Gravar-mail: Beyond total treatment effects in RCTs: why we need to measure outcomes at baseline when investigating mediation