METHODS

Study design

The design of the study has been described in detail elsewhere.²³ Every Danish citizen and resident is signed up to one general practitioner. The number of patients assigned to each practice is fairly equally distributed. The study population, aged 50–89 years old, was recruited from randomly selected general practitioners situated in the same local area. To optimise statistical power the sampling was stratified by 10 year age groups, and we attempted to invite at least 150 patients in each age decade. The only inclusion criterion was age between 50 and 89 years. Exclusion criteria were an inability to cooperate in the examinations (for example, because of dementia), permanent residency in a nursing home, or lack of response to two written invitations. An invitation to participate in the study was sent to 1088 people between 50 and 89 years of age (3.5% of the background population in the age group) and 764 people participated, giving an overall response rate of 70.2%. The study was conducted from September 1997 to February 2000. The study was performed according to Helsinki II declarations and approved by the local ethics committee. All participants signed a written informed consent form.

Questionnaire

The questionnaire provided information on, firstly, known heart diseases and other major diseases known to be associated with cardiac complications (such as hypertension, diabetes, pulmonary diseases); secondly, hospital admissions for major cardiovascular diseases; thirdly, symptoms of HF, including the degree of shortness of breath according to a modified World Health Organization classification²⁴ and ankle swelling, angina, and cough; fourthly, medication; and lastly, alcohol and tobacco consumption. During evaluation six grades of dyspnoea were used (defined in appendix 1).

Blood pressure and heart rate

Blood pressure and heart rate were measured twice after 15 minutes of rest in the sitting position with an automatic oscillometric device with an appropriate cuff applied on the upper arm.

Laboratory analyses

Blood and urine samples were collected for routine biochemical and haematological analyses including red and white cell counts, sodium, potassium, creatinine, random blood glucose, and plasma glycosylated haemoglobin A_1c.

Electrocardiogram

A 12 lead standard ECG was obtained according to standard procedures and evaluated according to the Minnesota code.²⁵ The following abnormal findings were recorded during evaluation: presence of pathological Q waves, left bundle branch block, significant ST segment abnormalities (ST elevation or depression > 1 mm) in two or more contiguous leads, T wave abnormalities (T wave inversion ⩾ 3 mm in two or more contiguous leads or flattening of T waves in two or more contiguous leads), LVH (SV1 + RV5–6 > 35 mm or RaVL + SV3 > 20 mm), atrial fibrillation or flutter, heart rate < 50 or > 100 beats/min, and the appearance of three or more ventricular extrasystoles in 10 seconds. Normal ECG was characterised by sinus rhythm with no pathological signs.

Echocardiography

Each participant underwent a transthoracic echocardiographic examination with state of the art equipment (either HP Sonos 5500 (Hewlett-Packard, Andover, Massachusetts, USA), Acuson 128/10c (Acuson Corp, Mountain View, California, USA), or Vingmed 750 (Vingmed, Horten, Norway)) according to standard protocols. Two dimensional apical four and two chamber and apical long axis views were used for all participants, supplemented by parasternal long and short axis views. Systolic function was evaluated off line in a blinded fashion by two experienced cardiologists (FSH, PH) by means of both the ninth and 16th segment wall motion index (WMI) score. LV ejection fraction (LVEF) was calculated by multiplying the ninth segment WMI score by 30.^26,27 LVEF was expressed as the average values between observers. Normal investigations were set to a value of 2.0. LV systolic dysfunction was graded as mild (LVEF 41–45%), moderate (LVEF 36–40%), or severe (LVEF ⩽ 35%). LV end diastolic diameter was measured perpendicularly from the interventricular septum to the lateral wall through the plane of the tip of the mitral valve leaflets.²⁸ In case of visual signs of LVH, the interventricular septum and posterior wall thickness was measured with a parasternal M mode scan according to standard procedures,²⁹ with LVH defined as wall thickness exceeding 11 mm. The transvalvar flow was evaluated by standard colour Doppler technique and possible valvar dysfunction was further assessed by grading the valvar incompetence on a four grade scale as negligible, mild, moderate, or severe. All echocardiograms were recorded on videotapes. The inter-observer variation for systolic dysfunction was 2.9% (expressed as a median percentage error) and the inter-observer coefficient of variation was 4.9%. In a random sample of 198 echocardiograms, 22 (11.1%) were characterised as low quality echocardiograms. In these 22 examinations the inter-observer coefficient of variation was 6.8%.

