How to demonstrate similarity by using noninferiority and equivalence statistical testing in radiology research

摘要

Demonstrating similarity between compared groups-that is, equivalence or noninferiority of the outcome of one group to the outcome of another group-requires a different analytic approach than determining the difference between groups-that is, superiority of one group over another. Neither a statistically significant difference between groups (P < .05) nor a lack of significant difference (P>=.05) from conventional statistical tests provides answers about equivalenceoninferiority. Statistical testing of ecfuivalenceoninferiority generally uses a confidence interval, where equivalenceoninferiority is claimed when the confidence interval of the difference in outcome between compared groups is within a predetermined equivalenceoninferiority margin that represents a clinically or scientifically acceptable range of differences and is typically described by DELTA. The equivalenceoninferiority margin should be justified both clinically and statistically, considering the loss in the main outcome and the compensatory gain, and be chosen conservatively to avoid making a false claim of equivalenceoninferiority for an inferior outcome. Sample size estimation needs to be specified for equivalenceoninferiority design, considering A in addition to other general factors. The need for equivalence/ noninferiority research studies is expected to increase in radiology, and a good understanding of the fundamental principles of the methodology will be helpful for conducting as well as for interpreting such studies.