Clinical background
The GRACE calculator meets a concrete need: to estimate the risk of in-hospital death on arrival in acute coronary syndrome (ACS) and thereby to guide both triage and the timing of an invasive strategy. Unlike TIMI, which was derived in selected trial populations, GRACE rests on a multinational registry reflecting real-world practice across the whole ACS spectrum (STEMI, NSTEMI and unstable angina) [1]. This makes it broadly applicable, but it also means that the treatment landscape has changed since its derivation, which affects calibration in today's populations.
Calculating the GRACE score
The score is a weighted sum of eight variables collected on arrival:
where the variables are age, heart rate, systolic blood pressure and creatinine (each through stepwise intervals), Killip class, cardiac arrest at presentation, ST-segment deviation on the ECG, and raised initial cardiac biomarkers. Each variable contributes points determined by a scoring table or interval banding. Killip class is scored directly: I (no heart failure) = 0, II (rales/S3) = 20, III (pulmonary oedema) = 39, IV (cardiogenic shock) = 59. Cardiac arrest at presentation gives 39 points, ST-segment deviation 28 points and raised initial cardiac biomarkers 14 points.
The derivation cohort consisted of 11,389 patients with ACS with and without ST elevation, enrolled in the Global Registry of Acute Coronary Events between April 1999 and March 2001, of whom 509 died during the hospital stay [1]. The eight variables accounted for 89.9% of the prognostic information for in-hospital mortality. The strength of the model lies in the fact that all the variables are clinically available within minutes of arrival.
Later versions of GRACE (1.0 and 2.0) use the same eight variables but with different weightings and address other outcomes, among them mortality at 6 months and the composite of death or myocardial infarction. These are generated through separate nomograms and web calculators [2, 5].
Interpretation in practice
The calculator assigns the patient to one of three risk bands for in-hospital mortality:
| Total score | Risk category | Clinical action |
|---|---|---|
| ≤108 | Low risk | Standard management; invasive investigation can be planned electively if indicated |
| 109–140 | Intermediate risk | Individual assessment; an early invasive strategy should be considered, particularly with other risk markers |
| >140 | High risk | Early invasive strategy within 24 hours according to current European and American guidelines [5] |
The threshold of >140 is the one that guidelines explicitly use to indicate high ischaemic risk and to warrant coronary angiography within 24 hours. It is important to note that different GRACE versions (original, 1.0, 2.0) can place the same patient in different risk bands, since the weighting of the variables differs [5].
Validation and performance
In the derivation study a c-statistic of 0.83 was achieved for in-hospital mortality [1]. In a confirmation cohort of 3,972 subsequent GRACE patients the c-statistic was 0.84, and in the external GUSTO-IIb cohort (12,142 patients) 0.79 [1].
For the 6-month model, derived in 43,810 patients from 94 hospitals in 14 countries, the prospectively validated c-statistic was 0.81 for death and 0.73 for the composite of death or myocardial infarction from admission to 6 months after discharge [2].
In a modern external validation of 9,803 patients in three independent cohorts (Heidelberg, Newcastle and SPUM-ACS, enrolled 2009–2017), the original GRACE performed less well than in its derivation, with an AUC of 0.741 for in-hospital mortality in the derivation cohort and 0.780 in the pooled population [3]. When the binary biomarker variable was replaced with continuous high-sensitivity troponin T, discrimination improved markedly: the AUC rose to 0.835 in the derivation cohort and 0.878 in the pooled population for in-hospital mortality, with a net reclassification improvement (NRI) of 0.097 [3].
A large-scale sex-specific evaluation of GRACE 2.0 in 420,781 patients with NSTE-ACS from the United Kingdom and Switzerland showed that discrimination was lower in women (AUC 0.82) than in men (AUC 0.86, p<0.0001) [4]. The score underestimated in-hospital mortality in women, so that they were more often incorrectly stratified to low or intermediate risk, where an early invasive strategy is not recommended. A revised model (GRACE 3.0) that takes sex differences into account improved the AUC to 0.91 for men and 0.87 for women in external validation [4].
Limitations
Binary biomarker variable. Handling raised initial cardiac biomarkers as yes/no is a weakness in an era when high-sensitivity troponin is measured continuously. The binary categorisation loses prognostic information, and replacing it with continuous troponin appreciably improves discrimination [3].
Sex bias. GRACE was derived in a predominantly male population and underestimates risk in women with NSTE-ACS, which can lead to undertreatment with respect to an early invasive strategy [4].
Several incompatible versions. The guidelines give GRACE >140 as the threshold for an early invasive strategy without specifying which version should be used. In a comparative study of 8,618 patients, up to 63.8% of patients were reclassified into a different risk category depending on the GRACE version used, which has direct clinical consequences for the proportion of patients sent for early angiography [5].
Derived in a different treatment era. The cohort was collected in 1999–2001, before the routine use of an early invasive strategy, modern P2Y12 inhibitors and high-sensitivity troponin. Mortality in ACS has fallen since then, which means that the score today tends to overestimate the absolute risk, while the relative ranking is largely preserved.
Killip class. In clinical practice Killip class is sometimes approximated from the NYHA class and blood pressure, which introduces subjectivity and variability into the score [5].
References
- Granger CB et al. Predictors of hospital mortality in the global registry of acute coronary events. Arch Intern Med 2003. PMID: 14581255
- Fox KAA et al. Prediction of risk of death and myocardial infarction in the six months after presentation with acute coronary syndrome: prospective multinational observational study (GRACE). BMJ 2006. PMID: 17032691
- Georgiopoulos G et al. Modification of the GRACE Risk Score for Risk Prediction in Patients With Acute Coronary Syndromes. JAMA Cardiol 2023. PMID: 37647046
- Wenzl FA et al. Sex-specific evaluation and redevelopment of the GRACE score in non-ST-segment elevation acute coronary syndromes in populations from the UK and Switzerland: a multinational analysis with external cohort validation. Lancet 2022. PMID: 36049493
- Jobs A et al. GRACE scores or high-sensitivity troponin for timing of coronary angiography in non-ST-elevation acute coronary syndromes. Clin Res Cardiol 2024. PMID: 37421436
