supplementary Table 7 show the 504lys allele frequencies of by location. Most alcoholics with inactive ALDH2 alleles were heterozygous; homozygous lys-lys was rarely found in the alcoholic subjects, although it was often observed in controls. As expected, the recessive model based on the lys-lys genotype could not be applied in the present analysis due to insufficient power. Regarding the discrepant results in the individual association studies included in this meta-analysis, there are some possible explanations besides ethnic heterogeneity (e.g., one study (Chen et al. 1997) reported different direction of odd ratio in Ami): first, in most studies, subjects with AD or alcohol abuse were diagnosed according to the ICD (World Health Organization) or DSM (American Psychiatric Association) system. ICD-10 criteria for alcohol and drug dependence are more stringent than DSM-IV criteria, which in turn are more stringent than DSM-III-R (Schuckit et al. 1994). Two studies (Yan et al. 2003; Yu et al. 2002) applied local criteria and six studies had no specified criteria. Since these classification criteria differ, certain subjects would be eligible for inclusion as “affected” in some studies but not in others. Less stringent diagnostic criteria could differentiate the patient and control groups to a smaller extent, potentially
