Let us suppose we have a treatment and we want to find out if it works. Call this treatment drug X. While we have observational data that it works—that is, patients say it works or, that it appears to work given certain tests—observational data can be misleading. As Edzard Ernst writes:
Whenever a patient or a group of patients receive a medical treatment and subsequently experience improvements, we automatically assume that the improvement was caused by the intervention. This logical fallacy can be very misleading […] Of course, it could be the treatment—but there are many other possibilities as well.
So we decide to hold a Randomised Control Trial (RCT). An RCT takes account for the non-specific effects of the treatment—these misleading possibilities Ernst is so interested in. The RCT has three arms to the experiment: the first arm receives the active treatment (in our case, drug X); the second arm receives a placebo, a treatment with mimics the drug X except in that it is devoid of the active ingredients (in our case, it’ll be a sugar pill); the third arm receives no treatment whatsoever. Following the RCT, we’ll know: first, whether the treatment is doing anything at all; and second, whether or not it is the active elements of the treatment that are doing the work.
The results of our RCT say that the first and second arms of the treatment are effective (that is, those patients receiving drug X and the sugar pill); however, there doesn’t appear to be any difference between those patients receiving drug X and those receiving the sugar pill.
The views, opinions and positions expressed by these authors and blogs are theirs and do not necessarily represent that of the Bioethics Research Library and Kennedy Institute of Ethics or Georgetown University.