| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ann Thorac Surg 2006;81:2105-2114
© 2006 The Society of Thoracic Surgeons
a Heart Research Centre, Melbourne, Australia
b Department of Cardiothoracic Surgery, The Royal Melbourne Hospital, Melbourne, Australia
Accepted for publication January 3, 2006.
* Address correspondence to Dr Ernest, Heart Research Centre, Box 2137 Post Office, The Royal Melbourne Hospital, Victoria 3050, Australia. (Email: christine.ernest{at}heartresearchcentre.org).
 |
Abstract |
|---|
 Top Abstract IntroductionPatients and MethodsResultsCommentAcknowledgmentsReferences |
|---|
METHODS: Consecutively listed candidates for elective bypass were randomly assigned to either off-pump or on-pump techniques (n = 107). A battery of 11 standardized neuropsychological tests was administered before surgery, and again at 2 and 6 months after surgery. The two groups were compared using a range of statistical procedures, including growth modeling.
RESULTS: There were no significant differences in cognitive test scores between the off-pump and on-pump groups using t tests at any of the time points. There were no differences between off-pump and on-pump groups in the incidence of cognitive deficits at 2 months or 6 months, with the exception that fewer off-pump patients showed impairment on one test of verbal fluency at 6 months. When the pattern of cognitive change over time between the two groups was compared using sophisticated modeling techniques, the two groups were again comparable, except for results on the test of verbal fluency, in which the off-pump group showed more rapid postsurgical cognitive gains than the on-pump group.
CONCLUSIONS: The off-pump group appears to be generally comparable to the on-pump group in terms of short-term and long-term postsurgical neurocognitive outcomes.
|
Introduction |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentAcknowledgmentsReferences |
|---|
A search of the literature identified nine studies that have assessed cognitive outcomes in patients randomly assigned to either off-pump or on-pump bypass surgery [9–17]. Two studies show some instances of better cognitive outcomes in the off-pump group on certain tests, at certain time points, or when certain statistical methods were used [12, 16]. Lee and colleagues [12] reported better cognitive outcomes in the off-pump group, specific to a test of verbal memory, but no group differences in the incidence of cognitive decline. Van Dijk and colleagues [16] conducted a large randomized trial (n = 281) with a comprehensive battery of neuropsychological tests and demonstrated better outcomes in off-pump patients at 3 months but not at 12 months after surgery. Again, there was no difference between the groups in the incidence of cognitive decline [16].
Two studies [10, 17] reported a definitive finding of better cognitive outcomes for off-pump surgery. The study by Diegeler and colleagues [10] was small (n = 40) and did not include any of the commonly recommended and applied cognitive tests. In addition, the postsurgical follow-up was limited to one assessment conducted 7 days after surgery. The Statement of Consensus on Assessment of Neurobehavioral Outcomes after Cardiac Surgery [18] recommends avoiding cognitive assessments in the period immediately after surgery, with one assessment at least 3 months after surgery when cognitive performances have stabilized. The study by Zamvar and colleagues [17] had a relatively larger sample (n = 60) and showed a lower incidence of cognitive impairment for off-pump patients. However, once more, follow-up assessments were performed at 1 week and 10 weeks after surgery, making long-term conclusions difficult. Finally, five randomized studies showed no differences between the off-pump and on-pump groups in terms of short-term and long-term cognitive outcomes [9, 11, 13–15], although only two of these five studies were published in their entirety in peer reviewed journals [13, 14]. Therefore, a clear cognitive advantage for the off-pump surgical technique has not been convincingly demonstrated.
The present study was designed to address some of the methodologic shortcomings that may have confounded the results of previous studies. Our study used a large standardized battery of neuropsychological tests covering key cognitive domains and had follow-up time intervals consistent with the consensus statement [18]. The methodologic and experimental design also accounted for practice effects and fatigue. The specific aim of the study was to compare cognitive outcomes in off-pump and on-pump patients at 2 and 6 months after surgery.
|
Patients and Methods |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentAcknowledgmentsReferences |
|---|
Of the 117 patients eligible for the cognitive component of the trial, 112 patients were randomly assigned to either off-pump or on-pump surgery between July 2001 and April 2004. Of these, 5 patients were excluded owing to withdrawn consent, inadequate English, or presence of comorbid condition with cognitive deficits (Fig 1).
|
Most of the surgical interventions were performed by two cardiothoracic surgeons, with each performing approximately equal numbers of both procedures. Four other surgeons were also involved. Bypass surgery was performed through a median sternotomy. The anesthetic regimen, sternal closure, postoperative care, and pain management were identical for off- and on-pump patient groups.
