Bias Control – A Closer Look at Blinding and Randomization

Abstract

There are many uncontrolled factors in clinical studies and, in order to make sound treatment assessments, researchers need to impede their influence. The two pillars of the clinical trial that deter bias are blinding and randomization. Blinding keeps the subjects and researchers honest by concealing the mediation a subject takes and randomization is designed to produce equivalent treatment groups. But blinding isn't fool proof --- treatment identification may be exposed to researchers and/or subjects because of the unique action (e.g. dry mouth) of a drug and there are also many examples of deliberate attempts to un-blind clinical studies. Without randomization researchers would decide who gets a new treatment and who gets a control treatment t (e.g. placebo) and those choices would almost assuredly biased a study's results. Allowing treatment assignments to be made by a random process is a major way to overcome that possibility, but, there is no guarantee that randomization has succeeded in creating unbiased treatment groups.