Statistical analysis

Normal or log normal distribution of continuous data was verified with the use of histograms and normal plots. Measures of plasma creatinine showed a log normal distribution. Two sample t tests were performed for mean or geometric mean values of variables in men and women in case of normal or log normal distribution of data; otherwise non-parametric statistics were used (Mann-Whitney tests). Pearson χ² test was used to compare independent groups with categorical data. Mantel-Haenszel common odds ratio (OR) estimates were used for both sexes on each variable. Logistic regression analysis with backward stepwise elimination was used for the multivariate model to analyse independent variables in relation to systolic dysfunction and HF. Variables were arranged in a sequence based on their theoretical probability of being predictors of systolic dysfunction and HF. The analysis was repeated several times by pulling out variables to decline the equation until the number of variables was appropriate to the number of cases. All variables in the equations were tested for interaction, comparing models with and without interaction parameters. Probability values of p < 0.05 were considered significant. All tests were computed with SPSS software (SPSS 9.0 for Windows, SPSS Inc, Chicago, Illinois, USA).

Appendix 2 provides the definitions of HF used.

RESULTS

Table 1 shows basic demographics of the study population. Interestingly, symptoms (adjusted for age) of HF were reported more frequently in women than in men (p  =  0.014). Dyspnoea was reported in 239 (31%) participants, 33.6% of the women and 28.3% of the men, with about 40% (n  =  99) of those reporting dyspnoea grade 2 (appendix 1). Ankle swelling was reported in 22.6% (n  =  173) patients, more commonly in women (27.5%) than in men (16.3%, p  =  0.0001).

Echocardiograms were evaluated with WMI scores computed for 99.7% of patients. The distribution of LVEF in patients with wall motion abnormalities depicts a pattern similar to the lower half of a normal distribution curve with the 5th centile at 42% and the 2.5th centile at 35%.

Systolic dysfunction increased significantly and almost exponentially with age and was especially more pronounced in the male population (p  =  0.26 for men, p  =  0.39 for women; table 2). Moderate to severe systolic dysfunction (LVEF ⩽ 40%) was found in 7.6% of the men and 2.6% of the women (p  =  0.001) (table 2). No HF symptoms were reported by 44.0% of the men with LVEF ⩽ 40%, and only in 9.1% of the women had asymptomatic LV systolic dysfunction (table 2). The prevalence of asymptomatic participants with LVEF ⩽ 40% decreased with age. Among men, the prevalence of systolic HF was 4.2% versus 2.3% among women.

The age and sex specific prevalence proportions obtained in the study sample were applied to the entire background population aged 50–89 years, with its current age and sex structure. The estimated overall prevalence proportions of LV systolic dysfunction and systolic HF were 4.4% (95% confidence interval (CI) 3.0% to 5.9%) and 3.0% (95% CI 1.8% to 4.2%). The prevalence of mild (LVEF ⩽ 45%) and severe (LVEF ⩽ 35% systolic dysfunction was 7.0% (10.8% in men, 3.9% in women) and 3.1% (4.8% in men, 1.9% in women), respectively.

Table 3 shows the frequencies of medical conditions and cardiovascular medications within the cohort, stratified according to LVEF below or above 40%. “Definite” ischaemic heart disease (IHD) (defined in appendix 1) existed in 9.1% of the women with systolic dysfunction and in 1.2% of those with normal ventricular function. “Definite” IHD was very frequent (32%) in the group of men with systolic dysfunction and rather infrequent (1.6%) in the group of men with LVEF > 40%. Other important co-morbidities were “definite” hypertension, occurring in 45.5% versus 32.3% of women (decreased versus normal LVEF, respectively) and in 55% versus 33.6% of all men. Furthermore, definite diabetes mellitus occurred in 9.1% versus 4.8% of the women (decreased versus normal LVEF, respectively) and in 8.7% versus 8.2% of men.

Univariate analysis showed that patients with systolic dysfunction more often had a history of myocardial infarction than did normal participants (OR 14.7, p < 0.00001), a history of pulmonary congestion (OR 10.3, p < 0.00001), a history of coronary artery disease (OR 4.8, p  =  0.002), a history of HF (OR 4.3, p  =  0.006), and hypertension (OR 2.8, p  =  0.004). Dyspnoea was reported in 81.8% versus 32.4% of all women (LVEF ⩽ 40% v > 40%, respectively) and in 56.8% versus 25.8% of all men. Ankle swelling—as a symptom or as a cause for admission—was not more frequent in any of the groups except as a symptom among men with systolic dysfunction. Both men and women with systolic dysfunction had higher plasma creatinine concentrations than those without (table 3).

