Current research in the CKB group is focussed on the prospective associations of major non-blood-related factors (e.g. smoking, alcohol drinking, blood pressure, adiposity, diet) with many different conditions (e.g. stroke, ischaemic heart disease, diabetes, chronic obstructive pulmonary disease). Attention is paid to adjustment for potential confounding factors, and the periodic resurvey data can be used to correct for “regression dilution” biases. Genotyping data available in ~100,000 participants will be incorporated to establish (or refute), using Mendelian randomisation, causality of certain environmental factors (e.g. alcohol consumption) for certain diseases; to investigate pleiotropic effect of genes on multiple phenotypes; and to replicate associations of genetic signals previously observed for particular diseases (e.g. IHD) and phenotypes (e.g. blood pressure). Examples of ongoing projects are described below.
- Smoking and mortality
- Alcohol, cardiovascular disease and cancer
- Blood pressure in relation to general and central adiposity
- Blood pressure and cardiovascular disease risk
- Adiposity, cardiometabolic disease and cancer
- Long-term prognosis following stroke
- Diabetes, blood glucose and cardiovascular disease risk
- Lung function, COPD and other related disease risks
- Diet and health
- Physical activity and chronic disease risk
- Female reproductive history and health outcomes
- Depression and anxiety disorders: correlates and consequences
- Population structure and regional variation in common genetic variants
- MTHFR gene variants and cardiovascular disease
- Mendelian randomisation study of blood pressure and risk of stroke
- Mendelian randomisation of adiposity and type 2 diabetes
- Mendelian randomisation of vitamin D, bone density and risk of fracture
Smoking and mortality
Lead investigators: Zhengming Chen, Richard Peto
Men in China account for ~8% of the world’s population, but smoke about one third of the world’s cigarettes. About two thirds of young Chinese men become cigarette smokers in early adult life. By 1990, tobacco caused about 12% of adult male deaths and this is predicted to increase to 30% by 2030 unless there is widespread cessation among smokers. Although the overall hazard is likely to be about the same in China as elsewhere, the chief diseases by which tobacco causes death are very different. Patterns of tobacco death may also change substantially over time as a result of changes in other risk exposures. The CKB will provide direct evidence about the current effects of smoking on cause-specific mortality and major morbidity as well as the health benefits of smoking cessation. The preliminary prospective analyses of CKB mortality data show that the proportion of male deaths attributed to smoking has increased markedly during the past decade, from ~12% during early 1990s to ~18% (24% in urban men) during 2008-12, suggesting there were an estimated ~1 million annual deaths from smoking in China during 2010. The CKB is the only large nationwide prospective study in China that has included a high proportion of men who have started to smoke cigarettes persistently since young adulthood (and so can assess the full effects of smoking). The information generated in the CKB should help to inform the development of tobacco control policy and legislation in China.
Alcohol, cardiovascular disease and cancer
Lead investigators: Iona Millwood, Ling Yang
Prospective studies, mainly in Western populations, have shown that alcohol consumption (especially heavy drinking) can cause certain types of cancer (e.g. mouth, oesophagus, liver), liver cirrhosis, pancreatitis, tuberculosis, injuries and mental illness, and is associated with increased risk of hypertension. Light to moderate drinking, however, has been associated with a reduced risk of cardiovascular diseases in several populations. As well as amount, the patterns of drinking (e.g. heavy drinking episodes, beverage type, consumption with meals) and exposure to other risk factors (e.g. obesity) may modify the health effects of alcohol. The prevalence and patterns of alcohol drinking in CKB participants differ greatly by age, sex and geographic region. Regular drinking is rare in women, but ranges from 7% to 51% among men in the ten study areas. Heavy drinking episodes are widespread particularly in younger men. Spirits are the most commonly consumed beverage, especially in rural areas, although beer drinking is common in younger men. These characteristics give us a unique opportunity to study the age- and sex-specific associations of different drinking patterns with cause-specific mortality and morbidity. Initial prospective analyses will focus on certain types of cancer (e.g. liver and oesophageal) and vascular diseases, with particular interest in whether the associations differ between ischaemic heart disease and stroke and between stroke subtypes. In addition, alcohol metabolising gene variants which result in the flushing response and influence alcohol consumption have been genotyped in ~10,000 CKB participants, and can be used as instrumental variables to assess the causal effect of alcohol on disease risk.
