OBJECTIVE The objective of this report is to review the research methods that have been used in the design, analysis, and reporting of Phase I dose-escalation studies of high-dose chemotherapy (HDCT) with bone marrow or stem cell support and to propose new guidelines for such studies that incorporate emerging principles of pharmacology, toxicity assessment, statistical design, and long-term follow-up. METHODS We performed a search of original, English-language, peer-reviewed full-length reports of HDCT (with or without radiotherapy) and unmanipulated hematopoietic precursor support (autologous bone marrow or stem cells or allogeneic bone marrow) in which one or more drug doses were escalated to identify dose-limiting toxicities needed for the design of subsequent Phase II trials. We reviewed the design, execution, analysis, and reporting of these trials to develop a coherent set of guidelines for the initiation of new HDCT regimens. The primary elements included in our analysis were the technique of dose escalation, the choice and application of toxicity grading scale, and the pharmacologic correlates of dose escalation. We also evaluated the methods employed to define dose-limiting toxicities and to select the maximum tolerated dose and the dose recommended for further study. We then examined whether subsequent Phase II trials based on these definitions corroborated the findings from the prior Phase I studies and summarized the findings from pharmacologic analyses that were reported from a subset of these investigations. RESULTS Thirty-five reports met the criteria for our literature review. Two standard methods of dose escalation (fixed increments or modified Fibonacci increments) were described in detail and were employed in the majority (30/35) of the studies. In 5 studies, the details of dose escalation were either not provided or not adequately referenced. There was marked heterogeneity among toxicity grading methods; scales used included the National Cancer Institute Common Toxicity Criteria (or similar scales such as the United States cooperative group or World Health Organization scales) as well as substantially modified versions of those instruments. Wide variations in the methods used to identify dose-limiting toxicities were observed. Statistical considerations, applied to the identification of the maximum tolerated or Phase II recommended dose, were similarly heterogeneous. Phase II trial designs varied from a simple expansion of the Phase I trial to separate, formally conducted studies. Nine Phase I trials featured pharmacologic analyses, and these ranged from simple pharmacokinetic evaluations to more complex analyses of the relationship between drug dose and the molecular targets of drug action. CONCLUSIONS Phase I clinical trials in the HDCT setting have been designed, analyzed, and reported using heterogeneous methods that limited their application to Phase II and II investigations. Moreover, correlative pharmacologic analyses have not been routinely undertaken during this critical Phase I stage. We propose guidelines for the design of new Phase I studies of HDCT based on 4 essential elements: (1) rational preclinical and clinical pharmacologic foundation for the regimen and for the agent selected for dose escalation; (2) incorporation of analytical pharmacology in the design and analysis of the regimen under investigation; (3) clear, prospective definitions of the dose- or exposure-limiting toxicities that can be distinguished from modality-dependent toxicities; selection of an appropriate toxicity grading scale, including an assessment of cumulative, delayed, and long-term effects of HDCT, particularly when designing tandem or repetitive cycle regimens; and (4) statistical input into the design, execution, analysis, interpretation, and reporting of these studies.