Effects of 5% albumin plus saline versus saline alone on outcomes from large-volume resuscitation in critically ill patients

Abstract

Objective:

To compare 5% albumin to 0.9% saline for large-volume resuscitation (LVR, > 60ml/Kg within 24h), on mortality and development of acute kidney injury (AKI).

Design:

Retrospective cohort study.

Setting:

Patients admitted to intensive care units (ICUs) in 13 hospitals across Western Pennsylvania. We analyzed two independent cohorts, the High-Density Intensive Care (HiDenIC) databases: HiDenIC08 (July 2000 to October 2008, H08) and HiDenIC15 (October 2008 to December 2014, H15).

Patients:

Total of 18,629 critically ill patients requiring LVR.

Interventions:

5% albumin in addition to saline vs 0.9% saline.

Measurements and Main Results:

After excluding patients with AKI prior to LVR, 673/2,428 (27.7%) and 1,814/16,201 (11.2%) patients received 5% albumin in H08 and H15, respectively. Use of 5% albumin was associated with decreased 30-day mortality by multivariate regression in H08 (OR 0.65, 95% CI 0.49–0.85, p=0.002) and in H15 (0.52, 95% CI 0.44–0.62, p<0.0001), but was associated with increased AKI in H08 (OR 1.98, 95% CI 1.56–2.51, p<0.001) and in H15 (OR 1.75, 95% CI 1.58–1.95, p<0.001). However, 5% albumin was not associated with persistent AKI, and resulted in decreased MAKE at 30, 90 and 365 days. Propensity matched analysis confirmed similar associations with mortality and AKI.

Conclusion:

During LVR, 5% albumin was associated with reduced mortality and MAKE at 30, 90 and 365 days. However, a higher rate of AKI of any stage was observed that did not translate into persistent renal dysfunction.

Introduction

Albumin accounts for ~75% of proteins in plasma, and albumin concentration is the major determinant of colloid osmotic pressure within the intravascular space.¹ This is an important clinical consideration because, according to Starling’s law, colloid osmotic pressure is the main determinant of fluid reabsorption from the interstitium.² This understanding has fueled the appeal of albumin and colloids in general (e.g. gelatins, hydroxyethylstarches) as resuscitation fluids because theoretically, colloids are expected to increase colloid osmotic pressure in the intravascular space, and therefore be more efficient at expanding the intravascular volume than crystalloids. Data from healthy volunteers has confirmed that the expansion of the plasma volume achieved by colloids within a healthy vasculature is larger, lasts longer, and can be achieved with lower infusion volumes when compared to crystalloids.³ Furthermore, albumin may also exhibit potentially beneficial properties such as acting as an anti-inflammatory, anti-oxidant protein and improving microvascular flow.¹

However, recent advances in endothelial biology have revealed that the differential oncotic pressure between plasma and a layer of proteoglycans and glycosylated residues called the glycocalyx, and not the interstitium, is the true force preventing loss of fluid from the capillaries.² This has also underscored that the kinetics of colloid distribution vary significantly from health to disease. Loss of the glycocalyx and the disruption of the endothelial architecture during critical illness increases the trans-endothelial albumin escape rate up to 300%,^4,5 explaining why the volume expansion benefits demonstrated in healthy volunteers are absent in septic critically ill patients. Therefore, the efficiency of colloids as a resuscitation fluid in the critically ill has been questioned. At least six large randomized clinical trials (RCT) have shown that the use of 4% albumin^6,7, 20% albumin^8,9 or hyper-oncotic hydroxyethylstarch (e.g. 6% HES)^10,11 in different populations of critically ill, septic patients confers no survival benefit compared to either saline or balanced crystalloids, and that 6% HES may, on the contrary, increase the risk of acute kidney injury (AKI).^10,11

Although 4% and 20% albumin preparations have been shown to be safe in critically ill patients^6,8, observational data has raised doubts about the safety profile of albumin, especially for AKI.^12–14 Furthermore, the available evidence from RCT and observational studies is centered on patients requiring only modest volumes of fluid, and although the ALBIOS trial included patients receiving larger volumes of fluid⁸, the evidence in large-volume resuscitation (LVR) is lacking. Therefore, the aim of this study was to determine the effect of 5% albumin and 0.9% saline versus saline alone on mortality and on the development of AKI among patients receiving LVR. We hypothesized that exposure to 5% albumin plus saline would result in lower mortality and development of AKI as compared to saline alone in patients receiving LVR.

