Abstract
PURPOSE: To examine whether the choice of genome-wide association study (GWAS) summary statistics can yield invalid or misleading conclusions in Mendelian randomization (MR) studies of myopia. METHODS: The relationships between (1) years of full-time education and myopia, and (2) myopia and primary open-angle glaucoma (POAG), were used as exemplar test cases. MR analyses were performed with nine different sets of summary statistics for myopia: seven from sources widely used in published MR studies, plus two newly derived sets (a GWAS for myopia in either 66,773 unrelated participants or 93,036 participants that included relatives). RESULTS: Using the two newly derived sets of summary statistics from GWAS for myopia in unrelated and related samples, MR analyses demonstrated a positive causal relationship between education and myopia: odds ratio (OR) for myopia = 1.18, 95% confidence interval (CI) = 1.10 to 1.26 and OR = 1.16, 95% CI = 1.09 to 1.23 per additional year of education, respectively, and a positive relationship between myopia and POAG: OR = 1.11, 95% CI = 1.03 to 1.19 and OR = 1.12, 95% CI = 1.03 to 1.21, respectively. MR analyses performed using existing published GWAS summary statistics yielded inconsistent results, including MR estimates that suggested education protected against myopia and that myopia reduced the risk of POAG. Re-analysis of a selection of published MR studies of myopia confirmed that most published results were invalid. CONCLUSIONS: Care is required when designing MR analyses. Many published MR studies of myopia have reported misleading results.