Impact of a Potassium-enriched, Chloride-depleted 5% Glucose Solution on Gastrointestinal Function after Major Abdominopelvic Surgery: Results of a Randomized Controlled Trial

INCREASING focus is being placed on enhanced recovery after surgery (ERAS) protocols for various major surgical procedures aiming to speed up recovery and reduce the length of hospital stay (LOS).^1,2  One of the cornerstones of ERAS protocols is perioperative fluid management with avoidance of salt and water overload.²  In a previous randomized trial, we have shown that a simple hydration protocol focusing on a zero postoperative fluid balance significantly reduces overall major complications and gastrointestinal (GI) complications by more than 50%.^3,4 

GI recovery is a critical factor determining postoperative recovery and consequently LOS. Electrolyte disturbances can negatively impact GI function, especially low potassium and magnesium levels. A relevant number of patients receiving a balanced physiologic crystalloid solution (Ringer’s maleate) perioperatively experience serum electrolyte disturbances in the early postoperative period after cystectomy and urinary diversion and need substitution. Excessive administration of normal saline (NaCl, 0.9%) is associated with higher GI and renal complication rates compared to balanced solutions.^4–6  Thus, optimization of perioperatively administered crystalloid solutions is required.

Open radical cystectomy combined with pelvic lymph node dissection and urinary diversion was chosen as a major surgery model for this study because it is an extensive and time-consuming, but standardized, surgical intervention in a high-caseload tertiary center. In addition, despite improvements in surgical technique and perioperative care, radical cystectomy is associated with early postoperative complication rates of more than 50% and a 90-day mortality rate of 2 to 7%.^3,7 

We hypothesized that the perioperative administration of a novel potassium-enriched, chloride-depleted 5% glucose solution (G5K) improves GI recovery while reducing electrolyte disturbances when compared to a balanced Ringer’s maleate solution.

The study was approved by the local ethics committee (Kantonale Ethikkommission Bern, Bern, Switzerland; KEKBE 151/13) and by the Swiss Agency for Therapeutic Products, Bern, Switzerland (2014DR4097). It was prospectively registered at http://www.controlled-trials.com (ISRCTN32976792; principal investigator: Dr. Wuethrich; date of registration: October 30, 2013) and conducted in compliance with the Declaration of Helsinki and Good Clinical Practice. All patients gave previous written informed consent.

This is a prospective, randomized, parallel-group, assessor- and patient-blinded, high-caseload, single-center interventional superiority trial conducted at the Department of Urology, University Hospital Bern, Bern, Switzerland.

Consecutive patients presenting for open radical cystectomy, pelvic lymph node dissection, and urinary diversion (ileal conduit, orthotopic bladder substitute, continent catheterizable ileal reservoir) were screened for eligibility and recruited from July 2014 to May 2015. Inclusion criteria were age more than or equal to 18yr and American Society of Anesthesiologists’ physical status II or III. Exclusion criteria were pregnancy, congestive heart failure (New York Heart Association classification more than or equal to 3), severe hepatic disease, and estimated glomerular filtration rate less than 45 ml/min.

Patients were prospectively randomized 1:1 by a computer-generated list with 11 blocks of 4 patients. The random allocation sequence was implemented by a study coordinator blinded to the study codes and was generated by the hospital pharmacy. Patients were enrolled by the research coordinator or the senior surgeon. Patients were included in strict numerical order and assigned to the group mentioned in the sealed, nontransparent envelope with the corresponding number. The similar-looking infusion bags were specially prepared and provided by the hospital pharmacy in accordance with good manufacturing practice. The investigators who assessed the return of GI function and performed the statistical analysis were blinded to the randomization.

Preoperatively, no antegrade bowel preparation was administered, but two high enemas were given the evening before surgery. Patients had oral intake till midnight before surgery and were encouraged to drink clear fluid till 2 h before anesthesia induction.

Surgery was performed in a standardized fashion as previously described with the patient in a 30° head-down position and with one of three senior urologists present.^8–10  A gastrostomy tube was placed intraoperatively, and the orogastric tube was removed at the end of the procedure. The ureteral stents were exteriorized.

In the intervention group (G5K group), patients received a potassium-enriched, chloride-depleted 5% glucose crystalloid solution (G5K solution, Bichsel, Switzerland) and in the control group, a balanced Ringer’s maleate solution (Ringerfundin^®, B. Braun, Switzerland) as a baseline infusion during the entire period requiring intravenous fluid administration. The main differences in electrolyte concentrations between the two solutions (G5K vs. Ringerfundin^®) were as follows: chloride (65.0 vs. 127.0 mmol/l), sodium (50.0 vs. 145.0 mmol/l), potassium (30.0 vs. 4.0 mmol/l), and magnesium (2.0 vs. 1.0 mmol/l) concentrations. The G5K solution contained in addition 50 g/l glucose (277.4 mmol/l), lactate 18.0 mmol/l, and phosphate 8.0 mmol/l (table 1).

