Risk stratification is essential in the management of acute coronary syndromes (ACS), particularly in non-ST-elevation ACS. ACS patients are a heterogeneous group of patients in terms of risk, which has important implications for management strategies. Early risk stratification is therefore essential.

The Global Registry of Acute Coronary Events (GRACE) score is the most widely used risk score due to its high predictive accuracy.1 Two groups in the US have developed more recent risk scores for stratification of patients undergoing coronary angioplasty, including patients with ACS.2,3

We sought to validate these US risk scores in a European population of patients with ACS and to compare their predictive accuracy with that of the classical GRACE risk score.

All consecutive adult patients (aged ≥18 years) included in a single-center registry on ACS between January 1, 2005 and October 31, 2013 were eligible. This is a continuous, prospective and observational registry. Inclusion criteria were a history of chest pain at rest or other symptoms suggestive of an ACS within 24 hours before admission associated with new or presumed new ECG changes (significant ST-T wave changes or left bundle branch block) and/or dynamic changes in levels of biomarkers of myocardial necrosis. ST-elevation myocardial infarction (STEMI) was defined as persistent (>20 minutes) ST-segment elevation. All other cases were considered non-ST-elevation ACS.

Data were recorded in a database that included demographic, clinical and patient management characteristics, as well as hospital outcome. Hypertension, diabetes and hyperlipidemia were defined as either previously known or on specific therapy. Patients were classified as smokers if they had smoked during the previous six months and were self-reported.

Decisions regarding patient management strategy, including referral for coronary angiography and myocardial revascularization, either by percutaneous coronary intervention (PCI) or by coronary artery bypass grafting (CABG), were left to the discretion of the attending physician. Only patients undergoing PCI were eligible for the present study.

For each patient a score was retrospectively assigned according to the National Cardiovascular Data Registry (NCDR) risk score, the Mayo Clinic risk score (MCRS) and the Global Registry of Acute Coronary Events (GRACE) risk score.1–3

The primary endpoint was all-cause mortality during the index hospitalization.

Other outcomes were also recorded in the study population: stroke/transient ischemic attack (TIA) and major bleeding. Stroke/TIA was defined as the presence of new neurological symptoms with signs of ischemia or bleeding on computed tomography or magnetic resonance imaging. Major bleeding was defined according to the Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO) criteria as intracranial bleeding or bleeding that caused hemodynamic compromise requiring intervention.4

Informed consent was obtained from each patient. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the institution's human research committee.

A total of 2148 patients were included in the present study. Baseline characteristics are detailed in Table 1. Risk factors and previous cardiovascular disease distribution were similar to other studies and ACS registries. There was a predominance of STEMI. Treatment and outcome are presented in Table 2. There were 96 in-hospital deaths (4.5%) in the study sample.

The MCRS showed adequate goodness of fit (p=0.337) and discriminative accuracy (AUC 0.85, 95% confidence interval [CI] 0.81–0.90) (Figure 1). Calibration was also good, except for patients with high scores, in whom predicted mortality was lower than that observed (Figure 2). The NCDR risk score also presented adequate goodness of fit (p=0.474) and good discrimination (AUC 0.87, 95% CI 0.83–0.91). Calibration was almost perfect. Finally, the GRACE risk score also showed adequate goodness of fit (p=0.482) and the highest discriminative ability (AUC 0.94, 95% 0.91–0.96) with excellent calibration. Comparing the GRACE risk score with the other scores, GRACE enabled significantly better reclassification of patients in their risk class, with a NRI >75% (Tables 3 and 4). Pseudo R-squared statistics confirmed that GRACE is the best predictive risk score: GRACE R2=0.497; NCDR R2=0.337; MCRS R2=0.339.

Figure 1.

Area under the curve for the three models. GRACE: Global Registry of Acute Coronary Events risk score; MCRS: Mayo Clinic risk score; NCDR: National Cardiovascular Data Registry risk score.

(0.08MB).

Figure 2.

Calibration of the three scores (plots of predicted vs. observed probability of death for the three scores). NCDR: National Cardiovascular Data Registry.

(0.17MB).

Risk stratification is an important step in the management of ACS patients, who are a heterogeneous group in term of characteristics and risk, particularly the subgroup with non-ST-elevation ACS. It is thus essential to perform early risk stratification to optimize decision-making for appropriate management of these patients, including the best time for revascularization and the ideal combination of antithrombotic therapies. Although these treatments improve outcome, they can increase both complications and costs. For this reason, informed decisions are important, and for this risk scores are helpful tools.

Several risk stratification scores for ACS have been published since the 1990s. The first was the TIMI risk scores for STEMI and unstable angina/non-STEMI.10,11 These were developed from cohorts of patients included in randomized clinical trials (RCTs). The later PURSUIT risk score was also based on an RCT, which can limit its application and use because real-world cohorts have different characteristics.12 The GRACE risk score1 was published in 2003, based on the GRACE multicenter ACS registry database that included the whole spectrum of ACS, reflecting a real-world cohort of patients. Due to its high predictive accuracy it is nowadays the most widely used risk score in ACS.13

More recently, both the Mayo Clinic and the National Cardiovascular Data Registry developed risk scores for patients undergoing coronary angioplasty.2,3 Both scores included patients with ACS. Although these US scores were developed from PCI registries, the variables selected were mainly clinical in nature; no angiographic data was used for risk scoring. Thus, the variables selected for these new risk scores are very similar to those used in the GRACE risk score. Both scores have been validated in US populations,14 but not in European populations, particularly in the context of ACS, and so we sought to validate both risk scores and to compare them with the classical GRACE risk score.

All the scores showed moderate predictive accuracy for in-hospital all-cause mortality. However, the GRACE score had the best performance, with significantly better discriminative accuracy as shown by ROC curve analysis. Since ROC curve analysis is considered insufficient to compare different prediction models, it is recommended to perform a reclassification study of one model compared to another. Patient reclassification was better for GRACE according to their observed risk, that is, patients with events were more often correctly reclassified to a higher risk level and patients without events to a lower risk level.

There are several possible explanations for these results. Firstly, the US scores were not developed from cohorts of patients with ACS but from larger registries of patients undergoing PCI. In these study groups, only 25–40% of the patients had ACS.2,3,14 This explains several different clinical characteristics of the patients in both US registries compared to our population of patients with ACS: our population has fewer females, fewer diabetic patients and less previous history of myocardial infarction, PCI, peripheral arterial disease and cerebrovascular disease. There was also a significant proportion of patients with STEMI, which is also an important difference compared to other cohorts of ACS patients. In addition, the scores were developed for medium-term mortality (30-day and 1-year) and we validated them for short-term mortality (in-hospital).

Our study has some limitations. Although our study population was obtained from a clinical registry as for the Mayo Clinic, NCDR and GRACE risk scores, the observational and non-randomized nature of the study might limit our conclusions, particularly for other European cohorts with different characteristics. Only patients undergoing PCI were included in the study cohort. Thus our results do not apply to ACS patients who were not revascularized. Furthermore, our study population is somewhat different from other cohorts, particularly its predominance of STEMI patients (70.9%), and thus our results cannot be directly extrapolated to other cohorts.

In a large European cohort of patients with ACS, the NCDR score and the MCRS are useful for in-hospital mortality risk stratification in patients undergoing PCI. However, the GRACE risk score has the best predictive accuracy and as such it should remain the preferred risk stratification score.

The authors have no conflicts of interest to declare.