Greater intravenous fluid volumes are associated with prolonged recovery after colorectal surgery: a retrospective cohort study

Transcription

1 British Journal of Anaesthesia, 116 (6): (2016) doi: /bja/aew125 Clinical Practice Greater intravenous fluid volumes are associated with prolonged recovery after colorectal surgery: a retrospective cohort study Z. Aga 1,2, M. Machina 2 and S. A. McCluskey 2,3, * 1 University of Toronto, Faculty of Medicine, Toronto, Ontario, Canada, 2 Department of Anesthesia and Pain Management, Toronto General Hospital, University Health Network, Toronto, Ontario, Canada, and 3 Department of Anesthesia, University of Toronto, Toronto, Ontario, Canada *Corresponding author. stuart.mccluskey@uhn.ca Abstract Background: We carried out a retrospective assessment of whether perioperative fluid volume was associated with length of hospital stay (LOS) after colorectal surgery. Methods: A single-centre chart review was conducted on colorectal surgeries that took place between January 2008 and December The primary outcome was LOS, with prolonged LOS defined as greater than median LOS. Secondary outcomes included postoperative pulmonary oedema, acute renal failure, myocardial infarction, and mortality. Univariate analysis, multivariable logistic regression, and quantile regression analyses were conducted to examine the association between perioperative fluid volume and prolonged LOS. Results: Of the 1242 procedures, 57% were elective, 62% oncological, and 31% laparoscopic. The median LOS was 8.2 days (interquartile range 5.2, 14.7). Patients received 3.2 (SD 1.5) litres of fluid in the perioperative period (operating and recovery rooms), predominantly crystalloid. The volume (in litres) of perioperative fluid was independently associated with prolonged LOS (odds ratio 1.23, 95% confidence interval , P<1). This association persisted across the spectrum of definitions for prolonged LOS (10th 90th percentile). Logistic regression analysis also revealed that prolonged LOS was associated with age >65 yr, Charlson Co-morbidity Index 3, use of colloids, emergent surgery, estimated blood loss >200 ml, preoperative anaemia, erythrocyte transfusion, open surgeries, and surgical duration >4 h (C-statistic=0.79, Hosmer Lemeshow=0.36). Conclusions: Greater perioperative fluid volume was independently associated with prolonged duration of recovery across a spectrum of surgical risk profiles. Fluid restriction should be considered a part of the care package in enhanced recovery after surgery programmes for colorectal surgery. Key words: colorectal surgery; fluid therapy; length of stay; perioperative care Colorectal cancer remains a leading cause of morbidity and mortality worldwide. 1 Surgical resection is the primary treatment for non-metastatic colorectal cancer; therefore, it is imperative to identify modifiable perioperative risk factors that might influence recovery. The influence of perioperative fluid volume on recovery after major abdominal surgery remains poorly understood. Historical practice involves administration of large volumes of i.v. fluid with the rationale to replace deficits caused by preoperative fasting, to prevent anaesthesia-induced hypotension, and to adjust for haemodynamic changes influenced Accepted: March 24, 2016 The Author Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please journals.permissions@oup.com 804

2 Fluid therapy in colorectal surgery 805 Editor s key points The impact of perioperative i.v. fluid volume on patient outcomes, such as length of hospital stay (LOS), is unclear. A single-centre retrospective review of patients undergoing colorectal surgery found an independent association between greater fluid volume and increased LOS. Although requiring confirmation by prospective randomized trials, these data suggest that restrictive fluid approaches could enhance recovery after colorectal surgery. by bleeding and hypothetical third space losses. However, the body of evidence supporting this management has been strongly contested. 2 Despite inconclusive evidence, enhanced recovery after surgery (ERAS) programmes recommend fluid restriction within their bundle of care for colorectal surgery. 