Statistical efficiencies that the FDA should encourage

The Food and Drug Administration (FDA) has a dual responsibility to the public of protecting against unsafe or ineffective drugs while efficiently licensing new drugs that are safe and effective. Balancing these two responsibilities requires correct safeguards that are also efficient in minimizing time, cost, and patient exposure to ineffective or unsafe experimental conditions. Reliance on statistical probabilities that do not relate to the question being asked, conservatively biased tests of significance, statistically inconsequential interim-analysis procedures, and largely superfluous active control groups in three-arm designs are discussed as problems. In each of these cases, design and analysis methods that are both correct and more efficient are available. Their use can reduce sample-size requirements and enhance power for documenting meaningful treatment effects.