Department of Thoracic and Cardiovascular Surgery, West-German Heart Center, University of Duisburg-Essen, Essen, Germany

Society of Thoracic Surgeons Score Is Superior to the EuroSCORE Determining Mortality in High Risk Patients Undergoing Isolated Aortic Valve Replacement Daniel Wendt, MD, Brigitte R. Osswald, MD, Katrin Kayser, Matthias Thielmann, MD, Paschalis Tossios, MD, Parwis Massoudy, MD, Markus Kamler, MD, and Heinz Jakob, MD Department of Thoracic and Cardiovascular Surgery, West-German Heart Center, University of Duisburg-Essen, Essen, Germany Background. Major scores for the evaluation of procedural risk in cardiac surgery are the European system for cardiac operative risk evaluation score (EuroSCORE), the Society of Thoracic Surgeons (STS) score, and the Parsonnet score. The aim of our study was to analyze the predictive value of these scores in high risk patients undergoing isolated aortic valve replacement (AVR). Methods. Six hundred and fifty-two patients underwent isolated AVR from January 1999 through June 2007. Emergency and redo operations were included; acute endocarditis was excluded. Evaluation was performed by logistic regression analysis. Data collection was prospective. Results. The mean logistic EuroSCORE of all patients was 8.5 7.9%, the mean STS score was 4.4 3.9%, and the mean logistic Parsonnet score was 9.8 8.5%. Inhospital mortality was 2.5% (n 16). Freedom from all-cause death was 93.4% at 1 year, 90.2% at 2 years, and 75.8% at 5 years, respectively. A total of 182 patients had a logistic EuroSCORE greater than 10. For the group of patients with a EuroSCORE between 10% and 20% (n 130) the mean EuroSCORE was 13.9 2.8% and the STS score was 6.5 3.8%. Observed mortality was 4.6% in this group. For the 52 patients with a logistic EuroSCORE of at least 20 (mean 28.5 10.3%, STS score 10.1 7.3%) the observed mortality was 3.9% (n 2). By stepwise logistic regression, none of the EuroSCORE variables could be identified as an independent predictor in the high- risk group. Conclusions. The logistic EuroSCORE was primarily created to allow patient grouping for the total spectrum of cardiac surgery. In patients undergoing isolated AVR, the EuroSCORE highly overestimates mortality, whereas the STS score seems to be actually more suitable in assessing perioperative mortality for these patients. (Ann Thorac Surg 2009;88:468 75) 2009 by The Society of Thoracic Surgeons Aortic stenosis (AS) is the most frequent acquired valve disease in Europe. Surgical aortic valve replacement (AVR) is the treatment of choice in symptomatic patients with severe aortic stenosis according to the current guidelines [1, 2]. Surgical AVR represents the gold standard with approximately 275,000 procedures annually worldwide [3]. Various standardized risk scoring algorithms exist for the preoperative stratification of patients according to their expected surgical risk. These risk scores are based on patients preoperative cardiac and noncardiac status. Traditionally, risk scores were used to estimate perioperative mortality in surgical patients cohorts. Recently, the same scores have been used to identify high-risk subgroups of Accepted for publication April 14, 2009. Presented at the Forty-fifth Annual Meeting of The Society of Thoracic Surgeons, San Francisco, CA, Jan 26 28, 2009. Address correspondence to Dr Wendt, Department of Thoracic and Cardiovascular Surgery, West-German Heart Center Essen, University Hospital Essen, Hufelandstraße 55, Essen, 45122, Germany; e-mail: daniel.wendt@uk-essen.de. patients in order to offer an alternative treatment option even for the individual patient. With increasing comorbidities such as advanced age, poor left ventricular function, chronic obstructive pulmonary disease, or renal dysfunction, some of these high-risk patients were deemed too sick for surgery and more than 30% remain untreated [4]. Therefore, transcatheter aortic valve implantation techniques (TAVI) have evolved as an endovascular or minimally invasive alternative to offer treatment options for patients whose surgical risk