Clinical Trials: Sharing the Road With Real-World Evidence

June 4, 2018

In the era of real-world data and its growing role in oncology, panelists discussed collecting and using this information in combination with clinical trials to inform evidence-based care during a session at the 2018 American Society of Clinical Oncology Annual Meeting.

In the era of real-world data and evidence and its growing role in oncology, panelists discussed collecting and using this data in combination with clinical trials to inform evidence-based care during a session at the 2018 American Society of Clinical Oncology Annual Meeting, June 1-5, in Chicago, Illinois.

While clinical trials remain the gold standard for the context in which they were designed and developed for—evaluating efficacy in tightly controlled and highly annotated samples—there are some drawbacks to using clinical trial data alone, explained Kathryn Reeder-Hayes, MD, MBA, MSc, assistant professor, University of North Carolina at Chapel Hill.

“In this context, assuming they meet their accrual goals, there is almost never too little data for the job in terms of the level of detail and the completeness of the data,” she said. She also noted that clinical trials are a familiar and comfortable form of evidence generation for multiple key stakeholders who are involved in the development of novel therapies for oncology.

However, clinical trials can also be prohibitively expensive and may not be the best fit for every question relevant to clinical practice, Reeder-Hayes warned. In particular, they are not the optimal fit for questions of application of the innovations to broad and diverse populations; and they are not well-adapted to answer post-hoc questions about differences in efficacy across subgroups or application to populations outside the clinical trials, often because the sample is either too small, too homogenous, or both, for the data to be able to answer these questions.

“In those contexts, clinical trials may in fact waste time and resources, both in terms of our economic sources and our patients’ time,” said Reeder-Hayes. “In those contexts, real-world evidence, I would argue, may be less expensive and more appropriate for some questions as they relate to post-hoc analyses and diffusion into broad populations.”

According to Reeder-Hayes, real-world evidence can be beneficial in several places:

  • After a randomized clinical trial is conducted to test the dissemination of the findings and if they are being adopted, in which patients they’re being adopted, and which patient populations are being left behind.
  • Alongside randomized clinical trials to extend findings to broader populations and answer secondary questions about differences among subgroups.
  • Before anticipating trials to inform the important problems and questions, quantify effect sizes, and identify the right population.

She concluded by cautioning that big data studies do require expertise in handling and analyzing observational data with its unique challenges related to potential bias and the need for complex data management strategies. Similar to clinical trials, these studies are best performed by experienced cross-disciplinary teams, and these studies are most useful when they answer the questions important to physicians and patients.

Sean Khozin, MD, MPH, director, Information Exchange and Data Transformation (INFORMED), followed Reeder-Hayes with an introduction to the FDA’s recently launched INFORMED program.

“Big data has many difference dimensions,” said Khozin. To explore and address these dimensions, the FDA launched the INFORMED program in April as an incubator for collaborative oncology science research. The program pairs engineers and data scientists with medical reviewers and regulatory scientists to conduct regulatory research using a variety of data inputs including: clinical trials, electronic health records, biometric monitoring devices, and applications.

From these inputs, results come in the form of publications, abstracts, and more recently codes and algorithms that can be incorporated into decision support tools, explained Khozin. Outputs also include policy positions and guidance documents that can disseminate findings to the community and inform development programs, he added.