There is a common misunderstanding that 100% visual inspection (VI) of biopharmaceutical products should mean 100% detection of defects. This idea is repeated across geographies and arises within companies and externally (e.g., during internal audits and regulatory reviews).
Yet if you routinely work in VI, you know the process is inherently probabilistic and so, despite all units of a product undergoing full VI, identifying all defects is not guaranteed. Those who do not routinely work in VI but must ‘audit or review the inspection’ may misunderstand the probabilistic nature of VI and share the idea of equating inspection with detection.
“While the intent and goal of an inspection process is to remove all defects, it is not possible to guarantee it with all of the potential variables on a given inspection – regardless of the inspection method (auto, manual, semi-auto, etc.),” said Alan McDonagh, Senior Specialist (VI) at Merck. “A robust qualification process that challenges the inspection method is critical to reduce the likelihood of defect acceptance. Having a qualification process that requires 100% detection of all defects does not make a process more robust, as it may be less sensitive to other qualifications.”
Our new paper, 100% inspection does not mean 100% defect detection, clarifies the misunderstandings associated with VI. It does this through examples and by highlighting that the probabilistic nature of an inspection is recognized in regulatory guidance documents. inspection is recognized in regulatory guidance documents.
Are you considering all the right factors?
Detection capability depends on many factors, including the inspection process, product presentation, and defect type. Attempts to move towards 100% detection may even introduce quality and regulatory risks, even when prioritizing critical defects. Such a move would necessitate using defects with a higher probability of detection being used to train/qualify your people and equipment. This means those defects with a lower probability of detection are not used in training/qualification, making them more likely to be missed during a routine inspection.
This is why our paper talks about the key subjects of regulations, criticality and detectability, probability of detection, qualification vs. detection, and – importantly – why criticality and detectability of defects do not correlate.
“This paper provides an easy-to-understand introduction to the probabilistic nature of VI and can be used as a reference when onboarding new hires and/or stakeholders to VI,” said Antonio Burazer, Global Program Lead for Visual Inspection at Takeda. “It is based on a scientific foundation and topped off with year-long experience from SMEs across the pharmaceutical industry.”
There is an expectation that your manufacturing control strategy must be designed to prevent and detect visible defects in your manufacturing. Also, your limits must be established and monitored to identify any defects within batches – whether they are considered typical and within typical ranges, or atypical or atypical levels.
Our paper will give you confidence in your existing VI capability and its capability to detect different types and presentations of defects. It will allow you to use risk management and remediation in your atypical event investigations. It also helps you understand regulatory requirements and market expectations.
An inspection process that is challenged and qualified with more difficult-to-detect defects (i.e., those with a lower probability of detection) may be more robust than one which claims 100% detection.
What’s your approach?