For clinical trials of anti-infective drugs for the treatment of vascular access device-related bloodstream infections, patients should be identified and enrolled on the basis of current standards for the clinical diagnosis of such infections. To ensure comparability of patients, only those infected with staphylococci and Candida species should be included. A prospective, randomized, double-blind design is recommended. Future protocols may include abbreviated courses of therapy, treatment with combinations of drugs, or a progression from parenteral to oral therapy. Clinical response is judged as cure, failure, or indeterminate response; there is no "improved" category. Microbiological response is categorized as eradication, persistence, or relapse and is of paramount importance. Several months of follow-up may be necessary for the detection of late relapses or metastatic infections. This guideline does not apply to studies of bacteremia or fungemia secondary to non-device-related, organ-based primary infections (e.g., pneumonia, urinary tract infection), which should be assessed in relation to the primary disorder.