Table 2.

PRS association with Alzheimer's disease, amount of phenotypic variability explained, and recurrence risk conferred by PRS.

MODEL	NIA‐LOAD
	OR (of the PRS)	R 2	Recurrence risk
			99th percentile	95th percentile	90th percentile	80th percentile
AD ~ PRS 1 + AGE + SEX + PCs	1.6 (1.51–1.73)	0.068	0.22	0.19	0.18	0.17
AD ~ PRS.AD 2 + AGE + SEX + PCs	2.1 (1.92–2.25)	0.12	0.25	0.22	0.2	0.19
AD ~ PRS 1 + AGE + SEX + PCs within APOE ε4 carriers	1.53 (1.4–1.68)	0.093	0.23	0.21	0.19	0.18
AD ~ PRS 1 + AGE + SEX + PCs within APOE ε4 non‐carriers	1.74 (1.53–1.91)	0.074	0.22	0.2	0.19	0.17

MODEL 3	EFIGA
	OR (of the PRS)	R 2	Recurrence risk
			99th percentile	95th percentile	90th percentile	80th percentile
AD ~ PRS 1 + AGE + SEX	1.425 [1.31–1.53]	0.036	0.19	0.18	0.17	0.16
AD ~ PRS.AD 2 + AGE + SEX	1.433 [1.31–1.56]	0.036	0.19	0.18	0.17	0.16
AD ~ PRS 1 + AGE + SEX + APOE4	1.38 [1.27–1.5]	0.066	0.21	0.19	0.18	0.17

¹ PRS was computed genome‐wide but excluding variants in the 2 MB window around the APOE gene (human genome b38, chromosome 19: 42,905,791–46,909,393).

² PRS.AD represents the polygenic risk score that includes the effect of APOE ε4 and ε2 alleles. PRS.AD constructed with genome‐wide SNPs excluding the APOE region, but including the effect sizes SNPs that code for APOE‐ε4 (rs429358) and ‐ε2 (rs7412) alleles. The APOE‐ε4 SNP rs429358 was not genotyped or imputed in the Caribbean Hispanic dataset, we used rs769449 as a proxy for APOE‐ε4 genotype. The Pearson correlation between number of APOE‐ε4 alleles and rs769449 is 0.56 (p < 1e‐16).

³ The first three principal components (PCs) in the EFIGA families were significantly associated with clinical diagnosis of Alzheimer's Disease. The PRS was calculated by adjusting for PCs as covariates. To determine the amount of recurrence risk conferred by the PRS alone, we excluded PCs from the association analyses (and used the R² estimates of the regression model to compute recurrence risk).