approximately 22% prevalence (based on the maximum in Japan), 23 logistic functions were used for the main and high diabetes projections (Appendix Figure 2e). The main assumption began with the prevalence of known and newly diagnosed diabetes from the InterASIA Study in 2000 (using fasting glucose; agestandardized 4.4% for men and 4.2% for women aged years), and simulated a rise to a ceiling of 15% in men and women by In order to estimate the additional percent prevalence of diabetes that would be estimated using OGTT in place of fasting plasma glucose, we started with age-specific prevalence of diabetes estimated using OGTT compared with

31 FPG in Chinese DECODA Study populations. After restricting to men and women in the year age range and age-standardizing to the 2000 Chinese Census, the highest additional % prevalence with OGTT was observed in the Shunyi population sample was 24% higher prevalence (2% absolute % higher) in men, and 39% higher prevalance (1% absolute % higher) in women. The high trend projection began in 2000 with the maximal relative increased prevalence of known and newly diagnosed diabetes using OGTT compared with fasting glucose testing reported from the Chinese Shunyi cohort (6.4% in men, 7.4% in women aged years), 22 then projected an increase to a prevalence of 20.4% in men and 21.5% in women by The low trend started with the InterASIA 2000 prevalence and used a more conservative linear trend similar to forecasts for China by King et al. 52 and the International Diabetes Federation. 25

32 Table 1. Stroke Inputs used for the CHD Policy Model-China Sex/age Men years Women years Hemorrhagic event rate/100, ,028 1, ,254 Hemorrhagic 28 day casefatality Hemorrhagic Mortality/100, Ischemic event rate/100, ,574 2, ,074 Ischemic 28 day casefatality Ischemic Mortality/100,

33 Appendix Methods Table 2. Relative risks for CHD and stroke estimated from three cohort studies: the China Multiprovincial Cohort Study (CMCS, , the primary coefficients for the CHD Policy Model-China) from China and the Framingham Heart Study from the United States (1948-present). Risk factor Age (5 years) Systolic BP (20 mm Hg) Total Cholesterol (1 mmol/l) HDL Cholesterol (1 mmol/l) Diabetes* (yes/no) Body Mass Index (3 kg/m²) Current smoking (yes/no) Males CHD CMCS Framingham Heart Study 1.43 (1.29,1.58) 1.30 (1.25,1,35) 1.35 (1.17,1.55) 1.27 (1.17,1.39) 1.31(1.17,1.53) 1.33 (1.22,1.47) 0.79 (0.51,1.17) 0.40 (0.30,0.53) 1.18 (0.74,1.88) 2.77 (2.08,2.67) 1.19 (1.04,1.37) (1.49,2.70) 1.39 (1.30,1.59) Ischemic Stroke CMCS 1.51 (1.40,1.65) 1.72 (1.58,1.88) 1.12 (1.00,1.26) 0.68 (0.49,0.96) 1.64 (1.20,2.24) 1.16 (1.04,1.30) 1.33 (1.06,1.67) Hemorrhagic Stroke CMCS 1.38 (1.20,1.58) 2.01 (1.77,2.32) (0.81,0.75) Total Stroke Framingham Heart Study 1.33 (1.23,1.45) 1.32 (1.21,1.44) (1.09,1.80) (1.51,1.85) Females CHD CMCS Framingham Heart Study 1.44 (1.22,1.69) 1.30 (1.25,1,35) 1.25 (1.04,1.52) 1.27 (1.17,1.39) 1.17(1.04,1.31) 1.39(1.22,1.58 ) 0.31 (0.15,0.62) 0.40 (0.30,0.53) 1.70 (0.88,3.28) 2.77 (2.08,2.67) 1.07 (0.88,1.31) (1.30,1.59) Ischemic Stroke CMCS 1.40 (1.26,1.57) 1.49 (1.35,1.67) 1.12 (1.00,1.26) 0.79 (0.53,1.21) 1.37 (0.86,2.17) 1.17 (1.03,1.33) 1.33 (1.05,1.59) Hemorrhagic Stroke

