Time to publication of oncology trials and why some trials are never published

Very little is known about the proportion of oncology trials that get published, the time it takes to publish them, or the reasons why oncology trials do not get published.

We analyzed all clinical trials that closed to accrual at our cancer center between 2009-2013. Trials were categorized by study purpose (therapeutic vs. diagnostic), phase (pilot, phase I, II, or III), and sponsor (industrial, cooperative group, institutional, or peer-reviewed). Final publications were identified in MEDLINE and EMBASE by NCT numbers, or by querying the principal investigator. For trials not published, we surveyed the principal investigators to identify the reason for non-publication.

469 of 809 protocols (58%) had been published by November 2016. The calculated probability of publication 7 years after completing accrual was 70.4%; the calculated median time to publication was 47 months. Only 18.8% of protocols overall were estimated to be published within 2 years from completing accrual. The calculated probability of publication was higher for therapeutic trials than non-therapeutic trials, but there was no difference based on phase or sponsor. Among protocols not published, 45.3% had completed accrual, and among these, a majority had a manuscript in preparation or review, or the trial was still collecting data. Failure to publish due to a pharmaceutical sponsor was rare. 30.6% of unpublished trials had closed for various reasons before completing accrual, usually due to poor accrual or pharmaceutical sponsor issues.

Almost 30% of trials were calculated to be unpublished by 7 years after closing to accrual at our institution. Failure to reach accrual goals was an important factor in non-publication. We have devised new institutional policies that identify trials likely not to meet accrual goals and require early closure. We should be able to shorten the time from accrual completion to publication, especially for pilot and phase I trials for which long follow up is not needed.