Waters, Jonathan H. MD; Gottlieb, Alexandru MD; Schoenwald, Peter MD; Popovich, Marc J. MD; Sprung, Juraj MD, PhD; Nelson, David R. MS
Recent reports have investigated the impact of normal saline infusion on acid-base balance (1–4). These reports describe the development of metabolic hyperchloremic acidosis after the administration of 0.9% (“normal”) saline (NS) containing solutions. In this study, we hypothesized that NS would be associated with hyperchloremic metabolic acidosis and attempted to determine whether the acidosis affected hospital outcome. This outcome was evaluated by assessing the incidence of complications, blood product use, ventilator time, intensive care unit (ICU) stay, and hospital stay.
A single study has been published evaluating outcome differences in patients resuscitated with NS versus lactated Ringer’s (LR) solution (5). This study of Vietnam War casualties found no difference in survival between resuscitation fluids in a patient population composed of young, previously healthy soldiers, a population very different from the population routinely cared for in most hospitals today. Important changes in perioperative management have occurred since the Vietnam War; subtle outcome differences may not have been recognized in a study of this earlier era. These differences may currently be important.
To answer the primary hypothesis of this study, patients undergoing aortic surgery were used to assess the potential impact of these fluids. This population of patients was considered to be particularly vulnerable to any detrimental effects of these fluids. These patients possess a number of characteristics that could amplify any differences. For example, these patients receive large volumes of crystalloid fluids and in our institution, all patients undergoing these procedures are taken to the ICU intubated, all patients receive a general anesthetic with epidural opioids for postoperative pain management, and all have invasive monitoring. Because of the nature of the operative procedure and the severity of coexisting disease, they may have complicated and long hospital stays. This group of patients was therefore considered to offer an excellent opportunity to detect outcome differences that might be related to the study fluids.
Materials and Methods
After IRB approval and written and verbal consent, patients undergoing open aortic aneurysm repair were enrolled in the study. Patients were randomized by a computerized random number generator to receive either LR or NS solution as the predominant resuscitation fluid. Anesthetic management was standardized to sodium thiopental and cisatracurium for induction and isoflurane, fentanyl, and cisatracurium for maintenance. Total fentanyl use was restricted to 1 mg. All patients had a thoracic epidural catheter placed for postoperative analgesia. Patients were excluded from the study if the catheter did not function postoperatively. These catheters were dosed with preservative free morphine sulfate on placement. A standard epidural infusion of bupivacaine 0.0625% with fentanyl 1.25 μg/mL and epinephrine 1.25 μg/mL was started after aortic unclamping when hemodynamic stability had been achieved. An initial dose of 5 mL of the standard epidural solution was administered on starting the epidural infusion. All patients received mannitol 12.5 gms before aortic cross-clamping along with dopamine at 2 μg · kg−1 · min−1.
Patients with a history of abnormal renal function or who had abnormal serum blood urea nitrogen (<5.0 or >25 mg/dL) or creatinine levels (<0.7 or >1.4 mg/dL) were excluded. Patients with abnormal serum chloride levels (<97 or >105 mEq/L), or patients with preexisting acid-base abnormalities as assessed by base excess (BE) (>2 or <−2 mEq/L) and Pco2 (<35 or >45 torr) were excluded. Because albumin is a colloid volume expander, patients received it intraoperatively at the discretion of the anesthesiologist. Albumin 5% was used in preference to hetastarch because the hetastarch solution available at the initiation of the study was based in NS. The sodium and chloride concentration of albumin 5% was 150 mEq/L and 93 mEq/L, respectively (3).
The anesthesia and critical care providers were blinded by covering the labels of the crystalloid solutions. The study solution administration started on arrival in the operating room and ended on arrival in the ICU. Packed red blood cells and other blood products were administered without prior dilution, or were diluted with the blinded solution per the discretion of the anesthesia provider. Dilution of blood was performed with no more than 100 mL of diluent. Blood replacement followed the guidelines recommended by the American Society of Anesthesiologists (6). Allogeneic erythrocytes were transfused when the hemoglobin concentration was <10 gm/dL. Transfusion of fresh-frozen plasma (FFP) and platelets was based on clinical evidence of microvascular bleeding and guided by point-of-care prothrombin time, partial thromboplastin time, (CoaguChek, Roche Diagnostics GmbH, Mannheim, Germany) and Sonoclot® (Sienco, Inc., Wheat Ridge, CO) monitoring. All cell salvage blood was washed in NS using a 225 mL Latham bowl. Estimated blood loss was calculated from the volume of cell-salvaged blood returned to the patient.
