Heart failure diagnosis in the general community – Who, how and when? A clinical consensus statement of the Heart Failure Association (HFA) of the European Society of Cardiology (ESC)

Patients with suggestive symptoms and signs of heart failure

As mentioned above, the diagnosis of heart failure requires the presence of symptoms. Typical symptoms in the majority of patients are caused by congestion as a result of elevated cardiac filling pressures.^{6, 10-12} Table 1 describes the typical constellation of symptoms and signs of heart failure that should raise suspicion for the diagnosis, although the sensitivity and specificity of individual features may be limited.

While most aspects of the physical examination and radiographic assessment have low accuracy for detecting clinical congestion, the presence of an elevated jugular venous pressure combines both moderately high sensitivity (70%) and specificity (79%) to identify individuals with elevated left ventricular filling pressures.^13-18 Changes in body weight have a high specificity for progressive congestion (97% likelihood that a patient with known heart failure who gains 2–4 pounds [∼1–2 kg] in 24–36 h requires hospitalization), but sensitivity is only 9%.¹⁹ Undiagnosed heart failure should also be considered a possibility in patients who are prescribed diuretic therapy for peripheral oedema in the absence of an established diagnosis of heart failure.

Whilst overt symptoms and signs of heart failure (‘clinical’ congestion) may not always be present at rest, such patients may have subclinical elevated cardiac filling pressures (‘haemodynamic’ congestion).^{20, 21} The diagnosis of heart failure (especially HFpEF) can be challenging in these circumstances and may require specialist input and the use of provocative testing (e.g. exercise stress testing) and/or invasive haemodynamic evaluation to make the diagnosis of heart failure.⁹

Patients with risk factors for heart failure

Common risk factors for the development of heart failure are coronary artery disease, including a previous MI, hypertension, atrial fibrillation (AF), valvular heart disease, diabetes, chronic obstructive pulmonary disease (COPD), chronic kidney disease (CKD), obesity, a history of high alcohol intake, cardiotoxic drugs (e.g. trastuzumab), or a family history of cardiomyopathy.^{22, 23} Therefore, a diagnosis of heart failure should be thought of in patients with a history of these diseases (or a family history of cardiomyopathy) who present with one or more of the symptoms detailed previously. Many patients with one or more of these factors will have regular interactions with their primary care team and we strongly advocate that undiagnosed heart failure should be considered and investigated for in patients reporting new symptoms of breathless, fatigue, exercise intolerance and ankle swelling. There are currently no ESC guideline recommendations to support population-level screening for heart failure, however targeted screening is recommended in patients receiving cardiotoxic cancer treatments and there is increasing interest in prospectively testing the value of targeted screening strategies in high-risk populations such as those with diabetes and CKD.²⁴

Risk scores

As discussed previously, the symptoms of heart failure lack specificity and signs can be difficult to detect clinically. Therefore, the availability of easy-to-use clinical risk scores to aid identification of patients with a high likelihood of heart failure, or even to identify patients at high risk of developing heart failure in the future, may be desirable to help guide further targeted testing. Numerous such scores have been developed from large population cohorts such as the Framingham Heart Study, the Multi-Ethnic Study of Atherosclerosis (MESA), the Atherosclerosis Risk in Communities (ARIC) study or the Health Aging and Body Composition (Health ABC) study (Table 2).^43-46 A specific score – WATCH-DM (weight [body mass index], age, hypertension, creatinine, high-density lipoprotein cholesterol, diabetes control [fasting plasma glucose], electrocardiogram [ECG] QRS duration, MI, and coronary artery bypass grafting) – has been developed for patients with type 2 diabetes and has been shown to perform well in predicting incident heart failure in external validation cohorts and shown to have improved predictive capacity when used in combination with natriuretic peptide concentrations.^{47, 48}

Scores have also been developed to aid the diagnosis of heart failure (Framingham) and specifically for the identification of patients with HFpEF – a diagnosis which is frequently challenging (Table 2).⁴³ These include the H₂FPEF score and the HFA-PEFF score, both of which require echocardiographic data.^{9, 49} It is interesting to note that most risk models to identify heart failure share a set of common variables (age, sex, New York Heart Association [NYHA] class, blood pressure, diabetes, body mass and indices of renal function). However, none of the scoring systems have found widespread use in clinical practice and current European guidelines do not recommend the use of score-based approaches, with the exception of HFpEF, in which the use of scoring systems may facilitate diagnosis.⁶ However, the guidelines acknowledge that general physicians often do not have access to all tests and parameters required by the specific diagnostic algorithms in HFpEF and are largely limited to the realms of specialist cardiology clinics. Future research should focus on the potential benefit of identifying high-risk patients by routine implementation of these risk scores (e.g. WATCH-DM in patients with diabetes).

