Intention-to-treat analysis in randomized trials: who gets counted?

This article discusses the rationale and implications associated with the selection and use of analysis strategies for randomized clinical trials as they relate to protocol deviations. The topics addressed specifically are the conceptual and methodologic approaches and biases of clinical efficacy and effectiveness assessment. Examples are provided that highlight the consequences of different analytic strategies, particularly regarding intention-to-treat analysis. Favored by statisticians intention-to-treat analysis seeks to answer the question, "Is it better to adopt a policy of treatment A if possible, with deviations if necessary, or a policy of treatment B if possible, with deviations if necessary?" This is a relevant question, sometimes more relevant than "Is treatment A better than treatment B?" The authors suggest that different analytic strategies may be more or less appropriate depending on the intended audience.