Hyperoncotic colloids and acute kidney injury: a meta-analysis of randomized trials

This meta-analysis addressed the question: does infusion of hyperoncotic 20% to 25% albumin or 10% HES to prevent or correct hypovolemia increase the risk of AKI as compared with crystalloid, hypo-oncotic 4% to 5% albumin, or no fluid? Mortality was a secondary endpoint.

Parallel-group randomized controlled trials (RCTs) were eligible for inclusion if they evaluated the occurrence of AKI in patients receiving hyperoncotic solutions of 20% to 25% albumin or 10% HES for prevention or correction of hypovolemia. The control regimen could consist of crystalloid, hypo-oncotic 4% to 5% albumin, or no fluid but not a non-fluid intervention such as an active drug or a procedure. Hypo-oncotic colloids were also included as control fluids in the CRYCO study [5]. Trials employing an iso-oncotic 6% HES control arm or comparing two hyperoncotic colloids were excluded. The difference in colloid osmotic pressure (COP) between hyperoncotic colloids and iso-oncotic 6% HES of 12 to 18 mm Hg is smaller than that between hyperoncotic colloids and hypo-oncotic 4% to 5% albumin [8, 9]; as a consequence of the smaller difference, AKI and mortality comparisons versus iso-oncotic 6% HES as a control fluid would likely be relatively insensitive. Colloids were classified as hyperoncotic on the basis of final concentration at the time of use; hence, cardiac surgery trials of extracorporeal circuit priming in which hyperoncotic colloid was extensively diluted prior to use were not eligible. No restrictions were placed on trial time period or reporting language. Both published and unpublished trials were sought.

Computer searches were performed between April and July 2010 in Medline, Embase, records of published and unpublished trials in the Cochrane Library, the ClinicalTrials.gov website, and the abstract databases from major meetings in surgery, anesthesiology, intensive care, and hepatology. The search terms of inclusion were the following: kidney, renal, renin, mortality, injury, failure, complication, adverse, illness, outcome, cirrhosis, albumin, hetastarch, pentastarch, pentaspan, hyperoncotic, 10%, 20%, 25%, clinical trials, prospective studies, fluid therapy, random allocation, and humans. The search terms of exclusion were cohort study, observational, survey, pharmacokinetic, retrospective studies, RCTs as topic, practice guidelines as topic, animal, rats, pigs, swine, review, news, letter, comment, editorial, meta-analysis, hypervolemia, molecular adsorbent recirculating system [MARS], MARS, albumin-bound, chemotherapy, paclitaxel, methotrexate, nanoparticle, and microsphere. Roots and variants of the search terms were also used. Eligible trials were also sought by examining the reference lists of primary study publications and review articles and contacting resuscitation fluid suppliers. All investigators participated in determining the eligibility of candidate trials. Differences in interpretation were resolved through discussion.

The authors, time periods, patients, and methods of each trial report were scrutinized to avoid duplication and ensure the most complete possible data set. From the reports of the included studies, the following data were extracted: year reported, number of patients, clinical indication, blinding, allocation concealment, patient age and gender, fluid regimen, criteria for diagnosing AKI, and incidence of AKI and death on the basis of intent to treat.

The pooled odds ratios (ORs) for AKI and mortality and their 95% confidence intervals (CIs) were computed under a random effects model. Heterogeneity was evaluated by Cochran Q test and calculation of the I² statistic and publication bias by linear regression of standardized effect in relation to precision. Study quality was judged on the basis of blinding and allocation concealment. Analysis was performed with Comprehensive Meta Analysis version 2.2.048 (Biostat, Inc., Englewood, NJ, USA) statistical software.

The number of candidate RCTs identified and screened was 109 (Figure 1). Upon detailed evaluation of the candidates, 98 were excluded. The most common reasons for exclusion were a non-hyperoncotic test colloid, an iso- or hyperoncotic colloid, a drug or procedure control, and an investigational design other than a parallel-group RCT. Eleven RCTs with a total of 1,220 patients fulfilled all eligibility criteria and were included in the meta-analysis [17‐27]. All trials had been published. Three of the 11 included trials were reported in the 1980 s and four each in the 1990 s and 2000 s (Table 1). Two trials were blinded and two were not, whereas the use of blinding was unspecified for the remaining seven trials. Allocation concealment was adequate in five trials and unspecified in six. The indication for volume expansion with colloid was treatment of ascites as an adjunct to paracentesis in three trials and to diuretics in one, surgery in three, and sepsis and spontaneous bacterial peritonitis in two each. One trial encompassed separate protocols for acute treatment of ascites in the hospital and subsequent long-term outpatient maintenance therapy [22]. Only the results from the acute treatment protocol were used in the meta-analysis. The median number of patients per trial was 72, and the interquartile range (IQR) was 32 to 116. More than 100 patients were evaluated in four trials, but only one trial involved over 200 patients. With 537 patients, that trial [25] was by far the largest included in the meta-analysis. Mean ages of patients in the trials were comparatively homogeneous. Whereas age averaged 49.1 years in one trial, the mean values in the other 10 trials fell within the relatively narrow range of 55.7 to 64.7 years. The median percentage of male patients in the included trials was 60% (IQR 60% to 65%). Seven trials evaluated hyperoncotic 20% to 25% albumin, and four trials evaluated hyperoncotic 10% HES. Control treatments were no colloid in seven trials, crystalloid in three, and hypo-oncotic colloid in one.

The AKI diagnosis criteria adopted in four trials included an increase of 50% or more in serum creatinine or blood urea nitrogen (Table 1). The need for RRT was the criterion in three trials, whereas other or unspecified criteria were applied in two trials each. Across all 11 studies, 199 of 1,220 patients (16%) developed AKI (Figure 2). As shown in Figure 2, hyperoncotic albumin decreased the odds of AKI by 76% (OR 0.24, CI 0.12 to 0.48; P < 0.0001). Among the seven trials evaluating hyperoncotic albumin, there was no evidence of either heterogeneity (P = 0.81; I² = 0%) or publication (P = 0.74) bias with respect to the AKI endpoint. Hyperoncotic HES showed the opposite effect (Figure 2), increasing the odds of AKI by 92% (OR 1.92, CI 1.31 to 2.81; P = 0.0008). Neither heterogeneity (P = 0.89; I² = 0%) nor publication (P = 0.55) bias was detectable in the AKI data from the four trials of hyperoncotic HES.

Among all 11 included trials, 283 of 1,220 patients (23%) died. At least one death occurred in 10 of the 11 included trials, allowing those 10 trials (with a total of 1,185 patients) to be included in a meta-analysis evaluating the OR for mortality (Figure 3). The only trial with no deaths assessed hyperoncotic albumin in the treatment of ascites [20]. Hyperoncotic albumin reduced the odds of mortality by 48% (OR 0.52, CI 0.28 to 0.95; P = 0.035; Figure 3). No significant heterogeneity (P = 0.81; I² = 0%) or publication (P = 0.12) bias was present regarding mortality in the six trials of hyperoncotic albumin with at least one death. Conversely, hyperoncotic HES raised the odds of mortality by 41% (OR 1.41, CI 1.01 to 1.96; P = 0.043). The mortality data in the hyperoncotic HES trials did not display either heterogeneity (P = 0.89; I² = 0%) or publication (P = 0.14) bias.