Coronary artery disease is narrowing of the coronary arteries, which supply blood to the heart, leading to damage or malfunction of the heart. The most common heart disorders due to coronary artery disease are angina pectoris (chest pain due to insufficient oxygen reaching the heart) and myocardial infarction (heart attack).
The usual cause is atherosclerosis,in which fatty plaques develop on artery linings. An affected vessel can become totally blocked if a blood clot forms or lodges in the narrowed area. Atherosclerosis has many interrelated causes, including smoking, a high-fat diet, lack of exercise, being overweight, and raised blood cholesterol levels. Other risk factors include a genetic predisposition and diseases such as diabetes mellitus and hypertension.
In its early stages, coronary artery disease often produces no symptoms. The first sign is frequently the chest pain of angina, or an actual heart attack. The disease may also cause cardiac arrhythmias (abnormalities in the heartbeat); in severe cases, arrhythmia can cause cardiac arrest (in which the heart stops beating). In elderly people, it may lead to heart failure, in which the heart gradually becomes less and less efficient.
Drugs are given to help the heart to function more efficiently and to relieve symptoms. These include glyceryl trinitrate (GTN) and other nitrate drugs (e.g. isosorbide mononitrate); calcium channel blockers (e.g. nifedipine); beta-blockers; potassium channel activators; and vasodilator drugs. If drug treatment fails to relieve the symptoms of coronary artery disease, or if there is extensive narrowing of the coronary arteries, blood flow may be improved by balloon angioplasty, stenting or coronary artery bypass surgery.
- Pictures of the coronary arteries
- Anatomy of the coronary arteries in detail
- Coronary heart disease - influences acting in utero and early childhood
Coronary heart disease: epidemiology and prevention
Unlike many medical conditions that are common, disable, and kill, cardiovascular disease (CVD)—already the most common cause of death in the world, and expected to account for a growing proportion of all deaths—is almost entirely preventable.
Socioeconomic factors and habits of society, including (1) a diet high in saturated fat, (2) sedentary living, and (3) smoking, are important underlying determinants of the population rate of coronary disease. Myocardial infarction (MI) mortality rates vary widely between countries (e.g. more than 10-fold higher in Russia than Japan) and change rapidly over time within countries (e.g. 50% decline in 20 years in western Europe; increases in Russia in the 1990s). Coronary heart disease (CHD) is the most common cause of death in women, and while there is a male excess in MI incidence, women are less protected from angina pectoris.
There are strong, unconfounded relationships between several risk factors and CHD mortality and nonfatal myocardial infarction. Those with the strongest effect are (1) age, (2) country, and (3) presence of symptomatic or preclinical disease. Based on recent individual patient data meta-analysis, systolic blood pressure and cholesterol have a log-linear relation with CHD mortality, with no evidence of lower threshold at every age up to the ninth decade of life. A lower blood pressure is associated with a lower risk, whatever the starting level. The implications of this are profound: shifting the distribution of such a risk factor in the whole population by an apparently small amount has a major effect on population rates of disease, e.g. a 5-mmHg reduction in the population mean systolic blood pressure (e.g. achieved through reductions in dietary salt) is predicted to decrease event rates by 20%.
Observations in cohort studies that specific dietary components—including antioxidant vitamins (A, C, and E), B vitamins, folate, and omega 3 fats (from fish)—may reduce the rate of coronary events have not been supported by the available randomized trial evidence. Likewise, hormone replacement therapy appeared protective in observational studies, but the Women’s Health Initiative and other trials showed that it was not.
Nine simple risk factors together may account for 90% of the population attributable risk of myocardial infarction across 52 countries in the Interheart study. Yet, despite this apparent triumph of explanation, meta-analyses of peripheral blood markers of inflammatory, haemostatic, and other processes support medium to strong associations for a number of novel factors, independent of established risk factors. The causal significance, or the contribution to prediction, of these ‘biomarkers’ awaits clarification.
Genetic factors—meta-analyses support a small effect of more than 10 common polymorphisms in the aetiology of CHD, although how they interrelate with other genes, or with the environmental factors is not known. Even an apparently ‘simple’ (one gene, dominant inheritance, and complete penetrance) disorder, familial hypercholesterolaemia, probably has complex interactions with the environment in determining the distal phenotype of premature coronary death.
While simple risk factors combine to predict CHD events in individuals, most events occur among people who are not at high risk. For this reason, treatment of high-risk individuals is a palliative action in public health terms.
A major challenge for CHD epidemiology in the future is to understand the macroeconomic and societal forces which influence population rates in the context of the molecular and genetic mechanisms through which they operate.
Coronary Heart Disease - epidemiology and prevention in great detail
Coronary heart disease (CHD) has, in several senses, grown desperate: from a global perspective, more people are dying from it than ever before, with the largest numbers in India, China, and Russia. For decades we have known the major determinants of the population rate in coronary disease, but for complex reasons this knowledge has not averted exporting the Western epidemic of coronary disease around the world. From the perspective of high-income countries, more people are living with coronary disease than previously, despite the declines in mortality rate observed in western Europe, North America, and Australia since the late 1960s. From a life-course perspective, coronary disease is a paediatric condition, with atherosclerotic changes established in the first and second decades of life. How desperate that primordial prevention has failed to such an extent that, not only in the United States of America and Europe, but also in China and India, there is an epidemic of obesity.
What then is the epidemiological evidence for remedial strategies aimed at preventing CHD? This chapter has three objectives. First, to understand the implications, societally profound, for ‘remedy’ which are provided by recent studies of the large variations in population rate of coronary disease between countries, over time, and in different demographic groups. Second, to consider new large-scale evidence for causes, ranging from distal socioeconomic factors through behavioural and psychosocial factors to more proximal causes ‘under the skin’, including new insights offered by ‘-omic’ disciplines and imaging. Third, to draw implications for national and international public health policy and practice.
Two remedial strategies are required to avert, or mitigate, the onset of coronary disease: population-wide measures which seek to shift the mean distribution of risk factors in the whole population and lower the population rate; and high-risk approaches which seek to identify and modify the behaviour of a minority of individuals (the traditional health care approach).
