[How blind is a double-blind trial really? Validity of controlled investigations].

During a double blind trial lasting 4 weeks (Diazepam versus Loxapin), two investigators tried to identify the given drug prematurely. Only the more experienced investigator was able to identify the two drugs (p = 0.01). His judgement was based neither on observation of side-effects nor on a clinical improvement but apparently on a summary of particular subtle effects, the cognition of which requires a specific experience. The use of an adequate comparative drug instead of a placebo in a double blind study does not completely prevent a premature decoding but should reduce this risk and therefore increase the validity of the method.