The mission of the US Food and Drug Administration (FDA) is to protect and promote public health. FDA does this by ensuring the safety, efficacy, and quality of human drugs, biological products, and medical devices as well as other FDA-regulated products. FDA is also responsible for making sure that the public has access to accurate, comprehensible science-based information for optimal use of medical products. Postmarket safety surveillance—monitoring the safety of medical products once they reach the marketplace—is key to these efforts.
For decades, FDA has relied primarily on spontaneous reporting systems to monitor postmarket safety. These systems depend on the public—both healthcare practitioners and patients—to voluntarily report adverse events, errors, and quality problems that they observe during use to either the manufacturer or directly to FDA; medical product manufacturers are legally required to submit to FDA certain product-related adverse events reports that they receive. Spontaneous reporting approaches have some limitations, however, including underreporting of adverse events and incomplete information on the reports that are submitted to FDA. Furthermore, although these systems are excellent for generating hypotheses of potential product-associated adverse events, the number of events reported (numerator) does not represent the actual number that occurred, and there is no exposure (denominator) data. As such, direct calculation of adverse event rates cannot be performed. Recognizing the limitations of these existing surveillance systems, FDA has long worked to strengthen postmarket safety monitoring. Enhancing safety monitoring capacity that complements spontaneous reporting by developing active surveillance capabilities is a goal that has received increased attention during the past 5 years.[1, 2]
In 2008, responding to the congressional mandate in Section 905 of the FDA Amendments Act of 2007 (FDAAA),i FDA launched its Sentinel Initiative. This Initiative seeks to leverage existing health care information to enable FDA to conduct active postmarket safety surveillance to augment its existing surveillance systems. This concept of secondary use, or using data collected for other purposes, such as electronic health record data originally recorded for patient care or insurance claims data used for reimbursement, is not novel. However, with the expansion of the availability of these types of data, based on the passage of the US Health Information Technology for Economic and Clinical Health (HITECH) Act,ii more focus has been directed to leveraging these data for purposes not originally intended, such as to inform public health issues. Safety surveillance of medical products is such an issue. Using data from administrative and claims databases, electronic health record systems, and registries will make it possible for FDA to monitor regulated medical products in near real-time and to better understand product safety once the products are being used by a broader population than that studied during clinical trials.
The system being created under the auspices of the Sentinel Initiative (the Sentinel System) will help FDA identify and investigate postmarket safety signals, a concern about an excess of adverse events compared with what is expected to be associated with a product's use, through the processes of signal generation, signal refinement, and signal evaluation. Signal generation is an approach that uses statistical methods to identify medical product–adverse outcome associations that may be safety signals; no particular medical product exposure or adverse outcome is pre-specified. Signal refinement is a process by which an identified potential safety signal is further investigated to determine whether evidence exists to support a relationship between the medical product exposure and the outcome. Signal evaluation consists of the implementation of a full epidemiological analysis to more thoroughly evaluate the causal relationship between exposure to the medical product and the adverse outcome of interest.
Signal refinement, the initial focus of the Sentinel Initiative pilot programs, can be useful at various times during a product's lifecycle. Sometimes a safety signal is identified during the product's premarket review process. In such a case, the signal might be further refined through prospective monitoring and investigation at pre-specified intervals during the postmarket period in a defined population. In other situations, a safety signal may emerge de novo during the postmarket period. When this happens, signal refinement methodologies can be applied to help gain a better understanding of the potential safety signal.
Should a signal refinement investigation provide evidence supporting a potential association between product exposure and an adverse outcome, it will often be necessary to conduct additional analyses to validate the signal. The validation process is conducted to ensure that the medical product–adverse outcome relationship is not spurious. This process will also likely provide more information about the safety signal, particularly if source record verification is included. Because health outcomes of interest (HOIs) are being identified using coded information (e.g., diagnosis, procedure), source records may be sought to verify that patients indeed had the HOI.
In all cases, information gained from the Sentinel System will be considered in the context of all other data about the potential safety signal. To make a regulatory decision, FDA staff evaluate data from premarket development programs (e.g., preclinical and clinical studies, clinical pharmacology studies, engineering data; the available information would vary by the product type), spontaneous reports, any available postmarket studies, and relevant literature. The Sentinel System investigation will offer FDA an additional source of information to support regulatory decision making.
To foster development and implementation of the Sentinel System, FDA has embarked on a variety of activities, including pilot programs, to help develop scientific methodologies, learn about data infrastructure needs, and inform the agency about how to create an appropriate governance structure to guarantee the privacy and security of the data being accessed for these active safety surveillance investigations. A key learning from these early activities identified a distributed data system model with voluntary participants (i.e., data partners) as the preferred approach for organizing an active medical product safety surveillance systemiii for several reasons. This model will make it possible for data to remain in its local environment, in contrast to a centralized approach, which consolidates the data into one physical location. Other benefits of the distributed model include being better able to maintain patient privacy by keeping directly identifiable patient information behind local firewalls in its existing protected environment. Additionally, because of the data partners' awareness of the changes that have occurred in their healthcare systems that result in the unique qualities of each database, this model will enable data partner involvement in running the analyses and ensures an informed approach to interpreting results.
The initial efforts of FDA have been implemented through the Mini-Sentinel pilot project, which is intended to function as a kind of laboratory that can inform FDA on scientific and technical issues related to the development of the Sentinel System. Mini-Sentinel involves over 20 organizationsiv led by the Harvard Pilgrim Health Care Institute (HPHCI); they contribute both data and expertise related to using this type of data for public health surveillance. Mini-Sentinel collaborators have worked over the past year to explore some of the key issues for creating a US system for active medical product safety surveillance, including what statistical and epidemiological approaches to use, what data attributes and infrastructure will be needed to enable surveillance investigations, and what kind of governance structure will work best to support the pilot. FDA is also taking advantage of the experience gained from the Vaccine Safety Datalink's population-based, active surveillance program for monitoring the safety of new vaccines, as well as the H1N1 vaccine safety surveillance experience from in the Post-licensure Rapid Immunization Monitoring System (PRISM).[4, 5] In year two of Mini-Sentinel, PRISM has been adapted to function as a general purpose vaccine safety monitoring system.