The strength of this study was that our sample was drawn from a large and unselected population, which consisted of ethnically homogenous Norwegian individuals. Because no information about ethnic backgrounds was available in our data file, the relatively few numbers of non-Caucasians subjects could not be excluded in the statistical analyses. Because the distribution of the two COMT alleles differs between different ethnic groups [23], the risk of bias caused by ethnical admixture could not be ruled out.
