Cohort Profile: The Hong Kong Cardiovascular Risk Factor Prevalence Study (CRISPS) and the follow-up studies

Atherosclerotic cardiovascular diseases, especially coronary heart disease (CHD), have for decades been the major cause of mortality in most developed Western countries.1 In the 1990s, CHD mortality was low in many East Asian regions, including Hong Kong (HK), but in Singapore, a city with a similar level of socio-economic development to HK, the CHD mortality rate was already increasing.2 A rise in CHD mortality has conventionally been thought to be an inevitable consequence of economic development. Hence, it was postulated that HK could be experiencing a similar rise, with its ageing population, rapid socio-economic development, and marked changes in diet and lifestyle. Besides, the difference in the dietary habits between Chinese and Caucasians might explain the difference in CHD mortality.3 In the 1990s, heart disease was the second commonest cause of mortality after cancer in HK. As a single disease entity, CHD emerged as the biggest killer in HK (3181 deaths in 1992).4 Therefore, the Hong Kong Cardiovascular Risk Factor Prevalence Study (CRISPS) was initiated to generate a comprehensive database regarding the prevalence of cardiovascular risk factors in HK Chinese, followed by cohort studies with detailed dietary and biochemical data, to facilitate the development of appropriate population-wide CHD prevention strategies and to monitor their long-term impacts.5,6 To establish a database of cardiovascular risk factors based on a representative sample of the HK Chinese population, 10 age–sex specific groups (25–34, 35–44, 45–54, 55–64, 65–74 years old) were recruited for blood sampling and anthropometric measurements. The World Health Organization (WHO) Monitoring Trends and Determinants in Cardiovascular Disease (MONICA) recommended 200 participants in each of the 10 age–sex specific groups to facilitate long-term monitoring of trends of cardiovascular risk factors.7 Previous surveys carried out in HK using randomly generated telephone numbers for similar telephone interviews as CRISPS had shown that this method provided an excellent random sample of the population.8 In a pilot study using this method, 80% of those who answered the telephone completed the telephone interview and 40% of the interviewees agreed to attend for physical examination and blood tests. Therefore, it was estimated that ∼9000 completed telephone interviews would be necessary to obtain 3000 individuals who would attend a comprehensive physical examination and clinical and biochemical assessments. Between 1 December 1994 and 31 October 1996, 21 135 random telephone numbers were dialled, where 11 970 (57%) were non-residential or otherwise ineligible. 7174 (78%) of the 9165 eligible individuals completed the telephone interview. A total of 2664 (37%) of them agreed to attend physical examination and blood taking. In view of fewer older individuals in the general population and greater difficulties in recruiting older individuals to attend the session for physical examination and blood taking, a supplementary sample of individuals aged 55–74 years was specifically recruited, where 556 completed telephone interviews and 236 (42.4%) agreed to attend physical examination and blood taking.9 In total, 7730 Chinese individuals (telephone survey cohort) completed structured telephone interviews for demographic and lifestyle factors. They formed the first HK database for cardiovascular risk factors including details on smoking, drinking, exercise, and socio-economic information such as marital status, education levels, occupation status and type of housing. A total of 2900 individuals attended for comprehensive physical examination, clinical and biochemical assessments, forming CRISPS-1, the baseline cohort.9 Pregnant women and individuals with serious diseases such as cancer or who were hospitalized were excluded. The age and sex distributions of the attendees and non-attendees were shown to match those of the general population (Table 1). Smoking status, alcohol intake, exercise participation, marital status, housing type, birthplace, highest educational level attained, occupational status, job nature, self-reported prevalence of hypercholesterolaemia and CHD were comparable between the attendees and non-attendees, except for slightly higher rates of self-reported diabetes (5.1 vs 3.4%, P < 0.001) and hypertension (12.5 vs 10.5%, P < 0.01) among non-attendees. Nonetheless, these self-reported