Randomized clinical trials: alternatives to conventional randomization.

The randomized allocation format remains an exceedingly powerful tool for clinical research. Because humans are the subjects in clinical research, this area of scientific study must operate within the limits dictated by such basic principles as individual autonomy, justice, and beneficence. As randomized studies have become more common in clinical research, it has become apparent that there is often need for modification of the basic randomized format or for alternatives. The most widespread modification is the use of sequential analyses to monitor the progress of a trial and ensure early identification of either unanticipated adverse effects or more pronounced differences than expected. Although this modification does not affect the basic randomized allocation format, it does provide a protection that is lacking in the conventional trial and should be utilized whenever feasible. Modifications that do affect the basic structure of the randomized trial include adaptive allocation and pre-randomization. The former is attractive and useful but limited to studies in which results from early enrollees are known before late enrollees are allocated. The greater the linkage between preceding subject results and subsequent assignments, the greater is the protection afforded by such a format. Pre-randomization, more universally applicable than adaptive allocation, suffers from pronounced cross-over potential and has been criticized on ethical grounds, a combination of weaknesses that raises questions of whether pre-randomization truly offers advantages to conventional randomized formats. True alternatives to the randomized format include the self-controlled study and the historical control design. Both possess significant ethical advantages over the simple randomized study. Unfortunately, both are at some methodological disadvantage when the same comparison is made. The self-controlled study is limited to the study of conditions sufficiently stable or recurrent that they permit two or more treatment courses in a single patient. In emergency medicine and critical care, this description fits only a small proportion of the illness spectrum. This design is underutilized in clinical research focused on less severe problems of the type seen in ambulatory and primary-care settings. Such problems can be suitable topics for research by emergency medicine specialists. Historical control studies are eminently applicable in the emergency and critical-care setting.(ABSTRACT TRUNCATED AT 400 WORDS)