Fewer than half of medical clinical trials published in the last year in leading medical journals are reported on the government site ClinicalTrials.gov, according to a new study published by the Journal of the American Medical Association (JAMA).

According to the study, despite a 2005 requirement by medical journal editors that researchers register their studies online before acceptance in a major journal, just 147 out of 323 articles published in the past year were properly registered before the end of the trial. To meet the requirement, researchers are supposed to state what outcomes they will measure to figure out trial success.

Even among the articles that were registered, amost a third “had discrepancies between the outcomes described in the registry and the ones ultimately reported,” reports the New York Times. Eighty-nine of the trials were not entered in the ClinicalTrials.gov at all.

If researchers don’t report what outcomes they are measuring before they publish a final study, they can easily doctor the results for a positive spin. For example, reports the Times, “a clinical trial is designed to determine whether a drug increases survival rates but investigators change the primary outcome to assessing if the drug improves quality of life.”

This is the perfect case of “secret” data–data that technically the government is supposed to have, but is being kept from the public. In this case, the apparent diagnosis is that the requirement for publishing at ClinicalTrials.gov is based on a private incentive rather than a straight-out requirement.