Many of the causes put forward for angina pectoris in the century after its first description in 1768 were explored, confirmed, or refuted—at least for myocardial infarction (MI) and coronary death—during the second half of the 20th century, largely through the complementary approaches of observational epidemiological studies and randomized trials. See table below:
|Table: Field synopsis of CHD epidemiology: factors associated with CHD aetiology according to their frequency, size of summarized evidence, and relative risk|
|Size of effecta||Risk factorb||Definition of high-risk vs low-risk groupsc||Prevalence of risk group (%)d||No. of CHD cases||Relative risk (age adjusted)e||Meta-analysis (or large dataset) reference|
|Very large (RR ≥3.0)||Age||Top vs bottom thirds||33||5 070||23||Illustration from male population of Finland aged 40–84 years; approximately one-third of the population is aged 49 years and under, and one-third 59 years and older|
|Country||Russia vs Japan||–||105||14||Illustration from WHO Mortality database 2002, comparing male rates aged 65–74 years|
|Symptomatic disease||Second-year MI survivor vs no MI||5||1 237||5||Illustration of 60-year-old British man (Law et al. 2002)|
|Exercise ECG||ST depression vs none||3–16||102||2–21||US preventive task force(Fowler-Brown et al. 2004)|
|Coronary artery calcium||Agatston score >400 vs 0||20||143||10||Pletcher et al. 2004|
|Large (RR 1.75–2.99)||Social position||Low vs high education||77–92||263 000||1.7||Illustration of women aged 60–74 years (Huisman et al. 2005)|
|Sex (MI or CHD death)||Men vs women||50||12 292||2.2||Illustration from WHO mortality database, UK age 65–74|
|Smoking||Current vs never||20–30||12 461||2.9||Interheart (Teo et al. 2006)|
|Diabetes||Diabetic vs not||5||7 570||2.1||Effect is stronger in women (Huxley et al. 2006)|
|Depression||Depressed vs not||15||6 362||1.8||Nicholson et al. 2006|
|Systolic blood pressure||Top vs bottom third||33||34 000||1.9f||Prospective studies collaboration (Lewington et al. 2002)|
|Cholesterol||Top vs bottom third||33||34 000||1.9f||Prospective studies collaboration|
|ApoB/ApoAI ratio||Top vs bottom third||33||15 152||2.5||Interheart (Yusuf et al. 2004)|
|Waist-to-hip ratio||Top vs bottom third||33||12 461||2.0||Interheart (Yusuf et al. 2005) 1.37 after full adjustment)|
|Lipoprotein (a)||Top vs bottom third||33||5 436||1.7||Danesh et al. 2000|
|Fibrinogen||Top vs bottom third||33||7 118||2.3f||Danesh et al. 2005|
|Creatinine||15–60 ml/min/1.73 m2 vs >60||8||3 262g||1.8f||In blacks (Weiner et al. 2005)|
|Ankle brachial pressure index||<0.9 vs ≥0.9||10||388||2.3||Doobay et al. 2005|
|Medium (1.30–1.74)||Family history||CVD in first-degree relative vs none||20||15 152||1.6||Interheart (Yusuf et al. 2004)|
|Ethnicity||South Asian vs white in England||2||15 619||1.5||Wild et al. 1997|
|Exercise||Sedentary vs vigorous||45||1 646||1.6||Oguma et al. 2004|
|Von Willebrand factor||Top vs bottom third||33||1 524||1.5||Whincup et al. 2002|
|C-reactive protein||Top vs bottom third||33||2 459||1.5||Danesh et al. 2005|
|APOE E2, E3, E4||ε4ε4 vs ε3ε3||23||15 492||1.4||Song et al. 2004|
|eNOS Glu298Asp||Asp298++ vs Glu298++||11||13 876||1.4||Casas et al. 2006|
|Small (<1.30)||Sex (angina)||Men vs women||50||119 000||1.1||Hemingway et al. 2006|
|Alcohol||Never vs light/moderate||20||66 118||1.2||Corrao et al. 2000|
|Nondiabetic fasting glycaemia||Glucose 4.7–6.1 mmol/litre; per 1.1 mol/litre||2 033g||1.2||Levitan et al. 2004|
|Homocysteine||25% higher usual homocysteine||5 073||1.2||Homocysteine Study Collaboration 2002|
|APOB signal peptide ins/del||D+||52||6 007||1.2||Chiodini 2003|
|CETP TaqIB||B1+ vs B22||68||7 681||1.2f||Boekholdt et al. 2005|
|Paroxonase-1 Q192R||R192+ vs QQ||21||5 723||1.2||Wheeler et al. 2003|
|ACE insertion/deletion||DD vs I+||28||14 292||1.2||Morgan et al. 2003|
|None, inconsistent, or confounded||Vitamin E||Low vs high||1 491g||1.6||Hooper et al. 2001|
|Fish oils||Low vs high||20||1 929g||1.1||Hooper et al. 2006|
|Hormone replacement therapy||Current vs never||25||1.0||Adjusted for social status (Nelson et al. 2002)|
a Size of effect: the cut points separating small, medium, large and very large effects are somewhat arbitrary. Different subgroups (e.g. women vs men), adjustments, may change the size of effect of some risk factors, particularly between medium and large categories.
b Eligible risk factors defined by the availability of large-scale studies or systematic reviews with more than 1000 CHD cases from cohort or case control studies; scores of polymorphisms and biomarkers have been the subject of literature based meta-analysis, and only a selection are shown here; eligible preclinical measures of disease were defined by the availability of any systematic review, irrespective of size.
c Definition of risk groups: for continuous variables top vs bottom tertiles (thirds) have been used where available. For categorical variables country and sex, the effects vary, and illustrative examples are given.
d Overall prevalence of risk factor is given, but in all cases will vary within populations (e.g. by age, sex, social position and ethnicity), between populations, or both.
e Relative risks are age adjusted (except the effect for age) and are shown as risks, rather than as protective effects (e.g. alcohol, exercise).
f Individual participant data meta-analysis.
g Outcome included noncoronary events, such as stroke, and all-cause death.
For example, Heberden noted patients ‘inclining to be fat’, while Black in 1819 noted the paradox of the low rates of coronary disease in France, despite their rich diet. This illustrates an important phenomenon (apart from the observation that there are few new ideas!); the natural history of a risk factor hypothesis is long. It is about 100 years since the first blood pressure–mortality relationships were described in large populations. After the Second World War, cohort studies, such as Framingham and Seven Countries, were established as highly successful study designs and reported associations between, literally, hundreds of putative ‘risk factors’ and coronary events. This is remarkable: the apparent luxury of associations is not found for cancer and other chronic diseases, but poses a challenge—which ones are real and important?
