Pre-clinical studies suggest that the simultaneous blockade of the α1b and 5HT2A receptors may be effective in reducing alcohol consumption. This study aimed to assess the efficacy and safety of prazosin (α1b blocker) and cyproheptadine (5HT2A blocker) combination in decreasing total alcohol consumption (TAC) in alcohol use disorder (AUD). This was a double-blind, parallel group, placebo-controlled, Phase 2, randomized clinical trial conducted in 32 addiction treatment centres in France. A total of 108 men and 46 women with severe AUD took part. Participants were randomly assigned to one of the following 3-month treatments: (1) low-dose group (LDG) receiving 8mg cyproheptadine and 5mg prazosin extended-release (ER) formulation daily; (2) high-dose group (HDG) receiving 12 mg cyproheptadine and 10mg prazosin ER daily; and (3) placebo group (PG) receiving placebo of cyproheptadine and prazosin ER. A total of 154 patients were randomized: 54 in the PG, 54 in the LDG and 46 in the HDG. The primary outcome was TAC change from baseline to month 3. A significant main treatment effect in the change in TAC was found in the intent-to-treat population (P = 0.039). The HDG and LDG showed a benefit in the change in TAC from baseline to month 3 compared with PG: -23.6g/day, P = 0.016, Cohen's d = -0.44; -18.4g/day, P = 0.048 (Bonferroni correction P < 0.025), Cohen's d = -0.36. In a subgroup of very high-risk drinking-level participants (> 100 g/day of pure alcohol for men and > 60 g/day for women), the difference between the HDG and the PG in the primary outcome was -29.8g/day (P = 0.031, Cohen's d = -0.51). The high and low doses were well-tolerated with a similar safety profile. A randomized controlled trial of treatment of severe alcohol use disorder with a cyproheptadine-prazosin combination for 3months reduced drinking by more than 23 g per day compared with placebo. A higher dose combination was associated with a larger magnitude of drinking reduction than a lower dose combination while showing similar safety profile.