diagnoses were not necessarily confirmed. Hence, the non-attendance bias should be small and the baseline cohort CRISPS-1 obtained was a representative population sample.5 Age and sex distribution of the Hong Kong population (in the 1996 By-census) in comparison with the phone interviewees and the Hong Kong CRISPS cohorta Adapted from The Hong Kong Cardiovascular Risk Factor Prevalence Study, 1995–1996 (Janus ED, ed.), 1997, Queen Mary Hospital, Hong Kong (ISBN 962-8310-03-8) with written permission from the copyright owner. Age and sex distribution of the Hong Kong population (in the 1996 By-census) in comparison with the phone interviewees and the Hong Kong CRISPS cohorta Adapted from The Hong Kong Cardiovascular Risk Factor Prevalence Study, 1995–1996 (Janus ED, ed.), 1997, Queen Mary Hospital, Hong Kong (ISBN 962-8310-03-8) with written permission from the copyright owner. All those who participated in CRISPS-1 agreed to be contacted again for follow-up visits. They also gave consent for data linkage to the Clinical Management System (CMS) of the Hong Kong Hospital Authority. All participants from CRISPS-1 were invited to attend the subsequent visits (CRISPS-2 to -5). In 2000–04, 1944 individuals returned for CRISPS-2 (46.3% men and 53.7% women; mean age: 52 years, range 30–82). In 2005–08, 1803 individuals returned for CRISPS-3 (46.5% men and 53.5% women; mean age: 56 years, range 35–85). In 2010–12, 1618 individuals returned for CRISPS-4 (46.2% men and 53.8% women; mean age 59 years, range 39–90). In the latest CRISPS-5, 1428 individuals returned in 2016–18 (44.3% men and 55.7% women; mean age 62 years, range 43–91), with a median follow-up of 20 years. For those who did not attend the follow-up visit in CRISPS-5, clinical outcomes of interest including cardiovascular events were traced from the CMS using the International Classification of Diseases codes, while death events were retrieved from the Hong Kong Death Registry. Hence, clinical outcomes of an additional 1376 individuals, including death data on 439, were available as of CRISPS-5, totalling 2804 individuals (96.7% of the baseline cohort) with clinical outcomes at 20 years valid for analysis. Another wave of CRISPS has been launched. The New Hong Kong Cardiovascular Risk Factor Prevalence Study (N-CRISPS) has commenced in late 2019 with a protocol similar to CRISPS-1, but with the use of a household-based recruitment strategy, and enhanced technologies for assessing life-course exposures and physical activities (further information about N-CRISPS can be found at http://ncrisps.hku.hk/wp/). N-CRISPS will be funded by the Health Care and Promotion Fund, Health and Medical Research Fund (HMRF). As there have been major changes in HK in diet, lifestyle, level of health literacy and the overall standard of medical care over the past 20 years, and with cardiovascular diseases remaining a leading cause of mortality, the N-CRISPS is initiated as a new community-based study to determine the updated sex- and age-stratified prevalence of cardiovascular risk factors. The profile of the N-CRISPS will be compared with that of CRISPS-1 and the overall 7730 individuals who completed the telephone interview, to obtain valuable insights into the major changes in cardiovascular risk factors, to facilitate the development of appropriate population-wide cardiovascular disease prevention strategies. A total of 7730 Chinese were randomly selected for telephone interviews, conducted with the questionnaire used in the 1992 Singapore National Health Survey and a validated Chinese version of the WHO Rose-Angina Questionnaire.10 This method of telephone interview was validated in a morbidity survey in HK.8 Information collected included demographic data, perceived general health, lifestyle factors, medical history and family history of CHD, diabetes, hypertension and stroke.10 In total, 2900 individuals agreed to attend physical examination and blood tests. Anthropometric parameters, including body height and weight, waist and hip circumferences, and blood pressure (BP), were measured using standardized techniques.9 Waist circumference (WC) was measured half-way between the xiphisternum and the umbilicus,11 which highly correlated with the WHO recommendation of WC measurement half-way between the lower rib margin and iliac crest.12 Also, 1010 of the 2900 participants in CRISPS-1 completed the dietary survey with a validated Food Frequency Questionnaire.6 All participants had blood taken after an overnight fast for 12 