Original Article. Jose I. Iglesias 1 5, Laura DePalma 6, Deborah Hom 3 5, Maria Antoniotti 3 5, Sammy Ayoub 3 5 and Jerrold S. Levine 7.

Transcription

1 Nephrol Dial Transplant (2008) 23: doi: /ndt/gfm565 Advance Access publication 5 September 2007 Original Article Predictors of mortality in adult patients with congestive heart failure receiving nesiritide Retrospective analysis showing a potential adverse interaction between neseritide and acute renal dysfunction Jose I. Iglesias 1 5, Laura DePalma 6, Deborah Hom 3 5, Maria Antoniotti 3 5, Sammy Ayoub 3 5 and Jerrold S. Levine 7 1 Department of Nephrology, UMDNJ School of Osteopathic Medicine, Stratford, NJ, 2 Department of Nephrology, Robert Wood Johnson School of Medicine, New Brunswick, NJ, 3 Department of Nursing, 4 Department of Pharmacy and 5 Department of Nephrology, Community Medical Center, Toms River, NJ, 6 Department of Medicine, Philadelphia College of Osteopathaic Medicine, Philadelphia PA and 7 Department of Nephrology, University of Illinois at Chicago School of Medicine, Chicago, IL, USA Abstract Background. A recent meta-analysis has suggested that nesiritide (NES), a new agent for the treatment of congestive heart failure (CHF), is associated with an increased risk of short-term mortality. Methods. We retrospectively examined this issue among 1407 consecutive elderly CHF patients by Pearson s chi-squared test, and determined independent risk factors for 60-day mortality by multivariate analysis in a cohort of 682 patients for whom we had sufficient clinical and laboratory data. Results. Univariate analysis revealed that NES usage was associated with increased mortality (n ¼ 1407, 10 vs 6%, P ¼ 0.011; n ¼ 682, 19 vs 12.5%, P ¼ 0.046). However, by forward stepwise regression analysis, NES usage did not survive as an independent predictor of mortality. The following variables were independent predictors of mortality: development of acute renal failure (ARF) defined as an increase of serum creatinine (SCr) 0.5 mg/dl; lack of b-adrenergic blockade; increased admission blood urea nitrogen; digoxin use; and increased admission brain natriuretic peptide. When patients were stratified according to NES usage, ARF emerged as an independent risk factor for mortality only among patients who received NES. Strikingly, among CHF patients who developed ARF (n ¼ 102), NES usage emerged as the only independent predictor of mortality (P ¼ 0.006, OR ¼ 3.73, 95% CI ). Conclusion. We conclude that, while NES per se is not independently associated with an increased risk for mortality, the development of ARF in association with NES use may confer an increased risk of mortality. Correspondence to: Jose I. Iglesias, 11 Paulette Lane, Howell, NJ, 07731, USA. jiglesias@verizon.net Keywords: acute renal failure; brain natriuretic peptide; congestive heart failure; nesiritide Introduction Recombinant human B-type natriuretic peptide, or brain natriuretic peptide (BNP), is a new agent used in the treatment of congestive heart failure (CHF) [1,2], and has been given the generic name of neseritide (NES). NES is a potent vasodilator with several pharmacological effects desirable in the setting of CHF [1]. Despite the potential benefits of NES, there is reason for caution. Two recent meta-analyses of pooled data from five randomized double-blind, parallel-group trials have suggested that the use of NES is associated with an increased risk of both worsening renal function and short-term mortality [3,4]. However, a major limitation of both metaanalyses was that there were inadequate data to adjust for potential confounding differences between groups other than assignment of treatment. In a recent publication, we addressed this limitation with respect to acute renal failure (ARF) [5]. We examined the risk for ARF associated with NES usage among a cohort of 219 consecutive patients admitted to a community hospital with a diagnosis of acute decompensated CHF. ARF was defined as an increase of serum creatinine greater than either 0.3 or 0.5 mg/dl, occurring within 30 days of admission. We used multivariate analysis to adjust for differences between patients receiving vs not receiving NES. Despite a greater burden of comorbid factors among patients receiving NES, NES usage did not emerge as an independent risk factor for the development of ARF [5]. As a second approach to this question, ß The Author [2007]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org

2 Mortality in CHF with nesiritide 145 we determined separately the independent risk factors for the development of ARF among patients who received vs did not receive NES. Comparison of the risk factors in these two groups suggested that administration of NES in the setting of severely diminished renal perfusion may confer an increased risk for ARF [5]. Here, we address the role of NES in short-term mortality occurring within 60 days of hospital admission. As in our previous analysis of ARF, we took two complementary approaches to adjust for baseline differences and to identify potential risk factors for mortality [5]. First, we asked whether NES was an independent risk factor for mortality among consecutive patients admitted to a community hospital with a diagnosis of acute decompensated CHF (n ¼ 1407). We used multivariate analysis to determine clinical and demographic factors that were independent predictors of mortality. The 219 patients from our previous study were included among these 1407 patients [5]. Second, to