GLP Final Reports vs Raw Data: What the Lovelace Warning Letter Teaches Sponsors

On 13 October 2026 the United States Food and Drug Administration posted a warning letter to Lovelace Biomedical Research Institute, a contract research organisation in Albuquerque, New Mexico. The finding is one every sponsor and every testing facility should read carefully. The quality assurance unit did not make sure the final study reports matched the raw data. Abnormal animal observations that were recorded during the studies simply did not appear in the reports that would support a drug programme.

This is a Good Laboratory Practice failure under 21 CFR Part 58, not a manufacturing problem. It matters because the sponsor who relies on a nonclinical study is accountable for the integrity of that study, even when a CRO ran it. This article explains what the Lovelace letter found, why the rule it cites is the heart of GLP, why FDA rejected the firm's fix, and what Canadian sponsors and CROs should do now.

What GLP and the QA unit actually require

Good Laboratory Practice is the quality system that governs nonclinical safety studies, and the quality assurance unit is the function that independently checks those studies. It matters because regulators accept a nonclinical safety study only when it was conducted and reported under GLP. What companies should do is treat the QA unit's review of the final report as a hard control, not a formality.

Under 21 CFR Part 58, a testing facility must have a quality assurance unit that is separate from the staff running the study. The QA unit inspects the study, maintains records, and, under 21 CFR 58.35, reviews the final report to assure that it accurately describes the methods and standard operating procedures and that the reported results accurately reflect the raw data. The study director, under 58.33, is responsible for the overall conduct and for the final report, and 58.185 sets out what the final report must contain. Raw data are the records made during the study, and under 58.130 they must be recorded promptly and preserved. The final report is not a summary of what the study director remembers. It is a faithful account of what the raw data show.

The Lovelace findings: what was left out of the final reports

FDA found that abnormal observations recorded during two studies were omitted from the final reports. It matters because those observations are safety signals, and leaving them out changes what the data appear to say. What companies should do is understand the scale of the omission, because it was not a single missed line.

FDA cited 21 CFR 58.35(b)(6), the requirement that the QA unit review the final report so that the reported results accurately reflect the raw data. The agency stated plainly that it is concerned about the integrity of the data from nonclinical studies at the facility. When a regulator questions data integrity, every study that facility has produced is open to doubt.

Why the final report must reflect the raw data

The rule exists because the final report is what reviewers read, and the raw data are the truth it must match. It matters because a decision maker at FDA or Health Canada rarely sees the raw data. They rely on the report. What companies should do is make sure nothing that happened to an animal is lost between the cage card and the report.

Abnormal cageside observations, nasal discharge, abnormal stool, sunken eyes, laboured breathing, are the body telling you the test article may be doing something. In a safety study those signals can shape the no observed adverse effect level, the starting dose for a first in human trial, and the risk language a product carries for its life. Omitting them does not make a product safer. It hides the question. This is the same integrity principle we examined in why 60 to 80 percent of FDA warning letters still cite data integrity, and in the quality unit failures behind FDA's first AI warning letter.

Why FDA rejected the firm's corrective action

The firm revised its procedures and its reports, yet FDA still found the response inadequate. It matters because it shows what a real corrective action looks like. What companies should do is treat new SOPs as the start of a fix, not the end.

FDA said the response fell short because it lacked:

• Enough detail on the updated procedures. The firm updated SOP-1106 on preparing a final report and SOP-1882 on animal records, but did not describe the updates in enough detail for FDA to judge them.
• Training on the new procedures and on GLP. There was no training on the updated SOPs, and none on GLP requirements, including the QA unit's own role and responsibilities. A procedure no one is trained on does not change behaviour.
• Complete correction of the reports. The individual dosing issues were not fully resolved in the revised reports.
• A plan to verify the fix. The firm did not describe any audits to confirm the updated procedures are actually followed. Without effectiveness checks, a correction is a promise, not evidence.

What this means for Canadian sponsors and CROs

Canadian sponsors rely on GLP nonclinical data for submissions to Health Canada and FDA, and they carry the risk if a study's integrity fails. It matters because the Lovelace problem is not specific to one US facility. It is a universal GLP control. What companies should do is qualify and audit their testing facilities, not just trust the final report.

Health Canada expects nonclinical laboratory study data that support drug submissions to adhere to the principles of Good Laboratory Practice, as set out in its guidance on non-clinical laboratory study data supporting drug product applications and submissions. Canada follows the OECD Principles of Good Laboratory Practice, and the Standards Council of Canada runs Canada's GLP recognition programme. International safety studies are planned under ICH M3(R2). The through line is the same across all of these frameworks: the final report must reflect the raw data, and an independent QA unit must confirm it. A Canadian sponsor that files a study with omitted observations inherits the risk, whichever regulator reviews it.

The practical defence is oversight. Treat a CRO like any critical supplier: qualify it, audit it, and verify its QA unit actually reconciles reports to raw data. That is the same discipline we describe in supplier qualification programmes that survive an inspection, and the same accountability lesson as your contract lab's warning letter is your problem.

Reconstruct the study, not just read the report

A GLP study must be reconstructable from its records years after it ends. It matters because a report that cannot be traced back to raw data cannot be defended in an inspection. What companies should do is protect the raw data and the audit trail as carefully as the report itself.

