A machined learning-based upgrade to the Global Registry of Acute Coronary Events (GRACE) 2.0 score better predicts in-hospital mortality risk in both men and women with non–ST-segment elevation acute coronary syndromes (NSTE-ACS), new research shows.
Moreover, the so-called GRACE 3.0 score reclassified about 5% of women as high risk, increasing the number of women who might benefit from early invasive treatment.
About 2% of men were shifted from high risk to the low-to-intermediate risk groups, where no early invasive treatment is indicated.
Importantly, the absolute mortality risk did not increase in the lower-risk groups for either sex, co-lead authors Florian A. Wenzl, MD, and Simon Kraler, MD, University of Zurich, Schlieren, Switzerland, report in the September 3 issue of The Lancet.
The broadly applied GRACE 2.0 has a class 1a recommendation to guide treatment in current US and European NSTE-ACS guidelines. The score was, however, derived from and validated in predominantly male populations and doesn’t account for sex-based differences in the clinical characteristics of NSTE-ACS.
Previous studies report GRACE 2.0 has poorer diagnostic accuracy in women than men with STE-myocardial infarction, but its performance in women with NSTE-ACS is uncertain.
For the current study, the investigators examined data divided by sex from 400,000 patients with NSTE-ACS from England, Wales, and Northern Ireland and an external validation cohort of 20,727 patients from Switzerland, treated between 2005 and August 2020.
GRACE 3.0 was developed with the same eight GRACE 2.0 variables — age, heart rate, systolic blood pressure Killip class, creatinine concentration, cardiac arrest, presence of ST-segment deviation, and troponin elevation — but applied a machine-learning algorithm (XGBoost) that accounts for potential nonlinear effects of baseline variables in men and women.
Results show that the ability of GRACE 2.0 to predict in-hospital death was good in men, with an area under the receiver operating characteristic curve (AUC) of 0.86, but notably lower in women, at 0.82.
Calibration was “suboptimal in both sexes, with pronounced underestimation of in-hospital mortality risk in female patients,” Wenzl said when presenting the results at the recent European Society of Cardiology Congress.
He noted that previous clinical trials and patient-level meta-analyses have shown that “early invasive treatment benefits only patients who are at high in-hospital mortality risk but show no benefit in those at low-to-intermediate risk.”
The discriminatory performance of the GRACE 3.0 score, however, was “remarkably better” than GRACE 2.0, reaching AUCs of 0.84 in females and 0.88 in males in the UK cohort and 0.87 and 0.91, respectively, in the Swiss external validation cohort, he said.
Calibration was superior to that of GRACE 2.0 and was good overall in both cohorts in unreported data.
GRACE 3.0 reclassified 7.9% of female patients from the low-to-intermediate risk group to the high-risk group and moved 2.7% down from high-risk to low-to-intermediate risk. Among male patients, 3.6% were moved into the high-risk group and 5.8% reclassified as lower risk.
This corresponds to a net loss of -5.2% of women previously considered low-to-intermediate risk, and a net gain of 2.2% of men in the lower risk group.
“Our study suggests that female patients in the low-to-intermediate risk group might benefit from early invasive treatment,” Wenzl said.
Commenting further, he said GRACE 3.0 provides an “updated tool for risk stratification in NSTE-ACS” and that an online calculator will be available soon.
Room for Improvement
The revised score is a welcome addition in the “spectrum of undertreatment” and the 5% to 6% of women reclassified as high risk is “of great interest for daily clinical practice,” Eva de Miguel-Balsa, MD, Elche University General Hospital in Alicante, Spain, remarks in an accompanying editorial.
“However, the implications for prognosis are nuanced,” she suggests. “Although discrimination improves, the score still performs better in male patients, suggesting additional opportunities for improvement. Prospective studies are required to assess the effect of the revised score on clinical management and outcome once it becomes available for both clinical and research purposes.”
De Miguel-Balsa also highlights as limitations of the study the fact that only a third of patients were women and sex was considered solely as a binary variable.
Reached by email for comment, senior author Thomas Lüscher, MD, said, “This argument is not valid in my mind as a third means 150,000 patients! No one has studied more patients in this context so far.”
He noted that discussions are ongoing to validate GRACE 3.0 “in another large cohort from a different country,” but also points to the need to address the undertreatment of women.
“The management of women has to improve not only as regards to coronary angiographies and PCI (10% less than males), but also in terms of secondary prevention (less statin prescription than males),” said Lüscher, of the Royal Brompton & Harefield Hospitals, London, UK.
The study was funded by the Swiss National Science Foundation, Swiss Heart Foundation, Lindenhof Foundation, Foundation for Cardiovascular Research, and Theodor-Ida-Herzog-Egli Foundation. Wenzl and Miguel-Balsa report no relevant financial relationships. Lüscher holds leadership positions at the European Society of Cardiology, Swiss Heart Foundation, and the Foundation for Cardiovascular Research-Zurich Heart House.
The Lancet. Published in the September 3, 2022 issue. Full text, Editorial
Follow Patrice Wendling on Twitter: @pwendl . For more from theheart.org | Medscape Cardiology, join us on Twitter and Facebook .
Source: Read Full Article