The level of N-terminal pro-B-type natriuretic peptide (NT-proBNP) is a predictor of adverse events in patients with heart failure. We examined the relation between acute changes in NT-proBNP during a single hospitalization and subsequent mortality and readmission. The data from a cohort of 241 consecutive patients aged ≥25 years who had been admitted to an urban tertiary care hospital with a primary diagnosis of heart failure were analyzed. Creatinine and NT-proBNP were measured at admission and at discharge of the first admission. The patient demographics, co-morbidities, and length of stay were collected. The patients were prospectively grouped into 2 categories according to the acute changes in NT-proBNP: a decrease of ≥50% or <50% from admission to discharge. The primary composite outcome was readmission or death within 1 year of the first hospital admission. The unadjusted hazard ratio of readmission/death was 1.40 (95% confidence interval 0.97 to 2.01; p = 0.07) for those with a <50% decrease in NT-proBNP compared to their counterparts with a ≥50% decrease. After adjustment for age, gender, race, and admission creatinine and NT-proBNP, the risk of readmission/death was 57% greater for those with a <50% decrease (hazard ratio 1.57, 95% confidence interval 1.08 to 2.28; p = 0.02). An adjustment for co-morbidity, length of stay, and left ventricular ejection fraction did not significantly change this relation. Reductions in NT-proBNP of <50% during an acute hospitalization for heart failure might be associated with an increased hazard of readmission/death, independent of age, gender, race, creatinine, admission NT-proBNP, co-morbidities, left ventricular ejection fraction, and length of stay. In conclusion, patients with a <50% reduction in NT-proBNP might benefit from more intensive medical treatment, monitoring, and follow-up.
Multiple studies have examined the use of B-type natriuretic peptide (BNP), its pro-peptide, or the inert N-terminus (NT-proBNP) in the diagnosis, management, and treatment of heart failure (HF). NT-proBNP has been correlated with symptoms, and age-specific cutoffs for HF have been proposed. When the diagnosis of HF has been uncertain, the measurement of NT-proBNP has decreased hospital costs, emergency room visit time, and rehospitalizations by allowing for a more accurate and rapid diagnosis. A change in NT-proBNP during acute hospitalization can also predict adverse outcomes. For example, when changes in NT-proBNP during hospitalization were grouped into a decrease of >30%, an increase of >30%, and intermediate values, both the intermediate values and an increase of >30% in NT-proBNP were associated with increased adverse outcomes. However, these studies used widely varying cutoffs, selected cutoffs in a post-hoc fashion (e.g., greater than vs less than the median), or had a small sample size. Serial testing of NT-proBNP for the monitoring of HF suggests that a weekly change of 47% in NT-proBNP would be clinically significant. The objective of our study was to prospectively evaluate whether a decrease of <50% in NT-proBNP during acute hospitalization would be associated with an increased risk of 1-year readmission and mortality.
Methods
We consecutively enrolled 241 patients aged ≥25 years who had been admitted to Johns Hopkins Hospital from June 2006 to April 2007 with a primary admission diagnosis of HF and had received intravenous furosemide. All participants provided informed consent, and the Institutional Review Board approved the present study. Patients were excluded if they were unable to give informed consent or had end-stage renal disease requiring dialysis.
The NT-proBNP levels were assayed from blood drawn by hospital phlebotomists within the first 24 hours of admission and again at discharge. The patients were treated with standard dosages of nitrates, β blockers, angiotensin-converting enzyme inhibitors, and diuretics by the treating physician. The decision for discharge was determined by clinical judgment, and the treating physician was unaware of the discharge NT-proBNP level. The NT-proBNP was measured using a fully automated, sample-selective analyzer for heterogeneous immunoassays (Dimension RxL Clinical Chemistry System, Siemens Healthcare Diagnostics, Deerfield, Illinois).
The primary outcome was the interval to readmission or death within 1 year. The electronic medical records from Johns Hopkins Hospital and the affiliated Bayview Medical Center were reviewed to assess rehospitalization. The Social Security Death Index was used to obtain the mortality information. The participants and/or family members were interviewed up to every 3 months to assess the rate of rehospitalizations at other institutions and to supplement the mortality data.
We collected information on the demographics, co-morbidities, left ventricular ejection fraction (LVEF), and length of stay from the medical record. Data on co-morbidities were collected as dichotomous variables, either present or absent. The co-morbidities included hypertension, coronary artery disease, cerebrovascular disease, atrial fibrillation, peripheral vascular disease, chronic obstructive pulmonary disease, and diabetes mellitus. The most recent LVEF was recorded. The serum creatinine levels were measured daily.
The patients were prospectively grouped into 2 categories according to the acute changes in NT-proBNP during the first admission: a decrease of ≥50% or <50% from admission to discharge. The demographic, clinical, and laboratory variables were compared across the groups using Student’s t test, Pearson’s chi-square test, and the Wilcoxon rank-sum test, as appropriate. We calculated the incidence of readmission/death for each NT-proBNP category using a person-time approach.
The association between the change in NT-proBNP and the cumulative risk of readmission/death was first evaluated using the Kaplan-Meier method with log-rank tests to determine significance of differences in the cumulative incidence. Cox proportional hazards models were subsequently used to calculate the hazard ratios and 95% confidence intervals for the risk of readmission or death, comparing the groups with a NT-proBNP decrease of <50% and ≥50%, after adjusting for demographics, admission NT-proBNP and creatinine level, co-morbidities, LVEF, and length of stay. The proportional hazards assumption was assessed by hazard function plots and Schoenfeld and Martingale residuals. The tests of significance were 2-tailed, with an α level of 0.05. Statistical analysis was performed using Stata, version 10 (StataCorp, College Station, Texas).
