Methods

The BELFRAIL study is a prospective, observational, population-based cohort study of subjects aged ≥80 years in 3 well-circumscribed areas of Belgium. All participants in the study gave informed consent, and the Biomedical Ethics Committee of the Medical School of Université Catholique de Louvain in Brussels approved the study. The study design, methods, recruitment of participants, and characteristics of the cohort were previously described in detail. Briefly, 29 general practitioner (GP) centers were asked to include patients aged ≥80 years. No other inclusion criteria were specified, and only 3 exclusion criteria were used: known severe dementia (Mini-Mental State Examination score <15), receiving palliative care, and the occurrence of medical emergencies such as acute-onset or decompensated heart failure.

Each GP was asked to record known noncardiovascular and cardiovascular co-morbidities of the study subjects and perform a structured and standardized history and clinical examination. GPs were blinded to the results of the other study visits. Noncardiovascular co-morbidities were defined as the presence of thyroid problems, asthma, chronic obstructive pulmonary disease, osteoarthritis, documented osteoporosis, and malignancies. Cardiovascular morbidities were defined as the presence of hypertension, diabetes mellitus, hyperlipidemia, history of angina pectoris or myocardial infarction, known cardiomyopathy, history of transient ischemic attack or cerebrovascular accident, peripheral arterial disease, history of decompensated heart failure, chronic atrial fibrillation (AF) or known valvular heart disease. GPs reported important cardiovascular interventions, such as percutaneous transluminal coronary angioplasty or stenting; coronary, valvular, or arterial surgery; and the placement of a pacemaker. The Anatomic Therapeutic Chemistry classification system was applied to register medication use. Data on relevant cardiovascular medication, including cardiac glycosides, diuretics, β blockers, calcium antagonists, angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, and lipid-lowering agents, were used. The history included questions about symptoms of angina pectoris, dyspnea (Medical Research Council scale score ≥3), invalidating fatigue, orthopnea, cough (nocturnal or diurnal), wheezing, edema of the lower extremities, palpitations, loss of appetite, smoking status (former or current), alcohol intake (units per week), recent changes in symptoms, recent weight change, and alteration of mental and psychological status. The clinical examination consisted of rest heart rate and breathing rate measurements. Blood pressure was measured in the sitting position in both arms, and measurements were repeated after 2 minutes. The highest systolic and diastolic blood pressure (left or right) after 2 minutes was used in the analyses. Heart auscultation was performed to detect third or fourth heart sounds or cardiac murmurs. The apex beat was palpated and recorded when abnormal. Lung auscultation was performed to detect crepitation, wheezing, and rhonchi. The carotid arteries were auscultated to detect murmurs. Jugular venous pressure was measured and noted if increased, and the presence of hepatojugular reflux was also checked. GPs determined whether hepatomegaly or edema of the lower extremities was present and noted whether patients had ascites.

Body mass index was calculated on the basis of standardized measurements of weight and height. Twelve-lead electrocardiograms were recorded using a QRS Universal ECG device (QRS Diagnostic, Maple Grove, Minnesota). Each electrocardiogram was digitally stored and analyzed off-line by a single cardiologist (blinded to the results of all other examinations) according to the Minnesota Code Classification System. Patients with previous myocardial infarctions, AF, artificial pacemaker, left ventricular (LV) hypertrophy, left bundle branch block, right bundle branch block, sinus tachycardia, or sinus bradycardia were included. Moreover, cardiologists were asked to report whether ECG findings were completely normal.

