Table 3.

Testing for interactions between Alzheimer’s Disease PRS and age and age² on cognitive performance.

Cognitive outcome	F-value in model with PRS Jansen	p-value	FDR-adjusted p-value	F-value in model with PRS Wightman	p-value	FDR-adjusted p-value	F-value in model with PRS Schwartzentruber	p-value	FDR-adjusted p-value
Outcomes - classical cognitive tasks
Digit span forward [number of digits]	2.174	0.114	0.579	2.344	0.071	0.320	2.116	0.096	0.432
Digit span backward [number of digits]	1.898	0.150	0.579	1.304	0.271	0.617	1.280	0.279	0.731
Corsi forward [number of blocks]	1.164	0.312	0.782	1.295	0.274	0.617	1.028	0.379	0.731
Corsi backward [number of blocks]	1.222	0.295	0.782	1.071	0.360	0.730	1.101	0.347	0.731
AVLT – immediate recall [sum of recalled words]	5.780	0.003	0.042	4.256	0.005	0.070	4.125	0.006	0.084
AVLT – delayed recall [number of words]	5.091	0.006	0.056	3.526	0.014	0.097	3.507	0.015	0.099
Word fluency task [number of animals]	0.501	0.606	0.877	0.684	0.562	0.844	0.541	0.654	0.921
Trail-making test A [log s]	10.165	<0.001	0.001	7.046	<0.001	0.003	6.976	<0.001	0.003
Trail-making test B [log s]	2.163	0.115	0.579	1.631	0.180	0.540	1.634	0.179	0.589
MWT-B [sum of correctly recognised words]	1.423	0.241	0.782	1.029	0.379	0.730	1.060	0.365	0.731
Outcomes - eye movement tasks
Log of spatial error during fixation [log °]	0.266	0.766	0.877	0.288	0.834	0.925	0.274	0.844	0.925
Log of saccade rate during fixation [log N/s]	0.700	0.496	0.877	0.584	0.625	0.844	0.779	0.505	0.853
Log of blink rate during fixation [log N/s]	0.388	0.679	0.877	0.632	0.594	0.844	1.076	0.358	0.731
Smooth pursuit velocity gain [%]	0.026	0.974	0.974	0.129	0.943	0.943	0.158	0.925	0.925
Saccade frequency during smooth pursuit [N/s]	2.000	0.135	0.579	3.733	0.011	0.097	3.551	0.014	0.099
Prosaccade latency [ms]	1.144	0.319	0.782	2.393	0.067	0.320	1.876	0.131	0.507
Prosaccade amplitude gain [%]	1.018	0.361	0.813	0.679	0.565	0.844	0.710	0.546	0.867
Log of prosaccade spatial error [log %]	0.445	0.641	0.877	0.619	0.603	0.844	0.393	0.758	0.925
Prosaccade peak velocity [°/s]	0.159	0.853	0.886	1.348	0.257	0.617	0.870	0.456	0.820
Amplitude-adjusted peak prosaccade velocity	0.820	0.440	0.877	1.965	0.117	0.395	1.562	0.196	0.589
Antisaccade latency [ms]	0.331	0.718	0.877	0.618	0.603	0.844	0.500	0.682	0.921
Antisaccade amplitude gain [%]	0.403	0.669	0.877	0.290	0.833	0.925	0.386	0.763	0.925
Log of antisaccade spatial error [log %]	0.227	0.797	0.877	2.184	0.088	0.338	2.229	0.083	0.432
Antisaccade peak velocity [°/s]	0.208	0.812	0.877	0.257	0.856	0.925	0.174	0.914	0.925
Amplitude-adjusted peak antisaccade velocity	0.270	0.764	0.877	0.168	0.918	0.943	0.179	0.911	0.925
Antisaccade costs [ms]	0.397	0.672	0.877	0.305	0.821	0.925	0.302	0.824	0.925
Antisaccade error rate [%]	0.757	0.469	0.877	0.515	0.672	0.864	0.529	0.662	0.921

The table displays the results of the likelihood ratio test, which compares the model fit of model 1 (cognitive variable ~ b₀/β₀ + PRS* b₁/ β₁ + age + age² + sex + residual error) with the model fit of model 2 (cognitive variable ~ b₀/β₀ + PRS*b₁/ β₁ + age*PRS*b₃/β₃ + age²*PRS*b₄/β₄ + age + age² + sex + residual error). The null hypothesis says that the data are equally likely under both models and can be rejected if the p-value is <0.05; otherwise, it cannot be rejected. Interpretation of the F-value: the data are “F-value” times more likely if the interaction terms “age*PRS” and “age²*PRS” are included in the model (model 2) than if they are not included in the model (model 1). We conducted the likelihood ratio tests for each PRS score separately and present the resulting F-values and corresponding p-values and FDR-adjusted p-values next to each other. The FDR-correction is based on 27 comparisons. In bold are those associations with an unadjusted p-value below 0.05.

PRS polygenic risk score, SNP single-nucleotide polymorphism, FDR false discovery rate, AVLT Auditory Verbal Learning and Memory Test.