Abstract: 1991 and 1998 Framingham algorithms had only slightly different discriminatory power in our populations (c-statistics of 0.68 and 0.69, respectively in Belfast and France for the 1991 algorithm as compared to 0.66 and 0.68 for the 1998 one).This suggests in practice that the variability in the measurement of LDL-cholesterol only partially contributed to the "quality" of the algorithms.Second, the author pointed out the problem of the concordance between estimated individual's probability of CHD events and the actual outcomes.Although we did not develop this point in our paper, we gave c-statistics (area under the receiving operative characteristics curve) at Table 4, which precisely represents this concordance for the different algorithms (Framingham and PROCAM) and populations (Belfast and France).Finally, we agree with Dr Wierzbicki that individual risk calculation by algorithm may be largely spurious and should be interpreted critically in the light of clinical judgment.