ABSTRACT
Randomization is the keystone of modern evaluations of new treatments. In human subjects the potential for selection bias is so great that unless a treatment is overwhelmingly effective, the only way to reliably assess its value is via a randomized trial. The dangers of not randomizing treatment allocation have now been repeatedly observed in clinical studies where subsequent randomized trials have clearly shown that major inaccuracies were reported in the non-randomized precursor trials.
