Pharma Fundamentals: Digital Pathology in Drug Development: Quality Control and Validation of Pathology Data

In this episode of PharmTech’s Pharma Fundamentals series, Dr. Aleksandra Zuraw, senior veterinary pathologist at Charles River, discusses how chain of custody, good laboratory practice (GLP), quality control (QC), and validation apply when digital pathology data support drug development.

Automated systems reduce human error in sample tracking, says Zuraw, because computers do not tire or lose focus, though they require upfront infrastructure investment. Chain-of-custody problems can occur under either approach, and every error must be investigated and corrected. The difference is how quickly errors are found. With automated tracking, teams can query the system instead of reconstructing events from scratch. Faster detection improves the odds that a mistake can be absorbed within existing study timelines rather than extending them.

Tracking and Chain of Custody for Digital Pathology Data

Zuraw explains that chain-of-custody rigor is the same in discovery and preclinical work as in regulated studies. The same processes can be applied across both; what changes is regulatory oversight and the level of documentation required. Pre-regulatory work is also where the organization is most open to new methods. It can serve as a sandbox for digital pathology, where teams can learn where systems fail before bringing them into the GLP environment. Artificial intelligence (AI) algorithms can be qualified in pre-regulatory work, and only those with a clear use case may move into validated environments, because migrating everything would waste effort.

Quality Control and Validation Requirements

Automated QC, she says, plays the same role as manual QC: the new method must perform at least as well as the old one, while adding scale, speed, and freeing staff for other work. For digital slides, automated QC checks whether the correct tissue is present, whether there is enough of it, and whether the image is in focus, the last being a concern specific to digital workflows. The underlying question is whether an image is good enough for a pathologist to interpret.

QC does not drive dose selection; scientific evaluation does. A poor scan would not lead to a wrong dose, but it could lead to an unnecessary read before a pathologist sends the slide back to the lab. Zuraw highlights tissue presence as a high-leverage check, because toxicology studies require specific tissues, some of them small and easily missed. Catching gaps at scanning, rather than after a pathologist has started reading, prevents recut requests that threaten timelines.

Sponsors should validate automated QC systems against their intended purpose using standard computerized-system frameworks, stresses Zuraw, with tissue experts on the team because failure modes are tissue-specific. Program-level validation follows the same principles as single-algorithm validation but involves far greater scale, spanning multiple systems, interfaces, sites, and partner organizations.

Regulators, Zuraw notes, are signaling their expectations through published guidance on whole slide images, AI, and digital peer review, and through direct engagement. Since validating its primary-read systems in May 2023, Charles River has supported more than 300 GLP-compliant toxicology studies submitted to regulators. Auditors have asked how images and labels are protected from alteration and about cloud integrity, and the company answered with test scripts and documentation.

Click the video above to watch the episode.

Click here to watch part 1.

About the Speaker

Dr. Aleksandra Zuraw is a board-certified senior toxicologic pathologist at Charles River Laboratories and a long-time leader in digital pathology and image analysis. She has worked extensively with AI tools and digital workflows since the early stages of her career, contributing to the development and validation of image analysis solutions across several organizations. Dr. Zuraw is the coauthor of several peer-reviewed publications on digital pathology and image analysis, and the author of the book "Digital Pathology 101: Everything you need to know to start and continue your digital pathology journey." Originally from Poland, she completed her pathology training in Berlin, Germany, and has worked across six countries. She also founded the online educational platform Digital Pathology Place, where she creates content and training materials that help bridge the gap between digital pathology innovation and practical application. Through her global perspective and commitment to education, Dr. Zuraw continues to shape the integration of AI and digital pathology in veterinary pathology.