Early and delayed stroke after coronary surgery an analysis of risk factors and the impact on short- and long-term survival,

Transcription

1 European Journal of Cardio-thoracic Surgery 40 (2011) Early and delayed stroke after coronary surgery an analysis of risk factors and the impact on short- and long-term survival, Abstract Magnus Hedberg a, *, Patrik Boivie b, Karl Gunnar Engström a a Department of Surgical and Perioperative Science, Heart Center, Cardiothoracic Division, Umeå University Hospital, S Umeå, Sweden b Department of Surgical and Perioperative Science, Section for Hand and Plastic Surgery, Umeå University Hospital, S Umeå, Sweden Received 15 September 2010; received in revised form 25 November 2010; accepted 29 November 2010; Available online 17 February 2011 Objective: Stroke is a serious complication to cardiac surgery, and is generally considered as a uniform disease regardless of its temporal relationship to surgery. Our hypothesis suggests that stroke, in association with surgery, reflects other characteristics than stroke occurring with a free interval. This issue was here explored for risk factors and survival effects. Methods: Data were collected from 7839 procedures of isolated coronary artery bypass grafting (CABG), 297 off-pump CABG, and 986 combined CABG and valve procedures. Records of patients with any signs of neurological complications were reviewed to extract 149 subjects with stroke at extubation (early, 1.6%) versus 99 patients having a free interval (delayed, 1.1%). Survival data were complete, with a median follow-up time of 9.3 years (maximum 16.3 years). Independent risk factors were analyzed by logistic regression and survival by Cox regression. Results: Risk factors for early stroke were advanced age, high preoperative creatinine level, extent of aortic atherosclerosis, and long cardiopulmonary bypass time (all P < 0.001). Factors associated with delayed stroke were female gender (P < 0.001), unstable angina (P = 0.003), previous cerebrovascular disease (P = 0.009), inotropic support requirement (P < 0.001), and postoperative atrial fibrillation (P < 0.001). Stroke explained mortality not only in the early postoperative period (P < 0.001), but also at long-term follow-up (P < 0.001). Early and delayed stroke were associated with mortality hazard ratios (HRs) of 1.44 and 1.85 (P = 0.008, P < 0.001), respectively. However, for patients surviving their first postoperative year, early stroke did not influence long-term mortality (HR 1.07, P = 0.695). This was in contrast to delayed stroke (HR 1.71, P = 0.001). Conclusions: Early and delayed stroke differed in their related risk factors. The influence of stroke on short-term mortality was obvious and devastating. Mortality in association with early stroke mainly presented itself in the acute period, whereas for delayed stroke survival continued to be impaired also in the long-term perspective. Our report emphasizes that early and delayed stroke should be considered as two separate entities. # 2010 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved. Keywords: Stroke; Cardiac surgery; Risk factors; Survival 1. Introduction Stroke is a devastating complication to cardiac surgery, with a reported rate between 1.6% and 8.4% [1,2]. Previous studies have described numerous risk factors behind stroke, for example, advanced age, female gender, history of cerebrovascular events, impaired left ventricular function, and atherosclerosis of the ascending aorta [1,3]. Off-pump coronary bypass (OPCAB) has been reported to offer a lower risk of stroke [4]. In the majority of previous studies, stroke, in association to cardiac surgery, has been investigated without considering the timing details. However, only a few reports have analyzed to The article has been presented at the Scandinavian Conference in Cardiothoracic Surgery, Oslo, 27 August 2010 (abstract, 244 words, not published). Funding sources: This study had support from the Medical Faculty at Umea University and from the Heart Foundation of Northern Sweden. The study had no commercial funding. * Corresponding author. Tel.: / ; fax: address: magnus.hedberg@karolinska.se (M. Hedberg). what extent stroke occurring during operation differs in characteristics compared with those occurring in the postoperative period [1,5,6]. Hogue et al. reported the rate of early and delayed stroke to be 0.6% and 1.0%, respectively, with the two groups carrying different risk factors [1]. Perioperative stroke results in longer hospital stay and also in a higher in-hospital mortality [1,7,8]. Many studies have only focused on