Diagnosis of heart failure using urinary natriuretic peptides

Transcription

1 Clinical Science (2004) 106, (Printed in Great Britain) 129 Diagnosis of heart failure using urinary natriuretic peptides L. L. NG, S. GEERANAVAR, S. C. JENNINGS, I. LOKE and R. J. O BRIEN University of Leicester, Department of Cardiovascular Sciences, Clinical Sciences Building, Leicester Royal Infirmary, Leicester LE2 7LX, U.K. A B S T R A C T In the present study, we assessed the use of urinary natriuretic peptides [N-terminal proatrial natriuretic peptide (N-ANP) and N-terminal pro-brain natriuretic peptide (N-BNP) and C-type natriuretic peptide (CNP)] in the diagnosis of heart failure. Thirty-four consecutive hospitalized heart failure patients (median age, 75.5 years; 14 female) were compared with 82 ageand gender-matched echocardiographically normal controls. All subjects provided plasma and urine specimens. Plasma was assayed for N-BNP, and urine was assayed for N-ANP, N-BNP and CNP. The diagnostic efficiency of peptides was assessed using receiver operating characteristic (ROC) curve analysis. All three urinary natriuretic peptides were significantly elevated in heart failure patients (P < 0.001). Urine N-BNP was correlated with plasma N-BNP (r s = 0.53, P < ). Areas under the ROC curves for urinary N-ANP, N-BNP and CNP were 0.86, 0.93 and 0.70 and for plasma N-BNP was Correcting urinary peptide levels using urine creatinine produced ROC areas of 0.89, 0.93 and 0.76 respectively. A urine N-BNP level cut-off point of 11.6 fmol/ml had a sensitivity and specificity for heart failure detection of 97 % and 78 % respectively, with positive and negative predictive values of 64.7 and 98 %. In conclusion, although all three natriuretic peptides were elevated in urine in heart failure, urinary N-BNP had diagnostic accuracy comparable with plasma N-BNP. Use of urinary N-BNP for heart failure diagnosis may be suitable for high-throughput screening, especially in subjects reluctant to provide blood samples. INTRODUCTION Heart failure represents a major and increasing burden in industrialized nations. Although echocardiography is the criterion standard for establishing a diagnosis, waiting lists and times are long. Biochemical indicators of heart failure have been suggested as useful aids in the diagnosis of heart failure, and natriuretic peptides, in particular, have been useful in establishing the diagnosis of heart failure. Brain natriuretic peptide (BNP) and its N-terminal precursor [N-terminal pro-bnp (N-BNP)] have been shown to be elevated in left ventricular systolic dysfunction (LVSD) [1 4], with high negative predictive values for the diagnosis of LVSD. In addition, N-terminal pro-atrial natriuretic peptide (N-ANP) has also been shown to be elevated in LVSD [1,5], and recent evidence suggests that myocardium is also a source for C-type natriuretic peptide (CNP) [6]. However, plasma N-BNP may perform better than N-ANP in this regard [1]. In screening programmes, some patients may prefer to provide urine rather than plasma specimens. This would also facilitate high-throughput collection of specimens. There is limited evidence that natriuretic peptides are excreted in urine. In studies on patients with kidney diseases, N-ANP is detectable in urine [7]. CNP has also been described in urine of small numbers of normal and heart failure patients [8]. N-BNP has never been described in human urine and the use of any of the urinary Key words: diagnosis, heart failure, natriuretic peptide, urinary peptide. Abbreviations: BNP, brain natriuretic peptide; CNP, C-type natriuretic peptide; LVSD, left ventricular systolic dysfunction; N-ANP, N-terminal pro-atrial natriuretic peptide; N-BNP, N-terminal pro-bnp; ROC, receiver operating characteristic. Correspondence: Professor L. L. Ng ( lln1@le.ac.uk).

