Influence of renal dysfunction phenotype on mortality in the setting of cardiac dysfunction: analysis of three randomized controlled trials

Abstract

Aims

Renal neurohormonal activation leading to a reduction in glomerular filtration rate (GFR) has been suggested as a mechanism for renal insufficiency (RI) in the setting of heart failure. We hypothesized that RI occurring in the presence of renal neurohormonal activation may be prognostically more important than RI in the absence of renal neurohormonal activation.

Methods and results

Subjects in the Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE) trial (n = 429), Beta-Blocker Evaluation of Survival Trial (BEST) (n = 2691), and Studies Of Left Ventricular Dysfunction (SOLVD) trial (n = 6782) limited datasets were studied. The blood urea nitrogen to creatinine ratio (BUN/Creatinine) was employed as a surrogate for renal neurohormonal activation and the primary outcome was the interaction between BUN/Creatinine and RI associated mortality. Baseline RI (GFR < 60 mL/min/1.73 m²) was associated with mortality in all study populations (P < 0.001). In patients with higher BUN/Creatinine, the risk of mortality was consistently greater in patients with RI [adjusted hazard ratio (HR) ESCAPE = 2.8, 95% confidence interval (CI) 1.3–14.3, P = 0.019; BEST = 1.6, 95% CI 1.2–2.2, P = 0.002; SOLVD = 1.6, 95% CI 1.3–2.0, P = 0.001]. However, in patients with lower BUN/Creatinine, the risk of mortality was not elevated in patients with RI (adjusted HR ESCAPE = 0.94, 95% CI 0.35–2.4, P = 0.90, P interaction = 0.005; BEST = 0.97, 95% CI 0.64–1.4, P = 0.90, P interaction = 0.02; SOLVD = 1.0, 95% CI 0.8–1.3, P = 0.71, P interaction = 0.005).

Conclusion

The association between RI and poor survival observed in heart failure populations appears to be contingent not simply on the presence of a reduced GFR, but possibly on the mechanism by which GFR is reduced.

Introduction

Renal dysfunction in the setting of cardiac failure is one of the most powerful predictors of subsequent adverse events.^1,2 However, it is unclear whether this adverse prognosis is driven by the mechanism underlying the reduced glomerular filtration rate (GFR) or simply by the reduced GFR itself. Due to shared risk factors for cardiac dysfunction and intrinsic renal parenchymal disease, this differentiation has been difficult to achieve using traditional risk factors. Although the mechanistic basis for heart-failure-induced renal dysfunction has yet to be established, a prevailing theory is that excessive neurohormonal activation (i.e. increases in renin, angiotensin II, aldosterone, vasopressin, and sympathetic nervous system activity) is an important mediator.^1,3,4 This concept provides a plausible link between the kidney and heart given that these neurohormonal systems are central to both the pathophysiology of cardiac failure and the regulation of essentially all aspects of renal function. In fact, it has been previously demonstrated that neurohormonal parameters correlate better with GFR than with ventricular function in heart failure.⁵ Given the importance of neurohormonal activation in heart failure progression, we hypothesized that renal dysfunction associated with neurohormonal activation may have a disproportionately poor prognosis. Unfortunately, due to the complex interactions between neurohormonal systems, the large disconnect between their serum and local activity, and the methodological complexity with their direct measurement, identification of patients with heart-failure-induced renal dysfunction using systemic neurohormone levels is problematic.^6,7

The serum blood urea nitrogen (BUN) level has been shown in a large number of epidemiologic studies to predict adverse outcomes in congested heart failure, independent of renal function.⁸ This finding likely relates to the fact that the renal handling of urea is integral in fluid and sodium homeostasis and thus closely parallels aggregate renal neurohormonal activation.^6,9 In fact, the widespread clinical use of the BUN/Creatinine in the discrimination of ‘pre-renal’ renal dysfunction from intrinsic renal parenchymal disease is based on this physiology. Given that heart-failure-induced renal dysfunction is traditionally classified as a ‘pre-renal’ cause of renal dysfunction, the renal handling of urea can likely serve as a useful tool to differentiate renal dysfunction with a strong renal neurohormonal component (i.e. heart failure induced) as opposed to renal dysfunction resulting from other causes (i.e. coincident intrinsic renal parenchymal disease).¹⁰ The primary aim of this study was to provide proof of concept that different mechanistic/prognostic subtypes of renal insufficiency (RI) potentially exist, using the BUN/Creatinine as a surrogate for renal neurohormonal activation.