h before their visit. Biochemical parameters including fasting glucose (FG), insulin, lipid profile and fibrinogen were measured, and a 75 g oral glucose tolerance test (OGTT) was performed in all except those taking anti-diabetic medications. The laboratory used standard methods which met the requirements of international quality control programmes.5 Plasma and serum samples were stored in aliquots at –70°C for biomarker measurements. A sub-cohort of 1010 participants involved in the dietary survey in CRISPS-1 had urinary sodium, potassium and creatinine levels measured.6 DNA samples of 1938 individuals in CRISPS-2 were available for genotyping and were analysed in relation to the development of cardiovascular risk factors. Totals of 296 and 502 individuals attending CRISPS-2 and CRISPS-3, respectively, had intima-media thickness (IMT) measurements at the common carotid arteries using high-resolution B-mode ultrasound (ATL HDI 3000 and 5000 ultrasound system; Advanced Technology Laboratories) to assess carotid atherosclerosis.13 In the follow-up visits, participants were comprehensively reassessed, as in CRISPS-1. Body fat was analysed with a TBF-410 Tanita Body Composition Analyzer (Tanita Corp., Tokyo, Japan). In 1997, a sub-cohort of 322 individuals with impaired glucose tolerance (IGT, defined by the WHO 1998 criteria) and 322 sex- and age-matched individuals with normal glucose tolerance from CRISPS-1 participated in a reassessment to examine the natural history of IGT in Chinese, and hence had additional data at year 2, including 75 g OGTT results, anthropometric and BP measurements.14 Table 2 summarizes the variables captured in the baseline and follow-up studies. Summary of parameters measured in the Hong Kong CRISPS cohort Questionnaire: demographics, perceived general health, smoking, drinking, physical activities, personal and family history of cardiovascular diseases, dietary survey Anthropometric data: body height, body weight, waist circumference, hip circumference Blood pressure Biochemical parameters: fasting glucose, insulin, lipid profile, fibrinogen, 75 g oral glucose tolerance test Same measurements as in baseline (CRISPS-1) Bioelectrical impedance Electrocardiogram Carotid intima-media thickness Plasma and serum samples (for biomarkers and genetics) Questionnaire: demographics, perceived general health, smoking, drinking, physical activities, personal and family history of cardiovascular diseases, dietary survey Anthropometric data: body height, body weight, waist circumference, hip circumference Blood pressure Biochemical parameters: fasting glucose, insulin, lipid profile, fibrinogen, 75 g oral glucose tolerance test Same measurements as in baseline (CRISPS-1) Bioelectrical impedance Electrocardiogram Carotid intima-media thickness Plasma and serum samples (for biomarkers and genetics) Summary of parameters measured in the Hong Kong CRISPS cohort Questionnaire: demographics, perceived general health, smoking, drinking, physical activities, personal and family history of cardiovascular diseases, dietary survey Anthropometric data: body height, body weight, waist circumference, hip circumference Blood pressure Biochemical parameters: fasting glucose, insulin, lipid profile, fibrinogen, 75 g oral glucose tolerance test Same measurements as in baseline (CRISPS-1) Bioelectrical impedance Electrocardiogram Carotid intima-media thickness Plasma and serum samples (for biomarkers and genetics) Questionnaire: demographics, perceived general health, smoking, drinking, physical activities, personal and family history of cardiovascular diseases, dietary survey Anthropometric data: body height, body weight, waist circumference, hip circumference Blood pressure Biochemical parameters: fasting glucose, insulin, lipid profile, fibrinogen, 75 g oral glucose tolerance test Same measurements as in baseline (CRISPS-1) Bioelectrical impedance Electrocardiogram Carotid intima-media thickness Plasma and serum samples (for biomarkers and genetics) Over the past 20 years, CRISPS has contributed to >100 international peer-reviewed publications. These include cross-sectional and prospective studies on the epidemiology of cardiovascular risk factors and diseases in HK, their prediction and potential preventive strategies, and genetic and mechanistic studies on cardiovascular risk factors and diseases. We have identified the key risk factors of diabetes, hypertension, dyslipidaemia, CHD and stroke in HK Chinese and demonstrated how central obesity predisposes to cancer