Methodological problems in meta-analysis of association studies between bipolar affective disorders and the tyrosine hydroxylase gene.