In the 1980s Geoffrey Rose argued that public health importance relates to the mean distribution of risk factors in populations as a whole. By the 1990s it was realized that the now scores of cardiovascular cohort studies were individually small (e.g. reporting only hundreds of events), and might usefully be combined using meta-analytic techniques. In the early 21st century, the exponential proliferation of new information available in large population collections, principally from high-throughput blood-based measures, including single nucleotide polymorphisms (SNPs) and other biomarkers, and imaging capabilities (ultrasound, CT), poses the challenge of how to reconcile the molecular and the macroeconomic influences on coronary disease in populations.
Rates: disease burden: geographical, temporal, and demographic variations
What is ‘CHD’?
A clinician distinguishes between stable and unstable angina, between ST elevation and non-ST elevation MI (STEMI and non-STEMI), and between ventricular arrhythmias which cause sudden cardiac death, because the underlying vascular, myocardial, and electrophysiological pathology; the prognosis; and the treatment differ. Not so the epidemiologist. By contrast with the clinician ‘splitters’, epidemiologists have been ‘lumpers.’ Overwhelmingly, coronary disease in epidemiological studies refers to fatal coronary disease, or nonfatal MI, defined largely without troponins or electrocardiographic subtyping. Such aggregate markers of ‘CHD,’ or even more broadly cardiovascular disease (which includes atherothrombotic disease of the cerebral and peripheral circulations) have served public health well; improving risk factor burden in individuals and populations lowers CHD mortality. However, increasingly there is evidence that the causes of sudden as compared to nonsudden cardiac death, or acute as compared with chronic coronary syndromes, may differ.
People who die from coronary disease commonly suffer beforehand. Death may be sudden (presumed cardiac) in that there are no premonitory symptoms within 24 h, but in many cases this is a consequence of earlier, clinically manifest, coronary disease. The conglomerate of codes for coronary death in the International Classification of Disease has been widely used in epidemiological studies, but is a measure that conflates aetiology (first onset) and prognosis (case fatality). Age-specific incidence rates of nonfatal MI are declining in the West. However, the term ‘incident MI’ usually refers to the first MI, which may have been preceded by stable angina, in Framingham and other studies a common initial presentation of coronary disease. There is a lack of age-specific incidence data for stable angina and unstable angina, but first admission with chest pain, and unstable angina show an increasing trend from 1990 to 2000 among people aged under and over 75 years. The prevalence of stable angina pectoris, using standardized questionnaires, has shown little evidence of decline over the last two decades. The age specific burden of heart failure is also projected to increase over the next 20 years, due improved survival among patients with coronary disease.
Of all the estimated 57 million deaths in the world in 2002, CVD was the largest single cause—resulting in 17 million deaths. CVD, particularly if atherothrombotic, is an indicator of a society’s economic development. The widespread urbanization accompanying economic development is associated with the uptake of high-fat diets, tobacco use, and sedentary habits. Disability adjusted life-years (DALYs) lost to CVD are increasing globally. Developed countries experienced marked declines in communicable disease before their epidemic in CVDs; the epidemiological transition currently being experienced in developing countries is different, with a double burden of the emerging noncommunicable diseases coexisting with communicable disease. There are very large variations in rates of coronary mortality and MI incidence between countries, with approximate fourfold differences in European women. Among Russian men aged 65–74 years, rates of CHD mortality in 2002 were about 14 times higher than those in Japan. In the West, CHD mortality rates have declined by around 60% since the mortality peak in the late 1960s, though it remains the leading cause of death in women and men.
Societies with medium or high income per capita demonstrate an inverse relation between social position and coronary disease incidence and mortality. Coronary disease rates are higher with each step down the social strata, whether indexed by occupation, education or income. The inverse gradient according to social class was modest at the start of the decline in CHD mortality in 1971, but by 1981 the gradient had become considerably steeper in relative and absolute terms, and by 1991 the rate was more than double in unskilled manual workers (social class V) (230/100 000 men) compared to social class I (80/100 000 men). In 1991 the CHD mortality rate among unskilled manual workers was little different from 1971. In contrast, the CHD rate among professional workers was 60% lower than at the height of the epidemic. Using a different marker of social position, educational level, there are large differences across Europe; at all ages from 45 to aged 75 years and older, absolute mortality rate differences between those with high and low education were highest for coronary disease.
The causal mechanisms are beginning to be understood by which occupational status, education, or income influence coronary disease. Habits of living such as smoking are undoubtedly important, but independent of behavioural factors there is evidence that a range of psychosocial factors may be involved. Depression has been the most extensively investigated psychosocial factor in cohort studies and is associated with an effect of about 1.8 in a meta-analysis illustrating a potential bias whereby studies which reported effects adjusted for established risk factors had a higher unadjusted effect than studies which did not (1.9 vs 1.5). Randomized trials of cognitive-behavioural and other interventions which have reduced depression scores have not been associated with lower coronary event rates.
ESC report, cardiovascular diseases in Europe 2006. WHO data June 2006. P15. Report online at http://www.escardio.org/guidelines-surveys/ehs/Documents/EHS-CVD-report-2006.pdf
Prospective studies show that working conditions are associated with risk of incident CHD; low control, high demands, and low support at work, effort–reward imbalance, job insecurity, and organizational injustice have each been associated with cardiovascular risk. The Interheart study, a 52-country case control study of MI, found that stress at home and at work, financial stress, stressful life events, low locus of control, and depression were associated with CHD risk, accounting for 33% of the population attributable risk of MI.
Basic science describes a number of plausible physiological mechanisms that may translate chronic stress into increased coronary risk. Altered functioning of stress-related autonomic and endocrine axes has been identified among men with increased CHD risk and low self-reported control at work. Prospectively, chronic work stress increases the risks of metabolic syndrome and obesity in a dose–response manner after adjustment for smoking, diet, and physical activity levels.
Age is the strongest risk factor for all forms of coronary disease, with approximate doubling of mortality every 5 years, and doubling of incidence for each 10 years of age. The effect of age dwarfs that of other risk factors and, along with country of origin, has been proposed as the only robust measure of screening for individual risk in asymptomatic individuals. The effect of age, like that of sex, is not fixed and inevitable; it varies in different societies, e.g. with some rural populations showing little evidence of increasing blood pressure with age, within a high-rate country.
