Validity Assessment of Low-risk SCORE Function and SCORE Function Calibrated to the Spanish Population in the FRESCO Cohorts

Recomendaciones preventivas cardiovasculares. Actualización PAPPS 2018

“…Un estudio reciente de validación del SCORE realizado en España muestra cómo todas las versiones de las funciones SCORE disponibles en España sobreestiman significativamente la mortalidad cardiovascular observada en la población española 38 . Estos resultados podrán ser contrastados con otros estudios de validación del SCORE para países de bajo riesgo actualmente en marcha.…”

Machine learning to predict cardiovascular risk

“…Clinicians are advised to assess risk at 5‐year intervals in patients with at least one important risk factor, such as exposure to tobacco, diabetes, hypertension, or dyslipidaemia . However, SCORE overestimates 10‐year cardiovascular mortality in the Spanish population, resulting in false positives that lead to unnecessary treatment as well as false negatives that fail to reflect individuals’ real risk . In addition, SCORE for low‐risk countries seems to provide acceptable results when applied in a high‐risk country, whereas the version adapted for high‐risk countries overestimates true risk.…”

Are healthy workers made equal to the general young to middle-aged population in Spain, a southern European country?

“…3 Although it was fairly recently shown that both of these functions for Spain overestimate the cardiovascular mortality observed in the Spanish population, the discrimination capacity is acceptable in the face of a calibration (the prediction of the number of fatal cardiovascular events) that is significantly inaccurate. 4 This situation is not very different from that which we have easily tolerated for years when applying, erroneously, the Framingham 5 function to southern Europe. 6,7 The authors themselves 1 ask, 'Why one more?'…”

“…In all other ways, a good re-calibration could not be obtained. This is enough to show an overall need for the re-evaluation of most investigations in relation to the models applied [1][2][3][4] to Spain and/or the Iberscore that, without GNDD or another proper technique to statistically assess the potential differences, concluded that re-calibration is a fair method and might therefore be used or discarded with no specific caution. 9 A theoretical, and not just a practical, issue is inherent in these considerations, 8,9 because showing the overall similarity in the magnitude of multivariable coefficients could represent the possible identification of a general biological rule linking risk factor levels and events when everything else is equal.…”