Does Coronary Artery Bypass Graft Surgery Improve Survival Among Patients With End-Stage Renal Disease?

Does Coronary Artery Bypass Graft Surgery Improve Survival Among Patients With End-Stage Renal Disease? Todd M. Dewey, MD, Morley A. Herbert, PhD, Syma L. Prince, RN, Carrie L. Robbins, RN, Christina M. Worley, RN, Mitchell J. Magee, MD, and Michael J. Mack, MD Cardiopulmonary Research Science and Technology Institute and Medical City Dallas Hospital, Dallas, Texas Background. Cardiovascular disease remains the most frequent cause of death for patients with end-stage renal disease. To determine the long-term benefit of surgical revascularization in this high-risk population, we studied our patients with ESRD having coronary artery bypass graft surgery (CABG), comparing the results of off-pump to on-pump revascularization. As a baseline reference group, we used dialysis patients with a diagnosis of coronary artery disease who did not have surgical revascularization or percutaneous coronary interventions. The control group data set was obtained from the United States Renal Data System. Methods. From January 1995 through July 2003, 158 patients with end-stage renal disease who were on hemodialysis (excluding those in cardiogenic shock, needing resuscitation, and with emergent or salvage status) underwent CABG. Fifty-nine patients (37.3%) had offpump revascularization, and 99 patients (62.7%) had bypass grafting utilizing extracorporeal circulation. Preoperative risk factors and operative results were analyzed, and longitudinal survival data obtained. Results. The mean follow-up time was 39.1 months (median, 33.1) for the on-pump patients and 18.3 months (median, 14.7) for off-pump. The total number of anastomoses per off-pump patient was 2.4 1.0, and with cardiopulmonary bypass (CPB), it was 3.3 0.9 (p < 0.001). Patients revascularized off-pump had an operative mortality rate of 1.7%, whereas patients grafted using CPB had an operative mortality of 17.2% (p 0.003). The predicted risk of mortality for the off-pump group (9.3% 7.4%) was not statistically different from the on-pump cohort (9.1% 7.7%, p not significant). Logistic regression analysis indicates that CPB use was an independent risk factor for early death (p 0.01, odds ratio 13.6, 95% confidence interval: 1.7 to 110). Long-term follow-up demonstrated that the patients revascularized using CPB had improved survival compared with the off-pump patients and the control population. Conclusions. Off-pump CABG improves early mortality rate when compared with conventional revascularization. Despite a greater operative mortality, however, long-term survival is improved in the patients revascularized with CPB as compared with the off-pump cohort, suggesting possible advantages from a more complete revascularization in this population. (Ann Thorac Surg 2006;81:591 8) 2006 by The Society of Thoracic Surgeons Cardiac disease remains the most frequent cause of death for patients with end-stage renal disease (ESRD) on long-term dialysis. Forty-four percent of allcause mortality within this high-risk population can be directly related to cardiac disease, with approximately 20% of deaths attributed to acute myocardial infarction (AMI) [1]. Patients with ESRD sustaining AMI have notoriously poor long-term survival. The 2-year mortality after AMI between the years 1990 and 1995 was 74% [2]. In the year 2000, there were approximately 281,000 patients on dialysis in the United States, and current projections show that number reaching to more than Accepted for publication Aug 25, 2005. Presented at the Fiftieth Annual Meeting of the Southern Thoracic Surgical Association, Bonita Springs, FL, Nov 13 15, 2003. Address correspondence to Dr Dewey, 7777 Forest Lane, Suite A323, Dallas, TX 75230; e-mail: tdewey@csant.com. 500,000 patients by the year 2010 [1]. Increasingly these patients are referred for surgical revascularization owing to symptoms of angina or hemodynamic instability while on dialysis. Additionally, coronary artery disease may be identified in asymptomatic patients undergoing evaluation and work-up for renal transplantation. In approximately two thirds of patients with chronic renal failure, the inciting etiology is either diabetes mellitus or longstanding hypertension. As such, these patients frequently present with an aggressive atherosclerotic process characterized by diffuse coronary artery disease as well as a virulent vasculopathic condition involving the entire body. Numerous reports have described the increased mortality and morbidity faced by these patients as compared with nondialysis patients while undergoing coronary artery bypass graft surgery (CABG) [3 6]. However, little is known regarding whether off-pump CABG (OPCABG) 2006 by The Society of Thoracic Surgeons 0003-4975/06/$32.00 Published by Elsevier Inc doi:10.1016/j.athoracsur.2005.08.048