Age at presentation is increasing; the upper age limit of MI in the MONICA study of 65 years is now about the mean age of MI in many countries. The percentage of the population over the age of 65 is increasing in low, middle, and high income countries. Aided by improved survival following MI, the absolute numbers living with CVD is increasing. From the age of 50 it has been estimated that 20% of remaining life expectancy is spent living with CVD. Much of the gains in life expectancy in developed countries in recent years have come from the reduction in CVD mortality.
Women, MI, and angina pectoris
CHD is the most common cause of death in women in most developed countries. The rate of decline in the death rate due to CVD has been less in women than in men. The incidence of nonfatal and fatal MI shows a male excess (albeit of varying magnitude) across countries with widely differing absolute rates of MI mortality. When CHD rates are plotted on a log scale, there is no evidence of a threshold increase in women around the age of the menopause; the male excess in CHD mortality diminishes continuously with age. By contrast, the male excess is not so clear for angina pectoris. Male sex is not associated with physician-defined angina occurrence in either a small (146 cases in women) healthy population study (Framingham) or in a large (67 832 cases in women) study of primary care patients in Finland. In the Finnish study angina was defined by new, filled prescriptions for nitrates (reasonably specific for the diagnosis of angina, upper panel) and, separately, among patients with ischaemic abnormalities on coronary angiography or exercise electrocardiography (lower panel). While both case definitions were associated with increased coronary mortality rates in women and men, the nitrate cases showed a slight female excess, and the test-positive cases a male excess. Evidence that the latter may be biased by access to investigation for women comes from a meta-analysis, based on almost 25 000 angina cases in women and men from 31 countries, using a standardized questionnaire for assessing angina symptoms in the general population, independent of diagnostic or treatment decisions. This found that women had a slightly higher prevalence of stable angina pectoris compared to men, with a pooled sex ratio of 1.20, consistent across countries with widely differing MI mortality rates. These findings suggest that, globally, symptoms of heart disease represent an important burden in women as well as men.
Standard risk factors poorly explain the male excess in CHD mortality. While meta-analyses of observational studies suggested that current hormone replacement therapy (HRT) might be protective for women, this effect was not seen among studies controlling for socioeconomic status. The Women’s Health Initiative trial found increased CHD events (HR 1.29) among women randomized to HRT. Since there is little sex difference in aortic calcium, or in intimal thickening in intracoronary ultrasound studies among heart transplant patients, factors related to plaque development and rupture may be important. The explanation for the high burden of anginal symptoms in women, despite their lower burden of obstructive disease in the epicardial coronary arteries, may relate to microvascular disease. Indeed, a direct measure of microvascular disease in the general population, the ratio of retinal arteriole:venule diameters was strongly related to coronary events in women, but not in men, in the ARIC study.
With continuing economic development, market economy countries attract migrants from poorer nations. For some migrant groups, coronary disease rates increase to approximate those of the host country (Japanese migrants to the United States of America), but this is not the case for two broad ethnic groups in the United Kingdom. Globally, migrants from South Asia have higher rates of CHD mortality than the indigenous population. South Asians living in England and Wales have about 50% higher coronary mortality than whites, and the rate of decline between 1971 and 1991 in coronary mortality was slower in South Asians. The 1999 Health Survey for England found that the prevalence of angina was higher, particularly among Bangladeshi men compared to whites. Despite similar blood pressures and lower total cholesterol levels than whites, South Asians have a higher prevalence of diabetes, higher fasting and postglucose serum insulin concentrations, higher plasma triglyceride, and lower high-density lipoprotein (HDL) cholesterol concentrations. However, these risk factors did not explain the excess risk among South Asians in one of the few prospective studies. Some South Asian communities are socioeconomically deprived, but the ways in which material or psychosocial factors, or access to preventive health care, might influence coronary incidence in South Asians are unclear.
By contrast, British people born in the Caribbean or West Africa have lower coronary mortality rates than whites. Afro-Caribbeans have higher mean blood pressure, left ventricular mass, and body mass index, and a higher prevalence of diabetes compared to whites. African Americans have a similar profile of risk factors and similar HDL and LDL cholesterol values to those in whites, and intriguingly the Multi Ethnic Study of Atherosclerosis showed that blacks in the general population have a lower prevalence of any coronary artery calcium and a lower mean distribution of calcium scores, after adjustment for risk factors. The prognostic validity of coronary artery calcium in different ethnic groups awaits demonstration, but this may shed light on different stages in the disease process. Incidence of nonfatal MI among people aged 60–74 years is lower in African Americans than whites in the ARIC study, yet mortality from CHD (which incorporates case fatality) shows a small excess among African Americans.
Risk factors and reversal: aetiological and prognostic factors
The current causal model for CHD in populations is that a modest number of risk factors (around 10) answers, with varying degrees of success, two fundamentally different questions: Why is the rate of CHD high in a given population? Why does a particular individual have CHD? In this section the risk factors which have been most extensively investigated for a causal role in CHD are discussed in relation to the strength of observational and, where available, randomized trial evidence.
All forms of tobacco consumption increase the risk of MI, with a relative risk approaching 3 between current and never smokers. This risk increases by about 6% for every additional cigarette smoked. Second-hand smoke (passive smoking) also has an effect of about 1.15. The population attributable risks of active smoking are particularly high in young men (58%) compared with older women (6%). Smoking continues to exert an effect among those who have stopped smoking more than 20 years earlier.
At the population level, the WHO Framework Convention on Tobacco Control, introduced in 2005, sets international standards on tobacco price, advertising, packaging, education, public awareness cessation measures, illicit trade, sales to minors, and support for economically viable alternatives. In England, bans on advertising (2003) and smoking in public places and workplaces (2007) are expected to further reduce the prevalence of smoking. Among the poorest people in the United Kingdom as much as 20% of total household expenditure is spent on tobacco. Effective interventions aimed at individual smokers, include telephone counselling, busproprion, and nicotine patches.
Unlike the categorical association of smoking and coronary disease—where the goal is abolition of smoke exposure—for some other coronary risk factors the relationship is continuous. In terms of global causal impact this is best understood for blood pressure and cholesterol. There is a continuous association between blood pressure values (down to 115 mmHg systolic) and the risks of CHD. At every age there is a straight line relationship between CHD risk (log scale) and blood pressure. There is no threshold, so a lower blood pressure is associated with lower risk, irrespective of the starting level. Risk approximately doubles between the 25th and 75th centile. Most events occur at those with ‘normal’ blood pressures.
