The human major histocompatibility complex (MHC) class II region is associated with genetic susceptibility to multiple sclerosis (MS). Roles for HLA class I loci have been supported in several case-control studies, but this methodology does not consider the known linkage disequilibrium (LD) between class I and II loci. In 1258 individuals from 294 MS families, we analysed class I and II interactions. Using transmission disequilibrium test and haplotype analyses, we found positive associations between MS and several HLA-DRB1*15-HLA-A haplotypes including HLA-DRB1*15-HLA-A*02 (P = 2.41 x 10(-5)) and -HLA-A*03 (P = 8.42 x 10(-6)) and several HLA-DRB1*15-HLA-B haplotypes including HLA-DRB1*15-HLA-B*07 (P = 2.23 x 10(-10)). HLA-DRB1*15 haplotypes divergent for reported HLA-A allelic associations were equally over-transmitted, illustrating no detectable effect of HLA-A or -B alleles in cis on susceptibility. HLA-A and -B alleles on haplotypes not bearing HLA-DRB1*15 were not over-transmitted. Similarly, general over-transmission of HLA-DRB1*15 haplotypes was independent of the HLA-B allele present. Furthermore, HLA-B*07 haplotypes from HLA-DRB1*X-HLA-B*X/HLA-DRB1*X-HLA-B*07 heterozygous parents were transmitted per random expectation giving no indication of HLA-B independence or trans complementation of HLA-DRB1*15 by HLA-DRB1*X-HLA-B*07 haplotypes. These results imply that many reports of class I allelic associations in MS are class II dependent, secondary to LD with class II loci. The lack of independent class I associations suggests that virus-related class I-antigen complexes are not T-cell targets in MS. The inability to replicate confirmed case-control associations highlights the importance of family-based analyses. The frequency of allelic associations not being replicated emphasizes the requirement for constructing multi-locus haplotypes in dissecting associations in regions of tight LD.