I read with interest the editorial1 in the August 2009 issue of Archives that tells us the need for effect size estimation in randomized controlled trials (RCTs). In addition to determining primary end points, estimating sample size, and identifying α and β levels that indicate the probability of type I or II error, prespecifying effect size is required for designing RCTs.2 As discussed in the article, Cohen d, defined as standardized mean difference and scaled and classified as small (d = 0.2), medium (d = 0.5), or large (d = 0.8), is commonly used as an effect size for continuous data following a normal distribution. However, an effect size of 0.2 could be considered large in some contexts, such as vaccination to prevent polio in the general pediatric population, and an effect size of 0.8 small in others.3 Consequently, we are faced with the crucial question of how large an effect size has to be to be clinically important. Although the article does not solve this problem, it is suggested that clinical significance of effect size is dependent on factors such as the severity of the condition and the convenience and cost of the intervention.
Fujita T. Clinical Importance of Effect Size in Randomized Controlled Trials. Arch Surg. 2010;145(4):400-401. doi:10.1001/archsurg.2010.47