Sentinel System's Story
The Sentinel Initiative
The Sentinel Initiative, launched in 2008, began as a Congressional mandate for the U.S. Food and Drug Administration (FDA) to establish a public-private partnership to develop a medical product safety surveillance system using existing data. The Initiative is now one of the FDA’s leading evidence-generation platforms. Sentinel proactively monitors medical product safety, provides for methodological innovation, and serves as a platform to advance the science of real-world evidence (RWE).1
Active Post Market Medical Product Safety
Included among the many responsibilities of the FDA is the ongoing monitoring of marketed medical product safety. The Sentinel Initiative was established to help the FDA assess the safety of approved medical products and identify their risks more reliably. The principal component of the Sentinel Initiative is the Sentinel System, a multi-site, privacy-preserving, curated distributed data infrastructure and suite of analysis tools. From its full deployment in 2016, the Sentinel System has become an important tool to help FDA safeguard patients. Since the completion of the Mini-Sentinel pilot phase, FDA has used the System to conduct more than 250 analyses, and it is now embedded in the regulatory review process through the Active Risk Identification and Analysis (ARIA) process. ARIA is comprised of pre-defined, parameterized, reusable routine querying tools that can be used with Sentinel’s distributed data. This allows analyses to be done quickly and efficiently for medical product safety surveillance.
Real World Evidence in Health Care
Interest in the use of using real-world data (RWD) to accelerate medical product development and bring new innovations more efficiently to patients is growing. RWD are data relating to patient health status or the delivery of health care routinely collected from a variety of sources, such as electronic health records (EHR) and administrative data; RWE is the clinical evidence derived from analysis of RWD.2 The 21st Century Cures Act tasked the FDA with developing a program to evaluate the potential use of RWE to support approval of new indications for approved drugs or to satisfy post approval study requirements3; FDA recently published a framework for its RWE program.4
If RWD and RWE are to be effectively leveraged for public health purposes, shared learning and collaboration across clinicians, patients, healthcare systems, pharmaceutical companies, and regulators are necessary. FDA now routinely uses RWD available through the Sentinel System to generate evidence about drug safety, drawing on data from claims, hospital stays, outpatient visits, and pharmaceutical dispensing data from more than 100 million individuals.2 More detailed information about Sentinel System data is available on the Sentinel Initiative website.
Evidence of Effectiveness
The FDA recognizes the interest in generating effectiveness evidence and is exploring the potential of the Sentinel System to support studies of efficacy. Databases with large amounts of patient data are particularly helpful to study rare events. The FDA is separately funding a study to explore whether observational methods can be used to replicate the results of approximately 30 clinical trials designed to provide evidence about the effectiveness of a drug. This project will assist the FDA in understanding how observational methods can be applied to evaluating drug effectiveness and may have the potential to provide evidence to inform regulatory decision-making.2 Although this project is not being undertaken in the Sentinel System, the results will inform how RWD, such as that available in the Sentinel System, might be used to evaluate effectiveness.
Using Real World Evidence
FDA is committed to continually increasing the scope of safety signals the Sentinel System evaluates by identifying opportunities to improve data, tools, and methods. Since the transition to the full-scale Sentinel System, FDA has completed or has underway several projects related to patient and product safety:
- Sentinel data have informed regulatory decisions made by the FDA’s Center for Drug Evaluation and Research and, in some cases, have eliminated the need for postmarketing studies.3 Sentinel’s Active Risk Identification and Analysis (ARIA) System has been used by FDA to examine eighteen potential safety issues associated with nine medical products, as an example, ustekinumab and serious infections.4 Additional ARIA analyses have provided one source of important evidence to support regulatory discussions or actions.5
- Sentinel is being used to develop more efficient algorithms to identify health outcomes of interest, beginning with cases of anaphylaxis and acute pancreatitis using machine learning and natural language processing.6,7
- The FDA is supporting the first randomized clinical trial in Sentinel to explore how randomized trials can be conducted in real-world settings. The IMPACT-Afib trial is testing an educational intervention to address underuse of effective medications to reduce the risk of stroke in patients with atrial fibrillation (Clinical Trials.gov number, NCT03259373).2,3,8
For a comprehensive list of ongoing and completed projects, please visit the Sentinel website.
Looking Ahead – New Tools and Projects
According to the Sentinel System Five-Year Strategy released by the FDA, the Agency envisions, by 2023, a robust Sentinel System that is a transformative, multi-purpose national data and scientific resource center for evidence-generation that a wide array of stakeholders use to inform all aspects of healthcare decision-making.1
The FDA lists the types of data improvements that will be prioritized over the next five years:
- Scaling capabilities related to the mother-infant linkage to evaluate in-utero exposure, medical product usage during pregnancy, and post-natal outcomes.
- Working to integrate national and state registry linkages including the National Death Index (NDI), Surveillance Epidemiology and Ends Results (SEER), and other rare-disease registries.
- Continuing to increase the number of validated Health Outcomes of Interest through medical record review, drawing from increased availability of EHR linkages.
- Expanding linkages to EHR data sources from Sentinel System Data Partners and exploring potential expansion to incorporate other data partners, such as the National Patient-Centered Clinical Research Network (PCORnet).
- Increasing the availability of full medical records, including improved access to the Medicare chart review process, prioritizing electronic sources from integrated delivery systems.
Acknowledging the rapidly changing healthcare system and ongoing scientific progress, FDA is positioning itself to fully realize emerging opportunities using healthcare data to learn, respond to legislative mandates, stay current with scientific advances, and fulfill the Agency’s mission to protect public health.
Reporting Safety Concerns
Any adverse event experience by patients should be reported to the FDA: FDA Adverse Event Reporting System (FAERS).
U.S. Food and Drug Administration. Sentinel System – Five-Year Strategy 2019-2023. 2019; https://www.fda.gov/downloads/Safety/FDAsSentinelInitiative/UCM629243.pdf. Accessed January, 2019.
Ongoing ARIA Assessments. https://www.sentinelinitiative.org/drugs/ongoing-aria-assessments. Accessed August, 2019.
How ARIA Analyses Have Been Used by FDA. https://www.sentinelinitiative.org/drugs/how-aria-analyses-have-been-used-fda. Accessed August, 2019.
Validation of Anaphylaxis Using Machine Learning. https://www.sentinelinitiative.org/sentinel/methods/validation-anaphylaxis-using-machine-learning. Accessed August, 2019.
Validation of Acute Pancreatitis Using Machine Learning and Multi-Site Adaptation for Anaphylaxis. https://www.sentinelinitiative.org/sentinel/methods/validation-acute-pancreatitis-using-machine-learning-and-multi-site-adaptation. Accessed August, 2019.
Implementation of an RCT to improve Treatment With Oral Anticoagulants in Patients with Atrial Fibrillation IMPACT-AFib (IMPACT-AFib). https://clinicaltrials.gov/ct2/show/NCT03259373. Accessed July, 2019.