The EuroSCORE – still helpful in patients undergoing isolated aortic valve replacement?☆ Free

Abstract

Background: The European System for Cardiac Operative Risk Evaluation (EuroSCORE) is one of the most prominent scores used for the evaluation of predicted mortality in cardiac surgery. The aim of our study was to analyze the logistic and additive EuroSCORE in view of its accuracy for patients undergoing isolated aortic valve replacement (AVR). Methods: A total of 652 patients underwent isolated AVR from January 1999 to June 2007. Emergency and redo operations were included. Acute endocarditis was excluded. Out of logistic regression analyses, receiver operating characteristic (ROC) curve statistics were calculated both for the logistic and additive EuroSCORE. Results: By using the identical variables used in the EuroSCORE, the area under curve was 70.7% for the logistic and 72.4% for the additive EuroSCORE, respectively. If age, which is by nature positively correlated with increasing cardiac and non-cardiac comorbidity, is calculated as a single parameter, the area under curve remains at 69.9% being very close to the result of the EuroSCORE. Conclusions: For the subgroup of patients undergoing isolated AVR, the use of the EuroSCORE provides a comparable precision concerning the estimation of early mortality compared with the simple factor ‘age’. The extended use of the EuroSCORE in view of percutaneous AVR, the insufficient accuracy of the score bears the risk of incorrect decision-making.

1. Introduction

The European System for Cardiac Operative Risk Evaluation (EuroSCORE), which was designed to predict operative mortality in cardiac surgery patients, is currently one of the most prominent risk calculators. This widely used risk calculation system was first introduced in 1999 and considers 17 risk variables, including patient-, cardiac-, and operative-related factors. The additive EuroSCORE was the first available model which offers an easy calculation by simply adding the values of each relevant variable, whereas the logistic EuroSCORE was released later but has to be calculated in quite a complex way [1]. Nevertheless, the logistic model of the EuroSCORE enables an appropriate weighting to various risk factors especially in ‘high-risk’ patients compared to the additive model [2, 3].

At present, the EuroSCORE and above all the logistic model is commonly used to identify potential ‘high-risk’ subgroups of patients who were considered for off-pump transcatheter aortic valve implantation (TAVI) procedures. However, the predictive value and accuracy of the EuroSCORE for patients undergoing isolated aortic valve replacement (AVR) is currently intensively discussed in view of TAVI and there is some recent evidence for over-prediction of procedural risk in those patients [4–8].

The present study, therefore, aimed to analyze the predictive value and the accuracy of the logistic and additive EuroSCORE in patients undergoing surgical isolated AVR in comparison to the simple risk factor ‘age’ which is by nature positively correlated with increasing cardiac and non-cardiac comorbidities.

2. Material and methods

2.1. Patients

A total of 652 consecutive patients underwent isolated AVR for pure or predominant aortic valve stenosis at the West-German Heart Center Essen, University Hospital Essen, Germany. All study patients were operated from January 1999 to June 2007. The present study obtained Institutional Review Board approval according to the Declaration of Helsinki. Patients were excluded from the study, if (1) significant aortic regurgitation was predominant, (2) had active endocarditis or (3) concomitant procedures like coronary artery bypass grafting (CABG) or other valve procedures 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 as well as decalcification of the anterior mitral leaflet were included.

2.2. Data collection

Patient demographics and operative data were recorded in an institutional database and were retrospectively extracted and evaluated. Perioperative deaths (30-day mortality) as well as long-term survival were tracked from the institutional database or by active follow-up and the calculation of completeness of follow-up was 99.2%.

2.3. Risk calculation

The online available EuroSCORE risk calculator was used both for additive and logistic EuroSCORE calculations. The required variables for calculation of the EuroSCORE are illustrated in Table 1 . All study patients had an isolated AVR and, therefore, fulfilled the item ‘operation other than isolated CABG’. Patients presenting with endocarditis were excluded from the present study. The additive and logistic EuroSCORE was calculated for each of the 652 patients. All variables and items used for the EuroSCORE were calculated and considered according to their exact definition of this score (Table 1).

2.4. Definition of age groups

In a next step, in order to identify different risk groups according to age, patients were classified according to their age presenting at time of surgery. Patients were separated into a group younger than 75 years and another group older than 75 years.

2.5. Outcome measures

The entire outcome measures within the present analysis are represented by in-hospital mortality (30-day mortality). Long-term survival was evaluated by active follow-up. Kaplan–Meier survival curves were generated for both age groups.

