- Email: [email protected]
The association of lowest hematocrit during cardiopulmonary bypass with acute renal injury after coronary artery bypass surgery
CARDIOVASCULAR
The Association of Lowest Hematocrit During Cardiopulmonary Bypass With Acute Renal Injury After Coronary Artery Bypass Surgery Madhav Swaminathan, MD, Barbara G. Phillips-Bute, PhD, Peter J. Conlon, MD, Peter K. Smith, MD, Mark F. Newman, MD, and Mark Stafford-Smith, FRCPC Departments of Anesthesiology, Medicine, and Surgery, Duke University Medical Center, Durham, North Carolina
Background. Acute renal injury is a common serious complication of cardiac surgery. Moderate hemodilution is thought to reduce the risk of kidney injury but the current practice of extreme hemodilution (target hematocrit 22% to 24%) during cardiopulmonary bypass (CPB) has been linked to adverse outcomes after cardiac surgery. Therefore we tested the hypothesis that lowest hematocrit during CPB is independently associated with acute renal injury after cardiac surgery. Methods. Demographic, perioperative, and laboratory data were gathered for 1,404 primary elective coronary bypass surgery patients. Preoperative and daily postoperative creatinine values were measured until hospital discharge per institutional protocol. Stepwise multivariable linear regression analysis was performed to determine whether lowest hematocrit during CPB was independently associated with peak fractional change in creatinine (defined as the difference between the preoperative and peak postoperative creatinine represented as
a percentage of the preoperative value). A p value of less than 0.05 was considered significant. Results. Multivariable analyses including preoperative hematocrit and other perioperative variables revealed that lowest hematocrit during CPB demonstrated a significant interaction with body weight and was highly associated with peak fractional change in serum creatinine (parameter estimate [PE] ⴝ 4.5; p ⴝ 0.008) and also with highest postoperative creatinine value (PE ⴝ 0.06; p ⴝ 0.004). Although other renal risk factors were significant covariates in both models, TM50 (an index of hypotension during CPB) was notably absent. Conclusions. These results add to concerns that current CPB management guidelines accepting extreme hemodilution may contribute to postoperative acute renal and other organ injury after cardiac surgery.
A
operative kidney function has not been similarly evaluated. The rationale for the renoprotective effect of hemodilution involves reduction of blood viscosity and improved regional blood flow in the setting of hypoperfusion and hypothermia [8, 9]. However several recent papers have reported independent associations of lowest hematocrit during CPB with morbidity and mortality after cardiac surgery [10 –12]. In this regard the cutoff for “safe” hemodilution has yet to be defined. Various “minimum acceptable” hematocrit values have been proposed based on animal experiments ranging from 9% to 18% [13, 14]. The relationship between current extreme hemodilution strategies and renal injury after cardiac surgery has not yet been investigated. Therefore we tested the hypothesis that lowest hematocrit during CPB is independently associated with acute renal injury after cardiac surgery.
cute renal injury is a ubiquitous finding after cardiac surgery with cardiopulmonary bypass (CPB). Numerous risk factors for perioperative renal dysfunction have been identified including age, preexisting renal disease, diabetes, hypertension, and low cardiac output states [1–5]. However intraoperative factors that are associated with renal dysfunction are less well understood. Although CPB is often necessary during cardiac surgery to optimize operating conditions it imposes several potential stresses on the kidney such as hemodilution, hypothermia, and nonpulsatile flow. In the setting of renal medullary hypoxia even minor disturbances in the relationship between oxygen supply and demand are believed to precipitate acute renal injury [6]. In a prospective study of the role of temperature management during CPB and the occurrence of renal injury after surgery [7] we could not demonstrate a major risk of normothermia versus hypothermia on renal function after surgery. However the relationship of current hemodilution management strategies during CPB to post-
(Ann Thorac Surg 2003;76:784 –92) © 2003 by The Society of Thoracic Surgeons
Patients and Methods Accepted for publication March 13, 2003. Address reprint requests to Dr Stafford-Smith, Department of Anesthesiology, Box 3094, Duke University Medical Center, Durham, NC 27710; e-mail: [email protected].
© 2003 by The Society of Thoracic Surgeons Published by Elsevier Inc
Population Selection After Institutional Review Board approval (Duke University Medical Center IRB registry #3048-01-8R0ER, ap0003-4975/03/$30.00 PII S0003-4975(03)00558-7
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
785
proved on August 3, 2001), demographic and outcome data were obtained from the Duke Cardiothoracic Surgery Database for all primary elective coronary artery bypass graft surgeries performed at Duke University Medical Center between February 1995 and February 1997. Patients with concomitant open chamber procedures and preoperative renal failure requiring dialysis were excluded from selection. Patients who required perioperative renal replacement therapy or who died within the first 2 postoperative days were also excluded as their serum creatinine values did not accurately reflect a perioperative renal insult. Perioperative variables were collected from the Duke Cardiothoracic Surgery Database, a prospectively entered quality assurance database that documents demographic data including the stage, severity history of ischemic heart disease, cardiac catheterization data, laboratory data, cardiac risk factors, and interventional cardiac therapy received within 48 hours of surgery. Intraoperative variables available in this dataset also include inotrope use, intraaortic balloon pump (IABP) therapy (preoperative need, intraoperative insertion, or postoperative use), duration of CPB and aortic cross clamp, highest prebypass serum glucose level, temperature management data, and number of coronary artery grafts performed. Postoperative data include transfusion of blood products, inotrope use, infectious postoperative complications, and hospital stay. In addition a Charlson comorbidity score was calculated for each patient using a list of preoperative demographic variables (ie, peripheral vascular disease, obstructive lung disease, peptic ulcer, diabetes, renal disease, liver disease, tumor/lymphoma/leukemia, metastatic cancer, dementia, and AIDS) [15].
surgery when compared other creatinine-derived markers of renal function [17], a consistent finding, regardless of preoperative renal status, that persists even in separate analysis of the group of patients who serum creatinine values never exceed the normal range. In our previous work we noted %⌬Cr after coronary bypass surgery to be remarkably similar among patient groups that differ by baseline renal function, creatinine values simply assuming proportionate but different absolute changes. CrmaxPost and lowest postoperative creatinine clearance were selected as secondary outcomes for analysis as they also highly correlate with postoperative morbidity and mortality but closely reflect impairment of renal filtration function [17].
Renal Function Assessment
Anesthesia and Surgery
Serum creatinine was measured as part of routine biochemical laboratory investigations for all elective coronary bypass surgery patients using a dry slide enzymatic reflectance technique (VITROS 950; Johnson and Johnson, New Brunswick, NJ) with a normal range of 0.7 to 1.4 mg/dL (62 to 124 mol/L).
Anesthesia was managed per the attending anesthesiologist’s preference. Use of agents with potential renal effects (eg, intravenous dopamine, diuretics, antifibrinolytic agents) were not controlled. Extracorporeal perfusion was performed using a Cobe CML Duo blood oxygenator with sealed hard-shell filtered venous reservoir (Cobe Laboratories, Lakewood, CO), a Cobe Century Perfusion System (Cobe Laboratories), and a 43-micron arterial line filter (Cobe Sentry arterial line filter with Primegard; Cobe Cardiovascular, Arvada, CO). Blood obtained by cardiotomy suction was routed by the roller pump into the integrated oxygenator-venous reservoir. Nonpulsatile perfusion was maintained at 2 to 2.4 L 䡠 min⫺1 䡠 m⫺2. The target CPB perfusion rate of 2 to 2.4 L 䡠 min⫺1 䡠 m⫺2 was maintained throughout the weight range. Other standard perfusion guidelines included titration of CPB perfusion rate to maintain a SvO2 of at least 60% during normothermic and 70% during hypothermic CPB. The bypass circuit was primed with mannitol (50 g of 20% solution) and crystalloid solution (0.9% normal saline). Typically, acceptable hematocrit ranged from 22% to 24% during bypass, with red blood cell transfusion usually occurring when values below 20% were observed. The arterial carbon dioxide tension was maintained
Definition of Variables The preoperative serum creatinine (CrPre) was identified as the value on the day before surgery for all inpatients, and was assessed within 1 week prior to surgery for all outpatients scheduled to undergo elective coronary artery bypass graft surgery. Peak postoperative creatinine (CrmaxPost) was defined as the highest of the daily in-hospital postoperative values. Peak percentage change in postoperative creatinine (%⌬Cr) was defined as the difference between the CrPre and CrmaxPost represented as a percentage of the preoperative value. Using the Cockroft Gault equation [16] preoperative creatinine clearance (CrClPre) was estimated from CrPre and lowest postoperative creatinine clearance (CrClPost) was estimated from CrmaxPost. The primary outcome variable, %⌬Cr, was selected as the marker that demonstrates the best association with mortality and major morbidity after coronary bypass
Hematocrit, Hemodynamic, and Transfusion Data Hematocrit was measured every 15 minutes during CPB as part of routine arterial blood gas analysis protocols using the AVL Omni Modular System blood gas analyzer (Roche Diagnostics, Indianapolis, IN). Hemodynamic parameters were available for each minute during CPB from the ARKIVE Information Management System (Arkive IMS, San Diego, CA). The TM50 is a previously reported, derived index of the degree and duration of low CPB perfusion pressure [18, 19]. The TM50 was defined as the time-pressure integral mean CPB blood pressure less than 50 mm Hg (min 䡠 mm Hg) or as the area under the CPB blood pressure curve below 50 mm Hg. Transfusion was defined in two variables: (1) RBC 48 hours: the number of units of packed red cells given within the first 48 hours postoperatively; and (2) transfusion: transfusion of more than 2 units of packed red cells and at least one other blood product.
CARDIOVASCULAR
Ann Thorac Surg 2003;76:784 –92
CARDIOVASCULAR
786
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
throughout bypass at 35 to 40 mm Hg (uncorrected for temperature) with the arterial oxygen tension maintained at 150 to 250 mm Hg. Target mean arterial pressure was between 50 and 70 mm Hg during bypass using intravenous phenylephrine or sodium nitroprusside as required. Typically patients were cooled to a nasopharyngeal temperature of 34°C to 28°C during bypass and rewarmed to a nasopharyngeal temperature of 37°C or a bladder temperature of at least 36°C prior to separation from bypass.
Statistical Analysis Simple descriptive statistics (mean, standard deviation) were used to describe the population demographics. We performed a primary stepwise multivariable linear regression analysis to test the association of lowest hematocrit during bypass with %⌬Cr. To assess robustness of the results and to increase generalizability, secondary multivariable regression analyses were also performed to test the association between lowest hematocrit during bypass and CrmaxPost and CrClPost. First we performed a full-model analysis in which we included all the demographic, comorbidity, and intraoperative data. To follow up this analysis we implemented a stepwise selection so that the final model included only variables that were significantly associated with (p ⬍ 0.10) the outcome (%⌬Cr). Of note the secondary analysis using CrmaxPost, while controlling for preoperative creatinine, is closely equivalent (and statistically identical) to assessing pre to peak change (ie, delta) in creatinine. All statistical analyses were performed using the SAS Statistical software (software version 8.0; SAS Institute, Cary, NC). A p value of less than 0.05 was considered significant.
Results A total of 1,404 patients undergoing primary elective CABG surgery met the criteria for analysis. Their demographic profile (Table 1) was similar to that of other previously reported populations [4]. Hematocrit, hemodynamic, and creatinine data are shown in Table 2. Among the excluded patients, 6 died within 48 hours of surgery, 7 were receiving preoperative dialysis, and 5 patients had acute renal failure requiring renal replacement therapy within 10 days of surgery. In the primary multivariable linear regression analysis of variables associated with %⌬Cr (Table 3), we found a significant two-way interaction between body weight and lowest hematocrit during bypass (parameter estimate [PE] ⫽ ⫺0.05; p ⫽ 0.007). With the inclusion of the hematocrit-body weight interaction in this model, there was a highly significant association between lowest hematocrit during bypass and %⌬Cr (PE ⫽ 4.5; p ⫽ 0.008). Figure 1 shows the influence of body weight on the association between lowest hematocrit during bypass and %⌬Cr. Other variables known to be renal risk factors including preoperative creatinine (CrPre), body weight, Charlson comorbidity score, prebypass serum glucose, perioperative blood transfusion, preoperative hematocrit, and postoperative IABP and inotrope use were also
Ann Thorac Surg 2003;76:784 –92
Table 1. Demographics Variable
Mean (SD)
Age (years) Female sex (%) African-American ethnicity (%) Weight (kg) Hypertension (%) Cerebrovascular accident (%) Diabetic (%) Previous myocardial infarction (%) Ejection fraction (%) Unstable angina (%) Carotid bruit (%) Preoperative congestive heart failurea (%) Cardiopulmonary bypass duration (min) Cross-clamp time (min) Preoperative intraaortic balloon pump (%) Postoperative intraaortic balloon pump (%) Preoperative inotrope (%)b Postoperative inotrope (%) Red blood cells 48 hours (units)c Transfusion (%)d Charlson comorbidity score
63.3 (11.0) 32.7 11.9 82.1 (16.6) 69.2 1.6 31.6 27.8 53.5 (14.0) 49.1 11.6 15.4 103.2 (39.9) 49.0 (18.2) 3.7 1.0 1.2 25.7 0.9 (1.6) 7.0 0.81 (1.02) (range 0 – 8) 132 (48)
Prebypass serum glucose (mg/dL) a
Congestive heart failure is defined as history of paroxysmal or nocturnal b dyspnea responsive to afterload reduction drug therapy. Inotrope use defined as postoperative infusion of either dopamine ⬎ 5 g 䡠 kg⫺1 䡠 ⫺1 ⫺1 ⫺1 min or dobutamine ⬎ 5 g 䡠 kg 䡠 min , or both, or epinephrine c ⬎ 0.03 g 䡠 kg⫺1 䡠 min⫺1. Number of units of packed red cells given d within the first 48 hours postoperatively. Transfusion of more than 2 units of packed red cells and at least one other blood product.
significantly associated with %⌬Cr in this multivariable model (Table 3). It is noteworthy that the association between lowest hematocrit during bypass and %⌬Cr was significant even after accounting for preoperative hematocrit. Figure 2 demonstrates the relationship of lowest hematocrit during bypass to %⌬Cr in patients weighing
Table 2. Hematocrit, Hemodynamic, and Creatinine Data Variable
Mean (SD)
Range
Preoperative Hct (%) Lowest Hct during bypass (%) TM50 (mm Hg 䡠 min) CrPre (mg/dL) CrmaxPost (mg/dL) %⌬Cr (%) CrClPre (mL/min) CrClPost (mL/min)
40.1 (9.1) 19.5 (3.8) 135.9 (139.7) 1.1 (0.5) 1.41 (0.7) 26.6 (45.0) 80.9 (33.4) 67.2 (28.8)
22–52 10 –33 ⫺1042– 0 0.5– 4.9 0.5– 8.2 ⫺74.4 –722.2 10.5–254.7 10.2–212.3
CrClPost ⫽ lowest postoperative creatinine clearance; CrClPre ⫽ preoperative creatinine clearance; CrmaxPost ⫽ peak postoperative creatinine; CrPre ⫽ preoperative creatinine; Hct ⫽ hematocrit; %⌬Cr ⫽ peak postoperative fractional change in creatinine; TM50 ⫽ integral of time and mean arterial pressure less than 50 mm Hg during cardiopulmonary bypass.
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
787
Table 3. Multivariable Linear Regression Analysis of Factors Associated With Peak Postoperative Change in Creatinine Variable Intercept Lowest Hct during bypass (%) CrPre (mg/dL) Charlson comorbidity score Body weight (kg) Lowest Hct during bypass*weight Preoperative Hct (%) IABP use RBC 48 hours (units) Inotrope usea Prebypass serum glucose (mg/dL) TM50 (mm Hg 䡠 min)
Parameter Estimate ⫺59.0 4.5 ⫺15.0 3.5 1.3 ⫺0.05 ⫺0.6 31.3 2.3 7.8 0.07 ⫺0.00
95% Confidence Limits ⫺126.0 1.2 ⫺20.7 0.9 0.5 ⫺0.1 ⫺1.0 4.6 0.6 1.7 0.02 ⫺0.02
p Value
8.1 7.8 ⫺9.3 6.2 2.1 ⫺0.01 ⫺0.2 57.9 4.0 13.9 0.1 0.01
0.08 0.008 ⬍0.0001 0.009 0.001 0.007 0.008 0.02 0.009 0.01 0.01 0.64
a Inotrope use defined as postoperative infusion of either dopamine ⬎ 5 g 䡠 kg⫺1 䡠 min⫺1 or dobutamine ⬎ 5 g 䡠 kg⫺1 䡠 min⫺1, or both, or epinephrine ⬎ 0.03 g 䡠 kg⫺1 䡠 min⫺1.
CrmaxPost ⫽ peak postoperative serum creatinine; CrPre ⫽ preoperative serum creatinine; Hct ⫽ hematocrit; pump; RBC 48 hours ⫽ number of units of packed red cells given within the first 48 hours postoperatively.
75 kg or more, with 95% confidence limits. To address the question of whether there is an “elbow” in the data where the slope of the relationship between lowest hematocrit during bypass and %⌬Cr changes significantly, we used a restricted cubic splines analysis of our regression model. We found no significant cutoff where the association between lowest hematocrit during bypass and %⌬Cr diverged from a linear relationship. The secondary multivariable linear regression model analyzing the variables associated with CrmaxPost con-
Fig 1. This figure is based on a predictive model for “typical” patients relating lowest hematocrit (Hct) during bypass to peak percentage change in creatinine (%⌬Cr), where the following covariates were adjusted for: prebypass glucose ⫽ 130 mg/dL, Charlson comorbodity score ⫽ 1, preoperative hematocrit ⫽ 40%, preoperative creatinine ⫽ 1.5 mg/dL, and no intraaortic balloon pump, inotropes, transfusion, or diabetes. The regression equations are %⌬Cr ⫽ for 75 kg (triangles), 3.499 ⫹ (lowest Hct during bypass*0.604); for 100 kg (asterisks), 38.16 ⫹ (lowest Hct during bypass*⫺0.791); for 125 kg (circles), 72.82 ⫹ (lowest Hct during bypass*⫺2.186); for 150 kg (squares), 107.48 ⫹ (lowest Hct during bypass*⫺3.581).
IABP ⫽ intraaortic balloon
firmed the findings of the primary analysis. Figure 3 shows the influence of body weight on the association between lowest hematocrit during bypass and CrmaxPost. There was a significant two-way interaction between lowest hematocrit during bypass and body weight in this model as well (PE ⫽ ⫺0.0007; p ⫽ 0.003). After inclusion of this interaction in the model there was an independent association between lowest hematocrit during bypass and CrmaxPost (PE ⫽ 0.06; p ⫽ 0.004; Table 4). Other known renal risk factors were also similarly noted in this analysis. Similar results were also seen in a multivariable analysis of factors associated with lowest estimated postoperative creatinine clearance (by the Cockroft-Gault equation), with a significant interaction between lowest hematocrit and body weight. The potential for interdependent variables exists in our
Fig 2. The relationship of lowest hematocrit (Hct) during bypass to peak percentage change in creatinine (%⌬Cr) in patients (n ⫽ 776) with weight 75 kilos or more, with 95% confidence limits (dashed lines). The regression equation is %⌬Cr ⫽ 50.62 ⫹ (lowest Hct during bypass*⫺1.10).
CARDIOVASCULAR
Ann Thorac Surg 2003;76:784 –92
CARDIOVASCULAR
788
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
Ann Thorac Surg 2003;76:784 –92
Fig 3. This figure is based on a predictive model for “typical” patients relating lowest hematocrit (Hct) during bypass to (CrmaxPost), where the following covariates were adjusted for: prebypass glucose ⫽ 130 mg/dL, Charlson comorbodity score ⫽ 1, preoperative hematocrit ⫽ 40%, preoperative creatinine ⫽ 1.5 mg/dL, and no intraaortic balloon pump, inotropes, transfusion, or diabetes. The regression equations are CrmaxPost ⫽ for 75 kg (triangles), 1.54 ⫹ (lowest Hct during bypass*0.007); for 100 kg (asterisks), 2.0 ⫹ (lowest Hct during bypass*⫺0.006); for 125 kg (circles), 2.46 ⫹ (lowest Hct during bypass*⫺0.03); for 150 kg (squares), 2.91 ⫹ (lowest Hct during bypass*⫺0.05).
analysis and may complicate interpretation of our results. Therefore, we performed further analyses to address this issue. It is possible that low hematocrit during bypass may be a marker for low hematocrit early after bypass, an effect that might influence the need for drugs that increase systemic vascular resistance, such as inotropes. We confirmed a relationship between low hematocrit during bypass and postoperative inotropes (a logistic regression indicates that low hematocrit during bypass predicts inotrope use, p ⫽ 0.05, c index ⫽ 0.53). We also found that inotrope use is a significant predictor of %⌬Cr (p ⫽ 0.0008, beta coefficient ⫽ 9.8). Because the predictors
are related to each other and also to the outcome, this raises the issue of confounding that cannot be completely addressed by our statistical models. However because the inclusion of inotropes in the multivariable model (see Table 3) does not alter the significance of our predictors of interest (weight, lowest hematocrit during bypass, and the interaction term) we conclude that these variables contribute an independent predictive ability beyond the role of postoperative inotropes. A second concern in interpretation of our data relates to the issue of transfusion and lowest hematocrit during bypass because it is difficult to separate the effects of transfusion from the need for transfusion as they would always be expected to occur together—that is, everyone who needs a transfusion gets a transfusion. We found an association between lower hematocrits during bypass and increased transfusion in a univariate regression model (p ⬍ 0.0001, beta coefficient ⫽ ⫺0.07). However because the inclusion of red blood cells transfusion in the model does not alter the significance of our predictors we again conclude that these variables contribute independent predictive ability. Because this relationship raises the issue of interdependent variables we performed tolerance and variance inflation diagnostics on our models designed to detect the presence of multicollinearity. When lowest hematocrit during bypass and transfusion are in a model together, the tolerance for the variables is 0.78 and 0.92 respectively (corresponding to variance inflation values of 1.3 and 1.1). Because the tolerance values are greater than 0.4 we conclude that the model is appropriate statistically. Using correlated variables as predictors in a model requires careful consideration and careful interpretation our diagnostics indicate that our models do not violate accepted guidelines for multicollinearity. Because hematocrit is directly influenced by concurrent blood transfusion we performed a subsequent analysis of the relationship between perioperative transfu-
Table 4. Multivariable Linear Regression Analysis of Factors Associated With Postoperative Peak Creatinine Variable Intercept Lowest Hct during bypass (%) CrPre (mg/dL) Charlson comorbidity score Body weight (kg) Lowest Hct during bypass*weight Preoperative Hct (%) IABP use RBC 48 hours (units) Inotrope usea Prebypass serum glucose (mg/dL) TM50 (mm Hg min)
Parameter Estimate ⫺0.9 0.06 ⫺0.2 0.06 0.02 ⫺0.0007 ⫺0.008 0.5 0.03 0.08 0.001 ⫺0.00003
95% Confidence Limits ⫺1.7 0.02 ⫺0.3 0.02 0.007 ⫺0.001 ⫺0.01 0.2 0.006 0.002 0.0002 ⫺0.0003
⫺0.008 0.1 ⫺0.1 0.09 0.03 ⫺0.0003 ⫺0.002 0.8 0.05 0.2 0.002 0.0002
p Value 0.05 0.004 ⬍0.0001 0.0007 0.0006 0.003 0.006 0.005 0.01 0.04 0.008 0.77
a Inotrope use defined as postoperative infusion of either dopamine ⬎ 5 g 䡠 kg⫺1 䡠 min⫺1 or dobutamine ⬎ 5 g 䡠 kg⫺1 䡠 min⫺1, or both, or epinephrine ⬎ 0.03 g 䡠 kg⫺1 䡠 min⫺1 or isoprenaline ⬎ 0.03 g 䡠 kg⫺1 䡠 min⫺1, or both.
CrPre ⫽ preoperative serum creatinine; Hct ⫽ hematocrit; IABP ⫽ intraaortic balloon pump; RBC 48 hours ⫽ number of units of packed red cells given within the first 48 hours postoperatively; TM50 ⫽ integral of time and mean arterial pressure less than 50 mm Hg during cardiopulmonary bypass.
Fig 4. The influence of blood transfusion on the relationship between lowest hematocrit (Hct) during cardiopulmonary bypass and peak postoperative fractional change in creatinine (%⌬Cr). RBCs: 9 (asterisks); 6 (triangles); 3 (circles); 0 (squares). (RBCs ⫽ number of units of packed red blood cells transfused intraoperatively.)
sion, lowest hematocrit during bypass, and %⌬Cr (see Fig 4). For any given lowest hematocrit during bypass we found a direct relationship between increased transfusion and greater %⌬Cr.
Comment We found a significant association between lowest hematocrit during bypass and creatinine rise after coronary bypass surgery that is influenced by body weight. The degree to which lowest hematocrit during bypass is adversely related to postoperative creatinine rise is proportional to increasing body weight (Fig 1). The following example illustrates this finding. In a 75-kg patient there is no association between lowest hematocrit during bypass and postoperative creatinine rise. However in a 150-kg patient there is a highly significant inverse association between lower hematocrit during bypass and greater postoperative creatinine rise (Fig 1). Our findings are consistent in multivariable analyses of the factors associated with both %⌬Cr and CrmaxPost. We also confirmed associations of previously recognized renal risk factors [3, 4] including elevated preoperative creatinine, Charlson comorbidity index, intraoperative highest prebypass serum glucose, postoperative inotrope and IABP use, preoperative hematocrit, blood transfusion and body weight with both, %⌬Cr and CrmaxPost. The absence of low bypass perfusion pressure (TM50) as a significant factor in both multivariable analyses is notable and confirms previous reports of lack of evidence of perfusion pressure as a predictor of renal dysfunction [20 –22]. The study design, large number of patients, and the use of valid markers of renal injury are the strengths of this study. The significant association of lowest hematocrit during bypass with both creatinine markers is consistent with an important contributory role of hemodilution in the pathophysiology of acute perioperative renal injury after cardiac surgery. The effects of hemodilution during bypass on various
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
789
organ systems are well described [14, 23, 24]. The range of lowest hematocrit during bypass in our study was almost identical to that observed in the two major previous publications relating hemodilution during bypass with postoperative complications involving 9,700 patients (10% to 33% versus 10% to 36% and 10% to ⬎35%) [10, 11]. In addition our mean lowest hematocrit during bypass (19.5%) was similar to those quoted in these studies, including one center reporting 2,738 patients with a mean lowest hematocrit of 18.8% [11]. Hemodilution reduces the need for addition of blood to the bypass circuit, both avoiding transfusion and reducing viscosity during hypothermic bypass. As acid-base abnormalities do not develop routinely during bypass, the resultant increase in regional blood flow during bypass has been thought to compensate for the reduced oxygen carrying capacity of the perfusate compared with prebypass levels. However the breakdown of the balance between oxygen supply and demand in this setting has not otherwise been extensively studied. Although the association of hemodilution or low hematocrit with adverse outcomes after cardiac surgery has been reported [10 –12] the degree of hemodilution that can be tolerated before an adverse effect occurs is not yet known. DeFoe and colleagues [10] reported an unfavorable association between lowest hematocrit during bypass and adverse in-hospital outcomes after cardiac surgery in a large group of patients in an observational study. However this study did not comment on adverse renal outcomes and did not investigate perioperative renal dysfunction as a primary outcome. In another retrospective analysis, Fang and coworkers [11] reported that lowest hematocrit during bypass was independently associated with mortality after cardiac surgery. This study also recognized preoperative renal failure as one of the risk factors for mortality in addition to lowest hematocrit but did not identify perioperative renal dysfunction as an adverse outcome related to lowest hematocrit during bypass. Ranucci and associates [12] found a hematocrit of less than 25% during bypass, low output syndrome, and homologous blood transfusions to be predictors of severe renal dysfunction in 316 patients undergoing cardiac surgery. However the results from this study were based on a regression analysis looking at three outcome variables in a subgroup of only 9 patients with severe renal dysfunction. These results also did not correct for preoperative creatinine, which is known to influence postoperative creatinine rise [3]. Our study investigates the independent association of lowest hematocrit during bypass with acute renal injury as a primary outcome in coronary bypass surgery patients. Our report has limitations related to study design and measurement of renal function. Although this study is retrospective, the data were taken from a prospectively gathered quality-assurance database. A second concern is that our study evaluated only one aspect of renal function, namely filtration, and not other renal homeostatic roles including osmolality, electrolyte and acidbase regulation and production, and release of enzymes and hormones. Although more sensitive markers of sub-
CARDIOVASCULAR
Ann Thorac Surg 2003;76:784 –92
CARDIOVASCULAR
790
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
tle renal injury than creatinine such as urinary N-acetyl -D-glucosaminidase [25] and cystatin C [26] have been reported, the link between these markers and adverse postoperative complications has not been established [27]. Moreover creatinine-based markers have consistently proved to be both sensitive indicators of renal injury and robust prognostic indicators of adverse outcome in patients with cardiovascular disease [28, 29] including those undergoing cardiac surgery [30]. We found a significant association between blood transfusion and acute renal injury. A potential concern is that extreme hemodilution may simply be a marker for increased likelihood of blood transfusion. Supporting this interpretation is the interaction we observed between hemodilution, body weight, and renal injury, with greater transfusion requirements to achieve a “target hematocrit” in an overweight patient. However in our multivariable analyses accounting for blood transfusion we still found lowest hematocrit during bypass to be significantly associated with postoperative creatinine rise (Tables 3, 4; Figs 1–3). Despite these limitations we feel that the use of valid markers of renal injury combined with data analysis in a large group of patients strengthen the significance of our findings. Although cause and effect are not demonstrated by our findings of an association of severe hemodilution during bypass and acute renal injury, it is important to explore possible physiologic foundations for this relationship. Because hematocrit is directly related to tissue oxygen delivery a simple interpretation of our findings is that severe hemodilution may contribute to renal injury by augmenting the local renal inflammatory response as a result of ischemia-reperfusion injury, particularly in the hypoxic environment of the renal medulla. However hemodilution has previously been thought to be renoprotective [31]. In an animal model of ischemic acute renal injury, Olof and colleagues [32] reported that trapping of red blood cells in the medullary vasculature and associated reductions in outer medullary blood flow observed with normal and elevated hematocrits are completely avoided with hemodilution (30%). Although other authors confirmed the avoidance “red cell trapping” with hemodilution, they could not confirm a benefit in terms of reduced renal injury [33, 34]. In a randomized study of 300 patients we did not find a difference between warm and cold bypass, with regard to renal injury, [7], suggesting a pathophysiology independent of oxygen demand. Other factors than ischemiareperfusion injury such as atheroembolism and systemic inflammatory response have been implicated as important contributors to post– cardiac surgery nephropathy. Koh and colleagues [35] speculated that alterations in the pattern of aortic blood flow related to cardiopulmonary bypass, potentially including the effects of hemodilution, may increase emboli delivery to the kidney and other organs. The relationship of hemodilution and inflammation during bypass is not known but it is possible that gut ischemia related to reduced tissue oxygen delivery may result in greater release of endotoxin and inflammationmediated renal injury [36]. Visceral fat is a key regulator
Ann Thorac Surg 2003;76:784 –92
site for the process of inflammation and obesity is associated with an upregulated inflammatory state [37]. The interaction between lowest hematocrit during bypass and increased weight that we observed in our study may be explained by the combining of effects that increase inflammatory stimuli and inflammatory response respectively. The significant association between lowest hematocrit during bypass and acute renal injury raises concerns regarding currently accepted bypass management strategies. Current guidelines for accepted levels of hemodilution during bypass do not lead to the development of acid-base disturbances. However occult cellular hypoxemia and injury may still occur. Such injury in the renal medulla, a tissue that is known to have a precarious balance between oxygen supply and demand even under normal conditions [38], may be the basis of the findings of our study. Strategies to avoid extreme hemodilution may include transfusion. However since transfusion of red cells is also independently associated with creatinine rise in our study (see Tables 3, 4; Fig 4) guidelines for optimal hematocrit during bypass that include transfusion cannot be simply defined. The decision to transfuse is usually based on clinical judgment, institutional protocols, and hematocrit “triggers.” Because transfusion has not been tested as a primary outcome of this study the authors cannot recommend changes in transfusion policy based on its effect on perioperative creatinine changes. However clearly minimizing bypass prime volumes should reduce anemia without transfusion and potentially reduce renal risk. Our data suggest that hemodilution during bypass should be carefully monitored and excessive falls in hematocrit avoided, especially in overweight patients (see Fig 2). We did not observe an “elbow” in the graph relating lowest hematocrit to %⌬Cr that would suggest a level below which hematocrit is detrimental to renal function. In summary we found an independent association between lowest hematocrit during bypass and postoperative creatinine rise that is influenced by body weight. This association was significant for all creatinine markers of renal injury that we assessed (%⌬Cr, CrmaxPost, and CrClPost). We did not observe an effect of low bypass perfusion pressure on creatinine rise. This is the first report highlighting the association of extreme hemodilution during bypass with acute renal injury as a primary outcome. Despite being a retrospective data analysis looking at one aspect of renal function we believe that the study design and the use of valid markers of renal injury justify our conclusions. Our findings question the wisdom of tolerating low levels of hematocrit during bypass. We have previously reported the lack of effect of bypass temperature on acute renal injury [7]. These findings suggest that optimal bypass management guidelines should emphasize the maintenance of adequate hematocrit levels during bypass regardless of temperature protocol employed. However the independent contributions of hemodilution and transfusion to the pathophys-
iology of perioperative renal injury require further examination in a prospective randomized trial looking at a variety of tests of renal function. This work was supported by the Cardiothoracic Division of the Department of Anesthesiology, Duke University Medical Center, Durham, North Carolina. The authors would like to gratefully acknowledge the contribution of Christopher Keith, and Cheryl Stetson, Department of Anesthesiology, Duke University Medical Center, in the preparation of this manuscript.
References 1. Ostermann ME, Taube D, Morgan CJ, Evans TW. Acute renal failure following cardiopulmonary bypass: a changing picture. Intens Care Med 2000;26:565–71. 2. Chertow GM, Lazarus JM, Christiansen CL, et al. Preoperative renal risk stratification. Circulation 1997;95:878 –84. 3. Conlon PJ, Stafford-Smith M, White WD, et al. Acute renal failure following cardiac surgery. Nephrol Dial Transplant 1999;14:1158 –62. 4. Mangano CM, Diamondstone LS, Ramsay JG, Aggarwal A, Herskowitz A, Mangano DT. Renal dysfunction after myocardial revascularization: risk factors, adverse outcomes, and hospital resource utilization. The Multicenter Study of Perioperative Ischemia Research Group. Ann Intern Med 1998; 128:194 –203. 5. Mangos GJ, Brown MA, Chan WY, Horton D, Trew P, Whitworth JA. Acute renal failure following cardiac surgery: incidence, outcomes and risk factors. Aust NZ J Med 1995; 25:284 –9. 6. Flemming B, Seeliger E, Wronski T, Steer K, Arenz N, Persson PB. Oxygen and renal hemodynamics in the conscious rat. J Am Soc Nephrol 2000;11:18 –24. 7. Swaminathan M, East C, Phillips-Bute B, et al. Report of a substudy on warm versus cold cardiopulmonary bypass: changes in creatinine clearance. Ann Thorac Surg 2001;72: 1603–9. 8. Messmer K. Hemodilution. Surg Clin North Am 1975;55: 659 –78. 9. Shah D, Corson J, Karmody A, Leather R. Effects of isovolemic hemodilution on abdominal aortic aneurysmectomy in high risk patients. Ann Vasc Surg 1986;1:50 –4. 10. DeFoe G, Ross C, Olmstead E, et al, for Group NNECDS. Lowest hematocrit on bypass and adverse outcomes associated with coronary artery bypass grafting. Ann Thorac Surg 2001;71:769 –76. 11. Fang WC, Helm RE, Krieger KH, et al. Impact of minimum hematocrit during cardiopulmonary bypass on mortality in patients undergoing coronary artery surgery. Circulation 1997;96(Suppl 2):194 –9. 12. Ranucci M, Pavesi M, Mazza E, et al. Risk factors for renal dysfunction after coronary surgery: the role of cardiopulmonary bypass technique. Perfusion 1994;9:319 –26. 13. Cook D, Orszulak TA, Daly R. Minimum hematocrit at differing cardiopulmonary bypass temperatures in dogs. Circulation 1998;98(Suppl):II170 –4; discussion II175. 14. Liam BL, Plochl W, Cook DJ, Orszulak TA, Daly RC. Hemodilution and whole body oxygen balance during normothermic cardiopulmonary bypass in dogs. J Thorac Cardiovasc Surg 1998;115:1203–8. 15. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis 1987;40: 373–83. 16. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron 1976;16:31–41. 17. Stafford-Smith M, Reddan D, Phillips-Bute B, Newman M, Winn M. Association of perioperative creatinine-derived variables with mortality and other outcomes after coronary bypass surgery. Anesth Analg 2001;92:SCA28.
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
791
18. Gold JP, Charlson ME, Williams-Russo P, et al. Improvement of outcomes after coronary artery bypass. A randomized trial comparing intraoperative high versus low mean arterial pressure. J Thorac Cardiovasc Surg 1995;110:1302–11. 19. Hill SE, van Wermeskerken GK, Lardenoye JW, et al. Intraoperative physiologic variables and outcome in cardiac surgery: part I. In-hospital mortality. Ann Thorac Surg 2000;69: 1070 –5. 20. Urzua J, Troncoso S, Bugedo G, et al. Renal function and cardiopulmonary bypass: effect of perfusion pressure. J Cardiothorac Vasc Anesth 1992;6:299 –303. 21. Bhat JG, Gluck MC, Lowenstein J, Baldwin DS. Renal failure after open heart surgery. Ann Intern Med 1976;84:677–82. 22. Abel RM, Wick J, Beck CH Jr, Buckley MJ, Austen WG. Renal dysfunction following open-heart operations. Arch Surg 1974;108:175–7. 23. Hall TS. The pathophysiology of cardiopulmonary bypass. The risks and benefits of hemodilution. Chest 1995;107:1125– 33. 24. Sungurtekin H, Cook DJ, Orszulak TA, Daly RC, Mullany CJ. Cerebral response to hemodilution during hypothermic cardiopulmonary bypass in adults. Anesth Analg 1999;89:1078 – 83. 25. Hsu WS, Kao JT, Chen JS. Clinical significance of urinary N-acetyl-beta-D-glucosaminidase and alanine aminopeptidase. Taiwan Yi Xue Hui Za Zhi 1989;88:407–9. 26. Fliser D, Ritz E. Serum cystatin C concentration as a marker of renal dysfunction in the elderly. Am J Kidney Dis 2001; 37:79 –83. 27. Hamada Y, Kanda T, Anzai T, Kobayashi I, Morishita Y. N-acetyl-beta-D-glucosaminidase is not a predictor, but an indicator of kidney injury in patients with cardiac surgery. J Med 1999;30:329 –36. 28. Friedman PJ. Serum creatinine: an independent predictor of survival after stroke. J Intern Med 1991;229:175–9. 29. Shulman NB, Ford CE, Hall WD, et al. Prognostic value of serum creatinine and effect of treatment of hypertension on renal function. Results from the hypertension detection and follow-up program. The Hypertension Detection and Follow-up Program Cooperative Group. Hypertension 1989; 13(Suppl):I80–93. 30. Stafford-Smith M, Phillips-Bute B, Reddan D, Milano C, Newman M, Winn M. The association of postoperative peak and fractional change in serum creatinine with mortality after coronary bypass surgery. Anesthesiology 2000; 93(Suppl):A240. 31. Rajagopalan PR, Reines HD, Pulliam C, Fitts CT, LeVeen HH. Reversal of acute renal failure using hemodilution with hydroxyethyl starch. J Trauma 1983;23:795–800. 32. Olof P, Hellberg A, Kallskog O, Wolgast M. Red cell trapping and postischemic renal blood flow. Differences between the cortex, outer and inner medulla. Kidney Int 1991;40:625–31. 33. Hellberg PO, Kallskog O, Wolgast M. Nephron function in the early phase of ischemic renal failure. Significance of erythrocyte trapping. Kidney Int 1990;38:432–9. 34. Andersson G, Jennische E. Lack of casual relationship between medullary blood congestion and tubular necrosis in postischaemic kidney damage. Acta Physiol Scand 1987;130:429–32. 35. Koh TW, Parker KH, Kon M, Pepper JR. Changes in aortic rotational flow during cardiopulmonary bypass studied by transesophageal echocardiography and magnetic resonance velocity imaging: a potential mechanism for atheroembolism during cardiopulmonary bypass. Heart Vessels 2001;16:1–8. 36. Gow KW, Phang PT, Tebbutt-Speirs SM, et al. Effect of crystalloid administration on oxygen extraction in endotoxemic pigs. J Appl Physiol 1998;85:1667–75. 37. Ziccardi P, Nappo F, Giugliano G, et al. Reduction of inflammatory cytokine concentrations and improvement of endothelial functions in obese women after weight loss over one year. Circulation 2002;105:804 –9. 38. Brezis M, Epstein FH. Cellular mechanisms of acute ischemic injury in the kidney. Annu Rev Med 1993;44:27–37.
CARDIOVASCULAR
Ann Thorac Surg 2003;76:784 –92
The Association of Lowest Hematocrit During Cardiopulmonary Bypass With Acute Renal Injury After Coronary Artery Bypass Surgery Madhav Swaminathan, MD, Barbara G. Phillips-Bute, PhD, Peter J. Conlon, MD, Peter K. Smith, MD, Mark F. Newman, MD, and Mark Stafford-Smith, FRCPC Departments of Anesthesiology, Medicine, and Surgery, Duke University Medical Center, Durham, North Carolina
Background. Acute renal injury is a common serious complication of cardiac surgery. Moderate hemodilution is thought to reduce the risk of kidney injury but the current practice of extreme hemodilution (target hematocrit 22% to 24%) during cardiopulmonary bypass (CPB) has been linked to adverse outcomes after cardiac surgery. Therefore we tested the hypothesis that lowest hematocrit during CPB is independently associated with acute renal injury after cardiac surgery. Methods. Demographic, perioperative, and laboratory data were gathered for 1,404 primary elective coronary bypass surgery patients. Preoperative and daily postoperative creatinine values were measured until hospital discharge per institutional protocol. Stepwise multivariable linear regression analysis was performed to determine whether lowest hematocrit during CPB was independently associated with peak fractional change in creatinine (defined as the difference between the preoperative and peak postoperative creatinine represented as
a percentage of the preoperative value). A p value of less than 0.05 was considered significant. Results. Multivariable analyses including preoperative hematocrit and other perioperative variables revealed that lowest hematocrit during CPB demonstrated a significant interaction with body weight and was highly associated with peak fractional change in serum creatinine (parameter estimate [PE] ⴝ 4.5; p ⴝ 0.008) and also with highest postoperative creatinine value (PE ⴝ 0.06; p ⴝ 0.004). Although other renal risk factors were significant covariates in both models, TM50 (an index of hypotension during CPB) was notably absent. Conclusions. These results add to concerns that current CPB management guidelines accepting extreme hemodilution may contribute to postoperative acute renal and other organ injury after cardiac surgery.
A
operative kidney function has not been similarly evaluated. The rationale for the renoprotective effect of hemodilution involves reduction of blood viscosity and improved regional blood flow in the setting of hypoperfusion and hypothermia [8, 9]. However several recent papers have reported independent associations of lowest hematocrit during CPB with morbidity and mortality after cardiac surgery [10 –12]. In this regard the cutoff for “safe” hemodilution has yet to be defined. Various “minimum acceptable” hematocrit values have been proposed based on animal experiments ranging from 9% to 18% [13, 14]. The relationship between current extreme hemodilution strategies and renal injury after cardiac surgery has not yet been investigated. Therefore we tested the hypothesis that lowest hematocrit during CPB is independently associated with acute renal injury after cardiac surgery.
cute renal injury is a ubiquitous finding after cardiac surgery with cardiopulmonary bypass (CPB). Numerous risk factors for perioperative renal dysfunction have been identified including age, preexisting renal disease, diabetes, hypertension, and low cardiac output states [1–5]. However intraoperative factors that are associated with renal dysfunction are less well understood. Although CPB is often necessary during cardiac surgery to optimize operating conditions it imposes several potential stresses on the kidney such as hemodilution, hypothermia, and nonpulsatile flow. In the setting of renal medullary hypoxia even minor disturbances in the relationship between oxygen supply and demand are believed to precipitate acute renal injury [6]. In a prospective study of the role of temperature management during CPB and the occurrence of renal injury after surgery [7] we could not demonstrate a major risk of normothermia versus hypothermia on renal function after surgery. However the relationship of current hemodilution management strategies during CPB to post-
(Ann Thorac Surg 2003;76:784 –92) © 2003 by The Society of Thoracic Surgeons
Patients and Methods Accepted for publication March 13, 2003. Address reprint requests to Dr Stafford-Smith, Department of Anesthesiology, Box 3094, Duke University Medical Center, Durham, NC 27710; e-mail: [email protected].
© 2003 by The Society of Thoracic Surgeons Published by Elsevier Inc
Population Selection After Institutional Review Board approval (Duke University Medical Center IRB registry #3048-01-8R0ER, ap0003-4975/03/$30.00 PII S0003-4975(03)00558-7
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
785
proved on August 3, 2001), demographic and outcome data were obtained from the Duke Cardiothoracic Surgery Database for all primary elective coronary artery bypass graft surgeries performed at Duke University Medical Center between February 1995 and February 1997. Patients with concomitant open chamber procedures and preoperative renal failure requiring dialysis were excluded from selection. Patients who required perioperative renal replacement therapy or who died within the first 2 postoperative days were also excluded as their serum creatinine values did not accurately reflect a perioperative renal insult. Perioperative variables were collected from the Duke Cardiothoracic Surgery Database, a prospectively entered quality assurance database that documents demographic data including the stage, severity history of ischemic heart disease, cardiac catheterization data, laboratory data, cardiac risk factors, and interventional cardiac therapy received within 48 hours of surgery. Intraoperative variables available in this dataset also include inotrope use, intraaortic balloon pump (IABP) therapy (preoperative need, intraoperative insertion, or postoperative use), duration of CPB and aortic cross clamp, highest prebypass serum glucose level, temperature management data, and number of coronary artery grafts performed. Postoperative data include transfusion of blood products, inotrope use, infectious postoperative complications, and hospital stay. In addition a Charlson comorbidity score was calculated for each patient using a list of preoperative demographic variables (ie, peripheral vascular disease, obstructive lung disease, peptic ulcer, diabetes, renal disease, liver disease, tumor/lymphoma/leukemia, metastatic cancer, dementia, and AIDS) [15].
surgery when compared other creatinine-derived markers of renal function [17], a consistent finding, regardless of preoperative renal status, that persists even in separate analysis of the group of patients who serum creatinine values never exceed the normal range. In our previous work we noted %⌬Cr after coronary bypass surgery to be remarkably similar among patient groups that differ by baseline renal function, creatinine values simply assuming proportionate but different absolute changes. CrmaxPost and lowest postoperative creatinine clearance were selected as secondary outcomes for analysis as they also highly correlate with postoperative morbidity and mortality but closely reflect impairment of renal filtration function [17].
Renal Function Assessment
Anesthesia and Surgery
Serum creatinine was measured as part of routine biochemical laboratory investigations for all elective coronary bypass surgery patients using a dry slide enzymatic reflectance technique (VITROS 950; Johnson and Johnson, New Brunswick, NJ) with a normal range of 0.7 to 1.4 mg/dL (62 to 124 mol/L).
Anesthesia was managed per the attending anesthesiologist’s preference. Use of agents with potential renal effects (eg, intravenous dopamine, diuretics, antifibrinolytic agents) were not controlled. Extracorporeal perfusion was performed using a Cobe CML Duo blood oxygenator with sealed hard-shell filtered venous reservoir (Cobe Laboratories, Lakewood, CO), a Cobe Century Perfusion System (Cobe Laboratories), and a 43-micron arterial line filter (Cobe Sentry arterial line filter with Primegard; Cobe Cardiovascular, Arvada, CO). Blood obtained by cardiotomy suction was routed by the roller pump into the integrated oxygenator-venous reservoir. Nonpulsatile perfusion was maintained at 2 to 2.4 L 䡠 min⫺1 䡠 m⫺2. The target CPB perfusion rate of 2 to 2.4 L 䡠 min⫺1 䡠 m⫺2 was maintained throughout the weight range. Other standard perfusion guidelines included titration of CPB perfusion rate to maintain a SvO2 of at least 60% during normothermic and 70% during hypothermic CPB. The bypass circuit was primed with mannitol (50 g of 20% solution) and crystalloid solution (0.9% normal saline). Typically, acceptable hematocrit ranged from 22% to 24% during bypass, with red blood cell transfusion usually occurring when values below 20% were observed. The arterial carbon dioxide tension was maintained
Definition of Variables The preoperative serum creatinine (CrPre) was identified as the value on the day before surgery for all inpatients, and was assessed within 1 week prior to surgery for all outpatients scheduled to undergo elective coronary artery bypass graft surgery. Peak postoperative creatinine (CrmaxPost) was defined as the highest of the daily in-hospital postoperative values. Peak percentage change in postoperative creatinine (%⌬Cr) was defined as the difference between the CrPre and CrmaxPost represented as a percentage of the preoperative value. Using the Cockroft Gault equation [16] preoperative creatinine clearance (CrClPre) was estimated from CrPre and lowest postoperative creatinine clearance (CrClPost) was estimated from CrmaxPost. The primary outcome variable, %⌬Cr, was selected as the marker that demonstrates the best association with mortality and major morbidity after coronary bypass
Hematocrit, Hemodynamic, and Transfusion Data Hematocrit was measured every 15 minutes during CPB as part of routine arterial blood gas analysis protocols using the AVL Omni Modular System blood gas analyzer (Roche Diagnostics, Indianapolis, IN). Hemodynamic parameters were available for each minute during CPB from the ARKIVE Information Management System (Arkive IMS, San Diego, CA). The TM50 is a previously reported, derived index of the degree and duration of low CPB perfusion pressure [18, 19]. The TM50 was defined as the time-pressure integral mean CPB blood pressure less than 50 mm Hg (min 䡠 mm Hg) or as the area under the CPB blood pressure curve below 50 mm Hg. Transfusion was defined in two variables: (1) RBC 48 hours: the number of units of packed red cells given within the first 48 hours postoperatively; and (2) transfusion: transfusion of more than 2 units of packed red cells and at least one other blood product.
CARDIOVASCULAR
Ann Thorac Surg 2003;76:784 –92
CARDIOVASCULAR
786
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
throughout bypass at 35 to 40 mm Hg (uncorrected for temperature) with the arterial oxygen tension maintained at 150 to 250 mm Hg. Target mean arterial pressure was between 50 and 70 mm Hg during bypass using intravenous phenylephrine or sodium nitroprusside as required. Typically patients were cooled to a nasopharyngeal temperature of 34°C to 28°C during bypass and rewarmed to a nasopharyngeal temperature of 37°C or a bladder temperature of at least 36°C prior to separation from bypass.
Statistical Analysis Simple descriptive statistics (mean, standard deviation) were used to describe the population demographics. We performed a primary stepwise multivariable linear regression analysis to test the association of lowest hematocrit during bypass with %⌬Cr. To assess robustness of the results and to increase generalizability, secondary multivariable regression analyses were also performed to test the association between lowest hematocrit during bypass and CrmaxPost and CrClPost. First we performed a full-model analysis in which we included all the demographic, comorbidity, and intraoperative data. To follow up this analysis we implemented a stepwise selection so that the final model included only variables that were significantly associated with (p ⬍ 0.10) the outcome (%⌬Cr). Of note the secondary analysis using CrmaxPost, while controlling for preoperative creatinine, is closely equivalent (and statistically identical) to assessing pre to peak change (ie, delta) in creatinine. All statistical analyses were performed using the SAS Statistical software (software version 8.0; SAS Institute, Cary, NC). A p value of less than 0.05 was considered significant.
Results A total of 1,404 patients undergoing primary elective CABG surgery met the criteria for analysis. Their demographic profile (Table 1) was similar to that of other previously reported populations [4]. Hematocrit, hemodynamic, and creatinine data are shown in Table 2. Among the excluded patients, 6 died within 48 hours of surgery, 7 were receiving preoperative dialysis, and 5 patients had acute renal failure requiring renal replacement therapy within 10 days of surgery. In the primary multivariable linear regression analysis of variables associated with %⌬Cr (Table 3), we found a significant two-way interaction between body weight and lowest hematocrit during bypass (parameter estimate [PE] ⫽ ⫺0.05; p ⫽ 0.007). With the inclusion of the hematocrit-body weight interaction in this model, there was a highly significant association between lowest hematocrit during bypass and %⌬Cr (PE ⫽ 4.5; p ⫽ 0.008). Figure 1 shows the influence of body weight on the association between lowest hematocrit during bypass and %⌬Cr. Other variables known to be renal risk factors including preoperative creatinine (CrPre), body weight, Charlson comorbidity score, prebypass serum glucose, perioperative blood transfusion, preoperative hematocrit, and postoperative IABP and inotrope use were also
Ann Thorac Surg 2003;76:784 –92
Table 1. Demographics Variable
Mean (SD)
Age (years) Female sex (%) African-American ethnicity (%) Weight (kg) Hypertension (%) Cerebrovascular accident (%) Diabetic (%) Previous myocardial infarction (%) Ejection fraction (%) Unstable angina (%) Carotid bruit (%) Preoperative congestive heart failurea (%) Cardiopulmonary bypass duration (min) Cross-clamp time (min) Preoperative intraaortic balloon pump (%) Postoperative intraaortic balloon pump (%) Preoperative inotrope (%)b Postoperative inotrope (%) Red blood cells 48 hours (units)c Transfusion (%)d Charlson comorbidity score
63.3 (11.0) 32.7 11.9 82.1 (16.6) 69.2 1.6 31.6 27.8 53.5 (14.0) 49.1 11.6 15.4 103.2 (39.9) 49.0 (18.2) 3.7 1.0 1.2 25.7 0.9 (1.6) 7.0 0.81 (1.02) (range 0 – 8) 132 (48)
Prebypass serum glucose (mg/dL) a
Congestive heart failure is defined as history of paroxysmal or nocturnal b dyspnea responsive to afterload reduction drug therapy. Inotrope use defined as postoperative infusion of either dopamine ⬎ 5 g 䡠 kg⫺1 䡠 ⫺1 ⫺1 ⫺1 min or dobutamine ⬎ 5 g 䡠 kg 䡠 min , or both, or epinephrine c ⬎ 0.03 g 䡠 kg⫺1 䡠 min⫺1. Number of units of packed red cells given d within the first 48 hours postoperatively. Transfusion of more than 2 units of packed red cells and at least one other blood product.
significantly associated with %⌬Cr in this multivariable model (Table 3). It is noteworthy that the association between lowest hematocrit during bypass and %⌬Cr was significant even after accounting for preoperative hematocrit. Figure 2 demonstrates the relationship of lowest hematocrit during bypass to %⌬Cr in patients weighing
Table 2. Hematocrit, Hemodynamic, and Creatinine Data Variable
Mean (SD)
Range
Preoperative Hct (%) Lowest Hct during bypass (%) TM50 (mm Hg 䡠 min) CrPre (mg/dL) CrmaxPost (mg/dL) %⌬Cr (%) CrClPre (mL/min) CrClPost (mL/min)
40.1 (9.1) 19.5 (3.8) 135.9 (139.7) 1.1 (0.5) 1.41 (0.7) 26.6 (45.0) 80.9 (33.4) 67.2 (28.8)
22–52 10 –33 ⫺1042– 0 0.5– 4.9 0.5– 8.2 ⫺74.4 –722.2 10.5–254.7 10.2–212.3
CrClPost ⫽ lowest postoperative creatinine clearance; CrClPre ⫽ preoperative creatinine clearance; CrmaxPost ⫽ peak postoperative creatinine; CrPre ⫽ preoperative creatinine; Hct ⫽ hematocrit; %⌬Cr ⫽ peak postoperative fractional change in creatinine; TM50 ⫽ integral of time and mean arterial pressure less than 50 mm Hg during cardiopulmonary bypass.
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
787
Table 3. Multivariable Linear Regression Analysis of Factors Associated With Peak Postoperative Change in Creatinine Variable Intercept Lowest Hct during bypass (%) CrPre (mg/dL) Charlson comorbidity score Body weight (kg) Lowest Hct during bypass*weight Preoperative Hct (%) IABP use RBC 48 hours (units) Inotrope usea Prebypass serum glucose (mg/dL) TM50 (mm Hg 䡠 min)
Parameter Estimate ⫺59.0 4.5 ⫺15.0 3.5 1.3 ⫺0.05 ⫺0.6 31.3 2.3 7.8 0.07 ⫺0.00
95% Confidence Limits ⫺126.0 1.2 ⫺20.7 0.9 0.5 ⫺0.1 ⫺1.0 4.6 0.6 1.7 0.02 ⫺0.02
p Value
8.1 7.8 ⫺9.3 6.2 2.1 ⫺0.01 ⫺0.2 57.9 4.0 13.9 0.1 0.01
0.08 0.008 ⬍0.0001 0.009 0.001 0.007 0.008 0.02 0.009 0.01 0.01 0.64
a Inotrope use defined as postoperative infusion of either dopamine ⬎ 5 g 䡠 kg⫺1 䡠 min⫺1 or dobutamine ⬎ 5 g 䡠 kg⫺1 䡠 min⫺1, or both, or epinephrine ⬎ 0.03 g 䡠 kg⫺1 䡠 min⫺1.
CrmaxPost ⫽ peak postoperative serum creatinine; CrPre ⫽ preoperative serum creatinine; Hct ⫽ hematocrit; pump; RBC 48 hours ⫽ number of units of packed red cells given within the first 48 hours postoperatively.
75 kg or more, with 95% confidence limits. To address the question of whether there is an “elbow” in the data where the slope of the relationship between lowest hematocrit during bypass and %⌬Cr changes significantly, we used a restricted cubic splines analysis of our regression model. We found no significant cutoff where the association between lowest hematocrit during bypass and %⌬Cr diverged from a linear relationship. The secondary multivariable linear regression model analyzing the variables associated with CrmaxPost con-
Fig 1. This figure is based on a predictive model for “typical” patients relating lowest hematocrit (Hct) during bypass to peak percentage change in creatinine (%⌬Cr), where the following covariates were adjusted for: prebypass glucose ⫽ 130 mg/dL, Charlson comorbodity score ⫽ 1, preoperative hematocrit ⫽ 40%, preoperative creatinine ⫽ 1.5 mg/dL, and no intraaortic balloon pump, inotropes, transfusion, or diabetes. The regression equations are %⌬Cr ⫽ for 75 kg (triangles), 3.499 ⫹ (lowest Hct during bypass*0.604); for 100 kg (asterisks), 38.16 ⫹ (lowest Hct during bypass*⫺0.791); for 125 kg (circles), 72.82 ⫹ (lowest Hct during bypass*⫺2.186); for 150 kg (squares), 107.48 ⫹ (lowest Hct during bypass*⫺3.581).
IABP ⫽ intraaortic balloon
firmed the findings of the primary analysis. Figure 3 shows the influence of body weight on the association between lowest hematocrit during bypass and CrmaxPost. There was a significant two-way interaction between lowest hematocrit during bypass and body weight in this model as well (PE ⫽ ⫺0.0007; p ⫽ 0.003). After inclusion of this interaction in the model there was an independent association between lowest hematocrit during bypass and CrmaxPost (PE ⫽ 0.06; p ⫽ 0.004; Table 4). Other known renal risk factors were also similarly noted in this analysis. Similar results were also seen in a multivariable analysis of factors associated with lowest estimated postoperative creatinine clearance (by the Cockroft-Gault equation), with a significant interaction between lowest hematocrit and body weight. The potential for interdependent variables exists in our
Fig 2. The relationship of lowest hematocrit (Hct) during bypass to peak percentage change in creatinine (%⌬Cr) in patients (n ⫽ 776) with weight 75 kilos or more, with 95% confidence limits (dashed lines). The regression equation is %⌬Cr ⫽ 50.62 ⫹ (lowest Hct during bypass*⫺1.10).
CARDIOVASCULAR
Ann Thorac Surg 2003;76:784 –92
CARDIOVASCULAR
788
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
Ann Thorac Surg 2003;76:784 –92
Fig 3. This figure is based on a predictive model for “typical” patients relating lowest hematocrit (Hct) during bypass to (CrmaxPost), where the following covariates were adjusted for: prebypass glucose ⫽ 130 mg/dL, Charlson comorbodity score ⫽ 1, preoperative hematocrit ⫽ 40%, preoperative creatinine ⫽ 1.5 mg/dL, and no intraaortic balloon pump, inotropes, transfusion, or diabetes. The regression equations are CrmaxPost ⫽ for 75 kg (triangles), 1.54 ⫹ (lowest Hct during bypass*0.007); for 100 kg (asterisks), 2.0 ⫹ (lowest Hct during bypass*⫺0.006); for 125 kg (circles), 2.46 ⫹ (lowest Hct during bypass*⫺0.03); for 150 kg (squares), 2.91 ⫹ (lowest Hct during bypass*⫺0.05).
