Improvement in renal function during the treatment of acute decompensated heart failure: Relationship with markers of renal tubular injury and prognostic importance

Abstract

Background:

Improvement in renal function (IRF) in acute decompensated heart failure (ADHF) is associated with adverse outcomes. The mechanisms driving this paradox remain undefined.

Methods:

Using the Renal Optimization Strategies Evaluation–Acute Heart Failure (ROSE-AHF) study, 277 patients were grouped according to renal function, with IRF defined by a ≥20% increase (N = 75), worsening renal function (WRF) a ≥20% decline (N = 53), and stable renal function (SRF) a <20% change (N = 149) in estimated glomerular filtration rate (eGFR) between baseline and 72 h. Three well-validated renal tubular injury markers, neutrophil gelatinase-associated lipocalin (NGAL), N-acetyl-β-d-glucosaminidase (NAG), and kidney injury molecule 1 (KIM-1), were evaluated at baseline and 72 h. Patients were also classified by the pattern of change in these markers.

Results:

Patients with IRF had the lowest admission eGFR (IRF: 37 [28, 51], WRF: 43 [35, 55] and SRF: 43 [32, 55] mL/min/1.73 m²; P_trend = 0.032) but greater cumulative urine output (IRF: 8780 [7025, 11208], WRF: 7860 [5555, 9765], and SRF: 8150 [6325, 10456] mL; P_trend = 0.024) and weight loss (IRF: −9.0 [−12.4, −5.3], WRF: −5.1 [−8.1, −1.3], and SRF: −7.1 [−11.9, −3.2] lb.; P_trend < 0.001) despite similar diuretic doses (P_trend = 0.16). There were no differences in the relative change in NGAL, NAG, or KIM-1 between renal function groups (P_trend > 0.19 for all). Patients with IRF had worse survival than patients with SRF (27% vs. 54%, HR 1.98, [1.10–3.58], P = 0.024).

Conclusions:

IRF during decongestive therapy for ADHF was not associated with improved markers of renal tubular injury and was associated with worsened survival, likely driven by the presence of greater underlying cardiorenal dysfunction and more severe congestion.

Introduction

Renal dysfunction is a common clinical occurrence when treating acute decompensated heart failure (ADHF).¹ The significance of worsening renal function (WRF) in ADHF varies according to the clinical context. In the setting of positive therapeutic interventions that achieve decongestion, WRF is associated with neutral to improved survival.^2–5 Although improvement in renal function (IRF) would intuitively seem desirable in ADHF, IRF has been independently associated with adverse outcomes and mortality compared with either stable renal function (SRF) or WRF.^6–8

The mechanisms behind the paradoxical association between IRF and adverse outcomes are poorly understood. IRF may be a sign of inadequate decongestion in some patients or a marker of disease severity in others, such that WRF and IRF may represent the same continuum of renal dysfunction depending on the perspective of the baseline kidney function.^{6, 8, 9} The assessment of well-validated urinary markers of renal tubular injury in addition to estimated glomerular filtration rate (eGFR) measurements may provide insight and may distinguish prognostic differences among patients with IRF.¹⁰

While renal tubular injury markers have not provided prognostic or mechanistic distinction of WRF during ADHF^{5, 11, 12}, renal tubular injury in the context of IRF during ADHF has not been studied. Therefore, we sought to investigate the role of renal tubular injury as a potential mechanistic driver of adverse outcomes in ADHF patients with IRF, who are an at-risk and relatively under-studied ADHF sub-group. We analyzed serial measures of eGFR and tubular injury markers across the ADHF admission in the Renal Optimization Strategies Evaluation–Acute Heart Failure (ROSE-AHF) clinical trial to understand the degree of tubular injury prior to the development of IRF, the pattern of change of injury markers when IRF occurs, and whether renal tubular injury markers can inform clinical outcomes in ADHF with IRF.

Methods

Patient Population

We analyzed data from ROSE-AHF and related research materials via the Biological Specimen and Data Repository Information Coordinating Center (BioLINCC) of the National Heart, Lung, and Blood Institute.¹³ All data, analytical methods, and study materials are publicly available to any researcher for the purpose of reproducing the results and can be accessed through the BioLINCC. Institutional review boards at each participating clinical site approved the study protocol, and all patients provided written informed consent prior to randomization. As this analysis was performed on de-identified data, approval by the institutional review board was not required. The ROSE-AHF trial design and results have been previously published.^{14, 15} Briefly, ROSE-AHF compared the effects of low-dose dopamine, low-dose nesiritide, or placebo in augmenting diuresis and preserving renal function in 360 patients with ADHF and renal dysfunction from randomization to 72 h. Baseline renal dysfunction was defined as an admission eGFR of 15-60 mL/min/1.73 m² as calculated by the Modification of Diet in Renal Disease (MDRD) equation. The median eGFR in the trial was 42 [interquartile range (IQR) 31, 54] mL/min/1.73 m². Patients received intravenous loop diuresis with a recommended total daily dose of 2.5x the total daily outpatient furosemide equivalent loop diuretic dose. Spot urine collections to measure renal tubular injury biomarkers were completed daily for the first 72 h. Compared to placebo, dopamine and nesiritide did not improve diuresis or preserve renal function when added to intravenous diuretics. No significant differences in the change in markers of renal tubular injury were observed among the three treatment arms.¹¹