592 DEWEY ET AL Ann Thorac Surg CABG IN END-STAGE RENAL DISEASE 2006;81:591 8 can decrease this mortality, and provide long-term results equivalent with conventional CABG. Furthermore, to elucidate whether revascularization improves the survival of these high-risk patients, a comparison with a population of patients on dialysis with coronary artery disease but who have not had documented surgical revascularization or percutaneous coronary intervention must be performed. The goal of this retrospective study was to compare early and long-term results of coronary revascularization using the alternative techniques of conventional revascularization with cardiopulmonary bypass (CPB) and off-pump bypass grafting. Additionally, the long-term survival of our surgical patients was judged against a cohort of patients identified from the United States Renal Data System (USRDS) database as having coronary artery disease but no interventions, either surgical or catheter based, to determine whether the natural history of this population can be significantly altered by coronary revascularization. Material and Methods Data Set The Cardiopulmonary Research Science and Technology Institute (CRSTI) maintains a Society of Thoracic Surgeons (STS) approved database for an 18-hospital, 22- surgeon practice, collecting data on all patients undergoing cardiac surgery. Since 1985, data have been collected on more than 32,000 patients. The current data collection form contains more than 300 data elements prospectively recorded on all patients undergoing cardiovascular procedures. In 2000, the field previously named chronic obstructive pulmonary disease (COPD) was renamed as chronic lung disease (CLD). Whereas COPD was a binary response variable (yes/no), CLD is now an ordinal variable (none/mild/moderate/severe). To be able to combine the data before 2000 with the current data, the CLD value was collapsed into a binary yes/no value and has been named COPD also. Institutional Review Board approval was obtained to identify 174 patients who had renal failure and were on dialysis at the time of CABG between January 1, 1995, and July 31, 2003. Patient identification was strictly controlled in accordance with the Health Insurance Portability and Accountability Act of 1996 regulations. Sixteen patients were excluded from the study population for cardiogenic shock and active resuscitation at the time of CABG, and for procedures classified as salvage/ emergent in nature. Patient follow-up involved collecting data from the Social Security Death Index and phone calls to the patient or their family. Changes in status such as death or kidney transplant were noted and the date of the occurrence recorded; the data were then combined with the patient records from the STS database. The control data set was obtained from the USRDS 2002 core and hospital data sets. This data set, derived from Medicare claims part A (hospitalization) and part B (physician/provider), contains information on approximately 1.2 million patients with ESRD on dialysis. The accuracy of these data has previously been validated by the National Institutes of Health [7]. Using the master data set, patients having a diagnosis of coronary disease (atherosclerosis or myocardial infarction) were extracted. Patients whose records showed that they had undergone CABG, percutaneous coronary intervention, or kidney transplant were then excluded. The resulting data set consisted of 340,037 patients with date of death or last follow-up records. Data were current as of October 2001. Statistics Chi-square statistics were used to test for differences in categorical variables between patient groups. When small numbers of values were observed, Fisher s exact test was utilized. Continuous variables were tested using a Student t test, while differences in medians were compared using the Wilcoxon rank sum test. In all cases, statistical significance required a p value of 0.05 or less. Survival curves were calculated using the Kaplan- Meier technique, with the survival distribution function plotted against the length of follow-up. The measured endpoint was death; patients alive at last follow-up were included as right censored values. Curves were compared using log-rank statistics, emphasizing the longer follow-up results. Using the CRSTI dataset, a Cox proportional hazards model was used to assess the impact of independent predictors on patient survival over time. Results One hundred and fifty-eight patients were identified as meeting entrance criteria, with 