In general, there is an association between methodological trial quality and estimation of effect, in that the stronger the quality of a trial, the more accurate the estimate of effect. Methodological quality can be measured by using various tools (24–26). In the case of randomized trials, these generally take account of factors such as the adequacy of randomization and concealment of allocation, blinding, presence or absence of differences between groups in potential confounders, intention to teach, and the validity and reliability of outcome measures. Thus, the presence of bias or systematic error, as indicated by weak randomization methods or by inadequate blinding, should result in an overestimation of effect. For example, the number needed to teach (NNT), which is akin to the number needed to treat in evidence-based medicine (27), represents one practical measure of effect when outcomes can be dichotomized, and it is calculated as the reciprocal of the absolute risk difference. The systematic review on stress-reduction programs for medical students by Shiralkar et al. (4) used NNT as an outcome measure. Although the association between effect size and outcome is open to empirical evaluation in any subject area, the NNT would expectedly become lower in association with a weaker study design. Thus, the validity of a study design should be taken into account in determining the importance of results.