3 With the wide uptake of ERAS programmes worldwide, it is increasingly important to conduct meaningful studies on the influence of perioperative fluid therapy on patient outcome and recovery. To address this, we conducted a single-centre (two-hospital) retrospective chart review capturing a 6 yr window of elective and emergent colorectal surgeries before the initiation of a comprehensive ERAS programme. Our primary objective was to determine whether the volume of i.v. fluid administered in the perioperative period was associated with the length of hospital stay (LOS) after colorectal surgery. Ultimately, this may help to inform practice related to the inclusion of fluid restriction in ERAS programmes. Methods This retrospective study was approved by our institution s Research Ethics Board (reference AE), which waved the requirement for informed consent. Study setting and patient cohort The study was conducted at two hospitals (Toronto General Hospital and Toronto Western Hospital) within the University Health Network (UHN), which is a tertiary referral centre in Toronto, Ontario, Canada. A chart review was conducted on consecutive, adult patients (>18 yr old) who underwent inpatient colorectal surgery between January 1, 2008 and December 31, Importantly, the epoch chosen corresponded to a period immediately before systematic initiation of an ERAS programme. 3 Minor procedures and procedures without significant bowel resection were excluded, including the following: colostomies, ileostomies, jejunostomies, laparoscopic adhesiolysis, and laparotomies without bowel resection. Patients admitted to the intensive care unit (ICU) before surgery would not qualify for ERAS management and were excluded. During the study period, fluid management was at the discretion of the attending anaesthetist, and cardiac output monitoring was not available. Data sources and linking of electronic data sets Data were retrieved from an electronic data warehouse using methodology described previously by Beattie and colleagues. 4 The three primary databases used were as follows: (i) electronic patient records; (ii) the operating room scheduling office system bookings database (ORSOS ; McKesson Corporation, San Francisco, CA, USA); and (iii) the UHN blood transfusion database HCLL (Hemocare Lifeline; Mediware, Chicago, IL, USA). Records were linked using patient medical record numbers and then de-identified. Predictor variables Predictor variables were selected based on previous studies and factors felt apriorito affect outcome. To assess systemic illness and co-morbidities, the Charlson Co-morbidity Index (CCI) 5 based on ICD-10 codes 6 and Revised Cardiac Risk Index were used. Other variables that were analysed included age, BMI, sex, preoperative haemoglobin and creatinine, and the following intraoperative variables: type of surgery (emergent or elective, laparoscopic or open laparotomy), surgical duration, epidural use, oncological status, i.v. fluids administered (in litres) and estimated blood loss (EBL). The i.v. fluid data included normal saline, balanced salt solution, hydroxyethyl starch, and albumin (5 and 25%). Red blood cell (RBC) transfusion were also collected. The perioperative period was defined as the duration from the commencement of the surgical procedure to discharge from the postanaesthetic care unit (PACU). For patients who were admitted directly to the ICU from the operating room (OR), the perioperative period was restricted to their surgical duration. Primary and secondary outcomes The primary outcome measured was LOS, defined as the duration from date of surgery to date of discharge. A prolonged LOS was defined as being greater than the median LOS for the entire cohort (8.2 days, interquartile range ). Values were truncated to the 90th percentile, and all patients who died in hospital were given an LOS of 26 days (90th percentile). Secondary outcomes included postoperative complications (myocardial infarction, acute renal failure, and pulmonary oedema) and postoperative mortality. Statistical analysis Statistical analysis was performed using SAS version (SAS institute Inc., Cary, NC, USA). Categorical variables were summarized as the frequency (percentage) and continuous variables as the mean (SD). Multivariable logistic regression The outcome of interest was prolonged LOS (LOS>median of 8.2 days). Univariate analyses and cubic spline curves 7 were conducted to assess the unadjusted association of population characteristics and candidate variables with