Blood pressure in relation to general and central adiposity
Lead investigators: Zhengming Chen, Sarah Lewington, Robert Clarke
Greater adiposity increases blood pressure and hypertension risk. In Western adult populations, 10 kg/m2 higher body mass index (BMI) appears to be associated, throughout most or all of the BMI spectrum, with about 10 mmHg higher systolic blood pressure (SBP). However, there is little reliable evidence about whether the strength of this association differs by age, sex, or other personal characteristics; nor is there much evidence about the strength of this association in China. This question is of particular importance to China, where high blood pressure is a leading cause of morbidity and premature mortality and where adiposity is rapidly increasing. Furthermore, there is conflicting evidence—both in China and elsewhere—about which simple measures of adiposity most strongly determine blood pressure. Information on which, if any, measures of adiposity most strongly predict blood pressure may be clinically useful, and might suggest whether general or central adiposity is a more fundamental determinant of blood pressure. We are undertaking detailed analysis of the cross-sectional associations of blood pressure, as a continuous variable, with several different measures of adiposity among individuals not taking any anti-hypertensive medications at baseline. Multiple linear regression is used to estimate the effects on blood pressure of measures of general (e.g. body mass index, height-adjusted weight, body fat percentage) and central (e.g. waist circumference, waist-hip ratio) adiposity, with and without adjustment for each other. Preliminary results show that in adult Chinese men and women, general adiposity is more strongly associated with blood pressure than central adiposity, and the association appears much stronger than that observed in many Western populations.
Blood pressure and cardiovascular disease risk
Lead investigators: Sarah Lewington, Yiping Chen, Ben Lacey, Zhengming Chen
About half of the world’s vascular burden is due to occur in the Asia Pacific region and blood pressure is an important determinant of vascular mortality and morbidity. The most reliable evidence to date on the age- and sex-specific associations of blood pressure with vascular diseases come from two meta-analyses: CTSU’s Prospective Studies Collaboration (PSC) involving 1 million participants in mainly Western populations with 56,000 vascular deaths, and the Asia-Pacific Cohort Studies Collaboration (APCSC) involving 400,000 participants with 15,000 vascular events. Both studies observed an approximate doubling in the risk of IHD and stroke mortality for every 20 mmHg higher usual SBP at ages 50-79 years. One of the major limitations of these meta-analyses, however, was the large proportion of strokes of unknown type (about one-half in PSC and one-third in APCSC) and, even among those with a reported pathological type, only a small proportion were known to be confirmed by CT/MRI. Consequently, many important details about the relationship between blood pressure and type-specific stroke remain unclear. Initially, the age- and sex-specific strength and shape of the relationship between usual blood pressure and cause-specific vascular disease will be assessed, with particular focus on IHD and stroke. With extremely high rates of CT/MRI assessment of stroke, we will be able to determine reliably if the age- and sex-specific associations with blood pressure differ between stroke subtypes. Preliminary analyses of the CKB data suggest that the associations are approximately 50% stronger for haemorrhagic than for ischaemic stroke.