Intraoperatively, the assigned solution was administered at a rate of 1 ml kg⁻¹ h⁻¹ until the bladder was removed, followed by 3 ml kg⁻¹ h⁻¹ until the end of surgery. Norepinephrine was titrated as needed from 2 to 8 μg kg⁻¹ h⁻¹ to maintain a mean arterial pressure of 60 to 100 mmHg. If this was not sufficient to correct hypotension less than 60 mmHg, boluses of 250 ml of Ringer’s maleate solution were infused in both groups. Blood loss exceeding 500 ml was compensated with an equal amount of Ringer’s maleate solution. Packed erythrocytes were transfused if hemoglobin values dropped less than 80 g/l (less than 100 g/l in patients with coronary artery disease). If hypotension persisted or if severe metabolic acidosis (base excess less than −5, pH less than 7.25) caused by hypovolemia occurred, additional boluses of Ringer’s maleate solution were given in both groups. Patients with one to two risk factors for postoperative nausea and vomiting (PONV) were prophylactically treated with antiemetics (ondansetron).

Postoperatively, patients received 1,500 ml of the crystalloid solution allocated by randomization per day until oral substitution was adequate. The crystalloid solutions were administered using a pump with limited volume per time because of the relatively high potassium concentration in the G5K solution. Patient management adhered to our standardized care pathway.¹¹  Patients were allowed to drink clear fluids immediately after surgery, were encouraged to chew gum, and were started on an oral liquid diet on postoperative day (POD) 1. Subcutaneous 0.5 mg neostigmine daily was administered from POD 2 on under cardiac monitoring. A gastrostomy tube was initially left on drainage at bed level, closed if patients were without nausea and vomiting for more than 24 h, and removed once patients passed stool. Antiemetic therapy (intravenous ondansetron and/or droperidol) was initiated if patients experienced PONV.

Electrolyte substitution was performed according to our internal guideline in case of hypokalemia (K⁺ less than 3.5 mmol/l), hypomagnesemia (Mg⁺⁺ less than 0.66 mmol/l; both with intravenous substitution), and moderate hyponatremia (Na⁺ less than 130 mmol/l; oral substitution). In case of hyperglycemia (more than 10 mmol/l), patients were treated with subcutaneous insulin injection (4 IU).

Bedside mobilization was encouraged as soon as possible, ideally the evening after surgery, or not later than the next morning. Active ambulation, including exercising in bed, sitting out of bed, and standing and walking in the room was started on POD 1, and prolonged mobilization as well as sitting in a chair were started on POD 2. Epidural analgesia remained until POD 5 combined with an opioid-free systemic analgesia (paracetamol and metamizol).

Primary endpoint was the time from end of surgery until first defecation (first portion of feces). Secondary endpoints were time to first flatus and normal GI function defined as defecation of a portion of normal stool (Bristol form scale 3 to 4).¹² As first defecation is the most widely used endpoint in GI recovery, we based our power analysis on this. However, we additionally considered the defecation of a normal portion of stool as a relevant secondary endpoint (i.e., return of normal GI function), as first defecation can be defecation of residual stool (i.e., rectal emptying in case of no preoperative bowel preparation) due to stimulation or paradoxical diarrhea.¹³  This endpoint was added after registration but before recruitment of the first patient.

Patients were instructed to report the occurrence of flatus and defecation, which was recorded twice daily by the study nurse. Stool consistency and odor were also assessed by the ward nurse and documented by a study nurse. Bodyweight was measured every morning at the same time. Postoperative ileus (POI) was defined as no return of bowel function after POD 6 requiring cessation of oral intake, intravenous support, or nasogastric tube placement.¹⁴  Incidence of PONV and antiemetic use were recorded.¹⁵ 

In addition, postoperative fluid balance (i.e., difference in bodyweight) and event of renal dysfunction according to the Acute Kidney Injury Network and Risk, Injury, Failure, Loss and End-stage kidney disease classifications were assessed.^16–18 

Safety endpoints included measurement of biochemical parameters (sodium, potassium, chloride, magnesium, hydrogen phosphate, osmolality, brain natriuretic peptide, renin, and aldosterone), urine electrolytes (sodium, chloride), and osmolality during the first 4 PODs.¹⁹  Plasma and urine samples were collected every morning at the same time (5:00 am). Dyselectrolytemia was defined as hyperchloremia (plasma value more than 107 mmol/l), hypo- or hyperkalemia (more than 4.5 or less than 3.5 mmol/l), hyponatremia (less than 135 mmol/l), or hypomagnesemia (less than 0.66 mmol/l). Perioperative normoglycemia was defined as glucose plasma levels between 4.5 and 10.0 mmol/l. Hypoosmolality was defined as lower than 280 mOsm/kg.

