| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Departments of Anesthesia (Drs. Wang, Dupuis, and Nathan) and Epidemiology (Ms. Williams), University of Ottawa Heart Institute, Ottawa, ON, Canada.
Correspondence to: Jean-Yves Dupuis, MD, Department of Anesthesia, University of Ottawa Heart Institute, 40 Ruskin St, Room H213, Ottawa, ON, Canada, K1Y 4W7; e-mail: jydupuis{at}ottawaheart.ca
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Design: Prospective, observational.
Setting: University hospital.
Patients: A total of 6,364 cardiac surgical patients.
Methods: The measured outcomes were postoperative renal failure requiring dialysis, and mortality and major morbidity. For each outcome, two multivariable risk models were developed, using either estimated ClCr as a measure of renal function, or plasma creatinine level. Risk-adjusted odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated for each outcome. Discrimination was compared using receiver operating characteristic (ROC) curves.
Results: For each 10 mL/min/1.73 m2 decrement of estimated ClCr, the ORs for renal failure requiring dialysis, mortality, and major morbidity in the whole population were 1.52 (95% CI, 1.35 to 1.67), 1.27 (95% CI, 1.19 to 1.35), and 1.18 (95% CI, 1.14 to 1.21), respectively; for each 0.2 mg/dL increment of plasma creatinine, ORs were 1.20 (95% CI, 1.15 to 1.26), 1.08 (95% CI, 1.04 to 1.13), and 1.12 (95% CI, 1.09 to 1.15), respectively. The areas under the ROC curves for prediction of renal failure requiring dialysis were 0.83 with both risk models. For prediction of mortality and major morbidity, areas under the ROC curves were 0.83 and 0.72, respectively, with the models using estimated ClCr, and 0.74 and 0.65, respectively, with the models using plasma creatinine level (p < 0.001 vs estimated ClCr for both outcomes). In patients with normal plasma creatinine levels (n = 4,603), estimated ClCr remained a significant predictor of each outcome with similar ORs, but plasma creatinine level was not a predictor of any outcome.
Conclusion: The risk-adjusted association between preoperative renal dysfunction and adverse events after cardiac surgery is stronger with estimated ClCr than with plasma creatinine level, particularly in patients with normal plasma creatinine levels. The routine preoperative estimation of ClCr may improve the identification of higher-risk cardiac surgical patients.
Key Words: clearance • creatinine • heart surgery • kidney • outcome • plasma • risk
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In this study, we estimated ClCr in a large cohort of cardiac surgical patients, using the formula developed by Cockroft and Gault.16 Previously evaluated in cardiac patients, this simple formula predicts ClCr with acceptable accuracy.17 To test our hypothesis, we developed preoperative multivariable risk models, using either estimated ClCr or plasma creatinine level as a measure of renal function, and assessed the ability of those models in predicting postoperative renal failure requiring dialysis, and mortality and major morbidity. We further determined the association between renal function and outcome by calculating the adjusted risks of major postoperative complications as the preoperative estimated ClCr decreased, considering separately patients with normal plasma creatinine levels and those with elevated levels.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Data Collection
All data were collected prospectively on a daily basis by research assistants who completed a preoperative and a postoperative datasheet. The preoperative datasheet contains 130 variables pertaining to the severity of disease and comorbid factors before surgery. The postoperative datasheet contains 92 variables related to outcome and patient evolution after surgery. The quality of the collected data were assessed on a regular basis by a physician (J.Y.D.) who verified the database information from randomly selected patients representing 5% of the whole population. The information was then compared with the data from the charts. An agreement rate of 98% between the database information and the data obtained from the charts has previously been reported.13
Plasma Creatinine Level and Estimation of ClCr
Plasma creatinine levels were measured in all patients by kinetic colorimetric testing (Jaffe method), using an automatic analyzer (Hitachi 917 Automatic Autoanalyzer; Boehringer Mannheim; Montreal, Canada). Plasma creatinine levels are reported in micromoles per liter by our laboratory. Those values were divided by a conversion factor of 88.4 to obtain values in milligrams per deciliter for the present study. After conversion, the upper limits for normal plasma creatinine levels in our institution are 1.0 mg/dL for women and 1.2 mg/dL for men. The ClCr was estimated using the equations developed by Cockroft and Gault,16
and adjusted for each 1.73 m2 of body surface area (BSA):
For men:
 |
 |
Outcomes
The studied outcomes were postoperative acute renal failure requiring dialysis, in-hospital mortality (regardless of length of stay), and major morbidity including: (1) cardiovascular: low cardiac output, hypotension, or both treated with intra-aortic balloon pump or with the infusion of at least two inotropes or vasopressors for > 24 h, malignant arrhythmia (asystole and ventricular tachycardia or fibrillation) requiring cardiopulmonary resuscitation, anti-arrhythmia therapy or automatic cardiodefibrillator implantation; (2) respiratory: mechanical ventilation for > 48 h, tracheostomy, reintubation; (3) neurologic: focal brain injury with permanent functional deficit, irreversible encephalopathy; (4) renal dysfunction7
: postoperative serum creatinine level of 
2 mg/dL and a minimum preoperative-to-postoperative increase in serum creatinine level of 0.7 mg/dL; (5) infectious: septic shock with positive blood culture results, deep sternal or leg wound infection requiring IV antibiotics and/or surgical debridement; and (6) other: any surgery or invasive procedure necessary to treat a postoperative adverse event associated with the initial cardiac surgery.
Statistical Analysis
The unadjusted association between patient characteristics and each outcome was assessed through univariate analyses, using the 
2 test. Odds ratios (ORs) with their 95% confidence intervals (CIs) were calculated for all variables. Receiver operating characteristic (ROC) curves were used to determine the best cutoff points for prediction of each complication, using estimated ClCr or plasma creatinine level as univariate risk predictor.
Stepwise multiple logistic regression analyses were performed to determine the association between each outcome and patient characteristics, with adjustments made for potential confounders. Two multivariable risk models were developed for each outcome, one using estimated ClCr and the other using plasma creatinine level as measures of renal function. Continuous and categoric variables were computed as such in the regression analyses. Left ventricular evaluation was dichotomized based on the presence or absence of severe dysfunction. This was done because left ventricular evaluation was missing in 5% of the cases or defined qualitatively in 35% of them. Severe left ventricular dysfunction was defined by one of the following: ejection fraction < 30% by echocardiography, nuclear imaging or radiographic planimetry, or severe global hypokinesis by echocardiography, or severe contractile anomaly of > 75% of the left ventricular circumference at ventriculography (right anterior oblique projection). Unknown left ventricular functions were considered as normal, as done in other studies19
20
estimating the risk of cardiac surgery. All variables with a p value 0.20 in the univariate analyses were considered for inclusion in the final multivariable models. As the equation to calculate ClCr includes age, gender, and weight, those risk factors were excluded from the risk models using estimated ClCr to avoid collinearity.21
The predictive accuracy of each multivariable risk model was determined and compared through ROC curves for each outcome. With this test, an area under the ROC curve of 1.0 indicates perfect accuracy, whereas an area < 0.5 (line of no discrimination) means that it is no better than chance. Areas of 0.5 to 0.7 suggest a low predictive accuracy, and values > 0.7 confirm the usefulness of the model as a risk predictor.22
Comparisons of the areas under the ROC curves and their SE were made using the pairwise deletion Mann-Whitney test. Calibration, which represents the precision of the probabilities generated by a prediction model, was assessed for each risk model using the Hosmer-Lemeshow goodness-of-fit test. The test compares the predicted with the observed rate of each outcome, for each decile of risk generated by the risk models. A small 2 value (or a p value > 0.05) indicates acceptable calibration.
The predictive value of estimated ClCr and plasma creatinine level as risk predictors in patients with normal plasma creatinine levels was assessed by repeating multiple regression analyses separately in patients with normal plasma creatinine levels and those with elevated plasma creatinine levels. The adjusted ORs for each outcome were calculated at various levels of estimated ClCr in each subgroup of patients, using the same reference population consisting of all patients having normal plasma creatinine levels and an estimated ClCr > 80 mL/s/1.73 m2.
SAS statistical software (version 8.02; SAS Institute; Cary, NC) was used in all analyses. For all statistics, a p value < 0.05 was considered significant.
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The characteristics of the study population and their association with outcomes as determined by univariate analyses are presented in Table 1
. The estimated ClCr ranged from 37 to 252 mL/min/1.73 m2 among patients with normal plasma creatinine levels and 8 to 105 mL/min/1.73 m2 in patients with elevated plasma creatinine levels. The distribution of estimated ClCr within each group of patients is presented in Table 2
.
|
|
|
|
|
|
|
There were 4,603 patients with normal plasma creatinine levels in the study. The rates of postoperative renal failure requiring dialysis, mortality, and major morbidity in those patients were 0.8%, 2.2%, and 17.9%, respectively. Multiple regression analyses repeated only in those patients with normal plasma creatinine levels showed that estimated ClCr remained a significant risk predictor of all outcomes. Each 10 mL/min/1.73 m2 decrement in estimated ClCr was associated with ORs of 1.59 (95% CI, 1.27 to 1.98) for renal failure requiring dialysis, 1.34 (95% CI, 1.19 to 1.51) for mortality, and 1.12 (95% CI, 1.08 to 1.17) for major morbidity. Emergency surgery, combined/complex surgical procedures, congestive heart failure, and age also remained significant risk factors for all studied outcomes in patients with normal plasma creatinine levels; however, plasma creatinine level was not a significant risk predictor for any outcome in that particular population: with each 0.2 mg/dL increase of plasma creatinine level, the ORs were 1.02 (95% CI, 0.65 to 1.57) for renal failure requiring dialysis, 1.02 (95% CI, 0.75 to 1.41) for mortality, and 1.08 (95% CI, 0.96 to 1.22) for major morbidity.
To further illustrate the advantage of estimated ClCr over plasma creatinine level as a risk predictor, the adjusted ORs for the studied outcomes were calculated separately in patients with normal plasma creatinine levels and in those with elevated plasma creatinine levels, at various levels of estimated ClCr. For each tested level of estimated ClCr, the ORs in patients with normal plasma creatinine levels were similar to the ORs in patients with elevated plasma creatinine levels (Fig 1 ).
|
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Most previous studies3 4 6 7 8 9 10 11 assessed the association between renal function and outcome after cardiac surgery by dichotomizing renal function, using a plasma creatinine level between 1.4 mg/dL and 1.7 mg/dL as a cutoff point to diagnose mild renal dysfunction. By using estimated ClCr and plasma creatinine level as continuous variables in multiple regression analyses, this study better defines and quantifies the relationship between those measures of renal function and outcome after cardiac surgery. After adjusting for confounding variables, the odds of developing renal failure requiring dialysis, dying, or having major morbidity after cardiac surgery increase by 52%, 27%, and 18% for each 10 mL/min/1.73 m2 decrement in estimated ClCr, respectively. Similarly, for each 0.2 mg/dL increment of plasma creatinine level, the risk-adjusted odds of having the same adverse outcomes increase by 20%, 8%, and 13%, respectively. Because estimated ClCr and plasma creatinine level are strong correlates of GFR, those results suggest that the rate of adverse outcomes after cardiac surgery is inversely proportional to GFR. This association is true with estimated ClCr whether the analyses include all patients or only those with normal plasma creatinine levels. With plasma creatinine level, the association is lost when the analyses are applied only to patients with normal plasma creatinine levels. The adjusted risk of adverse events is not significantly different whether patients have low or high normal levels of plasma creatinine. This probably reflects the inability of plasma creatinine level to detect significant differences of GFR in patients with low plasma creatinine levels. It also means that even if lower values of plasma creatinine were used to define renal dysfunction, estimated plasma ClCr would remain a better risk predictor in those particular patients.
Because estimated ClCr and plasma creatinine level are two different measures, it is impossible to compare their predictive weight on the basis of their respective adjusted OR for each outcome. This study used an alternate approach and compared the predictive performance of their respective multivariable risk model for each outcome. Both sets of risk models used the same confounding variables to predict each outcome. Thus, any difference in the predictive performance of the models for a particular outcome was likely due to a difference in the predictive strength between the two measures of renal function. The two sets of risk model performed equally well in predicting postoperative renal failure requiring dialysis; however, for prediction of mortality and major morbidity, the models using estimated ClCr had significantly better discrimination than the models using plasma creatinine level. The reason why the two sets of models performed similarly for prediction of renal failure requiring dialysis but not the other two outcomes cannot be explained with certainty. A decrease in estimated ClCr is a sensitive indicator of decreased GFR, while an elevated plasma creatinine level is a highly specific marker of the same condition. In this study, 98.5% of the patients with an elevated plasma creatinine level had an estimated ClCr of < 80 mL/min/1.73 m2 (Table 2) . The pathophysiologic link between decreased GFR and dialysis requirement is obvious; therefore, the high specificity of plasma creatinine level for low GFR probably compensates for its lower sensitivity in a way that significantly improves the discrimination of the risk model predicting dialysis. This is possible since ROC curves are simply plots of sensitivity over "1 - specificity." The association between renal dysfunction and mortality or morbidity may be more dependent on physiologic reserve than GFR. The equation to calculate ClCr may provide a better estimate of that reserve than plasma creatinine. This could explain the better predictive performance of the risk models using estimated ClCr for mortality and major morbidity. Those explanations are limited by the fact that comparisons of discrimination were made between the whole multivariable risk models, not between the individual risk factors.
This study suggests that the clinical appreciation of risk in cardiac surgery may be improved by the routine inclusion of estimated ClCr with laboratory reports of individual patients. Since the preexistence of renal impairment makes the kidney more vulnerable to ischemia and drug-related toxicity,7 11 23 24 25 26 27 the immediate availability of estimated ClCr could influence perioperative care. For example, in patients with decreased ClCr, clinicians may choose to improve renal perfusion (eg, optimization of cardiac output, and close control of fluids and electrolytes) and minimize exposure to potentially nephrotoxic agents (eg, radiographic contrast agents, nonsteroidal anti-inflammatory drugs, angiotensin-converting enzyme inhibitors, aminoglycoside antibiotics) before, during, and after surgery. Although the impact of those interventions on outcome is unknown, the increased morbidity and mortality associated with the postoperative deterioration of renal function probably justifies any attempt to minimize perioperative renal insults in patients with preexisting renal impairment.7 The use of estimated ClCr may also help in providing better informed consent to patients and in selecting participants for clinical trials. Most investigators use normal plasma creatinine level as an inclusion criteria defining normal renal function. Outcome results may be skewed by that approach because many cardiac surgical patients with normal plasma creatinine levels have low estimated ClCr and, as shown in this study, are at significantly higher risk of perioperative adverse events than those with normal estimated ClCr.
At least four important limitations must be considered in the interpretation of our results. The first one is the variability of the Cockroft and Gault16 formula in estimating ClCr as compared to values obtained from plasma and 24-h urine collection samples. The formula provides an acceptable estimate of ClCr in most stable cardiac patients, but it may underestimate GFR and conversely overestimate risk in obese patients and in patients with very low plasma creatinine levels.17 In contrast, the formula may overestimate ClCr and underestimate risk prediction in hemodynamically unstable patients with acute renal insult, since plasma creatinine levels may not have the time to reach its peak in those patients. Second, large prospective cohort studies are subject to errors due to the inadvertent entry of wrong data in the database and unavailability of certain data (eg, missing left ventricular function in some patients). A regular quality assessment of the data from randomly selected charts, as performed in this study, tend to reduce those errors, but does not eliminate them all. Observation bias, particularly for outcomes defined by clinical interventions (eg, dialysis, treatment of hemodynamic disturbances with two or more inotropes or with intra-aortic balloon pump, reintubation, and mechanical ventilation > 48 h depend on various treatment thresholds used by different clinicians) is another type of error found in studies derived from databases. Such bias are seen in almost all types of epidemiologic studies and are accepted as long as they remain random.28 A third limitation is the unexplored pathophysiologic link between renal function and outcome in this study. Decreased drug elimination, hypervolemia in oliguric patients, hyperkalemia, bleeding due to platelet dysfunction, anemia, and encephalopathy are all potential mechanisms whereby low GFR can complicate the perioperative period. Those pathologic conditions can also be induced by cardiac surgery-related factors including cardiopulmonary bypass (CPB), hypothermia, low cardiac output syndrome, and perioperative hemorrhage. In that context, preoperative renal dysfunction may simply be a nonspecific marker of decreased physiologic reserve rather than a cause of adverse events. Consequently, efforts to protect or improve renal function during the perioperative period must be undertaken with the understanding that they may not necessarily translate into improved outcomes. Finally, this study was performed in a single center, which may limit the generalization of its results to other centers; however, the mortality and rate of postoperative dialysis requirement found in our large surgical population compare very well with those from recent large multicenter trials in North America and Europe.7 29
This study shows that estimated ClCr has significant advantages over plasma creatinine level as a predictor of adverse events after cardiac surgery. This is particularly true for patients with normal plasma creatinine levels who, for any given level of estimated ClCr, have the same adjusted risk of renal failure requiring dialysis, and mortality and major morbidity as patients with elevated plasma creatinine levels. Overall, the results suggest that clinicians may improve their ability to identify higher-risk cardiac surgical patients by the routine use of estimated ClCr in their preoperative evaluation, at no extra cost or inconvenience to the patient.
|
Acknowledgements |
|---|
|
Footnotes |
|---|
This work was performed at the Department of Anesthesia, University of Ottawa Health Institute.
Supported by research funds of the Cardiac Surgical Unit and the Cardiac Anesthesia Division of the University of Ottawa Heart Institute.
Received for publication December 26, 2002. Accepted for publication April 16, 2003.
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  D. N. Wijeysundera, K. Karkouti, J.-Y. Dupuis, V. Rao, C. T. Chan, J. T. Granton, and W. S. Beattie Derivation and Validation of a Simplified Predictive Index for Renal Replacement Therapy After Cardiac Surgery JAMA, April 25, 2007; 297(16): 1801 - 1809. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Boodhwani, F. D. Rubens, D. Wozny, R. Rodriguez, A. Alsefaou, P. J. Hendry, and H. J. Nathan Predictors of Early Neurocognitive Deficits in Low-Risk Patients Undergoing On-Pump Coronary Artery Bypass Surgery Circulation, July 4, 2006; 114(1_suppl): I-461 - I-466. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Butler, C. Geisberg, R. Howser, P. M. Portner, J. G. Rogers, M. C. Deng, and R. N. Pierson III Relationship between renal function and left ventricular assist device use. Ann. Thorac. Surg., May 1, 2006; 81(5): 1745 - 1751. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Noyez, I. Plesiewicz, and F. W.A. Verheugt Estimated creatinine clearance instead of plasma creatinine level as prognostic test for postoperative renal function in patients undergoing coronary artery bypass surgery. Eur. J. Cardiothorac. Surg., April 1, 2006; 29(4): 461 - 465. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Leacche, W. C. Winkelmayer, S. Paul, J. Lin, D. Unic, J. D. Rawn, L. H. Cohn, and J. G. Byrne Predicting Survival in Patients Requiring Renal Replacement Therapy After Cardiac Surgery Ann. Thorac. Surg., April 1, 2006; 81(4): 1385 - 1392. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
O. V. Hein, J. Birnbaum, K. Wernecke, M. England, W. Konertz, and C. Spies Prolonged Intensive Care Unit Stay in Cardiac Surgery: Risk Factors and Long-Term-Survival. Ann. Thorac. Surg., March 1, 2006; 81(3): 880 - 885. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Asimakopoulos, A. P. Karagounis, O. Valencia, N. Alexander, M. Howlader, M. A. Sarsam, and V. Chandrasekaran Renal Function After Cardiac Surgery Off- Versus On-Pump Coronary Artery Bypass: Analysis Using the Cockroft-Gault Formula for Estimating Creatinine Clearance Ann. Thorac. Surg., June 1, 2005; 79(6): 2024 - 2031. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. M.A. van de Wal, B. L. van Brussel, A. A. Voors, T. D.J. Smilde, J. C. Kelder, H. A. van Swieten, W. H. van Gilst, D. J. van Veldhuisen, and H.W. T. Plokker Mild preoperative renal dysfunction as a predictor of long-term clinical outcome after coronary bypass surgery J. Thorac. Cardiovasc. Surg., February 1, 2005; 129(2): 330 - 335. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R de Vroege, F te Meerman, L Eijsman, W R Wildevuur, C. R. Wildevuur, and W van Oeveren Induction and detection of disturbed homeostasis in cardiopulmonary bypass Perfusion, September 1, 2004; 19(5): 267 - 276. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