Of the 360 patients enrolled in ROSE-AHF, we analyzed 277 (77%) (Figure 1). Subjects were excluded if: (1) eGFR calculations at baseline, 24 h, 48 h, or 72 h were not available; (2) measurements of urinary markers of renal tubular injury neutrophil gelatinase-associated lipocalin (NGAL), N-acetyl-β-d-glucosaminidase (NAG), and kidney injury molecule 1 (KIM-1) were not measured at baseline or 72 h; or (3) an intravenous diuretic was not administered. Slight differences in the number of patients from prior analyses of the ROSE-AHF trial are the result of these different exclusion criteria.

Figure 1. CONSORT (Consolidated Standards of Reporting Trials) diagram of patient selection into the study cohort.

eGFR indicates glomerular filtration rate; KIM-1, kidney injury molecule 1; NAG, N-acetyl-β-d-glucosaminidase; NGAL, neutrophil gelatinase-associated lipocalin; and ROSE-AHF, Renal Optimization Strategies Evaluation–Acute Heart Failure.

Measurement of Biomarkers

Plasma creatinine, cystatin C, and N-terminal pro-B-type natriuretic peptide (NT-proBNP) were measured at a core laboratory (Heart Failure Clinical Research Network Core Biomarker Laboratory, University of Vermont). Urinary KIM-1 and NGAL were measured with microbead-based assays at the Brigham and Women’s Hospital, where urine samples were incubated with microbeads coupled with NGAL (Enzo Lifesciences) and KIM-1 (R&D Systems) antibodies and quantified with the Bio-Plex 200 system (Bio-Rad). Urinary NAG was measured with the NAG kit per the manufacturer’s (Roche Diagnostics) instructions. All urine renal tubular injury biomarkers were normalized to urine creatinine due to urinary dilution resulting from intravenous diuresis.

Study Definitions

The ROSE-AHF trial employed the MDRD equation to calculate eGFR. Since the completion of the ROSE-AHF trial, the 2021 non-race-based Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) estimation equation has become the recommended method to calculate eGFR.¹⁶ We replicated all eGFR calculations with the 2021 CKD-EPI serum creatinine-based equation: no significant differences in eGFR were observed. Therefore, we used the MDRD eGFR equation throughout these analyses to be consistent with the parent study and trial database.

Consistent with prior analyses, IRF was defined as a ≥20% increase in the eGFR at any timepoint from baseline through 72 h, and WRF was defined as a ≥20% decrease in the eGFR at any timepoint between baseline and 72 h.^{2, 3, 6, 7, 11, 17–24} SRF was defined as a steady eGFR ± 19% from baseline throughout the 72-h intervention period. Improvement and worsening of each renal tubular injury marker were defined as a ≥20% reduction and ≥20% increase, respectively, in the measured urinary concentration from baseline to 72 h. Stable tubular injury markers met neither definition. For survival analyses, the percent change in the three renal tubular injury markers were averaged together and grouped as above based on the average percent change from baseline to 72 h.

Statistical Analyses

Continuous variables were described by the median and IQR or the mean ± standard deviation, and categorical variables were described with frequencies and percentages. Categorical variables were compared among the three groups with the Chi-squared test. Continuous variables were examined for normality with the Kolmogorov-Smirnov test, the Shapiro-Wilk test, and histogram plots. For normally distributed data, the three groups were compared with one-way analysis of variance (ANOVA). For non-normally distributed data, the Kruskal-Wallis test was used. Data were also analyzed for trends (referred to a P_trend) among groups of renal function trajectories (in order of worsened to stable to improved) with the Chi-squared trend test for categorical variables, linear regression for normally distributed continuous variables, and Spearman rho for non-normally distributed continuous variables. To evaluate the trends of biomarkers over time across the three groups of renal function, the percent changes of NGAL, NAG, and KIM-1 from baseline to 72 h were calculated. In addition, the average change in the three markers assessed collectively was calculated. The percent change was used due to the high variability within and among the groups across timepoints in the tubular injury markers. In addition, linear mixed models (two-level model with a random intercept by a variable identifying each patient) were used to assess the within-patient change of urinary injury biomarkers across groups. Fixed effects were time, groups, and their interaction term.