In patients with LVEF ⩽ 40%, angiotensin converting enzyme (ACE) inhibitors were given to only 18.2% of participating women and to 32.0% of the men, and 27.3% (women) and 26.5% (men) used loop diuretics. In those with a history of HF, 50.1% of the women were taking ACE inhibitors and 50.0% received loop diuretics, while 62.1% of the male participants with diagnosed HF received ACE inhibitors and 100% diuretics.

An abnormal ECG was found in 90.9% of women and 94.7% of men with LVEF ⩽ 40%. Significantly more participants with LVEF ⩽ 40% had moderate to severe valvar heart disease and ECG signs of LVH.

Participants with LV systolic dysfunction had a larger LV end diastolic diameter (mean 55 mm v 44 mm, p  =  0.001 for women, 55 mm v 48 mm, p < 0.0001 for men, LVEF ⩽ 40% v > 40%, respectively). No difference was found in blood pressure and heart rate in the two groups (table 3).

In 762 participants (variables were missing for two), independent factors associated with LV systolic dysfunction, which included medical history as presented to the doctor (model 1), were analysed in a logistic regression analysis (table 4). The multivariate analysis showed that a history of myocardial infarction (OR 11.4, p < 0.00001), pulmonary congestion (OR 6.9, p < 0.0005), male sex (OR 3.9, p  =  0.001), and dyspnoea (OR 3.1, p  =  0.007) were independent factors associated with LV systolic dysfunction (table 5).

A subanalysis (model 2) applied to 635 participants (129 with missing variables) included variables of cardiovascular and related diseases (as diagnosed by the doctor (appendix 2)), ECG, and blood tests (table 4). The multivariate analysis showed that abnormal ECG (OR 33.6, p < 0.00001), definite IHD (OR 10.7, p < 0.0001), plasma creatinine (OR 1.04 per unit, p  =  0.0001), and a low systolic blood pressure (OR 1.02 per mm Hg, p  =  0.002) were independent factors associated with LV systolic dysfunction (table 4). No interaction between variables in the two models was found.

Table 5 shows the prevalence of HF with preserved systolic function-symptomatic (dyspnoea) cardiac dysfunction (defined in appendix 1).

DISCUSSION

The main goal of the present study was to obtain valid estimates of the prevalence of cardiac dysfunction and manifest HF. In our sample of 764 patients aged 50–90 years recruited randomly from an urban population, 8% of the men and 3% of the women had LV systolic dysfunction defined by LVEF ⩽ 40% as estimated by echocardiography. Interestingly, 44% of the men and 9% of the women with LV dysfunction were asymptomatic.

Two types of selection bias may flaw our population sample. The first possible bias stems from the trend, although not significant, of a higher proportion of men among the elderly than in the background population, which would tend to increase the apparent prevalence of HF. The other potential bias originates from the possible differences between those volunteering to participate and those who did not respond. If the non-responders were more ill than the participants, this potential bias would tend to underestimate the true prevalence. In supplementary analyses, we found that the study population was comparable with the background population with respect to cardiac morbidity in the year preceding inclusion into the study and with respect to mortality one year after examination. However, the non-responders had a higher mortality rate than both the participants and the background population in the first year after invitation to participate.²³

Previous studies have tended to find a lower prevalence of LV systolic dysfunction when LVEF < 40% was chosen as the cut off value. In the Rotterdam study,¹⁷ the prevalence of systolic dysfunction was found to be 6% in men and 2% in women when LV fractional shortening of < 25% was used as the cut point (corresponding to LVEF < 42.5%). In the ECHOES (echocardiographic heart of England screening) study¹⁸ the prevalence was 3% in men and 1% in women when a cut off of < 40% was used for LVEF, which was evaluated primarily according to Simpson’s rule. Morgan et al¹⁹ used similar echocardiographic criteria and found a prevalence of 8% in men and 2% in women. Finally, the Glasgow MONICA (monitoring of trends and determinants in cardiovascular disease) study found a prevalence of 4% in men and 2% in women¹⁶ with LVEF ⩽ 30% as the cut off. In the first two of the cited studies, the age of the study population was comparable with our sample. The participants in the study by Morgan et al¹⁹ were older (70–84 years), and those in the Glasgow MONICA study were younger.¹⁶

