The distribution of the standardized mean difference is well understood. However, in many situations, researchers need to estimate an effect size to represent the relationship between a continuous outcome and a dichotomous grouping variable, adjusting for the effect of a covariate (or a set of covariates). Typically, this adjustment takes place via regression models. In this article, we consider five different estimators of standardized mean differences that could arise from regression models with one or more covariates. We demonstrate that an existing correction, believed to recover the pooled standard deviation, is in fact an approximation. In addition, we compared the performance of each standardized mean difference index. The function used to generate the data for the simulation is available in the Supplemental Material. Implications for comparing results from primary studies, as well as for meta-analysis, are also considered.