Materials and Methods

Study Design & Setting

We conducted a retrospective/observational study using the High-Density Intensive Care (HiDenIC)-08 database (H08) that includes data on critically ill adult patients admitted to ICUs in a single facility from July 2000 to October 2008. To confirm and probe the generalizability of our results, we reproduced the analysis in HiDenic-15 (H15), which includes data from adult patients admitted to the ICU at 13 facilities in western Pennsylvania from October 2008 to December 2014 (Table 1).

Table 1:

Cohort Characteristics

	HiDenIc-08	HiDenIc-15
Total Encounters	45,568	164,910
LVR-only Patients	4,710	39,918
Analytic Cohort	2,428	16,201
Enrollment Period	2000-Oct.2008^	Oct.2008^–2014
No. of Hospitals	1	13
No. of ICU	8	100+
Type of ICU	MICU%/SICU%%	MICU/SICU
SICU types	CT/Trauma, Neurotrauma/NICU	CT/Trauma/Neurotrauma/ NICU/Surgical
Hospital setting	Secondary, Tertiary, and Quaternary Care	Primary, Secondary, Tertiary, and Quaternary Care
Intensivist staffing1	Full Coverage	Full coverage/Consult only
Hospital size	Large@	Large
Hospital location	Urban	Urban/Rural
Patient Location	ICU	ICU
Primary Intervention	None	None
Primary Outcome	Mortality at 30 days	Mortality at 30 days
	Severe AKI*	Severe AKI
Secondary Outcome	Mortality at 90 days	Mortality at 90 days
	Mortality at 1 year	Mortality at 1 year
	Any AKI	Any AKI
		Length of stay (ICU)
		Length of stay (Hsp)
		MAKE at 30 days
		MAKE at 90 days
		MAKE at 1 year
Inclusion Criteria	Patients receiving large volume resuscitation, that is, ≥ 60ml/kg in a single 24-hour period	Patients receiving large volume resuscitation, that is, ≥ 60ml/kg in a single 24-hour period
Subgroup Analysis		Sepsis
		Cardiothoracic surgery
		Hepatic dysfunction

^*Severe Acute Kidney Injury (AKI) includes stage 2&3 only

^{^}Oct 2008-Mutually exclusive periods;

^@:More than 500 beds

^%MICU: Medical ICUs,

^%%SICU: Surgical ICUs

^1:Full coverage refers to dedicated ICU physicians available 24×7. Consult only refers to intensivists available by consult.

CT: Cadiothoracic, NICU: Neurovascular ICU.

Data Collection

The University of Pittsburgh Institutional Review Board approved the project under protocol PRO13120333. Various UPMC electronic health records (EHR) systems were combined to extract information on demographics, diagnosis, lab values, and procedure and diagnosis codes on all patients admitted at UPMC from 2000 to 2014. Post-discharge dialysis and mortality information was obtained from the US Renal Data System (USRDS), and National Death Index (NDI) database/Social Security Death Master File (SSDMF) respectively.

Study Cohorts and Sample Size Calculation

We analyzed data from H08 and H15 separately. In both datasets, the study cohort was composed of critically ill patients who received LVR, defined as ≥60ml/kg (regardless of the type of intravenous fluids) in a single 24-hour period. Time stamps for AKI diagnosis and LVR were used to exclude patients with AKI prior to the definition of LVR. To detect an effect size (odds ratio) of 0.52 in favor of albumin, and assuming a range of 5 to 30%, 30-day mortality among 14,387 patients receiving only 0.9% saline group, the total sample size required varied between 14,483 and 14,779. Therefore, we would need 96 to 392 patients to receive albumin in order to reach at least 80% power. Assuming a type 1 error rate of 5% at all mortality ranges, we would expect a range of 30-day mortality among those receiving albumin of approximately 2.5 to 14.5%.

