Cancer clinical trials designed to test the value of perioperative chemotherapy present special statistical problems for the investigator. Because chemotherapy must begin within hours of surgery, aspects of the design, conduct, and analysis of such trials are more complex than for many other types of clinical investigations. Special attention must be given to the logistics of patient entry. In some situations, patients may receive a treatment assignment prior to the determination of eligibility or even prior to histological diagnosis of cancer. Design problems may arise when important prognostic factors are not known at the time of randomization. A comparison of perioperative chemotherapy to other, conventionally timed adjuvant chemotherapy requires that adjuvant treatment be purposely delayed for some patients. Because patients enter a perioperative chemotherapy trial before or soon after surgery, the patient population is not the same as that treated in standard adjuvant trials. Thus it is not possible to compare results directly between perioperative and standard adjuvant trials. Analyses of toxicity will include surgical morbidity which would not be seen in conventional adjuvant trials. The analysis of perioperative chemotherapy trials requires care, as many case exclusion factors may make it difficult to determine the true effectiveness of the perioperative treatment plan. Analysis by intent to treat is recommended. These and other statistical and logistic aspects are discussed with reference to the current trial being conducted by the Ludwig Breast Cancer Study Group.