Skip to Main Content

To assess reporting performance on the National Institutes of Health website, STAT downloaded the site’s entire dataset of about 200,000 trial records on Sept. 10, 2015. The law governing the site contains many exceptions that allow results to be withheld, so STAT filtered the data to exclude all such trials. In other cases, legal deadlines to supply results had not yet arrived, and those trials were set aside as well. That left about 9,000 trials for which results were clearly due under the law.

Read More: Law ignored, patients at risk

STAT used a conservative standard, not counting any study that was not clearly required to report results. For example, pharmaceutical companies can request extensions to the reporting rules for experimental drugs. The FDA approves or denies those extensions, without making the decisions or their timing public. STAT obtained from the NIH a list of trials for which extension requests had been made and, in keeping with the conservative approach, assumed all of those trials had been granted extensions. It did not regard any of their results as overdue.


Performance was measured based on trials that an institution either collaborated on or sponsored. Collaborators don’t normally supply results to But STAT included collaborators’ trials because they act as funders, coresearchers, or principal investigators — roles that warrant some responsibility for ensuring that trial data is reported in a timely fashion. In nearly all cases, trials sponsored by an institution comprised all or the vast majority of its trials.

Explore: Clinical trial reporting lapses visualized

Data for pharma companies acquired by other firms since 2008 were included in the combined entity’s totals. For companies spun off from parent firms, trial data were assigned according to which company retained control over the trial.