Exposure

Exposure was defined by the fluid type received for LVR into: patients receiving only 0.9% isotonic saline (saline group) or patients receiving 5% albumin in addition to 0.9% saline (albumin group) (Table 1, Figure 1). The decision to use 5% albumin or saline was at the discretion of the treating physician. Clinical variables relevant to the exposure and outcome, as well as potential confounders, were identified. Details of these variables are provided in the supplementary methods (sMethod S1, Figure S1).

Figure 1:

CONSORT diagram for HiDenIC08 (A), and HiDenIC15 (B).

Outcomes

Our primary outcome was hospital mortality at 30 days. Secondary outcomes included 90-day and 1-year mortality, development of Kidney Disease Improving Global Outcomes (KDIGO) stage 2–3 AKI (using Cr and urine output criteria) within 72 hours after LVR, AKI recovery, defined as the absence of any KDIGO stage of AKI beginning on any day within 7 days of the first documented onset of AKI and sustained (stage 0 and no use of renal replacement therapy) though hospital discharge, and Major Adverse Kidney Events (MAKE) at 30, 90 days, and 1 year (sMethods S1).

Data Ascertainment and a priori Subgroup Analysis Plan

Variables such as serum creatinine, AKI,¹⁵ MAKE,¹⁶ acute physiology score (APS-III), and hypotensive index were computed using EHR abstraction as previously described.^17,18 The following subgroups were selected a priori for analysis in H15 and identified using International Classification of Disease (ICD9-DM) codes and septic patients using the sepsis-3 definition:¹⁹ cardiothoracic surgery (CT), sepsis, and liver disease. Analysis included the comparison of hospital mortality at 30 days and development of AKI between intervention groups.

Statistical Analyses

Categorical variables were summarized as number (percentage) and continuous variables were summarized as mean (SD) or median and interquartile range. After selecting variables using backward selection (p<0.10, sMethod S2–2a), models were built using multivariable logistic regression to compare treatments. Significance was set at the 0.05 level. Model fitness was assessed using the Hosmer-Lemeshow statistic and discrimination (area under the receiver operating characteristic curve). To further control for indication bias and confounding, we conducted an independent adjusted analysis using 1:1 nearest neighbor propensity score matching to the probability of receiving 5% albumin considering variables in sMethod S2–2b. Matches were created on logit scale without replacement using computation geometry based on distance between propensity scores (caliper 0.025). Ascertainment of covariates was determined using standardized difference of score between treatment and control. After matching, conditional logistic regression was performed with or without adjustment for total fluid volume. Survival analysis for time-to-mortality was performed on the propensity-matched cohort using Cox proportional hazards regression with robust standard errors to account for the induced correlation of the matched pairs. Length of stay was analyzed using zero-truncated negative binomial regression while adjusting for the total volume of fluid received. Assessment of residual confounding was performed calculating E-values for primary and secondary outcomes²⁰. Analyses were performed using SAS 9.4 (SAS Institute, Cary, NC, USA) and STATA 15.1 (Module “PSMATCH2”) (Stata Corp, TX, USA). Sample size calculations were performed using PASS Power Analysis and Sample Size Software 2020 (NCSS, LLC. Kaysville, Utah, USA, ncss.com/software/pass).

Results

Cohorts and Baseline Patient Characteristics

Among 45,568 encounters in H08 and 164,910 encounters in H15, we identified 4,710 and 39,918 patients who received LVR, respectively. For H08, the cohort included 2,428 patients after exclusions, of which 1,755 received saline, and 673 albumin (Figure 1A). For H15, the pre-propensity matched cohort included 16,201 patients after exclusions, of which 14,387 received saline, and 1,814 albumin (Table 1, Figure 1B). Median saline and albumin volumes within the 24-hour LVR window for patients in the saline and albumin groups were 2200 mL (IQR 1397–3000) and 750 mL (IQR 500–1250), respectively. Total volumes within the 24-hour LVR window were 4000 mL (2962–5330) and 4761 mL (3700–6353) for saline and albumin groups, respectively. Baseline characteristics for patients receiving LVR, stratified by fluid administered are shown in Table S1.

