Existing evidence remains inconclusive as to how the association between inactive ALDH2 and esophageal cancer (EC) depends on alcohol consumption. The study is based on the China Kadoorie Biobank cohort, with ten years follow-up of 0.5 million adults aged 30-79 years. ALDH2 activity was assessed by both self-reported flushing response and Glu504Lys (rs671 G>A) polymorphism. Among both male and female participants who consumed alcohol less than weekly (n=69,519; 211 EC cases), low active or inactive ALDH2 was not associated with increased EC risk [HRs (95% CIs): GA versus GG 0.75 (0.54, 1.04); AA versus GG 1.01 (0.46, 2.20)]. Among male weekly alcohol consumers, both flushing response [n=59,380; 501 EC cases; HRs (95% CIs): 'soon after drinking' versus 'no' flushing response 1.45 (1.05, 2.01)] and rs671 [n=10,692; 94 EC cases; GA versus GG 3.31 (1.94, 5.67)] were associated with EC risk. The increased EC risk associated with 'soon' response or rs671 GA was apparent in men consuming alcohol ≥30g/d. Among male daily consumers, the HRs (95% CIs) for EC associated with 15g/d of alcohol were 1.28 (1.15, 1.44) for 'soon' response [versus other responses: 1.12 (1.09, 1.15); P_interaction =0.047; n=36,401, 425 EC cases] and 1.41 (1.08, 1.82) for rs671 GA [versus GG: 1.16 (1.06, 1.27); P_interaction =0.493; n=6,607, 80 EC cases]. Self-reported flushing response had low sensitivity (56.8%) and high specificity (88.4%) in identifying rs671 A allele among male weekly alcohol consumers. In conclusion, low-activity ALDH2 was associated with increased EC risk among male heavy alcohol consumers. More accurate measurement of alcohol-related EC risk allows better achievement of precision prevention. This article is protected by copyright. All rights reserved.