the short-term consequences of stroke, whereas the knowledge regarding the long-term survival effects of stroke is limited. Dacey et al. analyzed coronary artery bypass grafting (CABG) and concluded that the greatest risk of death was within the first postoperative year, but with a sustained mortality effect during the following 10-year follow-up [9]. However, their study did not separate stroke into early versus delayed type. Other available studies have described this subdivision of stroke, but have not analyzed the potential group difference in terms of risk factors [6,10,11]. We hypothesized that early and delayed stroke represent two different entities with different influences on survival /$ see front matter # 2010 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved. doi: /j.ejcts

2 380 M. Hedberg et al. / European Journal of Cardio-thoracic Surgery 40 (2011) Therefore, the purpose of this study was to separately examine early and delayed stroke for risk factors, and to analyze their individual impacts on short- and long-term survival. 2. Materials and methods 2.1. Patient population Database inputs were collected from adult patients operated upon at the Cardiothoracic Surgery Department, Heart Center, Umeå University Hospital between January 1994 and December During this period, the recorded variables were managed consistently. Three surgical groups were extracted; isolated CABG, OPCAB, and coronary surgery Table 1. Patient characteristics. in combination with aortic and/or mitral valve procedures (COMB), in a total of 9235 procedures. Five of these patients were excluded due to foreign citizenship and hence were not available to survival follow-up. The analysis aimed at evaluating survival effects of stroke. A stroke diagnosis was required, and, therefore, patients dying in the operating theater (n = 18) or within the following 24 h (n = 32) were excluded. Patients with short episodes of neurological symptoms (i.e., transient ischemic attacks) were excluded (n = 28) to avoid uncertain events. A few patients had neurological deficits explained by other etiologies (e.g., spinal cord ischemia, preoperative symptoms, or global ischemia due to heart lung resuscitation). These were also excluded (n = 30), leaving a total of 9122 patients for analysis. Redo procedures accounted for 4.5% of the operations (Table 1). Therefore, some patients reoccurred All patients (n = 9122) CABG (n = 7839) OPCAB (n = 297) COMB (n = 986) General data Age, years 67.5 (60.0/73.3) 66.7 (59.3/72.6) 65.6 (57.4/71.6) 73.5 (67.9/77.6) Males, % Weight, kg 79 (71/88) 79 (71/88) 80 (72/88) 75 (67/84) Clinical preop data Previous cerebrovascular accident, % Diabetes mellitus, % Pulmonary disease, % Hypertension, % Anemia, % Serum creatinine, mmol/l 81 (69/95) 80 (69/94) 78 (66/95) 87 (72/106) Cardiac preop data Left ventricular function, 1 = Good, 3 = Bad 1 (1/2) 1 (1/2) 1 (1/1) 1 (1/2) NYHA class IV, % Main stem stenosis, % Unstable angina, % Previous cardiac operation, % Higgins score 2 (1/3) 1 (0/3) 1 (0/3) 4 (2.75/5) EuroSCORE 4 (2/4) 4 (2/6) 4 (2/5) 7 (5/8) Peroperative data Aortic wall quality, 1 = Normal, 3 = Severely 1 (1/2) 1 (1/1) 1 (1/1) 1 (1/2) Surgery time, h 3.1 (2.5/3.8) 3.0 (2.5/3.7) 2.2 (1.7/2.6) 4.2 (3.4/5.0) No. of anastomoses 3 (3/4) 3 (3/4) 1 (1/2) 2 (1/3) CPB time, min 79.0 (60.0/104) 75 (58/95) n.a. 143 (119/175) Aortic clamp time, min 42.0 (31/58) 39 (30/51) n.a. 99 (81/118) Complicated weaning CPB, % n.a Defibrillation required at weaning, % n.a Inotropic support, % IABP, % Total bleeding, ml 400 (300/500) 400 (300/500) 300 (200/500) 500 (400/800) Erythrocytes given, % Intensive care data Atrial fibrillation in ICU, % Mediastinal drain output 8 h, ml 520 (400/725) 520 (400/720) 460 (300/695) 548 (390/771) Erythrocytes given, % Time on ventilator, h 6.5 (4.5/10) 6.3 (4.4/9.5) 4.8 (3.5/7.0) 9.0 (6.0/17.3) Time in ICU, h 21.4 (18.2/23.0) 21.3 (18.0/22.8) 21.9 (17.3/23.5) 22.0 (20.5/45.5) Follow-up data Length of stay, days 9 (8/11) 9 (8/11) 8 (7/9.5) 11 (9/15) Hospital mortality, % a days mortality, % a One-year mortality, % a Stroke incidence Early stroke, % Delayed stroke, % Numeric variables are presented as median values and quartile range rather than categorized in order to save space. CABG: coronary artery bypass grafting; OPCAB: off-pump coronary artery bypass; COMB: coronary surgery in combination with aortic and/or mitral valve procedures; NYHA: New York Heart Association; CPB: cardiopulmonary bypass; IABP: intra-aortic balloon pump; and ICU: intensive care unit. a Mortality during surgery or within the first 24 h are excluded.

