Summary Background Cellulitis (erysipelas) of the leg is a common, painful infection of the skin and underlying tissue. Repeat episodes are frequent, cause significant morbidity and result in high health service costs. Objectives To assess whether prophylactic antibiotics prescribed after an episode of cellulitis of the leg can prevent further episodes. Methods Double-blind, randomized controlled trial including patients recently treated for an episode of leg cellulitis. Recruitment took place in 20 hospitals. Randomization was by computer-generated code, and treatments allocated by post from a central pharmacy. Participants were enrolled for a maximum of 3 years and received their randomized treatment for the first 6 months of this period. Results Participants (n = 123) were randomized (31% of target due to slow recruitment). The majority (79%) had suffered one episode of cellulitis on entry into the study. The primary outcome of time to recurrence of cellulitis included all randomized participants and was blinded to treatment allocation. The hazard ratio (HR) showed that treatment with penicillin reduced the risk of recurrence by 47% [HR 0·53, 95% confidence interval (CI) 0·26–1·07, P = 0·08]. In the penicillin V group 12/60 (20%) had a repeat episode compared with 21/63 (33%) in the placebo group. This equates to a number needed to treat (NNT) of eight participants in order to prevent one repeat episode of cellulitis [95% CI NNT(harm) 48 to ∞ to NNT(benefit) 3]. We found no difference between the two groups in the number of participants with oedema, ulceration or related adverse events. Conclusions Although this trial was limited by slow recruitment, and the result failed to achieve statistical significance, it provides the best evidence available to date for the prevention of recurrence of this debilitating condition.