also published in 2004, a) included 22 Asian studies including 1,980 cases and 2,550 controls (we included 9,678 cases and 7,331 controls); b) performed genotypic analysis but not allelic analysis; c) provided no specific P values; d) reported negative results (under the model of lys-lys vs. lys-glu); and e) found a difference between possessing one and two ALDH2 504lys alleles that approached but did not reach statistical significance. Compared with each published study, our meta-analysis a) included the largest sample size from 45 English and seven Chinese publications (it was important to include Chinese-language publications). Our analysis showed that there was no publication bias, thus, the Chinses literatutre did not introduce overestiamets of effect sizes in our meta-analysis (supplementary Table 2); b) performed both allelic and genotypic analyses using the random effects model; c) used both strict and extended criteria as well as carrying out subgroup analyses to detect the effects; d) applied comprehensive analysis methods, as shown in the results, to study additional questions not answered in those previous meta-analyses; and e) found strong evidence of association with both alcohol dependence and the alcohol-induced medical diseases. In addition, the procedure of ‘extended-quality score’ suggested in our previous study (Li