The future of screening for heart failure

It is foreseeable that in the future, artificial intelligence and machine learning assisted algorithms for the detection of unrecognized heart failure will become increasingly available. They will aid the identification of patients at risk for heart failure in whom targeted screening may be appropriate and potentially patients with undiagnosed heart failure.^{50, 51} Numerous such algorithms have already been developed. They follow two basic strategies: one is to evaluate multiple documented parameters, e.g. in electronic patient records, to identify patients at risk for developing clinical heart failure or heart failure events.^{47, 52} Another strategy with promising results is the use of machine learning techniques to interpret single diagnostic modalities such as the conventional ECG or echocardiogram in order to identify patients with unrecognized heart failure or to aid in the interpretation of results.^53-57

So far, none of these algorithms have found their way into routine clinical practice. Reasons include the ‘black box’ nature of machine learning algorithms, the fact that some of them require the inclusion of a multitude of variables limiting their usability, as well as potential ‘overfitting’ of the models to the derivation cohorts, which can limit generalizability to external data sets.⁵⁰ As long as artificial intelligence algorithms use the same parameters as the more traditional scores listed above, no significant improvement of their predictive power can be expected. However, the possibility of interpreting standard ECG recordings by machine learning algorithms in order to identify patients with a likelihood of heart failure as the cause of non-specific symptoms, or at high risk of developing heart failure in the future, may be a potentially valuable asset in community-based screening strategies. Future research therefore needs to determine whether such algorithms retain their accuracy when applied to patients with broader demographics, and to compare their diagnostic value to current standards, e.g. to the point-of-care measurement of natriuretic peptides and routine cart-based echocardiography.

Electrocardiography

The 12-lead ECG is an important part of the screening process in all patients with suspected heart failure as it is not only a sensitive screening tool but also provides additional information regarding potential personalized treatment options in patients confirmed to have heart failure. A normal 12-lead ECG is rare in patients with a diagnosis of HFrEF and virtually excludes a diagnosis of HFrEF.^62-65 In a cohort of 534 patients, 96 had left ventricular systolic impairment and all these patients had an abnormal ECG with 90 (94%) classed as having ‘major abnormalities’ (AF, previous MI, left ventricular hypertrophy, bundle branch block and/or left axis deviation).⁶² However, in the same cohort, major ECG abnormalities were present in 39% of those without heart failure. This highlights that the finding of ECG abnormalities is a highly sensitive marker for a diagnosis of HFrEF but lacks specificity. On the other hand, the sensitivity of the ECG in HFpEF is lower than in HFrEF, and a normal ECG may be present in 35–45% of patients with HFpEF.⁶⁶

In patients with signs and symptoms of heart failure, particular ECG abnormalities may provide clues as to the aetiology of heart failure and lead to specific investigations and treatment strategies. These include the presence of changes consistent with a previous MI, AF with an uncontrolled ventricular rate, high percentage of right ventricular pacing, and left bundle branch block. Specific treatments including anticoagulation in the context of AF, and cardiac resynchronization therapy in patients with a LVEF ≤35% and a left bundle branch block are recommended in patients with these specific ECG abnormalities.⁶

Chest X-ray

A normal chest X-ray does not rule out a diagnosis of heart failure. Approximately a third of patients with a depressed LVEF have a normal cardiothoracic ratio (<0.5) on chest X-ray.⁶⁷ Conversely, the presence of an enlarged cardiac silhouette on a chest X-ray does not necessarily signify the presence of left ventricular dilatation and systolic impairment. Alternative causes of radiographic cardiomegaly include pericardial effusion, pericardial fat, left ventricular hypertrophy and valvular heart disease. Several features on a chest X-ray including pulmonary venous congestion, interstitial oedema and bilateral effusions may support a diagnosis of heart failure, however these findings along with the presence of an increased cardiothoracic ratio offer little additive diagnostic value to that of an abnormal ECG.⁶⁸ In patients who are suspected to have heart failure, in addition to detecting signs which may support a diagnosis of heart failure such as cardiomegaly, there is value in a chest X-ray as a screening tool for alternative lung pathologies which may cause breathlessness (e.g. lung cancer or pulmonary fibrosis).

