We first examined all the then known alleles (1997) at the HLA-A, B, Bw, C, DRB1, 3, 4 and 5, and DQB1 loci in 55 late-onset (>65y) AD cases and 73 elderly controls from Oxford. We found an association of HLA-B7 with late-onset AD (odds ratio = 3.1, corrected P = 0.04) that was limited to apolipoprotein E epsilon4-negative subjects (odds ratio = 5.1, corrected P = 0.005). We then studied linkages with Class III genes and, finally, we sought to replicate our HLA-B7 result in cohorts from Montreal and Nottingham. Altogether, we used 299 histopathologically confirmed cases of late-onset AD and 175 controls. Our initial, clear finding was not replicated in Montreal and Nottingham, however. We also failed to support any other previously reported association of AD with an HLA gene. Though we cannot exclude distinct linkages in different cohorts as an explanation of the conflicting results of HLA/AD studies, we conclude that there is no compelling evidence of a strong, direct association between late-onset AD and any HLA Class I or II allele.