Genome-wide association studies for Alzheimer's disease: bigger is not always better

Escott-Price and John Hardy report that large genome-wide association studies are reported as identifying loci for Alzheimer's disease loci. They suggest that many are incorrectly designated because of misdiagnoses inherent in data collection. After genome-wide association studies of dementia have been performed, significant loci should be tested in neuropathologically confirmed data sets before they are designated as Alzheimer's disease loci. As the size of genome-wide association studies increase, the number of associated trait loci identified inevitably increase. One welcomes this if it allows the better delineation of the pathways to disease and increases the accuracy of genetic prediction of disease risk through polygenic risk score analysis. However, there are several problems in the continuing increase in the genome-wide analysis of 'Alzheimer's disease'. In this review, we have systematically assessed the history of Alzheimer's disease genome-wide association studies, including their sample sizes, age and selection/assessment criteria of cases and controls and heritability explained by these disease genome-wide association studies. We observe that nearly all earlier disease genome-wide association studies are now part of all current disease genome-wide association studies. In addition, the latest disease genome-wide association studies include (i) only a small fraction (similar to 10%) of clinically screened controls, substituting for them population-based samples which are systematically younger than cases, and (ii) around 50% of Alzheimer's disease cases are in fact 'proxy dementia cases'. As a consequence, the more genes the field finds, the less the heritability they explain. We highlight potential caveats this situation creates and discuss some of the consequences occurring when translating the newest Alzheimer's disease genome-wide association study results into basic research and/or clinical practice.