Clinical syndrome suggestive of heart failure is frequently attributable to non-cardiac disorders — population-based study☆
Abstract
Aim:
To assess how often the clinical syndrome (CS) of heart failure is attributable to alternative, including non-cardiac, explanations.
Methods and results:
Cross-sectional evaluation of 739 community participants aged ≥45 years. Subjects with ≥2 symptoms or signs (dyspnoea or fatigue, orthopnoea, nocturnal paroxysmal dyspnoea, third heart sound, jugular venous distension, rales and lower limb oedema) or who were receiving loop diuretics were considered to have the clinical syndrome of heart failure. Attributable fractions were derived based on adjusted odds ratios and the prevalence of underlying disorders among cases. CS was present in 28.0% of women and in 15.2% of men, p<0.001. The multivariate-adjusted fraction of CS attributable to female gender was 40.6%, to age≥65 years 28.5%, left ventricular systolic dysfunction, left ventricular dilatation or moderate–severe valvular disease 4.9%, diastolic dysfunction or atrial fibrillation 13.0%, obesity 22.6%, coronary heart disease 7.2% and chronic lung disease 6.9%. When additionally adjusting for depressive symptoms, the association with gender and age became much weaker, and 32% of cases were attributable to depressive symptoms. Forty-two percent of subjects with CS had cardiac abnormalities.
Conclusion:
In less than half of subjects with CS was systolic or diastolic heart failure confirmed. Female gender, older age, obesity and depressive symptoms accounted for the largest fraction of CS.
1. Introduction
In routine clinical practice, in a large proportion of patients with a clinical syndrome suggestive of heart failure, no cardiac dysfunction, either systolic or diastolic, is found and therefore the clinical syndrome is attributable to other conditions. Differential diagnosis is particularly difficult in elderly patients presenting chronic obstructive airways disease or obesity, common conditions late in life 1. Additionally, it is often difficult to distinguish pathologic conditions from mere physical deconditioning associated with ageing.
An objective demonstration of cardiac structural or functional abnormalities is currently required for the diagnosis of HF 2,3. The identification of a large proportion of HF cases without left ventricular systolic dysfunction (LVSD) 4, up to 50% in some studies, led to the proposal of diastolic HF or HF with preserved left ventricular systolic function as a separate clinical entity 5. Until recently, diastolic HF was essentially a diagnosis of exclusion 678, and alternative causes of dyspnoea or pulmonary congestion, such as pulmonary diseases and obesity, as well as conditions of circulatory overload such as renal failure had to be excluded 9,10, in addition to the demonstration of preserved systolic function. Currently, evidence of abnormal left ventricular relaxation, filling, diastolic distensibility or stiffness, beyond the typical clinical syndrome and absence of systolic dysfunction, is additionally recommended for diagnosis 5. Unfortunately, accurate diagnostic criteria for diastolic dysfunction based on non-invasive methods are not yet available 11.
The cause of complaints suggestive of heart failure varies depending on the prevalence and overlap of the underlying disorders. In some patients, the clinical manifestations may be the result of more than one contributing factor. Studies assessing factors causally involved in the clinical syndrome are rare, as are studies quantifying the proportion of the syndrome attributable to conditions or factors other than heart failure. We aimed to assess how often a clinical syndrome suggestive of heart failure is in fact attributable to alternative, including non-cardiac, explanations in a representative sample of urban middle-aged and elderly Portuguese adults.
2. Participants and methods
2.1. Study design
As part of a health and nutrition survey of the adult population of Porto, Portugal, we systematically assessed parameters of cardiac structure and function in middle-aged and elderly participants to estimate the prevalence of heart failure stages 12.
Using random digit dialling we selected households and with simple random sampling we selected one eligible subject within each household. Eligible subjects were all permanent residents aged ≥18 years. Refusals were not substituted. Participants were invited to visit our Department for an interview, which included a questionnaire on social, demographic, behavioural and clinical data. All participants had an anthropometric evaluation, fasting blood sample collected, spirometry and a resting 12-lead ECG. For the specific objective of this study, we assessed all participants aged ≥45 years who were recruited from January 2001 to December 2003. A few days after the first interview a second evaluation took place comprising a structured clinical interview by a physician, a cardiovascular physical examination and a transthoracic M-mode, 2D echocardiogram and pulsed Doppler evaluation of transmitral inflow. The proportion of participation was 70% 13. We assessed 739 community-dwellers aged ≥45 years: 443 were women with mean age 61.3 (10.3) years, and 296 were men aged 63.0 (11.0) years.
