ORIGINAL ARTICLE
European Journal of Cardio-Thoracic Surgery 47 (2015) e199–e205 doi:10.1093/ejcts/ezv023 Advance Access publication 16 February 2015
Cite this article as: Castelvecchio S, Menicanti L, Ranucci M, for the Surgical and Clinical Outcome Research (SCORE) Group. Development and validation of a risk score for predicting operative mortality in heart failure patients undergoing surgical ventricular reconstruction. Eur J Cardiothorac Surg 2015;47:e199–e205.
Development and validation of a risk score for predicting operative mortality in heart failure patients undergoing surgical ventricular reconstruction Serenella Castelvecchioa,*, Lorenzo Menicantia and Marco Ranuccib; for the Surgical and Clinical Outcome Research (SCORE) Group a b
Department of Cardiac Surgery, I.R.C.C.S. Policlinico San Donato, Milan, Italy Department of Cardiothoracic and Vascular Anesthesia & ICU, I.R.C.C.S. Policlinico San Donato, Milan, Italy
* Corresponding author. Department of Cardiac Surgery, I.R.C.C.S. Policlinico San Donato, Via Morandi 30, 20097 San Donato Milanese, Milan, Italy. Tel: +39-02-52774842; fax: +39-02-52774415; e-mail:[email protected] (S. Castelvecchio). Received 12 September 2014; received in revised form 12 November 2014; accepted 22 December 2014
OBJECTIVES: Different risk models have been introduced and refined in the past in order to improve standards of care. However, the predictive power of any risk algorithms can decline over time due to changes in surgical practice and the population’s risk profile. The present study aimed to develop and validate a risk model for predicting operative mortality in patients with ischaemic heart failure (HF) undergoing surgical ventricular reconstruction (SVR). METHODS: The study population included 525 patients with previous myocardial infarction and left ventricular remodelling referred to our centre for SVR. All patients underwent surgical reshaping; coronary artery bypass grafting was performed in 489 (93%) patients and mitral valve (MV) repair in 142 (27%). Operative mortality was defined as death within 30 days after surgery. All patients received an operative risk assessment using the logistic EuroSCORE and the ACEF score. RESULTS: Better accuracy was achieved by the ACEF score (0.771) compared with the EuroSCORE (0.747). On multivariable logistic regression analysis, forcing the ACEF score in the model, three additional factors remained as independent predictors of operative mortality: atrial fibrillation, NYHA Class 3–4 and MV surgery (odds ratio 2.2, 2.6 and 2.1, respectively) and were computed in the ACEF-SVR. The ACEF-SVR score demonstrated an improved accuracy in respect of the ACEF score (from 0.771 to 0.792) and a better calibration (Hosmer– Lemeshow χ 2 of 5.40, P = 0.714). CONCLUSIONS: The ACEF-SVR score, starting from a simplified model of risk enabled improvement in the accuracy and calibration of the model, tailoring the risk to a specific population of patients with HF undergoing a specific surgical procedure. Keywords: Heart failure • Operative mortality • Risk stratification • Surgical ventricular reconstruction
INTRODUCTION Risk stratification in cardiac surgery is an important issue within the processes aimed to improve the standards of care. Risk models aim to identify and weight different patient characteristics that may affect the probability of specific adverse outcomes. To this aim, different risk models have been introduced and refined within the last two decades, including either scores dedicated to isolated coronary artery bypass grafting (CABG) procedures, or scores for CABG with associated procedures [1]. Within Europe, the EuroSCORE I [2, 3] and II [4] are the most frequently used risk-stratification models and are intended to be used for ‘all adult cardiac surgical’ procedures. Conversely, the most recent STS-Prom risk models are designed in a specific way for different
operations, although some operations are not yet included [5]. Among others, surgical ventricular reconstruction (SVR) is not encoded in the above-mentioned models. SVR of the left cavity mostly combined with myocardial revascularization and mitral valve (MV) repair are some of the therapeutic strategies available in order to improve the life expectancy and quality of life of patients with ischaemic heart failure (HF) [6]. In particular, the aim of SVR is to exclude scar tissue from the left ventricular (LV) wall, thereby restoring the LV physiological volume and shape and improving LV systolic function and clinical status [7, 8]. The operative mortality observed for this type of operation has been reported to be between 2–13% [9], a wide but understandable range considering the heterogeneity of the populations undergoing this type of surgery and the complexity of
© The Author 2015. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.
ADULT CARDIAC
Abstract
e200
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
the underlying disease. Furthermore, the results from the Surgical Treatment for Ischemic Heart Failure (STICH) trial have challenged the additional benefit of SVR [10], even if the most recent guidelines have included this operation as a surgical option in selected HF patients [6, 11]. Therefore, there is a need to optimize the selection of patients including a proper stratification of operative risk, which is still lacking. The aim of this study is to develop and validate a risk model for predicting operative mortality in patients with ischaemic HF undergoing SVR.
METHODS Study design This study is based on a retrospective analysis of data prospectively collected in an Institutional Registry. The Registry contains clinical data of all the patients submitted to SVR from July 2001. The aim of this study is to develop a specific risk model for operative mortality of patients undergoing SVR. The study design was approved by the local ethics committee (San Raffaele Hospital, Milan, Italy).
Patient population The study population included 525 patients with previous myocardial infarction and LV remodelling referred to our centre for SVR. All patients underwent SVR; CABG was performed in 489 (93%) patients and MV repair in 142 (27%). Indications for surgery were HF, angina and/or a combination of the two. According to our usual practice, all patients had a complete clinical and echocardiographic evaluation before surgery, with the aim to collect demographic, clinical, laboratory and imaging data, and data on the presence of comorbidities. After surgery, operative details were retrieved. Operative mortality was defined as in-hospital mortality or mortality within 30 days after the operation for patients discharged from the hospital, including deaths occurring in rehabilitation centres, in secondary hospitals or at home.