Based on internal retrospective data in a similar surgical population (time until first defecation, 4.82 days; SD, of 0.82 day), we calculated that a sample size of 18 patients per arm randomized 1:1 would have a 90% power (β = 0.10) to detect a difference of 1 day between the groups at a two-sided significance level of 5% (α = 0.05) assuming a SD of 1 day. Presuming a drop-out rate of 20%, 22 patients per group were recruited.

Statistical analyses were conducted on an intention-to-treat basis. Data are expressed in medians with ranges for continuous variables or frequencies for categorical ones. Categorical data were compared with the Fisher exact or the chi-square test and continuous data with the Mann–Whitney U test for comparison of the two independent groups because of small sample sizes. Group differences in the primary outcome were evaluated using the Mann–Whitney U test for two independent groups. Nonparametric 95% CIs with Hodges–Lehmann (HL) estimator were used for differences of the two group medians.

A two-sided P < 0.05 was considered significant. Statistical analysis was performed by the Institute of Mathematical Statistics and Actuarial Science, University of Bern (Bern, Switzerland) using the Statistical Analysis System software (version 9.3; SAS Institute, USA).

Of 53 consecutive patients, 44 fulfilled the eligibility criteria, were randomized, and had complete follow-up data for the final analysis (fig. 1). Baseline demographic characteristics did not differ significantly between the groups (table 2).

There was no statistically significant difference between groups regarding surgical characteristics, LOS, intraoperative parameters, and fluid administration (G5K group: 750 ml [range, 500 to 1,700 ml] vs. control group 975 ml [400 to 1,600 ml]; P = 0.185). The amount of fluid administered postoperatively (G5K group: 4,750 ml [4,000 to 6,000 ml] vs. control group 5,250 ml [4,000 to 6,000 ml], P = 0.941) and fluid balance did not differ significantly between the groups during the first 4 PODs. There was a significant difference in the salt load administered between the two groups (table 3).

Time to first flatus did not differ significantly between groups: G5K group (44 h [13 to 118 h]) and control group (50 h [22 to 114 h]); HL estimator of the group difference, −2.5 (95% CI, −22 to 13); P = 0.716. Time to first defecation did not differ significantly between groups: G5K group (93 h [19 to 168 h]) and control group (120 h [43 to 241 h]); HL estimator of the group difference, −16 (95% CI, −38 to 6); P = 0.173. Time to return of normal GI function was significantly shorter in the G5K group (138 h [54 to 262 h]) than in the control group (169 h [108 to 318 h]); HL estimator of the group difference, −38 (95% CI, −74 to −12); P = 0.004 (fig. 2).

There was no significant difference in the incidence and duration of POI between groups: G5K 4.5% (1/22 patients), 3 days, and control 9.1% (2/22 patients), 3 and 4 days; P = 1.000. The incidence of at least one episode of PONV during PODs 1 to 4 was 22.7% (5/22 patients) in the G5K group and 45.5% (10/22) in the control group; P = 0.203. All patients with episodes of PONV received antiemetics (intravenous ondansetron and droperidol).

There was no significant difference in median plasma creatinine values and estimated glomerular filtration rate between the groups postoperatively. Median potassium and magnesium plasma values were significantly lower in the control group on PODs 3 and 4 even if a Bonferroni correction for multiple testing (6 time points) of these components is applied. Median sodium plasma values were significantly lower in the G5K group on PODs 3 and 4, and median sodium urine value was significantly lower on POD 4 even if a Bonferroni correction for multiple testing (6 time points) of this component is applied. Median chloride plasma values were significantly lower in the G5K group.

Urinary sodium values were significantly lower in the G5K group 6 h after surgery and on POD 4 even if a Bonferroni correction for multiple testing (6 time points) of these components is applied. Urine chloride values were only significantly lower in the G5K group 6 h after surgery. Urine osmolality did not differ significantly between groups (table 4).

The rates of potassium and magnesium intravenous substitution were significantly higher in the control (54.5 and 77.3%) than in the G5K (13.6 and 18.2%) group, P = 0.010 and P < 0.001, respectively. There was no difference between groups for any of the renal dysfunction classifications (table 5).

