Challenges of Non–Intention-to-Treat Analyses
Randomization is essential for objectively establishing causal associations between new interventions and outcomes. However, using randomization, a causal link with outcomes can be established only between assigned interventions rather than received interventions, and only then if there is complete follow-up and the ITT principle is applied. Most simplistic alternatives to ITT analysis reintroduce the confounding that randomization was intended to eliminate. It is unrealistic to imagine that simple statistical maneuvers can transform the trial that was conducted into the trial that ideally could have been conducted. To the greatest extent possible, all outcomes should be ascertained for all randomized participants, the primary analysis should be conducted according to the ITT principle, and the results be should be interpreted using sound informed scientific judgement—for which there is no statistical substitute.
Challenges of Non–Intention-to-Treat Ana
Microsoft Word Dokument