3 M. Hedberg et al. / European Journal of Cardio-thoracic Surgery 40 (2011) in the data set. These repeated patients were handled as separate cases in the analyses. The study protocol was approved by the local ethics committee Stroke review protocol and definitions The database had a prospective input of data. Data were entered on a daily basis from the time of patient referral until hospital discharge, including operating room, intensive care unit (ICU) and ward length of stay. Inputs were from surgeons, anesthesiologists, intensivists, perfusionists, and nurses. Overall, 53 variables were extracted for analysis, and their mode of categorization is listed in Table 2. In addition, the database contained 17 variables (not listed in Table 2) describing neurological symptoms. Patients with any type of neurological symptoms were screened for using these variables, which widely extracted potential stroke events. This search yielded 523 subjects for further analysis. The records of these patients were reviewed in detail, according to a protocol, to confirm the diagnosis of stroke and to separate the patients into early and delayed stroke groups. For most patients, the diagnosis was obvious at the time of the event. For patients with uncertain diagnosis, the pattern of symptoms, neurologist s report, and computed tomography (CT) findings were evaluated to confirm the stroke. Stroke criteria were fulfilled in 248 patients. The remaining patients for whom the stroke criteria were rejected were grouped together with the control subjects. The control group contained 8874 patients. Stroke definitions were according to routine guidelines [12 14], and included any new focal or global neurologic deficit lasting for more than 24 h. Early stroke was defined as symptoms observed at extubation (n = 149), whereas delayed stroke followed a symptom-free interval after extubation (n = 99). Of the reviewed patients, a neurologist was consulted in 52% of events, and for 84% of the patients, a CTscan was performed. Stroke appearing after discharge was not considered in this study. An overview of demographic details is listed in Table 1. The present study includes a subgroup of 2082 patients previously analyzed and published, with respect to stroke symptoms and CT findings [8,15] Database definitions In general, preoperative variables had a definition according to contemporary scores, Parsonet, Higgins, and EuroSCORE (European System for Cardiac Operative Risk Evaluation). All patients were evaluated for aortic quality by a semi-quantitative scale at three levels, based on palpation at surgery. In more recent years, epi-aortic scanning has also been used for evaluation. Left ventricular function was assessed by either ventriculography and/or echocardiography. The function was graded as good, reduced, or poor, approximately corresponding to ejection fractions of >50%, 30 50%, and <30%, respectively. Patients were monitored by continuous electrocardiography (ECG) during a minimum period of 48 h, postoperatively. Beyond this period, occurrence of arrhythmia was evaluated by nurses and physicians, at least every 8 h, or, more frequently, if rhythm disturbances were suspected. The follow-up data on all-cause mortality Table 2. Mode of categorization of data. General data Age, years <60, 60 70, 70 80, 80 Sex Male/female Weight, kg 65 90, <65, >90 Period of surgery , , Type of operation CABG, OPCAB, COMB Clinical preop data Diabetes mellitus Pulmonary disease Hypertension Peripheral vascular disease Previous cerebrovascular accident Hepatic disease Smoking Anemia Renal disease Dialysis, in use Serum creatinine, mmol/l <100, , >200 Cardiac preop data NYHA class I III/IV Left ventricular function Good, reduced, poor Preoperative cardiac rhythm Sinus rhythm, atrial fibrillation, pacemaker Main stem stenosis Unstable angina Coronary vessel disease, stenosis 50% 0, 1 2, 3 Previous cardiac operation Peroperative data Aortic wall quality/calcification Normal, moderately, severely 0, 1 2, >2 No. of anastomoses <3, 3 4, >4 Surgery time excluding CPB time, min <100, , >140 CPB time, min <90, , >150 Cardioplegic route Antegrade/retrograde Cardioplegic type Crystalloid/blood Aortic clamp time, min <60, 60 90, >90 Complicated weaning CPB Defibrillation required at weaning Peroperative urine production, ml/h >500, , <250 Total volume balance, ml <3000, , >6000 Units of erythrocytes required, No. 0, 1 2, >2 Units of plasma required, No. 0, 1 2, >2 Units of platelets required, No. 0/1 Total bleeding, ml <450, , >600 Inotropic requirement Vasoactive drugs requirement Cardiac rhythm at sternal closure Sinus rhythm, atrial fibrillation, pacemaker IABP requirement Postoperative data Perioperative stroke y No stroke, early, delayed Time on ventilator, h*,y <9, 9 18, >18 Time in ICU, h y <24, 24 72, >72 Mediastinal drain output 8 h, ml*,y <600, , >900 Units of erythrocytes required in 0, 1 2, >2 ICU, No. * Units of erythrocytes required in ICU/ward, No. y Units of plasma required in ICU, No.*,y 0, 1 2, >2 Units of platelets required in ICU, No.*,y 0/1 Atrial fibrillation in ICU/ward*,y Survival Follow-up per May 6th 2010 CABG: coronary artery bypass grafting; OPCAB: off-pump coronary artery bypass; COMB: coronary surgery in combination with aortic and/or mitral valve procedures; NYHA: New York Heart Association; CPB: cardiopulmonary bypass; IABP: intra-aortic balloon pump; and ICU: intensive care unit. First dichotomization level is used as reference. *Added postoperative variables for logistic regression of delayed stroke. y Added postoperative variables for survival analysis.

