Epidemiological data suggest that some genetic determinants of Alzheimer's disease (AD) might reside in those polymorphisms for the immune system genes that regulate immune inflammatory responses, such as the major histocompatibility complex (MHC). Therefore, MHC polymorphisms have been the focus of a large number of AD association studies. Class Ia, Ib (hemochromatosis gene (HFE)), class II and class III (complement, tumour necrosis factor and heat shock proteins) alleles have been studied. Nearly every positive result has been followed by several studies that have failed to replicate it or that have contradicted it. Several factors, including methodological biases, might explain these discordant results. However, the discordant results obtained with the same alleles in the various populations might also indicate linkage with another nearby locus, different in the diverse populations. In fact, the non-random assortment of alleles at neighbouring loci, i.e. ancestral haplotypes (AH), has been claimed to be maintained as the result of directional selection, i.e. molecular cooperation during the immune response. Thus, AH studies might contribute to explaining why discordant results are obtained with the same alleles in different populations. Hence, it has been suggested that the overall chance of a subject to develop AD might be profoundly affected by a 'susceptibility profile' reflecting the combined influence of inheriting multiple high-risk alleles. Discordant results may be due to other genetic factors not determined in these MHC studies and multivariate analysis in large patient cohorts considering both MHC and non-MHC genes are therefore necessary.