Methods

Our design was a retrospective cohort study of three clinical trial populations. Given that cardio-renal interactions may vary substantially with the severity of heart failure, and to provide opportunity to validate the findings, we studied three National Heart Lung and Blood Institute (NHLBI) sponsored clinical trial populations spanning a broad range of clinical heart failure and asymptomatic left ventricular dysfunction. The methods and primary results of the Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness (ESCAPE), Beta-Blocker Evaluation of Survival Trial (BEST), and the Studies Of Left Ventricular Dysfunction (SOLVD) prevention and treatment trials have been previously published.^11–15

The ESCAPE trial was a randomized, multicentre trial of therapy guided by pulmonary artery catheter vs. clinical assessment in hospitalized patients with acute decompensated heart failure.^14,15 Briefly, 433 patients were enrolled at 26 sites from January 2000 to November 2003. Inclusion criteria included an ejection fraction ≤30%, systolic blood pressure ≤125 mmHg, and at least one sign and one symptom of congestion. Exclusion criteria included an admission creatinine level >3.5 mg/dL.

The BEST trial was a randomized, placebo controlled trial investigating the impact of bucindolol in severe compensated chronic heart failure patients.¹³ Briefly, 2708 patients with New York Heart Association (NYHA) functional Class III or IV heart failure, a left ventricular ejection fraction of ≤35%, and use of an angiotensin converting enzyme inhibitor for ≥1 month (unless contraindicated) were randomized to bucindolol or placebo. Exclusion criteria included decompensated heart failure and a serum creatinine level of ≥3.0 mg/dL.

The SOLVD prevention and treatment trials were placebo controlled trials investigating the effect of enalapril on patients with asymptomatic and symptomatic left ventricular dysfunction and comprise the SOLVD limited dataset.^11,12 Briefly, 4228 patients were enrolled in the prevention trial and 2569 patients in the treatment trial at 23 international centres (total n = 6797). Inclusion in either trial required an ejection fraction ≤35%. Patients without evidence of overt heart failure that were not receiving heart failure medication were eligible for the prevention trial. Eligibility for the treatment trial required a diagnosis of heart failure and the use of medications for this condition. Exclusion criteria included a baseline creatinine level >2.5 mg/dL. The treatment and prevention populations were analysed as a whole to maximize power with planned sub-analyses of the population.

In all cohorts estimated GFR was calculated using the Modified Diet and Renal Disease equation.¹⁶ ‘Renal insufficiency’ was defined as a GFR < 60 mL/min/1.73 m². The ESCAPE, BEST, and SOLVD trials were conducted and supported by the NHLBI in collaboration with the ESCAPE/BEST/SOLVD study investigators. This analysis was conducted using a limited access dataset obtained from the NHLBI and does not necessarily reflect the opinions or views of the ESCAPE/BEST/SOLVD investigators or the NHLBI. This study was approved by the institutional review committee.

Statistical methods

Values reported are mean±standard deviation, median (Quartile 1–Quartile 3), and percentile. Independent Student's t-test and the Mann–Whitney U test were used to compare continuous parameters. Pearson's χ² was used to evaluate categorical variables. Correlation coefficients represent Spearman's rho. The primary objective of these analyses was to query whether BUN/Creatinine could function as a tool to differentiate potentially different mechanistic subsets of renal dysfunction. As such, the primary outcome of interest was the interaction between BUN/Creatinine and RI with respect to all cause mortality (as opposed to the main effect of BUN or BUN/Creatinine on all cause mortality). In the ESCAPE cohort, BUN/Creatinine was dichotomized as above or below median. Because the median BUN/Creatinine was in the normal range in the BEST and SOLVD cohorts, BUN/Creatinine was dichotomized and comparisons made between the top and bottom quartile. Proportional hazards modelling was used to evaluate time to event associations with all cause mortality. Candidate covariates for multivariable modelling were obtained by screening all baseline variables with missing data <5% and a univariate association with mortality (P≤0.2). Covariates were removed using backwards elimination (likelihood ratio) and variables with a P<0.2 were retained.¹⁷ Kaplan–Meier curves for death from any cause were plotted for the four combinations of groups between lower and higher BUN/Creatinine combined with presence or absence of RI. The x-axis was terminated when the remaining number at risk was <10% and statistical significance was tested using the log-rank statistic. Stratum-specific hazard ratios (HRs) were derived from proportional hazards modelling of the individual strata and the significance of the interactions was formally assessed using models incorporating terms for the main effect of RI, the main effect of high vs. low BUN/Creatinine, and the interaction between these variables. Proportional hazard models for the primary analysis and associated interaction models were subjected to 1000 bootstrap replications (with replacement) to derive P-values and 95% confidence intervals (CIs). Statistical analysis was performed with PASW Statistics version 18.0 (SPSS Inc., Chicago, IL, USA) and significance defined as two-tailed P < 0.05 with the exception of tests of interaction where significance was defined as P < 0.10.

