Methods
We computed the effective sample size and statistical power using a web browser program, Genetic Power Calculator developed by Purcell et al. [27] (http://pngu.mgh.harvard.edu/~purcell/gpc/), for both case-control and case-parent studies. We conducted power and sample size calculations under various assumptions about genetic models (i.e., allelic, additive, dominant, recessive, and co-dominant models), minor allele frequencies (MAFs), pair-wise LD, disease prevalence, case-to-control ratio, and number of SNP markers (i.e., single SNP, 500 K SNPs, and 1 M SNPs). The values tested for heterozygous odds ratio (ORhet) were 1.3, 1.5, 2, and 2.5. The power and sample sizes were calculated under different ranges of factors, such as MAF of 5%, 10%, 20%, and 30%; LD of 0.4, 0.6, 0.8, and 1; disease prevalence of 0.01%, 0.1%, 5%, and 10%; and case-to-control ratio of 1:1, 1:2, 1:3, and 1:4. We assumed Hardy-Weinberg equilibrium at the disease-susceptible allele.
The Bonferroni p-value that was specific to the number of SNP makers tested was applied to cover 3 billion base pairs of the human genome (i.e., p = 0.05 for a single SNP marker, p = 1 × 10-7 for 500 K SNP markers, and p = 5 × 10-8 for 1 M SNP markers). We fixed the proper range of sample sizes from 100 to 2,000 cases, because the power is too low when the sample size is below 100 cases (or trios), and the cost is too high to realistically collect samples when the sample size is above 2,000 [7, 22].