Unadjusted Outcomes

The use of albumin was associated with higher rates of AKI in both cohorts (Table 2). However, patients exposed to albumin had lower hospital mortality at 30 days in the H15 cohort (8.8% vs 10.9%; p<0.006, Table 2).

Table 2:

Unadjusted patient outcomes in both cohorts

Patient Outcomes in the HiDenIc-08 cohort
	Total	Saline	5% Albumin	p-value
No. of patients, N (%)	2,428 (100%)	1,755 (72.3%)	673 (27.7%)
In-Hospital use*
Mechanical Ventilation	1,706 (70.3%)	1,131 (64.4%)	575 (85.4%)	<0.001
Vasopressors Use	839 (34.6%)	543 (30.9%)	296 (44.0%)	<0.001
Mortality*
In-Hospital Mortality	499 (20.6%)	376 (21.4%)	123 (18.3%)	0.086
ICU Mortality	368 (15.2%)	276 (15.7%)	92 (13.7%)	0.21
Length of stay*
ICU	6 (4–12)	6 (4–12)	7 (4–13)	<0.001
Hospital	14 (8–25)	14 (7–23)	16 (10–29)	<0.001
Acute Kidney Injury (72h)**
No AKI	952 (44.5%)	799 (51.6%)	153 (25.9%)	<0.001
Stage 1	369 (17.3%)	253 (16.3%)	116 (19.7%)
Stage 2	590 (27.6%)	366 (23.6%)	224 (38.0%)
Stage 3	228 (10.7%)	131 (8.5%)	97 (16.4%)

Patient outcomes in the HiDenIc-15 cohort
	Total	Saline	5% Albumin	p-value
No. of patients, N (%)	16,201 (100%)	14,387 (88.8%)	1,814 (11.2%)
In-Hospital use*
Mechanical Ventilation	3,296 (20.3%)	2,898 (20.1%)	398 (21.9%)	0.073
Vasopressors Use	3,533 (21.8%)	2,896 (20.1%)	637 (35.1%)	<0.001
Mortality*
In-Hospital Mortality	1,726 (10.7%)	1,567 (10.9%)	159 (8.8%)	0.006
ICU Mortality	1,448 (8.9%)	1,315 (9.1%)	133 (7.3%)	0.011
Mortality by AKI recovery##
No AKI (n/total No AKI, %)	878/9,843 (8.9%)	832/9,016 (9.2%)	46/827 (5.6%)	<0.001
Recovery (n/Total AKI,%)	918/5,432 (16.9%)	823/4,558 (18.1%)	95/874 (10.9%)+	<0.001
No recovery (n/Total AKI,%)	431/730 (59%)#	379/633 (59.9%)$	52/97 (53.6%)$	0.27
Length of stay*
ICU	4 (2–9)	4 (2–9)	4 (2–8)	<0.001
Hospital	11 (7–18)	11 (7–18)	11 (8–19)	0.004
Acute Kidney Injury (72h)**
No AKI	10,093 (62.3%)	9,246 (64.3%)	847 (46.7%)	<0.001
Stage 1	2,181 (13.5%)	1,938 (13.5%)	243 (13.4%)
Stage 2	3,276 (20.2%)	2,687 (18.7%)	589 (32.5%)
Stage 3	651 (4.0%)	516 (3.6%)	135 (7.4%)
Major Adverse Kidney Events (MAKE)*
MAKE^30	3,258 (20.1%)	2,962 (20.6%)	296 (16.3%)	<0.001
MAKE90	4,417 (27.3%)	3,990 (27.7%)	427 (23.5%)	<0.001
MAKE365	6,121 (37.8%)	5,548 (38.6%)	573 (31.6%)	<0.001
Persistent Renal Dysfunction^*, Recovery from AKI***
30 day	767 (4.7%)	679 (4.7%)	88 (4.9%)	0.8
90 days	754 (4.7%)	657 (4.6%)	97 (5.3%)	0.14
365 day	842 (5.2%)	745 (5.2%)	97 (5.3%)	0.76

^{^:}For details, please refer to supplementary Methods ‘sMethodS1‘ section

^**Primary outcome,

^*Secondary outcomes,

^***For details please refer to the methods section.

^##In-hospital mortality.