gain insight whether NES may predispose to mortality under specific conditions, we compared the risk factors for mortality among patients who received NES vs those who did not. To adjust for baseline differences, we again used multivariate analysis. Our data suggest that, while NES usage per se is not independently associated with an increased risk for ARF or death, the development of ARF in association with NES usage may confer an increased risk of mortality. Methods Patients In order to identify and evaluate risk factors associated with mortality in acute decompensated CHF patients in general and in such patients receiving NES, we reviewed the clinical course of 1407 consecutive patients undergoing therapy for CHF. We focused our risk-factor analysis for 60-day mortality on a cohort of 682 patients (from the 1407 patients) for whom we had sufficient clinical and laboratory data. These patients were admitted between 2 February 2003, and 11 September NES was administered as an intravenous bolus of 2 mcg/kg followed by a continuous infusion at a rate of 0.01 mcg/kg/min. Study subjects were identified in one of the two ways: (i) Trendstar Õ decision support software and database (McKesson, San Francisco, CAUSA); and (ii) Joint Commission of the Accreditation of Health Care Organization (JCAHO)/Center for Medicare Services (CMS), Peer Review Organization of New Jersey (PRONJ), 7th Scope of Work involving CHF [6]. Patients with CHF were identified by the diagnosis-related group (DRG) code 127 and the 9th International Classification of Disease-Clinical Modification (ICD-9-CM) codes (428., , , , , , and ) [7]. All DRG and ICD-9-CM codes were entered by professional coders who were approved by the respective agencies. All patients entered into this study had at least one sign or symptom consistent with the diagnosis of CHF, including dyspnoea at rest, dyspnoea with minimal exertion, peripheral oedema, rales or radiological evidence of CHF. A diagnosis of coronary artery disease (CAD) required the presence of at least one of the following in the medical record: previous myocardial infarction, history of coronary artery bypass graft, history of percutaneous coronary angioplasty, positive coronary angiography or positive noninvasive cardiac testing. Patients under the age of 18 and those with end-stage renal disease on maintenance dialytic therapy were excluded from the analysis. Evaluation included the following admission variables: serum sodium (Na), blood urea nitrogen (BUN), haematocrit (Hct.), serum creatinine (SCr), plasma brain natriuretic peptide (BNP), peak SCr during hospitalization, medications, comorbidities, demographics and admission physiological variables. Medications, comorbidities and demographic variables were abstracted from the medical record. We stratified ARF into two categories based on the magnitude of the rise in SCr occurring any time within period of admission: increase of SCr 0.3 mg/dl and increase of SCr 0.5 mg/dl. Our choice of these thresholds, while arbitrary, is based upon our previous study, in which we analysed risk factors for the development of ARF in association with NES [5], as well as upon previous studies showing that a threshold increase in SCr of 0.3 mg/dl is associated with a longer hospital stay and an increased mortality (both in-hospital and long-term) [8,9]. Moreover, in the recent meta-analyses demonstrating an increased risk of ARF and mortality in association with NES, the threshold increase of creatinine for ARF was similarly defined as 0.5 mg/dl occurring any time with 30 days of NES administration [3,4]. Glomerular filtration rate (GFR) was estimated from the simplified equation developed by the Modification of Diet in Renal Disease Study [10]. Data analysis Primary outcome was mortality occurring with 60 days of hospitalization. We initially examined whether the use of NES was associated with 60-day mortality among 1407 consecutive CHF patients. We then performed risk factor analyses for 60-day mortality on a cohort of 682 of these 1407 patients for whom we had sufficient clinical and laboratory data. The groups we evaluated included the entire cohort (CHF, n ¼ 682), as well as the sub-cohorts receiving NES (CHF þ NES, n ¼ 178) and not receiving NES (CHF þ NO NES, n ¼ 504). Summary statistics were computed for the entire study cohort, and for the NES and NO NES sub-cohorts. We performed both univariate and multivariate analysis. Continuous variables were expressed as mean SD, and compared by the Student s t-test or the Wilcoxon rank-sum test, as appropriate. Categorical values were compared with Pearson s chi-squared test. Variables which were significant by univariate analysis at P < 0.05 were candidates for multivariate analysis. Logistic regression analysis with forward variable selection was performed to determine variables independently predictive of mortality. In logistic regression, step selections were based on the maximum likelihood ratio. For continuous variables, the odds ratio (OR) represents the relative amount by which the probability of obtaining the outcome variable will increase or decrease if the independent variable is increased by exactly one unit. ORs and their 95% confidence intervals