4 382 M. Hedberg et al. / European Journal of Cardio-thoracic Surgery 40 (2011) were collected from the Swedish population registry, using the 10-digit national identification number. The survival data presume 100% accuracy, and included the exact day of death. Survival follow-up was closed on 6 May The used variables are further described by their mode of categorization shown in Table Statistics Extracted database variables were tabulated (Excel, Microsoft Corp, Redmond, WA, USA) and analyzed (Statistical Package for Social Sciences (SPSS) release 16.0 for Windows, SPSS Inc, Chicago, IL, USA). Continuous variables were categorized into multiple levels. This was done to avoid identified non-linear characteristics. The reference category aimed to account for normal-type values and to include the majority of observations. The intervals for the remaining categories were adjusted in view of their distribution and clinical relevance. The period of surgery was added as a variable to compensate for possible variation in surgical management and care during the 11- year period of inclusion (Table 2). Some variables showed an obvious collinearity, the typical example being surgery time, which includes time during cardiopulmonary bypass (CPB). To reduce this effect, CPB time was subtracted from surgery time to create a new independent variable. Aortic clamp time showed a similar phenomenon against CPB, although these two variables remained unchanged to reflect the ischemic period versus general CPB effects. All three surgery groups shared coronary disease, but were obviously different in other aspects. The type of surgery was here treated as a predictor variable rather than analyzing the three groups separately. Univariate predictors of stroke were assessed using the chi-square or Fisher s exact test. All univariate predictors having a P value 0.15 were included in the multivariable logistic regression model. The two stroke groups were tested independently against the control subjects in both univariate and multivariable analyses. Regression analyses used a forward approach, but were manually confirmed in backward mode. Variation explained by the logistic models was evaluated by the Nagelkerke R 2 test. Furthermore, the goodness of fit was analyzed by the Hosmer Lemeshow method. The two stroke groups were compared in a univariate subanalysis to test their difference. The database contained missing observations, overall 6.7% among the 53 used variables. Missing observations were placed in a separate category in the analyses rather than applying case-wise deletion. If the missing-data category had a significant influence at multivariable testing, then, subjects with missing data were omitted by case-wise deletion. This condition applied for none of the variables in the logistic regressions and for three variables in the Cox models. For survival analyses, the starting point of follow-up was the stroke event. For non-stroke subjects, this corresponded to the day of surgery. Kaplan Meier estimates and log-rank testing were performed for mortality rate comparisons. Variables affecting survival in the univariate analysis (P value 0.15) were analyzed in a Cox regression model. The analysis focused on isolated CABG procedures only, to obtain a more homogeneous cohort and to avoid complex valverelated co-morbidities. Similarly, OPCAB patients did not share CPB-specific variables, which would have complicated the analysis. The Cox regression analysis was repeated to focus on mortality beyond the acute phase. This was done to avoid the impact of stroke on early mortality, which was substantial. Patients who survived their first postoperative year were then analyzed, with their survival status reset at 1 year. Moreover, the mortality impact of stroke was studied as a function of various follow-up periods. In this analysis, the follow-up time was shortened and the survival data were adjusted. Events beyond the shortened time period were disregarded. The impact of stroke was re-evaluated by multivariable backward Cox regression. Data are presented as median values with quartiles, odds ratio (OR), or hazard ratio (HR) with 95% confidence intervals (CIs). A P value 0.05 was considered significant. 3. Results 3.1. Relationship of type of surgery and stroke Patient characteristics are presented in Table 1. The overall stroke occurrence was 2.7%. Early and delayed stroke accounted for 1.6% and 1.1%, respectively (Table 1). Type of surgery was a significant univariate risk factor to explain early stroke (P < 0.001) but not for delayed stroke (P = 0.075, Table 3). For early stroke, CABG patients had a lower frequency compared with COMB (P < 0.001, Table 1). The frequency of stroke after OPCAB was significantly less as compared with that after CABG, when the overall stroke rate was considered (CABG vs OPCAB, P = 0.011). However, none of these findings was confirmed significant in multivariable analysis. The median free interval for delayed stroke was 3 days (quartile range 2/5) (Fig. 1). This interval did not differ significantly between CABG and COMB procedures. None of the OPCAB patients presented with delayed stroke. In comparison to the findings shown in Fig. 1, the median length of stay was 9 (8/11) days. [()TD$FIG] Fig. 1. Time distribution of delayed-stroke events.

5 M. Hedberg et al. / European Journal of Cardio-thoracic Surgery 40 (2011) Table 3. Univariate analysis versus control subjects. Only variables with P < 0.15 are presented. Early stroke Delayed stroke Indicator of risk P value Indicator of risk P value General data Age Advanced age <0.001 Advanced age <0.001 Sex Female sex <0.001 Female sex <0.001 Weight Low weight <0.001 Low weight Type of operation COMB <0.001 COMB Clinical preop data Diabetes mellitus Presence Pulmonary disease Presence Presence Hypertension Presence Peripheral vascular disease Presence Previous cerebrovascular accident Presence Presence Anemia Presence Renal disease Presence Presence Serum creatinine High level <0.001 High level Cardiac preop data NYHA class Class IV Class IV Left ventricular function Impaired <0.001 Impaired Unstable angina Presence Presence Peroperative data Aortic wall quality/calcification Poor quality <0.001 Poor quality Surgery time excluding CPB time Prolonged <0.001 CPB time Prolonged <0.001 Prolonged Cardioplegic route Retrograde <0.001 Cardioplegic type Blood <0.001 Blood Aortic clamp time Prolonged <0.001 Prolonged Complicated weaning CPB Yes <0.001 Yes Defibrillation required at weaning Yes Total