Survival was described with the Kaplan-Meier plot of groups based by renal function status, and differences in survival were compared with the Mantel-Cox log-rank test. Differences in survival were also analyzed across the three groups of renal function by linear trend with Cox regression (in order of observed best to worst survival, i.e., from SRF to WRF to IRF, as shown below). A multivariable Cox proportional hazards regression was performed to account for potential confounding by other variables on the associations of renal function categories with the risk of death. The following covariables were included: sex; race; history of ischemic heart disease; hypertension; type 2 diabetes mellitus; use of loop diuretics; thiazide diuretics; angiotensin-converting enzyme inhibitors or angiotensin receptor blockers; beta blockers; aldosterone antagonists; baseline measurements of left ventricular ejection fraction (LVEF); eGFR; blood urea nitrogen (BUN); serum sodium; serum chloride; NT-proBNP; NGAL; NAG; KIM-1; NT-proBNP at 72 h; change in weight at 72 h; and cumulative urine output at 72 h. BUN, NT-proBNP, NGAL, NAG, and KIM1 were log-transformed to generate an approximate normal distribution. Stepwise backward-elimination was used to select variables for inclusion in the regression model with an alpha-to-remove of 0.10. Survival analyses were also performed with groups categorized by changes in urinary renal tubular injury markers. We finally created groups of patients based on IRF or SRF and the biomarker change pattern: (1) IRF with improving markers; (2) SRF with stable markers; and (3) IRF with stable markers or SRF with improved markers.

Statistical analysis was performed with IBM SPSS Statistics version 28 (IBM Corp, Armonk, NY). Graphics were generated using GraphPad Prism version 9 (GraphPad Software, LLC, San Diego, CA). Statistical significance was defined as P < 0.05.

Results

Baseline Characteristics of the Study Subjects

Of the 277 patients included in our analysis, 75 patients (27%) had IRF, 53 patients (19%) had WRF, and 149 patients (54%) had SRF over the 72-h study intervention period. Baseline characteristics are presented in Table 1 for the entire included cohort and groups based on renal function status. Compared to patients with WRF or SRF, patients with IRF had worse baseline kidney function as measured by a lower eGFR (37 [28, 51] mL/min/1.73 m² vs. 43 [35, 55] mL/min/1.73 m² in WRF vs. 43 [32, 55] mL/min/1.73 m² in SRF; P_trend = 0.032), higher serum creatinine (1.80 [1.50, 2.24] mg/dL vs. 1.57 [1.30, 1.95] mg/dL in WRF vs. 1.70 [1.36, 2.00] in SRF; P_trend = 0.009), and higher BUN (44 [31, 61] mg/dL vs. 33 [24, 47] mg/dL in WRF vs. 37 [28, 51] mg/dL in SRF; P_trend = 0.013). Additionally, patients with IRF had lower systolic blood pressure, lower serum sodium and chloride concentrations, and lower urine sodium concentration than patients with WRF or SRF (P_trend < 0.05 for all). Patients with IRF appeared to have directionally higher NT-proBNP levels (6815 [3430, 12431] ng/L) compared to patients with WRF (4859 [1920, 9621]) ng/L) and SRF (4938 [2194–11016]) ng/L), though statistical significance was not met (P_trend = 0.06). No significant differences in physical examination findings of congestion or medication therapy were observed among the three groups.

Table 1.

Baseline characteristics of the study population and renal function cohort.

