Predicting the risk of coronary heart disease I. The use of conventional risk markers.

Description

Elaboration of appropriate risk prediction models is essential to identify high-risk individuals and therefore prevent fatal or nonfatal coronary heart disease (CHD). This paper reviews the different risk prediction models available. Two criteria must be taken into consideration to assess risk prediction models: calibration and discrimination. A well-calibrated model correctly estimates the average risk of a group of individuals. A model that discriminates well, ranks individual risk in the correct order according to the population as a whole. Such a model will have high sensitivity and specificity. Discrimination can be illustrated by receiver operator characteristic curves. There are also other approaches to test the performance of a model which are its ability to reclassify people into risk groups that are more accurate than alternatives and how much of the variation in risk it explains. The first research used to develop a risk prediction model was the Framingham Heart Study which included what have become to be known as conventional risk factors such as blood pressure, smoking, blood lipids and adiposity. It has been found that the Framingham equations tend to overestimate risk in low-risk populations and underestimate it in high-risk populations. In response to the deficiencies of the Framingham equations, addition of new variables was suggested in order to improve their performance. Recently, C-reactive protein (CRP) has been studied because of its reported positive association with CHD, but it was found that CRP did not add substantial predictive value beyond that of conventional risk factors for CHD. Other risk scores were developed over the years such as ASSIGN, which was the first risk score to incorporate socioeconomic deprivation and family history; PROCAM; the Reynolds risk score which is different for women and men; and more recently the QRISK based on a British population which appears to be superior to the Framingham risk score. However, it will need to be validated outside the UK and more evidence will be necessary before a change in practice can be recommended.