Neurological Outcomes in Coronary Surgery: Independent Effect of Avoiding Cardiopulmonary Bypass

Transcription

1 Neurological Outcomes in Coronary Surgery: Indeendent Effect of Avoiding Cardioulmonary Byass Nirav C. Patel, FRCS(C-Th), Anand P. Deodhar, MCh, Antony D. Grayson, BS, D. Mark Pullan, FRCS(C-Th), Daniel J. M. Keenan, FRCS, Ragheb Hasan, FRCS(C-Th), and Brian M. Fabri, MD, FRCS(Ed) Deartments of Cardiothoracic Surgery and Research and Develoment, The Cardiothoracic Centre Liverool, Liverool, and the Deartment of Cardiothoracic Surgery, Manchester Royal Infirmary, Manchester, United Kingdom Background. Recent studies examining neurorotective effects of off-um coronary artery byass grafting (CABG) have shown inconsistent results. We examined our database to quantify the indeendent effects of avoidance of cardioulmonary byass (CPB) and aortic maniulation on neurologic outcomes after CABG. Methods. A total of 2,327 consecutive cases undergoing isolated CABG between Aril 1997 and May 2001 were identified at our two institutions. Patients were divided into three grous: on CPB, off-um with aortic maniulation, and off-um without aortic maniulation. To control for the confounding effects of other risk factors, we erformed a multivariate logistic regression analysis. Potential covariates considered in the logistic model included age, sex, redo oerations, diabetes, chronic obstructive ulmonary disease, neurologic disease, eriheral vascular disease, ejection fraction, and riority of oeration. Results. A total of 1,210 cases were erformed on CPB, comared with 520 off-um with aortic maniulation, and 597 off-um without aortic maniulation. The incidence of focal neurologic deficit was 1.6% (n 19) in the on-um grou, 0.4% (n 2) in the off-um with aortic maniulation grou, and 0.5% (n 3) for the off-um without aortic maniulation grou ( for trend 0.027). The results of the multivariate logistic regression analysis demonstrated that use of CPB was a risk factor for focal neurologic deficit, with an odds ratio of 3.82 (95% confidence interval, 1.41 to 10.34; 0.005). Aortic maniulation did not significantly influence neurologic outcome in off-um atients. Conclusions. Off-um oeration, with or without aortic maniulation, reduces adverse neurologic outcomes comared with on-um rocedures. (Ann Thorac Surg 2002;74:400 6) 2002 by The Society of Thoracic Surgeons Presented at the Thirty-eighth Annual Meeting of The Society of Thoracic Surgeons, Fort Lauderdale, FL, Jan 28 30, Address rerint requests to Dr Fabri, The Cardiothoracic Centre Liverool, Thomas Drive, Liverool L14 3PE, United Kingdom; bfabr@ccl-tr.nwest.nhs.uk. Neurologic deficits after coronary artery byass grafting (CABG) still reresent the most devastating comlication. The mechanisms contributing to neurologic comlications are multifactorial. Hemodynamic fluctuations, cerebral embolization (atherosclerotic laque, air, fat, and latelet aggregates), cerebral hyerthermia, and other inflammatory and neurohumoral derangements associated with surgery have all been imlicated [1 3]. Recent studies examining the cerebrorotective effects of beating heart coronary revascularization versus conventional revascularization on cardioulmonary byass (CPB) have yielded inconsistent results [4 6]. Many of these reorts involved small numbers and did not differentiate between conventional roximal anastomoses on the ascending aorta (aortic maniulation) or the no touch aorta (no aortic maniulation) technique. We aimed to quantify the indeendent effect of avoidance of CPB and aortic maniulation on neurologic outcomes after CABG. Material and Methods Patient Poulation and Data A total of 2,327 consecutive atients undergoing CABG between Aril 1, 1997, and May 31, 2001, were identified from the Cardiac Surgery Databases at our two institutions (The Cardiothoracic Center Liverool and Manchester Royal Infirmary). Patients undergoing CABG that was incidental to heart valve reair or relacement, resection of a ventricular aneurysm, or other surgical rocedure were not included. These atients reresented the entire coronary revascularization ractice of 4 surgeons (2 surgeons from each institution: D.K., R.H., D.M.P., B.M.F.). All data were collected rosectively during the atient admission and entered into a Cardiac Surgery Database as art of routine clinical ractice. Methods of data collection and definitions have been ublished reviously [7]. Data were collected on the following variables: age, sex, body mass index, urgency of oeration, 2002 by The Society of Thoracic Surgeons /02/$22.00 Published by Elsevier Science Inc PII S (02)