Characteristic	Study Cohort (N = 277)	WRF (N = 53)	SRF (N = 149)	IRF (N = 75)	P	P trend
Demographics
Age, y	70 (62, 79)	71 (62, 81)	70 (62, 79)	70 (61, 81)	0.81	0.90
Male Sex, %	74	62	78	73	0.09	0.24
White, %	77	83	76	73	0.43	0.22
Vital Signs
HR, bpm	74 (65, 84)	75 (65, 84)	74 (65, 85)	72 (64, 81)	0.88	0.62
SBP, mm Hg	114 (103, 126)	115 (105, 131)	114 (104, 128)	109 (99, 120)	0.04*	0.014*
DBP, mm Hg	64 (59, 71)	62 (60, 69)	65 (59, 74)	64 (60, 70)	0.63	0.90
Clinical Congestion
JVP ≥8 cm H2O, %	96	98	96	94	0.57	0.30
≥2+ Edema, %	71	76	70	71	0.76	0.60
Rales, %	54	56	54	54	0.97	0.87
S3, %	26	24	25	29	0.71	0.45
Fluid Overload^ by CXR, %	72	63	77	68	0.13	0.74
Dyspnea, %	95	98	95	95	0.56	0.41
Orthopnea, %	89	92	88	88	0.69	0.49
Clinical History
LVEF, %	33 (20, 50)	33 (25, 47)	30 (20, 51)	29 (19, 52)	0.46	0.22
LVEF <50%, %	71	76	71	69	0.71	0.44
Duration HF Diagnosis, y	4.3 (1.8, 8.2)	4.2 (1.9, 11.4)	4.0 (1.7, 7.7)	5.0 (2.4, 10.0)	0.33	0.49
HFH in Past Year	68	69	67	68	0.96	0.87
ICM, %	60	57	64	53	0.29	0.57
HTN, %	83	87	86	76	0.13	0.08
DM2, %	56	64	57	49	0.24	0.09
AF/AFL, %	59	62	54	64	0.32	0.70
ICD, %	46	53	42	47	0.40	0.59
Home Medications
Loop Diuretic, %	95	96	94	96	0.72	0.96
Thiazide, %	18	21	16	21	0.56	0.83
ACEi or ARB, %	50	53	51	44	0.53	0.30
β-Blocker, %	84	85	83	85	0.91	0.90
MRA, %	29	26	26	35	0.39	0.26
Hydralazine, %	20	23	21	15	0.45	0.24
Nitrate, %	25	32	25	20	0.30	0.12
Digoxin, %	26	26	22	33	0.16	0.28
Laboratory Data
eGFR, mL·min−1·1.73 m−2	41 (32, 53)	43 (35, 55)	43 (32, 55)	37 (28, 51)	0.05	0.032*
Serum Cr, mg/dL	1.70 (1.40, 2.10)	1.57 (1.30, 1.95)	1.70 (1.36, 2.00)	1.80 (1.50, 2.24)	0.033*	0.009*
Serum CysC, mg/L	1.71 (1.44, 2.15)	1.71 (1.44, 2.23)	1.73 (1.45, 2.14)	1.70 (1.41, 2.16)	0.92	0.70
BUN, mg/dL	37 (28, 52)	33 (24, 47)	37 (28, 51)	44 (31, 61)	0.047*	0.013*
Serum Na, mEq/L	138.2 ± 3.8	138.7 ± 2.8	138.5 ± 3.8	137.2 ± 4.3	0.07	0.03*
Serum Cl, mEq/L	100.3 ± 5.1	101.3 ± 4.5	100.4 ± 5.3	99.3 ± 4.9	0.09	0.022*
HCO3, mEq/L	27 (24, 30)	27 (24, 30)	28 (25, 31)	26 (24, 29)	0.08	0.55
NT-proBNP, ng/L	5307 (2345, 10509)	4859 (1920, 9621)	4938 (2194, 11016)	6815 (3430, 12431)	0.17	0.06
Hct, %	36 (32, 39)	35 (31, 39)	35 (32, 39)	37 (32, 40)	0.36	0.15
Ur Na, mmol/L	52 (33, 77)	59 (42, 83)	53 (35, 78)	49 (26, 68)	0.031*	0.008*

Data presented as median and interquartile range, except for Serum Na and Serum Cl, for which the median and standard deviation are shown. ACEi indicates angiotensin-converting enzyme inhibitor; AF, atrial fibrillation; AFL, atrial flutter; ARB, angiotensin receptor blocker; BUN, blood urea nitrogen; Cl, serum chloride; Cr, serum creatinine; CXR, chest x-ray; CysC, serum cystatin C; DBP, diastolic blood pressure; DM2, type 2 diabetes; eGFR, estimated glomerular filtration rate; HCO3, serum bicarbonate; Hct, hematocrit; HFH, heart failure hospitalization; HR, heart rate; HTN, hypertension; ICD, implantable cardiac defibrillator; ICM, ischemic cardiomyopathy; IRF, improvement in renal function; JVP, jugular venous pressure; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonist; Na, serum sodium; NT-proBNP, N-terminal pro-B-type natriuretic peptide; SBP, systolic blood pressure; SRF, stable renal function; Ur Na, urine sodium (pre-diuretic); and WRF, worsening renal function.

^{^}Fluid overload by CXR: presence of pulmonary vascular congestion or pleural effusion.

^*indicates statistical significance of P < 0.05. The P value refers to ANOVA and Kruskal-Wallis testing for parametric and non-parametric data, respectively. The P_trend value refers to Chi-squared trend test for categorical variables, linear regression for normally distributed continuous variables, and Spearman rho for non-normally distributed continuous variables, with the three groups organized in an ordinal fashion (WRF, SRF, and IRF).

