The present is a multimarker subanalysis of the previously reported7 total cohort studied. The subanalysis was performed in those patients with availability of all the examined biomarkers (sNEP, NT-proBNP, hsTnT and ST2). From May 2006 to May 2013, ambulatory patients treated at a multidisciplinary HF clinic were consecutively included in the study. Referral inclusion criteria and blood sample collection have been described elsewhere.7,8 All biomarkers were analyzed in the same blood sample stored at −80°C without previous freeze-thaw cycles. All samples were obtained between 09:00 am and 12:00 pm.

All participants provided written informed consent, and the study was approved by the local ethics committee. All study procedures were conducted in accordance with the ethical standards outlined in the 1975 Declaration of Helsinki, as revised in 1983.

All patients were followed-up at regular predefined intervals, with additional visits as required in the case of decompensation. The regular visits schedule included a minimum of quarterly visits with nurses, biannual visits with physicians, and elective visits with geriatricians, psychiatrists, and rehabilitation physicians.9 Patients who did not attend the regular visits were contacted by telephone.

The primary outcome was a composite of CV death or HF hospitalization. Cardiovascular and all-cause deaths were also explored as secondary outcomes. A death was considered CV in origin if it was caused by HF (decompensated HF or treatment-resistant HF in the absence of another cause), sudden death (unexpected death, witnessed or not, of a previously stable patient with no evidence of worsening HF or any other cause of death), acute myocardial infarction (directly related in time to acute myocardial infarction due to mechanical, hemodynamic, or arrhythmic complications), stroke (associated with recently appearing acute neurologic deficit), procedural (post-diagnostic or post-therapeutic procedure death), and other CV causes (eg, rupture of an aneurysm, peripheral ischemia, or aortic dissection). Hospitalizations were identified from the clinic records of patients with HF, hospital wards, and the Catalan electronic medical record. Fatal events were identified from the clinical records of patients with HF, hospital wards, the emergency room, general practitioners, and by contacting the patient's relatives. Data were verified by the databases of the Catalan and Spanish health systems. Events were adjudicated by 2 of the authors (M. Domingo and J. Lupón). Follow-up was closed at March 31, 2014.

Human NEP was measured using a modified sandwich immunoassay (HUMAN NEP/CD10 ELISA KIT [reference, SK00724-01; lot number, 20111893, Aviscera Biosciences; Santa Clara, United States). Several modifications were made to improve the analytical sensitivity of the method and obtain a lower limit of sample quantification, as reported elsewhere.7 The modified protocol displayed analytical linearity for 0.250-4ng/mL. Samples with concentrations higher than 4ng/mL were diluted to a final concentration between 0.250 ng/mL and 64ng/mL. At a positive control value of 1.4ng/mL, the intra- and inter-assay coefficients of variation were 3.7% and 8.9%, respectively.

The NT-proBNP levels were determined using an immuno-electrochemiluminescence method (Elecsys®, Roche Diagnostics; Basel, Switzerland). This assay has < 0.001% cross-reactivity with bioactive B-type natriuretic peptide (BNP), and in the constituent studies in this report, the assay had inter-run coefficients of variation ranging from 0.9% to 5.5%.

Troponin levels were measured by electrochemiluminescence immunoassay using an hsTnT assay on the Modular Analytics E 170 (Roche Diagnostics). The hsTnT assay had an analytic range from 3 ng/L to 10000ng/L. At the 99th percentile value of 13 ng/L, the coefficient of variation was 9%. The assays were run with reagents from lot 157123 and were not affected by the analytical issues that have emerged with Roche hsTnT assays.

ST2 was measured from plasma samples using a high-sensitivity sandwich monoclonal immunoassay (Presage® ST2 assay, Critical Diagnostics; San Diego, California, United States). The ST2 assay had a within-run coefficient of < 2.5% and total coefficient of variation of 4%.

Categorical variables are expressed as percentages. Continuous variables are expressed as means ± standard deviation or medians [interquartile range] according to normal or nonnormal distributions. Normal distribution was assessed with normal Q-Q plots. The correlation between sNEP and NT-proBNP concentrations with age, left ventricular ejection fraction, LVEF, eGFR, assessed by the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) equation, blood urea and body mass index (BMI) were analyzed using the rho Spearman coefficient due to skewed distribution. Statistical differences (P-value for trend) in NT-proBNP and sNEP concentrations among eGFR subgroups (≥ 60, 30 to < 60, and < 30mL/min/1.73m2) and BMI subgroups (< 20.5, 20.5 to < 25.5, 25.5 to <30, and ≥ 30Kg/m2) were determined using the Spearman test.

