Increase in power by obtaining 10 or more controls per case when type-1 error is small in large-scale association studies

Table 3 P-values for 4 SNPs as the number of controls per case increases in the PLCO GWAS Explorer data, by averaging 50 random samples of controls at each level of controls per case†. For the pancreatic SNP rs635634, we also calculate the percent of the 50 random samples that were statistically significant at p < 5 × 10^–8

	Melanoma rs605965		Melanoma rs871024		Prostate rs6983267		Pancreas rs635634
Controls per case	P-value	P-value ratio vs 4 controls per case	P-value	P-value ratio vs 4 controls per case	P-value	P-value ratio vs 4 controls per case	P-value	P-value ratio vs 4 controls per case	% Significant
1	2E-07		3E-07		2E-04		1E-04		2%
2	2E-09		3E-09		3E-05		5E-06		4%
3	7E-11		3E-10		7E-05		1E-06		14%
4	1E-11	1	6E-11	1	5E-05	1	4E-07	1	24%
5	1E-12	10	3E-11	2	4E-05	1	2E-07	2	30%
10	1E-13	108	3E-12	22	1E-05	4	3E-08	14	58%
25	1E-14	991	6E-13	96	4E-06	13	1E-08	36	90%
50	-	-	-	-	-	-	6E-09	65	100%
100	-	-	-	-	-	-	5E-09	85	100%

^† We chose individuals with SNPs genotyped on the Illumina Global Screening Array platform. We include the top 2 SNPs for melanoma (2093 cases; rs605965: OR = 1.47, MAF = 4.2%; rs871024: OR = 1.26, MAF = 49.6%), the top SNP for prostate cancer (2012 cases; rs6983267: OR = 0.86, MAF = 50%) and the top SNP for pancreatic cancer (578 cases; rs635634: OR = 1.48, MAF = 18.5%). There were 57,501 cancer-free controls and thus melanoma and prostate cancer had a maximum of 25 controls per case in the data

ISSN: 1471-2288