MEASURING EFFECTS WITHOUT RANDOMIZED TRIALS - OPTIONS, PROBLEMS, CHALLENGES

The idea of using information routinely generated in the course of hea lth care delivery for assessing the relative efficacy of alternative t herapies has an undeniable attraction. If feasible, it would enable th e difficulties of running a randomized controlled trial (RCT) to be by passed. But grave obstacles, both practical and theoretical, beset the effort. If, nonetheless, the data base approach is chosen, certain su ggestions may be helpful. First, make sure that necessity actually pre vents using an RCT. Reconsider that question. Special attention is due to two variants of the usual RCT: the ''firms'' approach and the larg e simple trial. If, after reconsideration, a data base approach still is chosen, then let the undertaking be carefully planned, with partici pation from those who will produce the data and also from future users of the results. Let the planning result in a protocol, and in provisi ons for a quality assurance program. Finally it is time to respond to a challenge left us by David Byar: arrange to record why the patient i s being given the therapy selected. This information should be a power ful adjustment variable; arranging to collect it will call for imagina tive thinking, experimentation, and patience, but it is an idea deserv ing much effort.