is reported as too high [5 7]. Widely used scores for the evaluation of procedural risk in cardiac surgery are the European system for cardiac operative risk evaluation (EuroSCORE) and the Society of Thoracic Surgeons predicted risk of mortality (STS-PROM), but even risk calculation using the older Parsonnet score is performed [8 10]. Recently, the STS risk algorithm was reported to be the most sensitive score in defining the risk of patients undergoing isolated AVR [11]. Despite limitations in actual scoring systems an accurate and AVR specific risk scoring tool, particularly for high-risk patients, is still missing. This study aimed to 2009 by The Society of Thoracic Surgeons 0003-4975/09/$36.00 Published by Elsevier Inc doi:10.1016/j.athoracsur.2009.04.059

Ann Thorac Surg WENDT ET AL 2009;88:468 75 STS VS EUROSCORE FOR ISOLATED AVR 469 Table 1. Baseline Patient Characteristics Variable All Patients EuroSCORE 10 20 EuroSCORE 20 (n 652) (n 130) (n 52) p Value a Demographics: Age, years 67.7 11.5 73.4 8.9 75.6 8.3 0.18 Age 70 years 311 (47.8%) 96 (73.9%) 43 (82.7%) 0.25 Age 80 years 70 (10.8%) 26 (20.0%) 17 (32.7%) 0.08 Gender, female 293 (45.0) 65 (50.0%) 28 (53.8%) 0.74 Weight, kg 77.9 14.8 76.2 14.5 72.0 12.0 0.009 Height, cm 169.6 9.5 168.1 8.7 167.4 9.0 0.47 BMI, kg/m 2 27 4.5 26.9 4.6 25.6 3.2 0.20 Obesity (BMI 30 kg/m 2 ) 140 (21.5%) 27 (20.8%) 3 (5.8%) 0.014 Risk factors & comorbidities: NYHA class 2.8 0.7 2.9 0.7 3.1 0.6 0.70 Systemic hypertension 473 (72.7%) 108 (83.1%) 45 (86.5%) 0.65 Diabetes mellitus 140 (21.5%) 39 (30.0%) 14 (26.9%) 0.72 Coronary artery disease 149 (22.9%) 46 (35.4%) 20 (38.5%) 0.73 Peripheral vascular disease 77 (11.8%) 23 (17.7%) 27 (51.9%) 0.001 Hyperlipidemia 260 (39.9%) 62 (47.7%) 13 (25.0%) 0.007 Prior stroke 49 (7.5%) 18 (13.9%) 10 (19.2%) 0.37 Neurologic disorders 58 (8.9%) 19 (14.6%) 12 (23.1%) 0.19 COPD 154 (23.7%) 38 (29.2%) 20 (38.5%) 0.19 Renal disease ( 200 mol/l) 123 (18.9%) 39 (30.0%) 19 (36.5%) 0.48 Emergency 5 (0.8%) 4 (3.1%) 1 (1.9%) 1.00 Smoking 324 (49.8%) 74 (56.9%) 16 (30.8%) 0.002 Cardiac history: LVEF 0.572 0.141 0.539 0.158 0.476 0.148 0.008 Preoperative AVA 0.7 0.21 0.69 0.20 0.64 0.29 0.64 Pulmonary hypertension 51 (7.8%) 25 (19.2%) 9 (17.3%) 0.84 Reoperation 47 (7.2%) 24 (18.5%) 14 (26.9%) 0.23 Atrial fibrillation 180 (27.7) 43 (33.1%) 25 (48.1%) 0.06 Risk scores: Additive EuroSCORE, % 6.45 2.6 8.9 0.8 11.4 1.2 0.001 Logistic EuroSCORE, % 8.5 7.9 13.9 2.8 28.5 10.3 0.001 STS-PROM, % 4.4 3.9 6.5 3.8 10.1 7.3 0.001 Additive Parsonnet score, % 17.8 8.9 24.2 8.7 25.8 7.4 0.05 Logistic Parsonnet score, % 9.8 8.54 15.9 11.2 17.8 11.1 0.002 a EuroSCORE 10 to 20 versus EuroSCORE 20. Data are presented as mean SD or number (%). AVA aortic valve area; BMI body mass index; COPD chronic obstructive pulmonary disease; EF ejection fraction; EuroSCORE European system for cardiac operative risk evaluation; LV left ventricle; NYHA New York Heart Association; STS-PROM Society of Thoracic Surgeons predicted risk of mortality. analyze the predictive value of the following: (1) the EuroSCORE, (2) the STS score, and (3) the Parsonnet risk scoring systems in high-risk patients undergoing isolated AVR, with regard to early outcomes. A long-term follow-up was also performed. Patients and Methods Patients Between January 1999 and June 2007, 652 consecutive patients underwent isolated AVR for pure or predominant aortic valve stenosis at the West German Heart Center, University Hospital Essen, Essen, Germany. The present study obtained Institutional Review Board approval according to the Declaration of Helsinki. Patients were excluded from the study if significant aortic regurgitation was predominant, if they had active endocarditis, or if concomitant procedures like coronary artery bypass grafting were performed. Reoperations or emergency operations, additional myectomy, aortic root enlargement in order to prevent patient-prosthesis mismatch, or simple wrapping-plication of the ascending aorta were included. Demographics and preoperative characteristics of this cohort are listed in Table 1.

