3 The epidemiology of heart failure

Over the last 30 years we have gone from famine to feast in terms of the epidemiological data now published for heart failure (HF). The field started with the seminal publication on the natural history of HF from the Framingham study in 1971 showing a prevalence of HF of 0.8% in those aged between 50 and 59, rising to 9.1% in those over 80 years with incidence rates of 0.2% at age 54 and 0.4% at age 85 (Fig. 3.1).¹ This was followed by a large European study, ‘The men born in 1913’, which gave similar figures of a prevalence of 2.1% at age 50 and 13% at age 67 and incidence rates of 0.15% and 1% respectively at ages 50 and 67.²

These landmark studies relied on a clinical diagnosis of HF, based on symptoms, signs, and scoring systems to identify cases. More modern epidemiological studies have used definitions of HF which include objective measures of cardiac function in their definition, in keeping with current European and United States guidelines for the diagnosis of HF. Initial studies focused on systolic dysfunction because they reported at much the same time as the HF treatment trials which also enrolled patients with systolic HF. More recently attention has turned to describing the epidemiology of HF with preserved systolic function, in addition.

When describing the epidemiology of HF, it is worth bearing in mind that estimates of incidence and prevalence will vary according to the definition of HF used and the type of cohort being studied. This is especially important when assessing work which has objectively measured left ventricular systolic function. Variables such as left ventricular ejection fraction are normally distributed, so the cut point chosen is a critical determinant of the eventual results.

The present chapter aims to outline the contemporary epidemiology of HF by describing its prevalence, incidence, aetiology and mortality as well as describing the trends which are occurring in the area. It will discuss hospitalization rates, prognosis and economic burden in both Europe and the United States.

Many studies have been conducted in primary care or across geographical health care communities to estimate the prevalence of HF. One of the first was conducted in north-west London and reviewed 30 204 case records, yielding a crude prevalence of HF of 3.8/100 cases in the general population with a marked rise from those under 65 to those above 65 years of age, where the rate rose from 0.6 per 1000 to 28.0 per 1000.³

More recent data is available from the Scottish Continuous Morbidity scheme which covers 57 general practices in Scotland with access to GP Read codes for HF in 307 741 patients.⁴ The calculated prevalence within the general population in Scotland was 7.1 per 1000, increasing in the population above 85 years old to 90.1 per 1000. The population identified by primary care was more elderly, and had more comorbidities than in population-based studies or clinical trial populations. The findings have been corroborated in a European study based in Utrecht, Netherlands. It found that patients with HF who were under the supervision of a cardiologist compared to a general practitioner were more likely to be male, younger (in their sixties), and to have an ischaemic aetiology.⁵ When considering such data it must be remembered that the signs and symptoms of HF are neither sensitive nor specific. Studies evaluating referrals from primary care, when compared to expert cardiological assessment, have revealed only approximately 30% of patients may actually have HF.^6,7

A recent study in Sweden reiterated this salient point. Random primary health care centres were picked from across the country. Medical records were interrogated and variables recorded. Approximately 30% of the participants had had an echocardiogram. The majority were labelled as having HF on the basis of signs and symptoms and basic investigations including chest radiographs and the electrocardiogram. There was also an underuse of evidence-based therapies.⁸

The North Glasgow MONICA Study was the first to report on the prevalence of left ventricular systolic dysfunction in a random sample of the general population of 2000 men and women aged 25–74 years.⁹ In this cohort, 2.9% had significant left ventricular systolic dysfunction, of whom just over half had symptoms of breathlessness or were taking a loop diuretic. The estimated prevalence of HF was thus 1.5%, with 1.4% having the important precursor of HF, asymptomatic systolic dysfunction (ALVSD). The prevalence rose with age and was higher in men than women (Fig. 3.2).⁹

Many studies have reported subsequently both in Europe and in the United States. Data from these cohorts is fairly consistent for the general population. Prevalence rates for left ventricular systolic dysfunction (LVSD) were 1.8–3.5% in the ECHOES study from the English Midlands, with 50% of the LVSD being asymptomatic; in the US Olmsted county study 2.2% had HF validated using the Framingham criteria, and of these 56% had systolic dysfunction.^10,11

When we look at population-based studies which have included much older subjects, the prevalence rates increase markedly. In the Helsinki Ageing Study of 501 subjects aged 75–86 years, clinical HF was found in 8.2% overall, 2.3% had systolic dysfunction, and 9% had ASLVD.¹² In the Rotterdam Study of 2267 men and women aged 55–95, 3.7% had fractional shortening of less than 25% (5.5% men and 2.2% women) and 2.2% had asymptomatic LVSD (see Fig. 3.3).¹³ Similar findings were reported in a United Kingdom study of 817 subjects aged 70–84 years from Poole (on the south coast of England) which demonstrated that 7.5% had LVSD (12.2% men and 2.9% women) and 52% of those with LVSD were previously undiagnosed.¹⁴

