In critical care medicine there is still a paucity of evidence on how to manage most of the clinical problems commonly encountered in critically ill patients. Randomized controlled trials (RCTs) are the most powerful instruments to evaluate the efficacy of a therapeutic intervention and to generate evidence for clinical practice. Unfortunately, the design and conduct of RCTs in our field are particularly complicated, because of some intrinsic and structural problems (e.g. lack of reliable nosography, concomitant use of different therapies, problems in the definition of end-points besides mortality) that will be discussed in this review. Further challenges are represented by the lack of tradition of large ICU networks, difficulties in linking or integrating physiologic and therapeutic objectives in designing clinical protocols, scarcity of independent or non-profit funds. A particularly stimulating opportunity of development is represented also by the relationship of critical care to EBM. Because of the above problems, metanalyses could be less informative than in other areas of medicine, as they are based on few trials which are often contradictory and of unsatisfactory quality. Few suggestions are formulated which could help looking forwards.