2 Ann Thorac Surg PATEL ET AL 2002;74:400 6 EFFECT OF AVOIDING CARDIOPULMONARY BYPASS 401 rior cardiac oeration, angina class, history of myocardial infarction, smoking, diabetes, hyercholesterolemia, hyertension, eriheral vascular disease, cerebrovascular disease, resiratory disease, renal dysfunction, as well as the extent of coronary disease and degree of left ventricular ejection fraction. The main outcome measure for this study, focal neurologic deficit, was defined as a new focal neurologic deficit or a comatose state occurring ostoeratively, that ersisted for more than 24 hours after onset and was noted before discharge or death. To avoid subjective bias, we did not include transient neurologic events, confusional states, or intellectual imairment. Resident and consultant medical staff made the diagnosis of focal neurologic deficits. In-hosital mortality was defined as death within the same hosital admission regardless of cause. All atients transferred from the base hosital to another hosital were followed u to confirm their status at discharge. Surgical Technique Patients were divided into two surgical grous: on-um and off-um. The off-um grou was further subdivided into atients with and without aortic maniulation. All off-um atients had a median sternotomy aroach. In the off-um without aortic maniulation grou, the in situ left internal mammary artery was used in all atients who needed grafting to the left coronary arterial system. The radial or right internal mammary arteries were used as a comosite T-graft on the in situ left internal mammary artery to avoid roximal anastomoses on the aorta. Sequential grafting was used to achieve comlete revascularization. In the off-um grou with aortic maniulation, in situ internal mammary artery grafts were used as aroriate and additional grafts were reared with long sahenous veins, radial artery, or free internal mammary artery as required. Proximal anastomoses were constructed on the ascending aorta with the alication of a side-biting clam. The clam was alied after lowering the arterial ressure to 80 mm Hg. In both off-um grous the target coronary artery was stabilized by using the Octous II tissue stabilization system (Medtronic, Minneaolis, MN). When aroriate, temorary roximal control of the target coronary artery was achieved with a 4-0 Prolene (Ethicon, Somerville, NJ) suture buttressed with Teflon (DuPont, Parkersburg, WV). After making a suitable arteriotomy, an aroriately sized intracoronary shunt (Medtronic Clearview shunt; Medtronic) was introduced in almost all cases to maintain distal erfusion and to achieve a relatively bloodless field. Visualization was aided with a humidified carbon dioxide blower (CTS Aires CO 2 Blower; Cardiothoracic Systems, Cuertino, CA). Hyothermia was revented by alication of a forced air blanket to the lower half of the body. In the on-um grou, standard CPB techniques were used with a median sternotomy aroach. Hollow fiber or flat membrane oxygenators were used according to institutional and consultant reference. The same was true for the use of roller or centrifugal um heads. All the atients had the in situ left internal mammary artery used along with sahenous vein grafts as aroriate. On comletion of all distal anastomoses, the aortic crossclam was removed and the roximal anastomoses were constructed with the single alication of a side-biting clam. Normothermia or moderate hyothermia (34 to 36 C) was maintained and myocardial rotection was achieved with antegrade induction of blood cardiolegia followed by continuous or intermittent retrograde blood cardiolegia. Statistical Methods Continuous variables are shown as median with 25th and 75th