Results

Decompensated inpatient cohort (ESCAPE trial)

Characteristics of the 429 patients in the ESCAPE trial with baseline values for BUN and GFR available are presented in Table 1 and Supplementary Table S1. Blood urea nitrogen to creatinine ratio demonstrated only a weak correlation with baseline GFR and baseline creatinine (Supplementary Table S2 and Supplementary Figure S1A). Levels of atrial natriuretic peptide, B-type natriuretic peptide, epinephrine, and norepinephrine levels were higher and serum sodium was lower in patients with a BUN/Creatinine above the median (Supplementary Table S1). Blood urea nitrogen to creatinine ratio was significantly associated with increased mortality (HR = 1.4 per 10 increase in BUN/Creatinine, 95% CI 1.1–1.6, P = 0.001). Similarly, a BUN/Creatinine above the median was associated with a significantly increased risk for death (HR = 2.2, 95% CI 1.4–3.9, P < 0.001). This association persisted after adjustment for baseline GFR (HR = 1.9, P = 1.2–3.0, P = 0.007). However, the strength of the direct association between BUN/Creatinine and mortality was not as strong as that of BUN or GFR (Supplementary Table S3).

Table 1.

Baseline characteristics of the trial populations

Characteristic	ESCAPE (n = 429)	BEST (n = 2961)	SOLVD (n = 6782)
Demographics
Age	56.1 ± 14.0	60.2 ± 12.3	59.3 ± 10.2
White race (%)	59.7	69.9	88.2
Male (%)	74.6	78.0	85.6
Past medical history
Hypertension (%)	46.8	58.9	39.0
Diabetes (%)	32.3	35.6	19.3
Ischaemic heart disease	49.2%	48.6%	74.70
Physical examination
Heart rate	82.4 ± 15.7	82.2 ± 13.4	76.2 ± 12.3
Systolic blood pressure (mmHg)	105.5 ± 16.3	118.5 ± 19.4	119.4 ± 16.8
Medications (baseline)
ACE inhibitor or ARB	90.4%	91.9%	N/A
Beta-blocker	61.9%	N/A	17.8%
Digoxin (%)	72.5	92.2	33.1
Loop diuretic (%)	98.4	91.8	32.5
Laboratory value
Serum sodium (mmol/L)	136.7 ± 4.4	139.0 ± 3.4	139.5 ± 3.0
GFR (mL/min/1.73 m2)	57.1 ± 24.9	65.5 ± 23.2	65.7 ± 19.0
GFR <60 (mL/min/1.73 m2) (%)	60.4	40.8	39.4
BUN (mg/dL)	34.9 ± 22.7	24.8 ± 15.4	18.7 ± 6.9
Creatinine (mg/dL)	1.5 ± 0.6	1.2 ± 0.4	1.2 ± 0.3
BUN/Creatinine	21.1 (16.4–26.5)	17.8 (14.3–22.7)	15.0 (12.5–18.5)
Functional status/ejection fraction
Left ventricular ejection fraction (%)	19.3 ± 6.6	23.0 ± 7.3	27.0 ± 6.3
NYHA class	3.9 ± 0.4	3.1 ± 0.3	1.7 ± 0.7

ESCAPE, Evaluation Study of Congestive Heart Failure and Pulmonary Artery Catheterization Effectiveness; BEST, Beta-Blocker Evaluation of Survival Trial; SOLVD, Studies Of Left Ventricular Dysfunction; ACE, angiotensin converting enzyme; ARB, angiotensin receptor blocker.

