What FDA's Data Integrity Remediation Demands Actually Reveal

When FDA issued its July 23, 2025 warning letter to Shiva Analyticals Private Limited, the agency demanded "a complete and independent assessment of documentation systems" and a corrective action plan ensuring records are "attributable, legible, complete, original, accurate, and contemporaneous." That demand was the consequence. But read in reverse, it is also a disclosure: FDA was telling the firm, and anyone else paying attention, exactly what a functioning quality system was supposed to look like before investigators walked in the door.

Most pharmaceutical quality teams spend considerable energy understanding what FDA cites in warning letters. Fewer take the time to work backward through the remediation demands and ask what those demands reveal about FDA's baseline expectations. That second exercise is more useful. The violations describe what went wrong. The remediation asks describe what right looks like.

The Shiva Analyticals case provides a detailed example. Investigators found torn and discarded original CGMP documents in the main waste disposal area, including balance printouts and method verification records. They also identified hundreds of MassLynx audit trail entries tied to privileges including the ability to add, modify, and delete chromatographic peaks and to alter existing files on disk. FDA's response required an independent assessment of the entire documentation system and a corrective action plan built around ALCOA principles. Parsed carefully, that requirement encodes four things FDA expected to find in place: audit trails that could not be quietly altered, privileges that were role-limited rather than unrestricted, original data that moved through formal retention rather than waste disposal, and a quality system capable of detecting those failures without an inspection trigger.

The January 16, 2025 warning letter to Global Calcium Pvt. Limited pushed the remediation demands further and revealed more. FDA found that the firm deleted multiple electronic records during the inspection itself and that management acknowledged the creation of falsified manufacturing records to misrepresent active pharmaceutical ingredient production activities. The required response was comprehensive: an investigation into the full extent of inaccuracies across data records and reporting, covering omissions, alterations, deletions, record destruction, and non-contemporaneous record completion, plus a patient-risk assessment for distributed product and a global corrective action and preventive action strategy for data integrity. Every one of those requirements points toward something FDA expected to find functioning in advance. The investigation requirement implies FDA expected the quality unit to have been actively reviewing data integrity signals, not waiting for an inspector to surface them. The patient-risk assessment requirement implies FDA expected release decisions to rest on verifiably reliable data. The global CAPA requirement implies FDA expected enterprise-level governance rather than site-specific workarounds.

Landy International's June 12, 2024 warning letter illustrates the computer-system dimension of those expectations. FDA found that a standalone gas chromatography system lacked appropriate controls including an audit trail and individual log-in access to prevent deletion of data. Laboratory staff were using a shared password stored in an unsecured drawer. FDA concluded that the firm's quality system did not adequately ensure the accuracy and integrity of data supporting the safety, effectiveness, and quality of drugs manufactured. That conclusion is significant because it shows how FDA connects weak technical controls to the integrity of the quality system as a whole. The expected baseline, made visible by the citation, is individual user authentication on every system that generates quality-critical data, audit trails enabled and configured to prevent unauthorized modification, and access privileges limited to what a given role actually requires.

A. Nelson and Co. Ltd. adds the quality unit dimension. FDA's February 12, 2026 warning letter described manufacturing records and finished product specification forms that had been improperly discarded, including a handwritten sticky note documenting an out-of-trend result that was never formally entered into the quality system. FDA required a comprehensive investigation into data accuracy and a management strategy for ensuring the reliability and completeness of microbiological, analytical, and manufacturing data. The investigation requirement reveals the expectation that the quality unit should be actively protecting inconvenient results from being lost, buried, or detached from formal decision making. An out-of-trend result on a sticky note is not a clerical oversight in FDA's framing. It is evidence that quality unit oversight was insufficient to capture, route, and investigate a signal before it disappeared.

Taken together, the remediation demands across these four warning letters map onto a short set of concrete operating expectations. FDA expects audit trails that are enabled, reviewed routinely, and protected from unauthorized modification. It expects individual user authentication and restricted access privileges on every system that produces or modifies quality-critical data. It expects original raw data to be retained and accessible, not subject to informal destruction. It expects quality units to be actively engaged in surfacing data integrity anomalies rather than passively reviewing completed packages. And it expects the quality unit's oversight to be independent enough to identify problems before external pressure forces the issue.

These are not aspirational standards. They are what FDA is treating as baseline CGMP compliance. When the agency requires an independent retrospective review, a patient-risk assessment, and an enterprise-level CAPA in response to data integrity findings, it is measuring the gap between what it found and what it considered the minimum acceptable operating state. The size of that remediation package is a rough measure of how far below baseline the site was operating.

For small pharmaceutical manufacturers and contract testing laboratories, the practical implications are direct. Sites that still run standalone instruments without audit trails, that allow shared credentials, that tolerate unofficial worksheets or informal records alongside controlled ones, or that classify missing out-of-trend documentation as a training issue rather than a governance failure are operating in territory FDA has already marked for escalation. The warning letters do not describe edge cases. They describe a pattern the agency is seeing repeatedly across geographies, manufacturing types, and firm sizes.

The starting point for any team that wants to pressure-test its position is to read FDA's remediation demands as a checklist and assess whether the quality system would pass each item today, not in response to a citation. Can the site independently reconstruct what happened to any specific data record over the past two years? Does the quality unit have the authority, access, and workflow to detect and escalate an unexplained deletion or undocumented out-of-trend result before it disappears? Are all systems that generate quality-critical data operating with individual user accounts, enabled audit trails, and access privileges tied to defined roles? If the answer to any of those questions is uncertain, the gap is not theoretical. FDA has already told four other firms what it cost them to find out the hard way.