development. These findings have guided public health policies on the prevention and control of these major health problems in our community, especially in the context of population ageing. The CRISPS database also allows investigations of novel genetic and protein biomarkers to enhance the identification of individuals at increased cardiovascular risk. The list of publications from CRISPS can be found at http://ncrisps.hku.hk/wp/index.php/crisps-publications/. The baseline characteristics of the participants, aged 25–74 years, in CRISPS are summarized in Table 3. Baseline characteristics of the participants (aged 25–74 years) in telephone interview and the Hong Kong CRISPSa Adapted from The Hong Kong Cardiovascular Risk Factor Prevalence Study, 1995–1996 (Janus ED, ed.), 1997, Queen Mary Hospital, Hong Kong (ISBN 962–8310-03–8) with written permission from the copyright owner. SEM, standard error of the mean. HDL, high density lipoprotein; LDL, low density lipoprotein. Baseline characteristics of the participants (aged 25–74 years) in telephone interview and the Hong Kong CRISPSa Adapted from The Hong Kong Cardiovascular Risk Factor Prevalence Study, 1995–1996 (Janus ED, ed.), 1997, Queen Mary Hospital, Hong Kong (ISBN 962–8310-03–8) with written permission from the copyright owner. SEM, standard error of the mean. HDL, high density lipoprotein; LDL, low density lipoprotein. In CRISPS-1, 9.8% of individuals had diabetes, an unexpectedly high prevalence for a relatively non-obese population. Another 14% had IGT and 10% had hypertension (defined by systolic BP ≥160 mmHg, diastolic BP ≥95 mmHg, or on anti-hypertensive). A total of 46% of men and 41% of women had total cholesterol of ≥5.2 mmol/L. Overweight and obesity were also prevalent, with 58% of men and 49% of women having body mass index (BMI) > 23.5 kg/m2.5 The clustering of these risk factors was demonstrated by a high prevalence of the metabolic syndrome, present in 14–21% of the individuals depending on the criteria adopted.15 We have demonstrated that anthropometric parameters reflecting central obesity appear to contribute to a greater extent, compared with overall adiposity, to the development of cardiovascular risks.12 Regarding biochemical parameters, OGTT retained its value in improving the detection rate of individuals at risk of diabetes, compared with FG alone.16 In fact, isolated IGT correlated with cardiovascular risk factors,17 and among normoglycaemic individuals, 2-hour post-load glucose levels were independently associated with cardiovascular risk factors.18 The metabolic syndrome was associated with incident hypertension, incident diabetes,19 increased carotid IMT20 and increased all-cause and vascular disease mortality.21 Using the CRISPS cohort we demonstrated, among Chinese, the application of the Framingham cardiovascular risk equation in cardiovascular risk calculation,22 the usefulness of the diabetes risk test recommended by the American Diabetes Association in identifying individuals at risk of diabetes (AUROC of 0.725 at a cut-off of 5), and more recently, the utility of a validated ‘Non-invasive Diabetes Score’, comprising age, BMI and hypertension, to triage individuals to proceed with diabetes testing if deemed high risk.23 We also demonstrated the links between dietary habits and cardiovascular risk development. For instance, dietary antioxidant intake (vitamins A, C and E) was shown to be independently and inversely associated with incident adverse cardiovascular outcomes on 22 years of follow-up.24 CRISPS data have been applied to evaluate the clinical relevance of biomarkers of cardiometabolic disorders including adipokines, mostly identified from our preclinical studies. Circulating levels of adipocyte fatty acid-binding protein (AFABP) correlated with carotid atherosclerosis in women, independently predicted the development of metabolic syndrome and cardiovascular diseases, and were associated with early mortality from ischaemic stroke.25 Hypoadiponectinaemia was reported for the first time to predict incident hypertension and the progression of carotid atherosclerosis.13 Circulating levels of fibroblast growth factor 21 (FGF21) were independently associated with carotid atherosclerosis.26 Circulating AFABP, FGF21 and adiponectin levels all predicted incident diabetes on long-term follow-up, independent of obesity and glycaemic indices. Among them, FGF21 appeared to have a superior predictive value.27 Plasma C-reactive protein (CRP) levels independently predicted the