Comments on "Effects of Cognitive Versus Mind-Motor Training on Cognition and Functional Skills in the Community-Dwelling Older Adults".

Sharma et al. examined the effects of cognitive versus mind-motor training on cognition and functioning in community-dwelling older adults. Their sample size estimation, as described in the article, is problematic. In two-group randomized controlled trials (RCTs), the sample size for a continuous outcome measure is estimated either by seeking to determine whether one group is superior to the other by a certain value for the mean and standard deviation (SD) or by seeking to determine whether one group is superior to the other by a certain effect size. In the context, Sharma et al. acknowledged that their objective was to compare outcomes between two groups and analyzed their data accordingly; however, they estimated sample size based not on their objective and plan of analysis but on the change in magnitude of the outcome variable in a single arm of the RCT. Then they doubled this value, with compensation for expected dropouts, to obtain the necessary sample size for the full RCT. This is incorrect.

Further, although they powered the study to detect a one-point change in either of the two outcome measures, they did not state what the SD might be for this change in score. The sample size for the mean difference cannot be estimated without knowing the SD. Additionally, the SD is unlikely to be identical for two different outcome measures. Thus, how they arrived at a value of n = 20 per group is unclear. In this context, it is worth asking whether a one-point change, the least possible change in the two outcome measures, is truly clinically significant. It is worth noting that when the detection of such a small change is set as a target, the necessary sample size is expected to be large and not n = 20 (because small samples are powered to detect large changes, and vice versa).