(Computation of) the standardized mean difference in clinical trials should be standardized.

چکیده اصلی

The standardized mean difference (SMD) is widely used in psychology clinical trials to convert intervention effect estimates from raw units to a common scale, enabling comparison across settings. However, multiple operationalizations of the SMD are possible for clinical trial data, with large potential impacts on estimate magnitudes. In this commentary, I review practices for operationalizing the SMD in the Journal of Consulting and Clinical Psychology and other journals that publish results of clinical trials. I find that only one third of studies report sufficient information to determine the operationalization. Among those, the operationalization varies widely, threatening the SMD's utility in providing a common scale. Given the common goals and design elements across many trials, it is feasible to establish a default operationalization for the SMD's standardizer. I outline eight desirable SMD properties and evaluate each option against these properties. The pooled baseline standard deviation meets seven of eight, far more than any alternative, with its main limitation being susceptibility to inflation from restrictive inclusion criteria. I conclude that raw effect estimates should be divided by the pooled baseline standard deviation as a default, one-size-fits-most choice. Adopting this standard will help the SMD meet its promise as a standardized metric, improving comparability of results across different measures, outcomes, studies, and populations. I also provide reporting recommendations to improve transparency and enable recomputation using alternative standardizers when needed. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

متن کامل اصلی