prolonged LOS. Variables that were significantly different between the two groups (LOS 8.2 days and LOS>8.2 days) on univariate analysis or those demonstrated to influence surgical outcomes in previous publications 5 were then included in a stepwise multivariable logistic regression analysis. The fit of the logistic model was assessed using the Hosmer Lemeshow test (where P>5 suggests a well-fitting model), whereas the performance of the model was assessed using the receiver operating characteristic curve C-statistic. Quantile linear regression Using the same variables included in the multivariable logistic regression analysis, quantile linear regression analysis was conducted by dividing the population into various quantiles of LOS (10th percentile, 25th percentile, 50th percentile etc.) and using least squares regression analysis to determine whether fluid volume had a similar influence on each quantile to provide

3 806 Aga et al. Table 1 Patient characteristics and perioperative variables. Categorical values are displayed as count (percentage). Continuous variables are displayed as mean (SD) or median (25th percentile, 75th percentile). *Preoperative anaemia was defined as follows: haemoglobin <120 g litre 1 (female) and haemoglobin <130 g litre 1 (male) Caracteristics and variables Entire cohort (n=1242) Length of hospital stay Normal (<8.2 days) (n=627) Prolonged ( 8.2 days) (n=615) Age [yr; mean (SD)] 64 (SD = 15) 62 (SD = 15) 67 (SD = 16) Sex [female; n (%)] 579 (47) 290 (46) 289 (47) Weight [kg; mean (SD)] 74 (19) 75 (18) 74 (20) BMI [kg m 2 ; mean (SD)] 26.9 (6.3) 27.0 (5.8) 26.7 (6.7) Co-morbidities Preoperative anaemia* [n (%)] 731 (59) 181 (29) 282 (46) Oncology related [n (%)] 558 (45) 309 (49) 249 (40) Charlson co-morbidity score [n (%)] (32) 204 (32) 189 (31) (37) 248 (40) 211 (34) (22) 134 (21) 140 (23) (9.3) 41 (6.5) 75 (12) Preoperative laboratory variables Haemoglobin [g litre 1 ; mean (SD)] 125 (21) 129 (19) 121 (22) Creatinine (serum) [μmol l 1, mean (SD)] 95 (80) 91 (85) 100 (76) Surgical variables [n (%)] Laparoscopic 382 (31) 240 (38) 142 (23) Epidural 78 (6.3) 33 (5.3) 45 (7.3) Emergent 529 (43) 167 (27) 362 (59) Oncology 774 (62) 427 (68) 347 (56) Anaesthetic Duration (min) 246 (109) 236 (95) 258 (121) 232 (170, 300) Estimated blood loss (ml) 300 (630) 190 (300) 410 (830) 200 (0, 300) RBC transfusions (1 3 units) [n (%)] postoperative day (14) 50 (8.0) 168 (27) Intraoperative fluids Crystalloid (ml) 2400 (1200) 2000 (1500, 3000) 2200 (1000) 2600 (1300) Hydroxyethyl starch (ml) 710 (370) (n=464) 5% Albumin (ml) 990 (720) (n=41) Total intraoperative fluid (ml) 2700 (1400) 2500 (1800, 3500) Perioperative fluids Crystalloid (ml) 2800 (1300) 2600 (2000, 3500) Hydroxyethyl starch (ml) 730 (430) (n=524) 5% Albumin (ml) 960 (720) (n=50) Total perioperative fluid (ml) 3200 (1500) 2900 (2100, 4000) 650 (290) (n=172) 700 (570) (n=5) 750 (410) (n=292) 1000 (740) (n=35) 2400 (1100) 3000 (1500) 2600 (1200) 3000 (1400) 640 (350) (n=209) 750 (520) (n=6) 800 (470) (n=315) 980 (740) (n=44) 2900 (1300) 3500 (1700) a more comprehensive assessment of the magnitude of any isolated influence of fluids on LOS for patients with variable surgical risk profiles. Sensitivity analyses Our definition of the perioperative period included the time spent in the PACU, but patients who were admitted directly to the ICU after surgery bypassed the PACU. To address this, we examined the relationship between intraoperative fluids and prolonged LOS. We examined the relationship between perioperative fluids and LOS separately in both elective and emergent subgroups. The volume of perioperative fluid administered was included in our logistic modelling because it is a modifiable variable and fluid therapy is a focus of ERAS programmes. However, the amount of fluid administered in the OR increases with blood loss, urine output, and surgical duration. Therefore, we looked at the relationship between fluid balance [i.e. fluid administered (crystalloid and colloid administered) minus blood loss and urine output] expressed per kilogram per hour of surgery. Furthermore, as fluids are often administered to help mitigate blood loss, we examined the subgroup that did not receive blood transfusions. Lastly, we examined a definition of prolonged LOS being greater than the 75th percentile, because this is a commonly used outcome.

