Pooled ORs were calculated using a random effects (meta-analytic) model according to the DerSimonian–Laird method.27 This model automatically assigns the weighting that each individual study contributes to the pooled OR depending on that study’s sample size and yielded OR. Forest plots were used to depict variant allele (G) and genotype (G/G) ORs for each individual study as well as overall OR. The degree of heterogeneity, that is the variation that cannot be explained by random chance,28 in the ORs across studies was assessed using the I-squared statistic and a χ2-test of goodness of fit.29 When high levels of heterogeneity were observed (I-squared values > 40%), meta-regression models were fitted to the data in an attempt to explain the heterogeneity.30 The following explanatory variables were included in the meta-regression models: 50% or more participants Caucasian (yes/no); 50% or more participants Asian (yes/no); mean age (continuous variable); percentage of males (continuous variable); type of dependence (opioid/heroin); and majority ethnicity (Caucasian/Asian). The potential for publication bias in the meta-analysis was assessed statistically using Egger’s test31 and visually using Begg’s funnel plot.32 The funnel
