In a randomised double-blind trial conducted between 1990 and 1994, 616 patients from 43 centres, pefloxacin (group P, 316 patients) and a cefazolin-oxacillin combination (group C, 300 patients) were compared in the prophylaxis of bone infection after grade 1 and 2 open leg fractures. Samples were obtained at emergency, before and during surgery, and from drain aspirates. Antimicrobial susceptibility, slime production and adherence properties of the bacteria were tested. Cultures at emergency and before surgery showed similar distributions of gram-positive and gram-negative bacteria in both groups, while wound closure and infecting isolates showed prevailing gram-positive bacteria in group P and gram-negative bacteria in group C. Positive cultures at each stage were correlated with the occurrence of infection but were not predictive of the infecting species, which were nosocomial bacteria in most cases. Positive cultures at wound closure warn of a higher infection risk. Twenty-one of 316 (6.6%) patients in group P and 24 of 300 (8%) in group C were considered infected within 3 months. The difference is not significant (chi-square test = 0.42; P = 0.51). Infecting strains were isolated from 38 patients (group P, 18; group C, 20). Infecting species, although not predictable, appear to be those escaping the spectrum of the prescribed antimicrobial prophylaxis.