In the great majority of surgical cases, fluid therapy is managed with crystalloids. There are 2 broad categories of crystalloid solutions, normal saline (NS) and the balanced salt solutions, e.g., Ringer’s lactate or Plasmalyte™ (Baxter, Mississauga, Ontario, Canada). Balanced salt solutions may be preferable since they offer an electrolyte composition approaching the composition of plasma. These balanced salt crystalloids include a weak acid (e.g., lactate), to decrease the concentration of chloride and provide buffering capacity. However, at our institution NS is the least expensive solution, and hence is the most commonly used crystalloid solution for perioperative fluid management.a 1 , 2 3
NS has a chloride concentration of 154 mmol/L, well above the serum chloride concentration (100–110 mmol/L), and its overuse may lead to hyperchloremic metabolic acidosis.4 , 5
METHODS
This study was approved by the Research Ethics Board (REB) of the University Health Network, Toronto, Ontario, Canada. The requirement for informed consent was waived by the REB.
Study Setting, Patient Sample, and Data Collection
This retrospective observational study was prepared to conform to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.6
After obtaining institutional REB approval, data were retrospectively collected on consecutive adult patients (>18 years of age) who underwent surgery between January 1, 2003 and December 31, 2008. Patients were identified from the Operating Room Scheduling Office System (ORSOS™, McKesson Corporation, San Francisco, CA) surgical bookings. The following groups of patients were excluded: outpatient surgery, cardiac surgery, organ transplantation, preoperative renal dysfunction (creatinine >175 mM, preoperative hyperchloremia [serum chloride >110 mmol/L]), or patients whose preoperative or postoperative chloride levels were not measured (Fig. 1 ). For those patients who underwent >1 surgery at the institution during the study period, the first surgical case was considered the index surgery for this study.
Figure 1: Patient population. Noncardiac, nontransplant inpatient surgical patients. Cr = creatinine; Cl = serum chloride.
Synthesis and Linking of Electronic Database
All data, including demographics, the nature of the surgery, International Classification of Disease, Tenth Revision (ICD-10 ) codes, laboratory tests, and date of death were retrieved from the electronic data warehouse by using previously published methodology.7 ICD-10 codes for preoperative comorbidities were analyzed after the method of Quan et al.8
Since many patients had multiple assessments, we analyzed all laboratory data to derive both the maximum or minimum values for hemoglobin, creatinine, and chloride for each of the first 5 postoperative days (PODs). The chloride data for the analysis consisted of the last preoperative value and the maximum value on POD 1 through 2 and POD 3 through 5.
Blood transfusion data were derived from the transfusion medicine database (Hemocare; Mediware Information Systems, Alton, IL), which includes data on all blood products (red blood cells [RBCs], fresh frozen plasma, [plasma] and platelets) issued and not returned from the in-hospital blood bank. Transfusion data were merged with the set of data derived from the electronic data warehouse using the patients’ unique identifying number and the date of surgery.
Definition of Primary and Secondary Outcomes
Acute postoperative hyperchloremia was defined as serum chloride concentration >110 mmol/L on either POD 1 or 2. The primary outcome was the incidence of in-hospital death within 30 days of the index surgery. Secondary outcomes included: length of hospital stay, defined as the number of days from the index surgery to discharge. The incidence of postoperative pulmonary edema, myocardial ischemia, cerebrovascular events, cardiac arrest, and atrial fibrillation were estimated using the ICD-10 outcome codes. Postoperative renal failure was determined based on the Consensus RIFLE Criteria (Risk, Injury, Failure, Loss and End stage kidney disease).9 10
Statistical Analyses
SAS™ version 9.1.3 (SAS Institute Inc., Cary, NC) was used for all statistical analyses. Categorical variables were summarized as frequencies and percentages and continuous variables as means and standard deviations.