The investigation conforms with the principles outlined in the Declaration of Helsinki. The local ethics committee approved the study and all participants provided written informed consent.
2.2. Definition of clinical syndrome
Dyspnoea or fatigue, orthopnoea, nocturnal paroxysmal dyspnoea, third heart sound, jugular venous distension, rales and lower limb oedema at the end of the day were considered suggestive of heart failure when at least two of them were present. Dyspnoea and/or fatigue counted as one symptom only. Lower limb oedema was only considered in the absence of signs of chronic venous insufficiency. Since loop diuretics may camouflage these clinical manifestations, we considered in the clinically symptomatic group those who fulfilled the above defined criteria or who were receiving loop diuretics. The reason for loop diuretic prescription was not considered.
2.3. Definition of underlying disorders
Echocardiograms were done by four cardiologists, using the same equipment (HP Sonos 5500) and recorded on videotape for later review by a single experienced cardiologist (C.A.L.), blinded to clinical data. Measurements of wall thicknesses and chamber dimensions were obtained according to the Penn convention and left ventricular mass was calculated as 1.04 [(interventricular septum thickness+LV end-diastolic diameter+posterior wall thickness)3−(LV end-diastolic diameter)3]−13.6 14. Left ventricular mass, wall thicknesses and chamber diameters were indexed to height. Left ventricular (LV) systolic function was assessed according to the subjective impression of the operating cardiologist (visual assessment), this was later validated by one single experienced cardiologist. Left ventricular ejection fraction was determined using Simpson's rule. Left ventricular systolic dysfunction was defined as either ejection fraction <45% or by visual assessment, this being the only parameter when ejection fraction could not be calculated (44 participants). Left ventricular dilatation was considered when end-diastolic left ventricular diameter was larger than 58 mm in men and 52 mm in women 15. Valvular abnormalities were considered only if moderate or severe. In participants in sinus rhythm, diastolic dysfunction was defined using the peak E wave/peak A wave ratio (E/A), E wave deceleration time (DT) and isovolumetric relaxation time (IRT) according to the recommendations of the European Society of Cardiology 5: if age<50 years, IRT>100 ms and/or E/A<1 and DT>220 ms; if age≥50 years, IRT>105 ms and/or E/A<0.5 and DT>280 ms.
Coronary heart disease was considered as self-reported medical diagnosis of angina pectoris or myocardial infarction (“Has a doctor ever told you …?”) or pathologic Q waves on at least two adjacent leads on resting 12-lead ECG. Obesity was defined as body mass index≥30 kg/m2. Chronic lung disease was defined as history of chronic bronchitis (chronic productive cough during at least 2 months each year, for at least 3 years) or a diagnosis of moderate-severe obstructive (FEV1<70% of predicted) or restrictive (vital capacity<70% of predicted) syndrome on spirometry. The creatinine clearance was estimated according to the Cockroft-Gault formula, using serum creatinine, age and body weight 16. A cutpoint of 50 ml/min/1.73 m2 was used to define moderate renal dysfunction. Participants' educational attainment and occupation were used as indicators of socioeconomic status. Educational attainment was measured as the highest level of schooling completed and was categorised as either <4 years or ≥4 years. Occupation was classified according to the British Registrar General social classification of occupations 17. Women with no occupation (housewives) were grouped in class V for the current analysis. Depressive symptoms were assessed using the Beck Depression Inventory 18. Participants were considered to have depressive symptoms if they scored higher than 15 19 or were receiving pharmacological antidepressant treatment.
2.4. Statistical analysis
Data are described as mean and standard deviation (SD) for quantitative variables, and counts and proportions for categorical variables. The prevalence of underlying disorders was compared between participants with and without the clinical syndrome using the chi-square or Fisher's exact tests as appropriate. Multivariate logistic regression was used to assess the independent association between the clinical syndrome and gender, age and underlying disorders. Two models were fitted, one including as independent variables gender, age and the underlying disorders that we considered as a plausible explanation for the clinical syndrome, and the other considering additionally depressive symptoms. Data on depressive symptoms were available for 423 subjects; therefore the second model refers only to this subsample. Odds ratios (OR) from these models and the prevalence (p) of each underlying disorder as well as gender and age among those with the clinical syndrome were used to estimate population attributable fractions (AF) using the formula AF=1−Σ(pi/ORi) for each stratum i of the exposure variable 20. The population attributable fraction expresses the proportion of all cases of the clinical syndrome of heart failure that can be attributed to a particular underlying disorder or characteristic 21.
3. Results
The clinical syndrome was present in 28.0% (n=124) of women and in 15.2% (n=45) of men, p<0.001. Participants with the clinical syndrome were significantly older (Table 1) and more commonly women regardless of age.
| Women | Men | |||||
|---|---|---|---|---|---|---|
| No clinical syndrome | Clinical syndrome | p | No clinical syndrome | Clinical syndrome | p | |
| n=319 | n=124 | n=251 | n=45 | |||
| Age, years | 60.1±9.9 | 64.3±10.6 | <0.001 | 61.8±10.6 | 69.8±10.4 | <0.001 |
| Education, years | 7.2±4.7 | 5.3±4.1 | <0.001 | 8.3±4.8 | 7.1±4.3 | 0.12 |
| Occupation (Registrar General) | 0.005 | 0.25 | ||||
| I-II | 77 (24.1) | 17 (13.7) | 109 (43.4) | 14 (31.1) | ||
| III non-manual | 84 (26.3) | 23 (18.5) | 54 (21.5) | 15 (33.3) | ||
| III manual-IV | 57 (17.9) | 32 (25.8) | 73 (29.1) | 12 (26.7) | ||
| V or no occupation | 101 (31.7) | 52 (41.9) | 15 (6.0) | 4 (8.9) | ||
| LVSD, LV dilatation or valvular disease | 21 (6.6) | 19 (15.3) | 0.007 | 19 (7.6) | 8 (17.8) | 0.04 |
| Diastolic dysfunction or atrial fibrillation | 70 (22.4) | 38 (31.9) | 0.05 | 63 (25.9) | 18 (43.9) | 0.03 |
| Ischaemic heart disease, n (%) | 19 (6.0) | 19 (15.3) | 0.003 | 24 (9.6) | 9 (20.0) | 0.07 |
| Angina, n (%) | 16 (5.0) | 16 (12.9) | 0.007 | 13 (5.2) | 6 (13.3) | 0.05 |
| Previous myocardial infarction, n (%) | 5 (1.6) | 5 (4.0) | 0.15 | 15 (6.0) | 6 (13.3) | 0.11 |
| Chronic lung disease | 37 (11.6) | 20 (16.1) | 0.26 | 40 (15.9) | 13 (28.9) | 0.06 |
| Smoking | 0.04 | 0.08 | ||||
| No | 247 (80.7) | 104 (89.7) | 67 (27.0) | 10 (22.7) | ||
| Current | 34 (11.1) | 9 (7.8) | 65 (26.2) | 6 (13.6) | ||
| Former | 25 (8.2) | 3 (2.6) | 116 (46.8) | 28 (63.6) | ||
| Obesity, n (%) | 83 (26.2) | 60 (48.8) | <0.001 | 37 (14.8) | 10 (22.2) | 0.30 |
| Creatinine clearance<50 ml/min/1.73 m2 | 26 (8.6) | 15 (13.0) | 0.24 | 20 (8.7) | 10 (23.8) | 0.01 |
| Depressive symptoms* | 70 (37.8) | 38 (60.3) | 0.003 | 19 (12.9) | 10 (35.7) | 0.01 |
- a Data are presented as mean±standard deviation for continuous variables and n (%) for categorical variables.
- * Data available for 423 subjects (248 women, 175 men).
LVSD, LV dilatation and/or valvular disease were found in 15.3% (n=19) of women with and 6.6% (n=21) without the clinical syndrome, p=0.007. In males the proportions were respectively 17.8% (n=8) and 7.6% (n=19), p=0.04. Table 1 presents the prevalence of the alternative or complementary underlying disorders that we considered might explain the clinical syndrome. The clinical syndrome was associated with diastolic dysfunction or atrial fibrillation, coronary heart disease, particularly angina, and obesity in both genders. Renal dysfunction was significantly associated with the clinical syndrome in men. Of note, the minimum creatinine clearance in this sample was 30 ml/min/1.73 m2, meaning that no one had severe renal dysfunction. The association between chronic lung disease and the clinical syndrome was borderline significant only in men. Lower educational attainment and less skilled occupations were associated with a higher prevalence of the clinical syndrome only in women. In both genders the prevalence of depressive symptoms was significantly higher among subjects with the clinical syndrome, and it was significantly higher among women regardless of the clinical syndrome (37.8% versus 12.9% of men, p<0.001, in the absence of the clinical syndrome; 60.3% versus 35.7% of men, p=0.004, in the presence of the clinical syndrome).