Multivariable Cox regression analyses were performed using the backward step method. To fulfill the assumption of linearity of the covariables sNEP, NT-proBNP, hsTnT, and ST2, the logarithmic functions of sNEP, NT-proBNP and hsTnT, and the quadratic term of log (hsTnT) and of ST2 were used in the Cox models. For hazard ratio (HR) calculation in the 3 logarithm-transformed variables, a 1 standard deviation increase was used, and ST2 analyses were performed per every 10ng/mL change. In patients with SNEP levels below the lower range of detection (0.250ng/mL), a concentration of 0.249ng/mL was introduced as a continuous variable. The following variables were incorporated into the model: age, sex, ischemic etiology of HF, LVEF, New York Heart Association functional class, presence of diabetes mellitus, heart rate, systolic blood pressure, hemoglobin, serum sodium, eGFR, NT-proBNP, hsTnT, ST2, beta-blocker therapy, angiotensin-converting enzyme inhibitor or angiotensin receptor blockers therapy, and sNEP. To compare the prognostic values of both sNEP and NT-proBNP, different measurements of performance were used (discrimination, calibration, and reclassification), as described previously.9,10 For calculation of net reclassification improvement, we used risk categories based on tertiles of risk for the composite endpoint (<21%, 21%-39%, and >39%) and CV death (<11%, 11%-23%, and >23%).

Statistical analyses were performed using SPSS 15 (SPSS Inc.; Chicago, Illinois, United States) and the R (version 2.11.1) statistical package (Foundation for Statistical Computing, Vienna, Austria). A 2-sided P < .05 was considered significant.

The median soluble sNEP concentration was 0.64ng/mL [0,42-1,23] ng/mL and the median NT-proBNP concentration was 1187 ng/L [476-278] ng/L. sNEP levels were below the analytical measurement range in 101 patients (12.7%). Both biomarkers significantly correlated with age (ρ=0.41, P<.001 and ρ=0.19, P<.001), and NT-proBNP significantly correlated with eGFR rate (ρ=–0.37, P<.001), blood urea (ρ=0.31, P<.001), BMI (ρ=–0.29, P<.001), and weakly with LVEF (ρ=–0.08, P=.02); but sNEP did not (ρ=–0.03, P=.44; ρ=0.002, P=.95; ρ=0.04, P=.23; and ρ=0.008, P=.88, respectively). According to predefined eGFR and BMI strata, the serum concentration of NT-proBNP increased as renal function worsened (Figure 1A) and decreased at higher BMI (Figure 1B), whereas sNEP remained unaffected by these comorbidities (Figures 1C and 1D).

Figure 1.

Boxplots for N-terminal pro-B-type natriuretic peptide and soluble neprilysin according to estimated glomerular filtration rate and body mass index subgroups. A: N-terminal pro-B-type natriuretic peptide concentration according to grouped estimated glomerular filtration rate. B: N-terminal pro-B-type natriuretic peptide serum concentration according to grouped body mass index values. C: soluble neprilysin serum concentration according to grouped estimated glomerular filtration rate. D: soluble neprilysin concentration according to grouped body mass index values. The central box represents the values from the lower to the upper quartile and the middle line represents the median. eGFR, estimated glomerular filtration rate; sNEP, soluble neprilysin; NT-proBNP, N-terminal pro-B-type natriuretic peptide. The whiskers extend from the minimum to the maximum value, excluding outside values, which are not displayed.

(0.23MB).

Moreover, NT-proBNP highly correlated with hsTnT (ρ=0.57, P<.001), and also significantly with ST2 (ρ=0.33, P<.001), whereas sNEP did not correlate with hsTnT (ρ<0.01, P=.83), and correlated modestly with ST2 (ρ=0.15, P<.001).

As previously reported,7 sNEP and NT-proBNP did not correlate in the whole sample (ρ=0.01; P=.68); and in the present multimarker substudy the results remained nonsignificant (ρ=0.04, P=.28).

In separate Cox regression multivariate analyses that included clinical parameters and the well-recognized HF biomarkers hsTnT and ST2, sNEP remained independently associated with both the composite endpoint (HR=1.14; 95% confidence interval [95%CI], 1.02-1.27; P=.03) (Table 2) and CV death (HR=1.15; 95%CI, 1.01-1.31; P=.04) (Table 3), but NT-proBNP did not, neither for the composite endopint (HR=1.09; 95%CI, 0.93-1.28; P=.27) (Table 2), nor for CV death (HR=1.18; 95%CI, 0.97-1.43; P = .1) (Table 3).