Many of the population-based cohorts reviewed above concentrated on finding systolic dysfunction, as it is, to date, the only type of HF for which we have evidence-based treatment. Many of the cohorts have also by default or design been able to comment on the prevalence of HF with normal ejection fraction (HeFNEF). Hogg et al. reviewed the epidemiological data for HeFNEF. The prevalence ranged for 1.5% to 4.8% depending on the study. There was a definite increase in the proportion of HF due to HeFNEF in cohorts that contained more elderly individuals.¹⁵ In the ECHOES study of the general population, 1.1% had definite HF and a LVEF greater than 50%,¹⁰ whereas in the Helsinki Ageing Study, 72% of all the HF identified occurred with a normal LVEF.¹¹ In the United States, the Rochester Epidemiology Project found similar results in a random sample of 2042 subjects over 45 years, with 44% of subjects having HF with a LVEF greater than 50%.¹¹

Even higher prevalence rates have been found in a recent large cross-sectional study from Portugal.¹⁶ This showed the prevalence of HF was 16.1% in the population above 80 years old. The prevalence was roughly split equally between normal and reduced ejection fraction.

The above studies all confirm one thing: a large prevalence of HF which increases exponentially with age. It is unsurprising, therefore, that the current burden of HF within the European Union member states is estimated to be approximately 15 million¹⁷ and, according to the American Heart Association, more than 5 million Americans have HF.¹⁸

Contemporary studies of incidence are far fewer than those for prevalence (Table 3.2).

The incidence of HF was ascertained in the west London district of Hillingdon. Here, all incident cases of HF were identified via either a specialist referral clinic or emergency admission.¹⁹ The district contained 151 000 patients covered by 82 general practices. Using both portals of entry in the study, 220 new cases were identified. Participants had a full clinical assessment, standard investigations including a chest radiograph and ECG and 99% of the study population had an echocardiogram. The results were then shown to a panel of three cardiologists who made the gold standard diagnosis. The documented incidence rose from 0.02/1000 per year in the 25–34 age group to 11.6/1000 in those aged over 85 (Fig. 3.4).¹⁹ There was a preponderance of impaired systolic function. The study confirmed that HF is predominantly a disease of elderly people, with a median age of first presentation of 76 years.

Incidence data for the United States are available from the Cardiovascular Health Study (CHS) showing an incidence rate of 19.3/1000 person-years in 5.5 years of follow-up.²⁰ Data are also available for incidence from general practice records. From the General Practice research database (GPRD) in the United Kingdom (administered by the Office of National Statistics), 696 884 potential patients aged above 45 years old were identified.²¹ The records were interrogated and were categorized on the basis of records and medication prescription patterns. Using this approach, 6478 patients with definitive HF, 14 050 with possible HF, and 6076 with diuretics but a non-heart-failure diagnosis were identified. The overall incidence of definitive HF was 9.3/1000 per year, but if the possible HF group was included, the incidence increased to 20.2/1000 per year. The mean age of the definite HF population was 77 years. More recently, data from the Scottish Continuous Morbidity Recording data set showed an overall incidence of 2/1000 population per year: it was 25/1000 per year in men over the age of 85 years.²²

The majority of epidemiological surveys have concentrated on white populations, with a bias towards relatively affluent areas of the Western world. However, data from more diverse populations are now emerging. Recent work from an elderly institutionalized population in Memphis and Pittsburgh showed some differences in incidence with race, at least in the United States. The annual incidence rate in African Americans was 1.63% and in white Americans 1.19%.²³ In addition, a recent study in a younger population of 5115 participants between 18 and 30 years old at baseline followed for 20 years from Oakland, California; Chicago, Illinois; Minneapolis, Minnesota; and Birmingham, Alabama showed a mean age of onset of HF of 39 years.²⁴ The cumulative incidence for black men and women respectively was 0.9% and 1.1%. There was also a high prevalence of asymptomatic echocardiographic LV

systolic impairment. Rates of 9% in whites and 13% in African Americans were documented. This work highlights the need for more studies of incidence in ethnically diverse populations.