A blood sample was collected in the morning after fasting, and plasma (ethylenediaminetetraacetic acid) and serum samples were stored at −80°C. Hemoglobin concentrations were measured on whole blood using the Sysmex XE-2100 automated hematology analyzer (Sysmex, Milton Keynes, United Kingdom). Anemia was defined as hemoglobin <12 g/dl for women and <13 g/dl for men. Total cholesterol, low-density lipoprotein and high-density lipoprotein cholesterol, triglycerides, ultrasensitive C-reactive protein, and cystatin C were measured using the UniCel DxC 800 Synchron (Beckman-Coulter, Brea, California). Glomerular filtration rate was estimated using the Chronic Kidney Disease Epidemiology Collaboration 2 cystatin C formula. Plasma levels of B-type natriuretic peptide (BNP) were measured using the Biosite kit (Biosite, San Diego, California) on a UniCel DxI 800 Immunoassay System (Beckman-Coulter), and serum N-terminal pro-BNP (NT-proBNP) levels were measured using the Dade-Dimension Xpand (Siemens Healthcare Diagnostics, Deerfield, Illinois). Coefficients of variation ranged from 5.4% to 6.7% and from 3.9 to 4.3%, respectively.

Echocardiography was performed at the home of each subject by a single cardiologist (blinded to all other test results) using a commercially available portable system (CX50; Philips Medical Systems, Andover, Massachusetts). A complete examination, in accordance with the recommendations of the American Society of Echocardiography and the European Association of Echocardiography, was performed. The method, prevalence of echocardiographic abnormalities, quality of echocardiographic images, and classification of subjects according the presence of cardiac dysfunction were previously described in detail. Briefly, LV function was calculated using Simpson’s biplane method. Systolic dysfunction was defined as an LV ejection fraction ≤50%. The function of the mitral and aortic valves was evaluated using color Doppler echocardiography after optimizing the gain and Nyquist limit. Stenotic and regurgitant valve diseases were evaluated according to semiquantitative and quantitative methods recommended by the American Society of Echocardiography. Subjects with prosthetic valves were evaluated separately. Clinically relevant valvular heart disease was defined as any mitral stenosis severity, severe aortic stenosis, moderate or severe mitral regurgitation, and moderate or severe aortic regurgitation. Diastolic function was assessed using mitral flow velocities obtained using pulsed Doppler and pulsed tissue Doppler at the level of the mitral annulus. Additional apical and parasternal views for the assessment of tissue velocity (color tissue Doppler) were recorded. The American Society of Echocardiography and the European Association of Echocardiography guidelines were used to assess the presence of diastolic dysfunction. Because of difficulties in interpreting diastolic parameters in subjects with AF or pacemaker, the population under study was split into 2 groups for analysis. In subjects without chronic AF or pacemaker (on electrocardiographic or by GP assessment), the target condition was clinically relevant cardiac dysfunction, including an LV ejection fraction ≤50%, valvular heart disease, or severe diastolic dysfunction. In subjects with chronic AF or pacemaker, the target condition was an LV ejection fraction ≤50% or valvular heart disease. A composite target condition was chosen because the main goal was to identify those patients with increased risk who could be referred for further testing.

Continuous variables are expressed as mean ± SD or as median (interquartile range). Categorical variables are expressed as numbers and frequencies. Comparisons between different categories of subjects were performed using Student’s t test or the Mann-Whitney U test (for nonparametric data) for unpaired data. The added value of natriuretic peptide levels and ECG findings was measured using logistic regression analysis and classification and regression tree (CART) analysis, as complementary tools. Two strategies were used: one used all clinical variables collected by the GP and blood analyses without natriuretic peptide levels (the clinical model), and the other included natriuretic peptide levels and/or ECG findings (the clinical+ model). For the logistic regression analysis, BNP, NT-proBNP, and ultrasensitive C-reactive protein were log-transformed, anemia was used instead of hemoglobin levels, cystatin C was used instead of estimated glomerular filtration rate, and body mass index was dichotomized (>25 or ≤25 kg/m ² ). Low-density lipoprotein was not used, because of a strong correlation with levels of total cholesterol (Pearson’s correlation = 0.94, p <0.001). First, a univariate logistic regression analysis was performed with all clinical variables. Subsequently, all variables with p values ≤0.10 were included in the multivariate analyses (stepwise, forward conditional) to determine which items were independently related to the presence of cardiac dysfunction. Afterward, natriuretic peptide levels and ECG variables were added to the clinical model. The C-statistic and 95% confidence interval (CI) was calculated using the probabilities calculated from the regression analysis and compared between different models. Tree-based models (such as CART) are nonlinear and nonparametric alternatives to linear models for regression and classification problems. There are multiple advantages to using CART analysis: the method deals with multicollinearity in an intuitively correct way, interactions are automatically taken into account in an objective way, weights can be attributed to false-negatives (FNs) and false-positives (FPs), missing data are dealt with in a natural way, and indicators of diagnostic accuracy can easily be calculated from the output. The minimum-cost tree was selected as the best tree. The cost represents the proportion of misclassified cases. A stepwise approach in tree generation was used. First, a tree with equal costs for FNs and FPs was calculated. Second, the cost of FNs was set to 3 times that of FPs to generate a tree that excluded cardiac dysfunction. Third, a tree to identify the different subtypes of cardiac dysfunction (equal costs for FNs and FPs) was calculated in the remaining population. For each tree, the area under the curve was determined, and the sensitivity, specificity, and post-test probability of negative and positive test results were calculated using 2 × 2 tables. Data analysis was performed using SPSS version 19.0 for Windows (SPSS, Inc., Chicago, Illinois) and Salford Predictive Model Builder version 6.6 (Salford Systems, San Diego, California).

