Article Text

Download PDFPDF

Interventional cardiology
EuroSCORE as predictor of in-hospital mortality after percutaneous coronary intervention
  1. E Romagnoli1,
  2. F Burzotta1,
  3. C Trani1,
  4. M Siviglia1,
  5. G G L Biondi-Zoccai2,
  6. G Niccoli1,
  7. A M Leone1,
  8. I Porto1,
  9. M A Mazzari1,
  10. R Mongiardo1,
  11. A G Rebuzzi1,
  12. G Schiavoni1,
  13. F Crea1
  1. 1
    Institute of Cardiology, Università Cattolica del Sacro Cuore, Rome, Italy
  2. 2
    Division of Cardiology, University of Turin, Italy
  1. Dr Enrico Romagnoli, Via Sorelle Marchisio 49, 00168 Rome, Italy; enromagnoli{at}


Objective: To date, no common risk stratification system is available to predict the risk of surgical or percutaneous myocardial revascularisation in patients with coronary artery disease (CAD). Thus, we sought to assess the European System for Cardiac Operative Risk Evaluation (EuroSCORE) validity to predict in-hospital mortality after percutaneous coronary intervention (PCI).

Design, setting and participants: EuroSCORE was prospectively and systematically assessed in 1173 consecutive patients undergoing PCI in a high-volume single centre between April 2005 and October 2006.

Main outcome measure: The receiver-operating characteristics (ROC) curve was used to describe performance and accuracy of the EuroSCORE risk model for the prediction of in-hospital mortality after PCI.

Results: The EuroSCORE model demonstrated an overall relation between EuroSCORE rank and the incidence of in-hospital mortality, showing consistency in predicting patient risk across many subgroups and levels of global risk. At multivariable logistic regression analysis the EuroSCORE value was an independent in-hospital mortality predictor (p = 0.002) together with left main disease (p = 0.005), procedural urgency (p = 0.001), ACC/AHA C type lesion (p = 0.02) and PCI failure (p = 0.01). The area under the ROC curve for the EuroSCORE system was 0.91 (95% CI 0.86 to 0.97), indicating a good ability of the model to discriminate patients at risk of dying during the index hospitalisation.

Conclusion: The EuroSCORE risk model, already extensively validated for the prediction of early mortality following open-heart surgery, can also be efficiently utilised in the setting of PCI. The introduction of the EuroSCORE assessment in patients with documented CAD may help to improve the revascularisation strategy decision-making process.

View Full Text

Statistics from

In the last decade cardiac surgeons have successfully developed and validated several risk models for the prediction of early postoperative mortality, particularly for the comparison of results from different institutions and countries. In particular, in 1995 the European System for Cardiac Operative Risk Evaluation (EuroSCORE) risk model was conceived to predict in-hospital mortality for patients undergoing open-heart surgery.1 Owing to its simplicity and objectivity, this model has been subsequently validated in different subsets of patients and in different population studies.26 This scoring method is composed of groups of weighted patient-related, cardiac-related and procedural-related risk factors and it is available online at the EuroSCORE web page ( Indeed, this is the most updated surgical risk model and has the best discriminatory capacity among the existing risk score algorithms.7

More recently, several small-sized studies suggested that EuroSCORE could also be used for baseline percutaneous coronary intervention (PCI) risk stratification in selected high-risk procedures such as left main coronary artery stenting.810 Yet, a broader validation of the EuroSCORE to predict outcome in unselected patients with coronary artery disease (CAD) undergoing PCI, similar to that obtained for coronary surgery, may provide an immediate, better stratification of individual revascularisation-related risks.

In this prospective study we analysed the predictive power of the EuroSCORE risk model in the prediction of peri-procedural mortality in 1173 consecutive patients undergoing PCI.


Study population

The study population included 1173 consecutive patients undergoing elective or urgent PCI at our institution between April 2005 and October 2006.

All the procedures were performed by interventional cardiologists who met minimal proficiency criteria of the performance of ⩾100 interventional cases per year. Percutaneous transluminal coronary angioplasty and stent placement were the predominant modes of revascularisation. Different revascularisation strategies (for example, atherectomy, thrombectomy) were adopted in selected cases according to the operator’s choice. All patients were treated, unless contraindicated, with double antiplatelet therapy for one month after bare-metal stent implantation or for ⩾6 months after drug-eluting stent implantation. The use of glycoprotein IIb/IIIa inhibitors was reserved for patients undergoing primary PCI or chosen by the operators according to clinical instability and complexity of the target lesion. Written informed consent was obtained from each patient and the study was approved by the internal ethics committee.

Data collection

Demographic, clinical and procedural data were prospectively collected on a dedicated database (Cardio-planet V.3.0.8, Ebit Aet SpA, Genoa, Italy) agreed on by all the operators of the centre. Medical record of any outcome during in-hospital follow-up was completed by the patient-referring physician at the centre. In order to ensure the highest possible quality of data entry, two independent research fellows verified completeness and accuracy of the database at the end of every month. This method enabled us to obtain complete forms in >97% of the patients included in the study.