Propensity Score Matching of Cohorts
Propensity scores estimating the probability of developing hyperchloremia were calculated for all patients. In the nonparsimonious multiple logistic regression model for the propensity score, the covariates used are shown in Table 1 . Individual patients with normal preoperative serum chloride who developed hyperchloremia on POD 1 or 2 were matched 1:1 using propensity scores to patients who did not develop hyperchloremia. A 5→1 computerized greedy matching technique was used for this matching process, whereby cases were first matched to controls that had a propensity score (logit transformation) that was identical in all 5 digits. Those that did not match were then matched to controls on 4 digits of the propensity score. This continued down to a 1-digit match on propensity score for those that remained unmatched.b
Table 1: Patient Demographics and Surgical Variables
If after the matching 1 or more variables remained unbalanced, interaction terms were included in the model and unmatched variables were forced into the match. This iterative process continued, blinded to outcomes, until the matched cohorts were balanced for the baseline covariates.
We assessed the balance between the 2 cohorts using a standardized mean difference. Standardized difference (d ) is defined to be equal to
where
and
are the mean values for the treatment and control groups, while S2 treatment and S2 control are the sample standard deviations, respectively. Differences of absolute value >5% are considered to indicate significant covariate imbalance.11
Sensitivity Analysis
To consider the possibility of selection bias in the propensity-matched cohort, 4 sensitivity analyses were conducted; first, by imputing group means from missing values of preoperative chloride, creatinine, and hemoglobin; second, by excluding all oncology cases to ensure the outcome was not influenced by the preoperative diagnosis or prognosis; and third, by excluding all patients with preoperative serum chloride <100 mmol/L. The fourth sensitivity analysis divided the surgical procedures into high-risk (neurosurgery, vascular, and thoracic surgery) and low- or moderate-risk surgery (ears, nose and throat surgery, general surgery, gynecology and oncology, orthopedics, plastic surgery, spine surgery, urology, and other). To consider whether surgical procedure was a confounding variable, the incidence of 30-day mortality was determined by surgical procedure in the propensity-matched population.
Multivariable Logistic Regression
Multivariable logistic regression modeling was also performed to assess the adjusted association of acute postoperative hyperchloremia with 30-day postoperative mortality. Since acute postoperative hyperchloremia is a postoperative outcome, as is myocardial ischemia or renal dysfunction, we felt that it was important to demonstrate that each was independently linked to 30-day mortality. The mathematical relationship between the continuous predictor variables selected and 30-day mortality was assessed with the cubic spline functions.12 P < 0.1 was the criterion for variable retention. A Pearson correlation matrix was used to identify collinear independent variables.13
RESULTS
A total of 22,851 patients who underwent an inpatient, noncardiac, nontransplant surgical procedure were included in the study (Fig. 1 ). Of note, only 870 patients had hyperchloremia preoperatively. Among the study sample, 4995 (22%) had acute postoperative hyperchloremia and 1670 (7.3%) had late postoperative (day 3–5) hyperchloremia; 1182 (5.2%) patients were hyperchloremic in both periods. Acute postoperative hyperchloremic patients were older, more often female, underwent longer surgical procedures, had more comorbidities, and received more blood products than patients with normal chloride levels (Table 1 ).
Before risk adjustment, patients who developed acute postoperative hyperchloremia had higher mortality rate (3.4%) than patients who did not develop acute postoperative hyperchloremia (1.3%) (Table 2 ). Acute postoperative hyperchloremic patients had a longer length of hospital stay and were more likely to have postoperative renal dysfunction based on both ICD-10 outcome codes and eGFR. Other ICD-10 outcome codes more common in acute postoperative hyperchloremic patients included postoperative pulmonary edema, pulmonary embolism, myocardial ischemia, myocardial infarction, atrial fibrillation, and cerebrovascular events.
Table 2: Outcomes for Entire and Propensity-Matching Cohorts
Propensity-Matched Cohort
The propensity score derivation model included the following variables seen in Table 1 : sex, emergent surgery (yes or no), last preoperative hemoglobin (grams per liter), main procedure service (low, moderate, and high risk), procedure time, minimum hemoglobin on POD 1, last preoperative creatinine, Charlson comorbidities score summary, age, RBC transfusions, plasma transfusions, and platelet transfusions as listed. The first 6 variables (sex, emergent surgery, last preoperative hemoglobin, main procedure service, procedure time, and minimum hemoglobin on POD 1) were forced into the final model for propensity matching.