volume load Increased Units of erythrocytes given Required <0.001 Required Units of plasma given peroperative Required <0.001 Units of platelets given peroperative Required <0.001 Total bleeding Increased <0.001 Increased Inotropic support Required <0.001 Required <0.001 Cardiac rhythm at the end of operation Pacemaker required Pacemaker required IABP Required Postoperative data (restricted for delayed stroke) Time on ventilator n.a. Increased Units of erythrocytes given postoperative n.a. Required Atrial fibrillation postoperative n.a. Presence <0.001 NYHA: New York Heart Association; CPB: cardiopulmonary bypass; IABP: intra-aortic balloon pump; COMB: coronary surgery in combination with aortic and/or mitral valve procedures; and n.a.: not applicable in the analysis of early stroke. P values refer to chi-square or Fisher-test comparisons of each variable. The categorizations of each variable are as described in Table 2. The term indicator of risk gives the direction of effect versus stroke occurrence Univariate and multivariable predictors of stroke Numerous variables affected the occurrence of early and delayed stroke (Table 3), exemplified by advanced age. The stroke rate was also higher in patients with a history of prior cerebrovascular disease. Multivariable logistic regression analysis was performed based on 29 or 25 variables for early and delayed stroke, respectively. The results are presented in Table 4. The two stroke groups did not share any risk factors. The Nagelkerke R 2 values were and for early stroke and delayed stroke, respectively. These values describe a relatively poor explanatory model, although highly significant. The Hosmer Lemeshow test supported an appropriate match between observed and predicted observations, exemplified by an output value of P = for delayed stroke. The groups of early and delayed stroke were tested against each other. The hypothesis was confirmed and the two groups showed significant differences in the characteristics of numerous variables (Table 5) Survival analysis The 30-day mortality was 1.1% for the entire study group. This percentage excluded deaths occurring within the first 24 postoperative hours. With this mortality considered, the corresponding figures were 1.6%. Survival data were complete for all included patients, with a median followup of 9.3 (6.7/12.2) years and a maximum of 16.3 years. This corresponded to patient-years. The 30-day mortality after early stroke was 14.1% and after delayed stroke 8.1%, which contrasted to 0.8% among control subjects (group P < 0.001). Within this short-term perspective, there was no difference in survival status between the two stroke groups (P = 0.189). Further analyses focused on CABG patients only. Fig. 2 shows the survival curves of the two stroke groups versus control subjects. Patients with early and delayed stroke had a significantly shorter survival compared with controls (P < 0.001). The 1-year survival was 80.2% and 87.1%

6 384 M. Hedberg et al. / European Journal of Cardio-thoracic Surgery 40 (2011) Table 4. Multivariable predictors of stroke. OR 95% CI P value Upper Lower Early stroke (n = 149) versus controls (n = 8874) Age, years <0.001 <60 (reference) < Serum creatinine, mmol/l <0.001 <100 (reference) <0.001 > <0.001 Aortic wall quality <0.001 Normal (reference) Moderately <0.001 Severely <0.001 CPB time, min <0.001 <90 (reference) > <0.001 Delayed stroke (n = 99) versus controls (n = 8874) Sex (male as reference) <0.001 Unstable angina Previous cerebrovascular accident Inotropic support <0.001 Postoperative atrial fibrillation <0.001 OR: odds ratio; CI: confidence interval; and CPB: cardiopulmonary bypass. Table 5. Difference between early and delayed stroke. Early stroke n = 149 Delayed n =99 stroke n (%) n (%) P value Sex Male 104 (65%) 56 (35%) Female 45 (51%) 43 (49%) Serum creatinine, mmol/l < (51%) 67 (49%) (72%) 23 (28%) > (87%) 2 (13%) Type of operation CABG 106 (55%) 85 (45%) OPCAB 1 (100%) 0 (0) COMB 42 (75%) 14 (25%) Aortic wall quality <0.001 Normal 55 (47%) 61 (53%) Moderately 50 (62%) 31 (38%) Severely 44 (86%) 7 (14%) Surgery time excluding CPB time, min < (51%) 38 (49%) (56%) 45 (44%) > (78%) 12 (22%) CPB time, min <90 53 (48%) 57 (52%) (63%) 31 (37%) > (79%) 11 (21%) Clamp time, min <60 78 (53%) 68 (47%) (60%) 19 (40%) >90 41 (77%) 12 (23%) Complicated weaning CPB No 117 (56%) 92 (44%) Yes 31 (82%) 7 (18%) CABG: coronary artery bypass grafting; OPCAB: off-pump coronary bypass; COMB: coronary surgery in combination with aortic and/or mitral valve procedures; and CPB: cardiopulmonary bypass. P values refer to Fisher s exact test or chi-square test. for early and delayed stroke, respectively, versus 98.0% for control subjects. At 5 years, the following figures were extracted: 64.5% and 69.1% versus 90.3% The corresponding survival at 10 year, for patients available for long-term follow-up, was 40.3% and 38.2% versus 75.1%. Log-rank tests at univariate level yielded 44 variables that influenced survival. In the Cox regression analysis and with the entire follow-up considered, stroke was a moderate predictor of mortality with HR 1.44 (CI 1.10/1.89) for early stroke and HR 1.85 (CI 1.39/2.46) for delayed stroke, respectively. Additional factors are presented in Table 6. To investigate the temporal impact of stroke, the followup time of survival was shortened stepwise. The influence of stroke on mortality presented the highest impact in the early period. At 6-month s follow-up, the HR was (CI 6.27/ 23.97) and 6.53 (CI 2.73/15.63) for early and delayed stroke, Table 6. Multivariable predictors for mortality after CABG. HR 95% CI P value Lower Upper Age, years <0.001 <60 (reference) < < <0.001 Perioperative stroke <0.001 No stroke (reference) Early stroke Delayed stroke <0.001 Diabetes <0.001 Previous cerebrovascular accident <0.001 Pulmonary disease <0.001 Dialysis Serum creatinine, mmol/l <0.001 <100 (reference) <0.001 > <0.001 Left ventricular function <0.001 Good (reference) Reduced <0.001 Bad <0.001 Unstable angina Aortic wall quality <0.001 Normal (reference) Moderately <0.001 Severely Cardioplegic type CPB time, min <90 (reference) > Inotropic support <0.001 Units of erythrocytes given in OR < (reference) <0.001 > Units of platelets given in ICU (reference) > <0.001 Time in ICU, h <0.001 <24 (reference) > <0.001 Atrial fibrillation postoperative <0.001 HR: hazard ratio; CI: confidence interval; CABG: coronary artery bypass grafting; CPB: cardiopulmonary bypass; OR: operating room; and ICU: intensive care unit.