The surgical risk score of each patient was calculated according to both the additive and logistic EuroSCORE models.1 4 The American College of Cardiology/American Heart Association (ACC/AHA) angiographic classification11 was adopted to discriminate the target lesion complexity; for patients with multivessel disease or multiple lesions, the worst treated lesion was recorded. PCI failure was defined as angiographic residual stenosis >20% at the end of the procedure. Peri-procedural myocardial infarction (MI) was defined as Q-wave MI in the presence of new pathological Q-waves on the electrocardiogram in more than two contiguous leads or non-Q-wave MI in the presence of an elevation of creatine kinase (CK) level or its MB isoenzyme to ⩾3 times the upper normal limit.12 In patients presenting with acute MI, a diagnosis of re-MI required a fall and rise of CK-MB of 50% above the previous level.13

Missing data

Baseline demographics, co-morbidities, procedural data and outcomes were recorded in all patients. Among the other data, echocardiographic left ventricle ejection fraction was not recorded in 2.5% of patients and missing values were entered using a linear regression model that included age, cardiogenic shock, history of previous MI or coronary artery bypass graft (CABG), sex and history of congestive heart failure.14 15

Study end-point

Assessment of in-hospital mortality prediction of the EuroSCORE risk model in patients undergoing PCI was the end-point of the study. All in-hospital deaths were considered cardiac unless a definite diagnosis of non-cardiac death was established by clinical means or by autopsy.

Statistical analysis

Continuous variables are presented as mean (SD), and compared using the Student unpaired t test. Categorical variables are presented as counts and percentages, and compared by means of χ2 tests or Fisher’s exact test. The standard univariate Mantel-Haenszel method was used to generate odds ratios (ORs) with 95% confidence intervals (CIs) to evaluate the relation between individual factors and in-hospital cardiac mortality. Potential risk factors for cardiac death included in the analyses are reported in table 1. In the case of continuous variables with a non-linear relation with outcome, we used the cut-off points previously established in the EuroSCORE system.1

Table 1 Baseline, angiographic and procedural characteristics of study population. Predictors of in-hospital mortality by univariate logistic regression analysis are shown in detail

Individual risk factors with a p⩽0.05 value on the univariate analysis, and present in at least 2% of the population study, were entered into a backward stepwise, multiple logistic regression analysis to determine the model that best predicted in-hospital mortality. The goodness-of-fit testing (Hosmer and Lemeshow χ2) was used to assess how well the final model was calibrated.

Crude mortality and odds ratios (ORs) were calculated for various EuroSCORE classes. The receiver-operating characteristics (ROC) curve was used to describe the performance and predictive accuracy for both mortality risk and procedural success rate: average area and 95% confidence interval (CI) were calculated.16 All statistical analyses were carried out using SPSS for Windows 12.0.


A total of 26 (2.2%) in-hospital deaths occurred during the study period. Twelve patients died during the procedure or within 24 hours, seven of whom had PCI failure. The other cases of cardiac death occurred during a complicated postoperative course and were not apparently directly related to the procedure or to suspected stent thrombosis. In-hospital non-cardiac deaths were not observed during the study.

Demographics and clinical characteristics

The main clinical characteristics of the study population are reported in table 1. The mean (SD) patient age was 65 (11) (range 40–91) years, with 23.2% females. A history of previous MI was present in 27.8% of patients, previous PCI in 18.5% and previous CABG in 14.6%. The prevalence of co-morbidities was 21.8% for diabetes mellitus, 3.5% for chronic lung disease, 5.3% for peripheral vascular disease, 2.3% for neurological dysfunction and 6.8% for chronic kidney disease or end-stage renal disease. A pre-procedural critical preoperative state, including cardiogenic shock, was present in 2.4% of cases.

Diagnosis at admission was acute coronary syndrome in 53.8% of patients, including unstable angina (17.6%), acute ST-elevation MI (11.8%), non-ST-elevation MI (11.5%) and recent MI (12.9%). Impaired left ventricular function was present in 25.4% of patients, and it was associated with pulmonary hypertension in 1.8% of the cases. Univariate associations between pre-procedural characteristics and in-hospital death, with ORs and confidence intervals, are shown in table 1.

Angiographic and procedural characteristics

The angiographic and procedural characteristics are summarised in table 1. High-risk angiographic characteristics were frequently present among the study patients, including multivessel coronary artery disease in 61.3%, left main disease in 4.6%, proximal left anterior descending artery disease in 47.0%, ACC/AHA C-type lesion in 34.9%, stent thrombosis in 1.3% and in-stent restenosis in 5.4% of cases.