4 Fluid therapy in colorectal surgery 807 Results Study population and surgical characteristics During the study period, 1615 adult patients underwent colorectal surgery at UHN. Patients were excluded if they did not have a bowel resection (n=314, 19.4%), were in the ICU before surgery (n=30, 1.8%), or had incomplete data (n=29, 1.8%). The final cohort included 1242 (77%) patients, with average age in the mid-sixties, BMI consistent with modest obesity, and 47% female (Table 1). The most common co-morbidity was preoperative anaemia. Within the study population, 57% had elective surgeries and 31% were laparoscopic. Epidural anaesthesia was used in a minority (6.3%) of patients. The most common procedure type was open colonic surgery (43%), followed by open rectosigmoidal surgery (26%), and laparoscopic colonic surgery (20%; Table 2). Patients received an average of 3.2 (1.5) litres of i.v. fluid (crystalloid, hydroxyethyl starch, and 5% albumin) in the perioperative period, with crystalloids being the predominant solutions. There were 464 patients (37%) who received colloid (hydroxyethyl starch) during surgery. Patients lost 300 (630) ml of blood, and 176 patients (14%) received 1 3 unitsofrbcon the day of surgery (no patient received >3 units). Length of hospital stay and outcomes of colorectal surgery The median LOS for the cohort was 8.2 days (interquartile range 5.2, 14.7). There were 57 (4.6%) patients who died within 30 days of surgery, and 85 (6.8%) patients died within 90 days of surgery. Less than 2% of patients had postoperative pulmonary oedema or myocardial infarction. Although most patients were transferred to the PACU after surgery, 123 (10%) were admitted directly to the ICU. The majority of ICU admissions (85%) took place at an institution without an intermediate (step-down) care unit. For these patients, postoperative PACU fluid data were not available, and the perioperative period concluded at the end of their surgery. Univariate analysis Age, CCI, preoperative haemoglobin (in grams per litre) and creatinine (serum micromoles per litre), anaesthetic duration (in minutes), EBL (in millilitres), RBC transfusions (postoperative day 0, 1 3 units), and perioperative i.v. fluid volume (in litres) were different between the two groups (Table 1). Although patients with acute renal failure and myocardial infarction were more likely to have a prolonged LOS, the number of these outcomes, and the mortality, was too small to assess independent associations with fluid volume. Spline curves Spline function analysis revealed a dose response relationship between fluid volume and the probability of having a prolonged LOS (Fig. 1). All candidate variables (including all those in the multivariable model) showed a positive relationship with prolonged LOS on spline function analysis. Multivariable logistic regression analysis Variables included in the model were age, sex, BMI, CCI, preoperative anaemia, preoperative creatinine, perioperative fluid volume (in litres), use of colloids (yes/no), type of surgery (emergent), oncological status, EBL (in millilitres), RBC transfusions ( 1 unit of RBC on the day of operation or postoperative day 1), procedure type (laparoscopic vs open), year of surgery, epidural use, and duration of surgery. Perioperative fluid volume was independently associated with prolonged LOS with an odds ratio of 1.23 [95% confidence interval (CI) , P<1; Table 2]. This suggests that for every litre of fluid administered in the perioperative period, there was a 23% increase in the risk of prolonged LOS. Additional risk factors found to be associated with prolonged LOS included age >65 yr, CCI 3, emergent surgery, preoperative anaemia, RBC transfusion, estimated blood loss >200 ml, open surgery, and duration of surgery. Sensitivity analyses To account for fluids that were administered immediately after surgery, but not recorded for patients who were admitted directly to the ICU, sensitivity analysis was completed by considering intraoperative fluid volumes only. This produced a similar logistic regression model with identical independent variables, and intraoperative fluid volume remained independently associated with a prolonged LOS (odds ratio 1.24, 95% CI , P<1). Fluid balance was 8.0 (5.9), 6.4 (3.6), and 10 (7.4) ml kg 1 h 1 for the entire population, elective population, and emergent population, respectively. Intraoperative fluid balance remained associated with prolonged LOS for the entire population (odds ratio 1.02, 95% CI , P<1) and the elective population (odds ratio 1.09, 95% CI , P<1). In the emergent population, fluid balance was not associated with prolonged LOS (odds ratio 1.00, 95% CI ). In the population that did not receive a Table 2 Multivariable logistic regression model: association between candidate variables and prolonged length of hospital stay (C-statistic=0.79, Hosmer Lemeshow=0.36). Hb, haemoglobin; RBC, red blood cell Variable Description Odds ratio 95% Confidence interval P-value Age >65 yr old <1 Charlson Co-morbidity class 3 or <1 Perioperative fluid volume Litres Perioperative colloid Yes/no Surgical type Emergent <1 Estimated blood loss >200 ml <1 Preoperative anaemia Hb<120 g litre 1 (female) Hb<130 g litre 1 (male) RBC transfusion 1 unit RBC on day of surgery or postoperative day <1 Procedure type Open (colonic or rectosigmoidal) <5 Surgical duration >4 h <1

