The aim of our study is to investigate the effect of spironolactone on 30-day outcomes in patients with acute heart failure (AHF) and the association between treatment and outcomes stratified by biomarkers. We conducted a secondary analysis of the biomarker substudy of the multicenter COACH (Co-ordinating Study Evaluating Outcomes of Advising and Counseling in Heart Failure) trial involving 534 AHF patients for 30-day mortality and HF rehospitalizations. Spironolactone therapy was initiated and terminated at the discretion of the treating physician; 30-day outcomes were compared between patients who were treated with spironolactone and those who were not. Outcomes with spironolactone therapy were explored based on N-terminal pro–B-type natriuretic peptide, ST2, galectin-3, and creatinine levels. Spironolactone was prescribed to 297 (55.6%) patients at discharge (158 new and 139 continued). There were 19 deaths and 30 HF rehospitalizations among 46 patients by 30 days. Patients discharged on spironolactone had significantly less 30-day event (hazard ratio 0.538, p = 0.039) after adjustment for multiple risk factors. Initiation of spironolactone in patients who were not on spironolactone before admission was associated with a significant reduction in event rate (hazard ratio 0.362, p = 0.027). The survival benefit of spironolactone was more prominent in patient groups with elevations of creatinine, N-terminal pro–B-type natriuretic peptide, ST2, or galectin-3. In conclusion, AHF patients who received spironolactone during hospitalization had significantly fewer 30-day mortality and HF rehospitalizations, especially in high-risk patients.
Adverse outcomes from acute heart failure (AHF) are a major health-care challenge worldwide, with an approximate 30% risk of 90-day mortality and rehospitalization. To better understand the short-term therapeutic effects of mineralocorticoid receptor antagonists (MRAs) in patients hospitalized for AHF, we performed a secondary analysis of patients who enrolled in the biomarker substudy of Co-ordinating Study Evaluating Outcomes of Advising and Counseling in Heart Failure (COACH) trial. Our analysis aimed to evaluate the association between MRA use in patients with AHF and 30-day mortality and HF-related rehospitalizations. A secondary aim was to evaluate the interaction between MRA and events by biomarker analysis to better understand the potential mechanism and/or patient profile associated with MRA and improved outcomes.
Methods
This study is a secondary analysis of the biomarker substudy of the COACH study. The study design and main results of the COACH trial and the COACH biomarker substudy were described in detail previously. Briefly, the COACH study was a multicenter, prospective randomized HF disease management study, enrolling patients hospitalized for AHF before discharge. Patients were included before hospital discharge for AHF. The patients were randomized to 1 of 3 groups: (1) usual care, (2) basic support, and (3) intensive support. Patients were enrolled irrespective of EF (ranging from severely reduced EF to normal EF; mean EF: 33%). The patients were initiated on standard HF medications as tolerated before discharge. Patients were excluded if they were enrolled in another study, unable to complete the questionnaires, invasive procedure, or cardiac surgery within 6 months of admission or planned within 3 months after discharge, ongoing evaluation for heart transplantation, or not able or willing to give informed consent. From the COACH study, 592 patients were enrolled in the COACH biomarker substudy. Serum samples were collected during the index hospitalization, just before discharge, after initial stabilization after the AHF admission. Mean follow-up was 18 months. Demographic and clinical data were collected during index admission from medical records and patient interviews; 534 patients with sufficient serum for investigational biomarker (ST-2, N-terminal pro–B-type natriuretic peptide [NT-proBNP], and galectin-3 [Gal-3]) analysis formed our present study population.
The primary end point for the current analysis was 30-day mortality and HF-related rehospitalization (defined as an unplanned overnight stay in the hospital because of AHF). A secondary end point was to investigate the effect of spironolactone therapy in relation to biomarker levels (NT-proBNP, ST2, Gal-3, and creatinine). These biomarkers were chosen for this analysis as they have all been associated with increased risk for short-term adverse outcomes in previous studies. All events were evaluated and adjudicated by an independent end point committee. This study complied with the Declaration of Helsinki and was approved by the local medical ethics committees or Institutional Review Board at each of the participating sites. All patients provided written informed consent.
The biomarkers analyzed in this study were Gal-3, ST-2, and NT-proBNP. Gal-3 was measured with an enzyme-linked immunosorbent assay (BG Medicine, Inc., Waltham, MA). This assay has a lower limit of detection of 1.13 ng/ml and exhibits no cross-reactivity with collagens or other members of the galectin family. Total imprecision for this assay was <6%. ST-2 measurements were made with the Presage ST-2 assay by Critical Diagnostics (San Diego, CA). The assay had a limit of quantitation of 2 units/ml and imprecision <4.0%. NT-proBNP was measured by electrochemiluminescence on the ElecSys 2010 analyzer (Roche Diagnostics, Indianapolis, IN). Performance in the laboratory for this assay included a limit of quantitation of 2.7 pmol/L and within-run imprecision of 5.7%.
