How Predictive Site Ranking Enhances Clinical Trial Planning, Site Selection, and Trial Rescue

Home / Clinical Minds Blog / Data & AI

3 min read

Aug 21, 2023

Clinical trial planning and execution are notoriously difficult, as they require consideration of a number of complex operational, clinical, and social factors. One of the largest hurdles biopharmaceutical companies and CROs face is clinical trial enrollment.

As of April 2022, it’s estimated that 11% of clinical research sites fail to enroll a single patient¹. Nearly 90% of clinical trials experience significant delays due to issues with patient recruitment¹, which lead to significant financial consequences. Research shows that for each day a trial is delayed, the approximate revenue lost is between $600K and $8M².

Choosing sites that can recruit the greatest number of patients in the shortest amount of time provides an efficient way of overcoming the enrollment challenge. Current industry practice is to select sites based on anecdotal experience or limited historic performance data. Unfortunately, these methods have proven to be inaccurate and can result in costly downstream adjustments.

Predictive modeling helps biopharmaceutical companies and CROs better understand future site performance so they can more confidently choose sites that will enroll patients. A model that predicts site enrollment should consider historical performance (e.g., site enrollment rate, number of completed studies at a site) and contextual information (e.g., country, population density, study congestion at sites) to best provide a list of high performing sites.

How does predicting site enrollment play a critical role in the different stages of a clinical trial?

Conceptual Stage

During the conceptual stage of clinical trial planning, biopharmaceutical companies and CROs are looking to accurately determine the total number of sites and countries they will need and each sites’ anticipated performance. The traditional practice is to apply a single, generalized historical study-level enrollment rate across all sites and countries. But without granular site and country-level performance data, it’s challenging to optimize the number of sites to target in each country.

Predictive modeling gives the user an understanding of how many sites will be high enrolling in each country, which guides the total number of sites and countries to use in a trial to achieve the target study enrollment.

Site Selection Stage

Traditionally, sites are selected for a trial based on their experience running studies in the target indication. This method assumes that higher historical trial participation correlates to larger enrollment potential, which is not always the case.

Predictive modeling, which takes into account factors like historical site-level enrollment, enrollment duration, time to first patient in, and site activation time, can more accurately predict future enrollment performance. With these predictive models, biopharmaceutical companies and CROs can avoid sites that are less likely to enroll patients and focus on potential high enrollers, resulting in a cost-effective strategy to approaching sites.

Execution Stage

In cases where studies are already underway, biopharmaceutical companies and CROs may see that study sites are behind on their expected enrollment. If there are several underperforming sites in a study that are delaying timelines, changes need to be made to make sure the study stays on track.

Through predictive analytics, biopharmaceutical companies and CROs can identify potential top performing sites as add-on or replacement sites, even when the study is already in progress. This step can also be taken as a preventative measure—with these insights, trial teams can choose back-up sites from the onset of the trial and run certain start-up activities in parallel. If a site underperforms during study execution, a back-up site can be activated quickly.

Next Steps: How Medidata’s Intelligent Trials Solution Can Help

Medidata offers a Study Feasibility solution to support biopharmaceutical companies and their strategic feasibility teams with the tools they need to avoid the repercussions of poor site selection. This solution includes predictive site ranking models for 127 indications.

Medidata’s predictive models are trained on its vast, real-time clinical database of over 30,000 sites. Models include both historical and contextual features and can predict future site enrollment performance with greater accuracy than traditional methods alone. By leveraging models like these, biopharmaceutical companies and CROs can make more data-driven decisions and confidently select high-performing sites to meet enrollment timelines.