The exact role of natriuretic peptides in general practice remains to be defined.1 While raised plasma concentrations of N-terminal atrial natriuretic peptide (N-ANP) or brain natriuretic peptide (BNP) are associated with left ventricular systolic dysfunction and heart failure,2-4 the prognostic value of these peptides outside hospital populations remains uninvestigated. We recently showed that the risk of left ventricular systolic dysfunction in heart patients from general practice can be assessed from abnormal N-ANP concentrations, abnormal ECG, and a heart rate > diastolic blood pressure.3 The importance of these prespecified variables, in particular a raised N-ANP, would be confirmed if they also predicted mortality. The present follow up study from general practice examines this question.

All subjects ⩾ 40 years of age from one general practice and all subjects ⩾ 50 years of age from two general practices in Copenhagen were screened from 1994 to 1996 by a procedure based on a questionnaire and review of general practice records. Of 2158 subjects, 357 had past or present signs or symptoms of heart disease; of these, 105 were excluded (and not invited for examination),126 withdrew, and 126 subjects were examined. Reasons for exclusion by priority were: 38 living in nursing homes, 36 questionnaire non-responders, 21 patients without definite heart disease for administrative reasons in an early study phase, and 10 patients with advanced heart failure. Half the remaining subjects withdrew: 32 declined the invitation, 1 died, 37 were disabled because of various medical and psychosocial conditions, and 56 patients, after various degrees of contact, did not show up.3 ,5 Those withdrawing were older than examined subjects (47% were ⩾ 80 years of age) but they did not have more recognised ischaemic heart disease, hypertension or diabetes as inferred from general practice case notes. As no self perceived deterioration of health was involved non-participation is understandable.

We thus examined 126 patients who were willing to, and fit for, attending an outpatient cardiological examination. Their mean age was 70 years (range 49–93 years), 55 were men, 30 had a previous myocardial infarction, 11 had atrial fibrillation, 69 had hypertension, 12 had diabetes, 31 had chronic obstructive pulmonary disease, and 43 had received treatment for suspected heart failure. On examination 15 had an echocardiographic left ventricular ejection fraction ⩽ 0.45, and 22 were normal according to a full cardiac assessment. The median plasma concentration of N-ANP in these 22 patients was 0.41 nmol/l (range 0.17–0.76 nmol/l; reagents OY NT-pro-ANP ¹²⁵I RIA kit, Biotop OY, Finland). A significantly raised plasma concentration of N-ANP was subsequently defined as > 0.8 nmol/l. Seventeen subjects had raised N-ANP, and 60 had an ECG showing QRS or ST changes, or both. The Danish Central Populations Register provided information on all deaths from the day of examination up to 2 May 2000. We used total death as the end point while waiting for a search on causes of death, the registration of which lags two years behind.

The approval of the hospital ethics committee was obtained, as was informed consent from those patients examined.

We observed 34 deaths during a median follow up period of 4.3 years (range 3.9–6.1 years). The three prespecified binary markers were examined in the multivariate Cox regression model along with sex and age (< or ⩾ 70 years). Significant predictors of mortality were raised N-ANP (hazard ratio 5.6, 95% confidence interval (CI) 2.7 to 11.3), abnormal ECG (hazard ratio 2.2, 95% CI 1.1 to 4.4), and age ⩾ 70 years (hazard ratio 2.9, 95% CI 1.3 to 6.4), while sex and “heart rate > diastolic blood pressure” were not significant predictors (z score < 1.96). A critical question is whether the relation between N-ANP and mortality is distorted by treatment with angiotensin converting enzyme inhibitor or loop diuretic, or both (n = 26). When treatment was entered into the final model the predictive power of all four variables remained strong and significant. We also tested the independent prognostic value of N-ANP in addition to recognised prognostic factors: N-ANP (nmol/l), age, ejection fraction, New York Heart Association (NYHA) (Ia, Ib, IIa, IIb, IIIb), and heart rate minus diastolic blood pressure (bpm − mm Hg) treated as continuous variables, and sex and abnormal ECG as dichotomised variables. The significant variables in this model were N-ANP (z = 4.1), NYHA (z = 2.5), age (z = 2.4), and male sex (z = 2.4).

Table 1 outlines survival in various groups formed by combining the three significant predictors from the prespecified Cox model. The effect of N-ANP and ECG was dependent on age. The table is therefore split at the mean age of 70 years. Data at 3 years and 6.1 years illustrate trends in survival (all subjects were followed for > 3 years). The younger reference group, with two normal test results, had a significantly better survival than the older reference group (96%v 71%, p = 0.0086). Raised N-ANP was seen in 10% of those with a normal ECG (6/[26 + 1 + 28 + 5]), and in 18% of those with an abnormal ECG (11/60). Compared with the age matched reference groups mortality was not significantly higher with a single abnormality of either a raised N-ANP or an abnormal ECG. Having two abnormal tests, however, was a strong predictor of death in both age groups.

The main finding of this study, along with its predecessor,2 is that when N-ANP is measured in patients with an abnormal ECG it is a powerful predictor of early deaths and a moderate predictor of left ventricular systolic dysfunction in heart patients from primary care.

This study is the first to provide prognostic data about natriuretic peptide in heart patients from general practice. It supports the idea of using natriuretic peptide measurement in selected patients rather than advocating it for an unrestricted use. The strength of this study lies in its validation of prespecified variables, and that N-ANP remained the strongest predictor of death after considering recognised prognostic factors. A larger study is required to show the prognostic value of N-ANP in subjects with a normal ECG. BNP measurements were not available for this study. Nor have we evaluated these variables in acute patients, which is a different situation.4

Just half the target group accepted our invitation for a full cardiac assessment. Those who withdrew were old, they were probably more disabled, and their prognosis may have been worse than for subjects who were examined. In view of the understandable withdrawal pattern, we believe our results are applicable to a scenario in general practice where a physician wants to make a risk assessment for a presumed heart patient without overt congestive heart failure. A normal ECG will convince the physician and patient of a favourable prognosis and a low likelihood of left ventricular systolic dysfunction. Our study suggests that measurement of natriuretic peptide is particularly useful in subjects with an abnormal ECG. If both N-ANP and the ECG are abnormal, further cardiac assessment seems sensible to determine whether treatable conditions exist, although we have no data on whether cardiac assessment will indeed change the outcome.