At least 1 episode of mild hyponatremia was present in 13 of 22 patients (59.1%) in the G5K group and in 4 of 22 patients (18.2%) in the control group; P = 0.012 (table 5). However, overall only five patients in the G5K and one in the control group had a sodium plasma value under 134 mmol/l from POD 1 to 4.

Plasma glucose values were similar between the two groups (table 4). Episodes of hyperglycemia were present in 3 of 22 patients (13.6%) in the G5K group (measured 6 h after surgery, treated with a single subcutaneous injection of 4 IU insulin per patient) and in 6 of 22 patients (27.3%) in the control group (measured 6 h after surgery [4 IU insulin per patient] and on POD 1 [4 IU insulin per patient]).

The in-hospital complication rate did not differ between groups with the exception of a higher rate of surgical complications in the control group (31.8%: 2 lymphoceles, 4 wound dehiscences, and 1 ureteral anastomotic leak) compared to the G5K group (4.5%: 1 lymphocele); P = 0.046 (table 5).

In addition, we found four patients (18.2%) with neurologic complications in the G5K group (one transient ischemic attack on POD 2 due to a preoperatively not diagnosed internal carotid artery stenosis and three transient sensomotoric dysfunctions of the obturator nerve after surgical lymph node dissection). In the control group, we found two patients (9.1%) with neurologic complications (one transient confusion on POD 1 [with normal plasma sodium level at this time point] and one sensomotoric lesion of the femoral nerve due to positioning).

We were unable to detect a statistically significant difference in the time from end of surgery until first defecation (first portion of feces), our primary endpoint. However, a relevant finding of this prospective randomized study is that the return of normal GI function (secondary endpoint) was accelerated by perioperative administration of a potassium-enriched, chloride-depleted 5% glucose crystalloid as a perioperative maintenance solution when compared to the usually used balanced crystalloids. In addition, the need for postoperative intravenous potassium and magnesium substitution was reduced in the G5K group, suggesting fewer perioperative electrolyte shifts in this group.

There is some evidence that the quantity of crystalloid solution administered perioperatively influences postoperative rehabilitation and morbidity as well as having an impact on postoperative recovery.^4,5,20  A positive salt and water balance delay return of GI function in patients undergoing colorectal surgery or cystectomy,^4–6  while patients with no weight gain on POD 1 have fewer overall and GI complications.^5,6  However, not only the quantity of crystalloid solution but also the type of electrolytes in the crystalloid solution infused is of relevance. Normal saline solution (i.e., 0.9% NaCl), still the most commonly used crystalloid solution worldwide, contains excessive amounts of chloride and sodium. Administration of large amounts of saline solution has been associated with increased postoperative morbidity compared to more physiologic balanced crystalloids.²¹  Saline solution overload results in metabolic acidosis, with decreased renal blood flow, renal dysfunction, prolonged GI recovery time, and increased infectious complication rates compared to balanced crystalloid solutions.^6,21–24 

After having demonstrated, in the context of an ERAS protocol,²  that a fluid management scheme based on a zero postoperative weight gain dramatically reduces complications and LOS, we now compared a balanced Ringer’s maleate solution to a novel potassium-enriched (30 mmol/l) and chloride-depleted (65 mmol/l) 5% glucose–based crystalloid solution.⁴  Both groups had a similar fluid balance of around zero (i.e., no weight gain postoperatively) as a prerequisite according to our previously published protocol.⁴  Using the novel solution, return of GI function was significantly reduced by more than 1 day; this is relevant and comparable to the published outcomes in other studies, which focused on optimizing postoperative GI function.²⁵  The substantially higher sodium and chloride load in the control group but similar urinary sodium and chloride concentrations postoperatively when compared to the G5K group suggest a relevant sodium and chloride retention in the control group. This is possibly explained by the kidneys’ limited ability to excrete water and electrolytes perioperatively as a response to surgical trauma and postoperative stress reaction.²⁴  The observed difference in perioperative electrolyte load could be an explanation for the delayed return of GI function and the nonsignificant trend toward a longer time to first flatus and defecation in the control group. Both hypokalemia, which occurred in the control group, and hyponatremia caused by fluid overload have been reported to be risk factors for delayed return of normal GI function and POI^26,27  by impairing intrinsic small bowel myogenic and neurogenic activity. Although patients in the G5K group had lower sodium plasma values and a higher incidence of mild hyponatremia, osmolality and fluid balance did not differ from the control group. This suggests a compensated fluid load and possibly explains why mild hyponatremia did not affect GI recovery. The higher incidence of mild hyponatremia in the G5K group had no apparent adverse effect, and it was not associated with fluid retention. It is likely that impaired GI function recovery as a consequence of hyponatremia, which has been reported by some studies, is more likely caused by the concomitant fluid overload than by the hyponatremia per se.