7 [()TD$FIG] M. Hedberg et al. / European Journal of Cardio-thoracic Surgery 40 (2011) Fig. 2. Kaplan Meier survival curves for CABG patients, subdivided into early and delayed stroke versus control group. A total of 2341 death events occurred among 7839 CABG patients. respectively. The corresponding figures at 1-year follow-up were 5.46 (CI 3.13/9.53) and 3.76 (CI 1.82/7.75), respectively. In the long term, the HR asymptotically reached the levels presented in Table 6, with reference to the entire follow-up period. Patients with early stroke showed an apparently higher early mortality than their delayed counterparts (Fig. 2). However, when this assumption was tested and the calculation was restrained to the first postoperative year, this potential difference was not confirmed (P = 0.100). The lack of significance might reflect the limited numbers of stroke patients and mortality events. It is obvious that mortality in near association with the stroke event was substantial. To analyze the effect beyond the acute phase, the Cox regression was repeated to include only patients who survived their first postoperative year. The revised cohort accounted for 98% of all CABG patients. In this analysis, the occurrence of delayed stroke remained an independent predictor of mortality with HR being 1.71 (CI 1.24/2.36, P = 0.001). This was in contrast to early stroke subjects, among whom the accumulated mortality did not differ against the control group (HR 1.07, CI 0.768/1.49, P = 0.695). The curve patterns in Fig. 2, beyond 1-year follow-up, illustrate these findings. 4. Discussion Neurological complications are a major concern in cardiac surgery, both for patients as well for health care in general. Our study was designed to identify underlying risk factors behind surgery-related stroke and to evaluate the impact of stroke on patient survival, as hypothesized. A difference between early and delayed stroke was anticipated in our hypothesis and these two groups were here separated. Our findings confirmed the prelaid hypothesis in showing significant differences between patients who experience early stroke from those who developed delayed stroke. Several demographic risk factors behind stroke have been suggested in previous studies [3,16]. CPB exposure has also been addressed, in relation to potential OPCAB benefits [4]. However, it can be speculated that many of these risk factors depend on how the stroke variable was defined and handled. In the majority of previous studies, including large cohort analyses, stroke was not subdivided into early and delayed form [4,16]. In our analysis, early and delayed stroke did not share any independent risk factors. The difference was statistically confirmed when the two stroke groups were tested against each other. Atherosclerosis and its destructive effect on the aortic wall is a known cause for early postoperative stroke. Severe aortic calcification was here associated with a fivefold risk for early stroke. Raised creatinine levels, even moderately, predicted early stroke, through a potential atherosclerotic mechanism. Early stroke, but not delayed stroke, was associated with surgery that is more complex. CPB time was an independent risk factor, rather than the type of surgery. CABG patients who underwent CPB had an apparent higher stroke rate compared with those who underwent OPCAB. This issue has been previously addressed [4,5]. Nevertheless, when background demographic variables were accounted for, the potential benefit of OPCAB was not confirmed in our study. The OPCAB patients did not share all of the intra-operative variables considered in the two other surgery groups. This fact was adjusted for in the calculations, although this correction did not change the results. However, it must be emphasized that OPCAB patients were few and the majority of OPCAB patients at our unit had a low-risk profile. Delayed stroke had a dual risk pattern, with both a demographic and a postoperative influence. Intriguingly observed, female subjects presented a twofold risk for stroke compared with males, an observation that is supported by previous reports [1,6]. The importance of prior cerebrovascular disease in predicting delayed stroke has also been historically reported [1]. However, it is a challenge to explain the role of unstable angina emerging in our statistical analysis as a risk factor. It can be speculated that anticoagulationtherapy withdrawal may have rebound effects during the postoperative period with thrombotic complications. Of the two postoperative risk factors, atrial fibrillation is intuitively understood as an embolic mechanism. The other factor, inotropic requirement, is also known to induce atrial fibrillation [17]. Nevertheless, we found that inotropic requirement was statistically independent of atrial fibrillation, the former more likely reflecting a poor cardiac function. Hogue et al. reported that atrial fibrillation, in combination with low cardiac output, was associated with delayed stroke [1], and their findings are in line with ours. Delayed stroke typically occurred in the early postoperative period, with a peak occurring in the third postoperative day. This observation is similar to that previously observed [5]. From the time distribution of delayed stroke, it can be presumed that the great majority of potential delayed-stroke events occurred within the time

