[andrewla]

In the US, this is already present, to some extent, with FDAAA 801 [1], which requires that all clinical trials be registered before they start. Missing things like required publication and the "human-readable" digest, but the absence of data can be factored in by the FDA when considering approval -- that is, studies that terminate without reaching a finding can be considered weak indicators of problems, and in the long run, potentially point to organizations or individuals who might be suppressing negative results.

Even required publication is no cure for this; a result saying "we inadvertently broke the blind" or "the protocol was poorly implemented" is perfectly possible even in good faith.

[1] http://www.clinicaltrials.gov/ct2/manage-recs/fdaaa