Trial results and publication bias