The objective of this study was to describe the variation in preparation and administration of cyclophosphamide, mesna, and hydration for the treatment of childhood malignancies within clinical trial protocol documents. All cyclophosphamide-containing cooperative group (Pediatric Oncology Group) protocols that were open at Dana-Farber Cancer Institute in April 1998 were evaluated. Among the 14 active protocols, there were 23 unique cyclophosphamide regimens. Marked variation existed in infusion rate, fluid type, and volume used for admixing cyclophosphamide and mesna, as defined in the "Treatment" section of the protocols that we evaluated. Further variation was found in the type, amount, and rate of infusion of prehydration and posthydration fluid. Internal inconsistency existed within the protocols pertaining to the administration methods described in the "Agent Information," "Treatment," and "Consent" sections of the written documents. Clinical trial protocol documents serve as reference material for health care providers who prescribe, dispense, and administer protocol chemotherapy. Misinterpretation of protocol documents and clinician orders are contributing factors in serious and deadly medication errors. Internal inconsistency within protocol documents and variation in drug administration across protocols is a potential source of error. We recommend improved accuracy, clarity, and internal consistency of protocol documents to improve patient safety and compliance with protocol specifications. In addition, the use of standard concentrations, volumes, and methods of administration of chemotherapeutic agents and accompanying fluids is recommended.