Values were expressed as means and SDs, median and interquartile ranges, or counts or percentages as appropriate. Groups were compared with independent-samples t tests, Fisher’s exact tests for 2 by 2 tables, chi-square for a 2 by 3 table, and Mann-Whitney U tests as appropriate. SSPS (IBM Inc.) was used for all statistical analysis. All analyses were exploratory, and p-value of <0.05 was considered to denote significant differences.
We constructed a Kaplan-Meier survival curve for the primary end point, comparing patients who were or were not discharged on spironolactone. Survival differences were evaluated with Cox regression analysis throughout. Univariate analysis of clinical parameters and biomarkers (NT-proBNP, ST-2, Gal-3, and creatinine) were performed to evaluate for significant predictors of the 30-day composite end point. Significant univariate predictors were added to a multivariate model. Differences based on spironolactone treatment at discharge were separately evaluated in groups receiving and not receiving spironolactone at admission. For subgroup analyses, the biomarkers were cut by predefined cutpoints. NT-proBNP cutpoint was 6,300 pg/ml based on findings from previous studies. The ST-2 cutpoint was the median split. The predefined cutpoint for Gal-3 was 17.8 ng/ml. This cutoff was derived from a previous analysis, which demonstrated that patients with Gal-3 levels >17.8 ng/mL were at significantly increased risk for short-term AHF-related readmissions. This cutpoint was also approved by the Food and Drug Administration. The median of 1.3 mg/dl was used as the cutoff point for creatinine. Each biomarker subgroup was further analyzed for interaction with spironolactone therapy.
Results
Demographic and clinical information for the 534 patients included in our current analysis based on spironolactone treatment is summarized in Table 1 along with univariate hazard ratios for the prediction of the primary end point. Demographic information for patient included in our current analysis and for patients who were excluded because of insufficient serum for biomarker analysis was summarized in Supplementary Table 1 . Patients treated with spironolactone had less frequent hypertension and more implantable defibrillators and lower systolic blood pressure, sodium, and LVEF than patients not receiving spironolactone. Within our patient population, 176, 178, and 182 were randomized to usual care, basic support, and intensive support, respectively. Within these groups, discharge spironolactone rates were similar (55.2%, 61.2%, and 50.5%, respectively, p = 0.123). Group assignment was also unrelated to end point events (p = 0.355). Patients were relatively well treated in regard to evidence-based chronic HF therapies, and rates did not differ significantly between those receiving and not receiving spironolactone except that those receiving spironolactone were less frequently discharged with calcium channel blockers (9.8% vs 19.4%, p = 0.002).
| Parameter | Spironolactone | p | Endpoint Events Hazard Ratio (CI) | p | |
|---|---|---|---|---|---|
| Yes n = 297 (55.6%) | No n = 237 (44.4) | ||||
| Treatment group | 0.123 | 0.355 | |||
| Usual care | 32.3% | 32.9% | Reference | ||
| Basic support | 36.7% | 29.1% | 0.672 (0.312–1.448) | 0.310 | |
| Intensive support | 31.0% | 38.0% | 1.154 (0.594–2.245) | 0.672 | |
| Men | 58.9% | 62.9% | 0.373 | 1.125 (0.618–2.048) | 0.699 |
| History of myocardial infarction | 39.7% | 42.6% | 0.536 | 1.342 (0.752–2.393) | 0.319 |
| Prior percutaneous intervention | 11.1% | 9.7% | 0.671 | 1.806 (0.843–3.872) | 0.129 |
| Coronary artery bypass grafting | 17.2% | 12.2% | 0.143 | 1.010 (0.452–2.259) | 0.980 |
| Atrial fibrillation | 46.5% | 43.5% | 0.540 | 1.463 (0.819–2.614) | 0.198 |