centiles and categorical variables are shown as a ercentage with 95% confidence intervals (CI). Comarisons were made with Kruskal-Wallis tests and 2 tests as aroriate. The EuroSCORE was derived to assess differences in atient case mix between the three surgical techniques [8]. To control for confounding variables, we used multivariable logistic regression analysis to examine the effect of CPB and aortic maniulation on ostoerative neurologic deficit. Forward stewise selection was used to identify significant risk factors. Potential confounding factors offered to the logistic model included atient age, unstable angina, history of diabetes, rior CABG, history of vascular disease, history of ulmonary disease, and history of neurologic disease. These factors have been described as imortant determinants of neurologic deficit by Newman and colleagues [9]. Also offered to the logistic model were any significant or closely associated ( 0.1) univariate risk factors for ostoerative neurologic deficit from our own exerience, along with the three surgical techniques. To control for treatment selection bias we obtained a roensity score, which was the robability that a atient would undergo CABG without CPB. We used all the variables listed in Table 1 to construct roensity score. This roensity score was included as a covariate in the multivariable logistic model, with the goal of adjusting for treatment selection bias (eg, emergency oeration, extent of disease). This method of adjustment is referable when the number of events is small [10]. The C statistic and the Lemeshow-Hosmer goodness of fit statistic were calculated to assess the erformance and calibration of the models, resectively [11]. In all cases a value less than 0.05 was considered significant. All statistical analysis was erformed with SAS for Windows Version 8 (SAS Institute, Cary, NC). Results Cardioulmonary byass was used in 1,210 (52%) atients. Off-um oeration was used in 1,117 (48%) atients (520 [22.3%] with aortic maniulation and 597 [25.7%] without aortic maniulation). No differences were noted between off-um and on-um atients according to age, sex, angina class, revious myocardial infarction, history of diabetes, eriheral vascular disease, renal dysfunction, resiratory disease, cerebrovascular disease, rior cardiac oeration,

3 402 PATEL ET AL Ann Thorac Surg EFFECT OF AVOIDING CARDIOPULMONARY BYPASS 2002;74:400 6 Table 1. Patient and Disease Characteristics Characteristic On-Pum (n 1,210) Off-Pum With Aortic (n 520) Off-Pum Without Aortic (n 597) (for trend) Age (y) at oeration (range) 62 (55 68) 63 (55 69) 61 (55 68) Female sex (%) Body mass index (kg/m 2 ) 27 (25 30) 27 (25 31) 28 ( ) (range) Previous MI (%) Current smoker (%) Diabetes (%) Hyercholesterolemia (%) Hyertension (%) Periheral vascular disease (%) Cerebrovascular disease (%) Renal dysfunction (%) Resiratory disease (%) Ejection fraction 0.30 (%) Three-vessel disease (%) Left main stenosis (%) Prior cardiac oeration (%) Emergent rocedure (%) EuroSCORE (range) 2 (1 4) 2 (1 4) 2 (1 4) Continuous variables are shown as median with 25th and 75th centiles. Categorical variables are shown as a ercentage. EuroSCORE, Euroean System for Cardiac Oerative Risk Evaluation; MI, myocardial infarction. reoerative intraaortic balloon um, ventilation suort, and left ventricular ejection fraction. Off-um atients were less likely to undergo emergency oeration (0.9% [95% CI, 0.5 to 1.7] versus 3.1% [95% CI, 2.3 to 4.3]; 0.001) and had a lower number of diseased coronary vessels (62.1% three-vessel disease [95% CI, 59.2 to 64.9] versus 79.1% three-vessel disease [95% CI, 76.7 to 81.3]; 0.001) than on-um atients. Additionally, off-um atients were more likely to be current smokers (24.9% [95% CI, 22.5 to 27.6] versus 14.7% [95% CI, 12.8 to 16.9]; 0.001), hyertensive (48.8% [95% CI, 45.8 to 51.8] versus 40.3% [95% CI, 37.6 to 43.2]; less than 0.001), hyercholesterolemic (80.6% [95% CI, 77.9 to 83.1] versus 73.3% [95% CI, 70.7 to 75.8]; 0.001), and had a higher body mass index (28 kg/m 2 [25th and 75th centiles, 25 to 31] versus 27 kg/m 2 [25th and 75th centiles, 25 to 30]; 0.008). The