Changes in Laboratory Studies and Markers of Congestion

After 72 h, patients with IRF had higher eGFR, lower serum creatinine, lower cystatin C, and lower BUN compared to patients with WRF and SRF (P_trend < 0.001 for all; Table 2). In the IRF cohort, median eGFR increased 12 [IQR 8, 16] mL/min/1.73 m² whereas the median change in eGFR was −10 [−14, −7] mL/min/1.73 m² among patients with WRF (P_trend < 0.001). At 72 h, patients with IRF had greater cumulative volumes of urine output (8780 [7025, 11208] mL vs. 7860 [5555, 9765] mL in WRF and 8150 (6325, 10456) mL in SRF; P_trend = 0.024), higher 24-h urinary sodium excretion (141 [86, 240] mmol vs. 78 [39, 174] mmol in WRF and 147 [78, 221] mmol in SRF; P_trend = 0.011), and greater weight loss (−9.0 [−12.4, −5.3] lb. vs. −5.1 [−8.1, −1.3] lb. in WRF and −7.1 [−11.9, −3.2] lb. in SRF; P_trend <0.001) despite similar 72-h cumulative intravenous loop diuretic doses (P_trend = 0.16; Table 3). Despite this greater diuresis and natriuresis, objective physical examination and reported subjective measures of congestion remained similar between the IRF and WRF cohorts, suggesting the IRF group had a greater total body hypervolemia at baseline. As with the baseline NT-proBNP levels, somewhat higher NT-proBNP levels were observed in the IRF cohort (4245 [1790, 8283] ng/L) at 72 h compared to the WRF group (3391 [1152, 5748] ng/L) and SRF group (3064 [1351, 8081]) ng/L), but statistical significance was not reached (P_trend = 0.07). There were no significant differences in the absolute change or percent change in NT-proBNP across the three groups (P_trend > 0.05 for both).

Table 2.

Laboratory data at 72 hours by renal function status.

Characteristic	WRF (N = 53)	SRF (N = 149)	IRF (N = 75)	P	P trend
eGFR, mL·min−1·1.73 m−2	33 (25, 40)	43 (32, 53)	49 (39, 64)	-	-
ΔeGFR, mL·min−1·1.73 m−2	−10 (−14, −7)	0 (−4, 2)	+12 (8, 16)	-	-
Cr, mg/dL	2.00 (1.70, 2.63)	1.70 (1.37, 2.09)	1.44 (1.20, 1.79)	<0.001*	<0.001*
ΔCr, mg/dL	0.50 (0.30, 0.67)	0 (−0.10, 0.15)	−0.36 (−0.58, −0.26)	<0.001*	<0.001*
CysC, mg/L	2.08 (1.65, 2.55)	1.86 (1.48. 2.33)	1.60 (1.29, 2.06)	<0.001*	<0.001*
ΔCysC, mg/L	0.34 (0.14, 0.52)	0.12 (−0.03, 0.25)	−0.08 (−0.36, 0.06)	<0.001*	<0.001*
BUN, mg/dL	48 (34, 70)	39 (31, 54)	36 (25, 52)	0.005*	0.001*
ΔBUN	12 (9, 19)	2 (−2, 7)	−6 (−12, −1)	<0.001*	<0.001*
ΔBUN/Cr	1.0 (−1.2, 4.4)	1.1 (−1.7, 3.8)	2.2 (−1.1, 5.3)	0.38	0.34
Na, mEq/L	137 (134, 139)	138 (135, 140)	138 (136, 141)	0.11	0.047*
ΔNa, mEq/L	−3 (−5, 0)	−1 (−3, 1)	0 (−2, 3)	<0.001*	<0.001*
Cl, mEq/L	98 (93, 101)	98 (94, 100)	98 (94, 100)	0.86	0.82
ΔCl, mEq/L	−4 (−5, −2)	−3 (−5, 0)	−2 (−4, 1)	0.003*	<0.001*
HCO3, mEq/L	28 (26, 30)	30 (27, 32)	30 (28, 33)	0.02*	0.01*
NT-proBNP, ng/L	3391 (1152, 5748)	3064 (1351, 8081)	4245 (1790, 8283)	0.18	0.07
ΔNT-proBNP, ng/L	−1499 (−3259, −34)	−1134 (−3348, −170)	−1639 (−4808, −283)	0.36	0.20
ΔNT-proBNP, %	−33 (−54, −5)	−35 (−51, −9)	−31 (−50, −11)	1.0	0.93
Ur Na, mmol/L	43 (27, 72)	53 (33, 78)	52 (39, 74)	0.18	0.19

Data presented as median and interquartile range. ΔBUN indicates change in blood urea nitrogen; ΔBUN/Cr, change in BUN-to-creatinine ratio; ΔCl, change in serum chloride; ΔCr, change in serum creatinine; ΔCysC, change in serum cystatin C; ΔeGFR, change in estimated glomerular filtration rate; ΔNa, change in serum sodium; ΔNT-proBNP, change in N-terminal pro-B-type natriuretic peptide; BUN, blood urea nitrogen; Cl, serum chloride; Cr, serum creatinine; CysC, serum cystatin C; eGFR, estimated glomerular filtration rate; HCO3, serum bicarbonate; IRF, improvement in renal function; Na, serum sodium; NT-proBNP, N-terminal pro-B-type natriuretic peptide; SRF, stable renal function; Ur Na, urine sodium; and WRF, worsening renal function.