Adiposity, cardiometabolic disease and cancer
Lead investigators: Zhengming Chen, Huaidong Du, Kun Tang, Iona Millwood
In many parts of China, average adiposity levels have increased markedly over the past decade, particularly in coastal and urban areas. Meanwhile, there are still many parts of China where undernutrition (e.g. <18.5 kg/m2) is highly prevalent. Reflecting these patterns, the wide distribution in measures of general and central adiposity seen in the CKB should help to assess reliably the associations of both undernutrition and moderate overnutrition with risks of a range of conditions. Initially we will focus on vascular diseases, especially stroke, to investigate whether adiposity has quantitatively, or qualitatively, different associations with different stroke subtypes and which measures of adiposity best predict the risks of stroke subtypes. Similar analyses will also be done for diabetes, IHD and cancer as follow up continues. In particular, we will assess whether, in contrast to previous decades, the increasing adiposity has detectable effects in China on overall cancer incidence, partly because the main cancers caused by obesity (e.g. colorectal cancer, renal cancer, oesophageal adenocarcinoma) are now accounting for a greater proportion of overall cancer burden. To help determine whether China’s current stringent guidelines for overweight and obesity are reasonable, we will examine which measures of adiposity best predict excess overall mortality; and what would be the optimal adiposity levels associated with minimal overall disease burden in the population.
Long-term prognosis following stroke
Lead investigators: Will Whiteley, Sarah Lewington, Yiping Chen
Stroke is one of the commonest causes of adult death and disability in China. Intracranial haemorrhage accounts for a greater proportion of overall stroke in Chinese-origin than in European-origin populations (33% vs. 12%), and lacunar strokes − a manifestation of cerebral small vessel disease − are also likely to be more common. Therefore the longer term prognosis of ischaemic and haemorrhagic stroke may differ between Chinese and European populations. Moreover, with widespread use of CT/MRI, a high proportion of asymptomatic lacunar strokes are detected by imaging, and there is limited evidence about their risk factors and long-term prognosis. The CKB has already recorded a very large number of first and recurrent strokes through the established electronic linkages to mortality and morbidity registries as well as to the nationwide health insurance system. Hence, it will provide a unique opportunity to assess the long-term prognosis following first stroke. We will compare the longer term outcomes (death, ischaemic stroke, and haemorrhagic stroke) after first strokes of different types, and the effect of other modifiers on risk of recurrence (such as age, sex, smoking, blood pressure and adiposity). The study findings may be important for the design of secondary prevention strategies.
Diabetes, blood glucose and cardiovascular disease risk
Lead investigators: Fiona Bragg, Michael Holmes, Zhengming Chen
Diabetes has been shown to be associated with a two- to four-fold increased risk of cardiovascular diseases in Western populations. There is, however, relatively little reliable evidence about whether the association extends to blood glucose levels below the threshold for diabetes, nor is there much evidence about the association in China, either generally or in population subgroups (e.g. age, sex, different levels of adiposity). These questions are of particular importance to China, where the incidence of diabetes is rising rapidly and where cardiovascular disease, in contrast to most Western populations, is characterised by higher rates of stroke, particularly haemorrhagic stroke, and relatively lower rates of ischaemic heart disease (IHD). The CKB, with 30,000 individuals with self-reported doctor-diagnosed or screen-detected diabetes at baseline, as well as measurements of blood glucose on almost all study participants, is well-placed to assess the association of diabetes, and of blood glucose levels amongst individuals without known diabetes, with IHD and stroke subtypes. Initial cross-sectional analyses have demonstrated an approximate doubling of the risk of prevalent IHD and stroke/transient ischaemic attack among individuals with diabetes, and a modest, positive association of prevalent IHD and stroke/transient ischaemic attack with blood glucose levels among individuals without diagnosed or screen-detected diabetes at baseline. Prospective analyses will be conducted to examine these same relationships with well-phenotyped incident cases of IHD and stroke subtypes. Genetic data in ~100,000 participants will also be incorporated for a more in-depth assessment of the relationships.
Lung function, COPD and other related disease risks
Lead investigators: Om Kurmi, Zhengming Chen, Richard Peto
Throughout China, COPD is a major cause of adult death and disability. Cigarette smoking is a well established risk factor for COPD, but it accounts for only about half of male COPD deaths in China and an even smaller proportion of female COPD deaths, indicating that other factors must play important roles in the development of COPD. Detailed information on smoking, prior history of respiratory diseases and a range of other potential risk factors (e.g. passive smoking, household air pollution) for COPD along with spirometry data (FEV1, FVC) are available. This provides an unusually large and comprehensive resource for investigating the main environmental determinants of impaired lung function and respiratory diseases. Given the large number of “never smokers” in the study (30,000 men and 287,000 women), these relationships can be investigated in the absence of potential confounding by smoking. There are already ~8,000 incident cases of COPD recorded in the database. Currently, we are investigating the associations of airflow obstruction at baseline with subsequent cause-specific morbidity and mortality, and in particular subsequent cardiovascular morbidity and mortality. We also plan to investigate associations between COPD hospitalisations in the prospective data and subsequent CVD events, to try to understand further the association between COPD exacerbations and subsequent CVD events.
