The article by Doshi-Meta and Bhad-Patil addresses an important topic in orthodontics (Doshi-Meta G, Bhad-Patil WA. Efficacy of low-intensity laser therapy in reducing treatment time and orthodontic pain: a clinical investigation. Am J Othod Dentofacial Orthop 2012:141:289-97). If low-intensity laser therapy can safely speed the rate of tooth movements, and with the added benefit of less pain, that achievement would be greatly appreciated. Research along this line is clearly necessary. However, this report contains the following problems.
Inconsistencies concerning the study’s end point. The authors introduced the problem by telling readers that, “Generally, the time required for fixed appliance treatment is 20 to 30 months. Reducing the treatment time requires increasing the rate of orthodontic tooth movement.” We then read in Material and Methods that the treatments continued “until complete canine retraction on the experimental side.” This statement about complete retraction is repeated a number of times. But in the Results section, Table I, we see that T2 (the end point of measurements) was “at 4.5 months.” This 4.5-month time period is obviously not at “complete canine retraction on the experimental side” because in Table I we see that, on the experimental side, mean distances between the canines and the first molars were still 14.32 to 16.02 mm. So, the claimed end point of “complete canine retraction on the experimental side” was clearly not reached.
Questionable statistical reporting.
Concerning sample size, the authors wrote, “The sample size was determined by power analysis based on the results of the pilot study that showed that the rate of tooth movement was twice that of the control side.” But the authors’ formula for determining the sample size for the study used “population proportions” and “sample proportions.” What do these unspecified population and sample proportions have to do with “rate of tooth movement”?
A secondary outcome measure compared pain scores on the experimental and control sides at 3 times. First, as seen in Figure 2, the scoring system used was the “smiley faces” scheme, which is an ordinal measure. Therefore, labeling this Figure as a “visual analog pain scale” is technically incorrect. Analog means continuous, not ordinal. Second, since the statistical analysis compared repeated ordinal measures, the authors properly used a nonparametric test for repeated measures, the Wilcoxon signed rank test. But, in Table IV, the authors give Z values and the associated P values. However, Z scores are not determined by the nonparametric Wilcoxon signed rank test. It determines whether there is a significant difference between medians, not between means. It is therefore not clear from Table IV just what is being reported.
Lastly, it is incorrect to classify significant differences as either “highly significant” or just “significant.” This distinction is made a number of times, implying that certain significant differences are more significant than others. This is not accepted statistical practice and is not clinically applicable.