34 CMCS 1.16 (0.99,1.36) 2.05 (1.77,2.37) (0.90,4.88) Total Stroke Framingham Heart Study 1.33 (1.23,1.45) 1.48 (1.36,1.62) (1.35,2.21) (1.54,1.88) The CMCS is described in detail on pages A3 and A4 of this appendix. Note that if age*risk factor interactions were found, these are modeled in age specific risk coefficients (more detail on Methods page 4 and elaborate on Appendix page A2 and A4. *Diabetes defined as having a diagnosis of diabetes, or taking medications for diabetes, or having a fasting glucose 7.0 mmol.l ( 126 mg/dl) For the Framingham Heart Study, diabetes was defined similarly except that a cutoff of a fasting glucose 140mg/dl was used. Total stroke in the Framingham Heart Study included transient ischemic attack (TIA, ICD-9 435) in addition to the stroke outcomes defined for the CMCS, which did not include TIA. There were too few stroke events in Framingham to allow for estimation of separate ischemic and hemorrhagic stroke multivariate risk functions. 95% confidence intervals were not reported for the published estimates, 11 so standard errors estimated from original Framingham data were used to generate 95% confidence intervals. The beta coefficient for current smoking for CHD in Chinese women was considered unreliable; 5 the CHD Policy Model-China uses the coefficient from the PRC-USA Study, but adjusts the coefficient to account for total CHD 8 (myocardial infarction, cardiac arrest, and angina) rather than the hard CHD (myocardial infarction and cardiac arrest only) outcome used by the PRC-USA Study. 15 The 95% confidence interval for this adjusted estimate is unknown.

35

36 Appendix Methods Table 3: MEDLINE search terms Broad search terms (i.e. MeSH terms) were combined with terms specific to adult Chinese populations over the years 1980 to Elimination of duplicate titles resulted in the list of 1,699 titles produced from the PubMed search. Appendix A. PubMed Search Terms Comprehensive CVD risk factor search: 368 Titles ("Cardiovascular Diseases"[Mesh] AND ("Cardiovascular Diseases/epidemiology"[Mesh] AND "Cardiovascular Diseases/etiology"[Mesh])) AND ("china"[mesh Terms] OR "china"[all Fields]) AND ("data collection"[mesh Terms] OR ("data"[all Fields] AND "collection"[all Fields]) OR "data collection"[all Fields] OR "survey"[all Fields]) AND ("1980"[PDAT] : "2008"[PDAT]) AND "humans"[mesh Terms] AND ("adult"[mesh Terms:noexp] OR ("middle aged"[mesh Terms] OR "aged"[mesh Terms])) Diabetes Mellitus search: 343 Titles "Diabetes Mellitus/epidemiology"[Mesh] AND ("1980"[PDAT] : "2008"[PDAT]) AND "humans"[mesh Terms] AND ("adult"[mesh Terms:noexp] AND ("1980"[PDAT] : "2008"[PDAT])) AND ("China"[Mesh] OR "Chinese"[All Fields]) Blood pressure and hypertension search: 343 Titles "Hypertension/epidemiology"[Mesh] AND ("1980"[PDAT] : "2008"[PDAT]) AND "humans"[mesh Terms] AND ("adult"[mesh Terms:noexp] AND ("1980"[PDAT] : "2008"[PDAT])) AND ("China"[Mesh] OR "Chinese"[All Fields]) Cholesterol search: 534 Titles ("Hyperlipidemias/epidemiology"[Mesh] OR "cholesterol"[mesh Terms]) AND ("1980"[PDAT] : "2008"[PDAT]) AND "humans"[mesh Terms] AND ("adult"[mesh Terms:noexp] AND ("1980"[PDAT] : "2008"[PDAT])) AND ("China"[Mesh] OR "Chinese"[All Fields]) BMI, obesity and overweight search: 178 Titles ("Obesity/epidemiology"[Mesh]) AND ("1980"[PDAT] : "2008"[PDAT]) AND "humans"[mesh Terms] AND ("adult"[mesh Terms:noexp] AND ("1980"[PDAT] : "2008"[PDAT])) AND ("China"[Mesh] OR "Chinese"[All Fields]) Cigarette smoking search: 256 Titles "Smoking/epidemiology"[Mesh] AND ("1980"[PDAT] : "2008"[PDAT]) AND "humans"[mesh Terms] AND ("adult"[mesh Terms:noexp] AND ("1980"[PDAT] : "2008"[PDAT])) AND ("China"[Mesh] OR "Chinese"[All Fields])