The variation in prevalence of LV systolic dysfunction may be explained by three factors: the echocardiographic method, the choice of cut off values, and the selection of the study population. With regard to the echocardiographic method, we used WMI evaluated by experienced observers, which is a well established method for the determination of LV systolic function^30,31 and which has been used in some of the large HF treatment trials. To obtain a more direct comparison with previous studies, we chose to transform WMI into LVEF, although the precision of this method has been reported with mixed results.^27,32 In the present study interobserver variation as assessed on both high quality and poor quality echocardiograms was acceptable and in line with other HF epidemiological studies.^16,33 The more objective measurements obtained by Simpson’s rule or M mode measurements of fractional shortening do have several drawbacks, which make these methods difficult to apply in a population based study. They tend more often to exclude older and obese patients and patients with pulmonary disease, often reaching a discard rate of 10–15%.^16,17 Both methods are flawed by the fact that they may overestimate LV systolic function in patients with regional wall motion abnormality. Both the exclusion of a certain category of patients and the possible overestimation of LV systolic function would tend towards an underestimation of the prevalence of systolic dysfunction.

Based on the results of many of the large HF pharmacological intervention studies in which LVEF ⩽ 40% was arbitrarily chosen as the cut off point for treatment, it appeared reasonable to use the same upper limit. In addition we chose to present the prevalence of three cut off values characterised as mild (LVEF 41–45%), moderate (LVEF 36–40%), and severe (LVEF ⩽ 35%), since patients with HF symptoms and mild systolic dysfunction should be treated for HF as recommended, which makes it compelling for the clinician to know how frequent the condition is. We found that every fourth man aged 80–90 years had an ejection fraction equal or below 45%. McDonagh et al¹⁶ and Davies et al¹⁸ used LVEF ⩽ 30% and < 40% as their limits, respectively. Nevertheless, Davies and associates reported a lower prevalence than McDonagh and colleagues, despite their older study population. This difference underscores the importance of selection of the study population and the variation of echocardiographic methods used. In the study from Glasgow the prevalence of systolic dysfunction was similar to ours in comparable age groups (45–74 years): 5.4–6.4% in men and 2.4–4.8% in women.

It seems that defining severe systolic dysfunction in the present study was more similar to the approach in the ECHOES study. McDonagh et al¹⁶ defined the prevalence of mild (possible) systolic dysfunction as LVEF ⩽ 35% with an overall prevalence of 7.7%, in which 77% were asymptomatic; Davies et al¹⁸ (LVEF 40–50%) found an overall prevalence of 5.3% for LVEF < 50%, and 21.7% of the men older than 85 years had this condition. McDonagh et al¹⁶ stated that patients with LVEF 31–35% may be cardiologically healthy. In contrast it has been shown that patients with LVEF 41–45% and even 46–50% have a poorer outcome than patients with higher ejection fractions.^34,35

Finally, an important aspect in the context of population studies is the response rate and the sex distribution. In our study, we had a response rate of about 70%, which compared favourably with other studies. We found no sex related difference between the cohort investigated and the background population, and consequently no distortion in sex distribution between non-responders and the background population. Previous studies have had response rates varying from 61–78%, and not all studies have reported the age distribution among non-responders. It is noteworthy in this context that both the Glasgow MONICA study and the Rotterdam study had to exclude predominantly older participants because of unreadable echocardiograms, which would tend to underestimate the true prevalence.

Asymptomatic LV systolic dysfunction was found with a prevalence of 3.3% in men and 0.23% in women of the total population sample. This is very much in accordance with previous studies, reporting prevalence between 2.4% and 2.9% for men and between 0.2% and 0.7% for women.^16,18

These small variations in prevalence can probably be ascribed to differences in the interpretation of reported symptoms. One important possible source of error is that shortness of breath in elderly people may in some cases not be a symptom but rather reflects a natural age related decline in physical capacity. In other cases shortness of breath is a result of obesity.³⁶

The major co-morbidity in LV systolic dysfunction was IHD, which, given the high prevalence of IHD and its direct impact on myocardial function, is not surprising. Definite or possible IHD was found in about every third woman and every other man with LVEF ⩽ 40%. The same high level of co-morbidity holds true for hypertension, occurring in 45% of men and 55% of women with LV dysfunction. Moreover, 18% of the women and 28% of the men with LV dysfunction had a history of both IHD and hypertension. Valvar heart disease was found in 18% of the women and 12% of the men with LVEF ⩽ 40%, and dilated cardiomyopathy was found in 64% of women and 40% of men with systolic dysfunction (7% of the total population). Taken together, these conditions reflect the known most common causes of HF.^12,37,38 Intriguingly, diabetes was almost twice as frequent in women with systolic dysfunction as in those with normal function, in accordance with a previous report.¹²