Diet and health
Lead investigators: Huaidong Du, Tim Keys
Diet plays a critical role in the development of many chronic diseases such as diabetes stroke, ischemic heart disease (IHD), and certain types of cancer, but little is known of the effects of diet on these conditions in China, where the dietary patterns differ greatly from Western populations. In the CKB, information on frequency of usual intake of ~20 main categories of food stuff (e.g. fresh fruit, dairy products, soya bean, meat, fish, and vegetables) was collected at baseline. The analyses of cross-sectional data show that greater intake of fruit is associated with lower blood pressure and blood glucose. Preliminary prospective analyses suggest that there is a strong inverse association of fresh fruit intake with risks of diabetes, IHD, and stroke, especially haemorrhagic stroke, with the association persisting after adjusting for blood pressure. As well as metabolic and cardiovascular diseases, we will extend our investigation to the associations of several other food groups (meat, fish, dairy products, soya bean, and preserved vegetables) with a range of other conditions (e.g. cancer, osteoporosis). We will also examine the effects of different dietary patterns using principal component analysis to enhance the informativeness of the collected dietary data.
Physical activity and chronic disease risk
Lead investigators: Huaidong Du, Derrick Bennett
Low physical activity has been identified as the fourth leading cause of global mortality accounting for an estimated 3.2 million deaths annually. Evidence, mainly from western studies, has shown that lack of physical activity is associated with increased risk of a range of chronic diseases (e.g. diabetes, cardiovascular diseases, fracture). Compared to Western populations the Chinese population has very different patterns of physical activity and chronic diseases. Thus far, little is known of the effects of physical activity on chronic disease risk in Chinese adults. In the CKB, information was collected from all participants about their usual type and duration of activities related to work, commuting, household chores, and leisure-time activities. To quantify the level of physical activity, the number of hours spent per day participating in each activity was multiplied by the MET (Metabolic Equivalent tasks) score for that activity. Recent cross-sectional analyses of the CKB data have found an inverse association between adiposity and total physical activity (overall mean 21.7 MET-hrs/day, mainly from occupational and household activities), with different physical activity domains showing diverse associations. Preliminary prospective analyses also suggest that a lower level of total physical activity is associated with increased risk of ischemic heart disease and stroke in both men and women. We will extend our investigation to the associations of physical activity, both overall and by different domains or intensities, with a range of other conditions (e.g. diabetes, cancer, osteoporosis, and depression).
Female reproductive history and health outcomes
Lead investigators: Ling Yang, Yiping Chen, Sarah Lewington
Several reproductive factors have been associated with the risks of certain cancers in women. It is estimated that the cumulative incidence of breast cancer in developed countries would be more than halved if they had the average number of births (6.5 in developing countries versus 2.5 in developed countries) and lifetime duration of breastfeeding (about 24 months versus 9 months for each child) commonly seen in many developing countries during the 1990s. For each year of delay in age at first birth, the relative risk of breast cancer increases by ~3%, whereas in premenopausal women for each year of delay in age at menarche the risk decreases by ~7%. In China, there have been rapid and dramatic changes in reproductive patterns during recent decades. Compared with older generations, young women tend to have a later age of first birth, fewer children, a shorter duration of breastfeeding and higher prevalence of induced abortion. At the same time, the mean age of menarche has fallen significantly as a result of improvements in nutritional status during early life, with the exception of the Great Famine in 1958-61. Because of the extremely large sample size (>300,000) and detailed reproductive information collected, the CKB will provide the first reliable prospective evidence about the relationship of reproductive factors with risks of breast cancer, stroke and other conditions in China. This will greatly improve our knowledge of the nature of the effects of reproductive history on women’s health in circumstances that differ substantially from those investigated previously in Western studies.