The matching process created 2 cohorts of 4266 patients, 1 with acute postoperative hyperchloremia and another without. The 2 groups were well balanced with respect to all collected variables including blood transfusions (Table 1 ). Mortality in patients with acute postoperative hyperchloremia was higher (3.0% vs 1.9%; odds ratio [OR] = 1.58; 95% confidence interval [CI], 1.25–1.98). The hyperchloremic group also had a longer hospital length of stay and were more likely to have acute kidney injury (defined as >25% reduction in creatinine clearance; Table 2 ).
Sensitivity Analysis
A total of 11,510 patients were removed from the starting dataset because of missing perioperative data. To ensure that this did not result in a selection bias, normal values were imputed using group means for preoperative serum chloride (105 mmol/L) and creatinine (male <104 mcmol/L and female 96 mcmol/L) and hemoglobin concentration (male 135 g/L and female 127 g/L). Each of the imputed values was included in the analysis separately and together. The number of patients matched increased to a maximum of 5983 patients, but there was no change in the association between acute postoperative hyperchloremia and increased 30-day mortality and length of hospital stay for patients with postoperative hyperchloremia.
The propensity analysis was also repeated for 2283 pairs of patients undergoing nononcological surgery, and the association between hyperchloremia and mortality was retained: 47 deaths in the acute postoperative hyperchloremic group and 31 deaths in the normal serum chloride group (OR = 1.53; 95% CI, 0.97–2.41). A similar result was found for those patients undergoing oncological surgery. The length of hospital stay was 6.2 (interquartile range 3.4–11.3 days) and 6.1 days (interquartile range 3.31–10.94 days) in the acute postoperative hyperchloremia and normal serum chloride groups, respectively.
A total of 537 (2.4%) patients had a postoperative serum chloride concentration of <100 mmol/L. When these patients were excluded from the propensity analysis and match, there were 30 fewer matches made and the number of deaths in the normal chloride group was reduced from 80 to 71 (1.7%) and the OR for mortality at 30 days increased as a result to 1.86 (95% CI, 1.39–2.49).
The propensity analysis was repeated including the high-risk surgical services (vascular and thoracic surgery) and low- or moderate-risk surgery. This propensity-matched analysis resulted in 4334 pairs, and there were 60 (1.4%) deaths in the normal chloride group and 125 (2.9%) deaths in the acute postoperative hyperchloremic group at 30 days. The OR for mortality at 30 days was in keeping with the primary analysis, 2.11 (95% CI, 1.55–2.89).
Finally, to determine whether the surgical procedure was associated with hyperchloremia as a confounder, the incidence of 30-mortality was reviewed by surgical procedure in the propensity-matched population (Table 3 ). In reviewing the 10 procedures associated with the most frequent incidence of 30-day mortality, we were able to account for 85% and 74% of the 30-day mortality in the hyperchloremic and normal chloride groups, respectively. The procedure types were similar in both groups with nephrectomy and peripheral vascular surgery being the only unique procedures to the hyperchloremic and normal chloride groups, respectively (Table 3 ).
Table 3: Procedures Associated with the Most Frequent Incidence of 30-Day Mortality
Biologic Gradient
The relationship between the postoperative chloride concentration and 30-day mortality was generated using a Cubic spline relationship, before and after risk adjustment (Fig. 2 , A and B ). There is a linear increase in the probability of mortality between 100 and 125 mmol/L in both before and after propensity-matched cohorts.
Figure 2: Spline function graph of the probability of dying within 30 days of surgery and the postoperative maximum serum chloride concentration on postoperative day 1 or day 2. A, Unadjusted figure. B, Adjusted by propensity match.
Logistic Regression Analysis
Postoperative hyperchloremia was included in logistic regression modeling for 30-day postoperative mortality (Table 4 ). Acute postoperative hyperchloremia was independently associated with an increase in risk of death, and the models had good discrimination and calibration. Variables included in the final model were age (>70 years), sex, high-risk surgical services (thoracic or vascular), emergent surgery, Charlson Comorbidity Class (>1), preoperative anemia (hemoglobin <120 g/L for females, <130 g/L for males), blood transfusion (RBC >0), and all outcome variables measured listed in Table 2 with the exception of mortality, length of hospital stay, and acute renal failure. Renal dysfunction was defined in the final model as a >25% decrease in eGFR.
