When making the decision to transfuse

Transcription

1 BLOOD MANAGEMENT Clinical predictors of postoperative hemoglobin drift Michael C. Grant, 1 Glen J. Whitman, 2 Will J. Savage, 3 Paul M. Ness, 4 and Steven M. Frank 1 BACKGROUND: The decision to transfuse red blood cells in surgical patients should be based on multiple clinical variables, rather than on isolated hemoglobin (Hb) measurements alone. An important but often unrecognized clinical variable is the postoperative downward drift in Hb concentration (Hb drift), but the etiology, predictors, and time course of Hb drift are not well understood. STUDY DESIGN AND METHODS: Data were retrospectively collected for patients who did not receive postoperative transfusion. Initially, 11 common surgical procedures from one institution (n = 3179) were compared to assess clinical predictors of Hb drift. Data were analyzed in detail for two procedures associated with a large Hb drift (Whipple [n = 82] and lumbar spinal fusion [n = 74]), to determine the clinical predictors and temporal pattern of postoperative Hb drift. RESULTS: Surgical procedures with greater intraoperative intravenous (IV) fluid and blood requirements had greater postoperative Hb drift. The maximum Hb drifts after the Whipple and spinal fusion procedures were 2.5 ± 1.1 g/dl (occurring on Day 4, p < ) and 1.8 ± 1.2 g/dl (on Day 3, p < ), respectively. After the nadir, a 0.6 g/dl upward Hb drift (p < ) occurred after both procedures, resulting in a total drift after Whipple and spinal fusion of 1.9 ± 1.2 g/dl (p < ) and 1.3 ± 1.2 g/dl (p < ), respectively. Type of surgery (p = 0.03), intraoperative blood loss (p = 0.003), and a positive perioperative fluid balance (p = ) were independent predictors of Hb drift. CONCLUSIONS: Postoperative Hb drift was greater after surgical procedures with greater intraoperative IV fluid and blood requirements. Recognition of Hb drift may be an important facet of perioperative patient blood management. When making the decision to transfuse red blood cells (RBCs) during the perioperative period, clinicians should take into account multiple clinical variables rather than rely on isolated hemoglobin (Hb) measurements alone. 1 During surgery, variables such as intravascular volume depletion and ongoing or potential blood loss effectively increase the Hb concentration at which RBC transfusion is considered to be appropriate. Another variable is the occurrence of postoperative Hb drift, defined as the downward trend in Hb concentration during the first few days after surgery. The clinical variables associated with postoperative Hb drift have been described recently for cardiac surgery, 2 but the literature contains little information regarding Hb drift after noncardiac surgical procedures. The recommendation of a restrictive transfusion strategy has evolved over the past decade based on a number of transfusion guidelines endorsed by various societies 1,3-5 as well as several well-designed large randomized trials These guidelines support a suggested Hb transfusion trigger of 7 to 8 g/dl, but this trigger assumes that patients are not actively bleeding and intravascular ABBREVIATIONS: AIMS = anesthesia information management system; IQR(s) = interquartile range(s); POD = postoperative day. From the 1 Department of Anesthesiology/Critical Care Medicine, the 2 Department of Surgery, Division of Cardiac Surgery, and the 4 Department of Pathology (Transfusion Medicine), The Johns Hopkins Medical Institutions, Baltimore, Maryland; and 3 Department of Pathology (Transfusion Medicine), The Brigham and Women s Hospital, Boston, Massachusetts. Address correspondence to: Steven M. Frank, MD, Department of Anesthesiology, Zayed 6208, The Johns Hopkins Medical Institutions, 1800 Orleans Street, Baltimore, MD 21287; sfrank3@jhmi.edu. Support was provided from institutional and/or departmental sources and from The New York Community Trust (New York, NY). Received for publication August 3, 2013; revision received September 23, 2013, and accepted September 30, doi: /trf TRANSFUSION 2014;54: TRANSFUSION Volume 54, June 2014

2 POSTOPERATIVE Hb DRIFT volume is replete. During surgery, when blood loss and hypovolemia can be substantial, it is sometimes appropriate to transfuse RBCs at a higher Hb trigger. An unanswered question is whether the perioperative transfusion trigger should be altered to account for the likelihood of a postoperative downward Hb drift, but this would require a better understanding of the causes, predictors, and magnitude of drift. In an effort to define the characteristics of postoperative Hb drift, we evaluated patients who underwent various noncardiac surgical procedures to quantitatively assess changes in postoperative Hb concentration and to determine the variables associated with these changes. We anticipate that if predictors of Hb drift were better understood, this information would be useful in making decisions to transfuse RBCs during and after surgery. MATERIALS AND METHODS Analysis 1: Hb drift after 11 common surgical procedures After receiving institutional review board approval, we obtained data from two separate databases designed to assess perioperative blood utilization. The first is a blood management intelligence portal (IMPACT online, Haemonetics Corp., Braintree, MA) that included all transfused blood components and Hb laboratory values for all inpatients (from admission to discharge) at Johns Hopkins Hospital over a 34-month period (January 2010 to November 2012). This represented the time period for which data were available from both databases at the time of analysis. These data were extracted from three different computer systems within our medical center and were made available through a Web-based application designed to provide data to assess and optimize patient blood management. The second database included only intraoperative data, derived from an anesthesia information management system (AIMS; Metavision, imdsoft, Needham, MA). This database records intraoperative blood components transfused and Hb laboratory values for all surgical procedures performed during the same time period defined above. After we excluded all outpatient surgeries from the AIMS database and all nonsurgical patients from the IMPACT online database, 43,400 surgical patients remained. By combining these two databases matched on unique hospital visit numbers, we evaluated all RBC transfusions and Hb values from both the intraoperative and the postoperative periods. To assess the natural course of postoperative Hb drift, we excluded 7890 patients who received RBC transfusions in the postoperative period, leaving 35,510 patients remaining. We then selected patients undergoing 11 common surgical procedures for which Hb measurements were typically obtained, resulting in the final population for analysis, which was 3179 patients. All Hb concentration measurements were performed either in the central core laboratory by the electronic sensing zone method (Z series Coulter counter, Beckman Coulter, Inc., Indianapolis, IN) or in the STAT laboratory by the cooximetry method (ABL 800 Flex, Radiometer America, Inc., Westlake, OH). We have previously determined that the difference in results between these two methods is negligible when the same sample is tested (<0.1 g/dl). Postoperative Hb drift was defined as the last Hb value measured before hospital discharge minus the last Hb value measured during surgery, provided that the latter was obtained after the last intraoperative RBC transfusion. We were able to identify whether the last intraoperative Hb value was measured after the last RBC transfusion by using the time stamps in the AIMS database, as we have described previously. 11 After calculating Hb drift, we compared the 11 surgical procedures in rank order from greatest to least Hb drift. Predictors of Hb drift were then assessed by univariate and multivariate analyses. Analysis 2: Hb drift after two high-drift surgical procedures We chose the two surgical procedures with the greatest postoperative Hb drift from Analysis 1 for further detailed analysis. After receiving institutional review board approval, we retrospectively collected data for 110 patients who had pancreaticoduodenectomy (Whipple) and 149 patients who had single or multiple vertebral level lumbosacral spinal fusion surgery with instrumentation. To assess the natural course of postoperative Hb drift, we excluded 103 patients who received postoperative RBC transfusion. Thus, the final analysis included 82 patients who had the Whipple procedure and 74 patients who had the spinal fusion procedure. Intraoperative data were obtained from the AIMS database, and postoperative data were obtained from the hospital s electronic medical records (Sunrise Clinical Manager, Eclipsys, Inc., Atlanta, GA). Daily measured Hb values were collected. If more than one measurement was obtained, the first measurement made after midnight was used to represent the Hb value for that day. The incidence of missing Hb values was 3%, which reflects the fact that Hb concentration was not measured every 24 hours for all patients. Postoperative Hb drift was assessed by three distinct methods. First, the maximum postoperative Hb drift was defined as the lowest Hb value in the postoperative period minus the first postoperative Hb value obtained upon admission to the postoperative care unit. Second, total postoperative Hb drift was defined as the last Hb value before discharge minus the first postoperative Hb value obtained upon admission to the postoperative care unit. Third, upward postoperative Hb drift was calculated as Volume 54, June 2014 TRANSFUSION 1461

3 GRANT ET AL. the discharge Hb value minus the lowest postoperative Hb value. Data from only the first 7 postoperative days were analyzed. If the length of stay exceeded 7 days, then Postoperative Day (POD) 7 was considered to be the day of discharge. Our preliminary study showed that the Hb nadir typically occurs on POD 3 or 4, 12 suggesting that analysis beyond POD 7 was unnecessary for the purposes of this study. Predictors of maximum and total postoperative Hb drift were assessed by univariate and multivariate analyses. Also, for the patients included in Analysis 2, the POD of nadir Hb and length of stay in days were determined for each patient. We also quantitatively assessed intraoperative and postoperative fluid and blood product administration by reviewing the electronic medical records. Fluid balance was calculated for the entire perioperative period (intraoperative plus postoperative until discharge or POD 7) by subtracting all output fluids (urine, estimated blood loss, and surgical drain output) from all input fluids (intravenous [IV] crystalloid, colloid, blood components, and medications). Oral fluid was also included as fluid intake for postoperative patients. Insensible fluid losses were not included in these calculations since they are difficult to measure and thought to be relatively small compared to other fluid requirements. Statistical analysis All data were processed and analyzed with computer software (Excel, Version , Microsoft, Inc., Redmond, WA; and JMP, Version 9.0.0, SAS Institute, Cary, NC). For Analysis 2, a preliminary power analysis was performed with an alpha of 0.05 and a power of 0.8; it was determined that a sample size of 128 patients would be needed to show a decrease in Hb concentration of 0.5 g/dl with a standard deviation (SD) of 1 g/dl. To ensure an adequate sample size for Analysis 2, we chose to analyze 156 patients. For normally distributed variables, comparisons between means were made with the paired or unpaired t test or analysis of variance (ANOVA) for repeated measures, where appropriate. For nonparametric data, medians and interquartile ranges (IQRs) were compared by the Wilcoxon signed-rank test. Dichotomous variables were assessed by chi-square or Fisher s exact tests, where appropriate. In an effort to determine the clinical variables that were predictors of postoperative Hb drift, we performed both univariate and multivariate analyses in which Hb drift was treated as a continuous variable. Univariate testing was carried out by simple linear regression for continuous independent variables. Multiple linear regression was used to determine independent predictors of maximum and total Hb drift. All variables of interest were entered into the model. We eliminated variables from the multivariate model in a stepwise fashion using backward elimination, while retaining all variables with p values of not more than Data are given as mean ± SD or median (IQR) as appropriate. p < 0.05 defined significance. RESULTS Analysis 1: Hb drift after 11 common surgical procedures The patient characteristics and intraoperative fluid and transfusion requirements for patients who had one of the 11 selected surgical procedures are shown in Table 1. Postoperative Hb drift was compared among these procedures and graphically displayed in rank order (Fig. 1). The surgical procedures associated with greater blood and fluid requirements were generally associated with a greater postoperative Hb drift. The predictors of increased postoperative Hb drift by both univariate and multivariate analyses were the type of surgical procedure, advanced Surgical procedure Age (years) TABLE 1. Patient characteristics (Analysis 1) Sex (% male) Body mass (kg) First Hb (g/dl) EBL (ml) Intraoperative crystalloid (ml) % given intraoperative colloid % given intraoperative RBCs Spinal fusion (n = 532) 40 ± ± ± ± ± Pancreatic (n = 589) 62 ± ± ± ± ± Liver resection (n = 263) 58 ± ± ± ± ± Open aortic (n = 29) 65 ± ± ± ± ± Open abdominal (n = 257) 49 ± ± ± ± ± Thoracoscopic (n = 229) 58 ± ± ± ± ± Endovascular aortic (n = 23) 72 ± ± ± ± ± Renal transplant (n = 229) 52 ± ± ± ± ± Lumbar discectomy (n = 17) 49 ± ± ± ± ± Craniotomy (n = 979) 46 ± ± ± ± ± Carotid endarterectomy (n = 32) 72 ± ± ± ± ± EBL = estimated blood loss TRANSFUSION Volume 54, June 2014

4 POSTOPERATIVE Hb DRIFT age, male sex, greater intraoperative blood loss, and greater intraoperative crystalloid (Table 2). Analysis 2: Hb drift after two high-drift surgical procedures A subset of patients who underwent one of two surgical procedures associated with a largest Hb drift in Analysis 1 (Whipple procedure and lumbosacral spinal fusion) was analyzed in further detail. A comparison of these two groups of patients is shown in Table 3. For the spinal fusion patients, the median (IQR) number of vertebrae that were fused was 2 (1-3), and the range was from one to six vertebrae. Those patients who had the spinal fusion surgery were significantly younger, had a greater body mass, and received less intraoperative crystalloid and colloid; however, these patients received more blood components (RBCs and fresh-frozen plasma [FFP]). Total intraoperative and perioperative positive fluid balance was greater in those who had the Whipple procedure than in those who had the spinal fusion. Maximum and total postoperative Hb drift were also greater in patients who had the Whipple procedure. Upward postoperative Hb drift was similar after both procedures. The median hospital length of stay for patients who had the Whipple procedure was twice that for patients who had spinal fusion. Mean Hb at the time of hospital discharge was lower in the Whipple patients than in the spinal fusion patients. The daily mean Hb concentrations in the two groups of patients are shown in Fig. 2. Patient distributions of maximum, total, and upward postoperative Hb drift are shown as histograms in Fig. 3. All 156 patients who had the Whipple or spinal fusion procedure without postoperative transfusion were grouped together for the purposes of this analysis. The range of maximum and total Hb drift among these patients was wide, and most had some upward Hb drift (mean, 0.6 g/dl; range, g/dl). Of the 156 patients included, 108 (69%) had an upward Hb drift after reaching the Hb nadir and were discharged from the hospital with a Hb concentration above their nadir. The upward Hb drift was similar between the two surgical procedures that were analyzed (Table 3). For all patients combined, the mean Hb concentration upon discharge from the hospital (9.8 ± 1.1 g/dl) was significantly greater than the mean nadir Hb concentration (9.2 ± 1.0 g/dl; p < ). Daily and total perioperative fluid balance are shown in Fig. 4. Although both groups of patients had a positive fluid balance on the day of surgery, the total positive fluid balance was significantly greater for those who had the Whipple procedure (p < ). For the entire perioperative period, the net positive fluid balance was Fig. 1. Postoperative Hb drift is compared among 11 surgical procedures. For this analysis, drift was defined as the last Hb concentration measured before hospital discharge minus the Hb concentration measured at the end of surgery. The surgical procedures associated with greater IV fluid and blood transfusion requirements (shown in Table 1) were likely to have a greater magnitude of postoperative Hb drift. The difference between Hb drift among surgical procedures was significant (p < ) by one-way ANOVA. Data are shown as mean ± SD. TABLE 2. Univariate and multivariate correlates of total postoperative Hb drift (Analysis 1) Univariate Multivariate Variable Coefficient (95% CI) p value* Coefficient (95% CI) p value Surgical procedure < < Age (per year) ( to 0.005) < ( to 0.008) < Male sex 0.23 ( 0.17 to 0.30) < ( to 0.116) < Body mass (/kg) Intraoperative EBL (/L) 0.24 ( 0.3 to 0.17) < ( 0.22 to 0.032) < Intraoperative crystalloid (/L) 0.11 ( 0.13 to 0.98) < ( to 0.039) < * Probability value based on simple linear regression. Probability value based on multiple linear regression. EBL = estimated blood loss. Volume 54, June 2014 TRANSFUSION 1463

5 GRANT ET AL. Characteristic TABLE 3. Patient characteristics (Analysis 2) Whipple procedure (n = 82) Lumbosacral spinal fusion (n = 74) p value Age (years) 64 ± ± Sex, number (% male) 42 (51) 35 (47) 0.62 Body mass (kg) 77 ± ± Intraoperative EBL (ml) 630 ± ± Intraoperative crystalloid (ml) 5550 ± ± Intraoperative colloid (%, ml/all pts.) 20, 125 ± 270 8, 40 ± Intraoperative RBC transfusion (%, U/all pts.) 16, 0.27 ± , 1.0 ± Intraoperative FFP transfusion (%, U/all pts.) 2.7, 0.03 ± , 0.29 ± Intraoperative fluid balance (input output; ml)* 4000 ± ± Perioperative fluid balance (input output; ml)* 9067 ± ± 4116 < Cumulative surgical drain output (ml) 1245 ± ± Length of stay (days; median, IQR) 8 (7-12) 4 (3-5) < Postoperative day of nadir (median, IQR) 4 (3-5) 3 (2-4) < Mean preoperative Hb (g/dl) 12.7 ± ± Mean first postoperative Hb (g/dl) 11.7 ± ± Mean hospital discharge Hb (g/dl) 9.6 ± ± Mean maximum Hb drift (g/dl) 2.5 ± ± Mean total Hb drift (g/dl) 1.9 ± ± Mean upward Hb drift (g/dl) 0.6 ± ± * Fluid balance = sum of all fluids and blood product volume administered minus the sum of blood loss, drain output, and urine output volumes. Maximum Hb drift = nadir postoperative Hb minus initial postoperative Hb. Total Hb drift = last Hb before discharge minus initial postoperative Hb. EBL = estimated blood loss; U/all pts. = all units of blood product divided by all patients. greater for patients who had the Whipple procedure (approx. 9 L) than for those who had spinal fusion surgery (approx. 3 L, p < ; Table 3 and Fig. 4). The variables significantly associated with a greater magnitude of total postoperative Hb drift by univariate and multivariate analyses are shown in Table 4. The variables associated with greater total Hb drift in the multivariate analysis were surgical procedure (Whipple; p = 0.03), greater intraoperative blood loss (p = 0.003), and a positive perioperative fluid balance (p = ). The variables significantly associated with a greater magnitude of maximum postoperative Hb drift by univariate and multivariate analyses are shown in Table 5. The variables associated with greater maximum Hb drift in the multivariate analysis were type of surgery (Whipple; p = 0.04) and a positive perioperative fluid balance (p = 0.04). DISCUSSION The results indicate that the magnitude of postoperative downward Hb drift differs depending on the type of surgical procedure, and that surgeries associated with greater intraoperative blood loss, transfusion, and IV fluid requirements are associated with a greater Hb drift. With these major surgical procedures, the mean magnitude of drift to the nadir Hb concentration was approximately 2.0 to 2.5 g/dl over the first 3 or 4 postoperative days, followed by a smaller upward Hb drift (approx. 0.6 g/dl). The primary predictors of greater Hb drift were the type of surgical procedure, greater intraoperative blood loss, and Fig. 2. Mean daily postoperative Hb concentrations are plotted for patients who had lumbosacral spinal fusion surgery ( ; n = 74) or the Whipple procedure ( ; n = 82). To determine the natural course of postoperative changes in Hb, only patients who did not require postoperative RBC transfusion were included. For both surgical procedures, starting from the baseline initial postoperative Hb value (admit PACU or ICU), there was a significant decrease in mean Hb beginning on POD1(p< ) that remained significantly lower than baseline (p < ) for the remainder of the hospital stay (by ANOVA for repeated measures). Data are shown as mean ± SD TRANSFUSION Volume 54, June 2014

6 POSTOPERATIVE Hb DRIFT Fig. 3. Histograms showing the number of patients from Analysis 2 who exhibited various degrees of maximum, total, and upward Hb drift after Whipple or spinal fusion surgery. Maximum Hb drift is defined as the nadir postoperative Hb concentration minus the initial postoperative Hb concentration. Total Hb drift is defined as the Hb concentration upon hospital discharge minus the initial postoperative Hb concentration. Upward Hb drift is defined as the discharge Hb concentration minus the nadir postoperative Hb concentration. The findings demonstrate a broad range of downward Hb drift after the surgical procedures and a much smaller upward drift in the majority of patients. For all three of these variables, the mean drift values (mean ± SD) are shown in Table 3. a positive perioperative fluid balance. The clinical implications of these findings are the recognition that patients with these risk factors are susceptible to downward Hb drift, but the decision to transfuse should also take into account the subsequent upward drift that occurs as overall fluid balance becomes negative. Our findings are similar to those described after cardiac surgery, for which the mean downward drift was 1.8 g/dl over 4 days, followed by 0.7 g/dl of upward drift. 2 Cardiac surgery, however, is inherently different from the surgical procedures that we assessed. During cardiac surgery, extracorporeal circulation is utilized, and the duration of cardiopulmonary bypass was shown to be a strong predictor of Hb drift. 2 Of note is the similar finding in the cardiac and noncardiac surgery studies that after the postoperative nadir Hb is reached, a small but significant upward Hb drift occurs after POD 3 or 4. In the current study, this upward Hb drift occurred in more than two-thirds of patients (ranging from 0 to 2.5 g/dl). If clinicians are cognizant of the potential for upward drift after the Hb nadir is reached, it is possible that RBC transfusions may be avoided in some patients. The predictors of Hb drift that were common to both the first and the second analyses were the surgical procedure, intraoperative blood loss, and IV fluid requirements. Two additional predictors, however, increased age and male sex, were predictors in the first analysis. This difference in results between the two analyses may be related to the different patient populations that were included. The first analysis, which included a wide variety of surgical procedures with a range of intraoperative fluid and blood requirements, resulted in a wide range of Hb drift. In the second analysis, we included two specific surgical procedures associated with a large Hb drift: one a blood loss procedure (spinal fusion) and the other an interstitial compartment, fluid-seeking procedure (Whipple). Under these circumstances, age and sex were not significant predictors of Hb drift. We can only speculate on the reason for an increased propensity for Hb drift in the elderly. There is evidence for a reduction in hematopoietic reserves in elderly patients, 14 possibly due to a decreased responsiveness to erythropoietin, 15 which could contribute to a greater postoperative Hb drift. Although the exact causes for the downward postoperative Hb drift are not well understood, there are several possible contributing factors. These include ongoing bleeding, blood loss due to laboratory testing, hemodilution, decreased erythropoiesis, and a shortened life span of transfused RBCs. Although our study does not allow us to determine the relative contribution of these five factors, they are worthy of discussion. Surgical bleeding does not necessarily stop when patients leave the operating room. Ongoing postoperative blood loss is likely to contribute to Hb drift, and it is plausible that certain surgical procedures with high intraoperative blood loss are more likely to have postoperative bleeding. This propensity for postoperative bleeding may explain why intraoperative blood loss was an independent predictor of total postoperative Hb drift, even after adjusting for other variables by multivariate analysis. It would be helpful if postoperative blood loss could be accurately measured by assessing output through surgical drains, but not all patients have drains, not all bleeding will exit the drains when they are present, and the fluid that does exit Volume 54, June 2014 TRANSFUSION 1465

7 GRANT ET AL. Fig. 4. Perioperative fluid balance is compared between patients who had lumbosacral spinal fusion surgery and those who had the Whipple procedure, two procedures associated with significant postoperative Hb drift. On the day of surgery (POD 0) and on POD 1, patients who had the Whipple procedure had a large positive fluid balance that transitioned to a negative fluid balance beginning on POD 2. The patients who had spinal fusion surgery had a positive fluid balance on POD 0 and then transitioned to a negative fluid balance on POD 1. For the entire perioperative period, the cumulative fluid balance was approximately 9 L for the Whipple patients and approximately 3 L for the spinal fusion patients (inset). Data are shown as mean ± SD. *p < versus spinal fusion; #p < 0.01 versus spinal fusion. the drains is often serosanguinous, with a lower hematocrit than that of the patient s blood. 16 Postoperative blood loss due to phlebotomy for laboratory testing can be substantial and contributes to hospital-acquired anemia, 17 especially in the intensive care units, where daily blood draws can be as much as 50 to 100 ml/day. 18,19 A daily blood loss of 50 ml represents approximately 1% of total blood volume or roughly the amount of RBCs that are produced each day by erythropoiesis. 20 Adding to blood loss from laboratory testing is the practice of discarding the initial blood drawn from arterial or central venous catheters, which can be reduced by using self-contained syringe sets to allow the return of this blood back to the patient in a sterile fashion. 19 Such devices are associated with a reduction in blood lost to testing, less decline in Hb concentration, and even reduced requirement for RBC transfusion. 21 One of the primary predictors of increased drift in our analysis was an overall net positive fluid balance, suggesting that hemodilution contributes significantly to postoperative Hb drift after some surgical procedures. Patients undergoing the Whipple procedure, for example, are approximately 8 L positive for fluid balance by the end of POD 1 (Fig. 3). Much of this fluid enters the third space (the interstitial space) in the form of edema in traumatized tissues. Postoperatively, this third-space fluid can be redistributed back into the vascular and may contribute to the downward Hb drift. 22 Additional evidence for fluid balance as a primary contributing factor is the upward Hb drift that occurs after overall fluid balance becomes negative and excess fluids are eliminated. Since the temporal pattern for Hb drift was similar in the spinal fusion patients despite the absence of a large positive fluid balance, this suggests alternative reasons for drift such as ongoing postoperative blood loss. Another potential reason for Hb drift is impaired erythropoiesis in postoperative patients. Evidence has shown that the inflammatory response in critically ill or postoperative patients is associated with increased levels of hepcidin, a peptide that blocks iron transport and thereby leads to reduced erythropoiesis. Hepcidin specifically reduces the absorption of iron across the gut mucosa and the release of stored iron from macrophages, consequences that are both associated with anemia. 25,26 Since roughly 1% of RBCs are destroyed and produced each day in healthy individuals, impaired erythropoiesis leads to a net loss in RBCs and likely contributes to postoperative Hb drift. 20 The relative contribution of impaired erythropoiesis to postoperative Hb drift remains to be determined and should be the focus of future investigations. The last proposed contributing factor to postoperative Hb drift is the decreased survival, or effective lifespan, of transfused RBCs, 27 especially those that have been stored for long time intervals (>25 days) between donation and transfusion. 28 Flow cytometry has shown that RBC survival 24 hours after transfusion is 86% after short-term storage and 73% after long-term storage. 28 If a patient has received a large number of RBC units, especially those stored for longer durations, one might expect a relatively large proportion of these RBCs to be removed from the circulation in the first few postoperative days, thus contributing to postoperative Hb drift. Our findings support this hypothesis, as the number of intraoperative transfused RBC units was a predictor of increased postoperative Hb drift. Whether the storage duration of RBCs is 1466 TRANSFUSION Volume 54, June 2014

8 POSTOPERATIVE Hb DRIFT TABLE 4. Univariate and multivariate correlates of total postoperative Hb drift (Analysis 2) Univariate Multivariate Variable Coefficient (95% CI) p value* Coefficient (95% CI) p value Surgery (Whipple) 0.30 ( 0.11 to 0.49) ( 0.12 to 0.5) 0.03 Age (/year) ( 0.02 to 0.01) Male sex ( to 0.263) Body mass (/kg) ( to 0.007) Surgical drain output (/L) ( 0.32 to 0.14) Intraoperative EBL (/L) 0.7 ( 1.17 to 0.3) ( 1.08 to 0.22) Intraoperative crystalloid (/L) 0.17 ( 0.27 to 0.07) Intraoperative RBC units 0.20 ( to 0.039) Perioperative fluid balance (/L) ( 0.1 to 0.027) ( 0.10 to 0.024) * Probability value based on simple linear regression. Probability value based on multiple linear regression; final model included surgical procedure, intraoperative EBL, and perioperative fluid balance. EBL = estimated blood loss. TABLE 5. Univariate and multivariate correlates of maximum postoperative Hb drift (Analysis 2) Univariate Multivariate Variable Coefficient (95% CI) p value* Coefficient (95% CI) p value Surgery (Whipple) 0.34 ( 0.16 to 0.52) ( to 0.462) 0.04 Age (/year) ( to 0.005) Male sex ( to 0.219) Body mass (/kg) ( to 0.008) Surgical drain output (/L) ( to 0.01) Intraoperative EBL (/L) 0.8 ( 1.17 to 0.36) ( 1.2 to 0.36) 0.13 Intraoperative crystalloid (/L) 0.17 ( 0.27 to 0.078) Intraoperative RBC units 0.22 ( 0.37 to 0.073) ( 0.37 to 0.073) 0.12 Perioperative fluid balance (/L) 0.07 ( to 0.035) ( to 0.002) 0.04 * Probability value based on simple linear regression. Probability value based on multiple linear regression; final model included surgical procedure, intraoperative EBL, intraoperative RBC units, and perioperative fluid balance. EBL = estimated blood loss. directly correlated with postoperative Hb drift remains to be determined and could be the focus of future studies. Certain limitations should be recognized when interpreting the findings from this study. First, to define the natural pattern of Hb drift, we excluded patients who received RBC transfusions in the postoperative period, so we are unable to comment on Hb drift in this subset of patients. It is plausible, however, that patients with the greatest drift would be more likely to require RBC transfusion, and thus if excluding transfused patients introduces bias, we would err on the side of underestimating the drift. Second, the retrospective nature of the study is a significant limitation. For this reason, the Hb measurements included in the analysis were those that were ordered by the clinicians caring for those patients; hence, some values were missing. The overall number of missing Hb measurements, however, was relatively small (3%) and should not have significantly altered the findings. Since data were obtained retrospectively for Analysis 2, we relied on manual review of electronic medical records, which may be less accurate than prospective data collection. Third, if patients were discharged from the hospital before POD 7, then Hb values would be missing for the days after discharge. This limitation could theoretically lead to an overestimation of postoperative Hb drift because patients who underwent less complicated surgeries (with perhaps less Hb drift) may have been discharged sooner, leaving those with more complicated surgeries in the analysis. It is likely, however, that the lowest postoperative Hb value would have occurred before discharge, and thus our analysis of maximum Hb drift would be reliable. In conclusion, our findings indicate that patients having surgical procedures associated with greater blood loss and fluid requirements experience a greater magnitude of postoperative downward Hb drift. After the Hb nadir was reached, a subsequent upward drift occurs, which should be taken into account when making the decision to transfuse. Increased awareness of the natural course of postoperative Hb changes is important in making informed blood management decisions. Volume 54, June 2014 TRANSFUSION 1467

9 GRANT ET AL. ACKNOWLEDGMENTS The authors thank Claire Levine for editorial assistance; Sharon Paul for assistance with data acquisition; Liz Dackiw for data collection and analysis; and Deborah Popoli for data coding, processing, and export. CONFLICT OF INTEREST The authors report no conflicts of interest or funding sources. REFERENCES 1. Napolitano LM, Kurek S, Luchette FA, et al. Clinical practice guideline: red blood cell transfusion in adult trauma and critical care. J Trauma 2009;67: George TJ, Beaty CA, Kilic A, et al. Hemoglobin drift after cardiac surgery. Ann Thorac Surg 2012;94: Carson JL, Grossman BJ, Kleinman S, et al. Red blood cell transfusion: a clinical practice guideline from the AABB. Ann Intern Med 2012;157: Ferraris VA, Brown JR, Despotis GJ, et al update to the Society of Thoracic Surgeons and the Society of Cardiovascular Anesthesiologists blood conservation clinical practice guidelines. Ann Thorac Surg 2011;91: Nuttall GA, Brost BC, Connis T, et al. Practice guidelines for perioperative blood transfusion and adjuvant therapies. Anesthesiology 2006;105: Carson JL, Terrin ML, Noveck H, et al. Liberal or restrictive transfusion in high-risk patients after hip surgery. N Engl J Med 2011;365: Hajjar LA, Vincent JL, Galas FR, et al. Transfusion requirements after cardiac surgery: the TRACS randomized controlled trial. JAMA 2010;304: Hebert PC, Wells G, Blajchman MA, et al. A multicenter, randomized, controlled clinical trial of transfusion requirements in critical care. Transfusion Requirements in Critical Care Investigators, Canadian Critical Care Trials Group. N Engl J Med 1999;340: Lacroix J, Hebert PC, Hutchison JS, et al. Transfusion strategies for patients in pediatric intensive care units. N Engl J Med 2007;356: Villanueva C, Colomo A, Bosch A, et al. Transfusion strategies for acute upper gastrointestinal bleeding. N Engl J Med 2013;368: Frank SM, Savage WJ, Rothschild JA, et al. Variability in blood and blood component utilization as assessed by an anesthesia information management system. Anesthesiology 2012;117: Grant MC, Rothschild JA, Frank SM. Predictors of postoperative hemoglobin drift. American Society of Anesthesiologists Annual Meeting, Washington, DC, Abstract A225; McDonald JH. Handbook of biological statistics. 2nd ed. Baltimore (MD): Sparky House Publishing; Garry PJ, Goodwin JS, Hunt WC. Iron status and anemia in the elderly: new findings and a review of previous studies. J Am Geriatr Soc 1983;31: Takasaki M, Tsurumi N, Konjiki O, et al. [Causes, diagnosis, and treatment of anemia in the elderly]. Nihon Ronen Igakkai Zasshi 1997;34: Walid MS, Abbara M, Tolaymat A, et al. The role of drains in lumbar spine fusion. World Neurosurg 2012;77: Thavendiranathan P, Bagai A, Ebidia A, et al. Do blood tests cause anemia in hospitalized patients? The effect of diagnostic phlebotomy on hemoglobin and hematocrit levels. J Gen Intern Med 2005;20: Alazia M, Colavolpe JC, Botti G, et al. [Blood loss from diagnostic laboratory tests performed in intensive care units. Preliminary study]. Ann Fr Anesth Reanim 1996;15: Peruzzi WT, Parker MA, Lichtenthal PR, et al. A clinical evaluation of a blood conservation device in medical intensive care unit patients. Crit Care Med 1993;21: Kruse A, Uehlinger DE, Gotch F, et al. Red blood cell lifespan, erythropoiesis and hemoglobin control. Contrib Nephrol 2008;161: Mukhopadhyay A, Yip HS, Prabhuswamy D, et al. The use of a blood conservation device to reduce red blood cell transfusion requirements: a before and after study. Crit Care 2010;14:R Bundgaard-Nielsen M, Secher NH, Kehlet H. Liberal vs. restrictive perioperative fluid therapy a critical assessment of the evidence. Acta Anaesthesiol Scand 2009;53: Scharte M, Fink MP. Red blood cell physiology in critical illness. Crit Care Med 2003;31:S Claessens YE, Fontenay M, Pene F, et al. Erythropoiesis abnormalities contribute to early-onset anemia in patients with septic shock. Am J Respir Crit Care Med 2006;174: Sihler KC, Raghavendran K, Westerman M, et al. Hepcidin in trauma: linking injury, inflammation, and anemia. J Trauma 2010;69: Nemeth E, Ganz T. Regulation of iron metabolism by hepcidin. Annu Rev Nutr 2006;26: Kickler TS, Smith B, Bell W, et al. Estimation of transfused red cell survival using an enzyme-linked antiglobulin test. Transfusion 1985;25: Luten M, Roerdinkholder-Stoelwinder B, Schaap NP, et al. Survival of red blood cells after transfusion: a comparison between red cells concentrates of different storage periods. Transfusion 2008;48: TRANSFUSION Volume 54, June 2014