median number of distal coronary anastamoses were 3 (25th and 75th centiles, 2 to 4) and 4 (25th and 75th centiles, 3 to 5) for the off-um and on-um grous, resectively ( 0.001). Table 1 lists atient and disease characteristics for the three surgical techniques. The overall risk score (Euro- SCORE) was similar for the three different grous. The incidence of in-hosital mortality was 2.5% (95% CI, 1.7 to 3.6; n 30) in the on-um grou, 1.0% (95% CI, 0.4 to 2.4; n 5) in the off-um with aortic maniulation grou, and 1.5% (95% CI, 0.7 to 2.9; n 9) in the off-um without aortic maniulation grou ( for trend 0.076). The median ostoerative length of stay was 7 days (25th and 75th centiles, 6 to 8), 6 days (25th and 75th centiles, 5 to 7), and 6 days (25th and 75th centiles, 5 to 7), resectively ( 0.001). The incidence of focal neurologic deficit was 1.6% (95% CI, 1.0 to 2.5; n 19) in the on-um atients, 0.4% (95% CI, 0.1 to 1.5; n 2) in the off-um with aortic maniulation atients, and 0.5% (95% CI, 0.1 to 1.6; n 3) in the off-um without aortic maniulation atients ( for trend 0.027). Comared with the off-um without aortic maniulation grou, univariate odds of focal neurologic deficit were 0.76 (95% CI, 0.13 to 4.59; 0.768) and 3.16 (95% CI, 0.93 to 10.72; 0.051) for the off-um with aortic maniulation and on-um grous, resectively. The crude odds ratio for focal neurologic deficit (onum versus off-um) was 3.55 (95% CI, 1.32 to 9.53; 0.007). The details of the tyes of neurologic deficits are listed in Table 2. No significant differences in outcomes were found between the two institutions or the 4 consultant surgeons involved in the study. Table 3 shows the reoerative risk factors for focal neurologic deficit identified by univariate analysis. These risk factors were added to the logistic model along with the risk factors identified by Newman and coworkers [9]. The three surgical techniques were also added as covariates. Multivariate logistic regression analysis identified the use of CPB as an indeendent risk factor for focal neurologic deficit, with an odds ratio of 3.82 (95% CI, 1.41 to 10.34; 0.005). Other indeendent risk factors were eriheral vascular disease and atient age (Table 4). Aortic maniulation did not significantly influence neurologic outcome in off-um atients. Table 5 shows the multivariable logistic regression

4 Ann Thorac Surg PATEL ET AL 2002;74:400 6 EFFECT OF AVOIDING CARDIOPULMONARY BYPASS 403 Table 2. Tyes of Neurologic Deficits Deficit analysis adjusted with the roensity score. The roensity score was not statistically significant in the multivariable logistic model. This occurrence rovides evidence that the adjustment for the treatment selection bias was effective [10]. Comment On-Pum (n 19) Off-Pum With Aortic (n 2) Off-Pum Without Aortic (n 3) Ahasia Loss of vision Monoaresis Hemiaresis Comatose state Table 3. Univariate Preoerative Risk Factors for Focal Neurologic Deficit Risk Factor Odds Ratio 95% Confidence Intervals Age at oeration (years) a Female sex Body mass index 30 kg/m Angina class IV Previous MI Current smoker Diabetes Hyercholesterolemia Hyertension Periheral vascular disease Cerebrovascular disease Renal dysfunction Resiratory disease Ejection fraction Three-vessel disease Left main stenosis Prior cardiac oeration Emergent rocedure a For each additional year. MI, myocardial infarction. Table 4. Indeendent Risk Factors for Focal Neurologic Deficit Risk Factor Odds Ratio 95% Confidence Intervals Periheral vascular disease Age at oeration (years) a Cardioulmonary byass C 0.80, Lemeshow-Hosmer goodness of fit a For each additional year. Adverse neurologic events after isolated CABG have remained constant desite various attemts to reduce their incidence. In a large-scale multicenter study, Roach and colleagues [12] documented a 6.1% incidence of serious adverse neurologic events in a survey of 2,108 atients undergoing isolated CABG. Three ercent of these atients had a erioerative stroke, whereas a further 3.1% had rolonged unconsciousness, seizures, or encehaloathy. Advanced age and duration of CPB were the strongest correlating factors for neurologic comlications. In