Ambiguous Dropouts and the Goals of this Chapter
Reports of randomized clinical trials (RCTs) frequently refer to “dropouts” from the study. Unfortunately, RCT authors may be referring to 1 of 2 different phenomena. The first phenomenon includes study participants who no longer comply with the intervention (typically the experimental treatment and typically a drug) but who are willing to continue with follow-up for the outcome events of interest. The second phenomenon includes study participants who, for a number of possible reasons (eg, moved away or refused further participation), are no longer available and are thus lost to follow-up.
How to deal with these 2 groups of patients in the analysis phase of a study has captured the attention of clinical trialists for several decades. The implications of these analysis decisions can be important for understanding the results of an RCT. This chapter deals with the issues of nonadherence and loss to follow-up and how RCT authors should, and should not, deal with such trial participants when it comes to analyzing their data.
How Should Randomized Trials Deal with Treatment Arm Patients Who do Not Receive Treatment?
If patients do not take their medication, they do not benefit from their medication. Furthermore, we do not need RCTs—or studies of any kind—to substantiate this lack of benefit. One might therefore reason that in an RCT investigators should compare patients in the experimental group who actually received treatment with patients in the control group who did not. As it turns out, however, doing so is usually a mistake. We need to know about all of the patients in a trial, including those in the trial groups who did not adhere to or complete therapy.1
One argument for incorporating all patients in the final analysis, including those who did not adhere to treatment, has to do with the effect of the treatment on the community of individuals for whom the treatment is prescribed. If one is interested in knowing the effect of a drug on a given population, one must include all members of that population. When patients do not adhere to a regimen, particularly if adverse effects have caused them to be nonadherent, reservations will arise about the effect of a medication on a community. This is important information for policymakers and other investigators—but perhaps not so much for clinicians.
As clinicians, we are typically more interested in the effect of our interventions on individual patients rather than on populations. Consider the viewpoint of patients who are determined to adhere to a treatment regimen and are destined to succeed. Let us assume that 50% of patients in the treatment arm of a trial did not comply with the treatment regimen. Do the motivated patients wish to know the average effect of the treatment in a group of people of whom 50% did not comply? No, they want the best ...