Natriuretic peptide measurements for the diagnosis of “nonsystolic” heart failure

Good news and bad

The clinical dilemma

When first learning that their patient with presumed HF has a normal or near normal ejection fraction, clinicians should ask themselves several questions. First, is the ejection fraction correct? All measurement techniques, whether quantitative or qualitative, are fallible, especially when the data are suboptimal. The distinction between 50% and 39% is not always obvious, but it is useful because of the demonstrated benefits that medications such as angiotensin-converting enzyme (ACE) inhibitors and beta-blockers have been in systolic dysfunction.

Second, was the diagnosis of HF correct? In the past, clinicians often, but in many cases incorrectly, excluded the diagnosis when informed that left ventricular function was normal. But there are many potentially confounding causes for the symptoms and signs of HF. Some of these are listed in Table 1. The answer to this question is important in determining whether additional diagnostic testing is required, and it also has therapeutic implications.

Third, was HF precipitated by a transient condition that may have impaired systolic function that has now resolved, such as myocardial ischemia, an arrhythmia, or episodic hypertension? Is further evaluation required?

A simple diagnostic test is needed

In this setting, great need exists for a simple, readily available, reliable, and cost-effective diagnostic test. Ideally, the test could answer three questions: 1) In a patient with symptoms and signs attributable to HF, whether systolic or diastolic in origin, will the test be abnormal? In other words, does it have high sensitivity? 2) Conversely, does an abnormal test in a symptomatic patient exclude other causes for the clinical presentation? For example, does the test have high specificity for the diagnosis? 3) Can the same test distinguish systolic HF from diastolic HF? A test with these characteristics would not only facilitate appropriate management but might also obviate an extensive and expensive evaluation that often includes repeating previously performed tests and invasive procedures.

The test most frequently employed to address these questions is echocardiography with Doppler-flow measurements (echo-Doppler) (9). How well does it fill the need? Although echo-Doppler is noninvasive and widely available, it often is not available “on demand,” and it is relatively expensive. If a technically adequate study can be obtained, echo-Doppler has high sensitivity. With regard to sensitivity, a completely normal study, showing normal chamber size, function, and structure, with normal mitral flow velocities excludes HF with a high degree of certainty. However, echo-Doppler is less specific for the diagnosis of HF. Patients with no evidence of HF may have abnormalities of cardiac function or structure (reduced systolic function, chamber enlargement or hypertrophy, valvular abnormalities, and abnormal diastolic flow patterns), and patients with these findings may have symptoms and signs consistent with HF owing to a noncardiac etiology. And with regard to the third need, echo-Doppler is the test clinicians depend upon to distinguish systolic HF from HFPSF. In this regard, it is the ejection fraction measurement, not measurements of diastolic function, that drives this classification. Many patients with HFPSF do not have clear-cut evidence of diastolic dysfunction.

It should also be noted that the value of echo-Doppler is not limited to its ability to diagnose or exclude HF. It provides additional invaluable information on the etiology of HF. Furthermore, many of the findings that may result in a false-positive diagnosis of HF are clearly pathological and may be clinically important, leading to effective interventions (such as ACE inhibitors or beta-blockers for asymptomatic systolic dysfunction, more rigorous treatment of hypertension, or even valve repair or replacement).

The performance of BNP in the diagnosis of HFPSF

The accompanying study by Maisel et al. (1)provides the first information from the Breathing Not Properly trial about patients with HFPSF. Using an ejection fraction of >45% to define this group (their mean EF was 59%), patients with HFPSF constituted 165 (36.5%) of the 452 subjects with an adjudicated diagnosis of HF who had echocardiograms within 30 days of the emergency room visit. The median BNP of the HFPSF patients was 413 pg/ml, compared to 34 pg/ml in the patients without HF (p < 0.001) and 821 pg/ml in patients with systolic HF (also p < 0.001). Thus, BNP appears to be a helpful test in patients with HFPSF.

However, a closer look at the data indicates important limitations exist to the use of BNP to diagnose HFPSF. Considerable overlap occurs in BNP levels in HFPSF patients and those without HF. The sensitivity of BNP at a threshold of 100 pg/ml remained acceptable at 85%; nonetheless, 15% of HFPSF patients were missed—a figure substantially higher than in systolic HF. The larger problem, however, is specificity, and this is driven by the higher BNP levels exhibited in older patients, and particularly in older women, whose median BNP without HF was above the 100 pg/ml cut point. Indeed, no significant difference was seen in BNP levels between women adjudicated to have HFPSF and those who did not have HF. This problem is further compounded by the fact that older women constitute the largest proportion of patients with HFPSF (16,17). Finally, BNP levels did not separate patients with systolic HF from those with HFPSF, although the former group tended to have higher levels.

The poorer performance of BNP in diagnosing HFPSF is, in fact, not surprising. With regard to its inability to distinguish systolic from nonsystolic HF, this should be expected because the symptoms in both conditions relate to elevated ventricular filling pressures, the mechanism responsible for rising BNP levels. The trend to higher measurements in systolic HF probably relate to the generally greater chronicity and severity of systolic HF. If this distinction is to be made, a measure of ejection fraction is required.

The higher BNP levels in older women and, to a lesser extent, older men probably reflect underlying diastolic dysfunction in these groups and may well represent the equivalent of asymptomatic left ventricular systolic dysfunction; the levels probably identify patients at risk for developing diastolic HF (16,18–20). Indeed, the same phenomenon is responsible for the relatively poor specificity of Doppler-derived indices of diastolic filling for the diagnosis of HFPSF. The use of age and gender-specific “normal” values would improve the specificity of the test, but with a loss of sensitivity, which is already lower for HFPSF and systolic HF. Rather, clinicians will continue to have to use a Bayesian approach to the diagnosis of HFPSF, carefully considering the likelihood that the patient has HF and using elevated natriuretic peptide levels as corroborative evidence.

Study implications

Natriuretic peptides are becoming an important adjunct to the diagnosis of HF. They are highly accurate for the diagnosis of systolic HF, but unfortunately perform less well for HFPSF. It is important to recognize that natriuretic peptide levels increase with age, especially in women, probably as a result of the changes in cardiac structure and diastolic function with aging in industrialized societies. Thus, a higher rate of false-positive results is to be expected. An elevated BNP measurement in these demographic groups should be interpreted in light of the clinical presentation, and an echocardiogram may be necessary to clarify the diagnosis. Unfortunately, natriuretic peptides cannot differentiate systolic from diastolic HF, thus, a measurement of ejection fraction is still required to guide the management, if not necessarily the diagnosis, of this syndrome.