another large rosective study, Newman and colleagues [9] observed that 3.2% of their atients after isolated CABG had unfavorable neurologic consequences. In their model they identified age, history of neurologic disease, diabetes, history of eriheral vascular disease, redo CABG, and unstable angina as major correlating reoerative variables for adverse neurologic outcomes. The mechanisms that contribute to oor neurologic outcome have been well documented. Cerebral embolization with macro- and microemboli has been shown to be the most common mechanism involved [13, 14]. Using transcranial Doler, Stum and colleagues [15] clearly demonstrated that emboli could be detected during instrumentation of the aorta and heart. With the introduction of off-um CABG, there exists a otential to reduce the incidence of adverse neurologic events. Off-um coronary artery byass rocedures do not use aortic cannulation and cross-claming and thereby avoid injury to the aorta and dislodgement of any atheroma during byass. Off-um CABG also eliminates the generation of microgaseous and microarticulate emboli from the byass circuit, thereby reducing the embolic load. However, the otential exists for an increased risk of injury to the aortic wall during artial claming of a tense aorta, esecially if the aorta is diseased. Recent reorts on the neurorotective effects of offum CABG have revealed inconsistent results. Several authors have reorted imroved neurologic outcomes in off-um coronary oerations when comared with onum techniques [5, 6, 16, 17]. However, others have not reorted imroved neurologic results when comaring off-um versus on-um atients [4, 18 20]. The reort of Ricci and colleagues [16] showed a significant reduction in cerebrovascular strokes in octogenarians under- Table 5. Indeendent Risk Factors for Focal Neurological Deficit With Proensity Score Risk Factor Odds Ratio a 95% Confidence Intervals Periheral vascular disease Age at oeration (years) b Cardioulmonary byass C 0.80, Lemeshow-Hosmer goodness of fit a Adjusted for the roensity score (odds ratio, 1.01; 95% CI, ; 0.744). b For each additional year.

5 404 PATEL ET AL Ann Thorac Surg EFFECT OF AVOIDING CARDIOPULMONARY BYPASS 2002;74:400 6 going off-um oeration. However, the on-um grou was not comarable to the off-um grou in terms of major risk factors. Other reorts showed a significant reduction in neurocognitive imairment in off-um atients in small randomized controlled trials [5, 6]. Cleveland and coworkers [17], in a large multicenter, retrosective analysis involving 126 centers throughout the United States, showed a reduction in the observedto-exected ratio of stroke from multivariate analysis of off-um atients comared with on-um CABG (0.62 versus 1.05). None of the above-mentioned reorts assessed avoiding CPB as a factor indeendent of aortic maniulation in their analyses. Our exerience includes a large number of atients with comarable grous. Inevitably all reoerative variables were not evenly distributed between the three grous. However, key variables influencing major outcomes were evenly distributed between the grous leading to identical EuroSCOREs. EuroSCORE has already been reorted as an accurate risk model for redicting mortality for isolated CABG in western Euroe [8]. Moreover, the key variables affecting neurologic outcomes as demonstrated by Newman and colleagues [9] were evenly distributed between the three grous. As has been the case in reorts with large numbers of atients [17], the in-hosital mortality in our exerience was lower in the off-um grou than in the on-um grou. We also found a significantly shorter hosital stay in the off-um grou, as did other researchers [4, 17 20]. We chose as our main outcome the incidence of fixed focal neurologic deficits. As with other reorts [5, 6, 16, 17], the crude incidence of neurologic deficits in our exerience was lower in the off-um grou. The reoerative risk factors identified by univariate analysis (age, hyertension, history of eriheral vascular disease, history of cerebrovascular disease, resiratory disease, emergency rocedure) were also similar to all the major reorts [10, 12, 21, 22]. As