Results

From November 2, 2008, to September 15, 2009, 567 subjects were included in the BELFRAIL study. Echocardiography was performed in 548 participants, of whom 91 (17%) had AF or pacemaker. The target condition was present in 78 subjects (17%) without and 28 (31%) with AF or pacemaker. Table 1 lists the background variables, and Table 2 lists the results of the index tests according to the presence of the target condition. The median time from echocardiography to examination by the GP was 29 days (interquartile range 14 to 43), from echocardiography to electrocardiography was 7 days (interquartile range 3 to 15), and from echocardiography to natriuretic peptide testing was 3 days (interquartile range 1 to 13). In total, 509 participants (90%) underwent all index tests and echocardiography ( Figure 1 ). Some subjects failed to undergo electrocardiography, natriuretic peptide testing, or echocardiography as a result of technical issues or had inconclusive results.

Study flow diagram.

Logistic regression analysis showed a C-statistic of 0.79 (95% CI 0.74 to 0.85) for the clinical model in subjects without AF or pacemaker ( Table 3 ). Of the additional tests, LV hypertrophy on electrocardiography, negative ECG findings, NT-proBNP, and BNP were univariate determinants of cardiac dysfunction. The combination of NT-proBNP or BNP with negative ECG findings had the largest added value, resulting in an increase in the C-statistic to 0.82 (95% CI 0.76 to 0.88, p = 0.50). In subjects with AF or pacemaker, the clinical model had a high C-statistic (0.90, 95% CI 0.82 to 0.98). No additional test was an independent correlate of cardiac dysfunction.

Table 3

Determinants of the presence of cardiac dysfunction in subjects with or without chronic atrial fibrillation or pacemaker from background variables, history, clinical examination, electrocardiographic findings, and natriuretic peptide levels with logistic regression analysis