59 having off-pump and 99 on-pump CABG. None of these patients underwent renal transplantation during the follow-up period. Despite the nonrandomized nature of this review, an analysis of 26 categorical preoperative risk factors showed few significant statistical differences between the off-pump and the onpump groups (Table 1). The on-pump group had a greater number of patients with a diagnosis of COPD, 30.3%, versus the off-pump group, 12.1% (p 0.009). Additionally, the on-pump cohort as compared with the off-pump group had a greater percentage of patients with previous myocardial infarctions: 61.6% versus 32.2% (p 0.001). Patients in the off-pump group had a greater incidence of preoperative inotrope use as compared with the on-pump patients: 6.8% versus 1.0% (p 0.05). There were no differences between the groups in regard to the percentage of women in each cohort, or the incidence of diabetes, hypertension, peripheral vascular disease, cerebral vascular accident, left main disease greater than 50%, or symptoms of congestive heart failure. Analysis of continuous variables showed that average age, ejection fraction, and the STS predicted risk of mortality also were not statistically different between the two groups (Table 1). The incidence of reported angina was equal in the on-pump group (88.9%) and off-pump group (86.4%). Of those 64 patients with unstable angina, however, 76.6% were operated on using CPB (p 0.003). The perioperative outcomes are noted in Table 2. The

Ann Thorac Surg DEWEY ET AL 2006;81:591 8 CABG IN END-STAGE RENAL DISEASE 593 Table 1. Preoperative Risk Factors Variable Off-Pump (59) On-Pump (99) Count % Count % Chi-Square p Value Female 29 49.2 34 34.3 ns (0.066) Current smoker 11 18.6 15 15.2 ns Diabetes mellitus 38 64.4 61 61.6 ns Angina 51 86.4 88 88.9 ns Unstable angina (of those with angina) 15 29.4 49 55.7 0.003 Hypertension 48 81.4 82 82.8 ns Cerebral vascular accident 9 15.3 12 12.1 ns Recent 1 11.1 1 8.3 ns Remote 8 88.9 11 91.7 Chronic obstructive pulmonary disease 7 12.1 30 30.3 0.009 Peripheral vascular disease 20 33.9 32 32.3 ns Previous coronary artery bypass graft 4 6.8 4 4.0 ns Myocardial infarction (MI) 19 32.2 61 61.6 0.001 Never 40 67.8 38 38.4 0.002 New ( 7 days) 7 11.9 20 20.2 Old ( 7 days) 12 20.3 41 41.4 Congestive heart failure 19 32.2 44 44.4 ns Inotropes 4 6.8 1 1.0 0.05 Left main disease 50% 16 27.1 27 27.3 ns Intra-aortic balloon pump 2 3.4 9 9.1 ns Intraoperative 0 0.0 3 33.3 ns Postoperative 0 0.0 1 11.1 Preoperative 2 100.0 5 55.6 Continuous variables (mean SD) Age 62.0 13.0 63.3 11.0 ns Ejection fraction (%) 46.9 14.9 46.8 12.7 ns STS predicted risk of mortality 0.093 0.074 0.091 0.077 ns ns not significant; STS The Society of Thoracic Surgeons. overall mortality was 18 of 158 (11.4%), with the on-pump group having a significantly increased operative mortality as compared with the off-pump cohort, 17.2% versus 1.7% (p 0.003). (Data were available on the cause of death for 15 of 18 patients: 7 [46.7%] were listed as cardiac; 2 each [13.3%] from infection, neurologic, and other causes; and 1 each [6.7%] in pulmonary and renal categories. For reference, we looked at our data on all 391 CABG patients who had renal disease, but were not on dialysis. In this group, 33 of 391 [8.4%] died perioperatively, with 8 (24.2%) listed as from a cardiac cause, 7 (21.2%) as neurologic, 6 (18.2%) as renal or other, and 3 (9.1%) as from infection or pulmonary causes.) Additionally, the on-pump patients were significantly more likely to have prolonged periods of ventilation postoperatively (namely, more than 24 hours), 28.3% versus 6.8% (p 0.001), and to require transfusion of blood products, 86.6% versus 49.2% (p 0.001). There were no differences between the two groups regarding the need for reoperation for bleeding, or the incidence of stroke, sternal wound infection, perioperative myocardial infarction, or readmission to the hospital. Additionally, no patients required conversion from an off-pump approach to CPB-supported revascularization. The patients having revascularization with CPB had significantly more grafts than the off-pump patients: 3.3 0.9 versus 2.4 1.0 (p 0.001). However, the off-pump patients had a greater number of arterial grafts than the on-pump bypass group: 0.9 0.5 versus 0.7 0.6 (p 0.03). While the on-pump group spent a greater number of days in the intensive care unit than the beating-heart patients, 5.2 6.7 versus 2.7 2.6 (p 0.002), the longer overall length of stay of 11.9 10.9 days for on-pump patients was not statistically significantly different from the 9.3 7.9 days observed for the off-pump patients. Logistic regression analysis demonstrated that female sex, need for a preoperative intra-aortic balloon pump, and use of cardiopulmonary bypass were independent risk factors for operative death (Table 3). Long-term follow-up data were then collected on all patients that survived the perioperative period by direct calls to the patient or their family, and by review of the Social Security Death Index. Mean follow-up times ranged from 39.1 months (median, 33.1) for the on-pump group to 18.3 months (median, 14.7) for the off-pump cohort. Survival curves were calculated using the Kaplan-Meier technique and plotting the survival distribution function against time of follow-up (Fig 1). A control group was identified from the