470 WENDT ET AL Ann Thorac Surg STS VS EUROSCORE FOR ISOLATED AVR 2009;88:468 75 Data Collection and Follow-Up Patient and operative demographics were recorded in a prospective institutional database and retrospectively extracted and evaluated. Perioperative deaths (30-day mortality) were tracked from the institutional database. Long-term survival was obtained by active follow-up. Calculation of completeness of follow-up was 99.2%. Risk Calculation and Definition of Risk Groups The EuroSCORE calculator available online (http://www. euroscore.org) was used for both additive (AES) and logistic EuroSCORE (LES) calculations. The EuroSCORE considers 17 variables [10]. All study patients had an isolated AVR and therefore fulfilled the item operation other than isolated coronary artery bypass grafting, whereas endocarditis was excluded. The STS-PROM mortality risk calculation for aortic valve procedures was performed by the online available STS score calculator (http://66.89.112.110/stswebriskcalc261/de.aspx). The additive (APS) and logistic Parsonnet (LPS) score research calculator available online was used (http:// www.sfar.org/scores2/parsonnet2.html) [9]. All scores were calculated for each of the 652 patients. All variables and items were calculated and considered according to their exact definition of each score. The calculated scores were used to evaluate early mortality. According to the definition of the scores, 30- day in hospital mortality was evaluated. Receiver operating curves were calculated for all scores. Based on the EuroSCORE risk calculation, patients were divided into high-risk (logistic EuroSCORE between 10% and 20%) and excessive risk (logistic EuroSCORE 20%) groups. Receiver Operating Characteristic (ROC) Curve Analysis Receiver operator curves were generated for all risk scoring systems. Due to the fact that the estimates describing the loading of each item within each score were not available for the proprietary STS score, only absolute values of all scores have been considered for the ROC analysis. Sensitivity and specificity of expected versus observed mortality were summarized by receiver operator curves and the area under the resulting curve (AUC). A decreasing value of this statistic from 1.0 toward 0.5 indicates decreasing distinctiveness or discrimination between patients living and dead. Results Table 2. Risk Calculations Variable All Patients EuroSCORE 10 20 EuroSCORE 20 (n 652) (n 130) (n 52) Observed mortality n (%), CI 16 (2.5), 1.4 3.9 6 (4.6), 1.7 9.8 2 (3.9), 0.5 13.2 Additive EuroSCORE: Mean (minimum-maximum) 6.45 (1 15) 8.9 (7 11) 11.4 (10 15) Expected mortality, n, % CI 42.1, 4.6 8.3 11.5, 4.0 13.8 6.0, 4.4 23.4 2 -value, p 0.0001 0.08 0.08 Observed-expected mortality 0.38 0.52 0.33 Logistic EuroSCORE: Mean (minimum-maximum) 8.46 (1.2 66.4) 13.9 (10.1 19.7) 28.5 (20.01 66.4) Expected mortality, n, % CI 55.2, 6.3 10.6 18.7, 8.4 20.4 14.8, 16.2 40.7 2 -value, p 0.0001 0.0015 0.0001 Observed-expected mortality 0.29 0.32 0.14 STS-PROM: Mean (minimum-maximum) 4.4 (0 33.8) 6.5 (1.5 31.4) 10.1 (1.7 33.8) Expected mortality, n, % CI 28.7, 2.8 6.0 8.5, 2.3 10.8 5.3, 2.0 18.4 2 -value, p 0.015 0.38 0.13 Observed-expected mortality 0.56 0.71 0.38 Additive Parsonnet score: Mean (minimum-maximum) 17.8 (5 53) 24.2 (6 53) 25.8 (9 42) Expected mortality, n, % CI 116.1, 14.9 20.7 31.5, 16.9 31.6 13.4, 13.9 37.7 2 -value, p 0.0001 0.0001 0.0003 Observed-expected mortality 0.14 0.19 0.15 Logistic Parsonnet score: Mean (minimum-maximum) 9.9 (0.3 52.1) 15.9 (1.6 52.1) 18.0 (1.6 50.2) Expected mortality, n, % CI 64.6, 7.6 12.2 20.7, 9.6 22.2 9.4, 7.6 28.5 2 -value, p 0.0001 0.0004 0.0077 Observed-expected mortality 0.25 0.29 0.21 Data are presented as number (%). CI confidence interval; EuroSCORE European system for cardiac operative risk evaluation; STS-PROM Society of Thoracic Surgeons predicted risk of mortality.