Data from the Framingham study has not shown any increase in incidence since the 1970s, dispelling the theory that we are experiencing an epidemic of HF. Similarly, data from Medicare records show a slight reduction in incidence from 57.5/1000 to 48.4/1000 person-years in the 80–84 year age group in the period 1994–2003. However, despite the slight reduction in incidence, the prevalence rate rose markedly from 90/1000 to 120/1000.²⁵ These trends will continue with the changing demography of most Western populations, with more elderly individuals and a greater number of survivors from cardiovascular disease earlier in life.

Determining the exact aetiology of HF in patients in epidemiological studies can be difficult. The commonest cause within the Western world currently is ischaemic heart disease (IHD), which represents a change in aetiology over time. When the Framingham study first reported, the predominant aetiological factor was hypertension. In this study, the influence of coronary heart disease has increased over time by 40% in men and 20% in women.

In the North Glasgow MONICA Study, over 95% of patients with symptomatic LV systolic dysfunction had some evidence of prior IHD, although hypertension was also common in this group, occurring in 68%. Other data from prevalence studies show similar results. In the ECHOES study, 53% of those with systolic dysfunction had evidence of IHD and 42% had hypertension; in the Helsinki Ageing Study it was 54% both for hypertension and for IHD. US data from the CHS confirm similar results with the population attributable risk for HF being 13.1%for coronary heart disease and 12.8% for hypertension. Both are clearly important aetiological factors.²⁰

In the original Hillingdon study of incident HF, 41% of the HF cohort was due to coronary artery disease and much less, 6%, had hypertension.²⁶ A subsequent study carried out in Bromley (south London) looked into putative ischaemic aetiologies in more depth. All incident cases of HF were identified and referred to a specialist dedicated clinic or identified via tracking patients during their hospitalization. In this study 332 patients were identified and 99 of the 136 cases under the age of 75 years also underwent coronary angiography. An ischaemic aetiology was eventually attributed to 52% of the 136 cases.²⁷

Hypertension as a cause of HF still seems to predominate in those with HF and normal ejection fraction, where IHD seems less prominent. Patients with HeFNEF tend to be older and are more likely to be female. Both ischaemia and hypertension are still common; a recent study by Zile showed a prevalence rate of hypertension of 82% and congestive heart disease (CHD) of 45% in patients with HeFNEF.²⁸

The traditional risk factors that lead to myocardial disease have been extensively studied in predominately white populations. Increasing awareness that the same diseases may manifest and act differently in certain ethnic populations has led to some recent studies. Hypertension is a contributory factor in young African Americans below the age of 39 and results in a HF incidence 20 times the rate of whites.

South Asians seem to have a susceptibility to premature accelerated coronary artery disease.^29,30 A series of studies looking at South Asians hospitalized with HF in Leicestershire, United Kingdom and Toronto, Canada has shown that admission rates for HF were higher than in the white population, and there was more evidence of diabetes and hypertension. These findings may mean that the coronary artery disease leads to the development of HF at a younger age in a South Asian population.^31,–33

Traditionally, HF in the developing world has been viewed as a result of either infective or nutritional disease. Rheumatic heart disease is still endemic in sub-Saharan Africa.³⁴ However, the epidemiological data from traditional developing areas of the world is limited. Certain areas of the world which were previously viewed as ‘developing’ are now undergoing epidemiological transition to a more affluent, sedentary, urbanized population. This also means an epidemiological transition in terms of cardiovascular disease from infective and nutritionally based heart disease to that of atherosclerosis and progressive coronary artery disease.

Because of its high prevalence within elderly populations, HF is commonly associated with numerous comorbidities, including renal impairment, anaemia, diabetes mellitus, obstructive sleep apnoea, and chronic obstructive pulmonary disease. These all have an adverse impact on survival when associated with HF.³⁵

Anaemia was present in 51% of patients with HF in the Rochester Epidemiology Project. Severely impaired renal function was present in 10%. These rates are increased in patients presenting with acute HF syndromes; renal dysfunction occurred in 20% of those admitted with decompensated HF in the Euroheart Failure Survey II.³⁶

The 32-year follow-up of the Framingham study signposted the awful mortality rate associated with HF. The probability of dying from HF was 62% for men and 42% for women at 5 years of follow-up from incident diagnosis. However, data from the Framingham study have shown consistent improvements in survival over time for both men and women.³⁷ In Europe, the mortality of incident HeFNEF also seems to be falling. In the initial Hillingdon study, 25% of patients were dead at 6 months, but in the more recent cohort of this study (from 2004–5), the figure had dropped to 14%.³⁸ The fall was independent of confounding variables and is linked to the increased usage of renin–angiotensin system inhibitors and β-blockers.