Table 4: Independent Predictors of 30-Day Mortality
DISCUSSION
Postoperative hyperchloremia occurs frequently in our postoperative noncardiac surgical population; 22% had serum chloride levels >110 mmol/L on POD 1 or 2. In contrast to assumptions about the benign nature of postoperative hyperchloremia, our analysis shows an association between acute postoperative hyperchloremia and increased morbidity, length of hospital stay and 30-day mortality. This association was demonstrated using a propensity score matched-pair analysis and confirmed in 3 planned sensitivity analyses. The independent association between acute postoperative hyperchloremia and 30-day mortality was confirmed using multivariable logistic regression analysis including other postoperative adverse outcomes variables, i.e., myocardial infarction, stroke, renal failure, and blood transfusion. We therefore have 2 observations that merit consideration in the management of surgical patients: (1) acute postoperative hyperchloremia occurs frequently after surgery and (2) acute postoperative hyperchloremia is associated with increased postoperative morbidity and mortality.
An association of NS with postoperative hyperchloremia is not often cited as a reason to limit NS administration, and this is likely true at our institution. NS is the crystalloid of choice based on purchasing records for the operating rooms. Throughout the study period, approximately 35,000 L/y of crystalloid solutions were purchased and >50% of this fluid was NS.c
Both animal and human studies show that NS administration results in hyperchloremia in a dose-dependent manner.5 , 14–20 21
Other factors that could account for the development of postoperative hyperchloremia include renal dysfunction, preoperative hyperchloremia, and previous surgery with unmeasured fluid administration. However, patients with these potential causes of acute postoperative hyperchloremia were specifically excluded from our analysis. Postoperative hyperchloremia can be a result of excess free water loss (e.g., evaporative loss, diabetes insipidus, diarrhea, burns, renal loss, osmotic or after obstructive diuresis) and increase renal reabsorption of chloride (e.g., renal tubular acidosis, early renal failure, ureteral diversion procedures, acetazolamide induced).22
Previous evidence of the harmful effect of hyperchloremia has also been reported in critically ill patients.23 , 24 25 26 20 27 , 28
NS is recommended for RBC transfusion by North American blood agencies, and therefore one might anticipate hyperchloremia to be associated with blood transfusions. Transfusions are, in turn, associated with perioperative anemia. However, in our analysis, the effect of transfusion was negligible since all aspects of transfusion, blood products, and perioperative anemia were well balanced after propensity score matching and were included in the logistic regression analysis. Our analysis demonstrates that the association between hyperchloremia and postoperative morbidity and mortality is independent of the effects of patient comorbidities, surgical service, or emergent surgery, because these were also well balanced in the propensity-matched analysis. Finally, using multivariable logistic regression analysis, we have demonstrated that hyperchloremia continues to be independently associated with mortality after accounting for perioperative risk factors and postoperative outcomes variables (Table 4 ).
Propensity-matching analysis gives an estimate of the detrimental effect of acute postoperative hyperchloremia on the population or a marginal treatment effect.29 28
Avoiding hyperchloremia may offer an explanation for the beneficial effects ascribed to balanced salt solutions, but until recently, the evidence has been unconvincing.19 , 25 , 30 , 31 32
The major limitation of our study is the lack of data on the perioperative use of crystalloid solutions, and therefore we cannot definitively comment on the cause of hyperchloremia. However, NS administration is a modifiable risk factor for the development of hyperchloremia. It is reasonable to speculate that substitution of balanced salt solutions for NS would result in a lower incidence of hyperchloremia since this is supported by evidence from studies using balanced salt solutions.5 , 16
We must also recognize that by balancing groups in the propensity match using surgical service rather than the specific type of surgery, e.g., open versus endovascular repair, we may have introduced an unrecognizable bias.33 Table 4 ). Furthermore, the 10 procedures associated with the most frequent 30-day mortality in the hyperchloremic and normal chloride group were similar and therefore hyperchloremia was not associated with a particular type of surgery (Table 3 ).
Other limitations of our study include the retrospective nature of this study and the inability to adjust for unmeasured covariables. However, with these limitations in mind, it would seem prudent to avoid chloride-rich solutions perioperatively until better evidence is available.34 http://clinicaltrials.gov/show/NCT01270854
Conclusions
Hyperchloremic metabolic acidosis is not a benign, self-limiting, metabolic disturbance.14 35
DISCLOSURES
Name: Stuart A. McCluskey, PhD, MD.