The odds ratios for quantification of the association of age, gender, education and each of the underlying disorders and the clinical syndrome are shown in Table 2. When adjusting for age, gender, education and underlying disorders simultaneously, the fraction of the clinical syndrome attributable to female gender was 40.6%, that for age≥65 years was 28.5%, for left ventricular systolic dysfunction, left ventricular dilatation or moderate-severe valvular disease was 4.9%, for diastolic dysfunction or atrial fibrillation 13.0%, for obesity 22.6%, for coronary heart disease 7.2% and for chronic lung disease 6.9%. When including a score of depressive symptoms as an additional independent variable, the association with female gender and with age became much weaker. Thirty-two percent of cases of the clinical syndrome were attributable to depressive symptoms (Table 2, Model 2).
| Model 1 | Model 2* | |||||
|---|---|---|---|---|---|---|
| OR (95% CI) | Prevalence among those with clinical syndrome (%) | AF (%) | OR (95% CI) | Prevalence among those with clinical syndrome (%) | AF (%) | |
| Female gender | 2.24 (1.46–3.43) | 73.4 | 40.6 | 1.67 (0.92–3.03) | 69.2 | 27.7 |
| Age | ||||||
| <65 years | 1 | 45.0 | 28.5 | 1 | 50.5 | 22.9 |
| 65–74 years | 1.80 (1.16–2.80) | 34.3 | 1.72 (0.94–3.16) | 34.1 | ||
| ≥75 years | 2.77 (1.59–4.82) | 20.7 | 2.26 (1.01–5.09) | 15.4 | ||
| LVSD, LV dilatation or valvular disease | 1.44 (0.79–2.62) | 16.0 | 4.9 | 1.75 (0.68–4.53) | 14.3 | 6.1 |
| Diastolic dysfunction or atrial fibrillation | 1.59 (1.05–2.40) | 35.0 | 13.0 | 1.34 (0.75–2.38) | 31.4 | 5.9 |
| Coronary heart disease | 1.76 (0.98–3.18) | 16.6 | 7.2 | 1.45 (0.57–3.69) | 14.3 | 4.4 |
| Chronic lung disease | 1.54 (0.93–2.57) | 19.5 | 6.9 | 2.63 (1.30–5.31) | 24.2 | 15.0 |
| Obesity | 2.18 (1.44–3.28) | 41.7 | 22.6 | 2.53 (1.46–4.38) | 44.0 | 26.6 |
| Education ≤4 years | 1.50 (1.00–2.23) | 65.1 | 21.6 | 1.38 (0.81–2.35) | 56.0 | 15.5 |
| Depressive symptoms | - | - | - | 2.58 (1.49–4.47) | 52.7 | 32.3 |
- a OR, odds ratio; CI, confidence interval; AF, population attributable fraction; LVSD, left ventricular systolic dysfunction; LV, left ventricular.
- * n=423 (91 (21.5%) with the clinical syndrome.
The underlying disorders considered as alternative or complementary explanations for the clinical syndrome are not mutually exclusive. Among subjects with the clinical syndrome, 72 (42.6%) had a structural cardiac abnormality, and 15 (8.9%) had a history of coronary heart disease without systolic or diastolic dysfunction. In 37 (21.9%) clinically symptomatic participants no underlying disorder could be identified — 27 (73.0%) were women, 5 (13.5%) had creatinine clearance <50 ml/min/1.73 m2, and 22 had information on depressive symptoms (of whom 13 (59.1%) had depressive symptoms).
4. Discussion
In this study we found that a clinical syndrome suggestive of HF was common in a sample of middle-aged and elderly adults. In less than half of these cases was a cardiac abnormality objectively demonstrated, most often in men. The clinical syndrome was more prevalent among women and increased with age. Age, female gender and obesity each accounted for a larger proportion of the burden of this clinical syndrome than systolic or diastolic dysfunction or atrial fibrillation.
