Six Lessons on Keeping Animal Health Study Data Clean and Audit-Ready

Insights from Chantal March, Prelude's Director of Compliance, who has spent 25 years in quality, 14 of them also leading regulatory compliance, on the sponsor side of a first-of-its-kind approval through FDA's Center for Veterinary Medicine. Shared during Prelude's Animal Health Insights webinar series.

‍

A regulator opening your study has one question: can they rebuild what happened from your records? The answer gets decided long before submission. It gets decided when a number moves from a paper form into a spreadsheet, when a tired crew decides it can push through one more hour, and when the rain starts and the pen stops writing.

Chantal March has spent her career on that question. She started as a research associate at Memorial University's Ocean Sciences Centre in Newfoundland, collecting data on a project that became her career for 27 and a half years: AquAdvantage salmon. When the company realized it would need FDA approval as a new animal drug, somebody had to be responsible for quality. In her words: "Tag, you're it, Chantal." She built the quality systems from scratch, then carried them through approval, facility inspections, post-approval work, and a stint in food safety. She learned data capture on a fish farm, which is why so many of her stories involve rain, boats, and wet paper.

In this Animal Health Insights session, Prelude VP of Marketing Ashley Thompson sat down with her for 40 minutes on what clean data means to a regulator and how to keep it clean while a study is running. Six lessons stood out.

1. Clean Data Means You Can Rebuild the Study

Ask Chantal to define clean data and she goes straight to the regulator's test: "the ability to reconstruct what happened with confidence, so that you know that what was approved as a study design was followed through, and the raw data supports the conclusion of the final report at the end."

A dataset can look tidy and still fail that test, if nobody can trace a number in the final report back to the moment it was captured. She also rejects the idea that passing it is inherently hard: "I would argue that it shouldn't be hard to achieve consistently if you have the right processes and controls in place."

The lesson for study teams: judge your data by whether an outsider could rebuild the study from it, and build those processes and controls in from the first day.

2. The Errors Hide in the Hand-Offs

Her biggest data miss came early, when her team was brand new to regulated work. Data went from paper, to Excel, to files exported for statistical analysis. In the rush to get things done, transcription errors crept in, and nobody caught them until the statistics were run and "things just weren't adding up."

The team went back to the paper records, found where it had all gone wrong, and had the statistics redone. It cost at least six months, on a team of fewer than 20 people. Then came the detail that should worry any sponsor: once the errors were fixed, the outcome of the reports was more favorable. The mistakes had been hiding a better result.

The fix was "the proper checks and balances" at every transfer between capture and analysis. If your studies still run from paper to spreadsheet to stats, here is how paper and EDC compare in animal studies.

3. More Haste, Less Speed

Chantal draws a sharp line between being fast and having haste: "more haste, less speed," as she puts it. Haste is what puts mistakes into a form at the moment of capture, which is the cheapest place to prevent them.

Her teams countered it with two habits. Everyone had to understand why the data they were collecting mattered. And observers, study directors, or principal investigators stayed in the room, watching the data forms while the work happened and spotting the staff who were just trying to push through.

She adds a second failure mode: going too long. "If people have been doing the same thing continuously for an extended period of time, mistakes get made," she said, and not only in the data. Around large animals, fatigue becomes a safety problem, too.

4. Measure It While the Study Runs

Scrubbing data at the end tells you what went wrong once it is too late to prevent it. Chantal's in-study checks are simple. Is data entered right the first time, or does someone have to go back and redo it? How much of the expected data is actually complete? Do your database controls flag values that fall outside your acceptance criteria?

Then comes the filter she applies to everything: "not all data is created equal." Identify the critical pieces of information you would need to recreate the study, and hold them to the highest standard: recorded properly, accurate, and consistent with the ALCOA principles. ALCOA is the shorthand FDA uses for data that is attributable, legible, contemporaneous, original, and accurate, and CVM's data quality resources are the place to start for a refresher.

5. Give People the Why, and Ask Them First

"I can never tell people how important they are," Chantal said, and for her the work starts with the why. Hand a study team a requirement with no context and the response is predictable: "Why? Why is it important?"