Healthcare, Waukesha, WI, USA) echocardiographic instrument, and all four-chamber, two-chamber, long-axis and short-axis views were analysed before in all patients. The parameters measured were the following: (i) Diastolic and systolic internal diameters in the parasternal long-axis view (millimetres). (ii) End-diastolic and end-systolic volume calculated by applying the Simpson method, indexed by the body surface area (ml/m2). (iii) Ejection fraction (EF) derived from volumes as end-diastolic volume − end-systolic volume/end-diastolic volume (%). (iv) Left atrial diameter in the long-axis view (millimetres). (v) Tricuspid annular plane systolic excursion (TAPSE) in the apical four-chamber view with an M-mode cursor placed through the lateral annulus in real time. TAPSE was measured as the total displacement of the tricuspid annulus (mm) from end-diastole to end-systole (vi) Pulmonary artery systolic pressures estimated by Doppler echocardiography recording the tricuspid regurgitant velocity from any view with continuous-wave Doppler (modified Bernoulli equation). Operative details collected include the number of distal anastomoses, the use of a Dacron patch to close the cavity after the LV reconstruction and MV repair. Details of SVR operation have been previously described in a comprehensive manner [8]. On the basis of the variables collected, all patients received an operative risk assessment using the logistic EuroSCORE and the ACEF score [12]. The EuroSCORE II was not applicable in this population due to the lack of some risk factors in the database. The ACEF score is a simple risk-stratification system for cardiac surgery that was introduced in 2009 and subsequently included in the ESC/ EACTS guidelines for myocardial revascularization in 2010 [6] and 2014 [13] with a level of evidence IIb. ACEF is the acronym of Age, Creatinine and EF, and is calculated based on these variables, as the age (years) divided by the LVEF (%), plus 1 point if the serum creatinine level is >2.0 g/dl. With an appropriate logistic equation [12], this score is translated into predicted operative mortality (%). Follow-up information was obtained from clinical records, death certificates and telephone contact or correspondence. The follow-up was 100% completed.
Statistical analysis Data collection and definitions The following variables were collected: age, gender, body surface area, hypertension, dyslipidaemia, diabetes mellitus under medication, atrial fibrillation, previous documented ventricular arrhythmias, New York Heart Association (NYHA) functional class, recent myocardial infarction, type of LV post-infarct remodelling (anterior versus posterior), stable or unstable angina, extracardiac arteriopathy, critical preoperative conditions, active endocarditis, serum creatinine value (mg/dl), haemoglobin value (g/l), chronic obstructive pulmonary disease, previous cerebrovascular events, previous cardiac surgery, type of operation, combined operation, thoracic aorta operation and post-infarction septal rupture. Additional variables collected include medications and presence of single- or multivessel coronary artery disease. If necessary, these conditions were defined according to the EuroSCORE definitions. Two-dimensional, M-mode and colour Doppler transthoracic echocardiography was performed using a GE Vivid 7 (GE
As a preliminary step, the EuroSCORE and the ACEF score were tested for accuracy and calibration in predicting operative mortality. Receiver operating characteristic (ROC) analyses were used, and the area under the curve (AUC) was considered as a determinant of accuracy. Calibration was tested with the Hosmer– Lemeshow test, and clinical performance by comparison between the expected and the observed operative mortality rate ( percentage with 95% confidence interval). This last analysis was repeated in the whole patient population, and in subgroups of patients defined according to the quartiles of distribution of the EuroSCORE as low risk (first quartile, predicted mortality 9.1%). The risk model with the higher accuracy between the EuroSCORE and ACEF score was selected as the starting model for developing the SVR-specific risk model.
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
RESULTS Table 1 reports the general data of the patient population. Forty-eight patients (9.1%) died in hospital or within 30 days from the operation after discharge, and were considered as operative mortality for the following analyses. The predicted mortality rates for the EuroSCORE and the ACEF score are depicted in Table 2. The EuroSCORE underestimated the operative mortality risk in the overall population and in the highand very high-risk subgroups, with observed mortality rates outside the upper 95% confidence limits of the predicted mortality rates. Conversely, the mortality risk was overestimated in the low- and moderate-risk subgroups, with observed mortality rates outside the lower 95% confidence limits of the predicted mortality rates. This poor calibration of the EuroSCORE was confirmed by the Hosmer–Lemeshow goodness-of-fit test (χ 2 of 19.7, P = 0.012). The operative mortality is correctly predicted by the ACEF score in the overall population, high- and very high risk subgroups, while it was overestimated in low- and moderate-risk subgroups,
Table 1:
General characteristics of the patient population
Variable
Mean (SD) or median (IQ range) or number (%)
Age (years) Body surface area (m2) Gender male Left ventricular ejection fraction (%) NYHA class NYHA Class 3–4 Serum creatinine (mg/dl) Haemoglobin (g/dl) Systolic pulmonary artery pressure (mmHg) TAPSE (mm) Left atrial diameter (mm) Peripheral arteriopathy Hypertension Dyslipidaemia Chronic obstructive pulmonary disease Previous cerebrovascular accident Diabetes Atrial fibrillation Unstable angina Recent myocardial infarction Emergency operation Redo operation Active endocarditis Critical preoperative state Site of LV remodelling Anterior Posterior Antero-posterior Mitral valve surgery EuroSCORE (predicted mortality, %) ACEF score Predicted mortality ACEF score (%)
66 (58–72) 1.83 (0.16) 432 (83) 30 (8.1) 2 (2–3) 250 (47.6) 1.14 (0.95–1.44) 13.1 (1.8) 36.0 (28–48) 19 (17–22) 46.7 (7.2) 145 (27.6) 305 (58.1) 290 (55.3%) 95 (18.1) 48 (9.1) 138 (26.3) 74 (14.1) 69 (13.1) 159 (30.3) 0 (0) 19 (3.6) 0 (0) 0 (0)
ADULT CARDIAC
A number of additional variables collected in our database and not included in the EuroSCORE or ACEF score was considered for association with operative mortality. These included: hypertension; dyslipidaemia; preoperative haemoglobin value; preoperative atrial fibrillation; NYHA Class 3–4; left atrial dimension (diameter); TAPSE; LV aneurysm location (anterior, posterior and antero-posterior) and surgical MV repair. The above-mentioned variables were tested for association with operative mortality using Pearson’s χ 2 test for binary variables, Student’s t-test for continuous, normally distributed variables and a Mann–Whitney U-test for continuous, non-normally distributed variables. Normality of distribution was checked with the Kolmogorov–Smirnov test. The variables being associated with operative mortality at a P-value of 0.7 and 41.9% an AUC of >0.8, thus confirming the good accuracy of the ACEF-SVR score in predicting operative mortality. Reclassification tables were produced for comparison of the ACEF-SVR score versus the ACEF score (Table 5). For the purposes of this analysis, we considered three risk classes, based on a predicted mortality rate according to the ACEF and ACEF-SVR scores: moderate risk (20%). Overall, 23.6% of the patients were reclassified into a new risk category by the ACEF-SVR score. This produced a better alignment of the risk classification in the categories of high and very high risk, with predicted values closer to the observed values for the ACEF-SVR score.
DISCUSSION The main findings of this study may be summarized as follows: (i) the ACEF score showed a better accuracy and calibration than the EuroSCORE; (ii) the ACEF-SVR score, starting from an extremely simplified model of risk, was able to improve the discriminatory power and calibration of the starting model, tailoring the risk on a specific population of patients undergoing a specific procedure.
The performance of the ACEF score versus the EuroSCORE in HF patients undergoing SVR In this population of patients with multiple comorbidities and poor prognosis undergoing a specific operation, the EuroSCORE failed to accurately predict the operative risk either in the overall population or in the high- and very high-risk subgroups. This finding is not surprising and confirms the well-known
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
e203
Table 4: Multivariable analysis for operative mortality risk Variable
Regression coefficient
Odds ratio (95% CI)
P-value
ACEF score Atrial fibrillation NHYA Class 3–4 Mitral valve surgery
0.842 0.814 0.953 0.740
2.3 (1.57–3.42) 2.2 (1.00–5.1) 2.6 (1.09–6.2) 2.1 (1.01–4.35)
0.001 0.050 0.031 0.047
CI: confidence interval; NYHA: New York Heart Association.
binary variable [12]. In this way, the relevance of the ACEF score is crucial in highlighting the importance of including every possible value of the EF in the model. In doing so, the model enables one to reach a higher accuracy in the range of very low EF values, which could reasonably identify the ‘high- and very high-risk’ subgroups of patients. Figure 2: The operative mortality risk according to the ACEF-SVR score. CI: confidence interval; SVR: surgical ventricular reconstruction.
shortcomings of the EuroSCORE [15, 16]. Moreover, contrary to that previously reported by many authors, in this population the EuroSCORE significantly ‘underestimated’ the risk of operative death in the high- and very high-risk subgroups. The ‘high-risk’ group has been always shown to be a predictive challenge for different reasons; first, the definition of the ‘high-risk patient’ and related variables used to better assess the risk. From a statistical point of view, it should be emphasized that the EuroSCORE models were created for stratifying risk in the overall cardiac surgical patient population (different risk groups and different procedures). Removing subgroups from this population for specific analysis will invariably result in the loss of calibration. With regard to the variables, although the EF remains the most widely used and time-saved measure of LV systolic function and HF, its value has been strongly criticized and its inclusion in risk models is misleading when it is entered as a categorical variable; this approach does not enable the adequate discriminate patients with different degrees of systolic dysfunction (a patient with an EF equal to 17% is considered at a similar risk as a patient with an EF equal to 29%, as having ‘poor LV function’ in a misleading way). The same limitation has been retained in the EuroSCORE II in which a further category was simply added to identify ‘patients with a very low LV function’ (EF equal to 20% or less). In spite of that, a recent study by Howell et al. [17] showed that the new EuroSCORE II does not improve risk prediction in high-risk patients undergoing adult cardiac surgery when compared with original additive and logistic EuroSCOREs. The ACEF score was developed using only three factors for predicting operative mortality risk, including age and EF—as ‘continuous variables’—and preoperative serum creatinine value—as a
The additional value of the ACEF-SVR: procedure-specific models for specific subset of patients In this study, we faced the possibility of using the ACEF score for improving the risk assessment ‘according’ to the population risk profile and to the procedure whose risk we want to estimate. Doing that, we found that three more variables had an additional weight in this specific subset of HF patients undergoing SVR, each one doubling, at least, the risk of operative death when present (OR for these factors ranged between 2.25 and 2.6). Atrial fibrillation is a very common arrhythmia affecting several million patients with a median age of 70 years. As the mean age for patients selected for cardiac surgery is increasing, it is nowadays not uncommon to have patients presenting with concomitant atrial fibrillation. Preoperative atrial fibrillation among patients undergoing CABG has been shown to be associated with higher mortality and morbidity [18, 19]. Nevertheless, atrial fibrillation was never introduced as a variable in the most commonly used risk calculators, such as the EuroSCORE in which the calibration of the model may have been influenced by ‘unmeasured variables’ that indeed can influence the risk. A similar impact has been found for NYHA functional class and MV surgery. These findings partially confirm previous results reported by our group in 2007 [7]. In a smaller subgroup of patients undergoing SVR, we found that the presence of moderate-to-severe mitral regurgitation associated with a surgical repair carries a higher operative risk. Furthermore, a mitral regurgitation of grade 2 or more alone did not increase the operative mortality risk; conversely, if associated with advanced functional class (NYHA Class 3–4), the risk was significantly increased, reinforcing the concept that the worse
ADULT CARDIAC
Figure 3: Accuracies of the ACEF score and ACEF-SVR score tested with ROC analysis.
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
e204
Table 5: Reclassification table of the patient population according to the ACEF score and the ACEF-SVR score Risk model
ACEF score ACEF-SVR score
ACEF score Moderate risk High risk Very high risk ACEF-SVR score Moderate risk High risk Very high risk
Moderate risk (20%
66 (12.5) 58 (11.1)
% Reclassification into a new category Lower
Higher
Total
69 (13.1)
55 (10.5)
124 (23.6)
Data are denoted as number (%). SVR: surgical ventricular reconstruction.
prognosis is related to a poorer LV function and mitral regurgitation is merely an indicator of this bad condition. This could explain the higher risk for operative mortality when MV repair is combined with SVR. Additionally, the longer cardiopulmonary bypass and aortic cross-clamp times required by MV repair may justify an increased operative mortality. The effect of mitral valve surgery on survival is still unclear. However, MV repair is likely to improve symptoms and prevents further deterioration of LV function, hence the need to properly estimate the additional operative risk in performing mitral surgery combined with SVR. Risk stratification for MV surgery is accurate both with the EuroSCORE and the ACEF score, but the ACEF score has a better calibration [20]; this further supports the choice of the ACEF as the starting model for risk assessment in SVR. In this regard, the EuroSCORE appears less adequate, since it categorizes surgeries in general classes, supporting the role of the number of procedures without differentiating among non-CABG procedures.
