To the Editor: With surprise we read the paper by Drs Rea and Middleton on the association between Human Leucocyte Antigens (HLA) and longevity.1 Although we agree that previous studies on this subject have yielded conflicting results, we think that the methodological difficulties of these studies are not adequately addressed by Drs Rea and Middleton. In our opinion, it is not possible to draw valid conclusions from the data because the number of subjects is too small, the geographic background of the subjects is not scrutinized, and the meta-analysis techniques are not correctly applied. As in the previous studies, the number of the subjects in the study of Drs Rea and Middleton is too small. The sample size of studies on HLA and longevity (including the study by Rea and Middleton) varies from 452 to 3823 persons. According to power calculations, however,4 a minimum size of 320 persons for each sample is required to detect the difference in frequency if an antigen occurs, for example, in 5% of sample A and in 15% of sample B. To detect the difference between frequencies of 20% and 25%, respectively, samples of about 2400 persons each are required. In these calculations, the power of the study is set at 0.9 and the significance level at .005. The significance level is lower than the usual .05 because the number of comparisons is mostly large in HLA studies. In the previous studies, the number of antigens (and so the number of comparisons) varies from 142 to 49.5 Drs Rea and Middleton performed at least 174 statistical tests. Since they apply the Bonferroni method to adjust for multiple comparisons, the significance level should be set at .0003 and, consequently, the required sample is even larger than calculated in the examples above. In conclusion, the increase in frequency of the phenotype AlB8Cw7DR3 in nonagenarian males is far from statistically significant in the paper of Drs Rea and Middleton (P = .05), and their study on HLA and longevity is too small to give reliable results. To our knowledge there are two studies with numbers of subjects that are large enough to draw sound conclusions about the association between HLA and longevity.6, 7 Unfortunately, however, these are omitted from the listing in the paper of Drs Rea and Middleton. We question if the young blood donors are really comparable to the nonagenarians inasmuch as the young donors are resident in Northern Ireland, whereas the elderly subjects are resident in the greater Belfast area. More important, data on the migration and geographic origin of the subjects are not reported. In our opinion, such a report is necessary because the frequencies of HLA antigens can be different between geographic areas. Therefore, differences in migration between age groups might lead to differences in antigen frequency. Many factors may lead to changes in antigen frequency.8 One example is a major epidemic. If the population's history is not scrutinized for such factors, it is not justified to conclude that there is an association between HLA and longevity. A found association may be merely the outcome of unique historical and local events. We think that a more detailed analysis of the data of Drs Rea and Middleton is also needed in this aspect. Meta-analysis, or systematic review, is an efficient scientific technique9 that might be of value in the field of HLA and longevity. We think however, that it is not correct to combine results from two studies “by meta-analysis techniques” as is done by Drs Rea and Middleton. To preclude bias, reviews need a clear design and should include an exhaustive search of the literature.10 Among others, the criteria used to select studies for inclusion in their analysis are not clearly specified by Drs Rea and Middleton, and their list of previous studies on the subject is not exhaustive. Although data on the phenotype AlB8Cw7DR3 are only reported by Proust,11 we suggest that other authors of previous HLA studies would be willing to supply their data for a new analysis. Therefore, we feel that the conclusion of Drs Rea and Middleton is not valid and that there is no evidence for an association between the phenotype AlB8Cw7DR3 and aging. In summary, we appreciate the work that is done by Drs Rea and Middleton, but we think that their primary goal of addressing the difficulties of previous studies is not fulfilled.