Ann Thorac Surg WENDT ET AL 2009;88:468 75 STS VS EUROSCORE FOR ISOLATED AVR 471 Fig 1. Ratio of estimated versus observed mortality. Patients were divided into three groups based on the logistic EuroSCORE (all patients, high-risk 10 to 20 and excessive risk 20). (AES additive EuroSCORE; APS additive Parsonnet score; LES logistic EuroSCORE; LPS logistic Parsonnet score; STS Society of Thoracic Surgeons). were given as AUC accompanied by 95% confidence intervals. Statistics Descriptive statistics are summarized for categoric variables as frequencies (%) and compared between groups using the Pearson 2 exact test. Continuous variables were reported as mean standard deviation and were compared between groups using the Student t test. Observed and expected number of events for the groups were compared using the Pearson 2 or Fisher exact tests as appropriate. A p value less than 0.05 was considered to indicate statistical significance. Survival curves were generated with the Kaplan-Meier method [12]. All statistical analyses were performed using SAS version 9.1 (SAS Inc, Cary, NC). Statement of Responsibility The authors had full access to the data and take full responsibility for their integrity. All authors have read and agreed to the manuscript as written. Results Risk Scores and Calculations A total of 652 patients were identified as having undergone isolated AVR January 1999 and June 2007. Mean age was 67.7 11.5 years, 45.1% (n 293) were female, 23.7% (n 154) had chronic obstructive lung disease, 11.8% (n 77) had extracardiac arteriopathy, 8.9% (n 58) had preoperative neurologic dysfunction or disorders, 0.8% (n 49) of the patients had prior stroke, 7.2% (n 47) had previous cardiac surgery, 18.9% (n 123) of the patients had a serum creatinine greater than 200 mol/l, 0.9% (n 6) of the patients were in a critical preoperative state, and 1.0% (n 7) of the patients had a recent myocardial infarction. The mean ejection fraction was 0.572 0.141 and 78% (n 51) of the patients had pulmonary hypertension. Emergency operation was performed in 0.8% (n 5) of the patients, and no patient had an aneurysm repair or postinfarction ventricular septum repair. Patients demographics, risk factors, and comorbidities are presented in Table 1. Preoperative risk assessment of the complete cohort revealed an additive and logistic EuroSCORE of 6.45 2.6% and 8.46 7.9%, a STS score of 4.4 3.94%, and a mean additive and logistic Parsonnet score of 17.8 5.2 and 9.9 8.54. Thirty-day in-hospital mortality for the entire group was 2.5% (16 of 652 patients). Patients with a LES less than10 had an observed mortality of 1.7% (8 of 470). A total of 130 patients could be identified as having a logistic EuroSCORE greater than 10% and less than 20%. Within this group, estimated mortality ranged from 6.5% for the STS-PROM calculation to 8.9% for the AES and 13.9% for the LES, to 20.7% for the LPS and 24.2% for the APS. Thirty-day in hospital mortality for this group was 4.6% (6 of 130 patients). Within the patients at excessive risk as defined by an LES greater than 20% (n Fig 2. Kaplan-Meier survival curves for patients with a logistic EuroSCORE between 10 and 20 (black bold line) including 70% confidence interval (light dashed line). Age- and gender-matched German population (grey bold line). Patients at risk. (AVR aortic valve replacement.)