There were 60 (5.1%) procedural failures according to the selected definition (residual stenosis >20% at the end of the procedure), 48 (4.1%) peri-procedural complications, including 12 (1%) deaths, 33 (2.8%) Q-wave MI, 12 (1.0%) acute/subacute stent thromboses, and one (0.08%) stroke. Complete revascularisation was achieved in 57.4% of patients, requiring multivessel PCI in 19.5% and left main treatment in 2.3% of cases.

Univariate associations between procedural characteristics and in-hospital death, with ORs and confidence intervals, are shown in table 1.

Independent predictors of in-hospital mortality

The independent predictors of in-hospital mortality at the multivariable analysis included severe left ventricle dysfunction (LVEF⩽29%, p = 0.02), left main disease (p<0.01), ACC/AHA C-type lesion (p = 0.02), stent thrombosis (p = 0.05), emergent PCI (p<0.01) and PCI failure (p = 0.01). EuroSCORE risk classification, when added to the multivariate model, proved to be an important independent predictor of in-hospital mortality (p<0.01, fig 1). Comparable results were obtained using both the additive or logistic EuroSCORE. The non-significant Hosmer-Lemeshow goodness-of-fit p value (p = 0.59) indicated that the model was adequate.

Figure 1 Multivariate predictors of in-hospital mortality after percutaneous coronary intervention (PCI). EF, ejection fraction; ACC/AHA, American College of Cardiology/American Heart Association.

Risk stratification based on EuroSCORE

Data were initially analysed according to the standard additive EuroSCORE algorithm and divided in three risk groups, as previously described.1 The distribution of additive EuroSCORE values in the study population is shown in figure 2. The median EuroSCORE value was 4 (interquartile range 2.0–6.0) with 31% of procedures performed in patients deemed at high risk according to the cut-off value of 6 commonly used by cardiac surgeons.

Figure 2 Distribution of EuroSCORE values in the study population. The dotted line identifies the cut-off value commonly adopted by cardiac surgeons to discriminate high-risk patients.

The analysis among the three classic risk classes of additive EuroSCORE model demonstrated an overall relation between EuroSCORE rank and the incidence of in-hospital mortality (table 2). The observed rates of in-hospital death in the three risk strata were as follows: 0.08% for low risk group (EuroSCORE 0–2), 0.5% for the medium risk group (EuroSCORE 3–5) and 6.7% for high risk group (EuroSCORE ⩾6; p =  <0.001).

Table 2 Correlation between the additive EuroSCORE algorithm and the incidence of in-hospital mortality across several risk classes

Similar results were also obtained when a more accurate analysis of high-risk patients was performed according to the six groups classification of the additive model,4 but with an exponential risk increase among patients in the highest risk class (table 2). The observed rates of in-hospital death were 2.2% for EuroSCORE 6–8 group, 5.5% for EuroSCORE 9–11 group, 28% for EuroSCORE 12–14 group and 72% for EuroSCORE >14 group (p = <0.001). Notably, 81% of patients with a EuroSCORE value >14 had a critical preoperative state at admission.

When the OR was calculated for various risk classes, the additive EuroSCORE demonstrated consistency in predicting patient risk across many subgroups and levels of overall risk. The EuroSCORE value ⩽5 was associated with extremely low in-hospital mortality rate after PCI, while the EuroSCORE value ⩾9 identified patients at increased risk (fig 3). Interestingly, non-significant correlation was present between in-hospital mortality and the EuroSCORE 6–8 risk class. This rank probably identifies the cut-off value of this model to discriminate the low-risk and the high-risk patients and it is similar to the cut-off value of 6 commonly used for patients undergoing CABG.

Figure 3 Odds ratio calculated for different EuroSCORE risk classes. A EuroSCORE value ⩽5 was associated with extremely low in-hospital mortality rate after percutaneous coronary intervention, while EuroSCORE value ⩾9 identified patients at increased risk.

The use of logistic EuroSCORE and its analysis according to the seven groups classification,4 demonstrated similar results with a comparable predictive power among the different risk classes.

The area under the ROC curve was 0.91 (95% CI 0.86 to 0.97) for both the additive and logistic EuroSCORE, which indicates a good ability to discriminate between patients dying during the index hospitalisation and the survivors. Notably, the ability of the EuroSCORE model to predict PCI failure was limited with an area under the ROC curve equal to 0.56 (95% CI 0.47 to 0.64), thus not significantly different from chance (fig 4). These data did not deviate significantly from the logistic model, as indicated by the non-significant Hosmer-Lemeshow goodness-of-fit test (p  =  0.945).

Figure 4 Receiver-operating characteristics (ROC) curves showing the sensitivity of prediction of in-hospital mortality and procedural success for the EuroSCORE risk algorithm. The grey line represents the absence of discrimination.