3 146 J. Iglesias et al. (CIs) were determined by exponentiation of the regression coefficient and its upper and lower 95% CIs, respectively. Results Association of NES with increased mortality of CHF patients by univariate analysis To determine whether NES usage is associated with an increased risk for mortality among patients presenting with acute decompensated CHF, we retrospectively evaluated 1407 consecutive patients, of whom 334 (24%) received NES (Table 1). Mortality within 60 days of admission was significantly increased among patients who received NES vs those who did not (10 vs 6%, P ¼ 0.011, OR ¼ 1.041, 95% CI ¼ , Pearson s chi-squared test). In evaluating this association, it is important to adjust for baseline differences other than the choice whether or not to use NES. We also wished to identify additional risk factors besides NES that may be associated with increased mortality. To address both these issues, we focused our analysis on a representative cohort of 682 patients, for whom we had complete clinical and laboratory data. All 99 non-survivors had complete data available. NES was given to 178 patients (26%) of this cohort. Although this cohort had an overall increased incidence of mortality as compared with the entire population (14.5 vs 7%), NES was also associated with an increased risk for 60-day mortality (19% vs 12.5%, P ¼ 0.046, OR ¼ 1.60, 95% CI ¼ , Pearson s chi-squared test). Previously we showed that NES usage was not an independent risk factor for the development of ARF [5]. All 219 patients analysed in that study were included in the present cohort of 682 patients. The lack of association between NES usage and the development of ARF is replicated in our present larger cohort of 682 patients, irrespective of whether ARF is Table 1. NES usage and survival in 1407 consecutive CHF patients and a representative cohort of 682 CHF patients Entire cohort (n ¼ 1407) Survivors Nonsurvivors No NES (n ¼ 1073) 1010 (94%) 63 (6%) NES (n ¼ 334) 301 (90%) 33 (10%) Study cohort (n ¼ 682) Survivors Nonsurvivors No NES 441 (87.5%) 63 (12.5%) (n ¼ 504) NES (n ¼ 178) 145 (81%) 33 (19%) defined as a rise of SCr 0.3 mg/dl (34 vs 31%, P ¼ 0.388, OR ¼ 1.17, 95% CI ¼ , Pearson s chi-squared test) or a rise of SCr 0.5 mg/dl (16 vs 14%, P ¼ 0.54, OR ¼ 1.15, 95% CI ¼ , Pearson s chi-squared test). The latter definition of ARF is comparable to that used in the recent metaanalyses of NES usage [3,4]. Patient deaths occurred at a mean time of days following admission, with a range of 1 59 days. Non-survivors who received NES died later than non-survivors who did not receive NES ( vs days, P ¼ 0.007, Wilcoxon rank sumtest). It does not appear that NES was used as a salvage therapy (in other words, after failure of diuretics and failure of other therapies) in those patients who received NES and died. Of the 178 patients who received NES, 124 (70%) received NES on the first day of hospitalization. There was no statistical difference in the day of initiation of NES treatment for patients who died vs those who survived ( vs days, P ¼ 0.19, Wilcoxon rank-sum test). However, duration of treatment with NES was significantly longer in non-survivors vs survivors ( vs days, P ¼ 0.004, Wilcoxon rank-sum test). In multivariate analysis, duration of therapy did not survive as an independent predictor of mortality (see subsequently). Additional risk factors associated with increased mortality of CHF patients Table 2 gives the results of a univariate analysis among all CHF patients in our cohort (n ¼ 682) for admission characteristics associated with 60-day mortality. The following demographic, clinical and laboratory variables at admission were significantly associated with mortality: increased age (P ¼ 0.002), increased baseline SCr (P ¼ 0.028), increased baseline BUN (P ¼ ), increased baseline BUN/SCr ratio (P ¼ ), decreased baseline estimated GFR (P ¼ 0.012), increased serum BNP P ¼ 0.001), lack of use of b-adrenergic blockers (P ¼ 0.001, OR ¼ 0.45, 95% CI ¼ ), lack of use of angiotensinconverting enzyme inhibitor/angiotensin-2 receptor blocker (ACEI/A2RB) (P ¼ 0.005, OR ¼ 0.55, 95% CI ¼ ), use of an inotrope (P ¼ , OR ¼ 3.9, 95% CI ¼ ), use of NES (P ¼ 0.043, OR ¼ 1.6, 95% CI ¼ ) and use of digoxin (P ¼ 0.005, OR ¼ 1.86, 95% CI ). In addition, the development of ARF during hospitalization was also significantly associated with 60-day mortality. This was true irrespective whether ARF was defined as an increase of SCr 0.3 mg/dl (P ¼ 0.001, OR ¼ 2.27, 95% CI ¼ ¼ ) or 0.5 mg/dl (P ¼ , OR ¼ 3.10, 95% CI ¼ ). While both definitions of ARF were associated with increased mortality on univariate analysis, it should be emphasized that NES itself was not associated with the development of ARF, as we

4 Mortality in CHF with nesiritide 147 Table 2. Univariate analysis of risk factors at admission associated with mortality in all CHF patients Survivors (n ¼ 586) Non-surivors (n ¼ 96) Age Sex (M) 279 (47%) 44 (46%) Race (Caucasian) 573 (97%) 95 (98%) Na (meq/l) SCr (mg/dl) BUN (mg/dl) BUN/SCr GFR (ml/min) Hct BNP (pg/ml) ARF (SCr 0.3 mg/dl) 174 (30%) 47 (49%) ARF (SCr 0.5 mg/dl) 75 (13%) 30 (31%) b-blocker 393 (67%) 46 (48%) ACEI/A2RB 362 (62%) 45 (47%) Diuretic 571 (97%) 92 (96%) Inotrope 37 (6%) 20 (21%) NES 145 (25%) 33 (34%) Nitrates 397 (68%) 61(63%) Digoxin 210 (36%) 49 (51%) CAD 343 (58%) 48 (50%) COPD 229(39%) 46 (47%) Atrial fibrillation 264 (45%) 46 (48%) Hypertension 263 (45%) 53 (55%) Diabetes 215 (37%) 32 (33%) Values are expressed as mean SD or number (percent). ORs and 95% CI are presented only for discrete binary variables. Conversion factors for SI units: SCr: mg/dl 88.4 ¼ mmol/l; BUN: mg/dl ¼ mmol/l. All variables, except for the development of ARF, were determined