risk of glycaemic progression among individuals with IGT, 14 as well as incident hypertension.28 Measurements of circulating adipokines, such as adiponectin, were shown to have a potential role in improving diabetes risk prediction when incorporated into clinical risk prediction models.29 Furthermore, circulating levels of interleukin-6 and soluble tumour necrosis factor-alpha receptor 2 independently predicted incident cancers. WC, but not BMI, was an independent predictor of incident cancers, supporting an increased risk of cancer mediated by the chronic low-grade inflammation associated with central obesity.30 CRISPS database allows investigations of genetics in relation to cardiovascular risk factors, in part by contributing subjects without CHD as controls. With exome-wide association analysis, variants in three novel loci associated with blood lipid levels in Chinese were identified.31 Furthermore, exome-chip meta-analysis identified novel loci and East Asian-specific coding variants contributing to lipid levels and CHD.32 The finding of an Asian-specific missense variant in PAX4 associated with type 2 diabetes in Chinese from exome-chip association analysis supported the role of PAX4 in the pathogenesis of type 2 diabetes.33 Seven of the 13 obesity-related genetic variants identified from genome-wide association studies in Caucasians were shown to impact on the development of obesity in Chinese. Among them, the strongest associations with obesity were observed in the fat mass and obesity associated (FTO), glucosamine-6-phosphate deaminase 2 (GNPDA2), and melanocortin-4 receptor genes.34 CRISPS data were used to demonstrate the clinical relevance of novel findings from animal studies on diabetes and obesity, such as the role of angiopoietin-like protein 4 (ANGPTL4) and adiponectin. ANGPTL4, a circulating protein predominantly expressed in adipose tissue and the liver, potently decreased blood glucose in mouse models. In human subjects, circulating ANGPTL4 levels inversely correlated with glucose levels and homeostatic model assessment of insulin resistance (HOMA-IR), and ANGPTL4 levels were lower in diabetic patients compared with healthy individuals, suggesting that reduced ANGPTL4 could be a causative factor for type 2 diabetes.35 Adiponectin was found to alleviate fatty diseases in and among from an independent was found between serum adiponectin and levels, suggesting the potential clinical relevance of adiponectin in fatty CRISPS is the representative study performed in HK. this is a Chinese population. with a median follow-up of 20 years it has the of follow-up studies conducted in a valuable cohort to examine the of cardiovascular risk factors and diseases in Chinese, especially for those in The prevalence of diabetes of 9.8% observed in CRISPS-1 in the diabetes in where the prevalence in was but to in and in in with the rapid socio-economic growth and in the which would impact also on cardiovascular risk factors. from the CRISPS follow-up studies could into the cardiovascular risk in a of the participants were to follow-up for including and the of participants who attended of the data of the participants were to the CMS and the Hong Kong Death to the data of those to to ageing of the participants in CRISPS-1, the follow-up not be representative of the general population. The Hong Kong Cardiovascular Risk Factor Prevalence Study (CRISPS) is a community-based cohort randomly recruited from the Hong Kong population to determine the prevalence of cardiovascular risk factors Hong Kong Chinese in followed by over a median follow-up of 20 years to the development of incident cardiovascular risk factors and cardiovascular diseases in this A total of 7730 Chinese individuals had structured telephone A total of 2900 individuals attended the baseline CRISPS-1 and comprehensive clinical and biochemical assessments. were performed years from CRISPS-2 to CRISPS-5, the study data, lifestyle factors, and biochemical and genetic parameters were as well as DNA and blood are has been obtained from the of the of Hong Hong Kong CRISPS-1 was supported by the Hong Kong Health Research Fund the of Hong Kong on Research and the Hong Kong Research and the Hong Kong for the follow-up studies were supported by Health Care and Promotion Fund, Health and Medical Research Fund the and the The data collected in the CRISPS are not Nonetheless, are for of Hong of Hong of Hong of Hong of of Hong of Hong of Hong Promotion Hong Hong of Hong of Hong of Hong Hospital, Hong and of Hong