Keywords: prediction, clinical trial, enrollment
Clinical trials often fail to meet their enrollment targets, resulting in costly delays to market authorization or label extensions. Typically, site identification has relied on institutional knowledge and enrollment experience from previous studies. However, even when it’s available, prior experience may not be very useful because study eligibility criteria can vary significantly across studies. Furthermore, these precedents may become less useful as competing new therapies are used in clinical care. The rapidly-changing patient mix of comorbidities and standard of care are reflected in real-world data which we have used in a number of studies to generate a prioritized list of clinical sites and their expected eligible patient population. Now we are developing an enrollment simulator that combines site-specific target patient population size, insights on how frequently patients have face-to-face encounters with their providers, and historical enrollment rates. In this presentation we will share our recent experiences using real-world data to simulate clinical trial enrollment. We will highlight the statistical algorithms we have utilized, illustrate the benefits of this data-driven approach, and describe the challenges we have faced in implementation and value demonstration. We believe this tool will be especially useful in the planning of pragmatic clinical trials and other post-marketing studies.