on the day of admission. A2RB, angiotensin-2 receptor blocker; ACEI, angiotensin-converting enzyme inhibitor; ARF, acute renal failure; BNP, brain natriuretic peptide; BUN/SCr, blood urea nitrogen to creatinine ratio; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; GFR, glomerular filtration rate; Hct, haematocrit; NES nesiritide SCr, serum creatinine. previously showed [5], and confirmed in this study (see above). These predictors of mortality fall into several categories, which are in accordance with previous studies [9,11 16]. In addition to NES usage, these risk categories include: increased age; severity of CHF (increased serum BNP, use of an inotrope, use of digoxin, increased BUN/SCr ratio); pre-existing renal insufficiency (increased SCr, increased BUN, decreased estimated GFR); and the development of ARF. We speculate that the association between increased mortality and the lack of use of ACEI/A2RB and b-adrenergic blockers is a reflection of the severity of CHF and previous intolerance to these agents. Absence of NES as an independent risk factor for mortality of CHF patients Our univariate data are consistent with those of the recent meta-analysis showing an association between NES usage and increased mortality. However, a major limitation of the meta-analysis was a lack of statistical adjustment for baseline differences between the NES and NO NES groups. To address this issue, we performed logistic regression analysis with forward variable selection. As shown in Table 3, NES usage did not survive as an independent predictor of mortality. The independent predictors of mortality were (in descending order of coefficient of determination): development of ARF (SCr 0.5 mg/dl), lack of use of a b-adrenergic blocker, increased BUN, use of digoxin and increased serum BNP. One potential reason for the failure of NES to emerge as an independent risk factor for mortality is that patients administered NES may have had more severe illness, and that NES usage did not confer additional risk beyond its being a marker of more severe underlying illness. Consistent with this possibility, patients receiving NES differed in all of the following characteristics, indicative of more severe CHF and illness: decreased serum Na, increased SCr, increased BUN, increased BUN/SCr ratio, decreased estimated GFR, decreased haematocrit, increased serum BNP, increased use of inotropes, increased prevalence of CAD and increased prevalence of diabetes (Table 4). Risk factors associated with increased mortality in the presence and absence of NES Given that NES emerged as a univariate predictor of mortality in both the recent meta-analysis [3] and our retrospective analysis (Table 1), we used a second strategy to assess the role of NES as an independent

5 148 J. Iglesias et al. Table 3. Multivariate analysis of risk factors on admission associated with mortality Regression coefficient SE Chi-square Coefficient of determination P-value OR 95% CI Entire cohort ARF (SCr 0.5) b-blocker BUN (admission) Digoxin BNP NES ARF (SCr 0.5) Digoxin BNP ACEI/A2RB NO NES Inotrope b-blocker BUN/SCr (admission) A2RB, angiotensin-2 receptor blocker; ACEI, angiotensin-converting enzyme inhibitor; ARF, acute renal failure; BNP, brain natriuretic peptide; BUN/SCr, blood urea nitrogen to creatinine ratio. All variables, except for the development of ARF, were determined on the day of admission. Table 4. Baseline characteristics of CHF patients grouped according to NES use NES (n ¼ 178) No NES (n ¼ 504) Age Sex (M) 96 (54%) 227 (45%) Race 177 (99%) 494 (97%) (Caucasian) Na (meq/l) 136 4, SCr (mg/dl) BUN (mg/dl) BUN/SCr GFR Hct BNP (pg/ml) b-blocker 115 (64%) 324 (64%) ACEI/A2RB 109 (62%) 298 (59%) Diuretics 168 (95%) 495 (98%) Inotrope 42 (24%) 15 (3%) Nitrates 128 (72%) 330 (65%) Digoxin 65 (37%) 194 (38%) CAD 116 (66%) 275 (54%) COPD 67 (38%) 208 (41%) Atrial 73 (41%) 237 (47%) fibrillation Hypertension 77 (44%) 239 (47%) Diabetes 89 (50%) 158 (31%) Values are expressed as mean SD or number (percent). ORs and 95% CI are presented only for discrete binary variables. Conversion factors for SI units: SCr: mg/dl 88.4 ¼ mmol/l; BUN: mg/dl ¼ mmol/l. All variables were determined on the day of admission. A2RB, angiotensin-2 receptor blocker; ACEI, angiotensinconverting enzyme inhibitor; BNP, brain natriuretic peptide; BUN/SCr, blood urea nitrogen to creatinine ratio; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; GFR, glomerular filtration rate; Hct, haematocrit; SCr, serum creatinine. predictor of mortality. To determine whether NES may have had an independent adverse effect on our patients, we compared the risk factors associated with 60-day mortality among those patients who received NES (n ¼ 178) vs those who did not receive NES (n ¼ 504). For patients who received NES, by univariate analysis, the following admission characteristics were significantly associated with mortality (Table 5): increased BUN (P ¼ 0.008), increased BUN/SCr ratio (P ¼ 0.003), increased serum BNP (P ¼ 0.004), lack of use of ACEI/A2RB (P ¼ 0.012, OR ¼ 0.38, 95% CI ¼ ), use of an inotrope (P ¼ 0.001, OR ¼ 3.6, 95% CI ¼ ) and use of digoxin (P ¼ , OR ¼ 5.50, 95% CI ¼ ). In addition, the occurrence of ARF during hospitalization emerged as a predictor of mortality, irrespective of whether ARF was defined as an increase of SCr 0.3 mg/dl (P ¼ , OR ¼ 4.0, 95% CI ¼ ) or 0.5 mg/dl (P ¼ , OR ¼ 6.3, 95% CI ¼ ). Of these, the independent predictors, by multivariate analysis, were (in descending order of coefficient of determination): development of ARF (SCr 0.5 mg/dl), use of digoxin, increased serum BNP and lack of use of ACEI/ARB (Table 3). For CHF patients who not receive NES, by univariate analysis, the following admission characteristics were significantly associated with mortality (Table 6): older age (P ¼ 0.016), increased BUN (P ¼ 0.002), increased BUN/SCr ratio (P ¼ 0.005), increased serum BNP (P ¼ 0.015), lack of use of a b-adrenergic blocker (P ¼ 0.001, OR ¼ 0.42, 95% CI ¼ ), use of an inotrope (P ¼ 0.013, OR ¼ 3.71, 95% CI ¼ ) and absence of CAD