5 808 Aga et al. Probability of prolonged LOS Perioperative intravenous fluid volume (litres) Fig 1 Perioperative fluid volume and length of hospital stay (LOS). Unadjusted spline function graph displaying the influence of perioperative fluid volume (in litres) on the probability of having a prolonged LOS. The 95% confidence interval is displayed. blood transfusion (n=1024), fluid administered (odds ratio 1.35, 95% CI , P<1) and fluid balance (odds ratio 1.05, 95% CI , P<1) remained strongly associated with prolonged LOS. Another definition of prolonged LOS used previously is LOS >75th percentile. 8 When this definition was considered as our primary outcome, the volume of fluid administered in the perioperative period remained associated with prolonged LOS (odds ratio 1.22, 95% CI , P<1), in agreement with the quantile regression analysis. Quantile regression Quantile regression results suggest that each litre of fluid administered in the perioperative period increased hospital stay by days per litre of fluid administered. This is demonstrated graphically in Fig. 2A, with the 95% confidence interval always remaining above zero, suggesting that perioperative fluid volume was associated with prolonged hospital stay across the entire spectrum of LOS values. The upward trend suggests that the influence of fluid volume on LOS may be greater in patients who had longer recovery periods. For example, for patients within the 10th percentile of LOS, each litre of fluid increased their hospital stay by approximately half of a day, but for patients within the 75th percentile of LOS, each litre of fluid added almost 1 day to their hospital stay (Fig. 2A). When quantile regression was repeated for sensitivity analysis using intraoperative fluid volume only (Fig. 2B), the association between administered fluid volume and LOS was similar, with a more consistent upward trend. In both quantile regression analyses, we controlled for the same patient, surgical, and anaesthetic variables included in the logistic regression model. All variables significantly associated with prolonged LOS in logistic regression analysis remained significant in quantile analysis. The relationship between EBL and LOS was significant only for patients with a LOS below the 80th percentile. Discussion After reviewing a 6 yr window of colorectal surgeries before ERAS implementation, we found that the volume of i.v. fluid Additional postoperative hospital days A B Length of hospital stay quantile Fig 2 Quantile regression analysis: additional postoperative days per litre of fluid administered and length of hospital stay. The influence of each litre of i.v. fluid administered on quantiles of length of hospital stay. (A) Perioperative fluid. (B) Intraoperative fluid. administered in the perioperative period was independently associated with prolonged LOS. This independent association persists if the volume of fluid administered is restricted to the intraoperative period and regardless of whether prolonged LOS is defined as being greater than the median or 75th percentile. Moreover, quantile regression results suggest that there is an

6 Fluid therapy in colorectal surgery 809 association between perioperative fluid volume and LOS across the entire range of LOS from the 10th to the 90th percentiles of LOS. Additionally, the influence of administered fluid volume may be more important in patients anticipated to have a longer LOS (i.e. sicker patients), but this requires further evidence. The upward trend of quantile regression suggests a dose-dependent association; a finding also supported by the spline curve analysis. Together, these results suggest that reducing the volume of fluids administered in the perioperative period may be a meaningful, modifiable strategy to enhance postoperative recovery for colorectal surgery patients. Our sensitivity analysis found that patients undergoing elective surgery (i.e. those who are the focus of ERAS programmes) benefit more from fluid restriction than those undergoing emergent surgery. However, it is also reasonable to suggest that the fluid status of the emergent patient is more difficult to assess than that of the elective patient, and fluid management in these patients should be tailored to the patient s clinical status. The mechanism to explain the association between fluid volume and recovery time could be that larger fluid volumes increase tissue oedema. Increased tissue oedema has been correlated with adverse outcomes, such as poor wound healing, compromised pulmonary function, and delayed resumption of bowel function. 910 Given that normal saline is the most commonly used crystalloid globally, one must also consider the influence of large volumes of i.v. fluid on salt balance. Although the clinical significance is still unclear, renal dysfunction 11 and hyperchloraemic metabolic acidosis 12 have also been attributed to administration of large volumes of normal saline. 