Some of the practices that are believed to enhance the quality of science may produce bias. Studies with unexciting results may never be published, or results are selectively reported to highlight positive outcomes. Investigators often measure multiple outcomes while only reporting those with statistically significant findings. The best remedy for this problem is to require prospective declaration of study plans through study registration, such as the primary and secondary outcome variables and data analysis plans. Failure to report results of completed studies remains a serious problem. Further, results from many studies remain unpublished and the probability of publication is higher for positive results, leading to overestimates of treatment benefit. It is possible that some encouraging clinical trial findings are actually false positive results. For US Food and Drug Administration evaluations, data from a significant portion of relevant completed trials remain undisclosed at the time the pharmaceutical products are under evaluation.