Investigating treatment mediators when simple random assignment to a control group is not possible

As the study of psychological treatments begins to mature beyond questions of efficacy, investigators are focusing more on treatment mediators in an effort to aid revision and dissemination of existing treatments. Often, such studies are conducted without wait-list or no-treatment control groups. However, designs without a control group constrain the examination of treatment mediators and preclude statements about the overall effects of treatment. We review the limitations of treatment-treatment and single-treatment designs for studying mediation and suggest design modifications to strengthen future investigations. Specifically, we present the benefits and limitations of regression discontinuity and interrupted time series in testing mediation when ethical or organizational considerations preclude simple random assignment to a no-treatment condition.