13 In addition to perioperative fluid volume, several other variables were found to be associated with increased LOS, namely age, CCI, perioperative colloid use, emergent surgery, EBL (>200 ml), preoperative anaemia, RBC transfusions, open surgery, and duration of surgery (>4 h). These results are comparable to previously published studies, 5 14 as one would expect that older, sicker patients, undergoing more surgery with greater blood loss, would take longer to recover. As colloid use is limited to situations where clinicians anticipate a large deficit in circulating volume, the use of a colloid is indicative of a greater i.v. fluid requirement and should not be considered an independent cause of prolonged LOS. Study strengths and implications Several aspects of this study make it an important addition to the evidence to date. Firstly, by using a retrospective cohort design, we bypassed the need to define liberal and restrictive fluid volumes, allowing us to capture any influence of fluids across a wide spectrum of volumes clinically administered. These results provide insight into the inconsistencies in previous trials. For example, our spline function analysis depicted the steepest increase in probability of prolonged LOS after administration of 3 litres of i.v. fluids. In a previous randomized controlled trial where patients in the restrictive group received <2 litres, 15 patients were observed to benefit from being randomized to the restrictive group. In contrast, patients in a similar study who received more than 4 litres in the restrictive group were observed not to benefit from fluid restriction. 16 With LOS as our primary outcome, these results can be used as a baseline for the future assessment of ERAS programmes, which aim to reduce LOS. This study is the first to use quantile regression analysis to look at the relationship between perioperative fluid volume and LOS. Using this form of analysis allowed us to assess the influence of fluid volume on LOS, considering the wide range of factors that might influence a patient s baseline LOS;some measured(e.g.age and BMI) and others unmeasured, such as psychosocial factors. Limitations The limitations of this work are inherent to its retrospective observational approach when using a single-centre institutional data set. For example, the incidence of complications (pulmonary oedema, acute renal failure, and myocardial infarction) in our population was relatively small (n=42) and might have been underestimated by the data abstraction used by the ICD-10 coders. Furthermore, although we included all the predictor variables available in our data set, there were numerous unmeasured variables that we were unable to consider (e.g. intraoperative hypotension) Patients with hypotension are often treated with additional fluids; therefore, fluid administration could be a surrogate for more complex or challenging surgeries resulting in intraoperative hypotension. Lastly, by combining both crystalloid and colloid fluid volumes in our analysis, we did not account for differences in the types of solutions administered. The main objective of this work was to instruct or guide the development of ERAS programmes. The primary outcome (LOS>median) and the fluid variable (fluid volume administered) were selected to be adapted easily to ERAS. From our sensitivity analysis using fluid balance per kilogram per hour, we learned that although fluid therapy might influence patient LOS, the strength of this influence is affected by other surgical risk factors inherent in longer surgical procedures and emergent surgery. Despite these limitations, the robust nature of the association between i.v. fluid volume and LOS demonstrated in our study supports prospective studies to confirm the impact of limiting or restricting the volume of i.v. fluids administered in the perioperative period. Our results cannot be used to define a fixed fluid restriction protocol because fixed protocols might not be associated with improved patient outcomes unless some components are made adaptable to a patient s clinical status. Restrictive fluid therapy might be beneficial to the euvolaemic, but harmful to the hypovolaemic patient. As such, while we await the results of the landmark RELIEF Trial (restrictive vs liberal fluid therapy in major abdominal surgery, NCT ) we must keep in mind that a lack of any difference between the treatment groups in previous small trials might be attributable to the inadequately informed application of a fixed fluid protocol, forming a treatment misalignment bias. 19 Conclusions This large retrospective observational study found a strong positive association between the volume of fluid administered in the perioperative period and LOS in patients undergoing colorectal surgery, before the implementation of ERAS programmes. Our results suggest that i.v. fluid volume is an independent predictor of duration of recovery and increases postoperative LOS for patients with varying surgical risk profiles. These data can be used to explain discrepancies in previous randomized controlled trials using fixed treatment protocols for either liberal or restrictive fluid therapy and provide baseline results for further evaluation and consideration of recommended fluid restriction therapy in ERAS programmes for colorectal surgery. Authors contributions Design of the study and data collection: Z.A. Data analysis and editing the manuscript: M.M. Statistical analysis: S.A.M.