A decrease in renal function may either result from or cause failure of cardiac function. Underlying diseases such as hypertension and diabetes cause both of these conditions. Reduced cardiac output decreases perfusion of the kidneys. On the other hand, volume overload as a consequence of renal dysfunction may induce a reduction in cardiac function.²⁰

HF with preserved LV systolic function (HF symptoms combined with cardiac dysfunction) presents a special problem and was found in 3.7% of women and 1.2% of men (4.5% and 3.4%, respectively, when atrial fibrillation was included). Our findings on symptomatic valvar heart disease and atrial fibrillation are in accordance with those of Davies et al,¹⁸ who reported a prevalence of symptomatic valvar heart disease of 0.4% and atrial fibrillation of 0.6%. Cowie et al³⁸ reported that in patients with preserved systolic function, six of 220 patients with a new diagnosis had LVH. Fox et al³⁷ found that patients with LV dilatation had extensive coronary artery disease with no evidence of prior infarction. Taken together, the reported findings indicate that hypertension and chronic IHD without prior infarction are important contributors to non-systolic HF. Importantly, in the Framingham heart study, Vasan and colleagues³⁹ showed that patients with clinical HF and preserved LV function had a slightly higher mortality than healthy controls.

In an attempt to deduce diagnostic recommendations from our data, the multivariate analysis clearly indicates that patients who report a history of HF or prior hospital admissions for pulmonary congestion or myocardial infarction, or who have IHD combined with an abnormal ECG, are at high risk of having LV systolic dysfunction. These patients, along with the ones who complain of shortness of breath, should be referred for an echocardiographic examination. These findings concur with the findings in prior studies.^16–17,19 Since LV systolic dysfunction is associated with poor prognosis and since more than half of the patients are not sufficiently treated, it is essential that patients be referred for further investigations and treatment.

Conclusions

The present study suggests that the prevalence of LV systolic dysfunction and clinical HF have been underestimated in previous epidemiological studies, which is actually suspected by many of the authors of other studies. This is primarily due to bias in the stratification of the study population and to differences in cut off values used to define the disease. The disturbingly large number of patients with asymptomatic LV systolic dysfunction, viewed by many as an early stage of HF, and of patients with undetected and untreated HF clearly advocates an increased effort to trace these patients.

Appendix 1

1. Dyspnoea: grade 0–1 is no dyspnoea or dyspnoea with considerable physical activity; grade 2 is dyspnoea caused by vacuum cleaning or climbing stairs to the second floor; grade 3 is when walking on an even road; grade 4 is with minimum exertion or when dressing; grade 5 is orthopnoea; grade 6 is at rest).

2. IHD: “definite IHD” was a (1) discharge diagnosis of myocardial infarction (MI) or (2) self reported MI and the following ECG changes: Q waves in two or more contiguous leads or T wave inversion ⩾ 3 mm in two or more contiguous leads or left bundle branch block or (3) angina pectoris (AP) discharge diagnosis/coronary artery disease (CAD) discharge diagnosis/self reported exertional chest pains and ST depressions ⩾ 1 mm in two or more contiguous leads.

3. Hypertension: “definite hypertension” was (1) self reported hypertension and blood pressure more than 140 mm Hg systolic or 90 mm Hg diastolic or antihypertensive medication or (2) blood pressure exceeding 140/90 mm Hg and current treatment or (3) blood pressure exceeding 161 mmHg systolic or 112 mm Hg diastolic.

4. Diabetes: “definite diabetes” was (1) self reported diabetes and antidiabetic treatment or (2) random blood glucose > 11.1 mmol/l or (3) haemoglobin A1c > 7.5%.

5. Chronic obstructive pulmonary disease (COPD): “definite COPD” was self reported COPD and (1) cough or (2) treatment for asthma (ATC code r03-).

Appendix 2

Definitions of heart failure used:

• “Systolic heart failure” was defined according to the European Society of Cardiology criteria: symptoms of heart failure (dyspnoea or ankle swelling) and objective evidence of LV systolic dysfunction (LVEF ⩽ 40%).

• “Heart failure with preserved systolic function” was defined by (1) dyspnoea and (a) LVH or (b) aortic/mitral dysfunction or (c) dilated left ventricle and LVEF > 50% and no definite COPD.

• “Dilated left ventricle” was defined as > 97.5th percentile of the healthy population (no history of cardiovascular disease, no diabetes mellitus, no COPD, blood pressure < 140/90 mm Hg, and normal ECG) in each sex.

• “History of heart failure” was self reported diagnosed HF or admission for HF or pulmonary congestion or HF discharge diagnosis.