Many clinical trials of antimicrobials to prevent surgical infection are poorly designed and may, therefore, indicate policies that are misleading. This paper provides guidelines for trial design including statistical planning and analysis. Satisfactory trials have been carried out in most branches of surgery, but there is still a need to define patients at special risk of infection and to design trials within these groups. There is also a need to perform trials to validate current practice and to establish the cheapest, least toxic, and most effective antimicrobial regimens. Well-designed trials require little analysis, but the results of a poorly planned trial cannot be retrieved even by statistical methods.