Salvaging information from paused or stopped clinical studies

This letter aims to aid research teams in preserving the information from clinical studies that have been paused or stopped for unanticipated reasons. Our goal is to provide guidance to researchers seeking to honor the contributions of participants by increasing the chances of ultimately being able to answer the primary study questions. ¹ We note that these considerations are especially relevant in settings where study funding can be unstable.

For settings where trials may end before they can reach conclusions (e.g. transient infectious outbreaks), Dean et al. ² advocate for core protocols that can be modified for different settings but are designed to provide compatible data. Furthermore, platform trials, which can operate like core protocols, use a single control group for multiple interventions that enter the trial at different times (e.g. see Longini et al. ³ ). This shared control improves efficiency and facilitates combining information across trials. Developing platform trials remains the most effective strategy for generating reliable evidence when studies are at risk of unexpected stopping.

Below, we consider options when no such core protocol exists prior to the start of studies. In addition to maintaining trial databases,^4,5 it is crucial to implement policies that are necessary for these studies to meet their scientific objectives, including the following. (1) No release of trial data after stopping. (2) If the study has a Data Safety and Monitoring Board (DSMB), it conducts an interim review of the study for both efficacy and futility. (If there is no DSMB, the study team should consider assembling one for this purpose.) (3) Interim analysis should follow standard alpha-spending rules, generally based on the fraction of accrued information.

If the DSMB does not recommend stopping, then either plan to continue the study after a pause if conditions permit or run a separate trial with a protocol that is as similar as possible to the original. If the second trial is also unlikely to reach its information target, an interim analysis should be conducted. If the DSMB does not recommend stopping, a team with no access to information from either trial (except perhaps summary information combined across arms from the first trial) could recommend combining results into a single “merged” trial.

When there is no possibility of completing individual or merged trials, meta-analyses can provide overall summaries of results, understanding that this approach cannot preserve overall type I error control regarding primary hypothesis tests. Here, teams conduct separate trials and then use standard methods to combine results. ⁶ Trial designs should be standardized as much as possible to facilitate interpretation of results.

Our proposals are consistent with principles of ethical study closure. One exemplary model was the closure of three open-label phase 1 trials of HIV prevention in pregnant women following United States Agency for International Development (USAID) stop-work orders (Sharon Hillier, personal communication). The MATRIX study team developed tailored closeout plans, which included recovery and destruction of study products; timely notification with ethics boards and regulators; transparent communication with participants regarding safety results and trial terminations; assurance that biospecimens would be used only as originally intended, then destroyed; and final review of data integrity by medical and data experts before database lock. We propose adding a sixth element: explicit plans for future study data use, including policy regarding release of data. A current inability to meet study goals does not preclude the possibility of achieving them in the future.

Finally, we note that new methods might also aid in our goals. Perhaps a core protocol could be reverse-engineered from a paused or stopped study—using a platform trial design where appropriate. Such methods should facilitate the development of successor studies, obviating the need to design entirely new trials from scratch. If the paused study were able to restart, that study would be analyzed as originally planned, but the controls from the restarted portion of the study could also serve as controls for other studies in the platform.

Our recommendations have important implications for Food and Drug Administration (FDA) policy. Specifically, the options we propose for completing paused or stopped trials require that study data not be prematurely released. This recommendation could appear to conflict with the FDA requirement to report results of non-phase I trials within 1 year of termination. Nevertheless, a “good cause extension” of this deadline can be requested. We believe that preserving the scientific integrity of trials by following our proposals for data release should be considered as one of the listed good causes for granting extension requests. ⁷

In summary, we propose methods to protect the scientific potential of studies to answer important questions and thereby maintain the trust of the people who participate in them. We believe that including our proposals as part of rigorous ethical closure practices will enhance the integrity of the research enterprise and honor the contributions of participants and will potentially address the needs of people who would benefit from trial results.