Improving renal function during diuresis: useful information or just noise?

This article refers to ‘Decongestion discriminates risk for one-year mortality in patients with improving renal function in acute heart failure’ by N. Wettersten et al., published in this issue on pages 1122–1130. Researchers and clinicians continue attempts at decoding the meaning of changes in creatinine during the treatment of acute decompensated heart failure (ADHF). Over the past two decades, data have accumulated indicating that, on a population level, worsening renal function (WRF) is associated with reduced survival.1, 2 Based on these consistent observations, many of us assumed a causal relationship between the reduction in kidney function and worse outcomes. However, recent data have challenged these assumptions by demonstrating that WRF in the setting of beneficial therapies, such as effective decongestion or neurohormonal antagonist titration, is associated with neutral or even improved outcomes.3-5 Two potential explanations for this paradox are that (1) the therapeutic benefit of the intervention overwhelms the direct negative effect of WRF, or (2) acute changes in creatinine during ADHF primarily serve as a marker for disease severity or haemodynamic perturbations, without direct negative effect of WRF. The more recent observation that improved renal function (IRF) is associated with worsened outcomes is a strong indicator that #2 is a primary factor.6 While many theories and even animal experiments support hypotheses why WRF could have direct causal negative effect, it is hard to postulate how improvement in renal function could directly worsen survival. We also know we can improve renal function at the expense of the overall heart failure (HF) status. In a clever study illustrating this point, patients had all HF medications suspended for 48 h. Here we saw significantly improved renal function that came at the expense of a doubling of natriuretic peptides and of New York Heart Association class.7 The emerging narrative from these data is that the presence or directionality of small to moderate sized changes in kidney function provides minimal actionable information. Rather, the clinical context underlying the changes in creatinine is paramount in determining the outcome associated with the change. In this issue of the Journal, the study by Wettersten et al. 8 adds significantly to the discussion. Evaluating 760 ADHF patients enrolled into the AKINESIS trial, the authors sought to understand the relationship between incomplete decongestion and IRF. IRF was defined as a ≥20% increase in estimated glomerular filtration rate, which occurred in 22% of the population. Adequate decongestion was defined as a ≥40% improvement in B-type natriuretic peptide (BNP), and this occurred in approximately 44% of the population. Similar to prior studies, IRF appeared largely the product of pre-admission WRF and was associated with a significantly higher rate of post-discharge mortality. WRF and IRF in ADHF appear to share similar trajectories with different categorization based on the clinician's perspective of the baseline renal function (Figure 1). When the outpatient renal function in the past several months is unknown, selection of the serum creatinine at admission as the ‘baseline serum creatinine’ alters the diagnosis of WRF threefold compared to other definitions of baseline renal function.9 The new and important observation from Wettersten et al. is that the risk associated with IRF was entirely dependent on the level of decongestion achieved in the hospital. Notably, patients with IRF and inadequate decongestion had a dramatically increased risk of death compared to their well decongested non-IRF counterparts. However, in patients with IRF that achieved adequate decongestion, there was no statistically significant risk for worsened survival. These observations reenforce the concept that adequate decongestion is critical during ADHF treatment, and we as clinicians should not modify therapy based on small to moderate sized changes in creatinine. To that end, a small bump in creatinine in isolation should not dissuade continued diuresis of a congested patient. Additionally, the current findings support the somewhat counterintuitive notion that an improvement in kidney function may imply substantial residual congestion and a need for additional diuresis. Of the 166 patients with IRF in this study, 109 (66%) did not achieve at least a 40% reduction in BNP. This is surprising since IRF will lead to increase clearance of natriuretic peptides, thus further lowering BNP levels independent of any change in congestion. This would support speculation that there was substantial remaining congestion in most IRF patients and help to explain the observation that the majority of the risk associated with IRF was explainable by the BNP achieved during diuresis. Although limited by small numbers, the current analysis of transient vs. persistent IRF allows some additional speculation on the intersection between congestion and renal function. Of the 166 patients that met criteria for IRF at any time during hospitalization, 60 of these patients no longer met criteria for IRF at the time of discharge. Similar to a previous study by our group of transient vs. persistent IRF, patients with transient IRF had a longer length of stay.6 Thus, one can assume that there was either additional therapy (i.e. decongestion or titration of neurohormonal antagonists) or worsening HF (e.g. refill of the intravascular compartment from the extravascular space) leading to the re-worsening of kidney function in patients with transient IRF. In our prior study, there was substantially greater diuresis in the patients with transient IRF (6.9 vs. 4.8 L, P = 0.004) and also a 30% better long-term survival in these patients.6 However, in the current study, the authors found the opposite; persistent IRF having significantly better survival than transient IRF. By definition IRF–mortality associations are confounding; since IRF is a marker, not a cause, of worsened outcomes, these findings should not be assumed to be contradictory. Rather, upon close inspection of the subgroups based on congestion, a potentially consistent signal emerges. Specifically, in patients with transient IRF and inadequate decongestion (n = 45), mortality was 38%, the worse group in the cohort. Here we could speculate that this group had in-hospital worsening of HF leading to deterioration in kidney function. In patients with transient IRF and adequate decongestion (n = 15), mortality was only 7% and very similar to the patients with adequate decongestion without IRF (n = 274, 9%). WRF in the setting of adequate diuresis has repeatedly been associated with neutral to improved outcomes, so this should not be surprising. Although limited by small number, these observations are once again in line with the consistent observation that a change in creatinine, whether it be WRF, IRF or post IRF-WRF, can only be interpreted within the clinical context of the patient. Despite the fact that a relatively consistent message has now emerged from the large observational literature on change in creatinine during ADHF, if anything we now have more rather than fewer questions on how to respond to these data clinically. For example, it is now clear that in the setting of a small increase in creatinine that occurs during aggressive decongestion, the change in kidney function by itself is free from prognostic significance. Additionally, the observations from Wettersten et al. also suggest that when a decrease in creatinine occurs in the absence of effective decongestion, the change in kidney function by itself is again without prognostic significance. But what do we do with this information? Does this mean we should ignore any and all changes in creatinine during diuresis? Is there a threshold at which point the change in kidney function becomes too severe, and it starts to have prognostic significance? Here we enter a completely data free zone. The reality is the observational literature from which these conclusions have been drawn was generated largely in an era when physicians were working hard to preserve or improve kidney function. As a result of this nephroprotective approach to treating ADHF, there are sparse observational data on what happens in patients with more dramatic changes in creatinine. However, if we change the rules and now completely abandon any concern about modest changes in creatinine, do these same conclusions hold up? Just like when shooting for no change in creatinine we ended up with somewhere around one-third of HF admissions having a small to moderate increase, if we now shoot for a small change in creatinine, we will assuredly end up accidentally causing some patients to have large changes in creatinine (Figure 2). How this will ultimately impact outcomes is unknowable. Short of waiting another 20 years to see what observations during this new era of ‘permissive hypercreatininaemia’ produce, randomized trials evaluating different strategies are required to gain definitive understanding on how to approach kidney function in ADHF.