594 DEWEY ET AL Ann Thorac Surg CABG IN END-STAGE RENAL DISEASE 2006;81:591 8 Table 2. Perioperative Outcomes Variable Off-Pump (59) On-Pump (99) Count % Count % Chi-Square p Value Operative mortality 1 1.7 17 17.2 0.003 Reoperation for bleeding 0 0.0 2 2.0 ns Perioperative myocardial infarction 0 0.0 0 0.0 ns Deep sternal infection 1 1.7 1 1.0 ns Septicemia 0 0.0 5 5.1 ns Stroke, permanent 0 0.0 2 2.0 ns Stroke, transient 2 3.4 3 3.0 ns Prolonged ventilation 4 6.8 28 28.3 0.001 Pneumonia 2 3.4 5 5.1 ns Cardiac arrest 0 0.0 5 5.1 ns Tamponade 0 0.0 4 4.0 ns Gastrointestinal complications 1 1.7 5 5.1 ns Multisystem failure 0 0.0 3 3.0 ns Atrial fibrillation 8 13.6 19 19.2 ns Readmission to hospital 8 17.4 7 8.0 ns Blood products used 29 49.2 84 86.6 0.001 Continuous variables (mean SD) Length of stay, days 9.3 7.9 11.9 10.9 ns Median 7 8 0.04 Total number of grafts 2.4 1.0 3.3 0.9 0.001 Number of distal arteries 0.9 0.5 0.7 0.6 0.03 Number of distal veins 1.5 1.0 2.5 1.1 0.001 Intensive care unit, days 2.7 2.6 5.2 6.7 0.002 ns not significant. USRDS database consisting of dialysis patients with a diagnosis of coronary artery disease and no history of any coronary interventions or surgery. Survival in the on-pump group dropped rapidly during the first month, with a higher rate of perioperative mortality. The patients then showed better survival than the control or off-pump groups over the rest of the follow-up period. Off-pump patients showed the same survival as the USRDS controls for the first 10 to 12 months, then began dropping more rapidly. Comparison of the curves using log-rank statistics showed the differences between the offpump patients and the USRDS control group, as well as the on-pump to off-pump differences, to be statistically significant (p 0.03). (The early crossing of the curves that results from perioperative mortality may make the statistical probability less reliable.) Calculated over a 4-year period, this is an annual death rate of 38.1% for the off-pump group, 22.9% Table 3. Logistic Regression for Perioperative Mortality Variable Odds Ratio (95% Confidence Interval) p Value Sex (female/male) 4.4 (1.2 15.5) 0.02 Pump status (on/off-pump) 13.6 (1.7 109.8) 0.01 IABP use (Preoperative use/ never used) 17.6 (2.5 124.3) 0.01 IABP intra-aortic balloon pump. per year for the control group and 19.0% per year for the on-pump patients. Comparing just nondiabetic patients (Fig 2), the patients revascularized with CPB had a survival advantage over the control population (p 0.5, nonsignificant) and the offpump group (p 0.03). (The early crossing of the curves that results from perioperative mortality may make the statistical probability less reliable. After the curves cross, the on-pump data lie above the USRDS curve at all points.) The available data show the survival curve for the off-pump nondiabetic patients falling faster than either the on-pump patients (p 0.03) or the control diabetic patients (p 0.01). Comparing only patients with diabetes (Fig 3), the offpump diabetic patients are very close in survival to the control group, whereas the on-pump curve lies above the other two except for the short initial perioperative period. The differences are not statistically significant, with small numbers of patients in the on- and off-pump groups. A Cox proportional hazards model was used to identify variables affecting the rate of death. The risk of death increased with age (p 0.042; hazard ratio [HR] 1.02; 95% confidence limits [CL]: 1.00 to 1.04) and decreased with increasing body surface area (p 0.017, HR 0.35, 95% CL: 0.15 to 0.83); increasing ejection fraction also decreased the risk slightly (p 0.026; HR 0.98; 95% CL: 0.97 to 1.00). Pump as a factor had a hazard ratio of 0.64 (95% CL: 0.40 to 1.02) comparing on-pump to off-pump groups, indicating approximately a 35% reduction in risk