The rise of blood pressure with age is related to four dietary related factors: salt (NaCl) intake, low potassium intake, obesity, and excessive alcohol consumption; these are the targets for intervention aimed at shifting the mean blood pressure in the whole population. Some 26% of the world’s population (972 million people) are estimated to have elevated blood pressure, of whom 639 million are in developing countries. At an individual level, dietary change, reduced alcohol, and reduced weight can all lower blood pressure. Pharmacological blood pressure lowering among people with blood pressure >140/90 clearly reduces cardiovascular events, and probably coronary events.
Diet and lipids
Dietary habits cause coronary disease; the evidence is strong for saturated fat (via effects on serum cholesterol), good for the multiple dietary influences which affect blood pressure, but lacking—or controversial—for antioxidant vitamins (β-carotene, vitamin C and vitamin E), B vitamins and folate, and fish oils.
Saturated fat, cholesterol, and other lipid subfractions
Total cholesterol, LDL cholesterol
Saturated fat consumption is associated with serum cholesterol levels. Attempts to define ‘normal’ for serum cholesterol levels are misleading: there is a strong and continuous between-country correlation between average serum cholesterol and coronary mortality, ranging from 6 mmol/litre in Britain to 3–3.5 in rural China. LDL cholesterol, about 2 mmol/litre lower than total cholesterol, is responsible for much of the atherogenic effect of lipids, and the effects in observational studies and trials for LDL and total cholesterol tend to match closely. Within populations the relationship between total cholesterol and CHD mortality is log linear, which links the proportional change in the cholesterol to an absolute change in CHD mortality. There is no threshold; even in China, where the overall mean was low, there is a dose–response relation with CHD mortality.
Population reduction of saturated fat consumption by about 7% of total calories would lead to about a 0.6 mmol/litre reduction in serum cholesterol. This has occurred in the United States of America and in Finland. For example, the North Karelia project in Finland reduced the intake of saturated fat from liquid dairy products and spreadable fat by two-thirds between 1972 and 1992, with reductions in mean cholesterol levels. Denmark has banned industrially produced trans-fatty acids on the grounds that they adversely affect LDL levels and can be removed without effect on the taste, price, or availability of foods. At the international policy level it has been claimed that the European Union has maintained subsidies to producers of full-fat milk and beef while keeping the cost of fruits and vegetables high. Individual dietary change can achieve relatively modest (e.g. 5%) reductions in serum cholesterol. Statins reduce coronary event rates, among people with established CVD (Heart Protection Study) and among individuals without established CHD.
HDL cholesterol, apolipoproteins B and A1
High density lipoprotein (HDL) is protective for coronary disease; in cohort studies an absolute increase of 0.12 mmol/litre is associated with a 15% decrease in CHD and treatment with recombinant apolipoprotein A1 is associated with regression of coronary atherosclerosis. Statins increase HDL by only about 5%, so this effect on CHD events is small. HDL cholesterol is highly correlated with apoliporotein A1, and LDL cholesterol with apolipoprotein B; the apolipoproteins have the advantage of not requiring a fasting sample for measurement. The ratio of apoB/apoA is strongly related to MI with relative risk of about 2.5 for top vs bottom third (see Table 1).
Environmental factors influence an individual’s cholesterol level via pathways under genetic control, although for the most part the genes responsible are not known. Familial hypercholesterolaemia (prevalence 1 in 500) is an exception; caused by mutations in a single gene (for the LDL receptor), familial hypercholesterolaemia is inherited as a simple mendelian dominant trait, and has such large effects on cholesterol levels it might be considered as a linear, deterministic factor in causing CVD. Recent evidence suggests that this may not be the case. Based on a large pedigree of 250 individuals, with probands selected during screening (and not on the basis of having suffered CVD), mortality was not increased in carriers of the mutation during the 19th and early 20th century; it rose after 1915 and reached its maximum between 1935 and 1964 (standardized mortality ratio 1.78, 95% confidence interval 1.13–2.76), and fell thereafter. Furthermore, mortality differed significantly between two branches of the pedigree. This large variability over time and between branches of the pedigree points to a strong interaction with environmental factors, such as changing dietary patterns and medical care. In contrast with familial hypercholesterolaemia, CHD and stroke are complex diseases that are known to aggregate in families but not to segregate in a mendelian fashion. Under the common disease/common variant hypothesis a number of genes, each with modest effects, are thought to operate. People with the E4 allele of the apololipoprotein E (APOE) gene have higher serum cholesterol concentrations than those without the allele and this accounts for an appreciable proportion (about 7%) of the variability in cholesterol levels in the general population. E4 allele frequency differs widely between populations, with a lower prevalence in Japan (7%) compared to Finland (19%), and a gradient of decreasing prevalence from north to south Europe. The APOE gene E4 allele has been shown in a recent meta-analysis of 48 studies with over 15 000 cases to be associated with coronary artery disease with a relative risk of 1.42.
B vitamins, folate, homocysteine
Plasma homocysteine is associated with CHD, with meta-analyses suggesting a relative risk of 1.13 for a 2.6 µmol/litre higher homocysteine. In global public health terms this observation is of particular interest because there is a simple intervention—increasing folate consumption substantially decreases homocysteine levels. However, there is major scope for confounding (homocysteine levels are associated with smoking and low social position) and reverse causality (CVD may lead to elevated homocysteine levels). Recently common genetic variants have been used to help understand whether specific environmental exposures are causal. A common genetic variant in the MTHFR gene (677C to T substitution) is associated with homocysteine levels, such that those with the TT genotype have about 2.6 µmol/litre higher homocysteine levels than those with the CC genotype. Since, under Mendel’s second law, assortment of alleles at the time of gamete formation is random, confounders are likely to be evenly distributed by genotype, a situation analogous to a randomized controlled trial. Under the Mendelian randomization design, the association between genotype and coronary disease predicts that TT individuals have a risk of 1.16, which is very close to that observed, lending some support to causal inference. The complementary randomized trials of folate supplementation in a primary prevention (aetiological) setting are awaited, although trials in a secondary prevention (prognostic) setting have been negative.
Other dietary components
The search for other causal moieties in diet has seen a pattern of apparently promising associations in observational studies not being confirmed in randomized trials. The antioxidant vitamins β-carotene, vitamin C, and vitamin E have each fallen at the hurdle of intervention, and ω – 3 polyunsaturated fatty acids do not show clear and consistent associations in observational studies or trials. However, available evidence does not exclude an important effect of ω – 3 fats on coronary events; less biased trials reporting more events are required. Whole dietary patterns may be additionally important.