In the on-pump patient group, CPB involved the use of an arterial line filter, a membrane oxygenator, and a roller pump. Cardiac arrest was achieved by anterograde and retrograde blood potassium cardioplegia at 20°C. The flow was non pulsatile and was maintained at 2.5 L · min–1 · m–2. To institute CPB, the distal ascending thoracic aorta and right atrium were cannulated. The aorta was cross-clamped, and all anastomoses were constructed during the one period of cross-clamping. Patients were maintained at 32°C to 33°C. Rewarming was slow and with minimal gradients to 35°C to 36°C. The pH was maintained at 7.35 to 7.46, and the mean arterial pressure at 70 mm Hg or greater.
In the off-pump group, the induction and the initial phase of the operation were identical to those of the on-pump group. Appropriately placed pericardial stay sutures were used to visualize the pericardium. The aorta was avoided by extensive use of left internal thoracic-Y-radial artery composite grafts, pedicled and extended (with radial artery) right internal thoracic artery grafts. Saphenous vein grafts were not used. Side-biting clamps on the aorta were not used. The left anterior descending artery was anastomosed first, as this revascularized the anterior wall and septum of the heart, providing hemodynamic stability. Anastomotic stabilization was achieved with either the Octopus IV epicardial suction stabilizer (Medtronic, Minneapolis, Minnesota), or Platypus stabilizer (Foot plate compression style; Wolfram, Sydney, Australia). Systolic blood pressure was maintained at 110 mm Hg or more, and mean pressure at 70 mm Hg or more. The heart rate was kept at 60 beats per minute or higher.
Measures
Cognitive function was assessed using 11 standardized cognitive tests, covering a variety of domains, including tests recommended by the consensus statement [18] (Table 1). Test selection criteria included Murkin's guidelines [18], use in previous studies examining cognitive outcome after cardiac surgery, published norms, adequacy of psychometric properties (ie, reliability and validity), brevity, and ease of administration [18, 19]. Psychometrically comparable short forms were used wherever possible. Standard recommendations were followed for test administration and scoring [20, 21].
|
Assessments were conducted by interns undertaking doctoral training in neuropsychology, and they were supervised by a qualified neuropsychologist. Three assessors conducted most of the assessments, with another six being involved on an on-call basis to maintain strict time intervals and minimize dropout. All assessors were blind to randomization.
Sociodemographic, medical, and surgical characteristics were also recorded along with an estimate of intelligence, which was obtained using the National Adult Reading Test–Revised [22].
Data Analysis
In accordance with standard practice, data were analyzed by "intention to treat" [23, 24]. "Closest match" imputations were performed to compensate for missing data for patients who had completed the presurgical baseline assessment but only one of the two postsurgical assessments [25]. At the 2-month follow-up, the percentage of data imputed ranged from 8.7% to 11.8%, and at 6 months, it ranged from 5.9% to 7.7%.
Of the 11 cognitive tests administered, the Stroop test was excluded from the analysis owing to problems in the spectral resolution of the reproduced stimulus cards. As the tests generated a large number of variables, a total of 12 variables was selected as the final outcome measures (Table 1), based on statistical and theoretical grounds. The relative loadings of test variables on a principal components analysis were used to select variables. Additionally, summary scores were selected if shown to be highly correlated with component scores. Theoretical guidelines from the consensus statement, precedents set by previous studies, and expert opinion also guided the selection process [12, 16, 18].
Scores on the 12 selected variables for the off-pump group and on-pump group were compared through t tests using Statistical Package for the Social Sciences (SPSS) [26] at baseline, and at 2 and 6 months. The incidence of cognitive impairment in both groups was compared using 
2 analyses in SPSS for each individual test at baseline, and at 2 and 6 months. Incidence of cognitive impairment was defined as test scores that were greater than 1.5 SD below published normative means [20, 21].
In examining group differences in the change of cognitive test scores over time, the commonly used repeated measures analysis in SPSS was deemed to be inappropriate as it assumed linearity in test score changes over time. Therefore, a two-stage growth modeling exercise using Mplus [27] was carried out to describe the changes over time in cognitive functioning by comparing the trajectories of the off-pump and on-pump groups. The Mplus software was chosen primarily because it can estimate determinants of both linear and nonlinear change in cognitive test scores over time and model both off- and on-pump trajectories concurrently, allowing for comparison between the two groups [28, 29]. Technical details of the modeling process are available from the corresponding author.
