1999 Nature America Inc.• http://genetics. nature. com correspondence

20 nature genetics• volume 22• may 1999 truncated A2M fragments and A2M genotype (plasma/serum, 10 deletion carriers versus 21 insertion homozygotes; brain homogenates, 8 deletion carriers versus 3 insertion homozygotes; Fig. 1b). Plasma A2M levels measured by immuno-electrophoresis10 and nephelometry, which are known to be stable in a given individual11, were similar between 7 insertion homozygotes (mean±sd= 1.80±0.36 mg/ml) and 7 deletion homozygotes (1.61±0.37 mg/ml). Three possible explanations are envisioned for the disparity between these and previous results1. First, the association is restricted to a small subset of cases (for example NIMH site 3). Similar to all other data sets in this and the previous report, however, the NIMH 3 site subset was collected from a mixed population of European descent. Second, population stratification or a systematic diagnostic difference might have occurred in NIMH site 3, but the NIMH collection employed validated standard procedures similar to our own. Third, the reported association may be spurious. This would be concordant with:(i) the reduction in strength of association when more samples from the original NIMH data set are added;(ii) the biological implausibility of results in the Blacker et al. study showing random transmission of A2M alleles among sibs destined to be affected yet a tendency toward non-sharing among sibs destined to be discordant for AD;(iii) the failure to replicate the association in any of the four independent data sets examined here or elsewhere12; and (iv) the absence of a biochemical defect in A2M protein. The prior genetic evidence for an AD susceptibility locus on chromosome 12 (refs 3, 4, 7) would therefore seem most likely to arise from genetic variations other than in A2M* 2 alleles.