6 Mortality in CHF with nesiritide 149 Table 5. Univariate analysis of risk factors at admission associated with mortality in CHF patients receiving NES Survivors (n ¼ 145) Non-survivors (n ¼ 33) Age Sex (M) 80 (55%) 16 (49%) Race (Caucasian) 144 (99%) 33 (100%) Na (meq/l) SCr (mg/dl) BUN (mg/dl) BUN/SCr GFR Hct BNP (pg/ml) ARF (SCr 0.3 mg/dl) 44 (30%) 21 (63%) ARF (SCr 0.5 mg/dl) 17 (11%) 15 (45%) b-blocker 98 (67%) 17 (52%) ACEI/A2RB 95 (66%) 14 (42%) Diuretics 136 (94%) 32 (97%) Inotrope 27 (19%) 15 (45%) Nitrates 103 (71%) 25 (76%) Digoxin 42 (29%) 23 (70%) CAD 95 (66%) 21 (64%) COPD 51 (51%) 16 (49%) Atrial fibrillation 63 (44%) 10 (30%) Hypertension 60 (42%) 17 (52%) Diabetes 72 (50%) 17 (52%) Values are expressed as mean SD or number (percent). OR and 95% CI are presented only for discrete binary variables. Conversion factors for SI units: SCr: mg/dl 88.4 ¼ mmol/l; BUN: mg/dl ¼ mmol/l. All variables, except for the development of ARF, were determined on the day of admission. A2RB, angiotensin-2 receptor blocker; ACEI, angiotensin-converting enzyme inhibitor; ARF, acute renal failure; BNP, brain natriuretic peptide; BUN/SCr, blood urea nitrogen to creatinine ratio; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; GFR, glomerular filtration rate; Hct, haematocrit; SCr, serum creatinine. (P ¼ 0.046, OR ¼ 0.58, 95% CI ¼ ). In addition, the occurrence of ARF during hospitalization emerged as a predictor of mortality, but only when ARF was more strictly defined as an increase of SCr 0.5 mg/dl (P ¼ 0.023, OR ¼ 2.10, 95% CI ¼ ). Of these, the independent predictors, by multivariate analysis, were (in descending order of coefficient of determination): use of an inotrope, lack of use of a b-adrenergic blocker and an increased BUN/ SCr ratio (Table 3). The most notable difference between CHF patients receiving vs not receiving NES is that the development of ARF during hospitalization was an independent risk factor for mortality only among patients receiving NES. All other risk factors, for both patients receiving NES and those not receiving NES, appear to reflect increasing severity of CHF. Potential interaction between use of NES and occurrence of ARF as risk factors for mortality in CHF patients To assess the potential interaction between NES and ARF in CHF patients, we stratified patients who developed ARF during their hospitalization according to NES usage, and determined the effect of NES on mortality among these CHF patients with ARF. Out of our cohort of 682 patients, 221 (32%) developed ARF, as defined by a rise in SCr 0.3 mg/dl, while 105 (15%) developed ARF, as more strictly defined by a rise in SCr 0.5 mg/dl. As shown in Table 7, among patients who developed ARF, irrespective of definition, NES usage was significantly associated with 60-day mortality. The effect of NES on mortality in the presence of ARF was more apparent with the stricter definition of ARF (SCr 0.5 mg/dl), with mortality observed in 47% of patients receiving NES vs 21% of those not receiving NES (P ¼ 0.006, OR ¼ 3.40, 95% CI ¼ Pearson s chi-squared test). As a stronger association between NES and mortality is seen with a more strict definition of ARF, we limited our further analyses to ARF defined by an increase of SCr 0.5 mg/dl. To address the question whether NES is an independent risk factor for mortality among CHF patients developing ARF, we first performed a univariate analysis of all risk factors associated with mortality within the subset of CHF patients who developed ARF, irrespective of NES use. The following risk factors were associated with an increased risk of mortality (Table 8): increased BUN (P ¼ 0.022), increased BUN/SCr ratio (P ¼ 0.031) and use of NES (P ¼ 0.006, OR ¼ 3.40, 95% CI ¼ ). Remarkably, on multivariate analysis (Table 9), only

7 150 J. Iglesias et al. Table 6. Univariate analysis of risk factors at admission associated with mortality in CHF patients not receiving NES Survivors (n ¼ 441) Non-surivors (n ¼ 63) Age Sex (M) 199 (45%) 28 (44) Race (Caucasian) 433 (98%) 62 (98%) Na (meq/l) SCr (mg/dl) BUN (mg/dl) BUN/SCr GFR Hct BNP (pg/ml) ARF (SCr 0.3 mg/dl) 130 (29%) 26 (41%) ARF (SCr 0.5 mg/dl) 58 (13%) 15 (24%) b-blocker 295 (67%) 29 (46%) ACEI/A2RB 267 (60%) 31 (49%) Diuretics 438 (99%) 60 (95%) Inotrope 10 (2%) 5 (8%) Nitrates 294 (67%) 36 (57%) Digoxin 168 (38%) 26 (41%) CAD 248 (56%) 27 (43%) COPD 178 (40%) 30 (47%) Atrial fibrillation 201 (45%) 36 (57%) Hypertension 203 (46%) 36 (57%) Diabetes 143 (32%) 15 (24%) A2RB, angiotensin-2 receptor blocker; ACEI, angiotensin-converting enzyme inhibitor; ARF, acute renal failure; BNP, brain natriuretic peptide; BUN/SCr, blood urea nitrogen to creatinine ratio; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; GFR, glomerular filtration rate; Hct, haematocrit; SCr, serum creatinine. Values are expressed as mean SD or number (percent). ORs and 95% CI are presented only for