All patients were monitored via arterial and central venous catheters. Pulmonary arterial catheters were placed at the discretion of the anesthesiologist performing the procedure. Transesophageal echocardiography was not performed. Crystalloid solution was administered to maintain pulmonary arterial occlusion pressure or central venous pressure (CVP) within 10% of the baseline preoperative value. Colloid use was restricted to periods of rapid blood loss, which generally occurred after aortic clamping or unclamping. Sodium bicarbonate (1.0 mEq/mL) use was restricted to patients with BE <−5. Bicarbonate was dosed according to the formula [body weight (kg) × BE × 0.3]/2.
Patients were transported to the ICU intubated, sedated, and partially paralyzed. Reversal of the muscle relaxant took place on arrival in the ICU and the count of “Intubation time” was started. The day of arrival in the ICU was considered as day “0” of the ICU and of the hospital time. Returns to the ICU after initial discharge were not included in the measure of ICU length of stay; these times were included in the hospital time.
All patients were weaned from mechanical ventilation by respiratory therapists under the supervision of full-time intensivists. Weaning was usually accomplished by reduction in the synchronized intermittent mandatory ventilation rate until patients were breathing spontaneously with continuous positive airway pressure of 5 cm H2O and pressure support of 5 cm H2O. Ventilator changes were monitored by arterial blood gases and pulse oximetry. Patients whose mechanical ventilation needs were more complicated were maintained on flow- or pressure-limited ventilator modes and monitored by blood gases as necessary. Patients were extubated if they maintained comfortable spontaneous ventilation and acceptable blood gas measurements with minimal continuous positive airway pressure and pressure-support settings, and if they were able to generate a maximum negative inspiratory pressure of at least −25 cm H2O and a vital capacity of at least 10 mL/kg.
All patients were monitored with an arterial blood gas and ionized calcium (Chiron 800 Series blood gas analyzer, Chiron Corp., Emeryville, CA) and for electrolytes, (Beckman CX3, Beckman Labs, Brea, CA) and magnesium (AVL 988, AVL Medical Instruments, Schaffhausen, Switzerland), and lactate level (YSI 2300 STAT Plus glucose and lactate analyzer, YSI Biotechnology, Yellow Springs, OH) at the start of the surgical procedure, on arrival in the ICU, and every 24 h during their stay in the ICU or until normalization of the measured variable. All laboratory samples were processed by the institutional laboratory.
Continuous demographic and laboratory variables were compared with a two-sample Student’s t-test. The incidence of chronic disease and medication use was compared between the two groups using Fisher’s exact test. Fluid and blood product use was compared using Wilcoxon’s rank-sum test. A single omnibus hypothesis test combining the blood products used (albumin, cell salvaged blood, packed red blood cells, FFP, and platelets) determined if a difference existed between overall blood use between the LR and the NS group. This approach used the Wilcoxon’s rank-sum test approach described by O’Brien (7). Median differences between groups and their 95% confidence intervals were used to determine which of the specific variables differed between groups. Similarly, a χ2 test was used to determine whether the two treatment groups differed in overall complication rates, and 95% confidence intervals were used to evaluate specific complications.
A secondary analysis used univariate and multivariate analyses to determine which of the independent variables were related to the outcome measures. Durations of ICU and hospital stay and mechanical ventilation times were log-transformed before analysis. Log-transformations tend to make the regression variables difficult to interpret, so anti-logs were taken to calculate the percent change in the outcome in the presence of the risk factor. The significance level for all hypotheses was P < 0.05.
Because no data on outcome differences existed, the sample size was based on the primary outcome of change in BE for a sample of 25 patients having undergone abdominal aortic aneurysm repair. Preliminary data suggested that the sd of BE change from baseline to postoperative was approximately 3 mEq/L for patients treated randomly with NS or LR. An estimate of the difference between both groups was 6 mEq/L in this study. With 33 patients per treatment group, there was a 90% power to detect a difference of 5 mEq/L or more between the groups in change from baseline using a Student’s t-test with overall significance level of 0.05. Post hoc power calculations were performed to determine power for the overall outcome measure comparisons. The sample of 33 per group could detect with 90% power whether the LR group averaged 10% better or 10% poorer values for the continuous outcome measures.
