Centenarian controls increase variant effect sizes by an average twofold in an extreme case-extreme control analysis of Alzheimer's disease

Abstract

The detection of genetic loci associated with Alzheimer's disease (AD) requires large numbers of cases and controls because variant effect sizes are mostly small. We hypothesized that variant effect sizes should increase when individuals who represent the extreme ends of a disease spectrum are considered, as their genomes are assumed to be maximally enriched or depleted with disease-associated genetic variants. We used 1,073 extensively phenotyped AD cases with relatively young age at onset as extreme cases (66.3 ± 7.9 years), 1,664 age-matched controls (66.0 ± 6.5 years) and 255 cognitively healthy centenarians as extreme controls (101.4 ± 1.3 years). We estimated the effect size of 29 variants that were previously associated with AD in genome-wide association studies. Comparing extreme AD cases with centenarian controls increased the variant effect size relative to published effect sizes by on average 1.90-fold (SE = 0.29, p = 9.0 × 10^-4). The effect size increase was largest for the rare high-impact TREM2 (R74H) variant (6.5-fold), and significant for variants in/near ECHDC3 (4.6-fold), SLC24A4-RIN3 (4.5-fold), NME8 (3.8-fold), PLCG2 (3.3-fold), APOE-ε2 (2.2-fold), and APOE-ε4 (twofold). Comparing extreme phenotypes enabled us to replicate the AD association for 10 variants (p < 0.05) in relatively small samples. The increase in effect sizes depended mainly on using centenarians as extreme controls: the average variant effect size was not increased in a comparison of extreme AD cases and age-matched controls (0.94-fold, p = 6.8 × 10^-1), suggesting that on average the tested genetic variants did not explain the extremity of the AD cases. Concluding, using centenarians as extreme controls in AD case-control studies boosts the variant effect size by on average twofold, allowing the replication of disease-association in relatively small samples.

Fig. 1

Comparison of age at disease-onset and age at inclusion for cases and controls in previously reported case-control comparisons, and in our extreme phenotypes comparison. Weighted mean and (combined) standard deviation of the age at onset for AD cases and age at inclusion for controls. As weights, we used the sample sizes of each GWA study. Note that previous case-control studies of AD included samples from multiple cohorts, sometimes overlapping across studies. References to the cohorts reported in this figure are: [7, 8, 13, 25, 26, 30]

Fig. 2

Change in variant effect-size using extreme cases and centenarian controls relative to published effect-sizes, for 29 AD associated genetic variants. Dashed red line at E^k_EA−EC = 1 indicates same effect-size as reported in literature. Orange bars indicate nominal statistical significance for the association with AD (p < 0.05). Stars indicate significant changes of effect-size relative to previously reported effect-sizes (p < 0.05, two-sample z-test)

Fig. 3

Average increase in effect-size for the different comparisons. Average increase in effect sizes for: Extreme AD cases (N_EA = 1,073), of which early onset cases (N_eEA = 464), late onset cases (N_lEA = 609); centenarian controls (N_EC = 255); normal controls (N_NC = 1,664). 95% confidence intervals were estimated by random sampling (S = 10,000)