Article Text


Is BNP ready for use in clinical practice?
  1. K KHAN, Clinical Research Fellow
  1. Department of Medicine & Therapeutics
  2. Clinical Sciences Building, Leicester Royal Infirmary
  3. Leicester LE2 7LX, UK
  4. kmk2{at}

    Statistics from

    Editor,—Richards et alsuggest that brain natriuretic peptide (BNP) measured 24–96 hours after acute myocardial infarction (AMI) is a powerful, independent prognostic indicator for subsequent development of left ventricular failure and death.1

    In their multivariate analysis the site (anteriorv inferior) and type (Q wavev non-Q wave) of infarction do not appear to have been included. The important prognostic value of these indicators has been established.2 3 One might expect the anterior infarcts (39% of their study population) to demonstrate greater left ventricular dysfunction, higher BNP concentrations, and a poorer prognosis than the inferior infarcts (51%). A similar relation may exist between type of infarct and outcome, although figures for each type are not given. It would be interesting to see whether BNP is still a powerful prognostic variable if site and type of AMI were included in their analysis.

    Two further confounding variables that may have weakened the association between BNP and outcome are the timings of the radionuclide ventriculography and blood sampling (1–4 days after AMI). Assessment of ventricular function in the first 24–48 hours after AMI can lead to an overestimation of damage due to the phenomenon of myocardial stunning.4 The time course of BNP shows a peak at 16 hours followed by a significant decline in the next 48–72 hours.5 A narrower and standardised time window for ventriculography and venesection may have improved the correlations.

    It would have been useful to know the area under their receiver operating characteristic (ROC) curve for BNP, which is highly relevant in assessing the true value of a test.6 They demonstrated a negative predictive value of 100% for BNP at a threshold of 20 pmol/l, but they used different thresholds for looking at the sensitivity and specificity and predictive power for left ventricular ejection fraction and left ventricular failure (25 and 33 pmol/l, respectively). Although these values were derived from their ROC analyses, it is difficult to envisage how a BNP result should then be interpreted in clinical practice.

    If the additional prognostic value of BNP is confirmed once site and type of infarct are incorporated into their analysis, then its precise role still needs to be clarified. Measurement of BNP after AMI is unlikely to reduce the need for imaging of ventricular function because of its poor positive predictive value. Its potential use may lie in its ability to identify a high risk population in whom some sort of intervention is feasible before development of “clinical end points”. However, at present, there is no evidence to suggest that treating a high plasma BNP in the presence of a normal ejection fraction improves outcome. Clearly this should be an area for further investigation.

    BNP and its prohormone derivative (N-terminal proBNP) offer exciting prospects for non-invasive assessment of myocardial function and outcome following infarction. Richards et alhave clearly demonstrated the prognostic superiority of BNP over other neurohormonal markers thus supporting their initial hypothesis. Their concluding statement, however, suggesting that plasma BNP “could reasonably be included in the routine clinical workup of a patient following myocardial infarction” seems premature.1


    This letter was shown to the authors, who reply as follows:

    Dr Khan’s letter raises a number of interesting points. First, when the site of myocardial infarction (anterior or inferior) is included in multivariate analysis of data reported in our recently published paper, plasma BNP remains an independent prognostic indicator for both death and development of heart failure. This is not surprising as division of patients according to inferior or anterior site of infarction creates crude categories, which both include a broad spectrum of injury from mild to very severe. In contrast, plasma BNP is a continuous variable and is related to the severity of cardiac injury across the entire group. This finding is also predictable in view of the fact that already published reports have indicated BNP has prognostic power independent of left ventricular ejection fraction (LVEF).1-1 1-2 LVEF acts as a common indicator of degree of severity of injury regardless of site of infarction.

    Dr Khan also comments on a potential weakening of the association between BNP and outcome (although the context suggests he means LVEF rather than morbid or mortal outcomes) due to the timing of radionuclide ventriculography and blood sampling. We find a similar correlation between early postinfarction BNP(1–4 days) and both early (1–4 days) and late (3–5 months) radionuclide ventriculography.1-1 In addition, repeat BNP measurements at four months (unpublished data) continued to show a similarly strong correlation, albeit with an offset regression due to mean BNP falling somewhat from early postinfarction concentrations. In other words, a similar correlation is observed between BNP and ejection fraction regardless of early or late measurements of either variable. This suggests that the time window involved is not overly important provided reference data are established to allow for the systematic fall in plasma BNP over the months after infarction. However, multivariate analyses show that BNP and left ventricular ejection fraction areindependent predictors of death or later heart failure following myocardial infarction, and their independent nature implies their correlation will not be overly strong. The association of BNP with prognosis may well reflect the influence on plasma BNP concentrations of diastolic dysfunction and left ventricular mass as well as systolic function.1-3 1-4 For these reasons efforts to alter measurement method or timing simply to improve the correlation of BNP with LVEF are likely to be unproductive.

    The statement by Dr Khan “The time course of BNP shows a peak at 16 hours followed by a significant decline in the next 48–72 hours” does not take into account the conflicting nature of the literature. The pattern of BNP change is dependent on the severity of infarction, the exact nature of the BNP assay employed, and its degree of cross reactivity with pro-BNP 1–108 or its N-terminal deleted metabolites.1-5-1-8 In our hands the time profile shows a plateau between 24 and 72 hours and hence our election of a 1–4 day sampling window.

    Dr Khan’s comment that it is “difficult to envisage how a BNP result should then be interpreted in clinical practice” is disingenuous. Our published paper clearly points out that plasma BNP of less than twofold the upper limit normal within 1–4 days postinfarction has 100% negative predictive value for an ejection fraction of < 40% four months after MI. Our report also makes it very clear that the positive predictive value for BNP above this level is very weak (that is, BNP above the normal range within the early postinfarct period is a weak predictor of reduced LVEF), and we do notrecommend the routine use of plasma BNP as a substitute for measurement of left ventricular ejection fraction. However, theprognostic value of BNP is strong and, at the very least, early postinfarct BNP measurements will allow better risk stratification and therefore better targeting of surveillance in the follow up period.

    The most important and insightful part of Dr Khan’s letter addresses the possible potential for BNP together with ejection fraction to identify high risk populations in whom some sort of intervention is feasible before development of “clinical end points”. He states correctly that at present “there is no evidence to suggest that treating a high plasma BNP in the presence of a normal ejection fraction improves outcome”. In a manuscript now in preparation, the Christchurch Group is able to report follow up data on over 500 MI patients with a mean follow up period of approximately two years. This group has been divided according to both plasma BNP and radionuclide LVEF. Notably, over 20% of the group have an early postinfarct LVEF in excess of 40% but a concomitant BNP of over 25 pmol/l (2.5 times the upper limit of normal). This subgroup has a significantly greater risk of mortality and of developing heart failure than the group with LVEF above 40% and plasma BNP < 25 pmol/l. Furthermore, in patients who have ejection fractions below the 40% threshold but BNP < 25 pmol/l, there is very little increase in risk of either death or heart failure over two years compared with that group with low BNP concentrations and high ejection fraction. In other words, reduction in ejection fraction only predicts increased morbidity or mortality in the presence of neurohormonal activation as indicated by raised plasma BNP. Our findings concur with data from colleagues in Sweden and Norway (T Omland, C Hall, personal communication), and it is becoming clear that a randomised controlled trial of treatment in asymptomatic patients with LVEF > 40% but clear neurohumoral activation should be done.


    1. 1-1.
    2. 1-2.
    3. 1-3.
    4. 1-4.
    5. 1-5.
    6. 1-6.
    7. 1-7.
    8. 1-8.
    View Abstract

    Request permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.