Long-term Impact of Crystalloid versus Colloid Solutions on Renal Function and Disability-free Survival after Major Abdominal Surgery

Colloids, including hydroxyethyl starch solutions, represent a potential option to optimize hemodynamics in high-risk surgical patients and to prevent excessive positive fluid balance with its harmful consequences.^1–5  The safety of modern hydroxyethyl starch solutions remains debated, even though these solutions have been used in many goal-directed fluid therapy studies.^6–10  Although some studies in the intensive care setting have reported potential nephrotoxic effects,^11–13  others, performed in a surgical context, have not.^14–17  However, the quality and size of surgical studies, in contrast with those performed in intensive care, is a matter of concern. In addition, the short follow-up periods represent a major limitation of these studies. Therefore, there is an urgent need for longer-term safety evaluation of such solutions.

We recently conducted a prospective, double-blind randomized controlled trial comparing a balanced hydroxyethyl starch solution to a balanced crystalloid solution for intraoperative goal-directed fluid therapy in patients undergoing elective major open abdominal surgery.¹⁸  Our results showed that when fluid resuscitation was standardized and guided by a closed-loop system to limit the risk of investigator bias, use of a balanced hydroxyethyl starch solution was associated with fewer postoperative complications at 30-days postsurgery than was use of a balanced crystalloid solution. These results were associated with a lower intraoperative fluid balance, mainly related to a lower intraoperative fluid administration in the colloid group. Renal function assessed by quantifying the incidence of either acute kidney injury using the Kidney Disease Improving Global Outcomes classification or the need for renal replacement therapy revealed no evidence of a difference between groups at postoperative day 30. Two of our secondary registered outcomes (ClinicalTrials.gov registry number NCT02312999) were the 1-yr evaluation of the renal function based on serum creatinine and estimated glomerular filtration rate and the disability score at 1-yr after surgery using the World Health Organization Disability Assessment Schedule 2.0 (WHODAS).

The present study aimed to perform 1-yr follow-up of the patients included in this trial regarding renal and disability outcomes. Our hypothesis was that there would be no significant difference between groups in estimated glomerular filtration rate at 1 yr postdischarge between the original study groups.

The original study¹⁸  was conducted in two academic institutions in Brussels (Brugmann and Erasme hospitals) from April 2015 through November 2016. The trial was approved by the ethics committees of Brugmann and Erasme hospitals and registered on ClinicalTrials.gov on December 5, 2014 (NCT02312999). Ethics committee approval of the original study also allowed us to follow patients after surgery to evaluate long-term renal outcomes and disability. All patients provided written informed consent before surgery.

Detailed inclusion and exclusion criteria, as well as anesthesia management, were described extensively in the original manuscript.¹⁸  In summary, 160 patients having elective major open abdominal surgery were enrolled. All patients received a maintenance infusion of balanced crystalloid at 3 ml · kg⁻¹ · h⁻¹ (Plasmalyte, Baxter, Belgium). An automated closed-loop system^19–21  delivered additional 100-ml fluid boluses according to a predefined goal-directed strategy using the uncalibrated pulse contour EV1000 monitor (Edwards Lifesciences, USA). Patients were randomized to receive boluses of either a balanced crystalloid solution (Plasmalyte) or a balanced colloid (hydroxyethyl starch) solution (Volulyte, Fresenius Kabi GmbH, Germany). The primary outcome of the study was the postoperative morbidity survey score at day 2 after surgery. Secondary outcomes included all postoperative complications at 30 days after surgery. Lower postoperative morbidity survey scores and lower postoperative complications were observed in the colloid bolus group.

For the present study, we followed up with all patients for 1 yr after surgery. We used our hospitals’ databases (Mediview and Mediweb software) to collect results of estimated glomerular filtration rate, Scr, and urea measures taken during postoperative outpatient follow-up appointments. The estimated glomerular filtration rate was calculated by the laboratory using the Chronic Kidney Disease Epidemiology Collaboration (CKD – EPI) formula in both hospitals.²²  Results of renal function were collected at five time points for each patient: T0, preoperative; T1, postoperative day 2; T2, last lab results before hospital discharge (may have occurred day 3 or later); T3, renal function test closest to 6 months postoperatively; and T4, renal function closest to 1 yr postoperatively. If there were no results in our database at any time point, the patient’s general practitioner was contacted by telephone to obtain the renal function variables.

