A post-market safety surveillance system being piloted at five Massachusetts hospitals could inform FDA's efforts to broaden its Sentinel post-market safety system to devices, speakers at a recent Brookings Institution roundtable suggested.

According to Tom Gross, deputy director of the Office of Surveillance and Biometrics at FDA's device center, the center is exploring various means of active device surveillance, and hopes to learn from the Massachusetts pilot.

FDA recently launched its "mini-Sentinel" safety surveillance system to actively identify and analyze drug safety data from various sources, but it remains unclear how active surveillance will work for devices (¹ Also see "Major Work Ahead To Make "Mini" Sentinel System Work On A Larger Scale" - Medtech Insight, 18 January, 2010.).

Challenges for medical device safety surveillance include the lack of specific identifiers of the manufacturer and model involved in adverse events, data security issues and a lack of consensus regarding methods for detecting and interpreting potential safety problems, Fredric Resnic, M.D., Brigham and Women's Hospital, said during the May 7 Brookings webinar.

Resnic's team is evaluating the functionality of prospective, active device safety surveillance in a pilot study supported by an FDA research contract National Institutes of Health grant.

The Massachusetts Interventional Cardiology Device Safety Surveillance pilot project uses a web-based system to analyze data from the five hospitals, collected in a centralized database at Harvard Medical School.

The Delta (Data Extraction and Longitudinal Time Analysis) system is designed to generate alerts about potentially unsafe devices or procedures.

Now, five Massachusetts hospitals have implemented the Delta system locally, linking the software to their electronic medical records systems.

The hospitals send "well-structured" data to the central database at Harvard on a monthly basis, Resnic said.

The study will focus on interventional cardiovascular devices, including drug-eluting stents, vascular closure devices and embolic protection devices, and will include a propensity-matched control population, such as patients receiving products similar to the tested device.

Acute adverse events such as death, myocardial infarction, device failures and bleeding will be assessed.

Three Distributed Data Models Studied

One of the chief concerns regarding centralized data surveillance is patient privacy protection. The pilot study will test three methods of data access, with varying levels of privacy protection.

The first method sends de-identified, encrypted patient data to the central database, which then performs the Delta analysis.

The second applies the Delta model locally, and sends the encrypted case ID and outcomes, plus the Delta-predicted outcomes, to the central database. The third method sends only the results of the local Delta analysis to the central server at Harvard.

With each method, the researchers are looking for time savings gained through use of the automated surveillance system, "versus how much time we waste by exploring alerts that may not be significant," Resnic said.

"We have a goal of having a time-efficient, highly sensitive alerting system to trigger detailed investigation of possible safety concerns."

He emphasized that the alerts should be considered "hypothesis-generating," triggering a further clinical and statistical analysis of the specific cases. "Any tool that's developed, Delta or anything else, really ought to have the epidemiologic tools to allow the analysts to explore, 'What does the signal mean?'" Resnic added.

The ultimate vision, he said, is to reach "a day, hopefully in the not-too-distant future, when every new medical device will be subject to prospective surveillance with well-constructed parameters for triggering further review."

- Jessica Bylander ( ² j.bylander@elsevier.com )

Harvard Device Safety Surveillance Project Could Inform FDA Efforts

More from Archive

More from Medtech Insight