Biased reporting arises when two main decisions are made based on the direction and statistical significance of the data—whether to publish the trial at all, and if so, which analyses and results to report in the publication. Strong evidence for the selective publication of positive trials has been available for decades [7,8]. More recent cohort studies have focused on the misreporting of trials within publications by comparing journal articles either with documents from regulatory agencies [9–12] or with trial protocols from research ethics committees [13–16], funding agencies [17], research groups [18,19], and journals [20]. These cohort studies identified major discrepancies—favorable results were often highlighted while unfavorable data were suppressed; definitions of primary outcomes were changed; and methods of statistical analysis were modified without explanation in the journal article.

Overall, a substantial amount of primary outcome data submitted to the FDA was found to be missing from the literature. One quarter of trials in their sample were unpublished—predominantly those with unfavorable results. Not only were data suppressed for the unpublished trials, but an additional quarter of primary outcomes were omitted from journal articles of published trials. These findings are consistent with two recent reviews of FDA documents and journal articles [10,21], one of which was published in PLoS Medicine in September 2008 [21].

Bero and colleagues also identified important discrepancies between the primary outcomes, statistical analyses, and conclusions presented in NDAs versus those reported in journal articles. The vast majority of discrepancies favored the sponsor's new drug, suggesting biased reporting. While it is possible that the FDA requested modifications to the sponsor's analyses, these amendments should be mentioned in the FDA's statistical review; should not involve altering primary outcomes without explanation in the publication; and would not be expected to favor the sponsor's drug as often as was found in this study.

Regulatory agency submissions represent the final description of how the trial was conducted and analyzed prior to journal publication. However, details from these submissions are not publicly available in most countries. Although summaries of FDA reviews are posted on the FDA Web site, their content and availability is variable, and sections are often redacted [9,21,23]. Furthermore, regulatory agency submissions are prepared by companies after data analysis and may themselves be subject to biased reporting. Finally, only devices, pharmaceuticals, and biological agents require regulatory approval in the United States and other countries, meaning that trials examining other types of interventions (e.g., surgery, education)—which constitute 20% of published randomized trials [24]—would be excluded from reviews of regulatory agency documents. Pharmaceutical trials conducted post-approval would also be missed.

However, much remains to be done—not only to establish reliable, comprehensive registration and results disclosure processes worldwide, but also to start heeding the calls for increased access to full protocols and regulatory agency submissions [14,23,33,37,38]. As shown by recent examples and studies highlighted above, misreporting of trials can be difficult to detect without access to detailed documents beyond what is currently available on registries and results databases. Only with full transparency can the validity of a randomized trial be judged.