There is a J-shaped relationship between alcohol consumption and CHD, with risks elevated slightly in never drinkers, and those who drink excessively. However bi-directional confounding may be important in this relationship. People who report light alcohol consumption have a lower prevalence of 27/30 risk factors and behaviours compared to nondrinkers. By contrast, in heavy drinkers, confounding will obscure rather than exaggerate any coronary protection because of their heart-unhealthy behaviours. Thus the evidence for a coronary-protective effect is probably stronger for moderate to heavy drinking than for light to moderate drinking. In public health terms a more important question is not whether any given level of alcohol consumption is protective for coronary disease, but its overall relation with harms. Light to moderate drinking might have a small, unconfounded protective association with CHD; but any increase in the mean population per capita alcohol consumption will increase the prevalence of heavy drinking with its known (noncoronary) harms.
Physical activity, through pleiotropic effects on blood pressure, lipids, glucose and fat utilization, obesity, and cardiorespiratory fitness, is causally associated with the incidence of coronary disease. While the optimal type, intensity, duration, and frequency of exercise are continued grounds for debate, there is a dose–response relationship, with CHD risk decreasing with increasing amounts of physical exercise. Even walking for 1 h/week is associated with reduced risk of CVD outcomes. Higher resting heart rate, strongly associated with sedentary habits, is a consistent predictor of coronary events, particularly sudden cardiac death. Low heart rate rise with exercise, and fall after exercise, may also predict sudden death.
Obesity and diabetes
The combination of gluttony and sloth, more socially patterned than individually chosen, is causing an epidemic of obesity from childhood onwards in developed, and developing, countries throughout the world. The general thesis that obesity causes coronary disease is not contested, but there is much debate about how best to measure obesity and the mechanisms of causality. Based on 74 analytic cohorts with 60 374 deaths during follow-up, the associations between body mass index (BMI) and CHD mortality were modest with relative risk 1.51 in men for obesity (BMI >30) as compared to individuals with BMI 18.5 to <25. There was little evidence that overweight (BMI 25 to <30) was associated with mortality. In Interheart BMI showed a modest and graded association with myocardial infarction that was substantially reduced after adjustment for waist-to-hip ratio (1.12), and nonsignificant after adjustment for other risk factors. Among people with established disease a meta-analysis of 40 studies found lower cardiovascular mortality among the overweight and mildly obese groups. Waist-to-hip ratio is a simple surrogate measure of visceral obesity, a key determinant of metabolic abnormalities. Waist-to-hip ratio was associated with risk of MI (odd ratios for top vs bottom third 2.0), even after adjustment for other risk factors.
Diabetes and impaired glucose tolerance
The number of diabetics in the world is estimated to rise to 221 million by the year 2010 (from 124 million in 1997), with Asia and Africa experiencing the greatest increases. Diabetes is a strong risk factor for CHD, stronger in women than in men. There is some evidence that the 2-h blood glucose is a better predictor of deaths from CVD than is fasting glucose. Nondiabetic hyperglycaemia may have a weak (risk ratio 1.19) association with CHD. The environments that cause obesity also cause type 2 diabetes. Interactions with genes are likely to be important, with one promising candidate being transcription factor 7-like 2 gene (TCF7L2). Unlike many previous polymorphisms reported for chronic diseases, the association has been replicated in multiple populations, appears strong, (relative risk exceeding 2 for the homozygote) and population attributable risk is high (21%).
Interventions for obesity and diabetes
Obesity may represent ‘normal physiology within a pathological environment’ encompassing sedentary lifestyles and overly rich nutrition. Strategies that specifically reduce abdominal obesity have not been developed, but overall weight loss probably reduces abdominal obesity. Exercise is associated with improved CVD risk factors even if no weight is lost. Diabetes is potentially preventable by implementation of dietary changes at the prediabetic stage of impaired glucose tolerance.
Clusters of risk factors: the metabolic syndrome
Although it is clear that risk factors cluster within individuals, it remains unclear that the whole of the (variously) defined metabolic syndrome predicts risk of CHD events beyond the individual component variables. A recent meta-analysis of eight studies examining CHD incidence reported an effect for the metabolic syndrome of 1.5. In nine European population-based cohorts the prevalence of the metabolic syndrome differed widely according to definitions: WHO (27% in men, 20% in women), the National Cholesterol Education Program (26%, 23%), NCEP revised (32%, 29%) and the International Diabetes Federation (36%, 34%). With CVD mortality as the outcome, the respective hazard ratios were 2.1, 1.7, 1.7, and 1.5 in men, and consistently lower in women—1.6, 1.4, 1.1, and 1.5. With a few exceptions, the hazard ratios for full definitions of the syndrome were not significantly different from those for their single components. Serum triglycerides, which are strongly related to LDL and HDL cholesterol, may be associated with CHD risk, but the independence of this association remains contentious.
Combining risk factors to predict risk in individuals
Combinations of risk factors may be used to predict the risk of coronary events in individuals and the use of scores is increasing in clinical practice.The absolute risk of coronary events at a given level of one risk factor is determined by the levels of other risk factors. The premise of risk prediction is that those at higher absolute risk have the greatest individual benefits from risk factor reduction, and people who perceive themselves at risk may be more motivated to change their health related behaviours. A further advantage is one of explicit prioritization; patients at a given level of risk may be targeted, irrespective of how that risk was reached.
However, the identification of individuals at risk is not straightforward. Existing scores, based on combinations of risk factors, do not reliably identify a minority of high-risk patients who will experience the majority of CHD events. Most events occur in people who are not at high risk. Nor can risk factor screening in the population identify a group of individuals who will not benefit from reductions in risk factors. The technology of calculating risk in individuals has not been tested in randomized trials. Two simple pieces of information—older age and living in a country with a high rate of CVD—are the most robust identifiers of risk. To date the hope that novel biomarkers might substantially improve risk prediction has not been substantiated. If causal risk factors for disease have—thus far—been poor screening tools, then measuring the consequences of early, preclinical stages of disease may be more helpful.
