Determination and selection of the optimum number of sites and patients for clinical studies

Sample size estimation: an overview with applications to orthodontic clinical trial designs

Proper sample size estimation is an important part of clinical trial methodology and closely related to the precision and power of the trial's results. Trials with sufficient sample sizes are scientifically and ethically justified and more credible compared with trials with insufficient sizes. Planning clinical trials with inadequate sample sizes might be considered as a waste of time and resources, as well as unethical, since patients might be enrolled in a study in which the expected results will not be trusted and are unlikely to have an impact on clinical practice. Because of the low emphasis of sample size calculation in clinical trials in orthodontics, it is the objective of this article to introduce the orthodontic clinician to the importance and the general principles of sample size calculations for randomized controlled trials to serve as guidance for study designs and as a tool for quality assessment when reviewing published clinical trials in our specialty. Examples of calculations are shown for 2-arm parallel trials applicable to orthodontics. The working examples are analyzed, and the implications of design or inherent complexities in each category are discussed.

Methods for the statistical analysis of binary data in split-mouth designs with baseline measurements

Many split-mouth trials are characterized by the pairing of site-specific outcome and baseline data within each segment of a subject's mouth. However when the response variable of interest is binary, methods of statistical analysis for this design are not well developed. In this paper we present several analytic approaches that may be taken to the resulting data, showing how the efficiency of statistical inferences can be improved by appropriately incorporating the baseline information. The advantages and disadvantages of the different approaches are discussed in the context of an example from the published literature. The results from a limited simulation study are also presented.

Statistical issues in periodontal research

In the past 10 to 12 years, there have been several statistical issues identified in periodontal research which require and have generated non-standard or new statistical approaches. The purpose of this paper is to give an overview of these issues and approaches. Three general categories of issues are described: (i) statistical methods for detecting when disease progression occurs, and biological theories and corresponding statistical models which attempt to describe how the disease progresses; (ii) design issues in studies of therapeutic efficacy; and (iii) analytic issues arising from periodontal data analysis.