Ann Thorac Surg DEWEY ET AL 2006;81:591 8 CABG IN END-STAGE RENAL DISEASE 595 Fig 1. End-stage renal disease Kaplan-Meier survival curve for all patients. (USRDS United States Renal Data System.) when the patient is done on pump (p 0.058). Analysis of the data starting at 12 months shows the significance of the pump as a factor in long-term survival, with p 0.003 and a hazard ratio of 0.38 (95% CL: 0.20 to 0.71). To determine if the high perioperative mortality in the on-pump patients biased the survival comparison by eliminating the sickest patients from this group, we again contrasted the factors used to estimate preoperative risk for the patients in both cohorts who survived the perioperative period (Table 4). There were no significant differences between the two groups in regard to age, ejection fraction, incidence of diabetes or hypertension, or peripheral and cerebral vascular disease. Additionally, there was no statistical difference in the STS predicted risk of mortality between these two groups. Clinical differences between the groups included slightly more female patients remaining in the off-pump group than the on-pump, 48.3% versus 31.7% (p 0.05), and a higher frequency of COPD, 30.5% versus 12.3% (p 0.01), and previous myocardial infarction, 61% versus 32.8% (p 0.001), in the CPB patients. A Cox proportional hazards model was then used to identify independent predictors of death in these groups. The risk of death (excluding perioperative mortality) decreased 64% when the procedure was performed using CPB (p 0.001, relative risk [RR] 0.36, 95% confidence interval [CI]: 0.21 to 0.63). Additionally, female patients had a 44% decreased risk of death compared with males (p 0.03, RR 0.56, 95% CI: 0.33 to 0.96). Other factors associated with improved survival were increasing ejection fraction (p 0.001, RR 0.97, 95% CI: 0.95 to 0.99) and body surface area (p 0.006, RR 0.164, 95% CI: 0.057 to 0.474). Comment Coronary artery disease remains a major cause of death among patients with ESRD. In addition to coronary atherosclerosis, left ventricular hypertrophy, diastolic dysfunction, congestive heart failure, and complex ventricular arrhythmias leading to sudden death develop in patients with ESRD on dialysis [8]. The reported 2-year cumulative survival rate for patients with ESRD and congestive heart failure is as low as 33% [9]. ESRD patients also suffer from ventricular dysfunction caused by toxic waste byproducts [10] and diffuse coronary artery disease, the result of elevated calcium levels [11]. In this study, patients revascularized using CPB demonstrated a strikingly increased operative mortality compared with the off-pump patients, 17.2% versus 1.7%, despite Fig 2. End-stage renal disease Kaplan-Meier survival curve for nondiabetic patients. (USRDS United States Renal Data System.)

596 DEWEY ET AL Ann Thorac Surg CABG IN END-STAGE RENAL DISEASE 2006;81:591 8 Fig 3. End-stage renal disease Kaplan-Meier survival curve for diabetic patients. (USRDS United States Renal Data System.) having nearly the same STS predicted risk of mortality. The observed on-pump mortality rate falls within the wide range of previously reported mortality rates of 11.4% (Franga and colleagues [12]), 15% (Blum and coworkers [13]), and 20% (Rostand and coworkers [14] and Batiuk and associates [15]). Previous reports have shown a correlation between New York Heart Association class and operative mortality [13, 16], as well as a strong association between a recent preoperative myocardial infarction and operative death [15]. Our on-pump cohort had a higher frequency of new (less than 7 days) myocardial infarction than the off-pump group, 20.2% versus 11.9% (p 0.002), but an Table 4. Demographics of Patients Surviving the Perioperative Period Variable Off-Pump (58) On-Pump (82) Count % Count % Chi-Square p Value Female 28 48.3 26 31.7 0.05 Current smoker 10 17.2 13 15.9 ns Diabetes mellitus 37 63.8 50 61.0 ns Diabetes control Diet 1 2.7 1 2.0 ns Insulin 28 75.7 35 70.0 Oral 7 18.9 12 24.0 Hypertension 48 82.8 67 81.7 ns Cerebral vascular accident 8 13.8 11 13.4 ns Recent 1 12.5 1 9.1 ns Remote 7 87.5 10 90.9 Chronic obstructive pulmonary disease 7 12.8 25 30.4 0.01 Peripheral vascular disease 19 32.8 26 31.7 ns Previous coronary artery bypass graft 4 6.9 2 2.4 ns Myocardial infarction 19 32.8 50 61.0 0.001 Never 39 67.2 32 39.0 0.004 New ( 7 days) 7 12.1 16 19.5 Old ( 7 days) 12 20.7 34 41.5 Congestive heart failure 19 32.8 37 45.1 ns Inotropes 4 6.9 1 1.2 ns Left main disease 50% 16 27.6 18 22.0 ns Intra-aortic balloon pump 2 3.4 4 4.9 ns Intraoperative 0 0.0 2 50.0 ns Postoperative 0 0.0 0 0.0 Preoperative 2 100.0 2 50.0 Continuous variables (mean SD) Age 62.1 13.1 62.3 11.1 ns Ejection fraction (%) 46.8 15.0 46.3 12.4 ns STS predicted risk of mortality 0.093 0.075 0.078 0.059 ns ns not significant; STS The Society of Thoracic Surgeons.

