Introduction
Acute kidney injury (AKI) following cardiac surgery is prevalent and associated with considerable morbidity and mortality [
1,
2]. Perioperative resuscitation solutions in intensive care units shifted from crystalloids toward colloids in the hope of facilitating intravascular volume repletion [
3,
4]. However, synthetic starches are now recognized as an independent risk factor for AKI in critically ill patients [
5,
6]. The Kidney Disease Improving Global Outcome (KDIGO) Clinical Practice Guidelines for AKI favor isotonic crystalloids over colloids in patients at risk for or presenting with AKI in the absence of hemorrhagic shock [
7]. Whether this applies equally to different synthetic starches and albumin is uncertain.
Study results regarding administration of albumin-containing fluids have been conflicting. Hyperoncotic albumin has demonstrated a clear benefit in patients with cirrhosis [
8,
9]. However, in patients with shock, hyperoncotic albumin has been associated with a fivefold increased risk of AKI [
10]. A recent European consensus statement recommends withholding hyperoncotic albumin use in critically ill patients outside of clinical trials [
10,
11]. The recently published Albumin Italian Outcome Sepsis (ALBIOS) Study, a randomized controlled trial (RCT) on the effect of hyperoncotic albumin (20%) versus crystalloids in hypoalbuminemic patients with severe sepsis and septic shock, did not show any difference in mortality and AKI between the groups [
12]. However, the timing of AKI in relation to albumin administration >was not detailed. The use of iso-oncotic albumin in a heterogeneous population of critically ill patients has also been studied in a large, multicenter RCT: the Saline versus Albumin Fluid Evaluation (SAFE) study. In that RCT, the investigators concluded that iso-oncotic albumin administration is safe from a survival perspective [
4,
13] and found no differences in organ dysfunction or duration of renal replacement therapy (RRT). However, kidney function was not independently reported, and only severe cases of AKI were collected. No data specifically on the safety of albumin as it relates to kidney function after cardiac surgery exist.
In 2009, we reported a dose-dependent risk of AKI using pentastarch 10% (250 kDa/0.45) following cardiac surgery [
6]. Since then, our practice has shifted toward the use of hydroxyethyl starch (HES) 6% (130 kDa/0.4) and albumin for volume expansion in unstable patients, and more recently it has shifted toward the use of crystalloids with or without albumin. In the present study, we hypothesized that both synthetic starches and albumin-containing solutions are independently associated with AKI in a dose-dependent fashion following cardiac surgery.
Discussion
In our present study, we found that albumin administration was associated with a dose-dependent increased risk of AKI in patients undergoing cardiac surgery. To further assess the risk associated with albumin and adjust for potential indication biases, we paired individuals who received albumin to controls who received no albumin. On the basis of this propensity score, albumin administration was still associated with a twofold increased risk for AKI. We repeated this methodology in individuals without significant postoperative hemodynamic instability and found similar results.
To our knowledge, this is the first study to show a dose–response relationship between albumin administration and increased risk for AKI. Our study offers new insights into the association between albumin administration and kidney function. Current evidence regarding the beneficial or deleterious effect of albumin in this context is inconclusive, especially in surgical patients and those without hypoalbuminemia [
4,
8-
10,
12,
21-
23]. In recent international consensus statements and guidelines on AKI and fluid administration [
7,
11], experts have recommended the use of crystalloids ahead of albumin in patients at risk for or with AKI and advised against the use of hyperoncotic albumin solutions for fluid resuscitation. Most of the evidence in support of these recommendations originates from the CRYCO Study Group [
10]. In their observational study of 1,013 patients with shock, hyperoncotic albumin was associated with a fivefold increased risk of renal event, defined as a twofold increase in creatinine or need for dialysis [
10]. The authors did not report any dose–response relationship between albumin administration and risk for AKI. Of note, patients with cirrhosis, the group from which most of the evidence regarding the beneficial effect of albumin on kidney function derives, were excluded from the study [
8,
22,
23]. The results of other small studies have suggested a deleterious effect of albumin on kidney function, but they had limited statistical power [
21,
24-
26] or involved very large doses of albumin (more than 1,100 g per patient) [
24].
In contrast to these findings, other studies have shown a protective or neutral effect on kidney function due to albumin administration [
4,
9,
12,
27]. In a recent meta-analysis, the protective effect of albumin on kidney function seemed to be present in patients with cirrhosis [
9]. In the largest RCT conducted to date on albumin administration in critically ill patients (the Saline versus Albumin Fluid Evaluation (SAFE) study), albumin 4% was proven to be safe in terms of mortality and severe AKI compared to crystalloids [
4]. Severe AKI was defined as a creatinine level >300 μmol/L and/or need for dialysis. However, the safety profile of albumin was not consistent between subgroups, with patients with traumatic brain injury having increased mortality [
28] and patients with sepsis having decreased mortality [
29]. There were no reports on kidney function in the traumatic brain injury
post hoc study [
28] and no increased rate of severe AKI associated with albumin in the septic subgroup [
29]. In comparison, the ALBIOS study investigators looked at the effect of hyperoncotic albumin in patients with severe sepsis and septic shock and did not find any difference in either mortality as the primary outcome or in severe AKI, defined as a creatinine level >300 μmol/L [
12]. The authors also looked at the incidence of AKI based on RIFLE criteria in a
post hoc analysis and did not find a difference between the two groups [
12]. The timing of AKI and albumin administration were not defined in the study. Importantly, the ALBIOS study differed from the other studies, as albumin 20% was administered on a daily basis if the patient's albumin level was below 30 g/L and not according to the clinical context [
12]. In addition, the cumulative fluid balance was lower in the albumin group, which could have reduced mortality and morbidity in that group [
30-
32]. In our present study, we included in our propensity score analysis the amount of fluid received by 36 hours after surgery to address this bias.