Although mortality is higher in studies of incident HF, it is also high in prevalent cases. In the ECHOES study, the 5-year survival rate was 53% for those with HF due to impaired systolic function (Fig. 3.5).³⁹ Survival for those with HeFNEF was a little better, at 62%. In contrast, the Mayo Clinic data showed that survival in the community with HF was similar for those with systolic and nonsystolic HeFNEF.¹⁵ However, more recently the Mayo Clinic group reported on 4596 patients, of whom 47% had normal LVEF, between 1987 and 2001.⁴⁰ The survival rate was slightly better in the population with HeFNEF. However, the mortality rate declined in the population with systolic dysfunction during the study period, whereas patients with normal ventricular function had no change in mortality rates throughout the study period. More studies need to be done to clarify whether the mortality rates of these two types of HF do indeed differ.

The mortality rate for asymptomatic LV systolic dysfunction in the population is also high, with 21% of the population in the North Glasgow MONICA cohort dead at 4 years, with no significant difference in survival between those with symptoms of HF and those with asymptomatic LV dysfunction. This finding emphasizes the need for early detection and treatment of the precursor phase of HF.

Data derived from hospitalized patients in Scotland also show a trend to towards improved survival. Between 1986 and 2003, median survival after a first admission to hospital with HF improved in men from 1.3 to 2.3 years and in women from 1.3 to 1.8 years (Fig. 3.6).⁴¹ The poorer survival between those with acute HF syndromes requiring admission, compared to population-based surveys of prevalence is underscored by data from large European and US registries. In the EuroHeart Failure II survey, in-hospital mortality was 6.6%.³³ This varied with presentation but was nearly 40% in those presenting with cardiogenic shock. In a recent audit of acute hospital admissions within the United Kingdom results were poorer, with an in-hospital mortality of 15%.⁴² In-hospital mortality in the United States is better, running at 4% in the OPTIMISE HF registry.⁴³

In summary, mortality from HF remains high and the 5-year prognosis is worse than for either breast or bowel cancer.⁴⁴

Part of the enormous morbidity incurred by HF patients relates to frequent hospitalizations. Advanced HF patients who have been hospitalized experience rehospitalization rates at 6 months of 36–45%.^45,46 In the 1990s, studies in the Netherlands, Scotland, the United States and Sweden documented increasing trends of admissions relating to HF.^47,48 The rise in hospital admissions was accompanied by increasing expenditure. In Scotland, 0.2% of the population were hospitalized per year and admissions relating to HF accounted for more than 5% of all adult general medical admissions.⁴⁹ Some evidence has now emerged that HF admissions may have peaked in certain European countries during the mid 1990s. Data from Scotland on 116 556 patients identified from hospital discharge records during the period 1986–2003 showed that rates of admission rose and peaked in the mid 1990s and subsequently fell by 2003;³⁸ there are similar findings from the Netherlands (Fig. 3.7).⁵⁰

The most recent American data, however, initially seem to contradict this finding. For the period between 1979–2004, HF admissions were recorded using the National Hospital Discharge Survey. The rate of admission tripled from 1 274 000 in 1979 to 3 860 000 in 2004 (Fig. 3.8).⁵¹ However, lengths of stay and mortality have decreased in the United States according to data from the ADHERE registry.⁴²

Because of its high prevalence and hospitalization and rehospitalization rates, HF places a large economic burden on health care budgets. In the United States, the total expenditure on HF in 2007 was more than $33 billion (£21 billion, €24 billion).⁵² The statistics are mirrored in European settings. In the United Kingdom, HF consumes 1–2% of the National Health Service budget which is approximately £1.2 billion (€1.3 billion, US$1.8 million).⁵³ It is the leading cause of hospitalization in the elderly population in the United Kingdom. Approximately 60% of the total expenditure within the United Kingdom on HF is spent on hospital admissions. In Europe, the situation is similar with HF consuming approximately 1% of health care budgets. The length of stay also contributes to the expense, with median hospital stay for HF in Europe being 9 days.⁵⁴ The estimates of cost are likely to be underestimates as true costs should include all primary care consultations,

secondary care referrals, diagnostics, prescribing habits, further therapies including devices and care networks, and surgical intervention including transplantation.

Despite the advances that have been made in its treatment over the course of the last 20 years, which have seen mortality rates for those in clinical trials of HF therapies fall to less than 10% per year, epidemiological studies show that HF remains a common, lethal, disabling, and expensive condition. Its principal causes are IHD and hypertension, which often coexist. It also has a detectable asymptomatic precursor which is as prevalent as manifest HF itself. As the burden of HF increases with the ageing of our populations, we need to focus our efforts on better detection and prevention of the asymptomatic syndrome to improve its epidemiology in the future (Fig. 3.9).³⁹