Comparing prescribed and achieved treatment outcomes in digitally planned orthodontic treatment: statistical approaches

Abstract

Research on digital treatment planning has grown exponentially over the last decade. A question of interest is how closely do the prescribed and achieved tooth movement outcomes agree? Common approaches applied in attempt to answer this question include the use of the correlation coefficient, various accuracy formulas that calculate the percentage of the achieved treatment outcome compared to the prescribed one, or the classical paired t-test comparing the prescribed and achieved outcomes and claiming agreement in case of statistical non-significance. In this article, we explain why such approaches are problematic and should not be used. We describe alternative, more suitable methods based on the concept of clinical equivalence. Clinical equivalence means that differences between prescribed and achieved outcomes that fall in a small, pre-defined range are considered as clinically irrelevant. Appropriate methods include the use of the confidence interval for the mean difference, the two one-sided tests approach (TOST), and statistical modelling. The advantages and disadvantages of the methods are discussed, and applicability and recommendations are provided. Finally, sample size calculation for studies aimed at evaluation of the agreement between the prescribed and achieved tooth movement outcomes is discussed. Regardless of the chosen analytical approach, the selection of the equivalence threshold is paramount to the correct and clinically relevant interpretation of the study results.