^*indicates statistical significance of P < 0.05. The P value and P_trend values are as reported in Table 1.

Table 3.

Measures of decongestion at 72 hours by renal function status.

Clinical Congestion Metrics	WRF (N = 53)	SRF (N = 149)	IRF (N = 75)	P	P trend
Total Urine Output, mL	7860 (5555, 9765)	8150 (6325, 10456)	8780 (7025, 11208)	0.07	0.024*
24-h Na Excretion, mmol	78 (39, 174)	147 (78, 221)	141 (86, 240)	0.008*	0.011*
ΔWeight, lb.	−5.1 (−8.1, −1.3)	−7.1 (−11.9, −3.2)	−9.0 (−12.4, −5.3)	0.002*	<0.001*
JVP ≥8 cm, %	53	66	71	0.12	0.05
≥2+ Edema, %	26	35	32	0.42	0.58
Orthopnea, %	65	70	72	0.66	0.37
Clinically Worsening or Persistent HF, %	8	4	9	0.26	0.54
Dyspnea VAS	73 (52, 92)	80 (62, 90)	80 (64, 90)	0.60	0.38
Global VAS	73 (50, 90)	78 (61, 87)	80 (56, 90)	0.45	0.37
72-h Cumulative Diuretic Dose, mg IV Furosemide Equivalents	378 (273, 597)	479 (345, 676)	479 (304, 723)	0.19	0.16

Data presented as median and interquartile range. HF indicates heart failure; IRF, improvement in renal function; JVP, jugular venous pressure; SFR, stable renal function; VAS, visual analog scale; and WRF, worsening renal failure.

^*indicates statistical significance of P < 0.05. The P value and P_trend values are as reported in Table 1.

Patterns in Urinary Markers of Renal Tubular Injury

At baseline, there were no significant differences in urinary NGAL, NAG, or KIM-1 among patients with IRF, WRF, and SRF (P > 0.05 for all; Supplementary Table 1). At 72 h, the IRF group had lower urinary NGAL compared to the WRF group (40 [6, 22] ng/mg.uCr vs. 138 [30, 394] ng/mg.uCr; P_trend = 0.017). Conversely, a higher urinary NAG was seen in the IRF group compared to the WRF group, which was of borderline statistical significance (12.5 [6.5, 21.1] mU/mg.uCr vs. 8.9 [5.6, 16.5] mU/mg.uCr; P_trend = 0.049). There was no difference in urinary KIM-1 across groups at 72 h (P_trend = 0.83). There were no significant differences in the change in any of the three markers individually or when pooled together across the three renal function groups between baseline and 72 h (P_trend > 0.19 for all; Table 4, Supplementary Figure 1, and Figure 2). Likewise, there were no statistically significant changes in any biomarker by groups of renal function (IRF vs. SRF vs. WRF) when using linear mixed models (P > 0.05 for all).

Table 4.

Average percent change in individual and combined tubular injury markers from baseline to 72 hours.

Injury Biomarker	WRF (N = 53)	SRF (N = 149)	IRF (N = 75)	P	P trend
NGAL, %	−18 (−62, 262)	−32 (−82, 101)	−39 (−80, 119)	0.16	0.19
NAG, %	−5 (−34, 29)	−0.9 (−0.35, 56)	1 (−37, 59)	0.63	0.53
KIM-1, %	−4 (−37, 69)	−10 (−41, 48)	13 (−38, 89)	0.36	0.87
Combined, %	37 (−27, 127)	2 (−30, 68)	15 (−36, 138)	0.24	0.70

Data presented as median and interquartile range. The P value and P_trend values are as reported in Table 1.

“Combined” refers to the collective assessment of the three injury markers together. KIM-1 indicates kidney injury molecule 1; NAG, N-acetyl-β-d-glucosaminidase; and NGAL, neutrophil gelatinase–associated lipocalin.

Figure 2.

Biomarker Score: average percent change in combined tubular injury biomarkers from baseline to 72 hours. IRF indicates improvement in renal function; SRF, stable renal function; and WRF, worsening renal function.