So explain to the people collecting the data why each step matters. Then go one step further: before every T is crossed and every I dotted, talk to the people who will do the work on the ground. "They will tell you quite quickly" how much trouble a requirement will cause, and whether there is a better way to do it.

That matters most in livestock work, where everyone collecting data already has another job to do. In her words: "Them understanding and buying into the why will always contribute to you having better data."

6. Know What's Critical Before It Rains

On one GLP study, Chantal's team was on a boat collecting conventional salmon when it started to rain, and the pen stopped writing on the wet paper. The fish numbers had been randomized, and euthanasia times had to be recorded. Pencil was never allowed in GLP work, but someone had one, so the team captured the critical information with it, transcribed it in pen back at the sampling site, and kept all the papers together. "The question becomes, what's critical?"

Tablets inherit the same problem, which she used to call "wet, soggy sheets." "You're not going to be stuck in a rainstorm in a clinician's office somewhere," she said. "But in a farm… you could very well."

Field sites also test consistency. The GCP gaps she sees most in livestock are traceability and sample labeling, like a swine team whose samples got mixed up because people tagged them inconsistently. Her test for any SOP: "is it really reflective of how the people on the ground are actually doing those things?"

The Questions Sponsors Are Asking

The second half of the session turned to what comes next: CVM remote access, which Prelude's clients have been asking about, AI, and one question from the audience.

Does CVM remote access change how you build a study?

Yes, starting with how your data will appear to a reviewer. "Having that remote access is not a get-out-of-jail-free card," Chantal said. Decide before the study is built what access CVM will have, which data types reviewers can reach, and how much of your data lives in the EDC at all. Her example was flow cytometry: raw instrument output, without the software that reads it, would mean nothing to a reviewer. She has not been through the program herself, but after years of uploading raw data through FDA's portals, with their size and file-type limits, she calls it "a great idea." Prelude's explainer covers how it works.

Who should consider remote access?

"I think it's something everybody should consider," she said, on one condition: prepare in advance. Settle what reviewers will see, whether access is read-only, and whether data can be exported from the EDC. Her mindset comes from inspections in Canada, which arrived with a couple of weeks' notice. Once, the inspectors asked to come in the middle of a study sampling. "Everybody was like, oh no, oh no. And I was like, oh yes." If you have confidence in your systems and your processes, she said, "showcase it."

What are the challenges of collecting clean data in an EDC?

This one came from the audience. Connectivity comes first: with offline capture, how quickly does data upload, and do the timestamps survive? "What happens if your tablet gets dropped and a cow steps on it?" If it breaks before it syncs, what's the backup? The EDC also prevents problems, by prompting every site for the same required data and flagging entries that can't be right, the way her team's own system flagged a spawning fish entered at 5 grams. "It could have been 5 kilograms, but it was definitely not 5 grams." (Here's how Prelude handles offline capture.)

Where does AI fit in compliance?

As a tool, with limits. "I'm a fan of AI," Chantal said. She uses it in a closed system to help put her thoughts into words, and anything it turns up gets the treatment compliance gives everything: "you need to check and verify, which is kind of all the role of compliance is." She has watched this cycle before, with electronic systems in the early 2000s and then the cloud. AI, in her view, is "just another piece."

What This Means for Your Next Study

Every lesson in this session comes back to decisions made before the first animal is enrolled. Name your critical data, because that is what a reviewer will need to rebuild the study. Count your hand-offs, because each one is a place for errors to hide. Measure from the first entry, because scrubbing at the end only tells you what went wrong. And give the people on the ground the why, because they are holding the pen when the rain starts.

Her closing answer explained why any of it matters. Asked what one thing she would change about how sponsors approach compliance, she said: "compliance isn't a checklist. It's a mindset." Quality data, collected in compliance, supports the products that improve the health and well-being of our animals, "be it our pets or our livestock."

Watch the Full Conversation

This article is based on Prelude's Animal Health Insights webinar featuring Chantal March, Prelude's Director of Compliance, with 25 years in quality and 14 of them also leading regulatory compliance.

Watch the full replay here.

‍