Table 1. Descriptive statistics for contributing cohorts and previous meta-analyses of case-control and proxy-phenotypes of Alzheimer’s disease and LDSC output.

				LDSCa
	Phenotype	Mean χ2	λGC	h2liability (prevalence = 5%)	SE	Z	Intercept	Maternal Cases	Maternal Controls	Paternal Cases	Paternal Controls	Direct Cases	Direct Controls	N reported
Contributing Summary Data	IGAP—Stage I	1.117	1.089	0.073	0.012	6.083	1.025	-	-	-	-	21,982	41,944	63,926
	UKB Maternal	1.049	1.032	0.017	0.004	4.250	1.006	27,696	260,980	-	-	-	-	288,676
	UKB Paternal	1.031	1.029	0.012	0.006	2.000	1.009	-	-	14,338	245,941	-	-	260,279
Meta-analyses	Multivariate Model	1.139	1.079	0.069	0.008	8.625	1.001	27,696	260,980	14,338	245,941	21,982	41,944	-
	Marioni et al. (2019)	1.142	1.085	0.023	0.003	7.333	1.025	27,696	260,980	14,338	245,941	25,580	48,466	388,324
	Jansen et al. (2019)	1.114	1.068	0.020	0.002	9.500	1.000	47,793b	328,320b	-	-	24,087	55,058	455,258

^aLDSC-derived results are restricted to HapMap3 SNPs only, excluding the MHC and APOE (chr:19, bp 45,116,911: 46,318,605) regions. LDSC-derived results are derived from for the multivariate model are obtained by submitting the summary statistics from our multivariate meta-analysis to LDSC.

^bThese Ns refer to the total numbers of proxy cases and proxy controls (i.e., maternal + paternal cases and maternal + paternal controls).

Note: Heritability estimates reported here are on a liability scale, based on a population prevalence of 5%. Elsewhere in this paper, we consider sensitivity of liability scale heritability estimates to other values, and to the possibility of undiagnosed AD within control participants, and we specifically compare estimates to those obtained with a population prevalence of 5% used by Wightman et al. [4] to produce their estimate of h²_liability of 2.5%. Heritability presented for GWAX phenotypes and Marioni and Jansen meta-analyses are naïve, and have not been corrected for the indirect nature of the GWAX.