2 130 L. L. Ng and others natriuretic peptides as a diagnostic aid for LVSD has never been tested. We have developed immunoassays for detecting urinary natriuretic peptides (N-ANP, N-BNP and CNP) and have compared their performance with a recognized biochemical test for heart failure, namely plasma N-BNP, in the diagnosis of LVSD. METHODS Study population We studied 34 consecutive patients admitted to the Leicester Royal Infirmary with LVSD. Patients had radiographic evidence of pulmonary oedema and/or severe impairment of systolic function on echocardiography, as judged visually (equivalent to ejection fractions of under 35 %). All patients were in New York Heart Association (NYHA) class IV on admission to hospital, and NYHA class III on discharge. To reflect normal clinical practice, no particular exclusion criteria were made with regard to age, gender, the perceived need for inotropic support or prior medical history. None of the recruited patients had a second life-threatening condition that might have been expected to affect prognosis. Patients with acute myocardial infarction were excluded from the study. Age- and gender-matched normal subjects (n = 82) were also recruited as part of a screening study for heart failure. All had normal ECGs, normal echocardiography (ejection fractions > 50 % and no valvular disease), had no past medical history of ischaemic heart disease, hypertension or diabetes and were not on any medication. All patients gave informed consent to participate in the study, which was approved by the local Ethics Committee. This study conforms to the principles of the Declaration of Helsinki. Urine and blood sampling All subjects were venesected after 15 min bed-rest, and 20 ml of peripheral venous blood was drawn into pre-chilled sodium/edta (1.5 mg/ml of blood) tubes containing 500 international units/ml aprotinin. After centrifugation at 1500 g at 4 C for 15 min, plasma was separated and stored at 70 C until assayed. Subjects also provided a urine sample, and 20 ml was collected into tubes with the above additives, centrifuged and stored as above. Urine samples from patients were collected whilst patients were on therapy, after the diuresis from the morning dose of loop diuretics had abated. These samples were in general about 3 4 h after breakfast, when the patients had been stabilized on therapy and were ready to be discharged from hospital. Assay of N-ANP, CNP and creatinine These competitive immunoluminometric assays were devised using antibodies to the N-terminal section of pro-anp and its respective ligand (Peninsular Laboratories, San Carlos, CA, U.S.A.) and CNP (Bachem Ltd, St Helens, Merseyside, U.K.). Pro-ANP peptide (1 30 nmol) or CNP (10 nmol) were labelled with N-hydroxysuccinimidobiotin (200 nmol; Calbiochem, Nottingham, U.K.) in 0.1 mol/l sodium phosphate buffer, and the tracers were purified on reversed-phase HPLC using an acetonitrile gradient. Urine specimens were extracted on StrataX solid-phase extraction columns (Phenomenex Ltd, Macclesfield, Cheshire, U.K.) using protocols supplied by the manufacturer. A sample (4 ml) of urine was loaded on to the columns and, following washes with 0.1 mol/l NaH 2 PO 4 (ph 6.0), the peptides were eluted with 85 % (v/v) methanol. Extracts were dried on a centrifugal evaporator, reconstituted into immunoluminometric assay (ILMA) buffer before analysis (composition as described previously [9]). Individual wells of ELISA plates were coated with 100 ng of anti-(rabbit IgG) sera (Sigma, Poole, Dorset, U.K.). Wells were then filled with 50 µl of the extracts or N-ANP or CNP standards (from fmol/well) and 50 µl of ILMA buffer containing 50 ng of the anti- (N-ANP) or -CNP antibodies was added. Following incubation at 4 C for 24 h, 0.5 nmol of the biotinylated N-ANP or CNP tracer was added and incubated for a further 24 h at 4 C. The plates were washed three times in 0.1 % Tween-20 in PBS and streptavidin labelled with methylacridinium ester ( relative