Inflammatory, haemostatic, and other circulating biomarkers
The hypothesis that markers of inflammation may be causally associated with coronary events has had some empirical support, although the effects of C-reactive protein (CRP) appears modest, and Mendelian randomization studies suggest that the effect might be confounded by its strong associations with established risk factors. Fibrinogen is the only biomarker which has been the subject of large individual patient data meta-analysis. While fibrinogen is associated more strongly (2.3) with CHD events, similar genetic evidence suggests these effects might be confounded. A further challenge in assessing the causal contribution of these factors comes from the absence of specific therapeutic agents which alter CRP or fibrinogen activity. A potentially promising line of enquiry involves relating the risk of inflammatory disorders, e.g. psoriasis or rheumatoid arthritis, to risk of subsequent disease. Although issues of confounding still pertain, the reported risks are of intrinsic clinical interest.
Lipoprotein (a) (Lp(a)) is a putative biomarker subject to less confounding by environmental factors because of its high heritability (0.93–0.98). With a homology to plasminogen and therefore a potential role in haemostasis, Lp(a) is associated with coronary events with a relative risk of about 1.7. Serum creatinine level, a ubiquitously performed test in which there is little direct commercial interest, may offer clues to aetiology of importance to clinical practice. In an individual patient data meta-analysis, a glomerular filtration rate between 15 and 60 ml/min per 1.73 m2 was associated with CVD events in black individuals more strongly than in whites. This may reflect more frequent or severe subclinical vascular disease secondary to hypertension or diabetes in black individuals.
Over 100 circulating biomarkers have been associated with coronary disease aetiology or prognosis, and more than 10 of these biomarkers have been subjected to meta-analytic scrutiny; emerging findings suggest that individual biomarker effects tend to be modest, confounded, or subject to biased reporting. Even if a biomarker were found with a very large effect (odds ratio of 3 or more), this may add little to risk prediction. Combining sets of biomarkers in panels, and taking a systems biology approach may offer more promise, although there has been little empirical support to date. A recent study found that measuring multiple acute phase reactants, proinflammatory molecules and markers of endothelial activation in the post-MI setting added little beyond established risk factors to risk prediction.
Parental history of CHD confers an approximate doubling of risk of CHD among the offspring and, in the Framingham study, those with a sibling with CVD were also at increased risk beyond that conferred by parental CVD and established risk factors. The possibility that the sibling effect may be stronger than the parental effect suggests that early life environment, as well as shared genes, may be important. Many quantitative traits including measures of disease phenotype—proximal coronary artery stenosis, coronary artery calcification—and intermediate phenotypes (including blood pressure, BMI, cholesterol) show modest to high heritability coefficients (0.2–0.8). Genome-wide association studies investigating tens of thousands of SNPs have identified a small number of genes significantly associated with MI. Candidate gene approaches, summarized in literature based meta-analysis, have identified SNPs in more than 10 genes that are associated with MI. The risk alleles are common, but the reported effects are small, commonly less than 1.3. How these genes inter-relate with other genes or with environmental factors is not known. Within a generation, heritable changes in function and expression of genes not involving DNA sequence (epigenetics) are emerging as important mechanisms by which the environment might influence disease.
Trajectories of change in blood pressure, cholesterol, and markers of carbohydrate metabolism have their origins in early life; values track within individuals over time and measurements made in the second and third decade of life predict manifest coronary disease in mid and later life. Correlations between ischaemic heart disease mortality with both neonatal and postneonatal mortality stimulated the hypothesis that impaired fetal growth may influence adult coronary disease. Identifying which aspects of growth and development are most important is ongoing, although recent meta-analyses suggest that birthweight is not causally associated with either systolic blood pressure or cholesterol in adult life.
Imaging and screening for preclinical coronary disease
Measuring arterial, myocardial, or electrophysiological disease processes before the first symptomatic manifestation of disease is important, both to understand disease mechanisms and potentially as screening tools to improve the identification of high risk individuals. Although individual studies were small and take variable account of established risk factors; coronary event rates are lowest among those without subclinical disease, intermediate among those with subclinical disease, and substantially higher among those with manifest disease. A growing number of measures are available, although optimal combinations or sequential testing strategies await clarification.
The resting ECG is the most widely assessed measure of subclinical disease, although robust estimates of the predictive value of its many parameters have not been subjected to systematic review. At least one (categorical) ischaemic abnormality on the resting electrocardiogram is found in about 25% of the ‘healthy’ population, and even changes that have been termed ‘minor’ are associated with subsequent coronary event rates. Left ventricular hypertrophy, a less common abnormality, is associated with an approximate threefold increase in events. Continuous measures, such as QRS/T angle and high QRS nondipolar voltage, may further add predictive information. For exercise ECG, population-based studies of asymptomatic, healthy middle-aged people show that ST depression is a strong predictor of CHD events, with relative risk of 2.8 to 10 (based on 9 studies).
Coronary artery imaging for calcium
Currently the only direct population-based measure of the presence and extent of coronary artery disease comes from calcium scores, using electron beam CT. Although atherosclerosis may be present in the absence of calcium, the presence of calcium in the coronary arteries is highly specific for atherosclerosis. Among asymptomatic middle-aged individuals about 70% of white men and 45% of white women have some coronary artery calcium. Early evidence suggests a strong dose response relationship between increasing score and increasing coronary event rates. Compared to individuals with no coronary artery calcium, the odds ratios (coronary artery calcium scores) in a recent small meta-analysis were 2.1 (1 to 100), 5.4 (101–400), and 10 (>400).
Structure and function of peripheral arteries
Population studies increasingly incorporate measures of large and small artery structure and function outside the coronary circulation. The ratio of the systolic blood pressure assessed in the brachial and ankle arteries is associated with cardiovascular mortality, independent of other risk factors, with a relative risk of about 2. The ankle brachial index is a simple, low-cost measure, with high reliability. Common carotid intima media thickness, assessed with B mode ultrasound, is associated with risk factors and predicts future coronary events. For example, using the cut point of 1 mm, the relative risk in the ARIC study was 5.07 in women and 1.85 in men. Systemic arterial stiffness and pulse wave velocity have also been related to coronary events. Small-artery compliance or stiffness is a marker for endothelial dysfunction that induces a functional change in the microcirculation and is a predictor of adverse cardiovascular events.
Synthesis and strategy: implications for policy
High-risk approach to prevention
Geoffrey Rose presented two complementary strategies for prevention. The ‘high-risk’ strategy, the traditional medical approach to prevention, identifies individuals at high risk of subsequent CVD events who are then offered behavioural or pharmacological interventions. However, there are limitations to the high-risk approach to preventing coronary disease, the first being that current risk prediction tools have modest discriminatory performance.