Female gender and increasing age were the strongest determinants of the clinical syndrome. This observation is concordant with previous data reporting a higher prevalence of HF symptoms in women 3,22, frequently false positive cases. The association with age also illustrates a common problem in routine clinical practice, distinguishing between pathological situations and expected physical deconditioning with age. When adjusting additionally for depression, using the subsample for whom data on depressive symptoms were available, the association of the clinical syndrome with female gender and the association with age became weaker implying that part of the gender difference and of the age difference in symptoms and signs is explained by psychological factors or that the difference is only in the importance that subjects give to their complaints. Psychological factors are probably among the strongest confounders in the association between the clinical syndrome and underlying organic disorders. Psychological characterisation focused only on depressive symptoms, which is insufficient, and was not complete since we only obtained data on depressive symptoms from 423 participants due to non-response. The Beck Depression Inventory was not completed during the interview but was self-administered at home in order to shorten the interview time, which is a burden for community participants. Despite instructions many participants did not return responses. We did not compute the scores if the questionnaires were incomplete, thus further reducing the final number of participants with data available on this scale. Subjects with Beck Depression Inventory data were mainly men, younger, more educated, with more skilled occupations and either married or living as married. Among these variables, only education, occupation and marital status were independently associated with the probability of responding. Importantly, use of antidepressant medication was similar in both groups. There is probably residual confounding by emotional and other psychological factors which affects both odds ratios and estimates of population attributable fractions.
Although the clinical syndrome as defined in this study, is not a clinical diagnosis of HF, the proportion of subjects with systolic or diastolic dysfunction was unexpectedly low. Diastolic dysfunction was probably underdiagnosed. In the first place, we used only non-invasive methods to obtain data on diastolic function and we did not perform tissue Doppler because at the time the study was done the technology was not available at our institution. Furthermore, we did not assess pulmonary vein flow, which could help identify diastolic dysfunction in participants with pseudonormal patterns of transmitral inflow. However, the inter-observer agreement in performing such studies is low and good quality images are difficult to obtain in a large proportion of people, particularly if they are obese. Some have argued against the requirement of evidence of diastolic dysfunction for the diagnosis of diastolic HF, based on a study that showed that most patients meeting the Framingham criteria for the clinical diagnosis of HF, with ejection fraction >50% and echocardiographic evidence of LV hypertrophy, had diastolic dysfunction evidenced by invasive studies (catheterisation) 23. However, in this group of patients the pre-test odds of diastolic dysfunction was very high, given the use of stringent specific clinical diagnosis criteria (Framingham) and the requirement of LV hypertrophy, thus resulting in a high positive predictive value of diagnosis without assessment of diastolic function.
The prevalence of coronary heart disease was also probably underestimated, as is usual in population-based epidemiologic studies, because stress tests or angiography were not performed. The fact that angina was more strongly associated with the clinical syndrome can be explained by the fact that it is a chronic condition, while myocardial infarction is an acute event. If systolic or diastolic dysfunction occur as a consequence of an infarction, then the clinical syndrome of HF might develop, whereas angina or equivalents can easily be confused with dyspnoea or fatigue by patients and physicians.
Chronic lung disease was more prevalent among men, as a consequence of smoking which was much more frequent in men in this age group and probably also due to occupational exposure. Respiratory disorders are important in the differential diagnosis of dyspnoea, but not as important for the clinical syndrome as defined in this study, which included at least one other symptom or sign other than dyspnoea or fatigue. Most cases of chronic lung disease in this sample had only mild lung function disturbances and this could reduce the probability of an association with the clinical syndrome. If many severe cases were included, an association with the clinical syndrome might be found that could be mediated by cor pulmonale, but right chamber dilatation or systolic dysfunction were not considered in the current study. Mild lung function abnormalities may be a consequence of pulmonary congestion and could be the main explanation for a higher prevalence of chronic lung disease in those with the clinical syndrome. In men, having quitted smoking was directly associated with the clinical syndrome while current smoking was not. Reverse causation is a plausible explanation for this finding, given the cross-sectional design of this study. The fact that subjects being treated with a diuretic increased the pool of ex-smokers argues in favour of this hypothesis, that is, people quitted smoking as a consequence of symptoms, which were not necessarily due to smoking.
In conclusion, in less than half of all subjects with a clinical syndrome suggestive of heart failure was systolic or diastolic heart failure confirmed. Female gender, age≥65 years and obesity had the largest impact on the prevalence of the clinical syndrome. The gender and age effects were in part mediated by depression. The quantitative results of this study on the impact of characteristics of subjects and underlying disorders on the clinical syndrome cannot be extrapolated to the clinical context. However, our results may help to identify patient groups where false positive clinical diagnoses of heart failure tend to be more common.
5 Acknowledgment
The authors gratefully acknowledge a grant from Fundação para a Ciência e a Tecnologia (POCTI/SAU-ESP/61492/2004).