Contribution: This author helped design and conduct the study, analyze the data, and write the manuscript.
Attestation: Stuart A. McCluskey has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.
Name: Keyvan Karkouti, MSc, MD.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Keyvan Karkouti has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: Duminda Wijeysundera, PhD, MD.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Duminda Wijeysundera reviewed the analysis of the data and approved the final manuscript.
Name: Leonid Minkovich, PhD, MD.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Leonid Minkovich reviewed the analysis of the data and approved the final manuscript.
Name: Gordon Tait, PhD.
Contribution: This author helped design the study, analyze the data, and write the manuscript.
Attestation: Gordon Tait has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Name: W. Scott Beattie, PhD, MD.
Contribution: This author helped design and conduct the study, analyze the data, and write the manuscript.
Attestation: W. Scott Beattie has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
This manuscript was handled by: Sorin J. Brull, MD, FCARCSI (Hon).
FOOTNOTES
a Personal communication with hospital stores supplied by KPMG Plexus (Toronto, Ontario, Canada), November 2011. Normal saline $1.23/L, Ringer’s lactate $1.79/L, and Plasmalyte $2.17/L. Cited Here
b Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. In: Proceedings of the Twenty-Sixth Annual SAS User Group International Conference. Cary, NC: SAS Institute, 2001. Cited Here
c Personal communication and review of the operating room purchasing records supplied by KPMG Plexus, March 2012. Cited Here
REFERENCES
1. Canadian Blood Services. . Circular of information for the use of human blood components. 2011
2. Brecher ME Technical Manual. 200214th ed Bethesda, MD American Association of Blood Banks Press:476–8
3. Moritz ML, Ayus JC. Water water everywhere: standardizing postoperative fluid therapy with 0.9% normal saline. Anesth Analg. 2010;110:293–5
4. Scheingraber S, Rehm M, Sehmisch C, Finsterer U. Rapid saline infusion produces hyperchloremic acidosis in patients undergoing gynecologic surgery. Anesthesiology. 1999;90:1265–70
5. 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
6. Vandenbroucke JP, von Elm E, Altman DG, Gøtzsche PC, Mulrow CD, Pocock SJ, Poole C, Schlesselman JJ, Egger MSTROBE Initiative. . Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Epidemiology. 2007;18:805–35
7. Beattie WS, Karkouti K, Wijeysundera DN, Tait G. Risk associated with preoperative anemia in noncardiac surgery: a single-center cohort study. Anesthesiology. 2009;110:574–81
8. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care. 2005;43:1130–9
9. Bellomo R, Kellum JA, Ronco C. Defining and classifying acute renal failure: from advocacy to consensus and validation of the RIFLE criteria. Intensive Care Med. 2007;33:409–13
10. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron. 1976;16:31–41
11. Austin PC. Propensity-score matching in the cardiovascular surgery literature from 2004 to 2006: a systematic review and suggestions for improvement. J Thorac Cardiovasc Surg. 2007;134:1128–35
12. Harrell FE. SAS macros and data step programs useful in survival analysis and logistic regression. 1986 Available at:
http://biostat.mc.vanderbilt.edu/wiki/pub/Main/SasMacros/survrisk.txt . Accessed April 2013
13. Halm EA, Wang JJ, Boockvar K, Penrod J, Silberzweig SB, Magaziner J, Koval KJ, Siu AL. Effects of blood transfusion on clinical and functional outcomes in patients with hip fracture. Transfusion. 2003;43:1358–65
14. Prough DS, Bidani A. Hyperchloremic metabolic acidosis is a predictable consequence of intraoperative infusion of 0.9% saline. Anesthesiology. 1999;90:1247–9
15. Rehm M, Finsterer U. Treating intraoperative hyperchloremic acidosis with sodium bicarbonate or tris-hydroxymethyl aminomethane: a randomized prospective study. Anesth Analg. 2003;96:1201–8
