To the Editor:
We believe that some of the arguments by Grau et al in a recent article deserve reflection:
1. In order to emphasize the usefulness of coronary risk functions on lipid-lowering treatment, they maintain that they have only demonstrated its efficacy in primary coronary prevention, but not on the reduction of cerebrovascular accidents (CVA), while recent meta-analyses2 have analysed the evidence on primary and secondary prevention separately with results that confirm their favourable effect.
2. They sustain that the coronary risk calculation is preferable to the cardiovascular risk calculation since this involves "overtreatment." However, all of the recent prestigious clinical guidelines (societies, NICE, etc) recommend using cardiovascular risk for stratification in primary prevention, though they later differ in the methodology.
3. In diabetics, they confirm that VERIFICA3 has demonstrated that the adapted REGICOR function (RF) "precisely" estimates the rate of coronary events at 5 years, which is true. But does this mean it is clinically sufficient? The answer does not need to be negative, since one factor is "precision" and the other, which is defined as "reliability" or classificatory validity, is what is clinically interesting. The VERIFICA study does not provide data on sensitivity and specificity; however, other studies do with discouraging results.4
4. After years of insisting on the overestimation of risk by the Framingham function (FF), the VERIFICA study comes along and confirms the hypothesis5 that would justify its use in our environment. The FF, though overestimating risk on population, maintains its limited validity for classification. Calibration of the RF substantially improves the populational predictive validity, but it barely changes the classificatory validity in terms of sensitivity and specificity. Therefore, the relevance of its clinical use is not to use one or the other but rather to define a cut-off point.
5. The choice of the cut-off point is biased, in our opinion, towards reducing spending. For Comin et al6 using data from VERIFICA, the RF (cut-off point >10%) obtained a sensitivity of 36.8% and a specificity of 78.5% for coronary events; for the FF (cut-off point >20%), these values were 57.3% and 78.5% respectively. But, which one is preferable: a sensitivity of 36.8% with a specificity of 88.3% (we treat a few patients but the reach of primary prevention will be more limited) or a sensitivity of 57.3% with a specificity of 78.5% (we treat more patients but we avoid more events)? Using their data and assuming that statin treatment reduces coronary events by 33% (New Zealand guidelines), the differences in clinical efficiency are much less (NNT=28 with FF and NNT=34 with RF) than the differences in the percentage of population identified as high risk (22.4% with FF vs 12.4% with RF). Stratification of risk is crucial for adequate primary prevention, but there needs to be a review in which there is sufficient debate on the reason for choosing a given concept and the method for calculating, as well as the possible practical results of the proposed cut-off points.