^#Odds ratio for in-hospital mortality of No recovery vs. No AKI as reference: 27.9, p<0.01.

^$OR for in-hospital mortality of No recovery vs. No AKI as reference in saline and albumin groups: 27.4, p<0.01 and 39.7, p<0.01, respectively.

⁺No patients with AKI recovery in the albumin group died (thus an OR could not be calculated).

Primary & Secondary Outcomes

Mortality:

The use of 5% albumin was associated with lower adjusted hospital 30-day mortality (H08: aOR 0.65, 95% CI 0.49 – 0.85, p<0.002; H15: aOR 0.52, 95% CI 0.44–0.62, p<0.001), 90-day and 1-year mortality (Table S2, for further details see Table S3) as compared to the use of saline.

AKI:

Despite the beneficial effect on survival, the use of 5% albumin was associated with an increased risk of developing any AKI (aOR=1.9, 95% CI 1.6–2.5, p<0.001 in H08, and 1.8, 95% CI 1.6–2.0, p<0.001 in H15) and stage 2–3 AKI (aOR=1.6, 95% CI 1.3–2.1, p<0.001 in H08, and 1.9, 95% CI 1.7–2.1, p <0.001 in H15) within 72h after LVR (Table S2, for details see Table S4).

Survival Estimates and Cox Regression

Kaplan Meier unadjusted survival estimates in the propensity matched cohort showed lower 30-day and 1-year mortality in patients receiving 5% albumin compared to saline in both cohorts (Figure S3). We then performed covariate-adjustment from a Cox model for survival using H15. As in the unadjusted analysis, the use of 5% albumin was associated with higher 1-year survival (580 vs 465 patients, aHR 0.7, p<0.001, Figure 1, S4).

Propensity Matched Analysis and Estimates of Unmeasured Confounding

To control for indication bias, 1:1 propensity matching to the probability of receiving 5% albumin was used in both datasets. Adequate bias reduction and balance between covariates was attained (Figure S2). After matching, results from conditional logistic regression showed lower 30-day hospital mortality among patients receiving 5% albumin compared to saline (H08: OR 0.6, 95% CI 0.4–0.8, p<0.001; H15: OR 0.6, 95% CI 0.5–0.7, p<0.001), albeit higher odds of developing stage 2–3 AKI (H08: OR 1.6, 95% CI 1.2 – 2.0, p<0.001; H15: OR 1.7, 95% CI 1.5–2.0, p<0.001). These analyses were repeated by adjusting for total fluid intake within the first 72h of admission and showed similar results for both 30-day mortality and AKI (Table 3). E-values for the point estimate and upper limit of the confidence interval for hospital mortality at 30 days were 1.79 (1.81), for H08, and 2.12 (2.72) for H15 (Table S2). E-values for mortality at 90 and 365 days, and for the development of any or stage 2–3 AKI, are reported in Table S2.

Table 3:

Primary Outcomes in Propensity Matched Cohort

Cohort	Outcome	Pair (N)	Adjusted by	Conditional odds	95% CI	p-value
HiDenIc -08	Mortality at 30D	537	Unadjusted	0.6	0.4–0.8	<0.001
			Total Fluid Intake*	0.54	0.4–0.7	<0.001
	Severe AKI	461	Unadjusted	1.6	1.2–2.0	<0.001
			Total Fluid Intake	1.4	1.1–1.9	0.01
HiDenIc -15	Mortality at 30D	1,814	Unadjusted	0.6	0.5–0.7	<0.001
			Total Fluid Intake	0.6	0.5–0.7	<0.001
	Severe AKI	1,814	Unadjusted	1.7	1.5–2.0	<0.001
			Total Fluid Intake	1.8	1.6–2.1	<0.001

^*Total fluid intake within 72 hours

Major Adverse Kidney Events (MAKE) and AKI recovery

The use of 5% albumin was associated with a lower rate of MAKE at days 30, 90 and 365 compared to saline alone (Table 2). This effect was driven by mortality because differences in MAKE between groups disappeared when mortality was excluded (data not shown). Importantly, the use of 5% albumin was not associated with persistent renal dysfunction or a higher rate of new onset dialysis (Table 2). AKI recovery was not different between 5% albumin and saline (90.3% vs. 88.7%, p=0.11). The protective effect of albumin on mortality was larger in patients with AKI compared to patients without AKI (Table 2). Among patients with AKI, albumin was protective only in patients who recovered (Table 2).