7 810 Aga et al. Drafting the manuscript: Z.A., S.A.M. All authors read and approved the final manuscript. Acknowledgements Allan Okrainec and Robin McLeod participated in editing the manuscript. Declaration of interest None declared. Funding Department of Anesthesia and Pain Management at Toronto General Hospital, part of the University Health Network, Toronto, Ontario, Canada. References 1. O Dwyer. Randomized clinical trial of the effect of postoperative intravenous fluid restriction on recovery after elective colorectal surgery. Br J Surg 2006; 93: Corcoran T, Rhodes JEJ, Clarke S, Myles PS, Ho KM. Perioperative fluid management strategies in major surgery: a stratified meta-analysis. Anesth Analg 2012; 114: McLeod RS, Aarts M-A, Chung F, et al. Development of an enhanced recovery after surgery guideline and implementation strategy based on the knowledge-to-action cycle. Analysis 2015; 2: Beattie WS, Karkouti K, Wijeysundera DN, Tait G. Risk associated with preoperative anemia in noncardiac surgery: a single-center cohort study. Anesthesiology 2009; 110: Andrews E, McCourt M, O Ríordáin M. Enhanced recovery after elective colorectal surgery: now the standard of care. Ir J Med Sci 2011; 180: Sessler DI, Sigl JC, Manberg PJ, Kelley SD, Schubert A, Chamoun NG. Broadly applicable risk stratification system for predicting duration of hospitalization and mortality. Anesthesiology 2010; 113: Devlin TF, Weeks BJ. Spline functions for logistic regression modeling. Proceedings of the Eleventh Annual SAS Users Group International Conference. Cary, NC, 1986; Collins TC, Daley J, Henderson WH, Khuri SF. Risk factors for prolonged length of stay after major elective surgery. Ann Surg 1999; 230: Brandstrup B, Tønnesen H, Beier-Holgersen R, et al. Effects of intravenous fluid restriction on postoperative complications: comparison of two perioperative fluid regimens: a randomized assessor-blinded multicenter trial. Ann Surg 2003; 238: Grocott MP, Mythen MG, Gan TJ. Perioperative fluid management and clinical outcomes in adults. Anesth Analg 2005; 100: Hadimioglu N, Saadawy I, Saglam T, Ertug Z, Dinckan A. The effect of different crystalloid solutions on acid-base balance and early kidney function after kidney transplantation. Anesth Analg 2008; 107: Morgan TJ, Venkatesh B, Hall J. Crystalloid strong ion difference determines metabolic acid base change during acute normovolaemic haemodilution. Intensive Care Med 2004; 30: Myburgh JA, Mythen MG. Resuscitation fluids. N Engl J Med 2013; 369: Kelly M, Sharp L, Dwane F, Kelleher T, Comber H. Factors predicting hospital length-of-stay and readmission after colorectal resection: a population-based study of elective and emergency admissions. BMC Health Serv Res 2012; 12: Lobo DN, Bostock KA, Neal KR, Perkins AC, Rowlands BJ, Allison SP. Effect of salt and water balance on recovery of gastrointestinal function after elective colonic resection: a randomised controlled trial. Lancet 2002; 359: MacKay G, Fearon K, McConnachie A, Serpell M, Molloy R, O Dwyer P. Randomized clinical trial of the effect of postoperative intravenous fluid restriction on recovery after elective colorectal surgery. Br J Surg 2006; 93: Walsh M, Devereaux PJ, Garg AX, et al. Relationship between intraoperative mean arterial pressure and clinical outcomes after noncardiac surgery: toward an empirical definition of hypotension. Surv Anesthesiol 2014; 58: Monk TG, Bronsert MR, Henderson WG, et al. Association between intraoperative hypotension and hypertension and 30-day postoperative mortality in noncardiac surgery. JAm Soc Anesthesiol 2015; 123: Deans KJ, Minneci PC, Suffredini AF, et al. Randomization in clinical trials of titrated therapies: unintended consequences of using fixed treatment protocols. Crit Care Med 2007; 35: Handling editor: H. C. Hemmings