Separate growth models were examined for each of the cognitive tests. The assumption of linearity was relaxed to allow for the possibility that test scores may change in a nonlinear fashion. The Mplus software produces statistical information about the fit of linear and nonlinear models and generates significance tests to establish the extent to which the trajectories for the two groups differ. Model fit was assessed using the Comparative Fit Index [30], with higher values indicative of a better fit of the model to the data.
Owing to the relatively large number of crossovers, the data were also analyzed using a "treatment received" classification in which crossovers were assigned to the group consistent with the surgery that they actually received.
|
Results |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentAcknowledgmentsReferences |
|---|
|
|
|
|
The growth modeling results for the functional forms that best fit the data are presented for each cognitive test in Table 5. The coefficients for the "intercept regressed on group" describe group differences in presurgical cognitive test performances. A positive coefficient for the intercept regressed on group indicates that baseline test scores were higher for the off-pump patients. The "slope regressed on group" coefficients reflect group differences in cognitive function over time after surgery. A positive coefficient for the intercept regressed on slope indicates that the test scores for the off-pump patients increased more quickly than those for the on-pump group. To test for nonlinear test score trajectories, the variable representing time (number of months) was transformed. For linear models, time could take three possible untransformed values: 0, 2, and 6 (months). For nonlinear models, time was transformed to 0, 4, and 36 (quadratic models) or 0, 1.414, and 2.449 (square root models). Nine of the cognitive tests showed nonlinear trajectories, with five having a square root form (more rapid change in the early months) and four having a quadratic form (more rapid change in the later months). Again, significant differences in the trajectories between off-pump and on-pump groups were observed exclusively on the COWAT (Table 5), in which the off-pump group showed a faster rate of improvement in verbal fluency over time.
|
|
Comment |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentAcknowledgmentsReferences |
|---|
The COWAT has not been widely used in previous studies comparing off- and on-pump surgery, except for a study by Zamvar and colleagues [17] that reported no differences between the two groups on this test. However, the method of analysis and classification of impairment used in that study was quite different, making direct comparison difficult. The study by Lee and colleagues [12] reports a similar pattern of improvement in the off-pump group, albeit on a test of verbal memory, the Rey Auditory Verbal Learning Test (RAVLT). It is of interest that both the COWAT and the RAVLT are language-based tasks, which are predominantly subserved by the left hemisphere, particularly the frontal and temporal lobes [20]. Using single photon emission computed tomography, Lee and colleagues [12] report reduced cerebral perfusion in the left temporal lobe in on-pump patients compared with off-pump patients. This finding could possibly explain subtle postsurgical cognitive benefits in off-pump patients in tasks that are subserved by this cortical area. It is interesting that two studies that compared cognitive test performances of candidates for bypass with healthy control groups showed that cardiac patients had significant impairments in verbal fluency as measured by the COWAT before bypass surgery [31, 32]. More studies using tests of verbal fluency are required to further investigate the association between the neural systems underlying this cognitive function and cardiac disease-related factors or the use of CPB.
The strengths of this study include the adherence to the recommended guidelines [18], the use of a large standardized test battery, appropriate timing intervals, compensation for practice effects and fatigue, and the randomized design. This study predominantly used continuous test scores, which allows for more powerful and reliable methods of statistical analysis than dichotomous classifications, and reduces unnecessary manipulations of the data that could artificially influence outcomes [4, 7, 12, 16, 19]. Many previous studies that examine cognitive outcomes after bypass surgery tend to report exclusively on the incidence of cognitive decline, which is a dichotomous outcome, and is reliant on the definition of decline that is selected for the analysis [4, 33]. Varying definitions of cognitive decline have been used when measuring change in cognitive test performances after surgery, many of which may misclassify changes that are within expected fluctuations in repeated test administration [16]. The use of change scores to measure outcome is relative to presurgical test performances, which are assumed to be entirely accurate and stable. Although the use of incidence rates as primary outcomes is not recommended, these have intrinsic clinical relevance. Therefore, we included an analysis of incidence rates, reporting on the incidence of cognitive impairment on individual tests at each time point, instead of the widely used incidence of cognitive decline. Unlike previous studies that defined cognitive decline in relation to patients' baseline performances, this study adopted a definition of cognitive impairment in relation to published normative data. It is noteworthy that the group differences on the verbal fluency task were seen in the growth modeling exercise and the incidence of impairment analysis, but not on the comparison of the two groups on raw cognitive test scores. This observation highlights the fact that different methods of analysis can produce discrepant findings. Hence, we have reported on all three methods of examining the data in an attempt to present a complete picture of cognitive outcomes.
