Background Aldehyde dehydrogenase 2 ( ALDH2 ; rs671, Glu504Lys) and alcohol dehydrogenase 1B ( ADH1B ; rs1229984, His47Arg) polymorphisms have a strong impact on carcinogenic acetaldehyde accumulation after alcohol drinking. To date, however, evidence for a significant ALDH2 –alcohol drinking interaction and a mediation effect of ALDH2 / ADH1B through alcohol drinking on gastric cancer have remained unclear. We conducted two case–control studies to validate the interaction and to estimate the mediation effect on gastric cancer. Methods We calculated odds ratios (OR) and 95% confidence intervals (CI) for ALDH2 / ADH1B genotypes and alcohol drinking using conditional logistic regression models after adjustment for potential confounding in the HERPACC-2 (697 cases and 1372 controls) and HERPACC-3 studies (678 cases and 678 controls). We also conducted a mediation analysis of the combination of the two studies to assess whether the effects of these polymorphisms operated through alcohol drinking or through other pathways. Results ALDH2 Lys alleles had a higher risk with increased alcohol consumption compared with ALDH2 Glu/Glu (OR for heavy drinking, 3.57; 95% CI 2.04–6.27; P for trend = 0.007), indicating a significant ALDH2 –alcohol drinking interaction ( P interaction  = 0.024). The mediation analysis indicated a significant positive direct effect (OR 1.67; 95% CI 1.38–2.03) and a protective indirect effect (OR 0.84; 95% CI 0.76–0.92) of the ALDH2 Lys alleles with the ALDH2 –alcohol drinking interaction. No significant association of ADH1B with gastric cancer was observed. Conclusion The observed ALDH2 –alcohol drinking interaction and the direct effect of ALDH2 Lys alleles may suggest the involvement of acetaldehyde in the development of gastric cancer.