Reporting of trial outcomes is incomplete and biased

The medical literature represents a selective and biased subset of study outcomes. Chan and Altman (p 753) analysed all journal articles of randomised trials indexed in PubMed whose primary publication occurred in December 2000. They identified unreported outcomes as those mentioned in the methods but not the results and also by asking authors. In 519 trials with 553 publications and 10 557 reported outcomes, over 20% of the measured outcomes were incompletely reported, and non-reporting was associated with statistical non-significance. Non-significant outcomes of both efficacy and harm were, on average, twice as likely not to be fully reported than were statistically significant outcomes. Protocols of trials should be made publicly available, the authors say.