Baseline RI (GFR<60 mL/min/1.73 m²) was also associated with increased mortality (HR = 2.5, 95% CI 1.5–4.3, P < 0.001) and was significantly more common in patients with a BUN/Creatinine above the median [odds ratio (OR) =2.4, P < 0.001] (Supplementary Table S1). In the strata of patients with a BUN/Creatinine above the median, RI was associated with a substantial risk for subsequent death (Table 2). This association persisted after adjusting for baseline factors associated with mortality (hypertension, ischaemic aetiology, age, systolic blood pressure, serum sodium, loop diuretic dose, thiazide diuretic use, beta-blocker use, and angiotensin converting enzyme inhibitor or receptor blocker use) (HR = 2.8, 95% CI 1.3–14.3, P = 0.019). However, in the strata of patients with a BUN/Creatinine below the median, the hazard for mortality associated with RI was no longer present (Table 2, Figure 1A). This lack of association remained after adjusting for baseline characteristics (HR = 0.94, 95% CI 0.35–2.4, P = 0.90, P interaction = 0.012). Similar results were found when GFR was analysed as a continuous variable (low BUN/Creatinine adjusted HR = 0.93 per 10 mL/min/1.73 m² increase, 95% CI 0.78–1.1, P = 0.45; high BUN/Creatinine HR = 0.77, 95% CI 0.64–0.93, P = 0.007, P interaction = 0.04). The interaction remained significant when both estimated GFR (eGFR) and BUN/Creatinine were entered as continuous variables (P = 0.09). However, when these analyses were repeated substituting risk stratifying variables not directly related to renal neurohormonal activation, GFR had similar ability to predict mortality in both high and low risk strata (Supplementary Table S5).

Table 2.

Mortality risk associated with renal insufficiency in populations with varying degrees of heart failure severity stratified by blood urea nitrogen to creatinine ratio

Trial population	Low BUN/Creatinine		High BUN/Creatinine
	HR (95% CI)	P	HR (95% CI)	P	P interaction
ESCAPE	1.2 (0.58–2.4)	0.690	4.6 (2.2–22.3)	0.002*	0.005*
BEST	1.2 (0.87–1.7)	0.190	2.2 (1.7–2.7)	0.001*	0.008*
SOLVD
Overall SOLVD population	1.2 (1.0–1.5)	0.053	2.2 (1.9–2.7)	0.001*	0.001*
Only NYHA Class I and II	1.2 (0.93–1.4)	0.184	2.0 (1.6–2.6)	<0.001*	<0.001*
Prevention trial	1.1 (0.85–1.5)	0.421	2.5 (1.8–3.5)	<0.001*	<0.001*
Prevention trial and NYHA Class I only	1.1 (0.78–1.6)	0.858	2.6 (1.7–4.0)	<0.001*	0.002*
Prevention trial, NYHA Class I only, and no incident CHF prior to death	1.1 (0.72–1.7)	0.625	2.5 (1.5–4.2)	0.001*	0.022*
Prevention trial, NYHA Class I only, no incident CHF, and death unrelated to CHF	1.0 (0.63–1.7)	0.854	2.3 (1.3–413)	0.003*	0.036*

HRs represent the unadjusted risk for all cause death comparing patients with and without a GFR <60 mL/min/1.73 m². HR, hazard ratio; CI, confidence interval; BUN/Creatinine, blood urea nitrogen to creatinine ratio; NYHA, New York Heart Association; CHF, congested heart failure.

*Significant P-value. BUN/Creatinine dichotomized as above or below the median for the ESCAPE trial and top vs. bottom quartile for the BEST and SOLVD populations.

Figure 1.

Kaplan–Meier curves grouped by renal function and blood urea nitrogen to creatinine ratio in the decompensated inpatient cohort (A, ESCAPE), the severe outpatient cohort (B, BEST), and the mild outpatient cohort (C, SOLVD). eGFR: estimated glomerular filtration rate. BUN/Creatinine: blood urea nitrogen to creatinine ratio. BUN/Creatinine dichotomized as above or below the median for the ESCAPE trial and top vs. bottom quartile for the BEST and SOLVD populations.