light units/ml; synthesized as described previously [9]) was added for 2 h. Plates were read on a Dynatech MLX Luminometer, with sequential injections of 100 µl of 100 mmol/l HNO 3 containing 0.05 % H 2 O 2, followed by 100 µl of 250 mmol/l NaOH containing 0.25 % cetyl ammonium bromide [9]. The anti-(n-anp) and -CNP antibodies were specific for their respective peptides and had no cross-reactivity with human ANP, BNP or N-BNP. The lower limit of detection was 0.97 and 0.07 fmol/ml for N-ANP and CNP respectively. Intra- and inter-assay coefficients of variation were under 10 % for both assays. Urine creatinines were determined using the Jaffe reaction, measuring the change in absorbance at 505 nm following the addition of acetic acid. Assay of N-BNP The N-BNP non-competitive assay has been described previously [10]. Briefly, the rabbit polyclonal antibody directed to the C-terminal (0.5 mg in 100 µl for each well) was immobilized on to ELISA plates. The N-terminal antibody was affinity-purified and biotinylated using N-hydroxysuccinimidobiotin. Samples (20 µl of plasma or 50 µl of urine) or N-BNP standards were incubated in the wells for 24 h at 4 C. ELISA plates were washed with 0.1 % Tween in PBS, and streptavidin labelled with methylacridinium ester was used to detect bound antibody, as described above. The lower limit of detection was 0.6 fmol/ml. Intra- and inter-assay coefficients of

3 Diagnosis of heart failure using urinary natriuretic peptides 131 Table 1 Patient characteristics Medians (range) are reported. P values were calculated using the Mann Whitney test by comparing normal and heart failure patients. ACE, angiotensin-converting enzyme; ns, not significant. Characteristics Normal controls Heart failure patients P value Number Gender 37 (45%) female 14 (41%) female ns Age (years) 73.6 ( ) 75.5 (46 87) ns Drug therapy None Diuretics 33 β-blockers 7 ACE inhibitors 21 Nitrates 11 Calcium channel blockers 3 Co-morbidity Ischaemic heart disease 20 Hypertension 14 Diabetes 4 Plasma peptides (fmol/ml) N-BNP 74.1 ( ) 5829 ( ) Urine peptides (fmol/ml) N-ANP 0.97 ( ) 5.38 ( ) N-BNP 0.60 ( ) ( ) CNP 0.35 ( ) 1.05 ( ) Urine peptides/creatinine (pmol/µg of creatinine) N-ANP 2.24 ( ) 9.99 ( ) N-BNP 1.74 ( ) ( ) CNP 0.73 ( ) 2.95 ( ) variation were under 5 %. There was no cross-reactivity with ANP, BNP or CNP. Statistics Data were analysed using SPSS Version 11.0 (Chicago, IL, U.S.A.). Data are presented as medians (range) for data with non-gaussian distribution, and comparisons were with the Mann Whitney test. Spearman correlation coefficients (r s ) are also shown. Peptide data were normalized by log-transformation before stepwise (forward and backward) linear regression analysis to detect independent predictors of urinary peptide levels. Logistic regression analysis was employed to evaluate independent predictors of the presence or absence of LVSD. The relative sensitivity, specificity and predictive value of peptides for the absence or presence of LVSD was assessed by construction of receiver operating characteristic (ROC) curves. Areas under the ROC curve were compared as described previously [11]. RESULTS Demographic and clinical characteristics of the patients and normal controls are summarized in Table 1. Aetiology of heart failure was often multifactorial, but a majority of the patients had ischaemic heart disease. Patients and normal controls were well matched for age and gender. As expected, plasma N-BNP was elevated in the patients compared with controls (P < ). Urinary natriuretic peptides were measurable in most of the patients, but many controls had levels below the detection limit of assays. The differences between controls and patients was significant (P < for N-ANP and N- BNP; P < for CNP). Correcting the urinary peptide concentration for urinary creatinine (in order to adjust for urine dilution) made no difference to these findings. Plasma N-BNP was correlated with urine N-ANP (r s = 0.49, P < ) and N-BNP (r s = 0.53, P < ) and weakly with urine CNP (r s = 0.20, P < 0.03). Urinary N-BNP was correlated with urine N-ANP (r s = 0.53, P < ) and weakly with urine CNP (r s = 0.19, P < 0.04). Urine N-ANP and CNP were also related (r s = 0.57, P < ). The plasma creatinine was correlated with both urinary N-ANP and N-BNP (r s = 0.31 and 0.33 respectively, P < ). On stepwise regression analysis to detect the best independent predictors of urinary N-BNP, plasma N-BNP (P < ) and plasma creatinine (P < 0.002) accounted for an r 2 of 0.43, with age and gender excluded as