Variable	No AF or Pacemaker			AF or Pacemaker
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	C-Statistic (95% CI)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	C-Statistic (95% CI)
Clinical model			0.79 (0.74–0.85)			0.90 (0.82–0.98)
Age	NS	—		1.2 (1.1–1.4)	1.3 (1.0–1.6)
Thyroid dysfunction	2.3 (1.1–4.8)	2.3 (0.92–5.7)		NS	—
Cardiomyopathy	2.2 (1.0–4.8)	NS		NS	—
Peripheral arterial disease	NS	—		4.9 (1.4–17)	5.7 (1.0–32)
Episode of decompensated heart failure	3.9 (1.9–7.9)	NS		NS	—
Reported valvular heart disease	3.6 (2.1–6.1)	2.7 (1.4–5.1)		2.3 (0.95–5.8)	NS
Coronary surgery	3.1 (1.3–7.3)	2.8 (1.0–7.5)		NS	—
Cardiac glycosides	NS	—		2.8 (0.98–8.2)	NS
Diuretics	1.5 (0.95–2.5)	NS		NS	—
Angiotensin-converting enzyme inhibitors or angiotensin II receptor blockers	1.7 (1.0–2.8)	NS		NS	—
Invalidating fatigue	1.8 (0.99–3.1)	NS		NS	—
Orthopnea	2.2 (0.92–5.2)	NS		NS	—
Loss of appetite	NS	—		3.2 (0.95–11)	NS
Alteration of mental and psychological status	1.7 (0.99–2.8)	1.8 (0.97–3.4)		2.7 (1.1–6.8)	4.2 (1.0–17)
S3	5.3 (1.8–15)	4.5 (1.2–17)		NS	—
Cardiac murmur	3.3 (2.0–5.9)	2.3 (1.2–4.2)		3.7 (1.4–9.4)	4.5 (1.1–18)
Abnormal apex beat	2.6 (0.94–7.1)	NS		NS	—
Rhonchi	2.4 (0.87–6.5)	NS		NS	—
Carotid murmur	2.3 (1.2–4.5)	NS		8.7 (1.6–47)	NS
Increased jugular venous pressure	2.4 (1.1–5.3)	NS		NS	—
Hepatojugular reflux	4.2 (2.1–8.5)	NS		2.8 (0.91–8.3)	5.4 (1.2–25)
Peripheral edema on clinical examination	NS	—		2.2 (0.89–5.5)	NS
Body mass index >25 kg/m 2	0.58 (0.35–0.96)	NS		NS	—
Anemia	2.6 (1.5–4.5)	1.8 (0.90–3.4)		NS	—
Total cholesterol	NS	NS		0.97 (0.96–0.99)	0.96 (0.94–0.98)
High-density lipoprotein	NS	—		0.94 (0.90–0.98)	NS
Total cholesterol/high-density lipoprotein	0.77 (0.59–0.99)	NS		NS	—
Triglycerides	0.99 (0.99–1.0)	0.99 (0.98–1.0)		NS	—
Log ultrasensitive C-reactive protein	NS	—		2.2 (0.97–5.2)	NS
Cystatin C	1.7 (1.1–2.6)	1.7 (1.0–2.8)		2.2 (0.87–5.6)	NS
Clinical+ model
Electrocardiography
LV hypertrophy	2.6 (1.2–5.6)	1.5 (0.60–4.0) ∗	0.79 (0.74–0.85)	4.1 (1.1–16)	3.5 (0.24–52) ∗	0.91 (0.82–1.0)
AF	—	—		3.9 (1.4–11)	3.5 (0.71–18) ∗	0.91 (0.83–0.99)
Negative findings	0.45 (0.26–0.79)	0.46 (0.25–0.87)	0.80 (0.74–0.85)	—	—
Log BNP	16 (7.5–35)	8.3 (3.3–21)	0.81 (0.75–0.87)	4.2 (0.83–21)	0.74 (0.089–6.2) ∗	0.90 (0.82–0.98)
Log NT-proBNP	7.1 (4.1–12)	6.5 (3.1–13)	0.81 (0.75–0.87)	5.3 (1.7–16)	1.0 (0.23–4.7) ∗	0.90 (0.82–0.98)
Clinical model + log BNP + negative ECG findings			0.82 (0.76–0.87)			—
Clinical model + log NT-proBNP + negative ECG findings			0.82 (0.76–0.88)			—

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Association of Serum Uric Acid and Cardiovascular Disease in Healthy Adults
2. Effect of Morbid Obesity on In-Hospital Mortality and Coronary Revascularization Outcomes After Acute Myocardial Infarction in the United States
3. Association Between Bilirubin and Mode of Death in Severe Systolic Heart Failure
4. Safety and Effect of Very Low Levels of Low-Density Lipoprotein Cholesterol on Cardiovascular Events

Stay updated, free articles. Join our Telegram channel

Join