Survival Analyses and Biomarker-Based Risk Stratification

At 180 d, 59 of the 277 patients (21%) patients had died. Compared to patients with SRF, patients with IRF had an approximately two-fold higher risk of death (unadjusted hazard ratio [HR] 1.98, 95% confidence interval [CI] 1.10–3.58, P = 0.024; Figure 3A). IRF remained significantly associated with increased mortality following multivariable analyses (adjusted HR 2.19, 95% CI 1.20–4.01, P = 0.011; variables remaining following backward elimination: patient race, use of beta blockers, baseline eGFR, baseline BUN, NT-proBNP at 72 h, and change in weight at 72 h). Compared to patients with SRF, patients with WRF did not have a significantly higher mortality risk (unadjusted HR 1.37, 95% CI 0.65–2.88, P = 0.41). There were no significant differences in mortality when comparing patients with IRF to patients with WRF (adjusted HR 0.73, 95% CI 0.35–1.54). When patients were grouped based upon whether renal function was dynamic (i.e., either the IRF or WRF definition was met) or stable, patients with dynamic renal function had increased mortality (adjusted HR 1.97, 95% CI, 1.12–3.48, P = 0.02; Figure 3B).

Figure 3. Kaplan Meier survival curves by renal function status.

A. Worsening renal function (WRF) vs stable renal function (SRF) vs improvement in renal function (IRF)

The linear test for trend for differences in survival was performed in order of observed best to worst survival (i.e., from SRF to WRF to IRF).

B. Dynamic renal function vs. SRF.

* indicates statistical significance of P < 0.05. RF indicates renal function.

Neither increasing nor decreasing concentrations of renal tubular injury biomarkers over 72 h was associated with 180-d mortality risk (Figure 4A). Individual assessment of each of the three tubular injury markers also did not find any association between pattern of change and mortality risk (Supplementary Figure 2). When patients were grouped considering both renal function status change and renal tubular marker change direction, patients with IRF and improved renal tubular injury markers had the worst mortality rates while patients with SRF and stable markers had the best survival (unadjusted HR 3.25, 95% CI 1.06–9.99, P = 0.04; Figure 4B).

Figure 4. Kaplan Meier survival according to renal tubular injury.

A. Survival based on change in renal tubular injury markers.

The linear test for trend for differences in survival was performed in order of observed best to worst survival (i.e., from Stable to Worsening to Improving).

B. Survival based on renal function status and pattern of change in renal tubular injury markers.

* indicates statistical significance of P < 0.05. IRF indicates improvement in renal function; and SRF, stable renal function.

Discussion

The principal findings of this post hoc analysis of the ROSE-AHF trial are as follows: (1) patients with IRF during ADHF hospitalization managed with intravenous diuresis had increased mortality relative to patients with SRF; (2) upon admission for ADHF, there was a greater degree of renal dysfunction and higher natriuretic peptides in patients who subsequently experienced IRF; and (3) no pattern of change in urinary biomarkers of renal tubular injury independently predicted long-term adverse events. We suspect that improved kidney function is highly unlikely to causally worsen outcomes. Therefore, the link between IRF and its associated poor clinical outcomes is best explained by the fact that IRF observed with aggressive intravenous diuresis likely identifies sicker patients who have worse underlying cardiorenal function and more extensive congestion, which may have been incompletely treated.

Although counterintuitive, we found ADHF patients with IRF were at higher risk of death compared to ADHF patients with SRF, which is consistent with prior studies.^{6, 7} Across multiple cohorts, SRF is associated with the highest survival rate, indicating that dynamic renal function, regardless of direction, is a negative prognostic marker in the setting of ADHF.^{7, 19} IRF seems to be an indicator of cardiorenal syndrome severity, and most commonly follows the resolution of WRF, which typically has occurred prior to hospital admission for ADHF.⁸ In this analysis, patients with IRF had laboratory markers of more severe renal dysfunction than other patients. Patients with IRF also had lower baseline urine sodium concentrations, suggesting greater renal sodium avidity. Furthermore, patients with IRF had modest markers of more severe heart failure, including lower systolic blood pressure and lower serum sodium and chloride concentrations. More advanced heart failure can also be supported by the presence of greater venous congestion, which is increasingly appreciated as a major driver of adverse renal and cardiovascular outcomes.^{4, 17, 25, 26} Although physical examination measures of congestion were similar across cohorts at baseline and 72 h, the IRF group likely had greater volume expansion at baseline, which may have persisted at 72 h. This conclusion is supported by the observations that patients with IRF had significantly more weight loss, urine output, and sodium excretion without a concomitant larger change in NT-proBNP or improvement in congestion status at 72 h compared with other patients. In fact, patients who had IRF trended toward higher NT-proBNP levels at 72 h compared with patients who had SRF or WRF, a pattern similarly observed at baseline.