The use of continuous test scores also challenges an erroneous assumption that bypass surgery is only associated with cognitive decline. Previous studies with long-term follow-ups have documented improvements in cognitive function after bypass surgery for the majority of patients, and attribute this to practice effects or selective attrition [4, 6, 12, 14, 16, 17, 20, 33]. Studies that have shown postsurgical deterioration in test scores generally report this pattern in the early postoperative period (less than 2 months after bypass), with deterioration being shown to be completely reversible at later follow-ups [4, 6, 17, 34]. This finding underlines the importance of using a control group in postbypass studies of cognitive outcomes in which the group of interest, in this case the off-pump group, is compared with the conventional treatment group, the on-pump group, thereby controlling for practice effects and regression to the mean [7].
The use of sophisticated modeling techniques that allowed for linear and nonlinear models of change is an important strength of this study, as most cognitive tests scores changed over time in a nonlinear fashion. It is imperative that future research examining cognitive outcomes after an acute event or surgical treatment considers nonlinear models when assessing postevent cognitive change. It is a concern that the majority of past studies have reported on dichotomous outcomes dependent on arbitrary definitions. In addition, those that used continuous test scores have utilized statistical analyses that assume linearity of data, when it is apparent that cognitive outcomes rarely meet this assumption.
Despite the methodologic strengths of this study, a larger sample size and fewer crossovers and dropouts might have enhanced the ability to detect subtle differences between the two groups. Although most of the growth models fitted the data well, the relatively small sample size (n = 107) resulted in estimation difficulties on some tests. Some of the patterns seen in Figure 2 might have been statistically significant with a larger sample. On the other hand, although 12 outcome variables were selected from the numerous scores generated by the large cognitive test battery, the significant finding in verbal fluency could be due to the multiple comparisons that were undertaken (type I error). For instance, there were no group differences on any of the cognitive tests when data were analyzed using the "treatment received" principle. Furthermore, as in the study by Lee and colleagues [12], off-pump patients had fewer grafts than on-pump patients. Although, this difference could have occurred purely by chance, it has been suggested that this may be due to technical limitations in the off-pump procedures that precluded revascularization of vessels that are not easily accessible [12]. As a greater number of grafts have been associated with worse cognitive outcomes [4, 35], any subtle cognitive gains found in the off-pump group could be a function of the number of grafts rather than the avoidance of CPB.
In conclusion, the findings of this study indicate that the off-pump procedure does not demonstrate a clear advantage over conventional on-pump procedures in terms of cognitive outcomes. Further randomized studies using tests of verbal fluency, larger samples, and sophisticated analyses are required to investigate the clinical relevance of these findings.
|
Acknowledgments |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentAcknowledgmentsReferences |
|---|
|
References |
|---|
|
TopAbstractIntroductionPatients and MethodsResultsCommentAcknowledgmentsReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  C. H. Moller, L. Penninga, J. Wetterslev, D. A. Steinbruchel, and C. Gluud Clinical outcomes in randomized trials of off- vs. on-pump coronary artery bypass surgery: systematic review with meta-analyses and trial sequential analyses Eur. Heart J., November 1, 2008; 29(21): 2601 - 2616. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. F. Marasco, L. N. Sharwood, and M. J. Abramson No improvement in neurocognitive outcomes after off-pump versus on-pump coronary revascularisation: a meta-analysis Eur. J. Cardiothorac. Surg., June 1, 2008; 33(6): 961 - 970. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 O. A. Selnes, M. A. Grega, M. M. Bailey, L. Pham, S. Zeger, W. A. Baumgartner, and G. M. McKhann Neurocognitive Outcomes 3 Years After Coronary Artery Bypass Graft Surgery: A Controlled Study Ann. Thorac. Surg., December 1, 2007; 84(6): 1885 - 1896. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Al-Ruzzeh and D. O'Regan Assessment of Neurocognitive Outcome After Cardiac Surgery Ann. Thorac. Surg., July 1, 2007; 84(1): 358 - 358. [Full Text] [PDF]
|
|
|
|
|
|
C. Ernest, M. Worcester, J. Tatoulis, P. Elliott, B. Murphy, R. Higgins, M. Le Grande, and A. Goble Reply Ann. Thorac. Surg., July 1, 2007; 84(1): 358 - 359. [Full Text] [PDF]
|
|
|
|
|
|
C. S. Ernest, B. M. Murphy, M. U.C. Worcester, R. O. Higgins, P. C. Elliott, A. J. Goble, M. R. Le Grande, N. Genardini, and J. Tatoulis Cognitive function in candidates for coronary artery bypass graft surgery. Ann. Thorac. Surg., September 1, 2006; 82(3): 812 - 818. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
