Phenotypical differences in the characteristics of a population affects both the mortality and the performance of a risk-scoring model.

Abstract

This commentary refers to ‘Ethnicity-dependent performance of the Global Registry of Acute Coronary Events risk score for prediction of non-ST-segment elevation myocardial infarction in-hospital mortality: nationwide cohort study’, by S.M. Moledina et al., https://doi.org/10.1093/eurheartj/ehac052 and the discussion piece ‘Refitting the predictor variables included in a model in a new cohort usually exaggerates its calibration performance’, by Y.-M. He, https://doi.org/10.1093/eurheartj/ehac476.

We thank He et al. for their interest in our article and note their observation that the application of a risk-scoring model on a new population often results in deterioration in the calibration of the model while maintaining its discriminative ability. It is important to note that the original (Global Registry of Acute Coronary Events) GRACE risk-scoring model was validated in 14 countries with predominantly Caucasian populations.1 While our population was highly heterogeneous, there were clear phenotypically differences between ethnic minorities compared with White patients. Notably, ethnic minority patients tended to present with non-ST segment elevation myocardial infarction on average 5 years younger than White patients and had worse cardiometabolic risk factor profile with significantly increased frequency of diabetes mellitus, hypertension, and hypercholesterolaemia.2 With the GRACE score previously being validated in patients from the MINAP database, the optimal calibration in White patients (who formed the majority of subjects) was not unexpected.