We also assessed disability at 1 yr after surgery using the WHODAS, which has been shown to be a clinically acceptable, valid, and reliable instrument for measuring postoperative disability across different surgical populations.^23,24  The WHODAS questionnaire was completed by telephone with the patient (patient’s self-assessment). The WHODAS evaluates limitations in six domains (cognition, mobility, self-care, interpersonal relationships, work and household roles, and participation in society) over the 30 days before testing. Numerical values are attributed to each item (none = 0; mild = 1; moderate = 2; severe = 3; and extreme = 4). The total score is between 0 and 144 (36 questions on a 4-point scale) and is then divided by 144 and multiplied by 100 to convert it to a percentage of the maximum disability score. We classified disability based on the WHODAS and World Health Organization International Classification of Functioning, Disability and Health: none (0 to 4%); mild (5 to 24%); moderate (25 to 49%); severe (50 to 95%); and complete (96 to 100%) disability.²³  We calculated disability-free survival as 100% − dead (%) − moderate disability (%), where moderate disability is a WHODAS score of more than 25%. The presence of pruritus was collected during the same phone call at 1 yr after surgery. Data were collected by J.M., who was not involved in the earlier publication and was blinded to study group allocation at the time of follow-up.

Main outcomes measures were estimated glomerular filtration rate using the CKD – EPI formula in both hospitals,²²  serum creatinine, urea, pruritus, and WHODAS score, all of which are secondary outcomes according to our original study registration. We chose a statistical approach that accounts for missing data because of patient deaths and patients lost to follow-up. First, data were tabulated and compared directly on a complete-case-analysis basis, using chi-square tests for tabular/binary outcomes and Mann–Whitney U tests for scalar data, because there is a rationale for complete-case-analysis reporting even when other imputation methods are used, particularly when binary outcome measures (like death) are involved.^25,26 

Then to account for missing data from patients lost to follow-up, missing data were first assessed for being missing completely at random versus missing at random or missing not at random,²⁷  using the methodology of Jamshidian and Jalal.²⁸  If the data were missing completely at random, then complete-case-analysis would likely be relatively unbiased and sufficient for analysis; if they were not missing completely at random, then the data would be assumed to be missing at random, and multiple imputation has been shown to provide more unbiased effect estimates than complete-case-analysis in this setting.^29,30 

When indicated, multiple imputation was performed by chained equations,^31,32  and the effect of fluid type was assessed using mixed-model regression and raw time-to-follow-up data for the assessment time points for each patient, with log transformation of the dependent variables (because the target outcome variables were heavily skewed). The mixed-model effects included were chosen a priori by the authors as those likely to be associated with long-term renal and disability outcomes: group, age, body mass index, baseline renal labs, surgery type, surgery duration, estimated blood loss, Physiological and Operative Severity Score for Enumeration of Mortality and Morbidity (POSSUM) physiology score, POSSUM morbidity score, and total intraoperative fluid volume. To test the stability of the imputed missing data in the final model, additional models were run using the same multiple imputations replacement, maximum likelihood estimation imputation, complete-case-analysis, and last observation carry-forward imputation with the times of observations discretized into preoperative, postoperative day 2, last-before-discharge, 6 months, and 12 months.

The ideal means of handling of missing data caused by patient death is still debated and likely depends on the nature of the measure in the context of the overall study.^33,34  For the present follow-up study in high-risk abdominal surgery, 71% of the original cohort had had cancer operations (more than half of these procedures were for pancreatic or biliary cancer) and another 10% had had high-risk vascular procedures, all of which carry a high year-on-year risk of death. For the present analysis, it was presumed that deaths that occurred were a result of the original presenting indication for surgery and were therefore missing at random relative to the intraoperative fluid-therapy groupings. These cases were thus included in the same process as the “missing due to lost follow-up” patients for the principal analyses. This assumption may introduce bias if deaths are attributable to group assignment. Therefore, a secondary sensitivity analysis was performed imputing all missing data for deceased patients using the worst observed value in the cohort as the replacement value to assess the possible impact of this assumption on the conclusions of the analysis.

No specific power analysis was performed because this was a secondary analysis of a parent trial.¹⁸  According to the primary outcome of the parent trial, which was a 1-point difference in the postoperative morbidity survey score at postoperative day 2, 80 patients in each group were enrolled. All of them were included in the present analysis. Statistical analyses were performed using R (www.r-project.org) and SPSS (IBM Corp., USA). Scalar data are reported as median and [25th to 75th] percentiles. Differences in groups are described by location difference (the median of the difference between two samples). Significance testing was two-tailed, and CIs are reported at the 95th percentile. Outliers were evaluated, but no action was found to be necessary.