Indications for PCI are progressively increasing including a growing proportion of high-risk patients previously scheduled for coronary artery bypass surgery. In order to facilitate the decision-making process and the risk-benefit analysis, several risk models have been developed to predict mortality and peri-procedural complications, both in interventional cardiology and in cardiac surgery. To date, however, none of the available scoring systems have been shown to predict adverse outcome for both revascularisation strategies.

The EuroSCORE is a simple integer-scoring system, based on clinical characteristics, designed by cardiac surgeons to predict in-hospital mortality in open-heart surgery.1 Among the several cardiac surgical risk scores, EuroSCORE showed superior performance and accuracy, and it is the most updated and well validated risk model.7

The EuroSCORE system has also been successfully applied in the identification of patients at increased risk of an unfavourable outcome after unprotected left main coronary artery stenting.10 Furthermore, recent studies suggested that EuroSCORE could be used as selection criterion for specific risk-reduction strategies in very high-risk patients, such as off-pump strategy during CABG17 and pre-procedural use of intra-aortic balloon pump during unprotected left main PCI.18

The purpose of this study was to assess the validity of the EuroSCORE risk model for prediction of in-hospital mortality in patients undergoing percutaneous coronary revascularisation. Our results showed good performance and accuracy of this “surgical” risk model to evaluate pre-procedural PCI risk, as testified by the large area under the ROC curve. At the multivariate logistic analysis, EuroSCORE was an independent in-hospital mortality predictor together with severe left ventricle dysfunction (LVEF⩽29%), left main disease, ACC/AHA C-type lesion, stent thrombosis, procedural emergency and PCI failure. Furthermore, EuroSCORE demonstrated consistency in predicting patient risk across many strata with a good correlation between in-hospital death rate and EuroSCORE values. In particular, using the addictive standard EuroSCORE, the absolute risk of death was <1% for patients with a EuroSCORE value <6, while it was 6.7% for patients with a EuroSCORE value ⩾6.

In our study, similar results were obtained applying both the additive and logistic EuroSCORE models; these findings were probably due to the small number of patients at very high risk in the study population (for example, addictive EuroSCORE >14, logistic EuroSCORE >30.00). Although the logistic EuroSCORE model is supposed to be more accurate for high-risk patients, it overestimates observed mortality19 20 and its accuracy at predicting risk varies in different surgical subgroups. Thus, additive EuroSCORE is probably the better risk model to compare different revascularisation strategies and patient populations.

Our study confirms the usefulness of clinical variables in estimating pre-procedural risk of patients undergoing PCI. Indeed, patient-related conditions (advanced age, presence of co-morbidities), severity of cardiac disease (previous MI, severe left ventricle dysfunction, left main disease) and acute clinical presentation combined to better define individual risk. Notably, single co-morbidities and classic risk factors like diabetes mellitus failed to predict in-hospital death. Nevertheless, similar results have been previously described2123 and are perhaps related to the acute clinical presentation of many of the patients treated.

In patients undergoing open-heart surgery, EuroSCORE is not only a reliable predictor of mortality, but has also been associated with the severity of post-surgical myocardial damage,24 prolonged hospital stay, postoperative complications25 and late outcome.2 Thus, in the era of large randomised trials matching surgical versus percutaneous revascularisation approaches, the use of a common risk-prediction model as EuroSCORE, might facilitate the objective comparison of procedural outcomes following PCI and coronary artery bypass surgery.

The EuroSCORE model was originally designed to predict surgical in-hospital mortality, thus explaining the absence of angiographic variables in risk score computing. Therefore, the accuracy of the EuroSCORE system could be further increased by adding to this model some specific angiographic features that are known to influence complication rate during the procedure.22 26 27 Accordingly, in our study high-risk angiographic variables such as left main disease, ACC/AHA C-type lesion and stent thrombosis were independent predictors of in-hospital mortality after PCI at the multivariate regression analysis. The need for EuroSCORE adaptation to the PCI setting is also suggested by the limited discriminatory power of the EuroSCORE model to predict procedural failure rate in our study population. Nevertheless, these results are consistent with previous studies of comparisons between angiographic lesion classification and a clinical-based risk algorithm.22 2830

In conclusion, the present study, within the limitations of being single centre and based on a restricted number of adverse events, is the first to prove the applicability of the EuroSCORE risk model in patients undergoing percutaneous coronary revascularisation. Thus, the EuroSCORE may help cardiologists and cardiac surgeons alike to individualise the risk profile of patients in order to better define the revascularisation strategy and to appropriately counsel the patient. Further validation in larger cohort of patients is warranted to confirm discriminatory ability of this algorithm in patients undergoing PCI, and in comparison to other risk scores with established performance in this setting.29 31 32


We are indebted to the staff members, including nurses and technicians, of the Interventional Cardiology Unit of Catholic University for their contribution to the study.


View Abstract


  • Competing interests: None.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles