RCT of B-type natriuretic peptide (BNP)

Randomised trial
Results
Comment

Bandolier 121 examined the evidence for the use of B-type natriuretic peptide in the diagnosis of heart failure. On the evidence then available, BNP was probably going to be useful. The challenge was to work out how. The obvious thing to do was to perform a randomised trial of using the test in primary care, with the outcome of improved diagnosis. That trial has been done [1], and confirms the utility of the test.

Randomised trial

The trial, conducted in New Zealand, enrolled patients aged 40 years or more presenting to family doctors with symptoms of dyspnoea and/or oedema of recent origin. They could have comorbid conditions and be receiving any treatment. Those who needed urgent admission to hospital were excluded.

At the initial visit the family doctor recorded whether or not heart failure was suspected on history and examination. The patients were then booked to attend a study visit (Figure 1) in which they were assessed clinically by a cardiologist, with chest X-ray and ECG were conducted, and blood taken for BNP tests. A panel of three cardiologists and one family doctor used rigorous application of standard pre-defined criteria to make a gold-standard diagnosis of whether heart failure was present or not. This panel did not have access to BNP results, and was independent of study procedures.

Patients were then properly randomised to one of two groups (Figure 1).

Figure 1: Design of RCT to evaluate BNP test for diagnosis of heart failure in primary care

In one the family doctor had the results of the BNP faxed, with a standardised interpretive comment (as below).

Standard interpretive advice with BNP results

Normal range in healthy subjects BNP <50 pmol/L
BNP >150 pmol/L strongly suggests heart failure in a newly symptomatic patient, but all clinical information should be taken into account.
In between these levels heart failure is still possible, but all clinical information should be taken into account. BNP may be elevated by renal impairment, atrial fibrillation, LVH, COPD, after myocardial infarction, in the elderly and by treatment with beta-blockers or digoxin. BNP may be decreased by hypothyroidism, treatment with diuretics, vasodilators, and ACE inhibitors.

The other group had a laboratory sheet faxed informing the family doctor that the results of the BNP test were not available. The unit of randomisation was the patient.

Their family doctor, with or without the BNP result, then reviewed patients and made a final diagnosis. All family doctors had previously received a standardised 30-minute education session on interpretation of BNP results.

The outcome was the accuracy of the diagnosis made by the family doctor at the second visit.

Results

There were about 150 patients in each group, with an average age of about 70 years, and about two thirds were women. Comorbid conditions of hypertension, history of myocardial infarction, diabetes, asthma, and COPD were common. The two groups were well matched at baseline.

Panel diagnosis of heart failure was made in 77 patients, and 228 were judged not to have heart failure. Plasma BNP levels in those with heart failure averaged 290 pmol/L, compared with 61 pmol/L in those without heart failure.

At the initial visit the diagnostic accuracy of the family doctor compared with the panel judgement was about 50% in both groups (Figure 2). At the second visit, family doctors without BNP results improved their diagnostic accuracy by 8%, with an overall accuracy rate of 60%. With BNP results they improved their diagnostic accuracy by 21%, to 70%.

Figure 2: Overall percentage of correct diagnosis of heart failure by family doctor at initial and final visit, with and without BNP result

Almost all of the improvement in both groups came from correctly ruling out heart failure (Table 1). With or without BNP results, family doctors were good at diagnosing heart failure when it was actually present. They had more difficulty in ruling out heart failure when it was not present. The difference in diagnostic accuracy with BNP meant that about 11% more patients with symptoms of heart failure would be diagnosed correctly. The number needed to diagnose was about nine.

Table 1: Diagnosis of heart failure or no heart failure by panel of cardiologists and family doctor versus family doctor, in patients with and without BNP result available

	With BNP			Without BNP
Diagnosis	Panel	Family doctor	Percent correct	Panel	Family doctor	Percent correct
Heart failure	43	39	91	34	31	91
Not heart failure	109	67	61	119	60	50

Comment

Here we have the missing link concerning the utility of BNP in primary care. On the basis of this well-conducted study, it looks worthwhile. One other interesting thought is that the large differences at presentation between those patients with and without heart failure for parameters other than BNP might be used to create even more accurate diagnostic algorithms, particularly rule-out algorithms.

This is important because the majority of patients referred for heart failure don't have it. In this study an initial diagnosis of heart failure was made in 215 patients: only 77 were correct. Reducing unnecessary referrals should bring big benefits where hospital capacity is limited, or waiting times long.

The study was supported by two New Zealand bodies, government and charity. The results will benefit patients, professionals and systems. Manufacturers of BNP testing kits will also benefit. Why is it, then, that diagnostic manufacturers do not undertake useful research to prove that their products really do make a difference?

Reference:

SP Wright et al. Plasma amino-terminal pro-brain natriuretic peptide and accuracy of heart-failure diagnosis in primary care. A randomized, controlled trial. Journal of the American College of Cardiology 2003 42: 1793-1800.