discrete binary variables. Conversion factors for SI units: SCr: mg/dl 88.4 ¼ mmol/l; BUN: mg/dl ¼ mmol/l. All variables, except for the development of ARF, were determined on the day of admission. Table 7. NES usage and survival among CHF patients who developed ARF during hospitalization ARF (rise in SCr 0.3 mg/dl) (n ¼ 221) Survivor Nonsurvivors No NES (%) 130 (83) 26 (17) (n ¼ 156) NES(%) (n ¼ 65) 44 (68) 21 (32) ARF (rise in SCr 0.5 mg/dl) (n ¼ 105) Survivor Nonsurvivors No NES (%) 58 (79) 15 (21) (n ¼ 73) NES (%) (n ¼ 32) 17 (53) 15 (47) P OR 95% Ci NES usage remained as an independent predictor of mortality (P ¼ 0.006, OR ¼ 3.73, 95% CI ¼ ). The coefficient of determination for NES is 0.07, indicating that NES usage accounts for 7% of the variability in mortality among CHF patients who developed ARF. To confirm the interaction of NES usage and mortality in CHF patients who developed ARF, we looked for a dose-dependent effect of the rise of SCr on mortality in patients who received NES vs those who did not. Patients were stratified into 4 groups, based upon the peak rise in their SCr: <0.3 mg/dl (n ¼ 348 NO NES, n ¼ 113 NES), mg/dl (n ¼ 62 NO NES, n ¼ 21 NES), mg/dl (n ¼ 35 NO NES, n ¼ 18 NES) and 0.70 mg/dl (n ¼ 59 NO NES, n ¼ 26 NES). As shown in Figure 1, for patients who received NES, the percent mortality increased progressively from 11% to 43% from the group with the lowest to greatest peak rise of SCr (P ¼ 0.001, Pearson s chi-squared test). In contrast, for patients who did not receive NES, percent mortality did not significantly vary according to peak rise of SCr among the four groups (P ¼ 0.11, Pearson s chi-squared test). An identical result was obtained when we repeated the analysis by stratifying the patients who developed ARF into tertiles of equal numbers of patients (NES, P ¼ 0.001: NO NES, P ¼ 0.052; Pearson s chisquared test). This dose-dependent effect of ARF on mortality, occurring only among CHF patients who received NES, provides strong evidence for a deleterious interaction between NES usage and the development of ARF as a predictor of mortality. Moreover, among patients who developed ARF in the setting of NES therapy, the duration of therapy was significantly longer for non-survivors vs survivors ( vs days, P ¼ 0.035, Wilcoxon rank-sum test).

8 Mortality in CHF with nesiritide 151 Discussion An important finding of our study is that, although the use of NES per se is not independently associated with an increased risk for mortality among CHF patients, the development of ARF in association with NES administration appears to confer an increased risk for mortality. The deleterious interaction between NES use and ARF was evident in several ways. First, a comparison of the independent risk factors for mortality between patients who received NES vs those who did not demonstrate that the development of ARF (defined as an increase in increase in SCr 0.5 mg/dl) was an independent predictor of mortality only among Table 8. Univariate analysis of risk factors at admission associated with mortality in CHF patients developing ARF (rise in SCr 0.5 mg/dl) Survivors (n ¼ 75) Non-surivors (n ¼ 30) Age Sex (M) 36 (48%) 10 (30%) Race 71 (95%) 30 (100%) (Caucasian) Na (meq/l) SCr (mg/dl) BUN (mg/dl) BUN/SCr GFR Hct BNP (pg/ml) b-blocker 50 (66%) 15 (50%) ACEI/A2RB 43 (57%) 12 (40%) Diuretics 72 (96%) 29 (97%) Inotrope 11 (15%) 8 (27%) NES 17 (22%) 15 (50%) Nitrates 55 (73%) 21 (70%) Digoxin 30 (40%) 14 (47%) CAD 35(47%) 13 (43%) COPD 40 (53%) 13 (43%) Atrial 30 (40%) 16 (53%) fibrillation Hypertension 28 (37%) 13 (43%) Diabetes 20 (27%) 10 (30%) Values are expressed as mean SD or number (percent). ORs and 95% CI are presented only for discrete binary variables. Conversion factors for SI units: SCr: mg/dl 88.4 ¼ mmol/l; BUN: mg/ dl ¼ mmol/l. All variables were determined on the day of admission. A2RB, angiotensin-2 receptor blocker; ACEI, angiotensinconverting enzyme inhibitor; BNP, brain natriuretic peptide; BUN/ SCr, blood urea nitrogen to creatinine ratio; CAD, coronary artery disease; COPD, chronic obstructive pulmonary disease; GFR, glomerular filtration rate; Hct, haematocrit; SCr, serum creatinine. patients who received NES (Table 3). Second, upon multivariate analysis of risk factors for mortality among those CHF patients who developed ARF, NES usage emerged as the only independent predictor of mortality (Table 9). Remarkably, patients receiving NES were 3.4-times more likely to die (95% CI ¼ ), with NES usage accounting for 7% of the variability in mortality among these patients. Third, a dose-dependent effect between the magnitude of the rise in SCr and mortality existed for CHF patients who received NES (Figure 1). For patients receiving NES, mortality increased progressively from 11%, for patients with a minimal change in SCr (<0.3 mg/dl), to 43%, for those with the greatest rise in SCr (0.7 mg/dl). In contrast, for patients who did not receive NES, a dose response relationship between elevation in SCr and mortality did not exist. Finally, the duration of treatment with NES was significantly longer for patients who died. This was true both for the population as a whole ( vs days) and for the limited population who developed ARF ( vs days). Among CHF patients in general, NES usage was not an independent predictor of mortality. Although by univariate analysis NES was associated with increased mortality (Tables 1 and 2), it did not survive on multivariate analysis (Table 3). Importantly, the