Ann Thorac Surg DEWEY ET AL 2006;81:591 8 CABG IN END-STAGE RENAL DISEASE 597 equivalent percentage of patients with symptoms of congestive heart failure. Logistic regression analysis identified pump use as an independent predictor of mortality. Potential difficulties associated with the use of extracorporeal circulation include maintaining fluid and electrolyte balance, sustaining adequate hemoglobin levels, and restoring hemostasis at the conclusion of the procedure. These problems were realized in that our on-pump group required significantly more blood product administration and had longer postoperative ventilatory times than did the offpump cohort. The off-pump approach resulted in noticeably better outcomes in the perioperative period than did the use of CPB for revascularization. Similar results have been described in much smaller series when comparing off-pump to on-pump outcomes in dialysis-dependent patients and other high-risk populations [17, 18]. Remarkably, this perioperative mortality benefit did not translate into equivalent or better survival on long-term follow-up. For off-pump patients, longevity was significantly worse than for the on-pump patients who survived the perioperative period. In fact, the off-pump group did not demonstrate improved survival over a control group with coronary artery disease who did not have revascularization. A possible explanation may relate to there being significantly fewer grafts performed in the off-pump group, despite a considerable beating-heart surgery experience in our practice. Many reports in the literature comparing the two methods also demonstrate fewer grafts in off-pump than in on-pump patients. Whether the reduced number seen means incomplete revascularization and whether that leads to long-term sequelae have been a concern with off-pump bypass surgery. Early studies demonstrating decreased perioperative and long-term survival with incomplete revascularization were performed in patients with arrested hearts on CPB [22]. Despite this, most studies show improved perioperative results in off-pump compared with on-pump patients, especially for high-risk patients [19 21]. Long-term follow-up may suggest that any effect from the reduced graft numbers may be observed after the initial perioperative period. Improved perioperative survival of off-pump patients may be explained by the reduced ischemic burden placed on these patients. Because a period of global cardiac ischemia is induced in arrested-heart surgery, a greater penalty is paid, usually in the form of increased early mortality, for incomplete revascularization. Conversely, since off-pump bypass grafting causes only intermittent periods of regional ischemia, incomplete revascularization may be better tolerated and not show any consequences in the perioperative period. However, long-term survival and freedom from repeat surgical or percutaneous intervention would continue to be adversely affected. The use of arterial grafts for revascularization has been shown to increase survival in both dialysis and nondialysis patients [23, 24]. Additionally, internal mammary artery patency has been found to be superior to saphenous vein in patients with ESRD owing to its resistance to atherosclerotic change [25]. In this study, the off-pump patients demonstrated worse long-term survival even though more arterial grafts were used per patient than in the on-pump group. Once more, fewer grafts overall may play a role in the survival of the patients in the beating-heart group. It is recognized that diabetes mellitus is an independent predictor of adverse long-term outcomes in patients undergoing CABG [19, 26]. Subgroup analysis of our data showed that the long-term survival of off-pump patients was still worse than that of on-pump patients or controls, although the effect was largest in the nondiabetic group. This association was also confirmed by the Cox proportional hazard model, which showed that the use of CPB was a predictor of survival. To determine if the sickest or weakest patients were dying during the perioperative period, we looked at those surviving beyond the operative mortality period. The off-pump group only decreased by 1 patient while 17 on-pump patients had died. In the remaining patients, there was no significant change in the risk factors, with the exception of a decrease in the STS predicted risk of mortality for the on-pump survivors that