The good, the bad and the ugly in randomised controlled trials

For millennia people have conducted rudimentary clinical trials to test the effectiveness of new treatments or interventions.

But the modern randomised controlled trial (RCT) with placebo control is a relatively recent development. It’s generally credited to Sir Austin Bradford Hill’s 1948 trial of the effects of the antibiotic streptomycin on pulmonary tuberculosis.

His study had all the hallmarks of the modern placebo-controlled RCT. Patients were randomly assigned to two groups (streptomycin treatment or placebo) so that the study started out with two equivalent groups. It was a controlled trial because all participants were treated identically in every way except that one group got the treatment and the other got the placebo.

Neither the participants (single blind) nor the researchers interacting with the participants (double-blind) knew who was getting the treatment and who was getting the placebo.

At the end of the study the difference between the two groups on various outcome measures could then be confidently attributed to the effects of the specific treatment – streptomycin was determined to be useful in the treatment of tuberculosis.

By the 1960s the US Food and Drug Administration required proof of efficacy through RCTs, and by the 1980s doubleblind RCTs were firmly established as the “gold standard” for evaluating the causal effectiveness of a treatment or intervention.

According to the “hierarchy of evidence pyramid”, which ranks strength of evidence based on study design, RCTs are the highest level of evidence, and are the foundation of our current “evidence-based medicine”.

RCTs have contributed enormously to the knowledge base of the medical and health sciences, and to the development of effective drugs and other health interventions, and undoubtedly will continue to do so.

But, RCTs also have serious shortcomings and, particularly as people are increasingly doing their own research on the internet, the limitations of RCTs should be understood.

A great many important research questions don’t allow the random assignment of participants, and so we cannot run an RCT. For instance, it’s not ethical, nor feasible, to randomly assign young