Bart’s Corner: The Sentinel and Mini-Sentinel Initiatives
In 2008 in response to criticism and some safety issues and product withdrawals, FDA took a look at itself – with some outside help such as the FDA Amendments Act (FDAAA). This new law mandated the FDA to establish an active surveillance system for monitoring drugs using electronic data from healthcare information holders. In response, FDA put forth multiple new initiatives covering the areas it regulated. Most of these areas have now undergone significant change or are in the process of doing so.
In the drug/pharmaceutical world one of the big projects is the Sentinel Initiative.
The goal of Sentinel is to create a linked system of databases that will pull insurance claims, and medical and administrative data from multiple, automated healthcare data systems around the country. The goal here is to have continuous, pro-active, real-time monitoring of these data for monitoring of safety issues seen with drugs, vaccines, biologics and medical devices.
Dr. Janet Woodcock of FDA summarized Sentinel and Mini-Sentinel (its initial pilot project) as follows:
“18 large health care organizations across the country serve as data partners for Mini-Sentinel. When FDA safety scientists have a safety question they can submit “queries” to the Mini-Sentinel data partners about the drugs being used by the patients cared for by their organization. Each partner organization maintains its own secure and privacy-protected data, in some but not all cases including EHR as well as claims data, but with the use of a common data model, the necessary information from all of the different systems can be analyzed in the same way. This capability enables Mini-Sentinel to provide answers to FDA questions about drug safety. The Mini-Sentinel system can survey more than 350 million person years of observation, 4 billion pharmaceutical dispensings, and 4.1 billion patient encounters. Thanks to the ability to access data from various sources, the Mini-Sentinel system can use the information from potentially more than 150 million covered lives in our nation’s health care system to help answer important drug safety questions.“
The fundamental goal is the active surveillance of regulated products by assessing electronic healthcare data. A key change from the current system of monitoring spontaneously reported adverse events and safety issues is that this new system does not require active reporting by consumers, patients or prescribers. That is, the FDA will “pull” adverse events from databases rather than relying on the public to “push” AEs to the FDA.
FDA will also use the data to assess whether changes put in place (such as label changes, black boxes, etc.) have made an impact. When all of this is in place, FDA will hopefully be able to act in days or weeks for acute issues. The findings (though apparently not the actual data) will be in the public domain.
Obvious issues of data security and privacy are addressed (hopefully) and “wherever possible” aggregate and individual data will be anonymized. The HHS Office of Human Research Protections (OHRP) has determined that Sentinel activities are public health practice and not research. Therefore they have determined that it is not necessary for FDA to obtain approval from Institutional Review Boards (IRBs) or to obtain waivers of authorization under HIPAA to participate in these activities. For further information on the privacy and data security aspects, see the “Principles and Policies” document covering Mini-Sentinel.
Obviously a lot of preparatory work was needed to get the system up and running. Many problems, bottlenecks and political issues had to be overcome. Merging the data proved to be complex and tricky. This has largely succeeded and the system is now in use. Tools and techniques to mine the data have had to be developed. Some of these include:
- Evaluation of potential data sources for Sentinel
- Evaluation of existing methods of safety signal identification
- Engagement of patients, consumers and HCPs
- Defining and evaluating possible database models to implement the initiative
- Developing a governance and operations structure
- Evaluation of timeliness of medical updates for surveillance in health care databases
- Evaluation of state privacy regulations
Several other projects are completed or still ongoing to further explore valid methodology for the “detection and analysis of AEs related to regulated products in automated health care data”. Wisely, FDA has engaged the stakeholders in the project.:
- “Convener on Active Medical Product Surveillance” (never heard the word “convener” before. It means “someone who organizes the meetings of a committee” and is related to the more common word convene). This turns out to be the creation of meetings and workshops to engage all stakeholders.
- OMOP – The Observational Medical Outcomes Partnership. This was a 5-year public private partnership managed and funded by the Foundation of the NIH. They developed evaluated the performance of analytical methods to identify true associations of products and safety issues in safety surveillance as well as developing better tools to pull data from disparate sources into a single “shared resource” for these data. OMOP recently became part of the “Innovation in Medical Development and Surveillance Program” under the Reagan-Udall Foundation.
So, after this long (6 years) and complex journey, what have we got?
Currently (July 2014) the Mini-Sentinel Distributed Database includes “quality-checked” data from 18 partner organizations including Aetna, Brigham and Women’s Hospital, Cincinnati Children’s Hospital, HMO Research Network, Humana, Kaiser, U of IL Chicago Medical Center, Rutgers, U of Penn and others. The database contains 178 million “members” with 356 million person-years of observation time, 48 million individuals currently enrolled and accumulating new data of which 36 million individuals have over 3 years of data. They have information on 4 million dispensings (of products) from 4.1 billion unique encounters (including 42 million in-patient stays) and 31 million members with at least one laboratory result (including alkaline phosphatase (ALP), alanine aminotransferase (ALT), absolute neutrophil count (ANC), total bilirubin, creatine kinase total (CK), creatine kinase MB (CK-MB), ratio CK-MB/CK, creatinine, fibrin d-dimer, glucose (fasting and randomized), hemoglobin, glycosylated hemoglobin (HbA1c), influenza virus (A, B, A+B, and not specified), international normalized ratio (INR), lipase, pregnancy test, platelet count, troponin I cardiac, and troponin T cardiac.).
Approximately 100 projects have been completed or are ongoing and reports have been issued. They cover both methodology queries as well as evaluation of certain products and AEs. Examples include:
- Immunoglobulin & hemolysis (I & II) assessing several immunoglobulin product groups for hemolytic reactions.
- Influenza vaccines and febrile seizures. The findings showed that “we did not find evidence of a statistically significant elevated risk for febrile seizures in children 6-59 months of age following TIV, PCV13 or DTaP-containing vaccine during the 2010-2011 season.”
- Gardasil vaccination and the risk of venous thromboembolism in 9-26 year old females. This is underway.
There are some controversies, as expected, that touch on control, ownership and funding. Much of the money comes from big pharma and a bit from FDA. Some see this as an industry project and question whether industry should ever fund such initiatives. In response, officials talk about transparency and the multiple stakeholders (not just industry) on the governing committees and boards.
See the article from NPR below for more on this.
On July 21, 2014 NPR (US National Public Radio) did a story on Mini-Sentinel. This is a good, high level lay summary of Mini-Sentinel. Perhaps the best comment made in this report is from Thomas Moore of the Institute for Safe Medication Practices who said: “I think it’s a good and important step that the FDA is moving in this direction…The problem is, I think, they have underestimated how far they have to go.”
The FDA and the pharma industry are now in their seventh year of development of the Mini-Sentinel program. This is one of the first initiatives to look at and harness “big data” for use in pharmacovigilance. As usual in such large scale, national projects with multiple stakeholders and competing interests, it is not simple and will take far longer than projected. There are dangers lurking in regard to privacy and data security as we see in any large database that can be hacked. Nonetheless, early results suggest that this will be a good addition to the methodologic toolkit that FDA and the pharmacovigilance world are developing. Check back in 5 and 10 years.
Tags: #sentinel, drug safety, EHR, FDA, healthcare, medical records, Pharmacovigilance, privacy