| Implantable defibrillator | 2.0% | 0.0% | 0.036 | 1.943 (0.268–14.09) | 0.511 |
| Diabetes mellitus | 18.2% | 19.8% | 0.657 | 2.189 (1.182–4.055) | 0.013 |
| Transient ischemic attack | 8.8% | 7.2% | 0.527 | 1.073 (0.385–2.993) | 0.892 |
| Stroke | 8.1% | 11.0% | 0.296 | 2.101 (0.980–4.503) | 0.056 |
| Chronic obstructive pulmonary disease | 25.9% | 30.4% | 0.286 | 1.536 (0.844–2.795) | 0.160 |
| Renal disease | 6.7% | 7.6% | 0.737 | 0.575 (0.140–2.374) | 0.445 |
| Liver disease | 2.7% | 2.1% | 0.781 | 0.890 (0.396–2.001) | 0.778 |
| Hypertension | 37.0% | 48.1% | 0.011 | 1.163 (0.651–2.078) | 0.609 |
| Peripheral arterial disease | 14.1% | 18.1% | 0.234 | 1.110 (0.518–2.379) | 0.788 |
| Resting dyspnea | 64.1% | 61.8% | 0.586 | 0.915 (0.506–1.654) | 0.768 |
| Orthopnea | 70.9% | 62.9% | 0.060 | 0.783 (0.429–1.431) | 0.427 |
| Exertional dyspnea | 93.5% | 93.6% | 1.000 | 0.551 (0.218–1.394) | 0.208 |
| Age (years) | 70.6 ± 11.0 | 71.4 ± 11.2 | 0.406 | 1.014 (0.986–1.042) | 0.338 |
| Body mass index (kg/m 2 ) | 26.9 ± 5.6 | 27.4 ± 5.5 | 0.328 | 1.003 (0.951–1.058) | 0.904 |
| Heart rate (beats/min) | 74.9 ± 12.9 | 73.7 ± 13.2 | 0.279 | 1.013 (0.992–1.035) | 0.230 |
| Systolic blood pressure (mm Hg) | 114.1 ± 19.4 | 123.5 ± 22.4 | <0.001 | 1.002 (0.989–1.016) | 0.770 |
| Diastolic blood pressure (mm Hg) | 67.8 ± 11.3 | 69.8 ± 13.5 | 0.063 | 0.988 (0.964–1.013) | 0.341 |
| Serum sodium (mmol/L) | 138.3 ± 4.4 | 139.3 ± 4.1 | 0.007 | 0.970 (0.908–1.035) | 0.358 |
| eGFR (mL/min/1.73 m 2 ) | 54.3 ± 19.6 | 53.0 ± 20.3 | 0.442 | 0.971 (0.956–0.987) | <0.001 |
| LV ejection fraction (%) ∗ | 30.2 ± 13.1 | 35.3 ± 14.9 | <0.001 | 1.006 (0.984–1.029) | 0.575 |
| Galectin-3 (ng/mL) † | 25.6 (20.7–32.3) | 25.1 (21.1–31.1) | 0.796 | 3.145 (0.560–17.66) | 0.193 |
| NT-proBNP (pg/mL) † | 2768 (1316–6780) | 2444 (1305–4610) | 0.267 | 4.064 (2.254–7.326) | <0.001 |
| ST-2 (ng/mL) † | 2.49 (1.42–5.41) | 2.43 (1.42–5.27) | 0.673 | 2.608 (1.392–4.885) | 0.003 |
| Creatinine (mg/dL) † | 1.28 (1.05–1.61) | 1.27 (1.05–1.61) | 0.282 | 18.652 (3.660–95.07) | <0.001 |
| ACE inhibitors | 74.7% | 67.9% | 0.100 | 0.737 (0.402–1.352) | 0.324 |
| Angiotension receptor blockers | 11.1% | 12.2% | 0.686 | 1.400 (0.626–3.131) | 0.412 |
| Beta blockers | 64.3% | 68.4% | 0.358 | 0.668 (0.373–1.196) | 0.174 |
| Calcium channel blockers | 9.8% | 19.4% | 0.002 | 0.270 (0.066–1.115) | 0.070 |
∗ Data available for 432 patients.
There were 19 deaths and 30 HF-related rehospitalizations among 46 patients (8.6%) in 30 days. Of the 534 patients, 297 were discharged with spironolactone. There were 21 events among patients who were discharged on spironolactone (7.1%). There were 28 events among patients who were not discharged on spironolactone (11.8%). Spironolactone significantly reduced the 30-day composite of mortality and HF-related rehospitalization (hazard ratio [HR] 0.538, 95% confidence interval [CI] 0.299 to 0.968, p = 0.039, see Figure 1 ). In a multivariable model adding the significant univariate predictors from Table 1 , spironolactone therapy remained a significant predictor of favorable outcomes (mortality and rehospitalization) (HR 0.455, CI 0.249 to 0.832, p = 0.011). To avoid multicollinearity, this model was repeated with creatinine or estimated glomerular filtration rate omitted and remained significant (p = 0.014 and p = 0.015, respectively). When adjusted for all the significant univariate predictors in Table 1 , the spironolactone effect approached significance (HR 0.518, CI 0.257 to 1.045, p = 0.066). However, 102 subjects were omitted from this analysis for missing EF. EF was not significant in the model (p = 0.564). After removing EF from the model, the beneficial spironolactone effect regained statistical significance (HR 0.445, CI 0.238 to 0.829, p = 0.011).