Sixty-six patients (33 patients in the LR group, 33 patients in the NS group) were enrolled in the study. The NS group consisted of 29 infrarenal, 1 suprarenal, and 3 thoracoabdominal repairs; whereas, the LR group consisted of 30 infrarenal, 1 suprarenal, and 2 thoracoabdominal repairs. Seventeen patients in the NS group were monitored with a CVP catheter and 16 were monitored with a pulmonary arterial catheter. In the LR group, 15 patients were monitored with a CVP catheter and 18 were monitored with a pulmonary arterial catheter. There was no difference in the patients with respect to demographic data nor did they differ in the incidence of chronic disease with the exception of hypertension (Table 1). Despite a difference in the reported incidence of hypertension, no difference existed between the number of patients who were receiving drug therapy for hypertension. Confidence interval analyses indicated that there were no differences in the volume of crystalloid fluids administered. No differences existed in the amount of estimated blood loss or the volume of packed red blood cells or FFP administered. There was a difference in the volume of platelets transfused, with the NS group requiring more. In the test of all of the continuous blood product measures, the LR group had significantly less blood product exposure overall than the NS group (P = 0.02). The urine output was significantly more in the NS group (Table 2). Significant differences were seen in the pre- to postoperative pH, BE, Bicarb, Na+, and Cl− (Table 3). No difference in the postoperative complications nor death was seen (Table 4). No difference was seen in the ventilator time (45.6 ± 147.2 h in the LR group versus 29.7 ± 61.8 h in the NS group), ICU time (4.1 ± 7.6 days in the LR group versus 2.8 ± 3.8 days in the NS group), nor hospital stay (10.1 ± 8.3 days in the LR group versus 8.9 ± 4.7 h in the NS group). A significant difference in the volume of bicarbonate (3.8 ± 15.5 mL in the LR group versus 40.2 ± 64.0 mL in the NS group) used during the operative period was seen but there was no difference in the postoperative period. No differences were found in the amount of furosemide used during the operative or postoperative period.
Multivariate analyses were performed to determine which of the independent variables were related to the outcome measure of ventilation time, surgical ICU (SICU) stay, and hospital stay. The times were log-transformed before analysis. For ventilation time, three variables (diuretic use, β-adrenergic blocker use, and age) were significantly related to time on ventilation. Regression estimates indicated that the diuretic use, β-adrenergic blocker use, or age older than 65 yr increased ventilation time 62.3% (P = 0.002), 24.6% (P = 0.04), and 24.9% (P = 0.04), respectively. For SICU stay, the only significant variable was chronic obstructive pulmonary disease. Estimates indicated that the presence of chronic obstructive pulmonary disease increased SICU time 23.8% (P = 0.007). Similar analysis revealed that asthma and age older than 65 yr were significantly related to hospital LOS, 33.4% (P = 0.001) and 13.8% (P = 0.008), respectively. No relationship was found between these times and the crystalloid fluid.
The goal of this study was to determine if the acidosis after NS would change a patient’s outcome after surgery. Because little prior data existed as to the extent of the impact of the hyperchloremic acidosis, we evaluated multiple endpoints of outcome. These included bicarbonate use, blood product use, patient complications, ventilation time, ICU time, and hospital LOS. We found little difference in these measures with the exceptions of bicarbonate and blood product use.
A potential criticism of this study design would be that the sample size chosen for these multiple outcome end points might be inadequate to detect any true difference in one of the end points. This challenge would be based on the sample size calculation being derived from previously observed measures of BE to estimate variability. This measure was used because it provided readily quantifiable information, whereas no reliable information existed for other outcome mea-sures when the study was designed. In essence, the BE data served as a surrogate to establish levels of variation that could be expected among patients. Once this study was completed, the post hoc power calculation used the outcomes observed within the two groups to determine that the sample size was sufficient to detect a 10% difference between groups. Therefore, the use of BE appears to have established sample sizes that provided sufficient power to detect a clinically meaningful outcome difference in this high risk group of patients.
With these considerations, the following findings are of interest: first, the acidosis after NS administration resulted in a larger amount of sodium bicarbonate being administered during the intraoperative period to treat acidosis. Despite this intraoperative difference, after the transfer of these patients to the ICU and the resumption of routine crystalloid use, no significant difference in bicarbonate use was seen. This would suggest that the acid-base change seen after NS infusion might be a transient effect.
The second area of interest relates to renal function. Williams et al. (8) showed that osmolarity differences after NS led to a delay in the time to which a group of volunteers first urinated. This was postulated to be predictable based on the antidiuretic hormone response to changes in serum osmolarity. In this study, differences in calculated serum osmolarity were seen but this appeared to have little impact on urine output, with urine output being larger in the NS group. This may relate to the fact that the patients who were given NS for resuscitation received, on average, 500 mL larger volumes of the crystalloid solution and 1500 mL more total fluid. It was also anticipated that the higher osmolarity after NS would lead to increased amounts of retained fluid, longer ventilation times because of this fluid, and longer hospitalizations. None of these outcome differences was demonstrated. In addition, no differences in postoperative serum creatinine levels were noted, nor was there any difference in the development of postoperative renal insufficiency. It is important to note that interpretation of urinary output is complicated by the use of dopamine during the surgical procedure.