Subgroup Analyses

We compared the effect of 5% albumin vs saline on 30-day mortality and stage 2–3 AKI in three predefined subgroups (cardiac surgery, sepsis, and liver disease) using H15. Although the mortality benefit of albumin followed the same direction as the primary analysis, this was only significant in the sepsis subgroup (Table S5). Overall, results were generally consistent with the primary analysis across subgroups as shown in Figure S5, Table S5 and in propensity matched cohorts Table S6.

Post-hoc Analyses

We compared fluid balance at 72 hours and 7 days, and ICU and hospital length of stay (restricting LOS analysis among survivors to avoid early death bias) between different fluid strategies. The use of 5% albumin was associated with lower 7-day fluid balance (−0.2L, (−3.8, 3.8) vs. 0.7L, (3.2, 4.8), p<0.001, Table S7), and shorter ICU LOS (incidence rate ratio: 0.91, 95%CI: 0.8–0.9, p=0.013, Table S8). Propensity score-matched analysis stratified by quartiles of total fluid administered within the LVR window showed that the protective effect of albumin on survival was consistent across strata (i.e., quartiles), with the exception of quartile 1. The risk of AKI associated with albumin was consistent across strata (Table S9).

Discussion

During large volume resuscitation, use of 5% albumin in addition to saline compared to saline alone was associated with a survival advantage at 30 days, 90-days and 1-year, but also with the development of AKI. These two opposing signals might coexist if albumin confers a benefit at the expense of producing more AKI. This would be analogous to the use of a known nephrotoxin such an aminoglycoside antibiotic, which might be both life-saving and detrimental to the kidney. However, it might also be possible that albumin makes AKI more apparent or that saline alone masks AKI. This could be true if larger volumes of saline were used such that changes in serum creatinine were obscured. This explanation is unlikely for two reasons. First, fluid balance in the first 72 hours was not very different between the two fluid strategies, just under 5ml/kg, and adjusting for this difference did not change our results. Second, differences in AKI rates were even seen with stage 2–3 AKI, which are less likely to be produced by changes in fluid volumes. A third, intriguing explanation is also possible. Although development of AKI is a known determinant of mortality in the critically ill,^21,22 recent data has shown that patients who can recover from an acute episode of AKI after acute illness have a similar 1-year mortality as patients who never developed AKI.²³ On the other hand, persistence of AKI seems to bear the worst consequences, as these patients have a significant increase in mortality.²³ Therefore, the distinction between transient or persistent episodes of AKI is crucial to the impact on outcome. In our study, 5% albumin use was associated with AKI at 72h after LVR, but not with persistent renal dysfunction, or with the need for hemodialysis at 30, 90 and 365 days. These findings suggest that 5% albumin may be associated with early, transient AKI rather than persistent dysfunction, which does not compete with the survival advantage. Supporting this, cumulative fluid balance at 1-week continued to favor albumin and increased to more than 10ml/kg.

The mechanism leading to a survival advantage with albumin is unclear. Although we suspect the advantage could have been secondary to an increase in the efficiency of resuscitation, that is achieving hemodynamic goals faster with less fluid, our datasets do not provide sufficient data to conduct such an analysis. The ALBIOS trial⁸ however, showed that patients resuscitated with albumin had a shorter time to achieve target MAP within the first 6 hours of resuscitation, and a shorter time to discontinuation of vasopressors or inotropic medications, suggesting that albumin was more efficient than saline at achieving resuscitation goals. Furthermore, patients in the albumin group had lower cardiovascular SOFA scores. Because these effects were achieved with similar volumes in both groups (i.e. albumin vs. saline), the authors suggested that the nitric oxide scavenging and anti-inflammatory properties of albumin could have played a role in reversing vasodilatation and reaching MAP targets faster. The ALBIOS study also found that, in the subgroup with shock, the use of albumin was associated with a lower mortality.