Magnesium is another factor potentially affecting recovery of GI function.²⁸  As the G5K solution contains twofold more magnesium than the balanced Ringer’s maleate solution (Ringerfundin^®) and patients in the control group required significantly more magnesium replacement postoperatively, this could be another factor favoring return of GI function in the G5K group.

Intracellular uptake of glucose (50 g/l in the G5K solution) and at the same time by uptake of water through osmosis probably results in a less pronounced interstitial edema in the GI tract. Due to the 5% glucose content and low sodium and chloride concentration, the G5K solution is more likely to shift water from intravascular to interstitial to intracellular, than a similar quantity of balanced Ringer’s maleate solution would (i.e., intravascular to interstitial). Interstitial edema, due to fluid overload or inadequate crystalloid choice (saline solution), is another factor known to affect GI function and cause prolonged POI.^29–31 

The goal of an intravenous maintenance solution is to substitute the daily needs for water and electrolytes and to reduce ketosis starvation by giving 50 to 100 g glucose, 1 mmol kg⁻¹ d⁻¹ potassium, and 1mmol kg⁻¹ d⁻¹ sodium.³²  These requirements are better met by the G5K crystalloid solution than the balanced Ringer’s maleate solution, a solution containing no glucose and substantially less potassium. Therefore, it is not surprising that patients in the control group needed significantly more potassium substitution and had statistically but not clinically relevant higher chloride and sodium plasma values. This is an important finding, as hypokalemia is another factor known to impair GI recovery.³³ 

Established postoperative strategies to accelerate return of GI function and reduce POI have already been established in patients undergoing radical cystectomy with urinary diversion: use of epidural analgesia to avoid opioid use, alvimopan administration, chewing gum, and early oral intake are examples.^15,25  The amount of perioperative fluid and the electrolytes substituted is considered another key element, albeit still controversial. This study suggests that the electrolyte composition of a balanced crystalloid solutions influences the return of normal GI function. This is a more objective parameter than the commonly used first defecation, as the latter may be influenced by neostigmine, suppositories, or enemas or only reflect rectal emptying.

This study has some limitations. Many factors impact GI function such as the surgery (open vs. minimally invasive, type of urinary derivation, extent of pelvic lymph node dissection, bowel segment selected, blood loss, duration of surgery), anesthesiologic factors (opioids, fluid overload), and individual risks (gender, age, comorbidities, Glasgow prognostic score).³⁴  However, surgical and anesthesiologic bias was limited in our study, as perioperative management adhered to the same center-specific pathways in all patients. In addition, this study was performed in a high–caseload center; thus, it remains unclear if these findings can be extrapolated to other centers and larger patient populations. Finally, this study was not powered to confidently assess safety. However, we could not detect a difference in in-hospital complications including neurologic disturbances associated with dyselectrolytemia (i.e., acute changes in sodium plasma level and its correction). It has to be emphasized that the G5K solution, mainly because of its relatively high potassium concentration, should only be administered as a maintenance fluid using a volumetric pump.

The administration of a potassium-enriched, chloride-depleted 5% glucose crystalloid as a perioperative maintenance solution in conjunction with a zero perioperative fluid balance did not significantly affect the time to first defecation (primary endpoint), but accelerated the return of normal GI function (secondary endpoint) after open radical cystectomy and urinary diversion. It reduced the postoperative need for potassium and magnesium substitution. However, this potassium-enriched, chloride-depleted 5% glucose crystalloid solution resulted in substantial postoperative mild hyponatremia mandating close monitoring of plasma sodium.

The authors thank Robert G. Hahn, M.D., Ph.D. (Research Unit, Södertälje Hospital, Linköping, Sweden; and Section for Anesthesia, Linköping University, Linköping, Sweden) for fruitful discussions. The authors acknowledge the Institute of Mathematical Statistics and Actuarial Science, University of Bern, Bern, Switzerland, for statistical work performed. Compensation consisted of an hourly salary.

Supported by Gottfried and Julia Bangerter-Rhyner Foundation, Basel, Switzerland, and institutional research funds of the Department of Anaesthesiology and Pain Therapy and the Department of Urology, University Hospital Bern, Bern, Switzerland. The funding sources had no role in the design and conduct of the study, the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.

The authors declare no competing interests.

Full protocol available from Dr. Wuethrich: patrick.wuethrich@insel.ch. Raw data available from Dr. Wuethrich: patrick.wuethrich@insel.ch.