We set out to evaluate the clinical efficacy of individual antibiotic agents for bone and joint infections in adults. Published and unpublished controlled trials reported between 1966 and 2000 were reviewed to determine if they involved random or quasi-random allocation to systemically administered antimicrobials or local antibiotic therapy for osteomyelitis and septic arthritis. Quiescence of infection after 1 year of follow-up was defined as the primary outcome measure. 22 trials containing 927 patients were eligible for final analysis. Varying proportions of the entire study population could be evaluated with respect to primary and secondary endpoints. Methodological quality was poor among most studies, and interpretability of results was further limited by small sample sizes, missing descriptions of patient populations and disease characteristics, and the frequent application of concomitant antibiotics. A trend towards improved, long-lasting infection control was observed in favour of a rifampicin-ciprofloxacin combination versus ciprofloxacin monotherapy for the treatment of staphylococcal infections related to orthopaedic devices (absolute risk difference [ARD] 28-9%; 95% CI -0.7 to 54.4%). Obviously unbalanced comparative studies showed some benefit of ticarcillin for bone infections caused by Pseudomonas species. No significant differences in therapeutic efficacy were found among trials comparing oral fluoroquinolones with intravenous beta-lactam drugs for both end-of-treatment (OR 0.8; 0.5 to 1.4) and long-term results (OR 1.3; 0.8 to 2.1). A variety of drugs was used as controls, thereby leading to inconsistent findings of drug-related side effects. Only one randomised trial was suitable to investigate the impact of polymethylmethacrylate gentamicin bead chains compared with parenteral antibiotics for skeletal infections, although this study was biased by patients receiving both combined local and systemic antibiotic therapy. Whereas intention-to-treat evaluation suggested a therapeutic advantage of systemic over local therapy, this trend diminished in the per-protocol analysis (1-year follow-up ARD -2.3;-17.5 to 10.8%). There exists little high-quality evidence on antibiotic therapy for osteomyelitis and septic arthritis. The observed heterogeneity among patient populations and medical and surgical treatment concepts preclude reliable inferences from the available data.