2.6. ROC analysis

Receiver operator curves were generated both for the additive and logistic EuroSCORE. Receiver operating characteristic (ROC) calculation was not based on the resulting scores of the patients, but included all items used for the EuroSCORE calculation, since all ‘estimates’ describing the loading of each item within the EuroSCORE have been released and were available online. In addition, receiver operator curves were calculated for the single factor ‘age’ including all patients. Moreover, ROC analysis was performed for the two age groups (<75 years and >75 years). For all calculations, 30-day mortality was used in order to define ‘events’ within the ROC analysis. Sensitivity and specificity of expected vs. observed mortality were summarized by receiver operator curves and the area under the resulting curve (AUC) expressed as c-index and accompanied by 95% confidence intervals (CI). A decreasing value of this statistic from 1.0 toward 0.5 indicates decreasing distinctiveness or discrimination between patients living and dead within 30 days after surgery.

2.7. Statistics

Descriptive statistics are summarized for categorical variables as frequencies (%). Continuous variables were reported as mean±standard deviation (S.D.). Survival curves were generated by the use of the Kaplan–Meier method. All statistical calculations including ROC analysis were performed using the SPSS System^®, version 16.0.1 (SPSS Inc, Chicago, Illinois, USA).

3. Results

3.1. Patient population

Between January 1999 and June 2007, a total of 652 patients underwent isolated AVR. According to the order of the EuroSCORE variables, mean age was 67.7±11.5 years, 45.0% (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 neurological dysfunction, 7.2% (n=47) had previous cardiac surgery, 18.9% (n=123) of the patients had a serum creatinine >200 μmol/l, a total of 51 patients presenting with active endocarditis were excluded from the present study, 0.9% (n=6) of the patients were in a critical preoperative state, none of the patients had instable angina, 33.5% (n=218) had a moderate and 10.7% (n=70) had a poor left ventricular ejection fraction (LVEF) and 1.0% (n=7) of the patients had a recent myocardial infarction (MI). Pulmonary hypertension was present in 51 of the patients (7.8%). An emergency operation was performed in 0.8% (n=5) of the patients, all patients fulfilled the criteria of ‘other than isolated CABG’, and no patient had an aneurysm repair or postinfarction ventricular septum repair. All patient demographics and comorbidities according to the EuroSCORE variables are presented in Table 2 . All operative data are listed in Table 3 .

3.2. Risk calculation and 30-day mortality

Preoperative risk calculation resulted in a mean additive EuroSCORE of 6.45±2.6% (median 6%) and mean logistic EuroSCORE of 8.46±7.9% (median 5.8%), whereas 30-day in-hospital mortality for all patients was 2.5% (16/652 patients). All 16 in-hospital deaths were not valve related. Causes of death were respiratory failure requiring prolonged mechanical ventilation and subsequent pneumonia in two, prolonged ICU stay because of low cardiac output syndrome in six, sepsis followed by multi-organ failure in four and pulmonary embolism in one patient. In three patients, cause of death was unknown.

A total of 492 patients could be identified as to be younger than 75 years. Within this group, nine patients died within 30 days of surgery (1.8%) and another 160 patients were older than 75 years presenting a 30-day mortality of 4.4% for these patients (n=7).

3.3. Long-term survival

Survival differed significantly between the different age groups: freedom from all-cause death in patients younger than 75 years was 94.2% at one year, 91.9% at two years and 81.2% at five years, respectively. Patients older than 75 years showed a survival of 89.9% at one year, 85.1% at two years and 65.2% at five years. Kaplan–Meier survival curves for both age groups are presented in Fig. 1 .

3.4. ROC analysis

The area under the curve respecting each variable of the EuroSCORE was 72.4 for the additive EuroSCORE (CI 61.5–80.9) and 70.4 for the logistic EuroSCORE (CI 61.7–80.0). Receiver operator curves of additive and logistic EuroSCORE calculations are illustrated in Fig. 2 . ROC analysis based on age only resulted in an area under curve of 69.9 (CI 58.4–81.3). The results of ROC curve statistics calculating the single factor ‘age’ are shown in Fig. 3 .

In addition, ROC analysis for the two age groups was calculated. For patients younger than 75 years (n=492), the area under the curve was calculated by 71.3 (CI 60.5–82.1), whereas the group of patients older than 75 years (n=160) resulted in an area under the curve of 84.7 (CI 74.7–94.7). ROC analysis for both age groups is illustrated in Fig. 4 .

4. Discussion

The main findings of the present study are that (1) both the additive and logistic EuroSCORE can be used to predict 30-day mortality in patients undergoing isolated AVR, (2) isolated AVR can be performed at very low-risk, (3) the simple factor ‘age’ results in a comparable precision concerning the estimation of early mortality compared with the EuroSCORE, however (4) a more precise and simple prediction of mortality can be achieved by the single factor ‘age’ especially in elderly patients.