16. McFarlane C, Lee A. A comparison of Plasmalyte 148 and 0.9% saline for intra-operative fluid replacement. Anaesthesia. 1994;49:779–81
17. Rehm M, Orth V, Scheingraber S, Kreimeier U, Brechtelsbauer H, Finsterer U. Acid-base changes caused by 5% albumin versus 6% hydroxyethyl starch solution in patients undergoing acute normovolemic hemodilution: a randomized prospective study. Anesthesiology. 2000;93:1174–83
18. Todd SR, Malinoski D, Muller PJ, Schreiber MA. Lactated Ringer’s is superior to normal saline in the resuscitation of uncontrolled hemorrhagic shock. J Trauma. 2007;62:636–9
19. Kellum JA. Fluid resuscitation and hyperchloremic acidosis in experimental sepsis: improved short-term survival and acid-base balance with Hextend compared with saline. Crit Care Med. 2002;30:300–5
20. Cambier C, Detry B, Beerens D, Florquin S, Ansay M, Frans A, Clerbaux T, Gustin P. Effects of hyperchloremia on blood oxygen binding in healthy calves. J Appl Physiol. 1998;85:1267–72
21. O’Malley CM, Frumento RJ, Hardy MA, Benvenisty AI, Brentjens TE, Mercer JS, Bennett-Guerrero E. A randomized, double-blind comparison of lactated Ringer’s solution and 0.9% NaCl during renal transplantation. Anesth Analg. 2005;100:1518–24
22. Yunos NM, Bellomo R, Story D, Kellum J. Bench-to-bedside review: chloride in critical illness. Crit Care. 2010;14:226
23. Boniatti MM, Cardoso PR, Castilho RK, Vieira SR. Is hyperchloremia associated with mortality in critically ill patients? A prospective cohort study. J Crit Care. 2011;26:175–9
24. Gunnerson KJ, Saul M, He S, Kellum JA. Lactate versus non-lactate metabolic acidosis: a retrospective outcome evaluation of critically ill patients. Crit Care. 2006;10:R22
25. Wilkes NJ, Woolf RL, Powanda MC, Gan TJ, Machin SJ, Webb A, Mutch M, Bennett-Guerrero E, Mythen M. Hydroxyethyl starch in balanced electrolyte solution (Hextend)–pharmacokinetic and pharmacodynamic profiles in healthy volunteers. Anesth Analg. 2002;94:538–44
26. Wilcox CS. Regulation of renal blood flow by plasma chloride. J Clin Invest. 1983;71:726–35
27. Kellum JA, Song M, Almasri E. Hyperchloremic acidosis increases circulating inflammatory molecules in experimental sepsis. Chest. 2006;130:962–7
28. Wan L, Bellomo R, May CN. The effects of normal and hypertonic saline on regional blood flow and oxygen delivery. Anesth Analg. 2007;105:141–7
29. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46:399–424
30. Gan TJ, Bennett-Guerrero E, Phillips-Bute B, Wakeling H, Moskowitz DM, Olufolabi Y, Konstadt SN, Bradford C, Glass PS, Machin SJ, Mythen MG. Hextend, a physiologically balanced plasma expander for large volume use in major surgery: a randomized phase III clinical trial. Hextend Study Group. Anesth Analg. 1999;88:992–8
31. Senagore AJ, Emery T, Luchtefeld M, Kim D, Dujovny N, Hoedema R. Fluid management for laparoscopic colectomy: a prospective, randomized assessment of goal-directed administration of balanced salt solution or hetastarch coupled with an enhanced recovery program. Dis Colon Rectum. 2009;52:1935–40
32. Shaw AD, Bagshaw SM, Goldstein SL, Scherer LA, Duan M, Schermer CR, Kellum JA. Major complications, mortality, and resource utilization after open abdominal surgery: 0.9% saline compared to Plasma-Lyte. Ann Surg. 2012;255:821–9
33. Dexter F, Dexter EU, Ledolter J. Statistical grand rounds: importance of appropriately modeling procedure and duration in logistic regression studies of perioperative morbidity and mortality. Anesth Analg. 2011;113:1197–201
34. Cusack RJ, Rhodes A, Lochhead P, Jordan B, Perry S, Ball JA, Grounds RM, Bennett ED. The strong ion gap does not have prognostic value in critically ill patients in a mixed medical/surgical adult ICU. Intensive Care Med. 2002;28:864–9
35. Bradford-Hill A. The environment and disease: association or causation? Proc R Soc Med. 1965;58:295–300
