[A critical review of the pragmatic clinical trial].

Schwartz and Lellouch proposed in 1967 to differentiate between pragmatic and explanatory clinical trials. The pragmatic approach has been understood in different ways and has also been the object of some criticism. Yet different authors have asked to perform pragmatic clinical trials, especially in issues requiring a therapeutical decision. In order to clarify the present situation of the pragmatic trial, a bibliographic review has been performed. Between 1976 and 2002, 95 original articles on clinical trials conducted following a pragmatic design have been found. Only four followed strictly the approach developed by Schwartz et al. In a pragmatic clinical trial, it is usually not possible to blind the interventions and for this reason randomization is imperfect. This limitation casts doubts about the validity of the frequentistic methods in the intention-to-treat evaluation of the results of a pragmatic trial. This experimental approach demands an evaluation according to the decision-making theory. Following the Bayes theorem, the credibility and the previous probability of a hypothesis conditions its posterior probability. It has been agreed that Bayesian statistics is a suitable instrument for the evaluation of a pragmatic clinical trial, but the lack of adequate informatics' programs has limited seriously its application. Recently,some new programs (WinBUGS, TreeAge) have been developed and applied to the decision analysis in some primary care therapeutic questions. It seems possible to predict that, thanks to the new informatics'programs on Bayesian statistics, the pragmatic clinical trial will experience, in the short term, an important revival.