Genomic studies developed to understand HIV-1 infection and pathogenesis have often lead to conflicting results. This is linked to various factors, including differences in cohort design and selection, the numbers of patients involved, the influence of population substructure, the ethnic origins of the participants, and phenotypic definition. These difficulties in the interpretation of results are examined through published studies on the role of polymorphisms in HLA and the chemokine receptors genes in AIDS. Our analysis suggests that the use of haplotypes will strengthen the results obtained in a given cohort, and meta-analysis including multiple cohorts to gather large-enough numbers of patients should also allow clarification of the genetic associations observed. A P-value of 0.001 appears to be a good compromise for significance on candidate genes in a genetic study. Due to the generally limited size of available cohorts, results will have to be validated in other cohorts.We developed a model to fit transversal case studies (extreme case-control studies) with longitudinal cohorts (all-stages patients) for observations on two gene polymorphisms of CCR5 and NQO1. Interestingly, we observe a protective effect for the CCR5-Delta32 mutant allele in 95% of the simulations based on that model when using a population of 600 subjects; however, when using populations of 250 subjects we find a significant protection in only 59% of the simulations. Our model gives thus an explanation for the discrepancies observed in the various genomic studies published in AIDS on CCR5-Delta32 and other gene polymorphisms: they result from statistical fluctuations due to a lack of power. The sizes of most seroconverter cohorts presently available seem thus insufficient since they include less than a few hundred subjects. This result underlines the power and usefulness of the transversal studies involving extreme patients and their complementarity to longitudinal studies involving seroconverter cohorts. The transposition approach of extreme case-control data into longitudinal analysis should prove useful not only in AIDS but also in other diseases induced by chronic exposure to a foreign agent or with chronic clinical manifestations.