Differences in blood loss between fluid regimens has been of interest. Martin et al. (9) in a study of patients undergoing surgery associated with substantial blood loss found larger blood loss in patients that received hetastarch in a NS solution when compared with patients who received hetastarch in a buffered electrolyte solution or to a third group who received only LR. This suggested that the metabolic hyperchloremic acidosis after the NS-containing solution or the small amount of calcium in the buffered electrolyte solution were important factors. Gan et al. (10) reported a difference in blood loss between two similar study groups but did not find this difference to be statistically significant. In the study presented here, we found that the LR group had smaller estimated blood loss but, like Gan et al. (10), we found no statistically significant difference.
Though no statistical difference in blood loss was found between these two groups, we observed a difference in the blood products used. The NS group in this study had an increased need for FFP and a statistically significant difference in platelet transfusion volumes. Because the blood product use and the blood loss trended toward being larger in the NS group, the results from all of the transfused blood products were compared between the two groups using an omnibus hypothesis test. This global assessment indicated that NS led to increased use of blood products.
In conclusion, a significant difference in acid-base balance developed between the two groups. This imbalance occurred in the NS group with the development of a hyperchloremic metabolic acidosis. This finding is consistent with other studies evaluating the acid-base effects of saline solutions. With this acidosis, little difference in traditional outcome measures was found; however, more blood products were transfused in the group receiving predominantly NS. Thus, predominant NS would seem to be a less desirable choice of fluid compared to LR for large blood loss procedures. Although this study did not evaluate procedures that involve minimal blood loss and fluid administration, it can be speculated that the observations from the present study would not be applicable to such cases.
1. Scheingraber S, Rehm M, Sehmisch C, et al. Clinical application of a new acid base theory in crystalloid infusion therapy. Anesthesiology 1999; 90: 1265–70.
2. Rehm M, Orth V, Scheingraber S, et al. Acid-base changes caused by 5% albumin versus 6% hydroxyethyl starch solution in patients undergoing acute normovolemic hemodilution. Anesthesiology 2000; 93: 1174–83.
3. Waters JH, Bernstein CA. Dilutional acidosis following hetastarch or albumin in healthy volunteers. Anesthesiology 2000; 93: 1184–7.
4. Waters JH, Miller L. Mechanism of acidosis in prolonged surgical procedures. Crit Care Med 1999; 27: 2142–6.
5. Lowery BD, Cloutier CT, Carfy LC. Electrolyte solutions in resuscitation in human hemorrhagic shock. Surg Gynecol Obstet 1971; 133: 273–84.
6. American Society of Anesthesiologists Committee on Transfusion Medicine. Questions and answers about transfusion practices. 3rd ed. Park Ridge, IL: American Society of Anesthesiologists, 1998.
7. O’Brien PC. Procedures for comparing samples with multiple endpoints. Biometrics 1984; 40: 1079–87.
8. Williams EL, Hildebrand KL, McCormick SA, Bedel MJ. The effect of intravenous lactated Ringer’s solution versus 0.9% sodium chloride solution on serum osmolality in human volunteers. Anesth Analg 1999; 88: 999–1003.
9. Martin G, El-Moalem H, Bennett-Guerrero E, Mythen MG, Gan TJ. Comparison of intraoperative blood loss in patients undergoing major surgery using hextend, hespan and lactated Ringer’s solution [abstract]. Anesthesiology 1999; 91: A166.
10. Gan TJ, Bennett-Guerrero E, Phillips-Bute B, et al. Hextend, a physiologically balanced plasma expander for large volume use in major surgery: a randomized phase III clinical trial. Anesth Analg 1999; 88: 992–8.
In the August 2001 article by Rau et al. (2001;93:382–4), “Propofol in an emulsion of long- and medium-chain triglycerides: the effect on pain,” Propofol-®Lipuro (B. Braun Melsungen AG) was the hypnotic that was compared with Diprivan® 10%. Unfortunately, “Lipofundin® MCT 10%” was erroneously substituted for “Propofol-®Lipuro” throughout the article and in the column headings of Tables 1 and 2. “Lipofundin® MCT 10%” correctly appears at the bottom of the first column of page 383, in the second paragraph of the Discussion section, in the sentence “However, mixing 20 mL of the original propofol preparation with 10 mL Lipofundin® MCT 10% reduced the concentration of free propofol in the aqueous phase by 39.9% (2) and decreased pain on injection.” In all other instances in the article, “Lipofundin® MCT 10%” should read “Propofol-®Lipuro.” The online version of the article is correct, and corrected reprints are available. The publisher regrets the error and apologizes for any resultant confusion.