Depression and anxiety disorders: correlates and consequences
Lead investigators: Yiping Chen, Derrick Bennett, Canqing Yu, Ken Kendler
It is estimated that a third of adult disability is attributable to mental illness and three-quarters of the global burden for these conditions is from low and middle income countries like China. Although general patterns of mental disorders are similar across countries the prevalence of these conditions appears to vary by 5-10 fold across different populations, for reasons that are not yet properly understood. In the CKB, detailed information about major depression (MD) and general anxiety disorders (GAD) were collected at baseline and the subsequent periodic resurveys of ~25,000 surviving participants, using a computerised Chinese version of the WHO CIDI-SF instrument administered by trained health workers. These, along with a wide range of other information collected in the study, will enable us to examine a variety of socio-demographic and life-style related correlates of MD and GAD. It will also allow us to examine MD and GAD as the risk factors for other major diseases such as cardiovascular diseases, cancer, and T2DM. Our analyses of cross-sectional data showed that MD was associated with an approximate doubling of the risk of prevalent T2DM. In addition to further cross-sectional analyses of social-demographic and lifestyle-related correlates and determinants of MD, prospective analyses will be conducted to examine MD and GAD with a range of well-phenotyped incident cases of major diseases. Research findings generated will improve our understanding of the causes and long-term consequences of MD and GAD in adult Chinese.
Population structure and regional variation in common genetic variants
Lead investigators: Robin Walters, Iona Millwood, Derrick Bennett
It is well-established in genetic association studies that uncorrected geographic variation in the frequency of genetic variants can give rise to false positive associations, especially if the phenotype or disease being studied also shows geographic variation. There is little understanding of the potential impact of such population stratification on studies into the genetic basis of disease in China; we are therefore investigating genetic diversity in China using genotyping data for 335 common SNPs (MAF≥0.05) in ~100,000 CKB subjects from 10 geographically diverse regions of China. These SNPs were selected on the basis of reported associations with disease risk or disease risk factors (e.g. blood pressure), and initial analyses show that 319 of which exhibit statistically significant variations in allele frequency between the 10 regions, confirming the importance of considering population structure for genetic association studies based on CKB data. Using principal component analysis, other measures of population differentiation, clustering algorithms and other ancestry analysis tools (e.g. STRUCTURE), we are further analysing these SNP data, in combination with existing data from other studies, to better understand the origins of the observed genetic diversity. These investigations will facilitate the identification of the most appropriate genetic models to use in forthcoming genome wide association analyses.
MTHFR gene variants and cardiovascular disease
Lead investigators: Derrick Bennett, Iona Millwood, Robert Clarke
The MTHFR C677T variant is associated with higher homocysteine levels and Mendelian randomisation studies of this variant may be used to predict the effects of folic acid supplementation on risk of cardiovascular disease. Previous studies reported a 30% higher risk of stroke for the TT genotype compared to CC, in Asian but not in Western populations. In Asian regions, where there is low dietary folate consumption, previous research has suggested that randomized controlled trials of folic acid supplementation and further Mendelian randomisation studies are required to investigate the causal relationship between homocysteine and cardiovascular disease. Preliminary results in >82,000 un-related CKB participants show there is substantial geographic variation in MTHFR C677T frequency across 10 study regions. With appropriate correction for the geographic variation, the CKB results do not appear to support the previous associations with stroke reported for Asian populations. The findings of previous published studies may well reflect an artefact due to geographic variation in MTHFR C677T or other methodological problems. Our results suggest that folic acid treatment is unlikely to have any clinically important beneficial effect on risk of stroke overall, stroke sub-types, or IHD.