Natriuretic peptides

The natriuretic peptides are a group of circulating peptides which are released in response to increased cardiac wall stress with cardioprotective vasodilatory, natriuretic, antifibrotic and sympatholytic properties.⁶⁹ The most well-known and studied of the natriuretic peptides is B-type natriuretic peptide (BNP) and its biologically inactive precursor fragment N-terminal prohormone of BNP (NT-proBNP) which is released in equimolar concentrations to bioactive BNP. Elevated concentrations of BNP and NT-proBNP reflect increased left ventricular wall stress. As well as being a powerful independent predictor of adverse outcomes in patients with established heart failure, elevated concentrations of the natriuretic peptides are a sensitive marker when investigating patients with suspected heart failure in both the acute hospital and non-acute community settings.^70-72 For this purpose of this article, we will focus on the utility of natriuretic peptide measurement in the non-acute setting.

In a study by Cowie et al.,⁷¹ atrial natriuretic peptide (ANP) and BNP were measured in 122 consecutive patients referred by their general practitioner with suspected heart failure. Mean ANP and BNP were significantly higher in those patients adjudicated by a panel of cardiologists to have a new diagnosis of heart failure based on clinical assessment, chest X-ray and transthoracic echocardiography. A BNP cut-off of ≥22.2 pmol/L (76 pg/ml) had a negative predicative value for heart failure of 98% and provided a greater degree of sensitivity (97%), specificity (84%) and positive predictive value (70%) than ANP or N-terminal ANP. The high sensitivity of natriuretic peptides for the community-based diagnosis of heart failure was further confirmed in a series of studies in a range of populations and countries.^{61, 73-76} In a randomized controlled trial in patients with suspected heart failure based on symptoms, a diagnostic strategy including NT-proBNP had a greater diagnostic accuracy than a strategy without NT-proBNP.⁷⁷ NT-proBNP may also have a role as a screening tool for undiagnosed left ventricular systolic dysfunction and/or heart failure in high-risk populations to guide preventative treatments.^{78, 79}

Based on these data, the measurement of BNP or NT-proBNP has been recommended as a key test in the assessment of patients with suspected signs and symptoms of heart failure. Current international guidelines use BNP and NT-proBNP cut-off values of ≥35 pg/ml and ≥125 pg/ml, respectively.^{6, 8} Concentrations below these thresholds have a negative predictive value of 95–99% indicating that a diagnosis of heart failure is unlikely, therefore eliminating the subsequent need for echocardiography.

There are some factors and comorbidities which influence circulating natriuretic peptide concentrations and should be considered when interpreting results. Age, AF, diabetes, and CKD are all associated with higher natriuretic peptide concentrations.^{80, 81} It has been suggested that specific cut-offs for age and AF status be incorporated into diagnostic guidelines, however such a strategy requires further testing in prospective clinical trials.⁸² Conversely, obesity, a frequent comorbidity in patients with HFpEF, is associated with lower circulating levels of natriuretic peptides.⁸⁰ It has been suggested that a substantial proportion of ambulatory patients with early/minimally symptomatic HFpEF may have normal natriuretic peptide levels despite increased filling pressures as measured by invasive exercise haemodynamic testing.^{83, 84} Indeed, the dynamic nature of elevated filling pressures (and therefore natriuretic peptides) in HFpEF warrants consideration, with many patients having normal filling pressures at rest with elevations occurring only during exercise. Subsequently, specific HFpEF diagnostic guidelines have been suggested which permit the diagnosis of HFpEF in the absence of elevated natriuretic peptide levels, however we would encourage such cases to be discussed with specialists in this area.^{9, 49}

Future use of NT-proBNP in the diagnosis of suspected heart failure may involve the use of diagnostic support tools incorporating artificial intelligence to utilize electronic healthcare records recording clinical characteristics such as age, sex, kidney function and AF status along with NT-proBNP concentrations assessed as continuous variables (rather than dichotomous cut-off values). Such an approach has been shown to be a more accurate diagnostic tool for the diagnosis of heart failure in patients presenting with acute dyspnoea rather than an individual cut-off value.⁸⁵ The utility of these instruments in patients presenting to their general practitioner in the community with symptoms of heart failure has yet to be reported.