472 WENDT ET AL Ann Thorac Surg STS VS EUROSCORE FOR ISOLATED AVR 2009;88:468 75 Fig 3. Kaplan-Meier survival curves for patients with a logistic EuroSCORE greater than 20 (black bold line) including 70% confidence interval (light dashed line). Age- and gendermatched German population (grey bold line). Patients at risk. (AVR aortic valve replacement.) 52 patients), the predicted mortality ranged from 10.1% as calculated by the STS-PROM algorithm to 11.4% for the AES and 18% for the LPS, to 25.8% for the APS and 28.5% for the LES calculations. Observed 30-day in hospital mortality was 3.9% in this group (2 of 52 patients). The expected mortality by numbers within the excessive risk group (logistic EuroSCORE 20) was 6.0 by the AES (p 0.08), 14.8 by the LES (p 0.0001), 5.3 by the STS-PROM (p 0.13), 13.4 by the APS (p 0.0003), and 9.4 by the LPS (p 0.0077), whereas only 2 patients died (3.9%). Furthermore, the observed-expected ratios for perioperative mortality in this group were 0.33, 0.14, 0.38, 0.15, and 0.21 as predicted by the AES, LES, STS, APS, and LPS, respectively. Preoperative risk calculations by each scoring system are shown in Table 1. All risk scores overestimated the risk for mortality. However, the STS-PROM algorithm showed the most accurate prediction for mortality, especially in the group presenting a LES greater than 20, whereas the logistic EuroSCORE revealed the highest overestimation of mortality (mean LES of 28.5%). The observed to expected ratios for perioperative mortality for each group calculated by the different algorithms are shown in Table 2 and Figure 1. ROC Analysis The AUC was 69.9 for the logistic EuroSCORE (confidence interval [CI] 60.9 to 79.0) and 69.9 for the additive EuroSCORE (CI 60.4 to 79.4), whereas the STS score showed an AUC of 71.8 (CI 59.3 to 84.3). The AUC for both the logistic and additive Parsonnet score was 69.4 (CI 59.2 to 79.6) and 66.6 (CI 54.6 to 78.7), respectively. Survival Freedom from all-cause death in all patients was 93.4% at 1 year, 90.2% at 2 years, and 75.8% at 5 years, respectively. In the high-risk group survival was 86.1% at 1 year, 80.8% at 2 years, and 54.8% at 5 years, whereas in the excessive risk group survival was 90.1% at 1 year, 84.5% at 2 years, and 53.5% at 5 years. Kaplan-Meier survival curves are given in Figures 2 and 3. Comment The present study clearly demonstrates that the predictive value of many of the currently available scoring systems is insufficient to allow a reliable risk assessment in patients undergoing isolated aortic valve replacement. The overestimation is most prominent in high-risk patients. Risk stratification using the STS score was accurate in predicting the risk of mortality in high-risk patients. Nevertheless, even this most recently built score systematically overestimates procedural risk. From the clinician s standpoint there is a need for an objective risk assessment tool. The increased application of TAVI at our institution has raised the awareness for evaluating outcomes of conventional AVR patients because risk models are currently used, and potentially misused, to create a subgroup of patients at an assumed high or highest risk for conventional AVR. The aim of the present study was therefore to evaluate the predictive value of different risk algorithms in patients undergoing isolated AVR with regard to their outcomes and critically discuss the validity of the scores used. Endocarditis was excluded in our present study. This exclusion was based on mainly two reasons: On the one hand, aortic valve replacement based on endocarditis represents a completely different issue compared with aortic valve replacement based on calcific stenosis and therefore results in different outcomes. On the other hand, endocarditis represents an exclusion criterion in transcatheter valve implantation techniques. Therefore, in order to get equal conditions when talking about surgical AVR and transcatheter techniques, endocarditis was excluded. In the present study the mean additive and logistic EuroSCORE in very high-risk patients with a LES greater than 20 (n 52) were 11.4% and 28.5%, respectively. The STS score for this group was calculated by 10.1%. Observed mortality, however, was 4.6% (6 of 130) in the group presenting a LES between 10% and 20%, and 3.9% (2 of 52) in the excessive risk group (LES 20). In order to control for any bias, and above all to allow discrimination between the used risk scoring systems,

Ann Thorac Surg WENDT ET AL 2009;88:468 75 STS VS EUROSCORE FOR ISOLATED AVR 473 receiver operator curves were calculated for each score. However, due to the proprietary design of the STS score, only the calculated scores without the estimates for each item have been incorporated into the analysis. The STS score showed a moderately higher AUC compared with the logistic and additive EuroSCORE (71.8 vs 69.9), whereas the EuroSCORE nearly reached a good overall predictive value. The lowest c-values were calculated for the additive Parsonnet score (66.6). Both the EuroSCORE and the STS score can be used to accurately predict mortality in patients undergoing isolated AVR; however, the EuroSCORE and Parsonnet score highly overestimate the operative risk of AVR. Other groups have reported excellent and highly convincing results in patients at high-risk for surgery [13 16]. Our overall 30-day in-hospital mortality in patients at high-risk was only 2.5% (despite the presence of peripheral vascular disease in 11.8% of the patients and many patients presenting an impaired left ventricular ejection fraction, which are known as independent predictors of hospital mortality [13]). In addition, mortality within the patients with a LES of at least 20% was only observed by 3.9%, although almost 27% of the patients in this group had previous cardiac surgery. Risk overestimation by the EuroSCORE has been described earlier [13, 16 18]. The STS score was shown to accurately predict mortality in patients undergoing transcatheter aortic valve implantation who had a LES greater than 30% and a STS score greater than 15% [5]. The EuroSCORE was created primarily to allow patient grouping for the total spectrum of cardiac surgery; however, in the EuroSCORE cohort the part of patients undergoing AVR was represented by 17% of the whole cohort [17]. Furthermore, the EuroSCORE is still based on a 1995 mortality across all of cardiac surgery and has yet not been updated or recalibrated. The present study confirms recent reports and analyses that the logistic EuroSCORE clearly overestimates the risk of mortality, especially in patients at very high risk. The logistic EuroSCORE estimated nearly 15 deaths within the excessive risk group compared with an estimated mortality of 5 deaths as calculated by the STS score, whereas only 2 patients died. The excessive risk group had a calculated additive EuroSCORE of 11.4 compared with an observed mortality of 2 out of 52 patients (3.9%) (p 0.08). These findings can be matched with results from Brown and colleagues [17], who observed a mortality of 7.4% in high-risk patients undergoing isolated AVR compared with an estimated mortality of 12.6% (p 0.109) by the use of the additive EuroSCORE. Thus, using the currently available algorithms, risk overestimation for surgical AVR is common [8, 10, 19]. None of the available scores seem adequate to justify transcatheter aortic valve implantation because of a suggested high risk. With this information in mind transcatheter aortic valve implantation studies in transcatheter patients with a mean logistic EuroSCORE of 11% or even as high as 27% may appear in a different light [6, 20]. However, especially considering our own policy for the indication of transcatheter aortic valve implantation, many of the patients carry risk factors that are not depicted by the current risk scoring systems. In our hands, patients undergoing conventional AVR and patients undergoing transcatheter aortic valve implantation are completely separate subgroups. To establish an indication for TAVI, factors like frailty, multiple heart valve disease, end stage liver disease, etc, may have to be considered. However, not only verifiable model results will serve as the sole determinant of patient risk but many other objective and subjective tests evaluating patients mobility, frailty, quality of life, housing support, or social integration must be taken into account when evaluating the individual risk. Furthermore, the so-called eyeball test of the experienced surgeon will be useful in the preoperative evaluation of patients risk. The present study thus shows that most current scores systematically overestimate the risk of patients undergoing conventional AVR, even for an individual patient. The current scores, therefore, do not seem to be adequate to identify patients not suitable for conventional AVR because of a high risk. The algorithms need updating for an apparent improvement in surgical results, especially in patients at higher risk. The explanation for the systematic overestimation of scores is, on the one hand, the lack of a more specific AVR score as is true for the EuroSCORE and the Parsonnet score. As shown earlier by our group, the application of the STS score in patients undergoing transcatheter aortic valve implantation seems to adequately predict mortality in these patients but may still not be used to establish an indication for transcatheter aortic valve implantation [5]. Furthermore, the data used for risk estimation require constant recalculation, which is achieved recently only by the STS score, whereas the EuroSCORE and the older Parsonnet score have not as yet been updated or recalibrated. As well, the current clinical practice and all of the present most modern therapeutic options should be implemented in actual scoring systems. The presented data clearly demonstrate that the EuroSCORE (above all the logistic EuroSCORE) risk calculation overestimates mortality in patients undergoing isolated aortic valve implantation and mainly in those patients at high risk. Out of the available scoring systems the STS score, as to be explained from the evaluation modalities, represents the most accurate score although still an overestimation of risk is observed. Limitations The present study was performed at a single tertiary care medical center. Hence, the generalizability of our findings may not extend to all centers worldwide. The cutoff point for determining excessive risk was set arbitrarily at a logistic EuroSCORE greater than 20, resulting in only a small number of patients (n 52). Furthermore, one can assume, according to the European experience, that nearly one third of patients (primarily patients at high or highest risk) were not referred to surgery [4].

474 WENDT ET AL Ann Thorac Surg STS VS EUROSCORE FOR ISOLATED AVR 2009;88:468 75 Conclusion In patients undergoing isolated AVR, scores evaluated in the present study overestimate the risk of perioperative mortality. The highest over prediction could be observed by the logistic EuroSCORE algorithm, whereas the STS score seems to be more suitable in calculating operative mortality in patients undergoing isolated AVR. By using these scores the definition of a high-risk patient population suitable for TAVI does not seem to be justified. The results of conventional AVR are excellent, even in highrisk patients. Nevertheless, an indication for transcatheter aortic valve implantation exists, which must, however, be better defined by considering population specific indications. References 1. Bonow RO, Carabello BA, Chatterjee K, et al. ACC/AHA 2006 Guidelines for the Management of Patients With Valvular Heart Disease: A Report of the American College of Cardiology/ American Heart Association Task Force on Practice Guidelines (Writing Committee to Revise the 1998 Guidelines for the Management of Patients With Valvular Heart Disease): Developed in Collaboration With the Society of Cardiovascular Anesthesiologists: Endorsed by the Society for Cardiovascular Angiography and Interventions and the Society of Thoracic Surgeons. Circulation 2006;114:e84 231. 2. Iung B, Baron G, Butchart EG, et al. A prospective survey of patients with valvular heart disease in Europe: The Euro Heart Survey on Valvular Heart Disease. Eur Heart J 2003; 24:1231 43. 3. Ambler G, Omar RZ, Royston P, Kinsman R, Keogh BE, Taylor KM. Generic, simple risk stratification model for heart valve surgery. Circulation 2005;112:224 31. 4. Iung B, Cachier A, Baron G, et al. Decision-making in elderly patients with severe aortic stenosis: why are so many denied surgery? Eur Heart J 2005;26:2714 20. 5. Thielmann M, Wendt D, Kahlert P, et al. Transcatheter off-pump aortic valve implantation in patients with very high risk for conventional aortic valve replacement. Circulation 2008;118:S944 5 (Abstract). 6. Walther T, Falk V, Borger MA, et al. Minimally invasive transapical beating heart aortic valve implantation proof of concept. Eur J Cardiothorac Surg 2007;31:9 15. 7. Webb JG, Pasupati S, Humphries K, et al. Percutaneous transarterial aortic valve replacement in selected high-risk patients with aortic stenosis. Circulation 2007;116:755 63. 8. Nashef SA, Roques F, Michel P, Gauducheau E, Lemeshow S, Salamon R. European system for cardiac operative risk evaluation (EuroSCORE). Eur J Cardiothorac Surg 1999;16: 9 13. 9. Parsonnet V, Dean D, Bernstein AD. A method of uniform stratification of risk for evaluating the results of surgery in acquired adult heart disease. Circulation 1989;79(6 pt 2):I3 12. 10. Roques F, Nashef SA, Michel P, et al. Risk factors and outcome in European cardiac surgery: analysis of the Euro- SCORE multinational database of 19030 patients. Eur J Cardiothorac Surg 1999;15:816 23. 11. Dewey TM, Brown D, Ryan WH, Herbert MA, Prince SL, Mack MJ. Reliability of risk algorithms in predicting early and late operative outcomes in high-risk patients undergoing aortic valve replacement. J Thorac Cardiovasc Surg 2008;135:180 7. 12. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457 81. 13. Grossi EA, Schwartz CF, Yu PJ, et al. High-risk aortic valve replacement: are the outcomes as bad as predicted? Ann Thorac Surg 2008;85:102 7. 14. Lund O. Preoperative risk evaluation and stratification of long-term survival after valve replacement for aortic stenosis. Reasons for earlier operative intervention. Circulation 1990;82:124 39. 15. Lund O, Pilegaard H, Nielsen TT, Knudsen MA, Magnussen K. Thirty-day mortality after valve replacement for aortic stenosis over the last 22 years. A multivariate risk stratification. Eur Heart J 1991;12:322 31. 16. Osswald BR, Gegouskov V, Badowski-Zyla D, et al. Overestimation of aortic valve replacement risk by EuroSCORE: implications for percutaneous valve replacement. Eur Heart J 2009;30:74 80. 17. Brown ML, Schaff HV, Sarano ME, et al. Is the European system for cardiac operative risk evaluation model valid for estimating the operative risk of patients considered for percutaneous aortic valve replacement? J Thorac Cardiovasc Surg 2008;136:566 71. 18. Shanmugam G, West M, Berg G. Additive and logistic EuroSCORE performance in high risk patients. Interact Cardiovasc Thorac Surg 2005;4:299 303. 19. Michel P, Roques F, Nashef SA, EuroSCORE Project Group. Logistic or additive EuroSCORE for high-risk patients? Eur J Cardiothorac Surg 2003;23:684 7. 20. Grube E, Laborde JC, Gerckens U, et al. Percutaneous implantation of the CoreValve self-expanding valve prosthesis in high-risk patients with aortic valve disease: the Siegburg first-in-man study. Circulation 2006;114:1616 24. DISCUSSION DR CLIFF K. CHOONG (Melbourne, Australia): I want to congratulate you and your coauthors for this very good and timely study. I have a few comments to make. My coauthors and I have recently published in the January 2009 issue of the European Journal of Cardio-Thoracic Surgery an editorial comment regarding EuroSCORE: The EuroSCORE risk stratification system in the current era: how accurate is it and what should be done if it is inaccurate? [Eur J Cardiothorac Surg 2009;35:59 61]. We are well aware that the EuroSCORE overestimates observed mortality. To understand that, one needs to understand more about the history of EuroSCORE. The additive EuroSCORE risk model was developed utilizing data from 14,781 patients from 128 surgical centers in eight European states who underwent surgery between September and November 1995, and was published in 1999. Subsequently the logistic EuroSCORE model was developed to provide a better risk predictor, especially in high-risk patients, and was published in 2003. The additive model is now over ten years old and the logistic model is now over six years old and they were developed based on patient database who had surgery in 1995, and hence the models may be outdated. From the various recent publications we are aware that EuroSCORE currently overpredicts mortality. Your study is therefore very timely because a number of centers currently use the EuroSCORE to define high-risk aortic valve replacement for the transcatheter program. I would like to congratulate you and your group for this important and timely publication and reminder. The other comment I want to point out is that The Society for Cardiothoracic Surgery in Great Britain and Ireland, who is aware of the issue with the overestimation of EuroSCORE, have

Ann Thorac Surg WENDT ET AL 2009;88:468 75 STS VS EUROSCORE FOR ISOLATED AVR 475 modified the risk scoring system so as to more accurately predict the actual mortality. They have responded to the overprediction of logistic EuroSCORE by undertaking a complex recalibration whereby they have looked at the comparisons between the observed mortality and that predicted in each operative group to derive a series of recalibration coefficients. These were then applied to the analyses of national data for hospitals and surgeons. This was first done for publication in 2007 based on data from 2002 to 2005. This is an important area because as health care practitioners our actual mortality is actually compared with the predicted mortality in the U.K., and that is published on the Internet for the public to view (http:// heartsurgery.healthcarecommission.org.uk/). So once again, well done on your study. DR WENDT: Thanks for your very supportive comments, and indeed the EuroSCORE, as you mentioned, is nearly 15 years old, and those over 19,000 patients who were examined in the original EuroSCORE study included only 3,200 patients who received an aortic valve replacement, so only 17% out of this cohort, as compared to over 30,000 patients within the STS score. Therefore, I totally agree with your comments. And, furthermore, I think this is one of the most lacking points of the EuroSCORE. It is a very old score, and a precise risk score for aortic valve replacement should be timely recalculated, perhaps every year. So the STS score was recalculated and the latest version came up last year and I think it is a more specific tool for risk calculation in aortic valve replacement. Therefore, at our institution we started with a staged approach towards transcatheter aortic valve implantation, and we performed our first transcatheter aortic valve implantation only on very high-risk patients, which is reflected by our poster concerning transcatheter aortic valve implantation in high-risk patients at this meeting. Our mean STS score of transcatheter aortic valve placement is about 20%, reflecting a really high-risk patient population. DR ALESSANDRO PAROLARI (Milano, Italy): I agree with the part of your conclusion that says that all these methods overestimate the risk, but I totally disagree with your statement that one method is better than another one from the data you present here. I am very concerned from this kind of information because the events reported are not so high. Usually, to compare performance of risk score you have to use ROC [receiver operating characteristic] curves comparison that can evaluate the discrimination of the different scores, but in this case, due to the low frequency of the events you cannot do that. So, I would be very concerned about stating that one method is better than another one. DR WENDT: That is right, but we mentioned this in our limitations. It is a single-center study and only 652 patients could be identified and only 52 patients were at highest risk with a logistic EuroSCORE above 20, but the other groups, like Dr Grossi presented 730 and Dr Brown presented 1,177 patients as well. Therefore, I think we need some big and large randomized trials to get more information about that. DR PAROLARI: The problem is not in the total number of patients studied but it is in the number of events, very few, that limit this kind of analysis. DR WENDT: Yes, that is right. For this reason we need a large volume study in order to increase the number of events and afterwards perform the ROC analysis. Unfortunately, the ROC analysis of the STS score cannot be performed to date because of its proprietary design and not published estimates. DR MARK WAYNE BURLINGAME (Lancaster, PA): I have tried to use the STS risk stratifier to separate these patients for those that I am going to send for transcatheter aortic valve replacement. The specific deficiency in that database is the lack of inclusion of pulmonary hypertension and RV [right ventricular] function, and I wondered if the EuroSCORE includes that, and certainly shouldn t this be included on both scores? DR WENDT: That s right, but the older version of the STS score before the beginning of 2007 included pulmonary hypertension as well. And I totally agree with your comment that pulmonary hypertension and right heart function play a major role in transcatheter aortic valve implantation.