Of the 160 patients enrolled in the original study, follow-up data were obtained for 129 (81%) for long-term renal function. Sixteen (10%) patients died before the 12-month follow-up, and 15 (9%) were completely lost to follow-up (no data obtained after discharge, no hospital or provider visits, and no response at any phone number despite repeated attempts). Complete data for WHODAS score at 12 months were obtained for 114 patients (71%). (Some patients had documented lab results from primary care during the follow-up period but did not respond on available contact numbers or addresses to obtain WHODAS results.) The study flow chart is shown in figure 1.

Of the confirmed deaths in the 12-month follow-up period, 10 occurred in the crystalloid group, and 6 occurred in the colloid group (P = 0.429; table 1). Eight additional patients were completely lost to follow-up in the crystalloid group, and seven were lost in the colloid group (P > 0.999). Pairwise testing of group differences in serum creatinine, urea, and estimated glomerular filtration rate showed no significant differences at both the 6- and 12-month follow-ups (table 1). Figures 2 and 3 show the estimated glomerular filtration rate and serum creatinine values in the two groups at the five time points. Appendix 1 provides the incidence of chronic kidney disease (and stages) at 12 months in the two groups. At 1 yr, disability as assessed by the WHODAS score was statistically significantly lower in the colloid than in the crystalloid group (2.7 [0 to 12%] vs. 7.6 [1.3 to 18%]; P = 0.015). Disability-free survival (table 1) was significantly higher in the colloid group (79%) than in the crystalloid group (60%) at 1 yr as well (P = 0.024; 95% CI, 2 to 39%).

Analysis of the pattern of the missing data using Hawkins test revealed a P value of less than 0.001, and the nonparametric test of homoscedasticity revealed a P value of 0.027. Data missing completely at random were thus rejected, and the data were treated as missing at random, and multiple imputation by chained equations was performed to impute the missing values. The results for main effect of group on 6- and 12-month follow-up labs in the models using different approaches to the missing data are shown in table 2 along with the 95% CI for group effect at 12-month follow-up. Testing using exact time to follow-up showed a nonsignificant effect of fluid group on long-term estimated glomerular filtration rate, urea, and serum creatinine values (table 2). Sensitivity analysis showed that these results were stable across alternate imputation/analysis methods using discretized time handling (table 2) and under replacement of deceased patients with worst observed measurement (table 2). With the multiply imputed data set, there was a nonstatistically significant difference in WHODAS scores in the crystalloid group compared with the colloid group (P = 0.059; 95% CI, 0 to 8). This effect was statistically significant under maximum likelihood estimation imputation (P = 0.016; 95% CI, 5 to 9). Complete-case analysis (P = 0.317; 95% CI, −3 to 10) and replacement of deceased patients with worst-observed value (P = 0.185; 95% CI, −3 to 15) did not show a statistically significant effect.

In the long-term follow-up of a double-blind, randomized controlled trial of fluid resuscitation in patients undergoing major open abdominal surgery, we found no evidence for a difference in renal function at 1 yr in patients who received a balanced crystalloid and those who received a balanced hydroxyethyl starch solution. However, patients in the hydroxyethyl starch group had a statistically significantly lower disability score after surgery and a significantly higher rate of disability-free survival. To the best of our knowledge, this study is the first to report long-term (at least 1 yr) effects of balanced hydroxyethyl starch versus balanced crystalloid administration in patients undergoing major abdominal surgery. Another recent study reported no change in 1-yr renal function in patients undergoing cardiac surgery who had received balanced hydroxyethyl starch for fluid resuscitation,³⁵  but there was no control group. In the intensive care setting, some studies have suggested a higher incidence of renal failure in critically ill patients resuscitated with hydroxyethyl starch compared with those resuscitated with crystalloids,^11,12  but renal function was not assessed beyond 90 days.

Our results support several recent systematic reviews and meta-analyses showing no evidence for a deleterious effect of hydroxyethyl starch solutions on short-term renal function in surgical patients.^36–39  Recently, Kammerer et al.¹⁴  reported comparable renal safety profiles of hydroxyethyl starch and 5% albumin in more than 100 patients undergoing major urologic surgery, including the ratio of serum cystatin C, estimated glomerular filtration rate, and neutrophil gelatinase-associated lipocalin up to 90-days postoperatively.