Results
A total of 241 patients were enrolled; 1 patient withdrew from the study. Also, 3 patients who died during the initial admission were excluded. Of the 237 remaining records, 20 were missing at least one major covariate and were not included in the present analysis. The covariates not documented in the medical record were considered missing.
Of the 217 patients included in the present study, 50% were men and 67% nonwhite. Their mean age was 63.3 ± 14.4 years, the median admission creatinine was 1.2 mg/dl (interquartile range 1.0 to 1.7), the median admission NT-proBNP was 5,913 pg/ml (interquartile range 1,831 to 10,989), median LVEF was 30% (interquartile range 15% to 55%), and the median length of stay was 5 days (interquartile range 3 to 8). The patients had a history of hypertension (70%), diabetes (48%), coronary artery disease (43%), atrial arrhythmia (30%), chronic obstructive pulmonary disease (23%), peripheral vascular disease (13%), and cerebrovascular disease (14%). A total of 134 events and 48 deaths (22%) occurred. Finally, 86 patients (40%) were readmitted at least once and 23 (11%) had had ≥3 readmissions within 1 year.
The patients were prospectively categorized into those with an acute decrease in NT-proBNP of <50% (n = 120) and those with an acute decrease of ≥50% (n = 97) during the index (first) admission. The characteristics of the 2 groups are summarized in Table 1 . The proportion of men in both groups was equal. Of the demographic variables collected, the ≥50% group had more nonwhites (p = 0.002) and was 3.9 years younger (p = 0.05) than the <50% group. The median admission creatinine and NT-proBNP levels were similar in both groups. The median discharge NT-proBNP was, however, significantly lower in the ≥50% decrease group (p <0.001). The distribution of co-morbidities between the 2 groups was similar, except for hypertension. More patients had hypertension in the ≥50% group (p = 0.02). The median LVEF, proportion with preserved LVEF, and length of stay were similar for both groups. The median decrease in NT-proBNP was similar among those with preserved LVEF (46%) and reduced LVEF (41%). Additionally, the percentage of patients with preserved LVEF who exhibited a ≥50% decrease in NT-proBNP (47%) was similar to those with a reduced LVEF (44%).
| Variable | Decrease in NT-proBNP | p Value ⁎ | |
|---|---|---|---|
| <50% (n = 120) | ≥50% (n = 97) | ||
| Men | 50% | 50% | 0.94 |
| Nonwhite | 58% | 78% | 0.002 |
| Mean age (years) | 65.1 ± 14.6 | 61.2 ± 13.9 | 0.05 |
| NT-proBNP level (pg/ml) | |||
| At admission | 0.97 | ||
| Median | 6,087 | 5,599 | |
| Interquartile range | 1,394–11,868 | 2,217–9,556 | |
| At discharge | <0.001 | ||
| Median | 5,496 | 1,221 | |
| Interquartile range | 1,429–11,137 | 363–2,898 | |
| Admission creatinine level (mg/dl) | 0.27 | ||
| Median | 1.3 | 1.2 | |
| Interquartile range | 1–1.7 | 0.9–1.7 | |
| Co-morbidities | |||
| Chronic obstructive pulmonary disease | 25% | 20% | 0.34 |
| Peripheral vascular disease | 14% | 11% | 0.53 |
| Diabetes mellitus | 46% | 52% | 0.40 |
| Cerebrovascular disease | 14% | 14% | 0.96 |
| Hypertension | 63% | 78% | 0.02 |
| Coronary artery disease | 45% | 40% | 0.48 |
| Atrial arrhythmia | 30% | 29% | 0.86 |
| Left ventricular ejection fraction (%) | |||
| Median | 30 | 30 | 0.88 |
| Interquartile range | 15–53 | 15–55 | — |
| Proportion preserved (≥50%) | 31% | 28% | 0.68 |
| Initial hospital length of stay (days) | 0.84 | ||
| Median | 5.5 | 5.0 | |
| Interquartile range | 3–8 | 3–8 | |
⁎ Groups compared using Student’s t test, Pearson’s chi-square test, or rank sum test.
The rate of readmission/death was 2.2 and 3.2 per 1,000 person-days in the ≥50% and <50% group, respectively (p = 0.035). Kaplan-Meier survival curves for the risk of readmission/death within 1 year showed a divergence of the curves at 1 month after discharge ( Figure 1 ). Using Cox proportional hazard modeling, the patients in the <50% group had an increased risk of the combined end point of death or rehospitalization ( Table 2 ). A decrease of <50% in NT-proBNP during admission was associated with a 40% increase in the risk of readmission/death within 1 year. Adjustment for demographics, admission creatinine level, and admission NT-ProBNP level increased this risk to 57%. Additional adjustment for co-morbidities, LVEF, and length of stay did not significantly change this risk. Including time in the Cox proportional hazard model did not suggest a significant interaction. Overlays of the Kaplan-Meier curve and Cox proportional hazards model showed a good correlation, and the proportional hazards assumption was met according to the Schoenfeld residuals (p = 0.80).