4 132 L. L. Ng and others Figure 1 ROC curves for plasma N-BNP and urinary N-ANP, N-BNP and CNP for the diagnosis of heart failure Table 2 Areas under the ROC curves for plasma N-BNP and urinary N-ANP, N-BNP and CNP Uncorrected levels and levels corrected for urine creatinine are shown. Asymptotic 95 % confidence interval Variable ROC area S.E.M. Lower bound Upper bound Plasma N-BNP Urine N-ANP Urine CNP Urine N-BNP Urine N-ANP/creatinine Urine CNP/creatinine Urine N-BNP/creatinine independent predictors. The same predictor variables accounted for an r 2 of 0.39 for the urinary N-BNP/ creatinine ratio. The ROC curve for plasma N-BNP for detection of LVSD yielded an area of (Figure 1, and Table 2; P < compared with the diagonal). Of all the urinary natriuretic peptides considered, N-BNP had the largest ROC area under the curve (0.933), and correcting the peptide levels with urinary creatinine made little difference to the ROC area under the curves (Table 2). From the ROC for plasma N-BNP, the optimal cut-off value (113.5 fmol/ml) had a sensitivity and specificity of 97 and 64.5 % for detection of LVSD, with positive and negative predictive values of 53.2 and 98 %. The overall accuracy was 74 %. Using the best urinary peptide (N- BNP), a cut-off point of 11.6 fmol/ml had a sensitivity and specificity for LVSD detection of 97 % and 78 % respectively, with positive and negative predictive values of 64.7 and 98 %. The overall accuracy was 83.6 %. Thus urinary and plasma N-BNP provide almost equivalent performance in the detection of LVSD. Figure 2 ROC curves for plasma N-BNP, urinary N-BNP and the predicted probabilities derived from a logistic model of plasma and urinary N-BNP for the diagnosis of heart failure Logistic regression analysis to predict presence of LVSD was performed using urinary N-BNP, plasma N- BNP and plasma creatinine. Both urinary N-BNP and plasma N-BNP were significant independent predictors (odds ratios for a 10-fold rise in peptide level were and 82.4, P < and P < respectively) accounting for a Nagelkerke r 2 of The C-statistic derived from the predicted probabilities of this analysis (equivalent to the area under the ROC curve) was (95 % confidence interval, ), with a sensitivity of 100 % and a specificity of 90.1 % (Figure 2). The overall accuracy of diagnosis was 93.1 %. DISCUSSION Plasma BNP and N-BNP have increasingly gained acceptance as diagnostic aids for detection of LVSD [1 4]. However, previous experience with detection of urine natriuretic peptides has been limited [7,8]. In the present study, we have confirmed that N-ANP and CNP are detectable in urine from patients with LVSD, but we have also demonstrated for the first time the possibility of detecting N-BNP in urine in these patients. Although CNP was elevated in the urine from patients with LVSD, there was much overlap with the values from normal controls. CNP, being a cyclic peptide with a disulphide bond, is prone to denaturation and degradation, particularly by endopeptidases. The proximal tubule cell brush-border is rich in these enzymes. This overlap was less evident with N-ANP and least with N-BNP, perhaps reflecting the resistance of these N-terminal fragments of natriuretic peptides to degradation by urinary peptidases. The source of these peptides could be from glomerular filtration, although an additional contribution from tubule secretion cannot be excluded [12].