The raw disability score assessed by the WHODAS questionnaire was significantly lower at 1 yr in the hydroxyethyl starch group compared with the crystalloid group, and more relevant to patients themselves, there was a higher rate of disability-free survival in the colloid group. These results were persistent in the maximum likelihood estimation imputation but not in the complete-case-analysis model or when deceased patients were replaced with the worst-observed score (the multiple imputation model is equivocal). Thus, caution is advisable in interpretation of a benefit to patients in the colloid group, but our results show no evidence at all of worsened outcomes in the colloid group. Using the MacNew questionnaire^40,41  we recently reported no deleterious effect of hydroxyethyl starch solution on quality of life in cardiac patients.³⁵  In the present study, we chose the WHODAS questionnaire, because the MacNew questionnaire was only designed to assess quality of life in cardiac patients. The impact of administration of hydroxyethyl starch solution on quality of life in the critically ill population appears more controversial. Wittbrodt et al.⁴²  reported a worse self-perceived quality of life in hydroxyethyl starch–treated patients compared with crystalloid-treated patients, but Taylor et al.⁴³  observed no difference in long-term quality of life between hydroxyethyl starch– and saline-resuscitated critically ill patients. The reasons for these differences are not clear; one possible explanation might be related to the degree of inflammatory response in these different populations, with its known effects on the glycocalyx and capillary permeability.⁴⁴  The lower long-term disability score observed in our hydroxyethyl starch group is in accordance with the lower incidence of 30-day postoperative complications reported in our initial publication.¹⁸  An association between the incidence of immediate postoperative complications and the development of worse long-term outcomes has been previously reported by Khuri et al.⁴⁵  and Moonesinghe et al.⁴⁶ 

The double-blind design represents the main strength of this study. However, the study also has some limitations. First, the population was relatively small, and renal function was assessed only by estimated glomerular filtration rate and serum creatinine. The present study was not powered to detect differences in long-term renal function, because this was a secondary analysis of a parent trial with the postoperative morbidity survey score at postoperative day 2 as primary objective. The relatively small sample size may therefore lead to inconclusive 95% CIs. Second, because patients with preexisting elevated serum creatinine (more than 2 mg/dl) were excluded from the study, our results may not be generalizable to patients with mild preoperative renal insufficiency. Third, we did not have a baseline preoperative WHODAS score that could be considered as a significant limitation. Rates of disability (or more specifically changes in disability) might be underestimated or misinterpreted, although age, American Society of Anesthesiologists physical status, and POSSUM scores were well balanced between both groups. We also acknowledge that differences in the WHODAS score and disability-free survival may not be directly attributable to the different solutions used in our study. Finally, we considered a 10 ml min⁻¹ 1.74 m⁻² difference in estimated glomerular filtration rate to be clinically significant. The 95% CIs for estimated glomerular filtration rate difference at 6- and 12-month follow-up (table 1) include differences of 10 or greater, so we cannot completely exclude the possibility that the colloid group may have had worse long-term renal function by this measure.

In patients undergoing major open abdominal surgery, there was no evidence of a statistically significant difference in long-term renal function between a balanced hydroxyethyl starch and a balanced crystalloid solution used as part of intraoperative goal directed fluid therapy, although there was only limited power to rule out a clinically significant difference. However, disability-free survival was significantly higher in the colloid than in the crystalloid group.

The authors thank the Departments of Anesthesiology from Erasme Hospital, Brussels and Brugmann Hospital, Brussels, Belgium for their help in this trial.

Supported by the European Society of Intensive Care Medicine Baxter Award 2015 and grants from the Belgium Society of Anesthesiology (2015) and the Brugmann Foundation (2015).

Dr. Joosten is a consultant for Edwards LifeSciences, Irvine, California. Dr. Van der Linden has received, within the past 5 yr, fees for lectures and consultancies from Fresenius Kabi GmbH, Bad Homburg, Germany, and Janssen-Cilag SA, Beerse, Belgium. Dr. Cannesson and Dr. Rinehart both have ownership interest in Sironis, Newport Beach, California, a company developing closed-loop systems, and have consulted for Edwards Lifesciences, Irvine, CA. The other authors declare that they have no conflicts of interest concerning this article.

Full protocol available at: Alexandre.Joosten@erasme.ulb.ac.be. Raw data available at: Alexandre.Joosten@erasme.ulb.ac.be.