To the Editor:
We appreciate the comments by Lou and Vicente on the review "Risk Functions and the Primary Prevention of Cardiovascular Disease."1 We will gladly discuss some of the opinions that the authors stated in their letter.
On the first 2 points, the PROSPER2 and ALLHAT3 studies with a majority of patients over 65 years of age, in which the major concentration is on cerebrovascular disease (CVD), statins did not demonstrate any benefit. Both meta-analyses have shown that these drugs reduce absolute risk for CVD (by 0.7%) and ischaemic heart disease (IHD) (by 2%) with number needed to treat (NNT) of 268 and 60 patients respectively.4,5 One meta-analysis is a statistical technique that is subject to many limitations which cannot be substituted, under any circumstance, with the results of an adequately designed clinical trial. No clinical trial has demonstrated that the use of lipid-lowering drugs would be effective in primary prevention of CVD. The uncritical application of the meta-analyses results, of doubtful clinical significance (though statistically significant), in the primary prevention of CVD leads to overtreatment of the population. On the other hand, emphasis should be placed on treatment and control of hypertension, whose contribution to the risk of CVD is much greater. It is reasonable to optimise prevention efforts, focusing on the use of statins by estimating coronary risk, which has demonstrated its efficacy in clinical trials on primary prevention. This is not "economism" but rather clinical-epidemiology based on scientific studies.
Regarding diabetics (third point), the VERIFICA study6 demonstrated that prediction by the adapted REGICOR function did not differ significantly from real everyday practice (rate of incidence of coronary events), but it did not include those which refer to discrimination (sensitivity and specificity using an ROC curve) due to questions of space. We gladly include these data in Figure. According to these data, it should be pointed out that despite correctly discriminating the REGICOR function, the results for discrimination are somewhat worse6 (not "disappointing"). In the near future, we expect that the addition to the function of variables such as time since onset, metabolic control, or microalbuminuria will improve both characteristics.
Figure. Area under the curve in the calibrated REGICOR function. AUC indicates area under the ROC curve.
In the fourth and fifth points, Lou and Vicente argue that the important point is not which function to use, but rather the choice risk header. This is incorrect. As we have argued in the review,1 each country must use the function that is most accurate in making a prediction. Faced with a risk of 20% using the Framingham formula, we can feel vindicated in making aggressive prevention decisions that have a REGICOR equivalent of approximately 8%. However, the real risk in Spain is 8%. The cut-off points must be established by consensus among experts in the real risk and not based on overestimated risks. We do not believe that these criteria can be considered to be "economism" but rather "veracity" of the estimation.
Lou and Vicente mistake the first appearance of 78.5% (specificity) since it should have been 88.3%. We do not know either how to estimate the NNT with a relative risk reduction of 33%, since the calculation must be made using the absolute reduction of risk.
In a country with a low incidence of myocardial infarction such as Spain, it appears to be more reasonable to be stricter with the specificity than with the sensitivity, in order to avoid unnecessarily treating the population that does not need it. In the future, screening for cardiovascular risk in the population will include new markers combined with advanced imaging techniques in order to provide a better approximation of the long-sought perfect sensitivity and specificity.