Researchers at the University of Zurich have developed a new artificial intelligence-enhanced version of the GRACE (Global Registry of Acute Coronary Events) scoring system that provides a personalized and accurate method for assessing coronary risk of patients with non-ST-elevation acute coronary syndrome (NSTE-ACS). Published in The Lancet Digital Health, the new findings revealed by the development of GRACE 3.0 suggest that current approaches to treatment stratification for NSTE-ACS may be misclassifying patients and could be improved by leveraging individualized treatment effect predictions.
“GRACE 3.0 is the most advanced and practical tool yet for treating patients with the most common type of heart attacks,” said senior author Thomas F. Lüscher, PhD, director of the Center for Molecular Cardiology at the University of Zurich. “The new score not only predicts risk more accurately but can also be used to guide more personalized treatment decisions. This could reshape future clinical guidelines and help to save lives.”
The GRACE scoring system is used worldwide to guide treatment decisions in patients with NSTE-ACS, which is the most common form of acute coronary syndrome and a major cause of morbidity and mortality worldwide.
Using data from more than 609,000 patients with NSTE-ACS collected between 2005 and 2024, the researchers developed the one-year mortality model and validated it across the datasets from multiple European countries. In addition, a separate machine learning model was trained using information from the Danish VERDICT trial (Very Early vs Deferred Invasive Evaluation using Computerized Tomography), VERDICT is a randomized clinical trial designed to evaluate the timing of invasive coronary angiography in patients with non-ST-elevation acute coronary syndrome (NSTE-ACS).
The updated GRACE 3.0 system includes three components: an in-hospital mortality model, a one-year mortality model, and a model to predict the individualized benefit of early invasive treatment to manage the disease and disease risk. Together, these allow for refined predictions of short- and long-term mortality and offer a way to personalize invasive treatment strategies based on individual patient profiles.
Both the in-hospital mortality model (AUC 0.90) and the one-year mortality model (time-dependent AUC 0.84) demonstrated high discriminative accuracy and calibration when validated against external data. Compared with the GRACE 2.0 model, the updated versions showed improved discrimination and risk classification.
“In the context of a comprehensive clinical evaluation, GRACE 3.0 can support clinical decision making on patient triage and in personalizing secondary prevention regimens,” the researchers wrote.
Of particular interest was the development of the GRACE 3.0 individualized treatment effect model, which identifies which patients are likely to benefit from early invasive management. This model showed that high-benefit patients had a significantly reduced risk of long-term adverse cardiovascular outcomes when treated early (HR 0.60), while those with low predicted benefit did not (HR 1.06, p=0.014 for interaction).
“The results were striking. While some patients gained substantial benefit from early intervention, others showed little or no benefit,” said first author Florian A. Wenzl, MD, a physician-scientist at the University of Zurich. “This may imply a shift in how we should be managing these patients.”
Prior versions of GRACE were not specific to NSTE-ACS and had not been updated to account for modern treatment practices, such as newer stent technologies and intensified lipid-lowering therapies. In addition, earlier versions assumed a linear increase in treatment benefit with higher risk, but GRACE 3.0 has revealed that risk is more complex and has non-linear relationships between baseline characteristics and treatment outcomes.
“Patients with NSTE-ACS who benefit from early invasive management present with clinical characteristics different from those previously described. Indeed, the clinical profile of patients with predicted benefit from early invasive management is characterized by more signs of myocardial ischemia, worse hemodynamics, and a higher likelihood of being female, but younger age and better renal function,” the researchers wrote.
While GRACE 3.0 is a promising tool for clinical practice, further validation is needed, particularly in non-European populations. The researchers also noted that differences in healthcare systems, treatment availability, and unmeasured confounders may influence the model’s applicability globally.
