Delivery of cancer chemotherapy is often limited by myelotoxicity, primarily neutropenia. As part of an effort to create a model to predict the risk of chemotherapy-induced neutropenia, we reviewed the reports of randomized clinical trials with more than 50 patients per arm in early-stage breast cancer (ESBC) and non-Hodgkin's lymphoma (NHL) published between 1990 and 2000. We observed that no hematologic toxicity data were reported in 39% and 34% of the ESBC and NHL trials, respectively. The remaining trials reported on hematologic toxicity in 16 different ways. When reported, rates of neutropenia, leukopenia, and hematotoxicity varied widely with the same and similar chemotherapy regimens. Dose-intensity data were not reported in 39% and 54% of ESBC and NHL trials, respectively. The majority of the remaining studies reported incomplete dose-intensity data such as percentages of patients completing all cycles or receiving a given percentage of planned dose intensity. Only 28% reported the mean or median relative dose intensity received by patients. Based on this review, we conclude that current practices for reporting chemotherapy treatments are inadequate for describing the risk of chemotherapy to patients or for quantitatively assessing the risk of treatment alternatives. We recommend that standard procedures for documenting and reporting hematologic toxicity and dose intensity in cancer chemotherapy trials be required for publication of chemotherapy trials.