Echocardiography

In patients with elevated natriuretic peptides suspected of having heart failure, several cardiac imaging modalities may be used to assess the presence of cardiac systolic and/or diastolic dysfunction along with indirect metrics of elevated filling pressures. Imaging also plays an important role in ruling out conditions (and directing appropriate treatments) which may also elevate natriuretic peptides and cause symptoms suggestive of heart failure including hypertrophic obstructive cardiomyopathy, valvular heart disease, and cardiac amyloidosis. The most widely used imaging modality is transthoracic echocardiography, but others include cardiac magnetic resonance imaging, radionuclide ventriculography and cardiac computed tomography.

Key echocardiographic measurements required to diagnose heart failure include LVEF, left ventricular volumes and wall thickness, left atrial volume, mitral valve inflow Doppler measurements, tricuspid regurgitation peak velocity, and inferior vena cava diameter among others.⁶ A full review of these measurements is beyond the scope of this article. Historically the most important of these measures has been LVEF. Based on the inclusion criteria of large, randomized clinical trials in heart failure, LVEF is used to phenotype patients into HFrEF (LVEF ≤40%), HFmrEF (LVEF 41–49%), and HFpEF (LVEF ≥50%). Guidelines provide individualized treatment recommendations for each of these three phenotypes, highlighting the importance of LVEF in guiding treatment decisions.^{6, 8} Recently, sodium–glucose cotransporter 2 (SGLT2) inhibitors have been shown to reduce the risk of cardiovascular death or heart failure hospitalization across the full spectrum of LVEF, therefore their use may be appropriate in eligible patients with heart failure as indicated by typical symptoms and signs along with elevated natriuretic peptides prior to estimation of LVEF.⁴

With a growing elderly population along with an increasing burden of risk factors associated with heart failure such as diabetes, obesity and AF, the incidence of heart failure is expected to rise in the coming years, therefore new strategies to facilitate access to diagnostic echocardiography and its time-consuming analysis are urgently required. Point-of-care ultrasound represents an important tool in the move of echocardiography away from large hospital-based cart machines to hand-held portable machines able to be used in the community clinic and in resource-limited settings.^86-88 The analysis and interpretation of these scans can be achieved using telehealth models or artificial intelligence assisted automated reporting tools. Such tools have been demonstrated to have similar accuracy to expert sonographer measurements and have the potential to increase access to echocardiography, reduce costs and increase throughput in heart failure screening and diagnosis.^{55, 89-91}

Initial diagnosis

Following implementation of the diagnostic pathway described in Figure 2, patients with confirmed or suspected heart failure should be referred to their local heart failure secondary care team for review and initiation/optimization of guideline-recommended therapy. Where the healthcare system does not provide for universal referral, it is therefore important to identify those patients who stand to benefit from early referral and review by specialist teams. Structured electronic referral pathways for echocardiography and clinical review in patients with elevated natriuretic peptides may help the flow of patients between primary and secondary care.

Importantly, referral to a heart failure specialist should not delay initiation of treatment to address congestion (e.g. with diuretics), potential precipitating factors (e.g. uncontrolled blood pressure, myocardial ischaemia, arrhythmias), and comorbidities (e.g. diabetes, infection, lung disease). Indeed, an improvement of symptoms in response to diuretics may help to confirm the diagnosis of heart failure; and, conversely, a lack of response should raise suspicion for differential diagnoses. Recent data supporting the benefits of SGLT2 inhibitors across the full spectrum of ejection fraction mean that initiation of this treatment may be appropriate prior to echocardiography.^{4, 92} This strategy may be beneficial for patients in two ways: firstly, the benefits of SGLT2 inhibitors have been shown in clinical trials to be evident within weeks of starting, and secondly, early initiation of SGLT2 inhibition may increase the likelihood of tolerance of other guideline-recommended heart failure treatments including mineralocorticoid receptor antagonists.^93-95 In select patients, consideration of advanced therapies including device therapy may be indicated and it is vital to facilitate early review of this group of patients and appropriate specialist input and follow-up.⁶