Under 21 CFR Part 58, raw data and specimens must be retained and archived, and a study is expected to be reconstructable from those records long after the final report is signed. The Lovelace case shows why. Because the abnormal observations existed in the raw data, FDA could compare the records to the reports and see what was missing. That traceability is the safeguard. If the raw data had been incomplete or altered, there would have been nothing to reconcile against, and the integrity concern would have been even deeper. For a sponsor, the test is simple: could an inspector pick any result in your final report, follow it back to the moment it was recorded, and find nothing added and nothing lost along the way.

GLP report integrity compliance checklist

Use this checklist whether you run studies in-house or place them with a CRO.

• An independent QA unit is in place, separate from study conduct
• The QA unit reviews every final report against the raw data, not just the summary tables
• All cageside and clinical observations in the raw data are traceable into the final report
• Dosing incidents, deviations and unplanned events are recorded and reported, not dropped
• SOPs for final report preparation and animal records are current and specific
• Staff are trained on those SOPs and on GLP, including the QA unit's role
• Corrections to any report are complete, documented and reconciled to the raw data
• Periodic audits confirm the procedures are actually followed
• Sponsors qualify and audit each testing facility before and during the study
• A data integrity concern at a facility triggers a review of every study it produced for you

Common mistakes that put nonclinical data at risk

• Treating the final report as a summary. Writing what the study director recalls or expects, rather than what the raw data record.
• A QA unit that reviews format, not content. Checking that the report looks complete without reconciling results to the underlying data.
• Dropping inconvenient observations. Leaving out abnormal findings or dosing incidents because they complicate the story.
• New SOPs without training. Updating a procedure and assuming behaviour will follow, with no training record.
• Sponsors who do not audit the CRO. Accepting a final report on trust and discovering the gap only when a regulator does.
• Correcting one report, not the system. Fixing the study FDA named while leaving the same failure mode in every other study.

Frequently asked questions

What is Good Laboratory Practice under 21 CFR Part 58?

It is the FDA quality system for nonclinical laboratory studies, mainly animal safety studies that support drug, device and other product submissions. It sets requirements for the study director, the quality assurance unit, protocols, raw data, and the final report, so that the results can be trusted.

What does 21 CFR 58.35(b)(6) require?

It requires the quality assurance unit to review the final study report to assure that the reported results accurately reflect the raw data of the study. In the Lovelace case, FDA cited this provision because abnormal observations in the raw data were missing from the reports.

Who is responsible when a CRO's report is wrong, the CRO or the sponsor?

Both carry risk. The testing facility is cited directly, as Lovelace was. The sponsor relies on the study and is accountable to its regulator for the integrity of the data it submits. That is why sponsor oversight of a CRO is not optional.

Why do cageside observations matter so much?

Cageside observations such as abnormal stool, laboured breathing or sunken eyes are early signals that a test article may be causing an effect. They can influence the no observed adverse effect level and the safe starting dose for human trials. Omitting them hides a safety question rather than answering it.

Does Good Laboratory Practice apply in Canada?

Yes. Health Canada expects nonclinical study data supporting drug submissions to adhere to GLP principles, Canada follows the OECD Principles of Good Laboratory Practice, and the Standards Council of Canada operates the national GLP recognition programme. The core expectations mirror 21 CFR Part 58.

What makes a GLP corrective action acceptable to FDA?

A credible response identifies the root cause, updates procedures in enough detail to be judged, trains staff on the new procedures and on GLP, corrects every affected report completely, and schedules audits to confirm the change holds. FDA rejected Lovelace's response because several of these were missing.

How MFLRC can help

MF License and Regulatory Consultants helps sponsors and testing facilities in Canada, the United States and Europe build and defend the quality systems behind nonclinical and laboratory data. Our work includes GLP and CRO qualification audits, quality assurance unit reviews and training, raw data to final report reconciliation checks, data integrity assessments, standard operating procedure development, and corrective action and preventive action programmes that stand up to a regulator. We help sponsors put real oversight around the facilities they depend on, before a warning letter does it for them.

If you place studies with a CRO, or run your own testing facility, we can audit the programme and tell you honestly where the gaps are. Explore our audit services, quality control services, and regulatory affairs, licensing and import and export support.

Conclusion

The Lovelace letter is a reminder that data integrity is not only a manufacturing concern. It starts in the laboratory, with the records made at the cage and the report built from them. A nonclinical study is only as trustworthy as the quality assurance unit that reconciled its final report to the raw data. For sponsors, the message is to qualify and audit the facilities you rely on, insist that every observation reaches the report, and demand corrective actions that include training and verification. The alternative is to discover, as a regulator reviews your submission, that the data you relied on left something out.

Sources and references

• FDA Warning Letter, Lovelace Biomedical Research Institute (ref. 27-HFD-45-10-01), issued 6 October 2026
• 21 CFR Part 58, Good Laboratory Practice for Nonclinical Laboratory Studies (eCFR)
• Health Canada, Non-clinical laboratory study data supporting drug product applications and submissions: Adherence to good laboratory practice
• OECD Principles of Good Laboratory Practice
• RAPS, FDA Warning Letters Raise Concerns About CGMP, GLP Violations (15 October 2026)