Despite the differential risk for mortality when stratified by baseline BUN/Creatinine, overall both groups had similar baseline characteristics and the estimated GFR difference between the RI groups was only 3 mL/min/1.73 m² (Supplementary Table S1). Similarly, serum creatinine was similar between BUN/Creatinine groups (Supplementary Table S4). Notable exceptions were the dose of loop diuretics, use of angiotensin converting enzyme or angiotensin receptor blockers, B-type natriuretic peptide, atrial natriuretic peptide, norepinephrine, and epinephrine levels; variables that directly affect or are directly related to renal neurohormonal activation (Supplementary Table S1).

Severe outpatient cohort (BEST)

Characteristics of the 2691 patients in the BEST with baseline values for BUN and GFR available are presented in Table 1 and Supplementary Table S6. In this cohort, BUN/Creatinine also demonstrated only a weak correlation with baseline GFR and baseline creatinine (Supplementary Table S2 and Supplementary Figure S1B). Levels of norepinephrine were higher and serum sodium lower in patients with a BUN/Creatinine in the top vs. bottom quartile (Supplementary Table S6). Blood urea nitrogen to creatinine ratio was associated with increased mortality (HR = 1.4 per 10 increase in BUN/Creatinine, 95% CI 1.3–1.5, P < 0.0001) and a BUN/Creatinine in the top compared with the bottom quartile was associated with increased mortality (HR = 2.4, 95% CI 2.0–2.9, P < 0.0001). This association persisted after adjustment for baseline GFR (HR = 2.1, 95% CI 1.7–2.6, P < 0.0001). However, the direct association between BUN/Creatinine and mortality was weaker than that of BUN or GFR (Supplementary Table S3).

Baseline RI was associated with increased mortality (HR = 1.9, 95% CI 1.7–2.2, P < 0.0001). A BUN/Creatinine in the top quartile was significantly more common in patients with RI (OR = 3.2, P < 0.0001) (Supplementary Table S6). In the strata of patients with a BUN/Creatinine in the top quartile, RI was associated with a significant risk for subsequent death (Table 2). This association persisted after adjusting for baseline factors associated with mortality (hypertension, diabetes, obstructive coronary disease, age, systolic blood pressure, sodium level, haemoglobin level, uric acid level, loop diuretic dose, digoxin use, bucindolol use, vasodilator use, angiotensin converting enzyme inhibitor use, gender, ejection fraction, Minnesota Living with Heart Failure score, and NYHA class) (HR = 1.6, 95% CI 1.2–2.2, P = 0.002). However, in the strata of patients with a BUN/Creatinine in the bottom quartile, the risk of mortality associated with RI was no longer present (Table 2, Figure 1B). This lack of association remained after adjusting for baseline characteristics associated with mortality (HR = 0.97, 95% CI 0.64–1.4, P = 0.90, P interaction = 0.02). Notably, these differences in mortality occurred despite only 4.3 cc/min lower GFR in the high BUN/Creatinine group (Supplementary Table S6). Serum creatinine was also similar between BUN/Creatinine groups (Supplementary Table S4). Similar results were obtained with GFR as a continuous variable (low BUN/Creatinine adjusted HR = 0.96 per 10 mL/min/1.73 m² increase, 95% CI 0.0.87–1.1, P = 0.36; high BUN/Creatinine HR = 0.84, 95% CI 0.78–0.90, P < 0.001, P interaction = 0.023) and with alternative risk stratification variables (Supplementary Table S5). Likely due to the large number of patients with equivocal BUN/Creatinine levels in the overall cohort (i.e. median value 17.8), the interaction was not statistically significant when both eGFR and BUN/creatinine were entered as continuous variables (P = 0.71).

