For example, in RCTs of mammography screening, one third of participants in the treatment group have refused screening and many participants in the control group have obtained screening outside the trial. However, these assumptions are often not met. 5 If these assumptions of the RCT are met, strong inferences can be drawn about the average causal effect of treatment t relative to treatment c on the outcome. For the RCT, the additional assumptions required are (1) the units are independent, (2) participants actually received the treatment as intended (e.g., complete treatment adherence), (3) attrition from posttest measurement did not occur, and (4) the existence of other treatment conditions did not affect the participant’s outcome. Additional assumptions are required depending on the choice of alternative to the ideal design. Unfortunately, this ideal can never be achieved in practice. Theoretically, comparison of these 2 outcomes provides the ideal design for causal inference. Where Y t(u) represents the response of unit u to treatment t, and Y c(u) represents the response of the same unit u to the control treatment c at the identical time and in the identical setting. 6 – 11 Two perspectives have guided this work. Advances in alternative designs and statistical analyses have also occurred.

Important advances addressing the challenges of nonadherence, attrition, and their combination have been made during the last half century. People can refuse to be measured or migrate to another locale (attrition). 5 Unlike the corn plants in Fisher’s agricultural studies, people can seek out alternative treatments or refuse treatments (nonadherence to treatment). However, other features of the public health context require additional assumptions when traditional RCTs are utilized. In the context of Fisher’s agricultural studies, these expectations guaranteed that the design would provide an unbiased estimate of the true causal effect.

Fisher’s key insight was that random assignment of units to treatment conditions leads to 2 related expectations: (1) the mean level for each of the treatment conditions is equal, on average, on any conceivable participant background variable prior to the beginning of the experiment and (2) treatment assignment is, on average, unrelated to any conceivable participant background variable. The RCT has its origins in the pioneering work of the English scientist and statistician Sir Ronald Fisher 4 in agriculture during the 1920s and 1930s. Parallel designs exist in which settings, time, or even dependent measures are the unit of assignment. With one exception, our focus is on designs in which participants are assigned to treatment or control conditions. Randomization may be precluded if the religious organization implementing the intervention strongly believes that all people desiring the faith-based intervention should receive it. Highly religious participants may refuse to be assigned to a non–faith-based treatment group, whereas non-religious participants may refuse or be unable to participate sincerely in a faith-based group. Only atypical participants may agree to participate in the evaluation of a faith-based intervention. Such problems might be a concern in studies of the effects of environmental tobacco smoke on non-smokers or of the effects of the severe disruption of Gulf Coast communities by Hurricane Katrina on HIV risk behaviors and medical care. When structural or policy interventions are being examined, practical problems in conducting RCTs may arise 2-for example, research participants may not want to be randomized, randomization may not be feasible or not accepted in the research context, or only atypical participants may be willing to be randomized. However, public health researchers are increasingly addressing questions for which the RCT may not be a practical (or ethical) option or for which the RCT can be complemented by alternative designs that enhance generalization to participants and contexts of interest. THE RANDOMIZED CONTROLLED trial (RCT) has long been the gold standard for clinical research, representing the best way to determine efficacy and effectiveness for many intervention and prevention programs. Far better an approximate answer to the right question, which is often vague, than an exact answer to the wrong question, which can always be made precise.

Final draft trial limitations