An important protective mechanism through which albumin could improve survival is the effect on the capillary endothelium, the integrity of the glycocalyx, and on microvascular flow. Jacob et al. has shown in an isolated perfused guinea pig heart model that albumin prevented extravasation of fluid to the interstitium more efficiently than crystalloid or hydroxyethylstarch.²⁴ They found that this effect was independent of colloid osmotic pressure, but was lost when vessels were treated with heparinase, an enzyme that breaks down the glycocalyx. This is important because using an isolated vessel preparation, Adamson et al. found that the colloid osmotic forces that oppose filtration from the intravascular space outward, were not across the vessel wall but rather, within the endothelial surface layer or ESL.²⁵ The ESL contains the glycocalyx, non-circulating plasma, and plasma proteins (particularly albumin). Similarly, Job et al. demonstrated in an in vitro model that albumin increased the thickness of the glycocalyx.²⁶ However, data on the effect of albumin on microcirculatory function in critically ill patients is still lacking, and more studies are needed.

The mechanism by which 5% albumin in addition to saline was associated with greater rates of AKI in our patients compared to saline alone is also unclear. Given that albumin is diluted in a balanced carrier, we expected to see a reduction in AKI similar to recent studies comparing balanced fluids to saline.^27,28 A high chloride load from the combination of 5% albumin (~128 mmol/L) and saline could account partially for the AKI signal in our data, however, because our comparator was saline, and the median albumin volume administered was less than 1 L, this is unlikely. Hyperoncotic fluid could theoretically reduce glomerular filtration and renal function by increasing intraglomerular oncotic pressure beyond the level of hydrostatic pressure, this is an unlikely mechanism to explain our findings because we only included patients receiving 5% albumin, which is iso-oncotic at best. Experimental studies have suggested that albumin filtered through the glomerulus can cause increased tubular²⁹ and interstitial inflammation,⁴ which could lead to AKI. However, in our study, AKI was transient, suggesting the absence of structural damage to the kidney and more likely, a functional alteration that corrected in time.

Our results were consistent across predefined subgroups including patients with sepsis, post-cardiac surgery and with liver dysfunction. Our data conflict with that of Lee et al., who showed in a single center RCT that administration of 20% albumin before off pump cardiac surgery was associated with an increase in urine output and a decrease in the rate of post-operative AKI.³⁰ However, the design of their study was significantly different from ours in that it was a prospective, randomized clinical trial. In addition, as they only included patients with serum albumin levels < 4 g/dL and randomized patients to receiving 20% albumin (not 5%) or saline before the surgical procedure, and not in the post-operative period.

Limitations

Our study has several limitations. First, our analysis involved patients treated with saline with or without 5% albumin. Recent trials^27,28 have raised significant doubts about the use of saline and more physiological fluids (e.g. lactated Ringer’s) may now represent the standard of care. Unfortunately, our database contained very few patients treated with both albumin and balanced crystalloids precluding this comparison. The retrospective nature of our analysis precludes any assertions about causality. However, sensitivity analysis using E-values shows that the observed OR of 0.65 for hospital 30-day mortality could only be explained by an unmeasured confounder with a risk ratio 2.12 above that of the confounders measured in this study, like age, gender, multiple comorbidities and diabetes. Only APACHE 3 scores >63 had higher OR than the E-value, suggesting it is unlikely that unmeasured confounders could explain the observed association. We did not have sufficient data to evaluate the impact of fluid resuscitation on the goals of resuscitation, on hemodynamics, acid-base balance or on tissue perfusion, which would have allowed us to explore the efficiency of fluid resuscitation more directly. Finally, we acknowledge that by virtue of the study design, the estimation of the effect of 5% albumin on mortality may be overestimated. Nevertheless, it is because of the strength of this association and the potential practice-changing impact that further exploration in randomized studies is necessary.

Conclusions

During large volume resuscitation, 5% albumin is associated with lower mortality when compared to the use of saline alone. However, albumin was also associated with a higher rate of AKI that did not translate into persistent renal dysfunction or into the need for dialysis. Further studies in large volume resuscitation are needed to better understand the effects of albumin on renal function and mortality.

Figure 2:

Adjusted survival (cox model) by treatment group in HiDenIC15