Sharmista Chatterjee, FDA’s CMC lead for QbD (Quality by Design), explained to participants at the American Association of Pharmaceutical Scientists’ annual conference in San Diego that RTRT is an example of a high-impact model, whereas medium-impact models include those that define design space, and then low-impact models support formulation optimization.
Chatterjee noted that it’s important to indicate clearly where a particular method will be used for batch disposition, and which model will be used for release and which for stability testing.
Gert Thurau of Roche also offered some industry perspective on RTRT, which FDA defines as the ability to evaluate and ensure the quality of in-process or final product is based on process data.
“You need to strike the right balance on method validation without writing a PhD thesis in each dossier,” Thurau said.
Chatterjee added that companies “need a robust and representative reference method.” Model validation is also a two step process: Internal validation using subsets of calibration data and external validation using a naïve data set, she said.
What to Consider
General considerations for implementing high-impact models before the launch of a product include making sure to establish relevant parameters, ensuring the model is appropriately validated and demonstrating the ability of the model to discern OOS (out-of-specification) batches.
Post-launch, Chatterjee suggests companies establish a model monitoring system to determine when it needs updating or revision, monitoring process to determine if it remains within its verified ranges, and including details for model maintenance within the site’s quality system.
A common question from sponsors is: Does FDA require parallel testing after approval (related to NIR (near infrared) calibration model) used for RTRT? And Chatterjee said “not necessarily, but FDA expectation is that applicant provide sufficient data for external validation to demonstrate NIR method is validated at commercial scale.”
As far as regulatory notification of post-approval changes to calibration models, Chatterjee said companies need to consider the impact of the change on the performance of the analytical procedure, and the role of the analytical procedure in the control strategy.
She noted that companies also need to develop and document procedures on how to evaluate and update the model, such as how to deal with OOS test results, and develop criteria for model re-calibration. “Establish the scope of the revalidation” and include “plans for model maintenance/update in the firm’s quality system,” she concluded.