As highlighted in previous studies, the effect of albumin administration seems to differ between subgroups [
27-
29]. In our present study, we included patients undergoing cardiac surgery, a population with very limited data and a different pathophysiological model of AKI than patients with severe sepsis [
21,
33,
34]. In a recent systematic review, authors highlighted the need for additional studies on albumin administration in surgical patients [
35]. In cardiac surgery, a small study suggested that albumin may have a deleterious effect on kidney function [
21], whereas another found no difference [
34]. In the most recent study on this subject, researchers compared pentastarch, 25% albumin (75 g) and Ringer’s lactate solution and did not find any differences in creatinine levels between the three groups [
33]. Of note, cardiac surgery patients were excluded from the SAFE trial [
4], and only 6% to 7% of patients in the ALBIOS study had elective surgery [
12]. The type of surgery was not mentioned in the ALBIOS study.
Our study was not designed to investigate underlying biological mechanisms to explain the effect of albumin on kidney function. Postmortem examinations of a limited number of patients who received large doses of 25% albumin showed no evidence of abnormal albumin storage [
36]. The results of basic science studies on the renal impact of albumin administration have been contradictory [
37]. Albumin may have anti-inflammatory and antioxidant properties to protect against organ damage [
35,
38,
39]. However, although there are a number of mechanisms by which albumin might exert a beneficial effect, they are as yet unproven. In opposition to popular belief, according to the Starling equation, glomerular filtration pressure decreases as intracapillary oncotic pressure increases more than hydrostatic pressure, a situation favored by the use of hyperoncotic colloids [
37]. This principle may partly explain our findings regarding the dose-relationship effect between the risk of AKI and the amount of albumin administered. In light of recent findings supporting the beneficial effect of chloride-restrictive fluid administration on kidney function [
40], we ensured that albumin solutions had similar or lower chloride concentrations compared to normal saline, and we confirm that the use of crystalloids with lower chloride content was minimal at our institution when the study was conducted.
Although our present study was focused on kidney function as a primary endpoint, perioperative kidney function is a very relevant parameter in predicting long-term outcomes [
41-
43]. AKI is an important independent predictor of postdischarge mortality in patients undergoing cardiac surgery [
41-
43]. Other large studies have confirmed these findings in a broader population [
44,
45] as well as the deleterious influence of AKI on the development of chronic kidney disease (CKD) and ESRD [
45,
46]. Even mild AKI, defined as an increase in serum creatinine by 27 μmol/L (0.3 mg/dl), is associated with an increase in mortality [
42,
44,
47,
48], development of CKD [
49], increased length of stay [
48] and costs [
47]. Importantly, some of these studies were conducted in patients undergoing cardiothoracic surgery, as in our study population [
42,
48].
Our study has several strengths. It is the largest study to date on the effect of albumin administration on kidney function in patients undergoing cardiac surgery, a clinical setting with very limited data and a different pathophysiological model of AKI than severe sepsis or shock. We were able to demonstrate a twofold increase risk in AKI after adjustment with a propensity score that included baseline characteristics, surgical aspects, severity of illness scores and amounts of blood products, colloids and crystalloids administered. As recently shown, propensity scores may underestimate the true effect size compared to RCTs in critically ill patients [
20]. The use of a propensity score in our study could therefore underestimate the effect of albumin on the risk of developing AKI. Importantly, we have shown, for the first time to our knowledge, a dose–response relationship between albumin administration and risk for AKI in this population.
Our study also has limitations inherent to its retrospective and single-center design. First, there is a concern regarding hemodynamic instability (and risk of AKI) and the need to optimize left ventricular end-diastolic filling pressure using colloids. However, 86% of surgeries were elective, and any unstable condition was likely to surface after ECC at the end of the surgery. Our propensity score included the postoperative cardiovascular SOFA score. In addition, we obtained similar results in patients who had cardiovascular SOFA scores of zero. Second, we assessed whether the use of albumin by clinicians wishing to avoid giving patients synthetic colloids in the setting of impending AKI could have been a consequence of early AKI rather than its cause. However, the timing of AKI was similar in both groups. Third, it is possible that the period between the onset of oliguria and the rise of creatinine was a window within which albumin may have been preferred over synthetic colloids, again in the setting of impending AKI. However, we would have expected at least a trend toward greater use of crystalloids in the albumin group; however, the doses of crystalloids were not statistically different between groups (and lower in absolute numbers in the albumin group), even when addressing diagnosis of AKI <36 hours, 36 to 48 hours and >48 hours after surgery separately. We also had limited data on albumin levels. Recent studies have suggested a relationship between hypoalbuminemia and risk for AKI [
50,
51]. In our center, the postoperative prescription of albumin does not rely on a preoperative serum albumin value, as albumin is not often measured. Whether this could still be a confounder remains unknown. Some authors have suggested that albumin administration may improve morbidity once the albumin levels are increased up to 30 g/L [
39,
52], whereas others did not report such results [
12]. As the majority of our patients had nonurgent surgeries, it is unlikely that their baseline albumin levels would have been <30 g/L. We did not have data on emerging biomarkers for AKI. Current observational studies still rely on serum creatinine for making an AKI diagnosis, as these emerging biomarkers are not largely available even today. Furthermore, as mentioned, AKI, as defined by serum creatinine changes, is a relevant endpoint because it is strongly associated with increased mortality and morbidity. Finally, we cannot rule out that the association between albumin administration and AKI may have resulted from a selection bias. There also might be unknown confounding factors for which we were unable to adjust.
Competing interests
The authors declare that they have no competing interests.