To determine the relative efficacy and cost-effectiveness of five of the most commonly used antimicrobial preparations for treating mild to moderate facial acne in the community; the propensity of each regimen to give rise to local and systemic adverse events; whether pre-existing bacterial resistance to the prescribed antibiotic resulted in reduced efficacy; and whether some antimicrobial regimens were less likely to give rise to resistant propionibacterial strains. This was a parallel group randomised assessor-blind controlled clinical trial. It was a pragmatic design with intention-to-treat analysis. All treatments were given for 18 weeks, after a 4-week treatment free period. Outcomes were measured at 0, 6, 12 and 18 weeks. Primary care practices and colleges in and around Nottingham and Leeds, and one practice in Stockton-on-Tees, England. Participants were 649 people aged 12--39 years, all with mild to moderate inflammatory acne of the face. Study participants were randomised into one of five groups: 500 mg oral oxytetracycline (non-proprietary) twice daily (b.d.) + topical vehicle control b.d.; 100 mg oral Minocin MR (minocycline) once daily (o.d.) + topical vehicle control b.d.; topical Benzamycin (3% erythromycin + 5% benzoyl peroxide) b.d. + oral placebo o.d.; topical Stiemycin (2% erythromycin) o.d. + topical Panoxyl Aquagel (5% benzoyl peroxide) o.d. + oral placebo o.d., and topical Panoxyl Aquagel (5% benzoyl peroxide) b.d. + oral placebo o.d. (the active comparator group). The two primary outcome measures were: (1) the proportion of patients with at least moderate self-assessed improvement as recorded on a six-point Likert scale, and (2) change in inflamed lesion count (red spots). The best response rates were seen with two of the topical regimens (erythromycin plus benzoyl peroxide administered separately o.d. or in a combined proprietary formulation b.d.), compared with benzoyl peroxide alone, oxytetracycline (500 mg b.d.) and minocycline (100 mg o.d.), although differences were small. The percentage of participants with at least moderate improvement was 53.8% for minocycline (the least effective) and 66.1% for the combined erythromycin/benzoyl peroxide formulation (the most effective); the adjusted odds ratio for these two treatments was 1.74 [95% confidence interval (CI) 1.04 to 2.90]. Similar efficacy rankings were obtained using lesion counts, acne severity scores and global rating by assessor. Benzoyl peroxide was the most cost-effective and minocycline the least cost-effective regimen (ratio of means 12.3; difference in means -0.051 units/GBP, 95% CI -0.063 to -0.039). The efficacy of oxytetracycline was similar to that of minocycline, but at approximately one-seventh of the cost. For all regimens, the largest reductions in acne severity were recorded in the first 6 weeks. Reductions in disability scores using the Dermatology Quality of Life Scales were largest for both topical erythromycin-containing regimens and minocycline. The two topical erythromycin-containing regimens produced the largest reductions in the prevalence and population density of cutaneous propionibacteria, including antibiotic-resistant variants, and these were equally effective in participants with and without erythromycin-resistant propionibacteria. The clinical efficacy of both tetracyclines was compromised in participants colonised by tetracycline-resistant propionibacteria. None of the regimens promoted an overall increase in the prevalence of antibiotic-resistant strains. Systemic adverse events were more common with the two oral antibiotics. Local irritation was more common with the topical treatments, particularly benzoyl peroxide. Residual acne was present in most participants (95%) at the end of the study. The response of mild to moderate inflammatory acne to antimicrobial treatment in the community is not optimal. Only around half to two-thirds of trial participants reported at least a moderate improvement over an 18-week study period; extending treatment beyond 12 weeks increased overall benefit slightly. Around one-quarter dropped out when using such treatments, and 55% sought further treatment after 18 weeks. Topical antimicrobial therapies performed at least as well as oral antibiotics in terms of clinical efficacy. Benzoyl peroxide was the most cost-effective and minocycline the least cost-effective therapy for facial acne. The efficacy of all three topical regimens was not compromised by pre-existing propionibacterial resistance. Benzoyl peroxide was associated with a greater frequency and severity of local irritant reactions. It is suggested that the use of a combination of topical benzoyl peroxide and erythromycin gives less irritation and better quality of life. There was little difference between erythromycin plus benzoyl peroxide administered separately and the combined proprietary formulation in terms of efficacy or local irritation, except that the former was nearly three times more cost-effective. The data on cost-effectiveness, and outcomes in patients with resistant propionibacterial floras, did not support the first line use of minocycline for mild to moderate inflammatory acne of the face. Three priority areas for clinical research in acne are: defining end-points in acne trials (i.e. what is a satisfactory outcome?); developing and validating better patient-based measures for assessing treatment effects on facial and truncal acne; and exploring patient characteristics that may modify treatment effects (efficacy and tolerability).