dropped by about 11%. Statistically, it was not different from the off-pump value, but may suggest that the stress of surgery was worse when done on pump. This study remains limited in that no information was collected regarding the duration of dialysis before CABG, which could affect the survival curves if a disproportionate number of patients with a long history of dialysis were in one group. Additionally, no assessment regarding the quality of the target vessels, vessel size, or lesion characteristics were collected on the STS database. We conclude that off-pump revascularization in patients with ESRD on hemodialysis demonstrates superior perioperative morbidity and mortality compared with conventional surgery with CPB. However, long-term survival is significantly better in the on-pump patients, possibly because of incomplete revascularization in the off-pump cohort. Surgical revascularization using CPB does appear to increase the life expectancy of patients with ESRD and coronary artery disease compared with patients who had no interventions. The increased operative mortality among these patients emphasizes the importance of patient selection and surgical approach when offering CABG to this high-risk group. Finally, these findings stress the need for follow-up of all patients who have undergone off-pump bypass grafting to ensure that the benefits seen in the perioperative period, translate into long-term results equivalent to conventional revascularization. The data used to create a control population for this manuscript were supplied by the United States Renal Data System, a section of the National Institutes of Health. The interpretation and reporting of these data were the responsibility of the authors, and in no way should be seen as an official policy or interpretation of the United States government. References 1. National Institutes of Health, USRDS 1999 annual data report, Bethesda, MD, p. 92.

598 DEWEY ET AL Ann Thorac Surg CABG IN END-STAGE RENAL DISEASE 2006;81:591 8 2. Herzog CA, Ma JZ, Collins AJ. Poor long-term survival after acute myocardial infarction among patients on long-term dialysis. N Engl J Med 1998;339:799 805. 3. Dacey LJ, Liu JY, Braxton JH, et al. Long-term survival of dialysis patients after coronary bypass grafting. Ann Thorac Surg 2002;74:458 63. 4. Horst M, Mehlhorn U, Hoerstrup SP, et al. Cardiac surgery in patients with end-stage renal disease: 10-year experience. Ann Thorac Surg 2000;69:96 101. 5. Agirbasli M, Weintraub WS, Chang GL, et al. Outcome of coronary revascularization in patients on renal dialysis. Am J Cardiol 2000;86:395 9. 6. Labrousse L, de Vincentiis C, Madonna F, et al. Early and long-term results of coronary artery bypass grafts in patients with dialysis dependent renal failure. Eur J Cardiothorac Surg 1999;15:691 6. 7. National Institutes of Health. USRDS 1992 annual data report. Bethesda, MD: National Institutes of Health, Publication No. 92-3176, 1992:61 82. 8. Parfrey PS, Harnett JD, Barre PE, et al. The natural history of myocardial disease in dialysis patients. J Am Soc Nephrol 1991;2:2 12. 9. Parfrey PS, Griffiths SM, Harnett JD, et al. Outcome of congestive heart failure, dilated cardiomyopathy, hypertrophic hyperkinetic disease, and ischemic heart disease in dialysis patients. Am J Nephrol 1990;10:213 21. 10. Foley RN, Parfrey PS. Cardiac disease in chronic uremia: clinical outcomes and risk factors. Adv Ren Replace Ther 1997;4:234 48. 11. Jahangiri M, Wight J, Edmonson S, et al. Coronary artery bypass graft surgery in dialysis patients. Heart 1997;78:343 5. 12. Franga DL, Kratz JM, Crumbley AJ, et al. Early and long-term results of coronary artery bypass grafting in dialysis patients. Ann Thorac Surg 2000;70:813 9. 13. Blum U, Skupin M, Wagner R, et al. Early and long-term results of cardiac surgery in dialysis patients. Cardiovasc Surg 1994;2: 97 100. 14. Rostand SG, Kirk KA, Rutsky EA, et al. Results of coronary artery bypass grafting in end-stage renal disease. Am J Kidney Dis 1988;12:266 70. 15. Batiuk TD, Kurtz SB, Oh JK, et al. Coronary artery bypass operation in dialysis patients. Mayo Clin Proc 1991;66:45 53. 16. Kaul TK, Fields BL, Reddy MA, et al. Cardiac operations in patients with end-stage renal disease. Ann Thorac Surg 1994; 57:691 6. 17. Hirose H, Amano A, Takahashi A. Efficacy of off-pump coronary artery bypass grafting for the patients on chronic hemodialysis. Jp J Thorac Cardiovasc Surg 2001;49:693 9. 