5 Diagnosis of heart failure using urinary natriuretic peptides 133 Urinary N-BNP was dependent on both plasma N- BNP and plasma creatinine. The dependence on plasma N-BNP may not be surprising as the peptide could be filtered. However, the urinary levels of N-BNP rise further with renal impairment, which may be expected to reduce filtration. It remains to be established with formal clearance studies whether damaged or underperfused renal tissue from reduced cardiac output could be an additional source of the urinary N-BNP, as BNP has been demonstrated to be produced by renal tubular cells [12]. The use of urinary N-BNP for the detection of LVSD was equivalent to that of plasma N-BNP. Both modes of testing had similar sensitivities, specificities and accuracies, and ROC area under the curves were similar. In addition, both had high negative predictive values and would be effective in the exclusion of a diagnosis of LVSD. The use of urinary N-BNP in detection of LVSD may be particularly efficient in a high-throughput screening scenario, where collection of specimens would be easier and more acceptable to patients. In addition, such a test could be used in patients who refuse to provide blood samples. However, logistic regression analyses suggested that both urinary N-BNP and plasma N-BNP were independent predictors for the presence of LVSD, and a combination of these measures may increase accuracy of diagnosis further. Conclusions In the present study, we have demonstrated that detection of urinary natriuretic peptides in patients with LVSD is feasible. Urinary N-BNP provides a similar accuracy in the detection of LVSD as plasma N-BNP and provides an alternative when blood samples are unavailable. Urinary N-BNP testing may be particularly useful in screening large numbers of subjects for LVSD. REFERENCES 1 McDonagh, T. A., Robb, S. D., Murdoch, D. R. et al. (1998) Biochemical detection of left ventricular systolic dysfunction. Lancet 351, Hobbs, F. D. R., Davis, R. C., Roalfe, A. K., Hare, R., Davies, M. K. and Kenkre, J. E. (2002) Reliability of N-terminal pro-brain natriuretic peptide assay in diagnosis of heart failure: cohort study in representative and high risk community populations. Br. Med. J. 324, Talwar, S., Squire, I. B., Davies, J. E., Barnett, D. B. and Ng, L. L. (1999) Plasma pro brain natriuretic peptide and the ECG in the assessment of left ventricular systolic dysfunction in a high risk population. Eur. Heart J. 20, Hunt, P. J., Richards, A. M., Nicholls, M. G. et al. (1997) Immunoreactive amino-terminal pro-brain natriuretic peptide (NT-proBNP): a new marker of cardiac impairment. Clin. Endocrinol. 47, Hammerer-Lercher, A., Neubauer, E., Muller, S., Pachinger, O., Puschendorf, B. and Mair, J. (2001) Head-to-head comparison of N-terminal pro-brain natriuretic peptide, brain natriuretic peptide and N-terminal pro-atrial natriuretic peptide in diagnosing left ventricular dysfunction. Clin. Chim. Acta 310, Kalra, P. R., Clague, J. R., Bolger, A. P. et al. (2003) Myocardial production of C-type natriuretic peptide in chronic heart failure. Circulation 107, Franz, M., Woloszczuk, W. and Horl, W. H. (2001) Plasma concentration and urinary excretion of N-terminal proatrial natriuretic peptides in patients with kidney diseases. Kidney Int. 59, Mattingly, M. T., Brandt, R. R., Heublein, D. M. et al. (1994) Presence of C-type natriuretic peptide in human kidney and urine. Kidney Int. 46, Ng, L. L., O Brien, R. J., Demme, B. and Jennings, S. (2002) Non-competitive immunochemiluminometric assay for cardiotrophin-1 detects elevated plasma levels in human heart failure. Clin. Sci. 102, Omland, T., Persson, A., Ng, L. et al. (2002) N-terminal pro-b-type natriuretic peptide and long-term mortality in acute coronary syndromes. Circulation 106, Hanley, J. A. and McNeil, B. J. (1983) A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 148, Mistry, S. K., Hawksworth, G. M., Struthers, A. D. and McLay, J. S. (2001) Differential expression and synthesis of natriuretic peptides determines natriuretic peptide receptor expression in primary cultures of human proximal tubular cells. J. Hypertens. 19, Received 14 July 2003/29 August 2003; accepted 18 September 2003 Published as Immediate Publication 18 September 2003, DOI /CS