We read the article in the June 2011 issue on the comparison of oral impacts (Wu A, McGrath C, Wong RWK, Wiechmann D, Rabie ABM. Comparison of oral impacts experienced by patients treated with labial or customized lingual fixed orthodontic appliances. Am J Orthod Dentofacial Orthop 2011;139;784-90), and we became interested in the statistical parameters.
The authors are to be commended for their aim to conduct an age- and sex-matched prospective longitudinal trial on the oral impacts of 2 orthodontic appliances. Even if the statistical tests used are correct, there seems to be a statistical study design issue. Our main concern is that it does not seem that the authors performed any pre-hoc sample size calculation before enrolling patients and undertaking this experimental study. In a prospective study, if statistical tests are used, their power should be determined a priori.
For a particular experimental result to be claimed as significant or not significant, sufficient statistical power must be obtained, and the sample size for each group should be calculated a priori.
Did the authors perform a pre-hoc sample size calculation? Since we don’t know whether this study was sufficiently powered, we don’t know whether these findings are statistically correct and sufficiently powered. For example, the authors stated that there were significant differences in pain experiences attributed to a change in tongue position and a reduction in tongue space. We don’t know whether these “significant differences” are really so.
Although this study is innovative, interesting, and important, an accurate experimental design is the only way to draw adequate conclusions about experimental findings, transforming good hypotheses into statistically supported and evidence-based scientific conclusions. We hope our suggestions will be useful to other authors who will be involved in similar studies in the future.