Mild outpatient cohort (SOLVD)

Associations related to baseline RI in the SOLVD population have been previously reported.¹⁸ Characteristics of the 6782 patients with baseline values for BUN and GFR are described in Table 1 and Supplementary Table S7. BUN/Creatinine was weakly correlated with baseline GFR and creatinine (Supplementary Table S2 and Supplementary Figure S1C). Blood urea nitrogen to creatinine ratio was associated with increased mortality (HR = 1.2 per 10 increase in BUN/Creatinine, 95% CI 1.1–1.3, P < 0.001). Similarly, the presence of a BUN/Creatinine in the top compared with the bottom quartile was associated with a greater risk for death (HR = 1.2, 95% CI 1.1–1.4, P = 0.004). This association was strengthened after adjustment for baseline GFR (HR = 1.3, 95% CI 1.2–1.5, P < 0.001). Interestingly, and as opposed to the other two cohorts, a BUN/Creatinine in the top quartile was significantly less common in patients with RI (OR = 0.61, P < 0.0001) (Supplementary Table S7). Similar to the ESCAPE and BEST populations, GFR and BUN demonstrated a stronger associated with mortality than BUN/Creatinine.

Baseline RI was associated with increased mortality (HR = 1.7, 95% CI 1.6–1.9, P < 0.0001). In the strata of patients with a BUN/Creatinine in the top quartile, RI was associated with a significantly greater risk for subsequent death (Table 2, Figure 1C). This association persisted after adjusting for baseline factors associated with mortality (age, race, ejection fraction, heart rate, diastolic blood pressure, NYHA class, serum sodium, history of diabetes, hypertension, stroke, or myocardial infarction, loop diuretic, potassium sparing diuretic, digoxin, beta-blocker use, and randomization to enalapril) (HR = 1.6, 95% CI 1.3–2.0, P = 0.001). However, in the strata of patients with a BUN/Creatinine in the bottom quartile, the risk of mortality associated with RI was small (Table 2) and no longer present after adjustment for baseline factors (HR = 1.0, 95% CI 0.8–1.3, P = 0.71, P interaction = 0.005). These differences in mortality occurred despite only 1.9 cc/min lower GFR in the high BUN/Creatinine group (Supplementary Table S7). Serum creatinine was also similar between BUN/Creatinine groups (Supplementary Table S4). Notably, significant interactions remained present when the analysis was restricted to progressively less severe heart failure populations (Table 2). Similar results were found using GFR as a continuous variable (low BUN/Creatinine adjusted HR = 1.0 per 10 mL/min/1.73 m² increase, 95% CI 0.94–1.1, P = 0.98; high BUN/Creatinine HR = 0.90, 95% CI 0.86–0.95, P < 0.001, P interaction = 0.026) and with alternative risk stratification variables (Supplementary Table S5). Likely due to the large number of patients with equivocal BUN/Creatinine levels in the overall cohort (i.e. median value 15.0), the interaction was not statistically significant when both eGFR and BUN/Creatinine were entered as continuous variables (P = 0.16).

Discussion

The principal finding of this analysis is the strong dependence of the risk for death associated with RI on the BUN/Creatinine across a broad spectrum of patients from asymptomatic left ventricular dysfunction to severe decompensated heart failure. In patients with a high BUN/Creatinine, RI was associated with a substantially greater incidence of death. However, in patients with a low BUN/Creatinine, RI had little to no adverse prognostic significance; associations that persisted after extensively controlling for potential confounders. These findings were replicated in three separate, independent cohorts, supporting the internal validity and generalizability of these results. Despite only small differences in GFR, subjects with RI and a high BUN/Creatinine exhibited multiple indices consistent with neurohormonal activation compared with those with RI and a low BUN/Creatinine. These findings provide proof of concept that the adverse prognosis associated with reduced renal function in heart failure may not simply be a consequence of a reduced GFR.

The renal clearance of urea is determined by the balance between urea filtration and tubular reabsorption.¹⁹ Indeed, the fact that serum BUN is influenced by factors related to both neurohormonal activation and GFR is a likely explanation for the large body of epidemiologic data demonstrating BUN to be a powerful independent prognostic indicator.⁸ Although changes in the filtration rate of urea can occur independently of neurohormonal activation (i.e. reduced nephron mass secondary to hypertension or diabetes), the rate of tubular reabsorption of urea continues to be strongly influenced by neurohormones such as vasopressin and angiotensin II.^20,21 As a result, BUN is disproportionately elevated in conditions with prominent renal neurohormonal activation as opposed to intrinsic renal parenchymal disease where the filtration of urea is reduced but the tubular reabsorption of urea is relatively unaffected. This physiology forms the basis for the clinical practice of differentiating ‘pre-renal’ from intrinsic renal parenchymal disease using the BUN/Creatinine as described in many basic medical texts.^19,22 Given that neurohormonal activation likely plays a key pathophysiological role in both renal and cardiac dysfunction in heart failure, it is plausible that the BUN/Creatinine, a routinely available, inexpensive, non-invasive marker, could serve as a tool to stratify patients with renal dysfunction at particularly high risk for adverse outcomes.^1,3,23