lack of an independent association between NES usage and mortality among all CHF patients occurred despite the fact that patients given NES tended to be sicker and to have more severe CHF (Table 4). Thus, even in the face of overall poorer health and more severe CHF, NES still did not predict mortality. These data suggest that NES per se does not predispose to mortality, but only in the setting of ARF. Independent predictors of mortality for CHF patients in general, as identified by our analysis, were consistent with those of previous studies [9,11 16]. These risk factors were also similar to those identified in the subgroups of CHF patients who received or did not receive NES, and fell into several major risk categories (Table 3). These categories included: severity of CHF (increased serum BNP, use of an inotrope, use of digoxin, increased BUN/SCr ratio), pre-existing renal insufficiency (increased BUN) and lack of use of a b-adrenergic blocker or an ACEI/ARB. It is important to emphasize that the observed interaction between NES and the development of ARF in predicting mortality appears to be direct. Our data do not support the notion of an indirect effect, that is, that NES increases the risk for ARF, and ARF then Table 9. Multivariate analysis of risk factors on admission associated with mortality for CHF patients developing ARF (rise in SCr 0.5 mg/ dl) Regression coefficient SE Chi-square Coefficient of determination P-value OR 95% CI NES

9 152 J. Iglesias et al. % Mortality NONES mortality (%) NES mortality (%) NONES P= 0.11 NES P=0.001 < >0.69 Rise in SCr Fig. 1. Mortality stratified by rise of serum creatinine (Scr) in patients receiving vs not receiving NES. Patients were stratified into 4 groups, based upon the peak rise in their SCr, and the percent mortality within each group was determined for patients who received or did not receive NES Mortality increased in a dosedependent manner for patients who received NES (P ¼ 0.001, Pearson s chi-squared test), while no significant relationship existed between rise of Scr and mortality for patients not receiving NES (P ¼ 0.11, Pearson s chi-squared test). increases the risk of mortality independently of NES. Thus, in both our previous study of 219 CHF patients and our current study of 682 CHF patients, NES use was not associated with the development of ARF, even by univariate analysis. The lack of an association between NES and the development of ARF occurred with two distinct definitions of ARF, both based on a rise in SCr from admission, either 0.3 or 0.5 mg/dl. The failure of NES to predict ARF, while at the same time predicting mortality, makes it extremely unlikely that the NES-associated increase of mortality we observed is mediated by NES-induced ARF. Thus, while NES did not confer an increased risk for ARF, it did confer an increased risk for mortality, but only among patients who develop ARF. The mechanism by which NES and ARF interact to increase the risk for mortality remains uncertain. Possible mechanisms by which the uraemic environment could increase the toxicity of NES include: altered pharmacokinetics or metabolism, post-receptor signalling abnormalities, altered cellular expression or function of NES receptors, ARF-mediated changes in cardiopulmonary function, and/or interaction of NES with an accumulated uraemic toxin. The fact that BNP and NES are cleared predominantly by kidney, so that levels of BNP increase with declining renal function, in both normal and CHF patients, lends support to these putative mechanisms. There are several potential limitations to our present study. First, the retrospective nature of our study limited our ability to determine and adjust for differences in all baseline characteristics and comorbidities. Second, we lacked sufficient data to adjust for unmeasured potential baseline confounders, such as degree of proteinuria, levels of C-reactive protein, and other factors that may be important to the risk of developing ARF and mortality. Third, we lacked sufficient data to adjust for clinical variables indicative of therapeutic response, such as diuretic dose or urine output, which likely affected the clinical decision whether or not to use NES. Fourth, we lacked sufficient data to determine whether the interaction of NES and ARF predisposed to death from specific cardiovascular causes vs unanticipated aetiologies from non-cardiovascular causes. Fifth, our definition of ARF, while in accord with that used by multiple other studies, does not permit us to differentiate renal from pre-renal causes of ARF, and therefore we cannot determine whether mortality depended upon particular aetiologies or types of renal failure. Finally, The Diagnosis of CHF was based solely on clinical criteria, and did not include direct measurement of ventricular function by echocardiography or other means. To address the possibility of selection bias that is, that the cohort of 682 patients for whom we had complete admission data were somehow not representative of the complete population of 1407 patients we adjusted for baseline differences in all 1407 patients, using only those admission variables for which we had complete information. These were: age, sex, race, medication usage b-blocker, ACEI/A2RB, diuretic, inotrope, NES, nitrates and digoxin), comorbidities (CAD, COPD, atrial fibrillation, hypertension and diabetes). The major missing admission data for the 702 excluded patients were indices of renal function. Upon univariate