LIMITATIONS Our study was a retrospective cohort analysis and has the inherent limitations of this design. Furthermore, the ACEF-SVR score was developed on a narrow subset of patients undergoing a specific operation in a single institution, hence precluding the generalizability of the model to other cohorts of patients. Some variables were not available for the entire population (i.e. the left atrial volume and the diastolic pattern), and were excluded from the multivariable analysis. We cannot exclude that they could offer additional prognostic information, even if multicollinearity and intercorrelation with other factors like atrial fibrillation are anticipated. Finally, no external validation was performed.
CONCLUSION The ACEF-SVR model, starting from an extremely simplified model of risk, enabled improvements in the accuracy and calibration of the model on a subset of the HF population undergoing SVR. This new model meets the current need for procedure-
specific models to be applied on specific subset of patients, avoiding the error of roughly assigning a probability of outcome of a population. Future directions should explore, first, the possibility to externally validate the ACEF-SVR. Secondly, it will be mandatory to extend the current analysis over a longer time interval looking, for instance, at 1-year mortality or, even better, including a different outcome more appropriate for this subset of patients. A major concern for a population affected by advanced HF is the risk of surviving with repeated hospitalizations for HF, which, in turn, worsen the quality of life and increase the public health burden in terms of financial resources and costs. Conflict of interest: none declared.
REFERENCES [1] Nilsson J, Algotsson L, Höglund P, Lührs C, Brandt J. Comparison of 19 preoperative risk stratification models in open-heart surgery. Eur Heart J 2006;27:867–74. [2] 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. [3] Roques F, Michel P, Goldstone AR, Nashef SA. The logistic EuroSCORE. Eur Heart J 2003;24:882–83. [4] Nashef SA, Roques F, Sharples LD, Nilsson J, Smith C, Goldstone AR et al. EuroSCORE II. Eur J Cardiothorac Surg 2012;4:734–44. [5] Jin R, Furnary AP, Fine SC, Blackstone EH, Grunkemeier GL. Using Society of Thoracic Surgeons risk models for risk-adjusting cardiac surgery results. Ann Thorac Surg 2010;89:677–82. [6] Wijns W, Kolh P, Danchin N, Di Mario C, Falk V, Folliguet T et al. Guidelines on myocardial revascularization: the Task Force on Myocardial Revascularization of the European Society of Cardiology (ESC) and the European Association for Cardio-Thoracic Surgery (EACTS). Eur J Cardiothorac Surg 2010;38:Suppl:S1–S52. [7] Menicanti L, Castelvecchio S, Ranucci M, Frigiola A, Santambrogio C, de Vincentiis C et al. Surgical therapy for ischemic heart failure: singlecenter experience with surgical anterior ventricular restoration. J Thorac Cardiovasc Surg 2007;134:433–41. [8] Menicanti L, Castelvecchio S. Left ventricular reconstruction concomitant to coronary artery bypass surgery: when and how? Curr Opin Cardiol 2011;26:523–27. [9] Castelvecchio S, Menicanti L, Di Donato M. Surgical ventricular restoration to reverse left ventricular remodeling. Curr Cardiol Rev 2010;6:15–23.
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
[15] Shanmugam G, West M, Berg G. Additive and logistic EuroSCORE performance in high risk patients. Interact CardioVasc Thorac Surg 2005;4: 299–303. [16] Bhatti F, Grayson AD, Grotte G, Fabri BM, Au J, Jones M et al. The logistic EuroSCORE in cardiac surgery: how well does it predict operative risk? Heart 2006;92:1817–20. [17] Howell NJ, Head SJ, Freemantle N, van der Meulen TA, Senanayake E, Menon A et al. The new EuroSCORE II does not improve prediction of mortality in high-risk patients undergoing cardiac surgery: a collaborative analysis of two European centres. Eur J Cardiothorac Surg 2013;44:1006–11. [18] Quader MA, McCarthy PM, Gillinov AM, Alster JM, Cosgrove DM III, Lytle BW et al. Does preoperative atrial fibrillation reduce survival after coronary artery bypass grafting? Ann Thorac Surg 2004;77:1514–24. [19] Ad N, Barnett SD, Haan CK, O’Brien SM, Milford-Beland S, Speir AM. Does preoperative atrial fibrillation increase the risk for mortality and morbidity after coronary artery bypass grafting? J Thorac Cardiovasc Surg 2009;137: 901–6. [20] Barili F, Pacini D, Grossi C, Di Bartolomeo R, Alamanni F, Parolari A. Reliability of new scores in predicting perioperative mortality after mitral valve surgery. J Thorac Cardiovasc Surg 2014;147:1008–12.
ADULT CARDIAC
[10] Jones RH, Velazquez EJ, Michler RE, Sopko G, Oh JK, O’Connor CM et al. Coronary bypass surgery with or without surgical ventricular reconstruction. N Engl J Med 2009;360:1705–17. [11] Yancy CW, Jessup M, Bozkurt B, Butler J, Casey DE Jr, Drazner MH et al. 2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force on practice guidelines. Circulation 2013;128: e240–327. [12] Ranucci M, Castelvecchio S, Menicanti L, Frigiola A, Pelissero G. Risk of assessing mortality risk in elective cardiac operations: age, creatinine, ejection fraction, and the law of parsimony. Circulation 2009;119:3053–61. [13] Windecker S, Kolh P, Alfonso F, Collet JP, Cremer J, Falk V et al. Guidelines on myocardial revascularization: the Task Force on Myocardial Revascularization of the European Society of Cardiology (ESC) and the European Association for Cardio-Thoracic Surgery (EACTS). Developed with the special contribution of the European Association of Percutaneous Cardiovascular Interventions (EAPCI). Eur J Cardiothorac Surg 2014;46:517–92. [14] Cook NR, Ridker PM. The use and magnitude of reclassification measures for individual predictors of global cardiovascular risk. Ann Intern Med 2009;150:795–802.
e205
Copyright of European Journal of Cardio-Thoracic Surgery is the property of Oxford University Press / USA and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
European Journal of Cardio-Thoracic Surgery 47 (2015) e199–e205 doi:10.1093/ejcts/ezv023 Advance Access publication 16 February 2015
Cite this article as: Castelvecchio S, Menicanti L, Ranucci M, for the Surgical and Clinical Outcome Research (SCORE) Group. Development and validation of a risk score for predicting operative mortality in heart failure patients undergoing surgical ventricular reconstruction. Eur J Cardiothorac Surg 2015;47:e199–e205.
Development and validation of a risk score for predicting operative mortality in heart failure patients undergoing surgical ventricular reconstruction Serenella Castelvecchioa,*, Lorenzo Menicantia and Marco Ranuccib; for the Surgical and Clinical Outcome Research (SCORE) Group a b
Department of Cardiac Surgery, I.R.C.C.S. Policlinico San Donato, Milan, Italy Department of Cardiothoracic and Vascular Anesthesia & ICU, I.R.C.C.S. Policlinico San Donato, Milan, Italy
* Corresponding author. Department of Cardiac Surgery, I.R.C.C.S. Policlinico San Donato, Via Morandi 30, 20097 San Donato Milanese, Milan, Italy. Tel: +39-02-52774842; fax: +39-02-52774415; e-mail:[email protected] (S. Castelvecchio). Received 12 September 2014; received in revised form 12 November 2014; accepted 22 December 2014
OBJECTIVES: Different risk models have been introduced and refined in the past in order to improve standards of care. However, the predictive power of any risk algorithms can decline over time due to changes in surgical practice and the population’s risk profile. The present study aimed to develop and validate a risk model for predicting operative mortality in patients with ischaemic heart failure (HF) undergoing surgical ventricular reconstruction (SVR). METHODS: The study population included 525 patients with previous myocardial infarction and left ventricular remodelling referred to our centre for SVR. All patients underwent surgical reshaping; coronary artery bypass grafting was performed in 489 (93%) patients and mitral valve (MV) repair in 142 (27%). Operative mortality was defined as death within 30 days after surgery. All patients received an operative risk assessment using the logistic EuroSCORE and the ACEF score. RESULTS: Better accuracy was achieved by the ACEF score (0.771) compared with the EuroSCORE (0.747). On multivariable logistic regression analysis, forcing the ACEF score in the model, three additional factors remained as independent predictors of operative mortality: atrial fibrillation, NYHA Class 3–4 and MV surgery (odds ratio 2.2, 2.6 and 2.1, respectively) and were computed in the ACEF-SVR. The ACEF-SVR score demonstrated an improved accuracy in respect of the ACEF score (from 0.771 to 0.792) and a better calibration (Hosmer– Lemeshow χ 2 of 5.40, P = 0.714). CONCLUSIONS: The ACEF-SVR score, starting from a simplified model of risk enabled improvement in the accuracy and calibration of the model, tailoring the risk to a specific population of patients with HF undergoing a specific surgical procedure. Keywords: Heart failure • Operative mortality • Risk stratification • Surgical ventricular reconstruction
INTRODUCTION Risk stratification in cardiac surgery is an important issue within the processes aimed to improve the standards of care. Risk models aim to identify and weight different patient characteristics that may affect the probability of specific adverse outcomes. To this aim, different risk models have been introduced and refined within the last two decades, including either scores dedicated to isolated coronary artery bypass grafting (CABG) procedures, or scores for CABG with associated procedures [1]. Within Europe, the EuroSCORE I [2, 3] and II [4] are the most frequently used risk-stratification models and are intended to be used for ‘all adult cardiac surgical’ procedures. Conversely, the most recent STS-Prom risk models are designed in a specific way for different
operations, although some operations are not yet included [5]. Among others, surgical ventricular reconstruction (SVR) is not encoded in the above-mentioned models. SVR of the left cavity mostly combined with myocardial revascularization and mitral valve (MV) repair are some of the therapeutic strategies available in order to improve the life expectancy and quality of life of patients with ischaemic heart failure (HF) [6]. In particular, the aim of SVR is to exclude scar tissue from the left ventricular (LV) wall, thereby restoring the LV physiological volume and shape and improving LV systolic function and clinical status [7, 8]. The operative mortality observed for this type of operation has been reported to be between 2–13% [9], a wide but understandable range considering the heterogeneity of the populations undergoing this type of surgery and the complexity of
© The Author 2015. Published by Oxford University Press on behalf of the European Association for Cardio-Thoracic Surgery. All rights reserved.
ADULT CARDIAC
Abstract
e200
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
the underlying disease. Furthermore, the results from the Surgical Treatment for Ischemic Heart Failure (STICH) trial have challenged the additional benefit of SVR [10], even if the most recent guidelines have included this operation as a surgical option in selected HF patients [6, 11]. Therefore, there is a need to optimize the selection of patients including a proper stratification of operative risk, which is still lacking. The aim of this study is to develop and validate a risk model for predicting operative mortality in patients with ischaemic HF undergoing SVR.
METHODS Study design This study is based on a retrospective analysis of data prospectively collected in an Institutional Registry. The Registry contains clinical data of all the patients submitted to SVR from July 2001. The aim of this study is to develop a specific risk model for operative mortality of patients undergoing SVR. The study design was approved by the local ethics committee (San Raffaele Hospital, Milan, Italy).
Patient population The study population included 525 patients with previous myocardial infarction and LV remodelling referred to our centre for SVR. All patients underwent SVR; CABG was performed in 489 (93%) patients and MV repair in 142 (27%). Indications for surgery were HF, angina and/or a combination of the two. According to our usual practice, all patients had a complete clinical and echocardiographic evaluation before surgery, with the aim to collect demographic, clinical, laboratory and imaging data, and data on the presence of comorbidities. After surgery, operative details were retrieved. Operative mortality was defined as in-hospital mortality or mortality within 30 days after the operation for patients discharged from the hospital, including deaths occurring in rehabilitation centres, in secondary hospitals or at home.
Healthcare, Waukesha, WI, USA) echocardiographic instrument, and all four-chamber, two-chamber, long-axis and short-axis views were analysed before in all patients. The parameters measured were the following: (i) Diastolic and systolic internal diameters in the parasternal long-axis view (millimetres). (ii) End-diastolic and end-systolic volume calculated by applying the Simpson method, indexed by the body surface area (ml/m2). (iii) Ejection fraction (EF) derived from volumes as end-diastolic volume − end-systolic volume/end-diastolic volume (%). (iv) Left atrial diameter in the long-axis view (millimetres). (v) Tricuspid annular plane systolic excursion (TAPSE) in the apical four-chamber view with an M-mode cursor placed through the lateral annulus in real time. TAPSE was measured as the total displacement of the tricuspid annulus (mm) from end-diastole to end-systole (vi) Pulmonary artery systolic pressures estimated by Doppler echocardiography recording the tricuspid regurgitant velocity from any view with continuous-wave Doppler (modified Bernoulli equation). Operative details collected include the number of distal anastomoses, the use of a Dacron patch to close the cavity after the LV reconstruction and MV repair. Details of SVR operation have been previously described in a comprehensive manner [8]. On the basis of the variables collected, all patients received an operative risk assessment using the logistic EuroSCORE and the ACEF score [12]. The EuroSCORE II was not applicable in this population due to the lack of some risk factors in the database. The ACEF score is a simple risk-stratification system for cardiac surgery that was introduced in 2009 and subsequently included in the ESC/ EACTS guidelines for myocardial revascularization in 2010 [6] and 2014 [13] with a level of evidence IIb. ACEF is the acronym of Age, Creatinine and EF, and is calculated based on these variables, as the age (years) divided by the LVEF (%), plus 1 point if the serum creatinine level is >2.0 g/dl. With an appropriate logistic equation [12], this score is translated into predicted operative mortality (%). Follow-up information was obtained from clinical records, death certificates and telephone contact or correspondence. The follow-up was 100% completed.
Statistical analysis Data collection and definitions The following variables were collected: age, gender, body surface area, hypertension, dyslipidaemia, diabetes mellitus under medication, atrial fibrillation, previous documented ventricular arrhythmias, New York Heart Association (NYHA) functional class, recent myocardial infarction, type of LV post-infarct remodelling (anterior versus posterior), stable or unstable angina, extracardiac arteriopathy, critical preoperative conditions, active endocarditis, serum creatinine value (mg/dl), haemoglobin value (g/l), chronic obstructive pulmonary disease, previous cerebrovascular events, previous cardiac surgery, type of operation, combined operation, thoracic aorta operation and post-infarction septal rupture. Additional variables collected include medications and presence of single- or multivessel coronary artery disease. If necessary, these conditions were defined according to the EuroSCORE definitions. Two-dimensional, M-mode and colour Doppler transthoracic echocardiography was performed using a GE Vivid 7 (GE
As a preliminary step, the EuroSCORE and the ACEF score were tested for accuracy and calibration in predicting operative mortality. Receiver operating characteristic (ROC) analyses were used, and the area under the curve (AUC) was considered as a determinant of accuracy. Calibration was tested with the Hosmer– Lemeshow test, and clinical performance by comparison between the expected and the observed operative mortality rate ( percentage with 95% confidence interval). This last analysis was repeated in the whole patient population, and in subgroups of patients defined according to the quartiles of distribution of the EuroSCORE as low risk (first quartile, predicted mortality 9.1%). The risk model with the higher accuracy between the EuroSCORE and ACEF score was selected as the starting model for developing the SVR-specific risk model.
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
RESULTS Table 1 reports the general data of the patient population. Forty-eight patients (9.1%) died in hospital or within 30 days from the operation after discharge, and were considered as operative mortality for the following analyses. The predicted mortality rates for the EuroSCORE and the ACEF score are depicted in Table 2. The EuroSCORE underestimated the operative mortality risk in the overall population and in the highand very high-risk subgroups, with observed mortality rates outside the upper 95% confidence limits of the predicted mortality rates. Conversely, the mortality risk was overestimated in the low- and moderate-risk subgroups, with observed mortality rates outside the lower 95% confidence limits of the predicted mortality rates. This poor calibration of the EuroSCORE was confirmed by the Hosmer–Lemeshow goodness-of-fit test (χ 2 of 19.7, P = 0.012). The operative mortality is correctly predicted by the ACEF score in the overall population, high- and very high risk subgroups, while it was overestimated in low- and moderate-risk subgroups,
Table 1:
General characteristics of the patient population
Variable
Mean (SD) or median (IQ range) or number (%)
Age (years) Body surface area (m2) Gender male Left ventricular ejection fraction (%) NYHA class NYHA Class 3–4 Serum creatinine (mg/dl) Haemoglobin (g/dl) Systolic pulmonary artery pressure (mmHg) TAPSE (mm) Left atrial diameter (mm) Peripheral arteriopathy Hypertension Dyslipidaemia Chronic obstructive pulmonary disease Previous cerebrovascular accident Diabetes Atrial fibrillation Unstable angina Recent myocardial infarction Emergency operation Redo operation Active endocarditis Critical preoperative state Site of LV remodelling Anterior Posterior Antero-posterior Mitral valve surgery EuroSCORE (predicted mortality, %) ACEF score Predicted mortality ACEF score (%)
66 (58–72) 1.83 (0.16) 432 (83) 30 (8.1) 2 (2–3) 250 (47.6) 1.14 (0.95–1.44) 13.1 (1.8) 36.0 (28–48) 19 (17–22) 46.7 (7.2) 145 (27.6) 305 (58.1) 290 (55.3%) 95 (18.1) 48 (9.1) 138 (26.3) 74 (14.1) 69 (13.1) 159 (30.3) 0 (0) 19 (3.6) 0 (0) 0 (0)
ADULT CARDIAC
A number of additional variables collected in our database and not included in the EuroSCORE or ACEF score was considered for association with operative mortality. These included: hypertension; dyslipidaemia; preoperative haemoglobin value; preoperative atrial fibrillation; NYHA Class 3–4; left atrial dimension (diameter); TAPSE; LV aneurysm location (anterior, posterior and antero-posterior) and surgical MV repair. The above-mentioned variables were tested for association with operative mortality using Pearson’s χ 2 test for binary variables, Student’s t-test for continuous, normally distributed variables and a Mann–Whitney U-test for continuous, non-normally distributed variables. Normality of distribution was checked with the Kolmogorov–Smirnov test. The variables being associated with operative mortality at a P-value of 0.7 and 41.9% an AUC of >0.8, thus confirming the good accuracy of the ACEF-SVR score in predicting operative mortality. Reclassification tables were produced for comparison of the ACEF-SVR score versus the ACEF score (Table 5). For the purposes of this analysis, we considered three risk classes, based on a predicted mortality rate according to the ACEF and ACEF-SVR scores: moderate risk (20%). Overall, 23.6% of the patients were reclassified into a new risk category by the ACEF-SVR score. This produced a better alignment of the risk classification in the categories of high and very high risk, with predicted values closer to the observed values for the ACEF-SVR score.
DISCUSSION The main findings of this study may be summarized as follows: (i) the ACEF score showed a better accuracy and calibration than the EuroSCORE; (ii) the ACEF-SVR score, starting from an extremely simplified model of risk, was able to improve the discriminatory power and calibration of the starting model, tailoring the risk on a specific population of patients undergoing a specific procedure.
The performance of the ACEF score versus the EuroSCORE in HF patients undergoing SVR In this population of patients with multiple comorbidities and poor prognosis undergoing a specific operation, the EuroSCORE failed to accurately predict the operative risk either in the overall population or in the high- and very high-risk subgroups. This finding is not surprising and confirms the well-known
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
e203
Table 4: Multivariable analysis for operative mortality risk Variable
Regression coefficient
Odds ratio (95% CI)
P-value
ACEF score Atrial fibrillation NHYA Class 3–4 Mitral valve surgery
0.842 0.814 0.953 0.740
2.3 (1.57–3.42) 2.2 (1.00–5.1) 2.6 (1.09–6.2) 2.1 (1.01–4.35)
0.001 0.050 0.031 0.047
CI: confidence interval; NYHA: New York Heart Association.
binary variable [12]. In this way, the relevance of the ACEF score is crucial in highlighting the importance of including every possible value of the EF in the model. In doing so, the model enables one to reach a higher accuracy in the range of very low EF values, which could reasonably identify the ‘high- and very high-risk’ subgroups of patients. Figure 2: The operative mortality risk according to the ACEF-SVR score. CI: confidence interval; SVR: surgical ventricular reconstruction.
shortcomings of the EuroSCORE [15, 16]. Moreover, contrary to that previously reported by many authors, in this population the EuroSCORE significantly ‘underestimated’ the risk of operative death in the high- and very high-risk subgroups. The ‘high-risk’ group has been always shown to be a predictive challenge for different reasons; first, the definition of the ‘high-risk patient’ and related variables used to better assess the risk. From a statistical point of view, it should be emphasized that the EuroSCORE models were created for stratifying risk in the overall cardiac surgical patient population (different risk groups and different procedures). Removing subgroups from this population for specific analysis will invariably result in the loss of calibration. With regard to the variables, although the EF remains the most widely used and time-saved measure of LV systolic function and HF, its value has been strongly criticized and its inclusion in risk models is misleading when it is entered as a categorical variable; this approach does not enable the adequate discriminate patients with different degrees of systolic dysfunction (a patient with an EF equal to 17% is considered at a similar risk as a patient with an EF equal to 29%, as having ‘poor LV function’ in a misleading way). The same limitation has been retained in the EuroSCORE II in which a further category was simply added to identify ‘patients with a very low LV function’ (EF equal to 20% or less). In spite of that, a recent study by Howell et al. [17] showed that the new EuroSCORE II does not improve risk prediction in high-risk patients undergoing adult cardiac surgery when compared with original additive and logistic EuroSCOREs. The ACEF score was developed using only three factors for predicting operative mortality risk, including age and EF—as ‘continuous variables’—and preoperative serum creatinine value—as a
The additional value of the ACEF-SVR: procedure-specific models for specific subset of patients In this study, we faced the possibility of using the ACEF score for improving the risk assessment ‘according’ to the population risk profile and to the procedure whose risk we want to estimate. Doing that, we found that three more variables had an additional weight in this specific subset of HF patients undergoing SVR, each one doubling, at least, the risk of operative death when present (OR for these factors ranged between 2.25 and 2.6). Atrial fibrillation is a very common arrhythmia affecting several million patients with a median age of 70 years. As the mean age for patients selected for cardiac surgery is increasing, it is nowadays not uncommon to have patients presenting with concomitant atrial fibrillation. Preoperative atrial fibrillation among patients undergoing CABG has been shown to be associated with higher mortality and morbidity [18, 19]. Nevertheless, atrial fibrillation was never introduced as a variable in the most commonly used risk calculators, such as the EuroSCORE in which the calibration of the model may have been influenced by ‘unmeasured variables’ that indeed can influence the risk. A similar impact has been found for NYHA functional class and MV surgery. These findings partially confirm previous results reported by our group in 2007 [7]. In a smaller subgroup of patients undergoing SVR, we found that the presence of moderate-to-severe mitral regurgitation associated with a surgical repair carries a higher operative risk. Furthermore, a mitral regurgitation of grade 2 or more alone did not increase the operative mortality risk; conversely, if associated with advanced functional class (NYHA Class 3–4), the risk was significantly increased, reinforcing the concept that the worse
ADULT CARDIAC
Figure 3: Accuracies of the ACEF score and ACEF-SVR score tested with ROC analysis.
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
e204
Table 5: Reclassification table of the patient population according to the ACEF score and the ACEF-SVR score Risk model
ACEF score ACEF-SVR score
ACEF score Moderate risk High risk Very high risk ACEF-SVR score Moderate risk High risk Very high risk
Moderate risk (20%
66 (12.5) 58 (11.1)
% Reclassification into a new category Lower
Higher
Total
69 (13.1)
55 (10.5)
124 (23.6)
Data are denoted as number (%). SVR: surgical ventricular reconstruction.
prognosis is related to a poorer LV function and mitral regurgitation is merely an indicator of this bad condition. This could explain the higher risk for operative mortality when MV repair is combined with SVR. Additionally, the longer cardiopulmonary bypass and aortic cross-clamp times required by MV repair may justify an increased operative mortality. The effect of mitral valve surgery on survival is still unclear. However, MV repair is likely to improve symptoms and prevents further deterioration of LV function, hence the need to properly estimate the additional operative risk in performing mitral surgery combined with SVR. Risk stratification for MV surgery is accurate both with the EuroSCORE and the ACEF score, but the ACEF score has a better calibration [20]; this further supports the choice of the ACEF as the starting model for risk assessment in SVR. In this regard, the EuroSCORE appears less adequate, since it categorizes surgeries in general classes, supporting the role of the number of procedures without differentiating among non-CABG procedures.
LIMITATIONS Our study was a retrospective cohort analysis and has the inherent limitations of this design. Furthermore, the ACEF-SVR score was developed on a narrow subset of patients undergoing a specific operation in a single institution, hence precluding the generalizability of the model to other cohorts of patients. Some variables were not available for the entire population (i.e. the left atrial volume and the diastolic pattern), and were excluded from the multivariable analysis. We cannot exclude that they could offer additional prognostic information, even if multicollinearity and intercorrelation with other factors like atrial fibrillation are anticipated. Finally, no external validation was performed.
CONCLUSION The ACEF-SVR model, starting from an extremely simplified model of risk, enabled improvements in the accuracy and calibration of the model on a subset of the HF population undergoing SVR. This new model meets the current need for procedure-
specific models to be applied on specific subset of patients, avoiding the error of roughly assigning a probability of outcome of a population. Future directions should explore, first, the possibility to externally validate the ACEF-SVR. Secondly, it will be mandatory to extend the current analysis over a longer time interval looking, for instance, at 1-year mortality or, even better, including a different outcome more appropriate for this subset of patients. A major concern for a population affected by advanced HF is the risk of surviving with repeated hospitalizations for HF, which, in turn, worsen the quality of life and increase the public health burden in terms of financial resources and costs. Conflict of interest: none declared.
REFERENCES [1] Nilsson J, Algotsson L, Höglund P, Lührs C, Brandt J. Comparison of 19 preoperative risk stratification models in open-heart surgery. Eur Heart J 2006;27:867–74. [2] 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. [3] Roques F, Michel P, Goldstone AR, Nashef SA. The logistic EuroSCORE. Eur Heart J 2003;24:882–83. [4] Nashef SA, Roques F, Sharples LD, Nilsson J, Smith C, Goldstone AR et al. EuroSCORE II. Eur J Cardiothorac Surg 2012;4:734–44. [5] Jin R, Furnary AP, Fine SC, Blackstone EH, Grunkemeier GL. Using Society of Thoracic Surgeons risk models for risk-adjusting cardiac surgery results. Ann Thorac Surg 2010;89:677–82. [6] Wijns W, Kolh P, Danchin N, Di Mario C, Falk V, Folliguet T et al. Guidelines on myocardial revascularization: the Task Force on Myocardial Revascularization of the European Society of Cardiology (ESC) and the European Association for Cardio-Thoracic Surgery (EACTS). Eur J Cardiothorac Surg 2010;38:Suppl:S1–S52. [7] Menicanti L, Castelvecchio S, Ranucci M, Frigiola A, Santambrogio C, de Vincentiis C et al. Surgical therapy for ischemic heart failure: singlecenter experience with surgical anterior ventricular restoration. J Thorac Cardiovasc Surg 2007;134:433–41. [8] Menicanti L, Castelvecchio S. Left ventricular reconstruction concomitant to coronary artery bypass surgery: when and how? Curr Opin Cardiol 2011;26:523–27. [9] Castelvecchio S, Menicanti L, Di Donato M. Surgical ventricular restoration to reverse left ventricular remodeling. Curr Cardiol Rev 2010;6:15–23.
S. Castelvecchio et al. / European Journal of Cardio-Thoracic Surgery
[15] Shanmugam G, West M, Berg G. Additive and logistic EuroSCORE performance in high risk patients. Interact CardioVasc Thorac Surg 2005;4: 299–303. [16] Bhatti F, Grayson AD, Grotte G, Fabri BM, Au J, Jones M et al. The logistic EuroSCORE in cardiac surgery: how well does it predict operative risk? Heart 2006;92:1817–20. [17] Howell NJ, Head SJ, Freemantle N, van der Meulen TA, Senanayake E, Menon A et al. The new EuroSCORE II does not improve prediction of mortality in high-risk patients undergoing cardiac surgery: a collaborative analysis of two European centres. Eur J Cardiothorac Surg 2013;44:1006–11. [18] Quader MA, McCarthy PM, Gillinov AM, Alster JM, Cosgrove DM III, Lytle BW et al. Does preoperative atrial fibrillation reduce survival after coronary artery bypass grafting? Ann Thorac Surg 2004;77:1514–24. [19] Ad N, Barnett SD, Haan CK, O’Brien SM, Milford-Beland S, Speir AM. Does preoperative atrial fibrillation increase the risk for mortality and morbidity after coronary artery bypass grafting? J Thorac Cardiovasc Surg 2009;137: 901–6. [20] Barili F, Pacini D, Grossi C, Di Bartolomeo R, Alamanni F, Parolari A. Reliability of new scores in predicting perioperative mortality after mitral valve surgery. J Thorac Cardiovasc Surg 2014;147:1008–12.
ADULT CARDIAC
[10] Jones RH, Velazquez EJ, Michler RE, Sopko G, Oh JK, O’Connor CM et al. Coronary bypass surgery with or without surgical ventricular reconstruction. N Engl J Med 2009;360:1705–17. [11] Yancy CW, Jessup M, Bozkurt B, Butler J, Casey DE Jr, Drazner MH et al. 2013 ACCF/AHA guideline for the management of heart failure: a report of the American College of Cardiology Foundation/American Heart Association Task Force on practice guidelines. Circulation 2013;128: e240–327. [12] Ranucci M, Castelvecchio S, Menicanti L, Frigiola A, Pelissero G. Risk of assessing mortality risk in elective cardiac operations: age, creatinine, ejection fraction, and the law of parsimony. Circulation 2009;119:3053–61. [13] Windecker S, Kolh P, Alfonso F, Collet JP, Cremer J, Falk V et al. Guidelines on myocardial revascularization: the Task Force on Myocardial Revascularization of the European Society of Cardiology (ESC) and the European Association for Cardio-Thoracic Surgery (EACTS). Developed with the special contribution of the European Association of Percutaneous Cardiovascular Interventions (EAPCI). Eur J Cardiothorac Surg 2014;46:517–92. [14] Cook NR, Ridker PM. The use and magnitude of reclassification measures for individual predictors of global cardiovascular risk. Ann Intern Med 2009;150:795–802.
e205
Copyright of European Journal of Cardio-Thoracic Surgery is the property of Oxford University Press / USA and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