8 386 M. Hedberg et al. / European Journal of Cardio-thoracic Surgery 40 (2011) of hospitalization. Delayed stroke has no generally accepted definition, and here referred to the period of hospitalization only. Our survival analysis focused on CABG patients only. In overview, the results illustrated dramatic effects of stroke on mortality, in particular in the early postoperative period. In multivariable analysis, patients with early and delayed stroke had an increased mortality risk (HR 1.44, CI 1.10/1.89 and HR 1.85, 1.39/2.46, respectively). These rather moderate hazards refer to the long-term comparison against control subjects. In the long-term perspective, other causes of death affect these hazards in the multivariable model. On the other hand, in short-term perspectives stroke was associated with a substantial mortality. This was statistically confirmed by means of a step-wise shortening of the follow-up time. At 6 months, the corresponding HRs indicated a 12-fold and sevenfold increased risk of death for early and delayed stroke, respectively. This phenomenon has statistical implications when the results from other studies are to be compared. A short follow-up generates a higher hazard from the impact of stroke-related early mortality. Early stroke showed a trend toward a higher mortality in the acute postoperative period compared with delayed stroke. However, this finding was not statistically confirmed. A type-ii error must be considered in view of the limited number of stroke events. The influence of stroke on mortality diminished for patients who survived their first postoperative year. Nevertheless, patients suffering from delayed stroke continued to have a higher long-term mortality compared with stroke-free controls, as evident from multivariable testing. By contrast, early stroke patients surviving their first postoperative year had a near-identical survival expectancy compared to that of control subjects. This finding indicates a potential difference between the two stroke groups, although challenging to interpret. It can be speculated that delayed stroke is affected by postoperative mechanisms that continue to influence the long-term survival. This goes in parallel with the observed difference of independent risk factors between the two stroke groups. An example for delayed stroke is atrial fibrillation and embolic events, which may reoccur after discharge. On the other hand, early stroke largely reflects surgical events to which the patient is not further exposed. In analyzing survival after stroke, the number and appropriateness of available parameters are of importance, which may explain inconsistencies between studies. Moreover, the subdivision into two stroke groups must also be considered. Dacey et al. reported a HR of 3.2 at 10-year follow-up, although early and delayed stroke was not subdivided in their analysis [9]. This risk level is slightly higher compared with our findings. There are additional and similar studies available in terms of survival rates. Filsoufi et al. reported survival data up to 5 years [6], which are in accordance with our results. Again, their results referred to overall stroke without subdivision into early and delayed form. The knowledge about survival beyond 10-year followup is rare, which justifies our study Study limitations Stroke is a complex clinical problem and involves multiple aspects. This study analyzed survival outcome but lacked information on functional status and quality of life. Possible stroke events occurring after discharge were not considered. Moreover, our study had no data regarding the anatomical localization of the lesions. It was also limited by its retrospective design, although data were collected in a prospective fashion. Furthermore, database studies do not reach beyond the quality of their documented data. This was partly compensated for by reviewing clinical records of all patients with neurological complications reported in the database. Noteworthy is that the logistic stroke model yielded a moderate explanatory level. This may illustrate two phenomena: either the available parameters were not sensitive enough to explain stroke fully, or, the event may just reflect bad luck, regretted by the surgeon and suffered by the patient. 5. Conclusions Postoperative stroke is a devastating complication to cardiac surgery. It not only devalues successful surgical results but also severely affects survival. Mortality in association with early stroke mainly occurred in the acute phase, whereas, for delayed stroke, the mortality was increased also beyond the first postoperative year. Of specific importance, different risk factors contribute to early and delayed stroke. It is here emphasized that early and delayed stroke should be considered as two separate entities. References [1] Hogue Jr CW, Murphy SF, Schechtman KB, Dávila-Román VG. Risk factors for early or delayed stroke after cardiac surgery. Circulation 1999;100: [2] Wolman RL, Nussmeier NA, Aggarwal A, Kanchuger MS, Roach GW, Newman MF, Mangano CM, Marschall KE, Ley C, Boisvert DM, Ozanne GM, Herskowitz A, Graham SH, Mangano DT. Cerebral injury after cardiac surgery: identification of a group at extraordinary risk. Multicenter Study of Perioperative Ischemia Research Group (McSPI) and the Ischemia Research Education Foundation (IREF) Investigators. Stroke 1999;30: [3] Borger MA, Ivanov J, Weisel RD, Rao V, Peniston CM. Stroke during coronary bypass surgery: principal role of cerebral macroemboli. Eur J Cardiothorac Surg 2001;19: [4] Bucerius J, Gummert JF, Borger MA, Walther T, Doll N, Onnasch JF, Metz S, Falk V, Mohr FW. Stroke after cardiac surgery: a risk factor analysis of 16,184 consecutive adult patients. Ann Thorac Surg 2003;75: [5] Peel GK, Stamou SC, Dullum MK, Hill PC, Jablonski KA, Bafi AS, Boyce SW, Petro KR, Corso PJ. Chronologic distribution of stroke after minimally invasive versus conventional coronary artery bypass. J Am Coll Cardiol 2004;43: [6] Filsoufi F, Rahmanian PB, Castillo JG, Bronster D, Adams DH. Incidence, topography, predictors and long-term survival after stroke in patients undergoing coronary artery bypass grafting. Ann Thorac Surg 2008;85: [7] Roach GW, Kanchuger M, Mangano CM, Newman M, Nussmeier N, Wolman R, Aggarwal A, Marschall K, Graham SH, Ley C. Adverse cerebral outcomes after coronary bypass surgery. Multicenter Study of Perioperative Ischemia Research Group and the Ischemia Research and Education Foundation Investigators. N Engl J Med 1996;335: [8] Boivie P, Edström C, Engström KG. Side differences in cerebrovascular accidents after cardiac surgery: a statistical analysis of neurologic symptoms and possible implications for anatomic mechanisms of aortic particle embolization. J Thorac Cardiovasc Surg 2005;129: [9] Dacey LJ, Likosky DS, Leavitt BJ, Lahey SJ, Quinn RD, Hernandez Jr F, Quinton HB, Desimone JP, Ross CS, O Connor GT. Perioperative stroke and

