The US Food and Drug Administration (FDA) is piloting a new statistics-based approach on the process for selecting device manufacturing facilities to conduct site inspections, agency officials told Focus.
Developed at the Division of Biostatistics within FDA’s Center for Devices and Radiological Health (CDRH), a new software tool would transform the way agency staff are informed on where and how frequently to conduct site inspections through the Bioresearch Monitoring (BIMO) program.
The pilot is part of ongoing efforts to extend FDA regulatory activities across the total product lifecycle.
Adoption of the tool has the potential to improve FDA’s resource allocation—an issue underscored by recent legislative actions, such as the 21st Century Cures Act
. This in turn could lead to significant time and cost savings for both the device industry and agency staff.
In Fiscal 2017, CDRH officials were involved in a total of 287 domestic inspections—up from 270 in Fiscal 2016, though down from 301 the year prior, according to BIMO inspection metrics
Challenges in the current selection process for conducting site inspections include lack of resources to inspect all sites involved in a given study that seem to deviate from the norm over factors such as an unusually large number of protocol deviations.
“Current site inspection methods are ad-hoc,” CDRH mathematical statistician, Zhihao Yao, explained at the 11th annual FDA-AdvaMed Medical Devices and Diagnostics Statistics conference. This is because they are “based on manual check or eyeballing on descriptive site-statistics, which is very difficult to do for many sites and variables.” These methods also “lack sound statistical rigor.”
Under the leadership of Ram Tiwari, director at CDRH’s Division of Biostatistics, a group of agency employees are proposing a technical framework, with statistical tests for each variable deemed of value in the site selection process, including site size and number of associated serious adverse events.
The results would then be compiled into matrixes using two different approaches—the Fisher combination approach and the likelihood ratio test (LRT) approach. The tool being piloted as part of the reimagined process is called “BLISS”—which stands for BIMO LRT Inspection Statistical Software—and ultimately provides visual representations of site ranking to facilitate interpretation.
“The tool is being validated through simulated datasets and applied to some real case device clinical studies,” Tiwari told Focus
. This work will have a “direct impact on pre- and postmarket review work.”
Next steps include developing metrics to evaluate BLISS performance. The vision for the new process could eventually involve automation to obtain results at a faster pace, Yao told Focus