Defining renal function based solely on markers of glomerular filtration (i.e., the serum creatinine) is inherently problematic as creatinine perturbations are poorly correlated with renal tubular function and the integrity of the kidney at the tissue level.^{11, 27} Consequently, markers of renal tubular injury have come recently under significant investigation to better characterize global kidney function. We found no significant differences in the baseline levels of or in the relative change in NGAL, NAG, or KIM-1, or pooled tubular injury biomarkers among ADHF patients with IRF, WRF, or SRF. Furthermore, no pattern of change in renal tubular injury markers predicted improved or worsened risk of survival. Indeed, we observed that among ADHF patients with IRF, patients with improvements in biomarkers had the worst survival. This finding reveals the inability of biomarkers to identify particularly at-risk patients and suggests that increased renal tubular injury is unlikely to be a mechanism explaining the increased mortality observed in the setting of IRF. Therefore, serum creatinine fluctuations are not able to accurately identify underlying or developing renal tubular injury in ADHF, and the consideration of renal injury markers does not appear to provide additional information to guide therapeutic interventions.

The findings of this study complement prior literature that questions the value of tailoring ADHF therapies to prevent small increases of the creatinine or target a down-trending creatinine as a therapeutic goal in the treatment of ADHF. ADHF complicated by WRF and increased renal tubular injury markers during aggressive decongestion is associated with neutral if not better long-term outcomes than patients without these laboratory markers of worsening kidney injury.^{11, 24} Many guideline-directed medical therapies that improve heart failure outcomes cause temporary decreases in kidney function.²⁸ Our findings indicate IRF is a marker of disease severity and congestion rather than a therapeutic target. Collectively, growing evidence suggests that targeting decongestion should be the primary goal of ADHF therapies rather than preventing rises of or aiming for improvements in creatinine or renal tubular injury markers. The data presented here call for further study into interventions that aim to restore cardiorenal homeostasis or at least ameliorate the cardiorenal interactions in this patient population among whom outcomes are among the poorest across the ADHF spectrum.

Limitations

This analysis has several important limitations. The associations from this post hoc analysis of a randomized clinical trial should not be interpreted as causality. The ROSE-AHF study specifically enrolled patients with pre-existing renal dysfunction, limiting generalization to patients without chronic kidney disease. It was at the discretion of the treating physician to initiate, adjust, or discontinue guideline-directed medical therapies, which can impact markers of kidney function and were not accounted for in our analyses. Outpatient assessments of kidney function before and after the study period were not available. Thus, we could not quantify patients’ pre-hospital renal function trajectory, the magnitude of renal function change prior to admission, or the change in kidney function over the 180-d follow-up period. Survival analyses could be limited by loss of sample size at 180 d beyond which can be attributed to mortality alone. Finally, because of the definitions used to create the renal function and tubular injury marker groups, the number of subjects per group is limited, increasing the likelihood for insufficient power to identify potential differences.

Conclusions

IRF during ADHF treated with intravenous diuresis is associated with increased mortality compared to SRF, as is any dynamic renal function. Renal tubular injury markers did not demonstrate different patterns of change among patients with improving, stable, or worsening eGFR. Our findings do not support tailoring ADHF therapies toward a goal of IRF. Rather, the findings here suggest the focus of care of this particularly high-risk advanced heart failure sub-group should relate to preventing the factors that resulted in cardiorenal destabilization and address the etiologies that may explain persistent cardiorenal perturbations.

Clinical Perspective.

What is new?

• Improvement in renal function (IRF) during treatment for acute decompensated heart failure (ADHF) is paradoxically associated with increased mortality.

• We demonstrate that IRF observed during ADHF management is a marker for greater disease severity, including more severe renal dysfunction and likely more congestion.

• Using established markers for renal tubular injury, we found no differences in the change in injury among ADHF patients with improved, stable, or worsening renal function.

What are the clinical implications?

• Trends in creatinine do not correlate with trends in renal tubular injury markers. Therefore, creatinine fluctuations observed during ADHF likely represent transient changes in renal filtration rather than histologic or structural kidney damage.

• ADHF therapies should not be tailored toward a goal of IRF. Rather, the focus of care should be identifying this high-risk advanced heart failure sub-group and preventing the factors that result in cardiorenal destabilization.

Non-standard Abbreviations and Acronyms

ADHF

acute decompensated heart failure

ANOVA

analysis of variance

BioLINCC

Biological Specimen and Data Repository Information Coordinating Center

BUN

blood urea nitrogen

CI

confidence interval

CKD-EPI

Chronic Kidney Disease Epidemiology Collaboration

eGFR

estimated glomerular filtration rate

HR

hazard ratio

IQR

interquartile range

IRF

improvement in renal function

KIM-1

kidney injury molecule 1

MDRD

Modification of Diet in Renal Disease

NAG

N-acetyl-β-d-glucosaminidase

NGAL

neutrophil gelatinase-associated lipocalin

NT-proBNP

N-terminal pro-B-type natriuretic peptide

ROSE-AHF

Renal Optimization Strategies Evaluation–Acute Heart Failure

SRF

stable renal function

WRF

worsening renal function