In the 1970s, one of the largest toxicology laboratories in the United States was found to have been fabricating safety data on an industrial scale.
Industrial Bio-Test Laboratories ran studies underpinning regulatory approvals for pesticides, industrial chemicals and pharmaceuticals. When an FDA inspector noticed the raw data didn’t match the reports, it triggered an audit of everything the laboratory had ever produced. A substantial proportion was invalid. Manufacturers had to go back and repeat work regulators had accepted years earlier. Congressional hearings followed, and so did criminal convictions.
Here’s the part that mattered for everyone else. The science at IBT was bad, but that wasn’t the real problem. The problem was that nobody reading the final reports could have known. A study report is a summary written by the people who did the work. If the underlying records don’t exist, or don’t match, or were never kept, the report looks exactly the same as an honest one.
Good Laboratory Practice was the response, and it explains a great deal about how regulatory studies are run today. For those who have wondered what GLP actually means, we’ve written about what it guarantees, what it doesn’t, and why it’s the reason a regulator in one country will accept data generated in another.
What GLP actually is
GLP is a quality system covering how non-clinical safety studies are planned, performed, monitored, recorded, archived and reported. Notably, that list doesn’t include anything about the scientific method itself.
This is the part worth understanding. GLP does not specify how good your assay is, whether your model is appropriate, or whether your endpoint is scientifically meaningful. Those questions are answered by the test guideline. What GLP guarantees is something narrower and, it turned out, more important: that if a regulator wants to reconstruct exactly what happened in your study years after the fact, they can.
Every mechanism in the system serves that one objective:
A named study director. One identifiable person holds responsibility for the study and its final report. Not a department, not a committee, not a signature block. When something is wrong, there is no question about who is accountable for it, and that person has to put their name to the report knowing that.
An independent quality assurance unit. QA inspects the work while it’s in progress and audits the final report against the raw data, line by line, checking that the report says what the records say. The independence is the whole point. QA reports to test facility management rather than to the study director or the team being audited, so nobody is in a position to sign off their own work. At IBT there was nothing playing this role, which is why the reports and the records could drift apart unnoticed for years.
Raw data, retained and archived. Raw data means the original observations, not the processed version. The instrument printout, the dated worksheet, the observation recorded at the time by the person who made it. Corrections are made so the original entry remains legible, with who changed it and why. The report is a summary of this material, and the material outlives the report. If a plate reader produced a number, that number survives whatever the report goes on to say about it.
Standard operating procedures. Written, versioned, approved procedures for how work is done, so a method is executed the same way regardless of who runs it or when. This is partly about consistency and partly about traceability: if a result looks odd three years later, you can establish exactly which version of the method was in force on the day.
A study plan, signed before work starts. The method, the test item, the controls, the acceptance criteria and the classification logic are all fixed in advance. Planned changes are made as formal amendments, signed by the study director. Unplanned departures are recorded as deviations. Either way the change is documented rather than absorbed silently, which is what prevents a protocol being quietly adjusted once results start appearing. That kind of drift is a much more common failure than outright fabrication, and considerably harder to spot.
Archiving. Everything, the plan, the raw data, the report, the correspondence, goes into a controlled archive with a named archivist and a defined retention period. Regulators can and do ask to see it.
Why it matters commercially
The practical value to anyone commissioning a study is the OECD’s Mutual Acceptance of Data arrangement. A GLP study conducted to an OECD test guideline in one participating country must be accepted by regulators in all the others.
That is genuinely significant. It means one study rather than one per market. Without it, the same test gets repeated for each jurisdiction, which is expensive, slow, and, when animal methods were the norm, meant the same test conducted on animals repeatedly for no scientific reason at all.
MAD is the reason GLP compliance is worth the overhead. The audit trail is what makes a regulator in another country willing to accept data they had no part in generating.
What it comes down to
GLP is essentially an answer to a question that turned out to matter enormously in 1976: how do you know?
Not how do you know the result is right, but how do you know the study happened the way the report says it did. It’s less a scientific standard than a paper trail rigorous enough that fabrication becomes very difficult and honest error becomes traceable.
For a regulator accepting data they didn’t generate, from a laboratory they’ve never visited, in a country they don’t regulate, that trail is the whole basis of trust.
If you’re working out whether your submission needs GLP data or whether non-regulatory testing will do, get in touch and we’ll talk it through.
Email info@x-cellr8.com or call +44 (0)1925 607 134.
———
Contact us if you’d like to learn how you can support our mission, become a corporate sponsor or donate to our crowdfunder.
Email info@x-cellr8.com
Call +44 (0)1925 607 134.
Follow Us!