9 M. Hedberg et al. / European Journal of Cardio-thoracic Surgery 40 (2011) long-term survival after coronary bypass graft surgery. Ann Thorac Surg 2005;79: [10] Salazar JD, Wityk RJ, Grega MA, Borowicz LM, Doty JR, Petrofski JA, Baumgartner WA. Stroke after cardiac surgery: short- and long-term outcomes. Ann Thorac Surg 2001;72: [11] Filsoufi F, Rahmanian PB, Castillo JG, Bronster D, Adams DH. Incidence, imaging analysis, and early and late outcomes of stroke after cardiac valve operation. Am J Cardiol 2008;101: [12] Edmunds Jr LH, Clark RE, Cohn LH, Grunkemeier GL, Miller DC, Weisel RD. Guidelines for reporting morbidity and mortality after cardiac valvular operations. J Thorac Cardiovasc Surg 1996;112: [13] Special report from the National Institute of Neurological Disorders and Stroke. Classification of cerebrovascular diseases III. Stroke 1990;21: [14] Thorvaldsen P, Asplund K, Kuulasmaa K, Rajakangas AM, Schroll M. Stroke incidence, case fatality, and mortality in the WHO MONICA project. World Health Organization Monitoring Trends and Determinants in Cardiovascular Disease. Stroke 1995;26: [15] Hedberg M, Boivie P, Edström C, Engström KG. Cerebrovascular accidents after cardiac surgery: an analysis of CT scans in relation to clinical symptoms. Scand Cardiovasc J 2005;39: [16] Janssen DPB, Noyez L, van Druten JA, Skotnicki SH, Lacquet LK. Predictors of neruological morbidity after coronary bypass surgery. Eur J Cardiothorac Surg 1999;15: [17] Mariscalco G, Engström KG. Are current smokers paradoxically protected against atrial fibrillation after cardiac surgery? Nicotine Tob Res 2009;11: Editorial comment Does the time of onset of postoperative stroke determine outcome? Keywords: Coronary bypass surgery; Stroke; Outcomes research The preceding article is a retrospective analysis of almost 9000 cardiac procedures performed during a 10-year period [1]. The records of patients with any signs of neurological complications were examined. The authors found 149 patients who exhibited signs of stroke at extubation versus 99 patients having a free interval prior to stroke symptoms while still in the hospital. Independent risk factors for stroke were analyzed by specific statistical methods as were survival data calculated from the Swedish population registry. Their conclusions were that stroke was a significant risk factor for death in the perioperative period. Strokes which occurred early did not influence long-term mortality which was in contrast to delayed stroke which appeared to have a significant effect on mortality persisting to 9+ years. On the surface, the authors conclusions about early and delayed strokes appear intuitive. Early stroke would be caused by events precipitated by surgical activity in the mediastinum, usually aortic manipulation [2]. Brain damage from embolization appearing in the perioperative period, on the other hand, is thought to be more related to the overall severity of patient s disease, especially atherosclerosis of the ascending aorta and brachiocephalic vessels and atrial arrhythmias [3]. With the delayed patients having more severe disease, their long-term life expectancy would be expected to be worse. Recent information, however, would support that many cases of delayed stroke are due to the combination of surgical manipulation and atherosclerosis. A paper from Japan demonstrated that epiaortic ultrasound examination of the ascending aorta and a portion of the transverse arch prior to closing the chest in routine cardiac surgery often found evidence of unstable atherosclerotic plaques, poorly closed cannulation sites and irregular anastomotic sites from coronary bypass grafts and aortotomies which could be the source of thrombotic or atherosclerotic embolic material [4]. Since the authors apparently did not routinely screen patients for these etiologies, we do not know what role they played in the pathogenesis of delayed stroke. It was of interest that in the authors analysis of risk factors, the risk factors for early and delayed stroke were almost identical with the exception of atrial fibrillation being a risk factor in the delayed group. In the first line of the abstract, the authors suggest stroke is a uniform disease which it is not. In an awake person the cerebral cortex receives approximately four times the blood flow as the white matter and subcortical structures. Therefore infarcts are commonly found in the cerebral watershed areas. During hypothermic cardiopulmonary bypass, there is little difference between cortical and subcortical flow. Emboli tend to take the path of least resistance and take the first major vessels off the Circle of Willis, the lenticulostriate arteries, and cause lesions in the structures involved in postural background movement and memory. The distribution of lesions is different for intraoperative stroke and the size and severity of the lesion is affected by the temperature and glucose levels at the time of the infarction. In the immediate post-operative period, the brain is swollen and the vessels demonstrating evidence of a systemic inflammatory response related to the passage of gaseous and lipid microemboli [5]. Therefore the internal environment is considerably different during the peri-operative period which affects the dimensions and presentation of a focal neurologic insult. The authors should be congratulated for their excellent results in this relatively low risk group of patients. The presence of such short cross clamp times would indicate that the authors probably used a multiple clamping protocol for patients receiving on-pump coronary bypass procedures. It has been shown in a randomized study that multiple clamping is associated with higher numbers of emboli and a poorer neuropsychological score postoperatively [6]. In addition,