analysis and may complicate interpretation of our results. Therefore, we performed further analyses to address this issue. It is possible that low hematocrit during bypass may be a marker for low hematocrit early after bypass, an effect that might influence the need for drugs that increase systemic vascular resistance, such as inotropes. We confirmed a relationship between low hematocrit during bypass and postoperative inotropes (a logistic regression indicates that low hematocrit during bypass predicts inotrope use, p ⫽ 0.05, c index ⫽ 0.53). We also found that inotrope use is a significant predictor of %⌬Cr (p ⫽ 0.0008, beta coefficient ⫽ 9.8). Because the predictors
are related to each other and also to the outcome, this raises the issue of confounding that cannot be completely addressed by our statistical models. However because the inclusion of inotropes in the multivariable model (see Table 3) does not alter the significance of our predictors of interest (weight, lowest hematocrit during bypass, and the interaction term) we conclude that these variables contribute an independent predictive ability beyond the role of postoperative inotropes. A second concern in interpretation of our data relates to the issue of transfusion and lowest hematocrit during bypass because it is difficult to separate the effects of transfusion from the need for transfusion as they would always be expected to occur together—that is, everyone who needs a transfusion gets a transfusion. We found an association between lower hematocrits during bypass and increased transfusion in a univariate regression model (p ⬍ 0.0001, beta coefficient ⫽ ⫺0.07). However because the inclusion of red blood cells transfusion in the model does not alter the significance of our predictors we again conclude that these variables contribute independent predictive ability. Because this relationship raises the issue of interdependent variables we performed tolerance and variance inflation diagnostics on our models designed to detect the presence of multicollinearity. When lowest hematocrit during bypass and transfusion are in a model together, the tolerance for the variables is 0.78 and 0.92 respectively (corresponding to variance inflation values of 1.3 and 1.1). Because the tolerance values are greater than 0.4 we conclude that the model is appropriate statistically. Using correlated variables as predictors in a model requires careful consideration and careful interpretation our diagnostics indicate that our models do not violate accepted guidelines for multicollinearity. Because hematocrit is directly influenced by concurrent blood transfusion we performed a subsequent analysis of the relationship between perioperative transfu-
Table 4. Multivariable Linear Regression Analysis of Factors Associated With Postoperative Peak Creatinine Variable Intercept Lowest Hct during bypass (%) CrPre (mg/dL) Charlson comorbidity score Body weight (kg) Lowest Hct during bypass*weight Preoperative Hct (%) IABP use RBC 48 hours (units) Inotrope usea Prebypass serum glucose (mg/dL) TM50 (mm Hg min)
Parameter Estimate ⫺0.9 0.06 ⫺0.2 0.06 0.02 ⫺0.0007 ⫺0.008 0.5 0.03 0.08 0.001 ⫺0.00003
95% Confidence Limits ⫺1.7 0.02 ⫺0.3 0.02 0.007 ⫺0.001 ⫺0.01 0.2 0.006 0.002 0.0002 ⫺0.0003
⫺0.008 0.1 ⫺0.1 0.09 0.03 ⫺0.0003 ⫺0.002 0.8 0.05 0.2 0.002 0.0002
p Value 0.05 0.004 ⬍0.0001 0.0007 0.0006 0.003 0.006 0.005 0.01 0.04 0.008 0.77
a Inotrope use defined as postoperative infusion of either dopamine ⬎ 5 g 䡠 kg⫺1 䡠 min⫺1 or dobutamine ⬎ 5 g 䡠 kg⫺1 䡠 min⫺1, or both, or epinephrine ⬎ 0.03 g 䡠 kg⫺1 䡠 min⫺1 or isoprenaline ⬎ 0.03 g 䡠 kg⫺1 䡠 min⫺1, or both.
CrPre ⫽ preoperative serum creatinine; Hct ⫽ hematocrit; IABP ⫽ intraaortic balloon pump; RBC 48 hours ⫽ number of units of packed red cells given within the first 48 hours postoperatively; TM50 ⫽ integral of time and mean arterial pressure less than 50 mm Hg during cardiopulmonary bypass.
Fig 4. The influence of blood transfusion on the relationship between lowest hematocrit (Hct) during cardiopulmonary bypass and peak postoperative fractional change in creatinine (%⌬Cr). RBCs: 9 (asterisks); 6 (triangles); 3 (circles); 0 (squares). (RBCs ⫽ number of units of packed red blood cells transfused intraoperatively.)
sion, lowest hematocrit during bypass, and %⌬Cr (see Fig 4). For any given lowest hematocrit during bypass we found a direct relationship between increased transfusion and greater %⌬Cr.
Comment We found a significant association between lowest hematocrit during bypass and creatinine rise after coronary bypass surgery that is influenced by body weight. The degree to which lowest hematocrit during bypass is adversely related to postoperative creatinine rise is proportional to increasing body weight (Fig 1). The following example illustrates this finding. In a 75-kg patient there is no association between lowest hematocrit during bypass and postoperative creatinine rise. However in a 150-kg patient there is a highly significant inverse association between lower hematocrit during bypass and greater postoperative creatinine rise (Fig 1). Our findings are consistent in multivariable analyses of the factors associated with both %⌬Cr and CrmaxPost. We also confirmed associations of previously recognized renal risk factors [3, 4] including elevated preoperative creatinine, Charlson comorbidity index, intraoperative highest prebypass serum glucose, postoperative inotrope and IABP use, preoperative hematocrit, blood transfusion and body weight with both, %⌬Cr and CrmaxPost. The absence of low bypass perfusion pressure (TM50) as a significant factor in both multivariable analyses is notable and confirms previous reports of lack of evidence of perfusion pressure as a predictor of renal dysfunction [20 –22]. The study design, large number of patients, and the use of valid markers of renal injury are the strengths of this study. The significant association of lowest hematocrit during bypass with both creatinine markers is consistent with an important contributory role of hemodilution in the pathophysiology of acute perioperative renal injury after cardiac surgery. The effects of hemodilution during bypass on various
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
789
organ systems are well described [14, 23, 24]. The range of lowest hematocrit during bypass in our study was almost identical to that observed in the two major previous publications relating hemodilution during bypass with postoperative complications involving 9,700 patients (10% to 33% versus 10% to 36% and 10% to ⬎35%) [10, 11]. In addition our mean lowest hematocrit during bypass (19.5%) was similar to those quoted in these studies, including one center reporting 2,738 patients with a mean lowest hematocrit of 18.8% [11]. Hemodilution reduces the need for addition of blood to the bypass circuit, both avoiding transfusion and reducing viscosity during hypothermic bypass. As acid-base abnormalities do not develop routinely during bypass, the resultant increase in regional blood flow during bypass has been thought to compensate for the reduced oxygen carrying capacity of the perfusate compared with prebypass levels. However the breakdown of the balance between oxygen supply and demand in this setting has not otherwise been extensively studied. Although the association of hemodilution or low hematocrit with adverse outcomes after cardiac surgery has been reported [10 –12] the degree of hemodilution that can be tolerated before an adverse effect occurs is not yet known. DeFoe and colleagues [10] reported an unfavorable association between lowest hematocrit during bypass and adverse in-hospital outcomes after cardiac surgery in a large group of patients in an observational study. However this study did not comment on adverse renal outcomes and did not investigate perioperative renal dysfunction as a primary outcome. In another retrospective analysis, Fang and coworkers [11] reported that lowest hematocrit during bypass was independently associated with mortality after cardiac surgery. This study also recognized preoperative renal failure as one of the risk factors for mortality in addition to lowest hematocrit but did not identify perioperative renal dysfunction as an adverse outcome related to lowest hematocrit during bypass. Ranucci and associates [12] found a hematocrit of less than 25% during bypass, low output syndrome, and homologous blood transfusions to be predictors of severe renal dysfunction in 316 patients undergoing cardiac surgery. However the results from this study were based on a regression analysis looking at three outcome variables in a subgroup of only 9 patients with severe renal dysfunction. These results also did not correct for preoperative creatinine, which is known to influence postoperative creatinine rise [3]. Our study investigates the independent association of lowest hematocrit during bypass with acute renal injury as a primary outcome in coronary bypass surgery patients. Our report has limitations related to study design and measurement of renal function. Although this study is retrospective, the data were taken from a prospectively gathered quality-assurance database. A second concern is that our study evaluated only one aspect of renal function, namely filtration, and not other renal homeostatic roles including osmolality, electrolyte and acidbase regulation and production, and release of enzymes and hormones. Although more sensitive markers of sub-
CARDIOVASCULAR
Ann Thorac Surg 2003;76:784 –92
CARDIOVASCULAR
790
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
tle renal injury than creatinine such as urinary N-acetyl -D-glucosaminidase [25] and cystatin C [26] have been reported, the link between these markers and adverse postoperative complications has not been established [27]. Moreover creatinine-based markers have consistently proved to be both sensitive indicators of renal injury and robust prognostic indicators of adverse outcome in patients with cardiovascular disease [28, 29] including those undergoing cardiac surgery [30]. We found a significant association between blood transfusion and acute renal injury. A potential concern is that extreme hemodilution may simply be a marker for increased likelihood of blood transfusion. Supporting this interpretation is the interaction we observed between hemodilution, body weight, and renal injury, with greater transfusion requirements to achieve a “target hematocrit” in an overweight patient. However in our multivariable analyses accounting for blood transfusion we still found lowest hematocrit during bypass to be significantly associated with postoperative creatinine rise (Tables 3, 4; Figs 1–3). Despite these limitations we feel that the use of valid markers of renal injury combined with data analysis in a large group of patients strengthen the significance of our findings. Although cause and effect are not demonstrated by our findings of an association of severe hemodilution during bypass and acute renal injury, it is important to explore possible physiologic foundations for this relationship. Because hematocrit is directly related to tissue oxygen delivery a simple interpretation of our findings is that severe hemodilution may contribute to renal injury by augmenting the local renal inflammatory response as a result of ischemia-reperfusion injury, particularly in the hypoxic environment of the renal medulla. However hemodilution has previously been thought to be renoprotective [31]. In an animal model of ischemic acute renal injury, Olof and colleagues [32] reported that trapping of red blood cells in the medullary vasculature and associated reductions in outer medullary blood flow observed with normal and elevated hematocrits are completely avoided with hemodilution (30%). Although other authors confirmed the avoidance “red cell trapping” with hemodilution, they could not confirm a benefit in terms of reduced renal injury [33, 34]. In a randomized study of 300 patients we did not find a difference between warm and cold bypass, with regard to renal injury, [7], suggesting a pathophysiology independent of oxygen demand. Other factors than ischemiareperfusion injury such as atheroembolism and systemic inflammatory response have been implicated as important contributors to post– cardiac surgery nephropathy. Koh and colleagues [35] speculated that alterations in the pattern of aortic blood flow related to cardiopulmonary bypass, potentially including the effects of hemodilution, may increase emboli delivery to the kidney and other organs. The relationship of hemodilution and inflammation during bypass is not known but it is possible that gut ischemia related to reduced tissue oxygen delivery may result in greater release of endotoxin and inflammationmediated renal injury [36]. Visceral fat is a key regulator
Ann Thorac Surg 2003;76:784 –92
site for the process of inflammation and obesity is associated with an upregulated inflammatory state [37]. The interaction between lowest hematocrit during bypass and increased weight that we observed in our study may be explained by the combining of effects that increase inflammatory stimuli and inflammatory response respectively. The significant association between lowest hematocrit during bypass and acute renal injury raises concerns regarding currently accepted bypass management strategies. Current guidelines for accepted levels of hemodilution during bypass do not lead to the development of acid-base disturbances. However occult cellular hypoxemia and injury may still occur. Such injury in the renal medulla, a tissue that is known to have a precarious balance between oxygen supply and demand even under normal conditions [38], may be the basis of the findings of our study. Strategies to avoid extreme hemodilution may include transfusion. However since transfusion of red cells is also independently associated with creatinine rise in our study (see Tables 3, 4; Fig 4) guidelines for optimal hematocrit during bypass that include transfusion cannot be simply defined. The decision to transfuse is usually based on clinical judgment, institutional protocols, and hematocrit “triggers.” Because transfusion has not been tested as a primary outcome of this study the authors cannot recommend changes in transfusion policy based on its effect on perioperative creatinine changes. However clearly minimizing bypass prime volumes should reduce anemia without transfusion and potentially reduce renal risk. Our data suggest that hemodilution during bypass should be carefully monitored and excessive falls in hematocrit avoided, especially in overweight patients (see Fig 2). We did not observe an “elbow” in the graph relating lowest hematocrit to %⌬Cr that would suggest a level below which hematocrit is detrimental to renal function. In summary we found an independent association between lowest hematocrit during bypass and postoperative creatinine rise that is influenced by body weight. This association was significant for all creatinine markers of renal injury that we assessed (%⌬Cr, CrmaxPost, and CrClPost). We did not observe an effect of low bypass perfusion pressure on creatinine rise. This is the first report highlighting the association of extreme hemodilution during bypass with acute renal injury as a primary outcome. Despite being a retrospective data analysis looking at one aspect of renal function we believe that the study design and the use of valid markers of renal injury justify our conclusions. Our findings question the wisdom of tolerating low levels of hematocrit during bypass. We have previously reported the lack of effect of bypass temperature on acute renal injury [7]. These findings suggest that optimal bypass management guidelines should emphasize the maintenance of adequate hematocrit levels during bypass regardless of temperature protocol employed. However the independent contributions of hemodilution and transfusion to the pathophys-
iology of perioperative renal injury require further examination in a prospective randomized trial looking at a variety of tests of renal function. This work was supported by the Cardiothoracic Division of the Department of Anesthesiology, Duke University Medical Center, Durham, North Carolina. The authors would like to gratefully acknowledge the contribution of Christopher Keith, and Cheryl Stetson, Department of Anesthesiology, Duke University Medical Center, in the preparation of this manuscript.
References 1. Ostermann ME, Taube D, Morgan CJ, Evans TW. Acute renal failure following cardiopulmonary bypass: a changing picture. Intens Care Med 2000;26:565–71. 2. Chertow GM, Lazarus JM, Christiansen CL, et al. Preoperative renal risk stratification. Circulation 1997;95:878 –84. 3. Conlon PJ, Stafford-Smith M, White WD, et al. Acute renal failure following cardiac surgery. Nephrol Dial Transplant 1999;14:1158 –62. 4. Mangano CM, Diamondstone LS, Ramsay JG, Aggarwal A, Herskowitz A, Mangano DT. Renal dysfunction after myocardial revascularization: risk factors, adverse outcomes, and hospital resource utilization. The Multicenter Study of Perioperative Ischemia Research Group. Ann Intern Med 1998; 128:194 –203. 5. Mangos GJ, Brown MA, Chan WY, Horton D, Trew P, Whitworth JA. Acute renal failure following cardiac surgery: incidence, outcomes and risk factors. Aust NZ J Med 1995; 25:284 –9. 6. Flemming B, Seeliger E, Wronski T, Steer K, Arenz N, Persson PB. Oxygen and renal hemodynamics in the conscious rat. J Am Soc Nephrol 2000;11:18 –24. 7. Swaminathan M, East C, Phillips-Bute B, et al. Report of a substudy on warm versus cold cardiopulmonary bypass: changes in creatinine clearance. Ann Thorac Surg 2001;72: 1603–9. 8. Messmer K. Hemodilution. Surg Clin North Am 1975;55: 659 –78. 9. Shah D, Corson J, Karmody A, Leather R. Effects of isovolemic hemodilution on abdominal aortic aneurysmectomy in high risk patients. Ann Vasc Surg 1986;1:50 –4. 10. DeFoe G, Ross C, Olmstead E, et al, for Group NNECDS. Lowest hematocrit on bypass and adverse outcomes associated with coronary artery bypass grafting. Ann Thorac Surg 2001;71:769 –76. 11. Fang WC, Helm RE, Krieger KH, et al. Impact of minimum hematocrit during cardiopulmonary bypass on mortality in patients undergoing coronary artery surgery. Circulation 1997;96(Suppl 2):194 –9. 12. Ranucci M, Pavesi M, Mazza E, et al. Risk factors for renal dysfunction after coronary surgery: the role of cardiopulmonary bypass technique. Perfusion 1994;9:319 –26. 13. Cook D, Orszulak TA, Daly R. Minimum hematocrit at differing cardiopulmonary bypass temperatures in dogs. Circulation 1998;98(Suppl):II170 –4; discussion II175. 14. Liam BL, Plochl W, Cook DJ, Orszulak TA, Daly RC. Hemodilution and whole body oxygen balance during normothermic cardiopulmonary bypass in dogs. J Thorac Cardiovasc Surg 1998;115:1203–8. 15. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chron Dis 1987;40: 373–83. 16. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron 1976;16:31–41. 17. Stafford-Smith M, Reddan D, Phillips-Bute B, Newman M, Winn M. Association of perioperative creatinine-derived variables with mortality and other outcomes after coronary bypass surgery. Anesth Analg 2001;92:SCA28.
SWAMINATHAN ET AL CPB HEMATOCRIT AND RENAL INJURY
791
18. Gold JP, Charlson ME, Williams-Russo P, et al. Improvement of outcomes after coronary artery bypass. A randomized trial comparing intraoperative high versus low mean arterial pressure. J Thorac Cardiovasc Surg 1995;110:1302–11. 19. Hill SE, van Wermeskerken GK, Lardenoye JW, et al. Intraoperative physiologic variables and outcome in cardiac surgery: part I. In-hospital mortality. Ann Thorac Surg 2000;69: 1070 –5. 20. Urzua J, Troncoso S, Bugedo G, et al. Renal function and cardiopulmonary bypass: effect of perfusion pressure. J Cardiothorac Vasc Anesth 1992;6:299 –303. 21. Bhat JG, Gluck MC, Lowenstein J, Baldwin DS. Renal failure after open heart surgery. Ann Intern Med 1976;84:677–82. 22. Abel RM, Wick J, Beck CH Jr, Buckley MJ, Austen WG. Renal dysfunction following open-heart operations. Arch Surg 1974;108:175–7. 23. Hall TS. The pathophysiology of cardiopulmonary bypass. The risks and benefits of hemodilution. Chest 1995;107:1125– 33. 24. Sungurtekin H, Cook DJ, Orszulak TA, Daly RC, Mullany CJ. Cerebral response to hemodilution during hypothermic cardiopulmonary bypass in adults. Anesth Analg 1999;89:1078 – 83. 25. Hsu WS, Kao JT, Chen JS. Clinical significance of urinary N-acetyl-beta-D-glucosaminidase and alanine aminopeptidase. Taiwan Yi Xue Hui Za Zhi 1989;88:407–9. 26. Fliser D, Ritz E. Serum cystatin C concentration as a marker of renal dysfunction in the elderly. Am J Kidney Dis 2001; 37:79 –83. 27. Hamada Y, Kanda T, Anzai T, Kobayashi I, Morishita Y. N-acetyl-beta-D-glucosaminidase is not a predictor, but an indicator of kidney injury in patients with cardiac surgery. J Med 1999;30:329 –36. 28. Friedman PJ. Serum creatinine: an independent predictor of survival after stroke. J Intern Med 1991;229:175–9. 29. Shulman NB, Ford CE, Hall WD, et al. Prognostic value of serum creatinine and effect of treatment of hypertension on renal function. Results from the hypertension detection and follow-up program. The Hypertension Detection and Follow-up Program Cooperative Group. Hypertension 1989; 13(Suppl):I80–93. 30. Stafford-Smith M, Phillips-Bute B, Reddan D, Milano C, Newman M, Winn M. The association of postoperative peak and fractional change in serum creatinine with mortality after coronary bypass surgery. Anesthesiology 2000; 93(Suppl):A240. 31. Rajagopalan PR, Reines HD, Pulliam C, Fitts CT, LeVeen HH. Reversal of acute renal failure using hemodilution with hydroxyethyl starch. J Trauma 1983;23:795–800. 32. Olof P, Hellberg A, Kallskog O, Wolgast M. Red cell trapping and postischemic renal blood flow. Differences between the cortex, outer and inner medulla. Kidney Int 1991;40:625–31. 33. Hellberg PO, Kallskog O, Wolgast M. Nephron function in the early phase of ischemic renal failure. Significance of erythrocyte trapping. Kidney Int 1990;38:432–9. 34. Andersson G, Jennische E. Lack of casual relationship between medullary blood congestion and tubular necrosis in postischaemic kidney damage. Acta Physiol Scand 1987;130:429–32. 35. Koh TW, Parker KH, Kon M, Pepper JR. Changes in aortic rotational flow during cardiopulmonary bypass studied by transesophageal echocardiography and magnetic resonance velocity imaging: a potential mechanism for atheroembolism during cardiopulmonary bypass. Heart Vessels 2001;16:1–8. 36. Gow KW, Phang PT, Tebbutt-Speirs SM, et al. Effect of crystalloid administration on oxygen extraction in endotoxemic pigs. J Appl Physiol 1998;85:1667–75. 37. Ziccardi P, Nappo F, Giugliano G, et al. Reduction of inflammatory cytokine concentrations and improvement of endothelial functions in obese women after weight loss over one year. Circulation 2002;105:804 –9. 38. Brezis M, Epstein FH. Cellular mechanisms of acute ischemic injury in the kidney. Annu Rev Med 1993;44:27–37.
CARDIOVASCULAR
Ann Thorac Surg 2003;76:784 –92