analysis, the following variables were significantly correlated with mortality: age (P ¼ 0.017), lack of use of b-blocker (P ¼ ), lack of use of ACEI/A2RB (P ¼ 0.003), use of inotrope (P ¼ ), use of nesiritide (P ¼ 0.008), use of digoxin (P ¼ ) and presence of hypertension (P ¼ 0.024). With the exception of hypertension, these are in full accord with the data shown in Table 2. By multivariate analysis, as in our cohort of 682 patients, NES failed to survive as an independent predictor. Only the following were independent predictors of mortality (in descending order of coefficient of determination): use of inotrope (0.022, P ¼ ), lack of use of b-blocker (0.013, P ¼ ), presence of hypertension 0.009, P ¼ ), use of digoxin (0.008, P ¼ ), lack of use of ACEI/A2RB (0.0007, P ¼ 0.001), and age (0.003, P ¼ 0.025). With the exception of baseline indices suggestive of preexisting renal insufficiency, for which admission data were lacking, these results are in general agreement with those of Table 3. Moreover, direct comparison of the excluded 725 patients with our cohort of 682 patients showed that they were demographically well-matched in terms of age, sex, and race (P > 0.83 for all three variables). In summary, by retrospective analysis, we report that the administration of NES to patients with CHF may be associated with an increased risk of death. This risk appears to be limited to patients who develop ARF, and does not seem to apply to CHF patients in

10 Mortality in CHF with nesiritide 153 general. A major strength of our study is that our study population is derived from a large community hospital, and is therefore clinically and demographically similar to CHF patients within the general population. The mechanism for this adverse interaction between NES administration and the development of ARF remains unclear. Based on our results, we recommend extreme caution in the care of CHF patients who develop ARF in the setting of NES administration. Prospective studies are needed to evaluate the effects of NES in this potentially high-risk group of patients with both cardiac and renal dysfunction. Acknowledgements. This work was supported by NIH grants DK59793 and by a Renal Innovations Program Award from Genzyme, Inc. Community Medical Center, Toms River, NJ, USA was the study institution. Conflict of interest statement. J.I.I has received speaker honoraria from Scios, Inc and Ortho Biotech, and J.S.L. has received research funding from Genzyme, inc. References 1. Abraham WT, Lowes BD, Ferguson DA et al. Systemic hemodynamic, neurohormonal, and renal effects of a steadystate infusion of human brain natriuretic peptide in patients with hemodynamically decompensated heart failure. J Cardiac Fail 1998; 4: Colucci WS, Elkayam U, Horton DP et al. Intravenous nesiritide, a natriuretic peptide, in the treatment of decompensated congestive heart failure. Nesiritide Study Group. N Engl J of Med 2000; 343: Sackner-Bernstein JD, Kowalski M, Fox M et al. Short-term risk of death after treatment with nesiritide for decompensated heart failure: a pooled analysis of randomized controlled trials. JAMA 2005; 293: Sackner-Bernstein JD, Skopicki HA, Aaronson KD. Risk of worsening renal function with nesiritide in patients with acutely decompensated heart failure. Circulation 2005; 111: Iglesias J, Hom D, Antoniotti M et al. Predictors of worsening renal function in adult patients with congestive heart failure receiving recombinant human B-type brain natriuretic peptide (nesiritide). Nephrol Dial Transplant 2006; 21: Centers for Medicare and Medicaid Services National Heart Failure Quality Improvement Project. United States Department Health and Human Services: 2001; US Health Care Financing Administration. Diagnosis-related Groups Definitions Manual. 3M Health Information Services, Wallingford CT: Butler J, Forman DE, Abraham WT et al. Relationship between heart failure treatment and development of worsening renal function among hospitalized patients. Am Heart J 2004; 147: Pitt B, Segal R, Martinez FA et al. Randomised trial of losartan versus captopril in patients over 65 with heart failure (Evaluation of Losartan in the Elderly Study, ELITE). Lancet 1997; 349: Rule AD, Larson TS, Bergstralh EJ et al. Using serum creatinine to estimate glomerular filtration rate: accuracy in good health and in chronic kidney disease. Ann Intern Med 2004; 141: Fonarow GC, Corday E. Overview of acutely decompensated congestive heart failure (ADHF): a report from the ADHERE registry. Heart Fail Rev 2004; 9: Forman DE, Butler J, Wang Y et al. Incidence, predictors at admission, and impact of worsening renal function among patients hospitalized with heart failure. J Am Coll Cardiol 2004; 43: Gottdiener JS, McClelland RL, Marshall R et al. Outcome of congestive heart failure in elderly persons: influence of left ventricular systolic function. The Cardiovascular Health Study. Ann of Intern Med 2002; 137: Gottlieb SS, Abraham W, Butler J et al. The prognostic importance of different definitions of worsening renal function in congestive heart failure. J Card Fail 2002; 8: Heywood JT. The cardiorenal syndrome: lessons from the ADHERE database and treatment options. Heart Fail Rev 2004; 9: Kittleson M, Hurwitz S, Shah MR et al. Development of circulatory-renal limitations to angiotensin-converting enzyme inhibitors identifies patients with severe heart failure and early mortality. J Am Coll Cardiol 2003; 41: Received for publication: Accepted in revised form: