Researchers led from Washington University in St Louis have reported three genome-wide significant loci associated with disease risk in people who carry the mutations that cause autosomal dominant Alzheimer's disease, published in The Lancet Neurology in its June 2026 issue. The genome-wide association study used whole-genome sequencing on 101 unrelated, non-Hispanic White, symptomatic participants carrying PSEN1, PSEN2 or APP mutations and 5050 asymptomatic unrelated controls, drawn from the Knight Alzheimer Disease Research Center, the Dominantly Inherited Alzheimer Network observational study and the Alzheimer Disease Sequencing Project release 4. The associations held irrespective of which of the three mutated genes a participant carried. Taken together, the risk alleles were present in 27 of the 101 mutation carriers, 27 percent.

The signal at the CNIH4 locus was driven by a missense variant, Gly54Ser (odds ratio 11.99, 95 percent CI 5.39 to 26.64, p<0.0001). The risk allele at the CCNG1 locus was associated with higher Alzheimer's risk (odds ratio 9.56, 4.29 to 21.24) and with an age at dementia onset about ten years earlier (beta -10.15 years, 95 percent CI -17.31 to -2.77, p=0.0068); it was also associated with higher plasma TDP-43 protein levels and a wider gap between chronological age and brain age estimated from structural MRI. The RHOJ risk allele (odds ratio 5.96, 3.42 to 10.36) was associated with higher cerebrospinal fluid total tau (p=0.0056) and phosphorylated tau 181 (p=0.0006), and a lower amyloid beta 42 to 40 ratio (p=0.016), comparing DIAN participants with and without the allele. Adding APOE e4 as an interaction term showed no significant effect, and a fourth signal, at ALK, did not survive sensitivity analyses.

The question the paper addresses is an old one for families: why relatives carrying the same near fully penetrant mutation can become ill years apart. Earlier work in single large kindreds pointed to modifiers such as the APOE3 Christchurch variant and RELN, but those findings came from founder populations and have been hard to generalise. This is an association study in a rare population, not an experiment, and it does not establish that the variants cause the differences in timing.

The authors set out the constraints themselves. Despite being, to their knowledge, the largest such cohort assembled, the rarity of inherited Alzheimer's leaves a small sample, with estimated statistical power between 48 percent and 95 percent to detect the loci found; they write that the CNIH4 and ALK associations should be interpreted with caution because of possible winner's curse inflation. A linked commentary in the same issue notes the cohort comprised 5151 unrelated non-Hispanic White participants in total, which limits what can be said about families of other ancestries.