Placebo control groups in randomized treatment trials: a statistician’s perspective

Because a statistical tie between standard treatment and an innovation is uninterpretable, most trials intended to demonstrate efficacy of innovations in psychopharmacology employ a placebo control group, despite the existence of standard medications for many disorders. In this review I consider the statistical issues that inform the ethics of the decision to use a placebo condition and make the following points: 1) the investigator is relying on the assumption that the effects of delayed standard treatment are neither long-lasting nor harmful; 2) the usual practice of truncating follow-up when a patient ceases to adhere to a study treatment makes it difficult to empirically test that assumption; 3) placebo control trials often suffer from methodological weaknesses (including nonrandom truncation) that reduce their inferential power; 4) these subtleties place a substantial burden on the informed consent process; 5) alternative designs are available but not well explored, due to the dominant role of “regulatory” trial methodology; and 6) researchers should consider other goals besides helping to introduce another treatment that improves on placebos but not the standard treatment.