The Effects of a Multi-component Lower Extremity Training Technique on Physical Function in Healthy Older Adults: A Randomized Controlled Trial

To the Editor,

First, we determined the calculated sample size (n = 14 in each group) based on the effect size = 1.3, power = 0.95, and α-error = .05. The sample size was consistent with the one calculated by G*Power software using a one-sided test (“one tail”). For one-sided tests in clinical trials, it is preferable to adopt the approach of setting type I errors at half the conventional type I error used in two-sided tests; this is recommended by the statistical principles for clinical trials by an international guideline (Lewis, 1999). Accordingly, we adjusted the significance level to <.025. We did not mention the percent of dropouts; however, the CONSORT 2010 checklist does not request for dropout percent information. Next, the representative data needs to be selected appropriately depending on the data distribution. For example, if the data follows normal distribution, the mean value is used; otherwise, the median value is used. In this study, distributions where the normality could not be assessed were observed for some variables, whereas normal distributions were observed for several variables. Therefore, to enhance readability, the mean and standard deviation values were used consistently. We believe that it is not customary to describe different representative values in a table for each distributor characteristic. We performed a two-way analysis of variance (ANOVA) with repeated measures to evaluate the interaction (group × time). ANOVA, which is derived under the assumptions of normality and homogeneity of variance, is extremely robust even under the violation of these assumptions (Ito, 1980; Kirk, 1995). Finally, we mentioned about the dropouts under the follow-up section twice to emphasize that the dropout occurrence times were different. The first mention was during the follow-up and the second was at the time of remeasurement.