The second limitation is that while there are effective methods to modify dietary and smoking behaviour, their influence on blood pressure, cholesterol or BMI has tended to be disappointing. Unfortunately, there are no satisfactory clinical trials that evaluate the effect of a 30% fat or a 20% low-fat diet on clinical CHD. Pharmacological modification of risk factors such as lipid levels (with statins) and blood pressure has been demonstrated to have an effect in randomized trials. Controversially, it has been proposed that individuals aged 55 and over should be offered a combination pill (‘polypill’) containing low doses of hydrochlorothiazide, atenolol, enalapril, simavastatin, aspirin, and folic acid. Although randomized trials have yet to be carried out, such an approach is predicted to lead to a reduction in ischaemic heart disease events by 88%.
A third issue is that even when quality of care can be defined, international surveys repeatedly show that it is suboptimal. For example, in a survey in 15 European countries of patients with coronary disease, behavioural modification and pharmacological control of cardiovascular risk factors was poor.
A fourth matter is that uptake of screening or interventions may favour higher socioeconomic groups, and this may contribute to the slower declines e.g. in smoking prevalence in more deprived groups.
It is also the case that compliance with treatment regimens in chronic asymptomatic diseases remains difficult to achieve, and most fundamental of all, the high-risk approach is palliative and temporary in that it does not seek to alter the underlying causes of the disease but to identify individuals who are particularly susceptible to those causes.
Population-based strategies for prevention
A more radical approach to prevention, the ‘population strategy’, seeks to control the determinants of incidence in the population as a whole. The risk associated with many clinical variables—blood pressure, cholesterol—is continuous with no evidence of threshold. Such variables are strong, consistent predictors of CVD in groups, but are poor predictors of susceptibility in individuals. The distributions of blood pressure (top panel) in five populations taken from the Intersalt study; with increasing median values their normal distributions shift rightward, towards higher risk. Interventions on the whole population will shift the distribution of a risk factor in a favourable direct (to the left in this case); a 5 mmHg lowering of the population mean blood pressure would correspond to about a 20% reduction in the coronary event rates.
Globalization and risk factors
An important mechanism for bringing about improvements of the cardiovascular health of whole populations lies in national and international fiscal, regulatory, and legislative change—so-called ‘health protection’—in which individual members of the population need have no knowledge or understanding of the changes, nor how they may be linked to health. The changing ways in which people now interact across physical, temporal, and cognitive boundaries, most commonly referred to as globalization, is increasing interdependence, integration, and interaction between people and companies in disparate locations and influencing cardiovascular health in disparate ways. Public health policies need to take into account the role that agriculture, trade, education, the physical environment, town planning, and transport have on CVD aetiology.
Explaining temporal and geographical variations with risk factors
The Interheart study concluded that nine risk factors—smoking, apoB/apoA, high blood pressure, diabetes, abdominal obesity, psychosocial factors, fresh fruit and vegetables, alcohol, and physical activity accounted for 90% of the population attributable risk for myocardial infarction. The WHO Multinational Monitoring of Trends and Determinants in Cardiovascular Disease (MONICA) Project found that trends in risk factors, measured since the 1980s in 38 populations from 21 countries, explained much of the variation in population trends in MI before age 65 years. However, several countries continue to offer paradoxes. The rapid increase, and then decrease, of CHD mortality in Russia during the 1990s occurred without major changes in smoking, blood pressure, and cholesterol; it may have reflected social and economic stressors, in combination with high alcohol consumption. Japan has had a substantially lower coronary mortality rate than other developed countries despite a prevalence of cigarette smoking among the highest in developed countries, cholesterol levels comparable to those of the United States of America, and increasing prevalence of obesity and diabetes.
Likely future developments
If globally the CVD epidemic is growing more desperate, future remedies must be commensurate; policy approaches to the prevention of CHD will increasingly take an international or global perspective. The WHO Commission on Social Determinants of Health, which reports in 2008, is expected to demonstrate the pre-eminent role of social and economic factors in defining the risk behaviours, and their psychosocial context within countries, and to make recommendations for achievable ways in which societies across the development spectrum, should change. WHO is shortly to announce its guidelines for the prevention of CVD.
A key tension for the future is the reconciliation of two perspectives: the cultural and the cellular, the macroeconomic and the molecular. Such wide variations in the rate of CHD across the globe and, particularly, such rapid changes in CHD rate over time among migrants and indigenous populations have, traditionally, been interpreted as leaving little scope for the play of genetic factors in global cardiovascular health. In the next decade this position is likely to be nuanced, if not recast. Epigenetics might prove a core discipline in this regard, with early reports suggesting strong associations e.g. between smoking and CpG island methylation, and animal models suggesting that environmental interventions (diet) can lead to epigenetic changes, which in turn lead to altered gene expression. Already population geneticists are reporting large genetic differences across current political borders over which there has been large and recent two-way migration. Black and white populations show large differences in frequency of common polymorphisms affecting the metabolism of an increasingly prevalent set of environmental exposures—therapeutic medication. Improvements in individualized risk prediction for CHD are likely, although the process of sifting the huge amount of information into knowledge is likely to be slow and complex. Using a panel of information derived from large numbers of ‘-omic’ variables might generate signatures that supplant and surpass current efforts to identify high risk individuals; expression studies are already claiming this for the prognosis of cancer.
However, without refining the measures of disease outcome—separating plaque development, plaque rupture, thrombotic and necrotic processes—efforts to dissect out the molecular causes may be blunted. Epidemiological studies will increasingly complement the study of the presence or absence of coronary disease (usually undifferentiated MI) with measures of how much coronary artery disease and which manifestation. Higher endpoint resolution of known phenotypes and ‘class discovery’ of new phenotypes are both important. As radiation dose decreases for CT techniques, it may be possible to image the coronary arteries in the general population for plaque burden and (with molecular imaging) composition.
Over the next decade epidemiology as applied to CHD is likely to become more complex. There has been much success with the epidemiological paradigm, likened to a game of billiards, where one exposure of interest (the cue ball) is linearly associated with one outcome, independent of confounders (a third ball getting in the way). It remains to be seen how this (simple) paradigm evolves in relation to new insights from a widening range of disciplines: from social science, through to systems biology and bioinformatics. From a public health perspective epidemiology is predicted to remain, as William Petty wrote in 1690, political arithmetic.