Alzheimer's disease (AD) is a genetically complex and heterogeneous disorder. To date, a large number of candidate genes have been associated with the disease, however none of these findings has been consistently replicated in independent datasets. In this study we report the results of family-based analyses for polymorphisms of five such candidates on chromosomes 2 (interleukin-1beta, IL-1B), 3 (butyrylcholinesterase, BCHE), 11 (cathepsin D, CTSD; Fe65, APBB1) and 12 (lipoprotein receptor-related protein-1, LRP1) that were all suggested to be associated with AD in recent case-control studies. To minimize the possibility of spurious findings due to population admixture, we used a family-based design applying the sibship disequilibrium test (SDT) as well as two-point parametric linkage analyses on families from the National Institute of Mental Health (NIMH) Genetics Initiative. Contrary to the initial reports, none of the polymorphisms that were analyzed showed evidence for association or linkage with AD in our families. Our results suggest that the previously reported associations from case-control studies are either (a) false positive results, e.g. due to type I error or population admixture, (b) smaller than initially proposed, or (c) due to linkage disequilibrium with an as yet unidentified polymorphism nearby.