The randomized, controlled trial is the benchmark for the evaluation of new drug therapies. Random allocation is a powerful means of controlling for the potential effects of confounders and serves to minimize bias (systematic deviation from the truth) on the part of physicians and patients. Concealing the treatment allocation from the patient and investigator (blinding), which reduces the potential for bias, is also a fundamental component of a rigorously designed clinical trial. Considerable controversy has arisen regarding the use of placebo controls in evaluating new drugs. Some authors have argued against their use on the basis that patients are denied treatment of proven efficacy and thus may experience some degree of morbidity by participation in a placebo-controlled study. However, a placebo comparison allows an evaluation of the new therapy against the alternative of no treatment, and thus is scientifically valid and ethical when the standard therapy has only modest efficacy or causes important adverse effects. In Phase III trials it is vital that investigators choose as a primary measure of response an outcome that is clinically meaningful. In the past there has been an over-reliance on surrogate markers of efficacy such as improvements in laboratory tests. In some instances these have been shown not to correlate with clinically meaningful events. Investigators should consider utilization of quality of life measures as measures of response in addition to the more conventional outcomes of death, occurrence of disease-related complications and clinical activity indices. Once an appropriate outcome has been identified, the planning of a clinical trial requires input from a biostatistician. Careful consideration is given to the number of patients required, which is dependent upon the alpha (false positive) and beta (false negative) error rates selected by the investigator, the size of treatment effect that is considered to be clinically meaningful and the estimated rate of occurrence of the outcome of interest in the placebo (or standard therapy) group. If interim analyses are planned, these must be defined prior to initiation of the study, and appropriate statistical techniques employed to account for the increase in the alpha error rate that results from the use of multiple statistical testing procedures.