Mendelian randomisation study of blood pressure and risk of stroke
Lead investigators: Robert Clarke, Sarah Parish, Zhengming Chen
A Mendelian randomisation approach is being used to assess the causal relevance, both qualitatively and quantitatively, of lifelong differences in blood pressure for stroke sub-types. Genotype data for 32 SNPs previously associated with blood pressure are available on 82,000 unrelated CKB participants and were combined to generate a genetic score for systolic blood pressure (GS-SBP). This score will then be used to assess the relative risks associated with lifelong differences in blood pressure for stroke and stroke sub-types (and subsequently for other diseases), overall and in age-specific groups. Overall, there is a 5mm Hg difference in the mean SBP between the top and bottom fifths of the GS-SBP. Among those aged <55, 55-65, and 65+ years at baseline, mean SBP increases with age by 16.5 mmHg, but the mean difference between the top and bottom quintiles of GS-SBP was similar in each age-group. Preliminary prospective analyses suggest that GS-SBP per 5 mm Hg lower SBP is associated with ~20% lower risks of both ischaemic stroke and intra-cerebral haemorrhage, which were comparable in magnitude with those observed for non-genetic studies in Western populations. There are also trends towards weaker RRs per 5 mmHg lower SBP at older ages, similar to trends seen in non-genetic prospective studies conducted in Western populations. Additional genetic data are required on all available strokes in this cohort to confirm the age-specific associations of GS-SBP for stroke and further analyses will also involve other major cardiovascular diseases.
Mendelian randomisation of adiposity and type 2 diabetes
Lead investigators: Iona Millwood, Robin Walters, Robert Clarke
Observational studies have shown that measures of general and central adiposity are associated with type 2 diabetes, and related phenotypes such as fasting blood glucose, blood pressure and blood lipids. However, observational findings could arise from the effects of confounding, or reverse causality. The Mendelian randomisation approach uses genetic variants associated with an exposure or intermediate phenotype, as an instrumental variable (i.e. unconfounded proxy for the exposure) to allow an investigation into the causal relationship between an exposure and outcome. This study uses 26 genetic variants previously associated with body mass index (BMI), waist circumference (WC) and waist:hip ratio (WHR) in large genome-wide meta-analyses, which have been genotyped in ~100,000 CKB participants. Genetic risk scores for BMI, WHR and WC will be used as instrumental variables to investigate the causal effect of general and central adiposity on diabetes and related outcomes to help to assess which measures of adiposity are causally related to diabetes and related outcomes. Results in Chinese will be compared to those seen in Western populations, and gene: environment interactions will also be explored.
Mendelian randomisation of vitamin D, bone density and risk of fracture
Lead investigators: Ling Lu, Iona Millwood, Robert Clarke
Osteoporosis accounts for substantial morbidity and mortality in older people, but whether chronic insufficiency of vitamin D is a reversible determinant of osteoporosis and related risk of fractures is controversial. Observational studies also indicate that low plasma levels of 25-hydroxyvitamin D [25(OH)D] are associated with higher risks of fractures, and vascular and non-vascular mortality, but it is unclear if these associations are causal. Randomized trials assessing the effects on fracture and other health outcomes have generally failed to demonstrate beneficial effects of vitamin D supplementation. Vitamin D deficiency is common in parts of China, where it is believed to occur in about one-third of men and one half of all women. Mendelian randomization studies of single nucleotide polymorphisms (SNPs) encoding plasma levels of 25-hydroxy-vitamin D (25[OH]D) can be used to investigate the effects of lifelong differences in 25(OH)D and risk of disease. Therefore, four SNPs encoding 25(OH)D identified in previous genome-wide meta-analyses were genotyped in ~100,000 CKB participants. A genetic risk score consisting of these SNPs will be used as an instrumental variable to explore the causal relevance of vitamin D deficiency for fracture and a range of other vitamin D-related health outcomes, including bone density measured in a subset of participants during resurvey. The results of this study should guide the design of clinical trials assessing the effects of vitamin D for the prevention of fractures, vascular and non-vascular disease outcomes.