Neither sNEP (HR=1.02; 95%CI, 0.92-1.14; P=.68) nor NT-proBNP (HR=1.09; 95%CI, 0.95-1.26; P=.23) were independently associated with all-cause death in the multivariate analysis. By contrast, in the analysis focused only on HF endpoints, sNEP remained independently associated with both HF-related death (HR=1.31; 95%CI, 1.11-1.55,; P=.001) and with the composite endpoint of HF-related death or HF hospitalization (HR=1.20; 95%CI, 1.06-1.35; P=.005), whereas NT-proBNP did not (HR=1.18; 95%CI, 0.97-1.43; P=.1 and HR=1.07; 95%CI, 0.89-1.29; P=.48, respectively).

Direct comparison of sNEP vs NT-proBNP in a predictive model containing 11 clinical variables plus hsTnT and ST2 showed no differences in discrimination (all differences in area under the curve, P>.05) (Figure 2) and reclassification (all integrated discrimination improvement and net reclassification improvement, P > .05) for both endpoints added to the reference model (Table 4). The calibration was good, with a nonsignificant Hosmer-Lemeshow test in all models (all P > .05), although the models containing sNEP showed slightly lower Aikaike information criterion and Bayesian information criterion values (better calibration) (data not shown). The addition of sNEP improved overall goodness-of-fit assessed by the likelihood ratio test for both the composite primary endpoint (P=.02) and CV death (P=.03), but NT-proBNP did not (P=.22 and P=.11 respectively). A significant P value in this test means that adding a new variable to the model significantly improves the accuracy of the model of reference.

Figure 2.

Area under the curve for the predictive models. Core model (red line): age, sex, ischemic etiology of heart failure, left ventricular ejection fraction, New York Heart Association functional class, presence of diabetes mellitus, hemoglobin, serum sodium, estimated glomerular filtration rate, high-sensitivity troponin T, ST2, beta-blocker therapy, and angiotensin-converting enzyme inhibitor or angiotensin receptot blockers therapy. The model also containing soluble neprilysin is depicted as a green line and the model also N-terminal pro-B-type natriuretic peptide is depicted as a blue line. A: composite endpoint of cardiovascular death and heart failure hospitalization (P=.83 for the comparison between the core model and the model also containing N-terminal pro-B-type natriuretic peptide; P=.24 for the comparison between the core model and the model also containing soluble neprilysin, and P=.22 for the direct comparison between the 2 models containing neurohormonal biomarkers). B: cardiovascular death (P=.87; P=.26, and P=.40 respectively for the same comparisons). AUC, area under the curve sNEP, soluble neprilysin; NT-proBNP, N-terminal pro-B-type natriuretic peptide.

(0.23MB).

Figure 3 shows the Cox regression event-free survival curve for the primary composite endpoint of CV death or HF hospitalization (Figure 3A) and survival curve for CV death (Figure 3B) according to the serum concentrations below or above the median of sNEP, hsTnT, and ST2.

Figure 3.

Cox regression survival curves according to the number of elevated biomarkers (soluble neprilysin, high-sensitivity troponin T and ST2; N=797). A: Event-free survival curve for the primary composite endpoint of cardiovascular death or heart failure hospitalization. B: Survival curve for cardiovascular death. hsTnT, high-sensitivity troponin T; sNEP, soluble neprilysin. A biomarker was considered elevated if equal or above the median values: soluble neprilysin, 0.64 ng/mL; high-sensitivity troponin T, 22.3 ng/L; ST2, 38.1 ng/mL.

(0.19MB).

The experimental assay for sNEP measurement described here has long incubation times, making it ill-suited for daily clinical use. We have no data on the stability of sNEP while frozen so we cannot rule out the possibility that the sNEP concentrations found would have been different in fresh samples. Samples were obtained during routine visits and no data on the clinical stability of patients (ie, possible decompensation during the 3 previous months) were collected. However, the sample is representative of ambulatory chronic HF patients in real life. Although the study population was a real-life HF population with different HF etiologies, it was treated at a specific multidisciplinary HF unit in a tertiary care hospital; most patients were referred from the cardiology department and, thus, were relatively young men with HF of ischemic etiology and reduced LVEF. As such, these results cannot necessarily be extrapolated to a more global HF population. In the near future, with the likely widespread use of NEP inhibitors in patients with HF and reduced LVEF, the prognostic value and use of sNEP and other circulating biomarkers may change.

Prospective studies are needed to assess tailored strategies for pharmacological NEP inhibition based on measurements of sNEP levels in patients with HF. The appropriate use of biomarkers supporting management of patients with HF should help to reduce the costs of a very costly disease in developed countries.28