To the Editor:

It was with interest that we read and discussed the remarks Campillo et al made with regard to our article.1 We would like to thank the authors for their comments, particularly those referring to the Kappa index, which assist us in understanding and contextualising our objectives.

The authors mention the overlapping confidence intervals for the resulting and validated diabetes prevalence rates. On this point, we do not agree with their conclusions due to the following: a) the overlap is only marginal; b) each confidence interval excludes the alternative score; and c) each difference is statistically significant (McNemar's Test, P<10-6).

With respect to the observation on the correctness of using the Kappa index as an agreement measurement between the information collected by the questionnaire and the corresponding biometric models, it is true that a study with these characteristics is not the normal setting for applying the index. However the k value is not the most important result in this analysis, although excessive attention may well have been drawn to it in the text. Rather, the most important topic for discussion is the different approach used in a population-based cardiovascular study like this one (based on interviews and on a small, highly select number of biometric tests performed a single time to reach a diagnosis, which are collectively valid although they may be individually inexact) and our focus on clinical use in a hospital setting with the possibility of requesting and repeating a high number of tests. Our objective was to evaluate the reliability of these diagnoses based on polled individuals' responses within the context of cardiovascular risk factor studies, which are typically included in transversal population-based and follow-up studies; furthermore, these studies examine the confusion variables necessary for adjustment, while the diagnosis is often of no interest. This is a common approach on the pages of the Revista Española de Cardiología, which maintains a permanent "Epidemiology and Prevention" section in its publication.

In keeping with our approach, the absolute validity of the questionnaire is defined by sensitivity and specificity values, while other indexes, whether apparent or not, may provide complementary information of varying degrees of interest for epidemiologically or clinically-minded readers. We feel that one of the study's strong points is having contributed enough data to be able to calculate alternative indexes.

We thank Campillo et al for their remarks which allowed us to clarify a distinction that might not otherwise have been observed.