37

38 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Table 4. Past population-based Chinese surveys, reviewed for risk factor data. Study Comment Risk factors Used for Time Interval measured risk factor trend China Study I & Total X X II53, 54 cholesterol Independent random sample of >8,000 men and women aged years from 130 villages in 69 counties throughout mainland China, 1983 & 1989 BMI, SBP, smoking, cholesterol, diabetes 2000 PRC-USA Study Sino-MONICA, Beijing36, 58 INCLEN 21 Tianjin Urban 59 China Health & Nutrition Surveys (CHNS) 60, 61 Chen et al. 62 Cohort of 5,084 men and women workers and farmers from urban and rural areas of Bejing and Guangzhou aged years at baseline with CVD risk factors measured in , , and Repeated independent random samples in the Beijing area, Multi-national survey including 1,188 men age from Chengdu and Shanghai in 1987 Two independent sample surveys of men and women age years in Tianjin city; 14,046 in 1989 and 4,000 in 1996 Household survey with repeated samples (see appendix text for detailed information) Baseline measurements from a prospective study of >222,000 men aged years from 45 BMI, SBP, smoking, cholesterol BMI, SBP, smoking, cholesterol SBP, cholesterol, BMI, smoking No X X X Total cholesterol No X X X X X BMI, SBP, smoking No X X BMI, SBP, smoking BMI, BP, smoking BMI, SBP, smoking No X XX X XX 27

39 Risk factor trends and future CVD in China Andrew Moran, MD, MPH China Hypertension Epidemiology Follow Up Study CHEFS4, 63 InterASIA45, Nationally Nutrition and Health Status of the Chinese People, randomly sites (selected from 145 total Chinese Disease Surveillance Points system sites) Nationally representative multistage, random clustering sample of 83,533 men and 86,338 women age 40 years in 1991 representative multistage cluster sample of 15,838 Chinese adults, aged years that occurred over National survey of 243,479 men & women >18 years old from 31 provinces, autonomous regions, and municipalities, urban and rural, multistep cluster sample BMI, SBP, smoking No X X BMI, SBP, smoking, cholesterol, diabetes using fasting glucose (ADA/WHO criteria) Obesity & overweight, smoking, diabetes, for cholesterol see below Intercept X for assumed 2000 risk factor levels except diabetes worst cases No X Nutrition and Health Status of the Chinese People, 2002 (cholesterol subsample) 41 Sub-sample of the above 2002 survey: 49,252 Chinese adults age 18 years, urban and rural Anhui Rural 68 18,630 participants ages years recruited from two rural areas of Anhui province, Cholesterol (though study measured multiple risk factors, see above) BMI, SBP, smoking, cholesterol, diabetes Cholesterol X No X 28

40 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Shanghai workers 18 Baseline Values of 2,923 participants ages years, part of a larger prospective study of male and female Shanghai factory workers, Cholesterol, SBP, smoking No X Wang WJ et al 69 Survey of 42,751 men & women age years from 11 Chinese provinces, urban and rural Obesity & overweight No X Beijing professionals 39 Two independent samples of >13,000 male and female professionals ages in Beijing, and 2003 Cholesterol Cholesterol X X China National Diabetes Survey (19 Province Survey) 47 Survey of >224,000 men and women >25 years old in 19 provinces of China, Diabetes using OGTT and 1985 WHO criteria No X 11 Province Survey of Survey of diabetes and IGT in 11 provinces of China, sampled 42,751 men and women in 11 provinces of China Diabetes (using OGT, WHO 1985 criteria) No X Wong CW and Wang Z 51 International Diabetes Federation Systematic review and metaanalysis of 21 studies of diabetes prevalence in Mainland China, Projections based on past prevalence and urbanization assumptions Diabetes Diabetes X X X X X Diabetes Diabetes 29

41 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Atlas 25 Wild et al. 52 GBD review of diabetes trend in China among other nations and regions Diabetes Diabetes Pan et al. 49 Da Qing IGT and diabetes study. Sampled 110,660 residents of Da Qing aged in 1986 (Da Qing city also part 1994 national survey) Diabetes using OGTT and 1985 WHO criteria X DECODA 22 INTERSALT 20 Wu X, Wu Y, Zhou B 70 Zhao XL et al. 71 Yu JJ et al. 31 Multinational adult diabetes survey including three Chinese cohorts, Multinational blood pressure study including 600 men and women age from Beijing, Tianjin, and Nanning Two surveys of 17,097 men and women aged years, urban and rural, and Survey of 29,076 men and women aged from 12 provinces, 1 autonomous region, 1 municipality across China, 2002 National survey of 29 provinces of China, 1984 Diabetes using OGTT and fasting glucose (1999 WHO) No X X SBP No X X Hypertension No X X Hypertension No X Smoking No X China National Smoking Survey Nationally representative Smoking No X National Health Services Nationally representative questionnaire survey conducted in Smoking No X X 30

42 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Surveys 23 Psychoactive substance use in three sites in China survey 23 Yang et al. 72 (DSP survey) China Seven Cities Study 34 WHO World Health Survey , We were not able to access smoking data for the 2003 survey. Survey of tobacco use in Helongjian, Hunan, and Jiangsu provinces, included urban and rural populations Survey of 16,056 men & women years, from 145 disease surveillance points (DSP) sites nationally, 2002 Survey of tobacco smoking in seven Chinese cities, 2002 China sample from international survey, urban and rural, 3,993 male and females age 18 years, 2002 Smoking No X Smoking No X Smoking No X Smoking No X 31

43 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Table day case fatality proportion assumed for China (main assumptions) and the United States (used in a sensitivity analysis with lower case fatality). All China case-fatality data were estimated from Beijing Sino-MONICA data (age-trended and smoothed from estimates for year olds, personal communication, Dr. Dong Zhao, Beijing Institute for Heart, Lung, and Blood Vessel Diseases). U.S. estimates for CHD are based on original analyses of the National Hospital Discharge Survey (NHDS) 73 for 2005 and the National Registry of Myocardial Infarction (NRMI) 74. U.S. stroke estimates are based on NHDS analysis and the Atheroscerosis in Communities Study (ARIC). 75 Assumption China (main assumptions) years CHD (including MI & arrest only, exclude angina) Ischemic stroke Hemorrhagic stroke Men Women Men Women Men Women United States (used for sensitivity analysis) years

44 Risk factor trends and future CVD in China Andrew Moran, MD, MPH 33

45 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Table 6a. Incremental changes in projected CHD and stroke events and deaths, and non-cardiovascular deaths attributable to projected trends in cardiovascular disease risk factors, Chinese men age years, Simulation CHD events CHD deaths Stroke events Stroke deaths Non-CVD deaths Men Base case Total 24,255,000 10,066,000 75,947,000 27,659,000 99,390,000 Trend cases: incremental events (% change from baseline) SBP Main High Low 1,006,000 (4) 1,332,000 (6) -845,000 (-4) 415,000 (4) 552,000 (6) -387,000 (-4) 6,571,000 (9) 8,720,000 (12) -6,300,000 (-8) 2,278,000 (9) 3,059,000 (12) -2,387,000 (-9) 311,000 (<1) 315,000 (<1) -233,000 (-1) TC Main High Low 3,800,000 (16) 7,554,000 (31) 1,334,000 (6) 1,667,000 (16) 3,346,000 (33) 581,000 (6) 2,482,000 (3) 4,658,000 (6) 907,000 (1) 533,000 (2) 994,000 (4) 196,000 (<1) -1,249,000 (<-1) -2,485,000 (<-1) -419,000 (<-1) Active Smoking Main Worst Best -1,940,000 (-8) -1,139,000 (-5) -2,691,000(-11) -937,000 (-9) -551,000 (-6) -1,298,000 (-13) -586,000 (<-1) -869,000 (<-1) -379,000 (-1) -106,000 (<-1) -59,000 (<-1) -71,000 (<-1) -1,424,000 (-1) -811,000 (<-1) -2,118,000 (-2) Diabetes Main High Low 464,000 (2) 740,000 (3) 108,000 (<1) 209,000 (2) 334,000 (3) 48,000 (<1) 1,608,000 (2) 2,585,000 (3) 366,000 (<1) 339,000 (1) 544,000 (2) 78,000 (<1) 980,000 (1) 1,514,000 (2) 241,000 (<1) BMI Mediated by SBP, TC, HDL, diabetes Main High Low 1,533,000 (6) 2,545,000 (11) 960,000 (4) 631,000 (6) 1,046,000 (10) 399,000 (4) 3,665,000 (5) 6,124,000 (8) 2,278,000 (3) 650,000 (2) 1,082,000 (4) 402,000 (2) 1,026,000 (1) 1,705,000 (2) 584,000 (1) Simultaneous main trends SBP, TC, diabetes Smoking decline No smoking decline 3,129,000 (13) 5,543,000 (23) 1,245,000 (12) 2,412,000 (24) 10,153,000 (13) 10,806,000 (14) 2,947,000 (11) 2,053,000 (11) -1,240,000 (-1) 146,000 (<1) 34

46 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Table 6b. Incremental changes in projected coronary heart disease and stroke events and deaths, and noncardiovascular deaths attributable to projected trends in CHD risk factors, Chinese women age years, Simulation CHD events CHD deaths Stroke events Stroke deaths Non-CVD deaths Women Base case Total 14,339,000 6,592,000 53,882,000 18,420,000 76,150,000 Trend cases: incremental events (% change from baseline) SBP Main High Low 435,000 (3) 590,000 (4) -317,000 (-2) 202,000 (3) 273,000 (5) -151,000 (-3) 4,769,000 (8) 6,554,000 (12) -3,668,000 (-7) 1,882,000 (10) 2,592,000 (14) -1,476,000 (-8) 1,227,000 (2) 1,624,000 (2) -1,263,000 (-2) TC Main High Low 1,169,000 (8) 2,122,000 (15) 336,000 (2) 554,000 (8) 1,013,000 (15) 159,000 (2) 1,738,000 (3) 3,084,000 (6) 509,000 (1) 349,000 (2) 617,000 (3) 102,000 (<1) -549,000 (<-1) -1,018,000 (-1) -155,000 (<-1) Active Smoking Main Worst Best -105,000 (<-1) 270,000 (2) -152,000 (-1) -59,000(<-1) 151,000 (2) -85,000 (-1) -37,000 (<-1) 86,000 (<1) -135,000 (<-1) -19,000 (<-1) 43,000 (<-1) -90,000 (<1) -128,000 (<-1) 300,000 (<1) -125,000 (<-1) Diabetes Main High Low 847,000 (5) 1,411,000 (10) 190,000 (1) 409,000 (6) 683,000 (10) 92,000 (1) 726,000 (1) 1,180,000 (2) 169,000 (<1) 141,000 (<1) 299,000 (1) 33,000 (<1) 700,000 (1) 1,098,000 (1) 174,000 (<1) BMI Mediated by SBP, TC, HDL, diabetes Main High Low 945,000 (7) 1,469,000 (10) 647,000 (5) 424,000 (6) 660,000 (10) 289,000 (4) 1,650,000 (3) 2,476,000 (5) 1,217,000 (2) 281,000 (2) 433,000 (2) 192,000 (1) 1,269,000 (2) 1,902,000 (3) 891,000 (1) Simultaneous main trends SBP, TC, diabetes Smoking decline No smoking decline 2,468,000 (17) 2,594,000 (18) 1,164,000 (18) 1,234,000 (19) 7,269,000 (14) 7,307,000 (14) 2,314,000 (13) 2,333,000 (13) 1,280,000 (3) 1,406,000 (3) 35

47 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Table 7a. Optimistic and pessimistic scenario analyses for total cholesterol, systolic blood pressure, smoking and diabetes assumptions in Chinese men aged years, Simulation CHD events CHD deaths Stroke events Stroke deaths Non-CVD deaths Men Base case (totals) 24,255,000 10,065,000 75,947,000 27,659,000 99,390,000 Trend cases: incremental events [absolute and (%) change from main base case; except low case fatality risk factor trend scenario compared with low case fatality base case] Optimistic Scenarios Base case with lower case-fatality starting 2010 (totals) 25,092,000 7,693,000 77,008,000 20,396, ,068,000 Main trends SBP, TC, diabetes+ lower case-fatality starting 2010* 3,230,000 (13) 941,000 (9) 10,135,000 (13) 1,953,000 (7) -198,000 (<-1) Smoking Aggressive anti-tobacco policy SBP, TC, diabetes main trends + aggressive anti-tobacco -5,456,000(-23) -1,455,000 (-6) -2,624,000 (-26) -953,000 (-10) -1,679,000 (-2) 7,857,000 (10) -253,000 (-1) 2,592,000 (9) -4,467,000 (-5) -4,733,000 (-5) SBP Lower SBP 3.6 mm Hg TC, diabetes, smoking trends, lower SBP 3.6 mm Hg TC, diabetes, lower SBP 3.6 mm Hg, no smoking change -1,164,000 (-5) 755,000 (3) 2,963,000 (12) -502,000 (-5) 244,000 (2) 1,313,000 (13) -7,253,000 (-10) -4,370,000 (-6) -3,824,000 (-5) -2,796,000 (-10) -2,270,000 (-8) -2,178,000 (-8) -283,206 (-1) -1,640,000 (-2) -257,000 (<-1) Pessimistic Scenarios TC Increase ceiling to 6.0 mmol/l 11,649,000 (48) 5,175,000 (51) 6,882,000 (9) 1,459,000 (5) -3,770,000 (-4) Diabetes Diabetes high trend and Framingham Heart Study diabetes β coefficients 3,285,000 (14) 1,394,000 (15) 2,377,000 (3) 812,000 (3) 798,000 (1) BMI Main (mediated by SBP, TC, HDL, diabetes) Additional non-mediated effects 1,533,000 (6) 2,403,000 (10) 631,000 (6) 992,000 (10) 3,665,000 (5) 3,379,000 (5) 650,000 (2) 609,000 (2) 1,026,000 (1) -956,000 (-1) *Incremental change compared with the lower case-fatality base case. 36

48 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Table 7b. Optimistic and pessimistic scenario analyses for total cholesterol, systolic blood pressure, smoking and diabetes assumptions, Chinese women aged years, Simulation CHD events CHD deaths Stroke events Stroke deaths Non-CVD deaths Women Base case (totals) 14,339,000 6,592,000 53,882,000 18,420,000 76,150,000 Trend cases: incremental events [absolute and (%) change from main base case; except low case fatality risk factor trend scenario compared with low case fatality base case] Optimistic Scenarios Base case with lower casefatality starting ,076,000 4,047,000 54,608,000 14,160,000 80,765,000 Main trends SBP, TC, diabetes+ lower case-fatality starting 2010* 2,711,000 (19) 799,000 (12) 7,373,000 (14) 1,725,000 (9) 1,938,000 (3) SBP Lower mean SBP 3.6 mm Hg TC, diabetes, smoking trends, lower SBP 3.6 mm Hg TC, diabetes, lower SBP 3.6 mm Hg, no smoking change -449,000 (-3) 1,455,000 (10) 1,575,000 (11) -208,000 (-3) 693,000 (11) 759,000 (12) -4,658,000 (-9) -2,437,000 (-5) -2,401,000 (-5) -1,812,000(-10) -1,416,000 (-8) -1,399,000 (-8) -1,395,000 (-2) -1,307,000 (-2) -1,181,000 (-2) Pessimistic Scenarios TC Increase ceiling to 6.0 mmol/l 3,417,000 (24) 1,634,000 (25) 4,842,000 (9) 966,000 (5) -1,610,000 (-2) Diabetes Diabetes high trend and Framingham Heart Study diabetes β coefficients 1,903,000 (13) 892,000 (14) 3,259,000 (6) 1,078,000 (6) 192,000 (<1) BMI Main (mediated by SBP, TC, HDL, diabetes) Additional non-mediated BMI effects 945,000 (7) 483,000 (3) 424,000 (6) 218,000 (3) 1,650,000 (3) 2,426,000 (5) 281,000 (2) 422,000 (2) 1,229,000 (2) -490,000 (-1) *Incremental change compared with the lower case-fatality base case. 37

49 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Supplemental Methods Figure and Figure Legends: Appendix Figure 1. Literature review, data abstraction, and analysis flow chart for the analysis of CHD risk factor trends in China, the CHD Policy Model-China project. Initial PubMed Search yielded 1,699 titles of studies of CVD risk factors in China, Screening step Application of exclusion criteria to 1,699 abstracts: , at least age years, male & female Population-based Standard, quality measurement methods Fulfilling risk factor definitions National/Multi-provincial scale OR Repeated samples of the same population(s) 98 full manuscripts reviewed 1,601 studies excluded by criteria Additional data from the WHO InfoBase (6 studies) 72 studies excluded by criteria on closer review Data abstraction and synthesis, selected from 27 studies from electronic search plus World Health Organization Infobase and International Diabetes Federation data (33 studies total): Plotting of age-standardized past survey data for ages over time for studies finally included Main linear trends in SBP, BMI, smoking from China Health & Nutrition Survey data Fitting of main trends for total cholesterol, diabetes from summary estimates of various studies Future Trend Projections: Main, High, and Low trend assumptions for the years entered into the CHD Policy Model-China 38

50 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Figures 2a-2e. Age standardized past cardiovascular disease risk factor trends for adults aged years from Chinese surveys and assumed projected trends. Survey data were plotted from the U.S. and Japan for reference. Age standardization of past trends in China, United States, and Japan was performed using the 2000 Chinese Census. 39

51 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Figure 2a. 40

52 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Figure 2b 41

53 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Figure 2c. 42

54 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Figure 2d. 43

55 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Figure 2e. 44

56 Risk factor trends and future CVD in China Andrew Moran, MD, MPH Appendix Methods Figure 3. Exponential decline in smoking prevalence in Chinese men to 20% by 2020 and 10% by 2030, modeled in a sensitivity analysis. 45