The randomized clinical trial: bias in analysis.

The realization that bias in patient selection may influence the results of clinical studies has helped to establish the randomized controlled clinical trial in medical research. However, bias can be equally important at other stages of a trial, especially at the time of analysis. Withdrawing patients from consideration in the analysis because of ineligibility on account of study entry criteria, lack of compliance to the protocol, or data of poor quality may be a source of systematic error. Examples to illustrate the possible consequences are taken from trials in the cardiovascular field. We recommended that reported study results should include outcome data from all subjects randomized in the group to which they were originally assigned.