. 2014 Sep 16;11(9):e1001713. doi: 10.1371/journal.pmed.1001713

Table 5. Association of lipid genotype risk scores with LOAD per one unit increase in lipid levels excluding population controls from the MRC-WTCCC2 group.

Trait	GRS (n SNPs)	Score Calculation Method	MRC (n = 4,409^a)			IOP+ (n = 663^a)			ADNI (n = 517^a)			MRC, IOP+, and ADNI Meta-analysis (n = 5,589^a)
			OR	95% CI	p-Value	OR	95% CI	p-Value	OR	95% CI	p-Value	OR	95% CI	p-Value	Percent I²
HDL-C	Full (69 SNPs)	Individual level data	0.987	0.70–1.40	0.941	1.559	0.39–2.56	0.265	1.514	0.39–2.55	0.394	1.104	0.82–1.49	0.517	0
	Full (69 SNPs)	Summary data	0.992	0.75–1.31	0.959	1.411	0.76–2.62	0.280	1.399	0.66–2.95	0.384	1.085	0.85–1.38	0.508	0
	Trait specific (45 SNPs)		1.299	0.74–2.27	0.365	2.744	0.80–9.42	0.109	0.840	0.19–3.66	0.827	1.388	0.86–2.24	0.18	0
LDL-C	Full (55 SNPs)	Individual level data	0.623	0.36–1.07	0.088	1.309	0.24–4.22	0.677	3.652	0.24–4.22	0.077	0.840	0.53–1.34	0.465	65
	Full (55 SNPs)	Summary data	0.773	0.57–1.05	0.094	1.145	0.60–2.22	0.701	2.081	0.94–4.60	0.069	0.913	0.70–1.18	0.491	65
	Trait specific (9 SNPs)		0.840	0.20–3.45	0.82	0.401	0.02–11.07	0.604	27.220	0.60–1242	0.090	1.089	0.32–3.73	0.892	38
TG	Full (40 SNPs)	Individual level data	0.910	0.64–1.29	0.600	0.619	0.39–2.58	0.236	1.795	0.39–2.58	0.228	0.992	0.68–1.25	0.600	31
	Full (40 SNPs)	Summary data	0.910	0.64–1.29	0.607	0.632	0.30–1.34	0.235	1.74	0.71–4.29	0.231	0.923	0.69–1.24	0.595	30
	Trait specific (16 SNPs)		0.839	0.29–2.44	0.761	1.395	0.13–14.84	0.795	3.194	0.187–54.4	0.430	1.043	0.42–2.61	0.928	0
TC	Full (70 SNPs)	Individual level data	0.768	0.52–1.13	0.181	0.961	0.36–2.78	0.935	2.066	0.35–2.78	0.165	0.882	0.63–1.23	0.451	37
	Full (79 SNPs)	Summary data	0.823	0.61–1.10	0.195	0.971	0.51–1.85	0.936	1.791	083–3.88	0.141	0.917	0.71–1.18	0.501	41
	Trait specific (18 SNPs)		2.301	0.86–6.20	0.099	0.800	0.10–7.26	0.853	0.127	0.01–1.78	0.126	1.035	0.99–1.09	0.151	60

Scores were calculated using all independent SNPs associated with each trait (full) and SNPs associated exclusively with each trait (trait specific) for all datasets and pooled together using inverse-variance fixed effects meta-analysis. Full allele scores were calculated using both raw genotype data and summary data. Trait specific allele scores were calculated using summary data only. Since there was some evidence for between study heterogeneity, random effects models were also tested but did not affect the meta-analysis results.

Maximum.