18. Tashiro T, Nakamura K, Morishige N, et al. Off-pump coronary artery bypass grafting in patients with end-stage renal disease on hemodialysis. J Card Surg 2002;17:377 82. 19. Magee NJ, Dewey TM, Acuff TA, et al. Influence of diabetes on mortality and morbidity: off-pump coronary artery bypass grafting versus coronary artery bypass grafting with cardiopulmonary bypass. Ann Thorac Surg 2001;71:776 81. 20. Perrault LP, Menasche P, Peynet J, et al. On-pump, beating heart coronary operations in high-risk patients: an acceptable trade-off? Ann Thorac Surg 1997;64:1368 73. 21. Magee MJ, Jablonski KA, Stamou SC, et al. Elimination of cardiopulmonary bypass improves early survival for multivessel coronary artery bypass patients Ann Thorac Surg 2002;73: 1196 203. 22. Cosgrove DM, Loop FD, Lytle BW, et al. Primary myocardial revascularization. Trends in surgical mortality. J Thorac Cardiovasc Surg 1984;88:673 84. 23. Loop FD, Lytle BW, Cosgrove DM, et al. Influence of the internal mammary artery on 10-year survival and other cardiac events. N Engl J Med 1986;316:1 6. 24. Blakeman BP, Sullivan HJ, Foy BK, et al. Internal mammary artery revascularization in the patient on long-term renal dialysis. Ann Thorac Surg 1990;50:776 8. 25. Ura M, Sakata R, Nakayama Y, et al. The impact of chronic renal failure on athersclerosis of the internal thoracic arteries. Ann Thorac Surg 2001;71:148 51. 26. Chychota NN, Gau GT, Pluth JR, et al. Myocardial revascularization. Comparison of operability and surgical results in diabetic and nondiabetic patients. J Thorac Cardiovasc Surg 1973; 65:856 62. DISCUSSION DR KEVIN D. ACCOLA (Orlando, FL): In your conclusions you differentiated between off-pump and on-pump results. I was curious how you selected which patients you were going to do off pump and which patients you were going to do on pump. Was this merely a surgical decision at the time of the procedure or do you have some type of decision-making process with this? DR DEWEY: Much of the selection was surgeon bias. Up until the late 1990s, most of these patients were done on bypass, but since approximately mid 1998, our approach has been to do off-pump bypass surgery because we felt that perioperative morbidity and mortality were reduced with that approach. DR ALAA Y. AFIFI (Albany, NY): I enjoyed your paper. These are certainly a very sick group of people. You have made it clear that in the long term, people with complete revascularization do better than those who have had incomplete revascularization. You have demonstrated that you have better results in your on-pump group versus the off-pump group. The big concern, though, is why is there a 10-fold increment in operative mortality in patients that you do on pump versus off pump? We certainly have seen similar concerns with some of these very sick patients. The way we have handled them on many occasions, is by going on pump, cooling them down to 25 C, and giving one dose of cold blood high potassium cardioplegia, and then throughout the remainder of the procedure we just give cold blood in a retrograde manner with a goal of performing complete revascularization. So the question becomes, how do you approach myocardial protection while you are on pump? And, once again, why do you think you have a 10-fold mortality difference in your on-pump versus off-pump group? Thank you. DR DEWEY: I think renal failure patients are a difficult group in which to perform bypass grafting, and the problems associated with cardiopulmonary bypass are primarily electrolyte or fluid in nature. Additionally, one can have problems maintaining hemoglobin levels because these patients typically are anemic. Furthermore, there can be problems with hemostasis as evidenced by greater need for transfusions not only red cells but also platelets and other clotting factors to control bleeding. Some of the perioperative deaths can be attributed to multiorgan failure, secondary to volume overload, leading to increased ventilator time, and some of the deaths are due to delayed tamponade in the intensive care unit. The problem with doing these patients on bypass is that the vasodilatory effects of the pump are vastly accentuated over patients who do not have renal disease. Our standard on-pump approach is to use blood cardioplegia, and dialyze these patients intraoperatively, then not dialyze them again until either the first or second postoperative day.