with other reorts [10, 12, 21, 22], we found age at oeration was an indeendent risk factor for adverse neurologic outcome. We also identified a history of eriheral vascular disease as an indeendent redictor, but a rior history of cerebrovascular disease was not redictive. In contrast, Roach and coworkers [12] found a history of cerebrovascular disease to be an indeendent redictor of adverse neurologic events after CABG but not the history of eriheral vascular disease, an exerience shared by others [21, 22]. These reorts, however, did not include off-um CABG atients in their study. The most significant finding in our study was that CPB is an indeendent risk factor for adverse neurologic outcomes. Bowles and colleagues [23] reorted a significant reduction in cerebral microemboli in the off-um grou comared with the on-um grou. More imortantly, by using transcranial Doler studies, they illustrated that although surgical maniulation does account for some emboli during CABG, the vast majority of emboli (84%) occurred while the atient was receiving CPB with no surgical maniulation identified [23]. Stum and Newman [15] showed that the embolic load during aortic cannulation/decannulation or cross-clam alication/removal is similar to the load seen when alying and removing the side-biting clam for roximal anastomoses. These two studies suort our finding that avoiding CPB is more imortant than avoidance of aortic maniulation for constructing roximal anastomoses. The retrosective nature is the main limitation of this study, as retrosective reviews inherently have many confounding variables and ossible selection bias. We also did not undertake detailed neurologic and neurosychiatric assessment of the atients. The academic urists would dictate that the best way to rovide conclusive evidence would be to conduct a large-scale, multicenter, rosective randomized trial. Many are of the oinion that such an endeavor is unlikely to haen. Recently, Blackstone [24] highlighted a helful breakthrough in statistical methods that may adjust for selection factors and allow causal inferences to be made from unrandomized clinical exeriences. He suggested the use of roensity scoring to adjust for selection factors [10]. We used roensity score to adjust the odds ratios obtained from multivariable logistic regression analysis. Even after adjustment with roensity score, the use of CPB remains a significant variable for adverse neurologic outcomes. Another limitation is the low event rate. With only 24 focal neurologic deficits recorded (1.03%), this low number may not be sufficient for an accurate rediction [25]. In conclusion, we have shown that off-um CABG, with or without aortic maniulation, reduces adverse neurologic outcomes when comared with conventional CABG erformed on CPB. We thank, for their considerable efforts, Suzanne Chaisty and Janet Deane, who maintain the quality and ensure comleteness of data collected in our Cardiac Surgery Registry. References 1. Mora CT, Murkin JM. The central nervous system: resonses to cardioulmonary byass. In Mora CT, ed. Cardioulmonary byass: rinciles and techniques of extracororeal circulation. New York: Sringer-Verlag, 1995: Hersokowitz A, Mangano DT. The inflammatory cascade: a final common athway for erioerative injury? Anesthesiology 1996;85: Barbut D, Gold J. Aortic atheromatosis and risks of cerebral embolisation. J Cardiothor Vasc Anesth 1996;10: Arom KV, Flavin TF, Emery RW, et al. Safety and efficacy of off-um coronary artery byass grafting. Ann Thorac Surg 2000;69: Deigler A, Hirsch R. Neuromonitoring and neurocognitive outcome in off-um versus conventional coronary byass oeration. Ann Thorac Surg 2000;68: Murkin JM, Boyd WD, Ganaathy S, et al. Beating heart surgery: why exect less central nervous system morbidity? Ann Thorac Surg 1999;68: Wynne-Jones K, Jackson M, Grotte G, et al. Limitations of the Parsonnet score for measuring risk stratified mortality in the north west of England. Heart 2000;84: Nashef SAM, Roques F, Michel P, et al. The EuroSCORE study grou. Euroean system for cardiac oerative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 1999;16: 9 13.

6 Ann Thorac Surg PATEL ET AL 2002;74:400 6 EFFECT OF AVOIDING CARDIOPULMONARY BYPASS Newman MF, Wolman R, Kanchuger M, et al. Multicenter reoerative stroke risk index for atients undergoing coronary artery byass graft surgery. Circulation 1996;94(sul 2):II Blackstone EH. Comaring ales and oranges. J Thorac Cardiovas Surg 2002;123: Hosmer D, Lemeshow S. Alied logistic regression. New York: Wiley; Roach GW, Kanchuger M, Mangano CM, et al. Adverse cerebral outcomes after coronary byass surgery. N Engl J Med 1996;335: Blauth CI. Macroemboli and microemboli during cardioulmonary byass. Ann Thorac Surg 1995;59: Pugsley W, Klinger L, Paschalis C, Treasure T, Harrison M, Newman DP. The imact of microemboli during cardioulmonary byass and neurosychological functioning. Stroke 1994;25: Stum DA, Newman SP. Emboli detection during cardioulmonary byass. In Tegler CH, Babikian VL, Gomez CT, eds. Neurosonology. St. Louis, MO: Mosby, 1996: Ricci M, Karamanoukain HL, Abraham R, et al. Stroke in octogenarians undergoing coronary artery surgery with and without cardioulmonary byass. Ann Thorac Surg 2000;69: Cleveland JC, Shroyer AL, Chen AY, Peterson E, Grover FL. Off-um coronary artery byass grafting decreases riskadjusted mortality and morbidity. Ann Thorac Surg 2001;72: Iaco AL, Contini M, Teodori G, et al. Off or on byass: what is the safety threshold? Ann Thorac Surg 1999;68: Lancey RA, Soller BR, Vander Salm TJ. Off-um versus on-um coronary artery byass surgery: a case-matched comarison of clinical outcomes and costs. Heart Surg Forum 2000;3: Hernandez F, Cohn WE, Baribeau YR, et al. In-hosital outcomes of off-um versus on-um coronary artery byass rocedures. A multicenter exerience. Ann Thorac Surg 2001;72: Puskas JD, Winston AD, Wright CE, et al. Stroke after coronary artery oeration: incidence correlates, outcomes and cost. Ann Thorac Surg 2000;69: Amassi GH, Sommers T, Moritz TE, et al. Stroke in cardiac surgical atients: determinants and outcome. Ann Thorac Surg 1999;68: Bowles JB, Lee JD, Dang CR, et al. Coronary artery byass erformed without use of cardioulmonary byass is associated with reduced cerebral microemboli and imroved clinical results. Chest 2001;119: Blackstone EH. Breaking down barriers: helful breakthrough statistical methods you need to understand better. J Thorac Cardiovasc Surg 2001;122: Concato J, Feinstein AR, Holford TR. The risk of determining risk with multivariable models. Ann Intern Med 1993;118: DISCUSSION DR LISHAN AKLOG (New York, NY): One concern that might be raised regards the secific techniques that were used in the on-um grou to minimize the neurologic risks, such as eiaortic scanning, maintaining a high erfusion ressure, and using a single clam technique instead of a side-biting clam. In other words, were the on-um atients given the best oortunity to minimize their neurologic risk? DR PATEL: No, we did not use any eiaortic scanning, but we used a single alication of a side-biting clam. These atients did not have single cross-clam alication to construct the roximal anastomosis. I doubt this technique would have made a significant difference, as our incidence of neurologic deficits in the on-um grou was 1.6%, which is much lower than most reorts. DR THOMAS Z. LAJOS (Buffalo, NY): I want to congratulate you for the excellent aer and statistics you have done. We have investigated the same roblem, actually only on reoerative coronary artery disease. We thought that reoerative coronary artery disease is carrying a higher incidence of neurologic adverse reactions, so they may show higher statistical significance. We have done 6,629 cases of sole coronary artery byasses between 1995 and Of these, only 667 cases reresented our study: reoerative coronary disease. Of the 667 cases that we have done, 342 were on-um and 325 cases were off-um. If one further breaks down the off-um cases, they were minimally invasive rocedures; the left anterior thoracotomy, the left osterolateral thoracotomy, and right gastroeiloic rocedures with subxihoid aroach; 46 (14.2%), 66 (20.3%), and 33 cases (10.2%), resectively. The incidence of neurological comlications on-um in reoerative coronary artery cases was 3.8% (13/324 cases) while off-um it was 0.6% (2/325 atients) But if one further breaks down the off-um cases to the 144 minimally invasive consecutive cases (MIDCAB, subxihoid gastroeiloic rocedures and left thoracotomy), we had no neurological comlications. So we say that as long as one does not ut on clams, does not use the um, and avoids maniulation of the heart, the atients have a good chance to avoid neurological comlications. I have one question to the authors, and that is: did you break down searately the reoerative grou? If you did, it would be interesting to know what the statistics are for the reoerative grou. DR PATEL: You just saw from our slide that the reoeration rate was retty low (around 5%) in our exerience, so I think it would have been difficult to achieve statistical significance, because our rate of neurologic deficit was also low (1.6%). That is why we did not break it down. All the atients in this study had median sternotomy aroach. No atients had a minimally invasive aroach. DR EUGENE A. GROSSI (New York, NY): Congratulations on a fine clinical exerience. I question the validity of your model, because although the model evaluates the comlication of stroke it does not include intraoerative echocardiograhic findings regarding atheromotous aorta, which we all know is the strongest redictor of oor neurologic outcome. DR PATEL: Eiaortic scanning is not routinely erformed in our institutions and so it does not exist in our database as a variable. Thus we did not have the state of the aorta as a factor in our model. DR HANI SHENNIB (Montreal, Quebec, Canada): I want to congratulate you on this study, which brings u an imortant issue, and that is that new technology and techniques are forcing us to rearaise our current ractice. One asect of this rearaisal is neurologic outcome. My concern with this study is that

7 406 PATEL ET AL Ann Thorac Surg EFFECT OF AVOIDING CARDIOPULMONARY BYPASS 2002;74:400 6 it has the inert limitations of a retrosective analysis, and I think we all agree that crude neurologic assessment is not the ideal way to determine the advantage of one technique over another. Are you now going to go back and start designing a study that will look at more details, including your sychological analysis? We live in an era in which lifestyle is imortant and we areciate the imact of a technique on quality of life and sychological outcome. Thank you. DR PATEL: Thank you for your comments, Dr Shennib. I fully agree that this retrosective study has its own limitations, but the reason we looked at only the gross neurologic deficit was to avoid any subjective bias. Gross neurologic deficits are more clinically imortant than transient difficulties for a atient being discharged. So we included only ermanent deficits. And I fully agree that there is a need for having a detailed neurosychiatric assessment in a systematic fashion, in a large, randomized, multicenter exerience. Such a study will robably rove the truth. DR THOMAS Z. LAJOS (Buffalo, NY): I want to congratulate you for the excellent aer and statistics. We have investigated the same roblem, actually only on reoerative coronary artery disease. We thought that reoerative coronary artery disease carries a higher incidence of neurologic adverse reactions, so our data may show higher statistical significance. We have carried out 6,629 sole coronary artery byasses between 1995 and Of these only 667 cases reresented our study, reoerative coronary disease. Of the 667 cases 342 were on-um and 325 were off-um. If one further breaks down the off-um cases, they were minimally invasive rocedures; the left anterior thoracotomy, the left osterolateral thoracotomy, and right gastroeiloic rocedures with subxihoid aroach comrised 46 (14.2%), 66 (20.3%), and 33 cases (10.2%), resectively. The incidence of neurologic comlications on-um in reoerative coronary artery cases was 3.8% (13/324 cases) while off-um it was 0.6% (2/325 atients) ( 0.007). But if one further breaks down the off-um cases to the 144 minimally invasive consecutive cases (MIDCAB, subxihoid gastroeiloic rocedures, and left thoracotomy), we had no neurologic comlications. So we say that as long as the surgeon does not ut on clams, does not use the um, and avoids maniulation of the heart, the atients have a good chance of avoiding neurologic comlications. I have one question to the authors: did you break down searately the reoerative grou, and if you did it would be interesting to know what were their statistics.