Although there may be biological plausibility to the above, this analysis can only describe association making it impossible to conclude causality and confirm or disprove the mechanism underlying the results. However, we have demonstrated highly significant and consistent results in three large and diverse heart failure populations, essentially eliminating the possibility that this is a chance finding. Perhaps most important is the finding that patients in these cohorts with a low BUN/Creatinine have little to no risk associated with a RI, a finding strengthened by adjustment for potential confounders. This is particularly noteworthy given that RI has been associated with adverse outcomes in essentially every population studied, including the general population.^24,25 As a result and regardless of the actual mechanism driving the association, there is a high probability that a low vs. high BUN/Creatinine is identifying biologically distinct groups of patients with RI.

Given that prognosis may not be determined solely by glomerular filtration, it may be that correction of the primary pathophysiological derangement (i.e. renal neurohormonal activation) could mitigate some of the associated risk with RI. As a result, strategies that aim to ameliorate renal neurohormonal activation rather than directly improving GFR may hold significant promise for these patients. The limitation of focusing primarily on GFR was recently highlighted by the failure of the adenosine antagonist rolofylline to produce clinically meaningful effects in a large multicentre heart failure trial.²⁶ Despite the fact that adenosine antagonists have been shown to increase GFR, they have no direct ability to decrease renal neurohormonal activation but paradoxically may actually lead to increased renin secretion.²⁷ In light of the substantial prevalence of RI in heart failure and the large number of existing and novel therapeutic strategies with which renal neurohormonal activation can be influenced, further study of this subject is of paramount importance.

Limitations

Given the post hoc nature of this study, the limitations inherent to retrospective analyses apply, uncontrolled confounding cannot be excluded, and causality is impossible to demonstrate. Patients with advanced RI (creatinine >3.5, 3.0, and 2.5 mg/dL in ESCAPE, BEST, and SOLVD, respectively) were excluded limiting generalization to these groups of patients where the influence of RI on mortality may be more pronounced. Moreover, the strict inclusion/exclusion criteria typical of clinical trials produce a population that may not be representative of general clinical practice. For example, the average age of the patients in these trials is significantly younger than typically found in registries and patients with preserved ejection fraction were excluded. Given that the elderly and heart failure with preserved ejection fraction makes up a large percentage of the heart failure population found in practice, these exclusions limit generalizability of these findings. In addition to the rate of clearance, urea production can be influenced by factors such as protein catabolism and diet, introducing additional noise and potential bias into the assessment of neurohormonal activation with BUN. Furthermore, although small, there was some degree of colinearity between BUN/Creatinine and both eGFR and serum creatinine, potentially introducing some bias into the analysis. The BEST, ESCAPE, and SOLVD trials were not designed to evaluate associations with renal function and thus treating physicians were not blinded to BUN or creatinine and may have modified treatment based on these values. Direct measures of neurohormonal parameters were unavailable in the majority of study participants limiting direct assessment of the value of these markers compared with the surrogate BUN. As a result of the above limitations, these findings should be regarded as hypothesis generating and, as such, should challenge rather than change current practice.

Conclusion

These data provide proof of concept that the presence of a reduced GFR alone is not necessarily indicative of a poor prognosis in patients with cardiac dysfunction. Rather, the setting in which reduced glomerular filtration occurs (i.e. neurohormonal activation) may be more important than simply the fact that it is reduced. Future studies are necessary to extend these findings to broader populations, identify more specific methods to differentiate these prognostically distinct forms of renal dysfunction, and potentially evaluate targeted therapeutic strategies for these groups.

Supplementary material

Supplementary material is available at European Journal of Heart Failure online.

Funding

This study was funded by National Institutes of Health grant 5T32HL007843-15.

Role of the funding source

The funding source had no role in study design, data collection, analysis or interpretation.

Conflict of interest: none declared.