The most valid way to determine the efficacy of a treatment for fatigue is to randomly assign patients to receive either the treatment or some alternate intervention for comparison. Random assignment minimizes the risk of systematic bias by distributing variables that may influence outcome between the study groups.

Random assignment of treatments is a prerequisite for a controlled trial. It does not, however, eliminate the possibility that the outcomes observed are related to placebo effects. To establish the efficacy of an intervention distinct from a nonspecific placebo effect, the intervention must be compared directly with placebo. If feasible, a study design that incorporates randomization, double-blinding of treatments, and a placebo control is most likely to provide evidence for or against efficacy.

Of course, a placebo control is not possible with most non-pharmacological interventions, such as exercise, cognitive therapies, or education about sleep hygiene. In these studies, patients are usually randomly assigned to receive the study intervention or a control selected to reduce some sources of bias. For example, contact with the investigator is a potential source of bias (i.e. positive outcomes may not be related to the intervention per se, but rather to nonspecific positive effects elicited through contact with the investigator) and a useful control group may include some degree of investigator contact without access to the study intervention. In more sophisticated trials, more than one control group of this type can be created in an effort to further clarify the intervention effect. For example, a study of a cognitive therapy (such as hypnosis) may incorporate a control group that receives some educational materials related to the problem of fatigue and another control group that is allowed to make an appointment with the investigator but will receive no intervention (so-called waiting list control).