There are several limitations of this study. First, as in any meta-analysis, this study is based on heterogeneous samples with differential assessment of measures. In addition there was statistical evidence of heterogeneity. Heterogeneity likely contributes to a lack of precision in this analysis. However, because misclassification is independent of genotype, any bias under the alternative hypothesis would decrease the power of the study (leading to an underestimate of the interaction effect).38 Since we had adequate power to detect an interaction on heavy vs light smoking between rs16969968 and age at onset of regular smoking, improved homogeneity of the studies should only strengthen this result.