Currently, surgical AVR is the golden standard treatment option in patients presenting with symptomatic aortic valve stenosis as there is no effective medical therapy [9, 10]. Surgical AVR can be actually performed at very low-risk, even in older patients [5, 6, 11, 12]. However, due to an increased life expectancy in the Western industrialized countries and increasing comorbidities, alternative treatment options like TAVI have been evolved in order to treat high-risk patients less invasively. Furthermore, nearly 30% of them were not referred to surgery due to an ‘unacceptable’ risk for surgery [4, 13]. Therefore, mainly elderly and patients at highest risk for surgical AVR were suitable candidates for new catheter-based techniques due to uncertainty regarding long-term durability and clinical outcomes compared to conventional AVR [14]. In order to identify patients at highest risk for surgery, the EuroSCORE represents one of the most common risk algorithms in cardiac surgery. However, there is some recent evidence for over-prediction of procedural risk by the use of the EuroSCORE, especially in high-risk patients [4, 6, 7]. In a previous analysis, we could demonstrate a more precise prediction of 30-day mortality in ‘high-risk’ patients (EuroSCORE >20%) undergoing isolated surgical AVR by the use of the STS risk calculator compared to the EuroSCORE [8]. As previously described, we decided to exclude the item ‘endocarditis’ within this present study in order to eliminate a potential bias when talking about aortic valve stenosis due to the fact that both were different disease patterns with different outcomes and morbidities. Furthermore, endocarditis was excluded to provide equal conditions when comparing results of conventional AVR to TAVI techniques where endocarditis represents a contraindication [8].

In the present study, we focused on age as a simple stratification factor for a group of elderly patients. Thirty-day mortality was predicted by the EuroSCORE (additive and logistic) and compared to the prediction by the single factor age regarding their accuracy by a ROC analysis. ROC curve analysis has been reported to represent a useful method to measure the ability of a clinical risk model to discriminate between hospital deaths and survivors [15].

The EuroSCORE risk calculator was primarily created to allow patient grouping for the total spectrum of cardiac surgery and was used to compare results of different centers, however, within the original EuroSCORE cohort only 17% of the patients underwent AVR and the majority was treated for coronary artery disease [4]. Over-prediction of procedural risk in patients undergoing isolated AVR, especially in high-risk and elderly patients has been recently reported [4, 6], however, the EuroSCORE is actually widely used to recruit patients for TAVI. The EuroSCORE which was released in 1999 is still based on 1995 mortality and has never been updated or recalibrated. This may explain a decreased accuracy of risk stratification by the use of the EuroSCORE during the last years compared to the current era [7]. In this entire context, these findings may reflect the actual shift in patients’ comorbidities and the increase of geriatric population as well as the medical and operative/technical progress in modern cardiac surgery and intensive care medicine. Nevertheless, the EuroSCORE is still strongly associated with mortality and still accurately stratifies procedural risk, even though it may be more useful for predicting mortality during medium-term follow-up than 30-day mortality in patients undergoing isolated AVR [6].

In the present study, ROC analysis showed acceptable values both for the additive (72.4%) and logistic (70.7%) EuroSCORE, however, when observing the single factor age the area under curve remains at 69.9%, being very close to the result of the EuroSCORE. These findings became more evident by recalculation according to an increasing age resulting in a c-index of 0.713 for patients younger than 75 years and 0.847 in elderly patients (>75 years).

Increasing age has been extensively reported as a significant independent risk factor of postoperative mortality even for AVR [5, 11]. Age has been reported to be one of the most important and independent factors for the prediction of postoperative mortality in patients undergoing isolated AVR and represents the strongest predictor for late mortality. In addition, there is some considerable evidence for several patient and procedural characteristics (e.g. NYHA class, aortic insufficiency, hypertension, mechanical valve, preoperative pacing or longer cardiopulmonary bypass time) which may affect early- and long-term survival in patients undergoing isolated AVR [16].

4.1. Limitations

The results of the present study were based on data of a single university hospital. Additionally, one can suggest that nearly one-third of patients, primarily patients at high or highest risk, were not referred to surgery by the cardiologist according to the findings of the European Heart Survey [17]. Furthermore, the small proportion of patients who died within the present study may affect results of the ROC analysis. Therefore, the cut-off point for age was set at 75 years in order to increase the small number of older patients getting more reliable results, since only nine patients died in the group younger than 75 years and seven died within 30 days after surgery within the group older than 75 years. Further large-scale, well-designed prognostic studies on post-AVR mortality must evaluate long-term outcomes and health-related quality of life assessments as well [4, 16].

5. Conclusion

In conclusion, surgical AVR can be performed at low to very low risk, even in elderly patients. Furthermore, the additive and logistic EuroSCORE remains useful so far for the prediction of 30-day mortality in patients undergoing isolated AVR, however, the single factor ‘age’ which is positively correlated with patients’ comorbidities can be likewise used for a simplified prediction of 30-day mortality in patients undergoing isolated surgical AVR. The accuracy of this single factor ‘age’ increases in elderly patients, and in the current era of TAVI this simple variable ‘age’ should be taken into account and may influence decision-guiding.

References