Data Integrity in Analytical Laboratories

Why data integrity matters in pharmaceutical laboratories

Pharmaceutical and chemical analytical laboratories generate the evidence used to assess material and product quality. Within pharmaceutical Quality Control (QC), accurate, complete and reliable records support decisions about compliance with specifications, batch disposition and patient safety. A result loses credibility when the laboratory cannot trace it to the sample, test conditions, original observations and review that produced it [1] [3].

In its 2018 guidance, FDA reported increased findings of data integrity violations during CGMP inspections and described regulatory actions including warning letters and import alerts. These findings explain why the reliability of laboratory records receives close attention during an inspection [1].

Weak quality culture, ineffective leadership, inadequate procedures, limited software knowledge and poorly controlled technology can leave opportunities for errors or misconduct. Deadline pressure can encourage shortcuts when staff do not feel able to report a problem. Management should provide workable processes, training and a culture that encourages prompt reporting and investigation. Technical controls and management oversight need to work together [1] [2].

ALCOA principles and complete laboratory records

In short: A complete record includes original evidence, the full testing history and the metadata needed to review it.

ALCOA describes five qualities of a trustworthy record: attributable, legible, contemporaneous, original and accurate. In a laboratory, that means identifying who performed the work, keeping the record readable, documenting it when the activity occurs, preserving the original evidence or an appropriate true copy, and checking that the record correctly represents the work [1] [2].

ALCOA+ adds complete, consistent, enduring and available. Failed and repeated tests belong in the testing history; timestamps and sequences should make sense; records should remain protected and retrievable throughout retention. A true copy must preserve the meaning and required metadata of the original. Where dynamic functionality is needed to reconstruct or evaluate the analysis, a static printout alone may be insufficient [1] [2].

For US finished-pharmaceutical CGMP, 21 CFR 211.194 requires complete laboratory data, including test methods, instrument graphs and spectra, calculations, results and second-person review. ICH Q7 section 6.6 addresses laboratory control records for active pharmaceutical ingredients. 21 CFR Part 11 governs applicable electronic records and signatures; it does not replace these underlying record requirements. The requirements that apply depend on the regulated activity and jurisdiction [3] [7] [9].

How small shortcuts become data integrity concerns

A shared password, an undocumented trial injection, a deleted analytical run or an unjustified manual integration can remove information needed to assess a result. A shortcut that appears to save a few minutes may leave no dependable account of who did the work or why data changed. Poor understanding of laboratory software can compound this problem [1] [4].

More serious examples include fabricating data, copying results and presenting them as new measurements, or backdating an analysis to misrepresent when it occurred. Weak access controls, missing audit trail reviews and investigation conclusions without supporting evidence can allow such problems to persist. These practices must be investigated on their facts; a discrepancy alone does not establish deliberate misconduct [1] [2].

A failed sequence that must stay visible

Hypothetical sequence

Data manipulation and selective reporting

Testing a sample repeatedly until it passes, while ignoring, excluding or deleting unsuccessful results, creates a misleading picture of the analysis. FDA's laboratory controls guidance specifically addresses unofficial trial injections of product samples. A product sample cannot be used to see whether it will pass and then disappear from the official testing record [4].

During a review or inspection, instrument and server data can be compared with laboratory notebooks, worksheets and final reports, with IT support where necessary. This reconciliation can identify runs that have no corresponding explanation, results copied between records, or gaps between acquisition and reporting. Preserve and investigate discrepancies instead of assuming that a complete-looking report represents the entire analysis [1] [3].

Failed, repeated and invalidated runs must remain traceable, with the reason for invalidation and any investigation retained. Excluding a result from the final calculation requires scientific justification; it does not authorize destruction of the original evidence. Incomplete records can undermine confidence in the laboratory's controls and support regulatory findings [1] [3].

Audit trail review and changes to standard weight

In short: Review critical changes alongside the associated record before final approval.

An audit trail is a secure, computer-generated, time-stamped history that supports reconstruction of record creation, modification and deletion. It can show who changed a critical value, when the change occurred and the recorded reason. Its usefulness depends on the system being appropriately configured, protected and reviewed; software should not be assumed to capture every action merely because an audit trail feature exists [1] [2].

QA personnel, supervisors and designated reviewers need to examine relevant changes and resolve unexplained entries. FDA recommends reviewing trails for changes to critical data with the associated record, before final approval. The review frequency follows the underlying record requirements; where no frequency is specified, a documented risk assessment should guide it. Disabled audit trails or ineffective review leave gaps in oversight, but one missed change does not establish that a trail was disabled [1].

Shared user accounts and individual accountability

When several analysts use a common analyst or administrator login, the system identifies the account rather than the individual who performed an action. A manual integration, data edit or approval under that shared identity may therefore be impossible to attribute reliably to one person. Password sharing undermines the same control [1].

Assign individual accounts and appropriate privileges for creating, changing and approving records. FDA distinguishes these actions from shared access used only to view data: a view-only account may be acceptable, but it cannot establish who performed a required second-person review. The laboratory must preserve individual accountability for that review [1].

Instrument date and time controls

Instrument and computer timestamps establish when testing and changes took place. Changing a system clock to make a sample appear to have been analysed within its approved stability window misrepresents the testing history. The apparent time of analysis can no longer demonstrate that the sample met the applicable timing requirement [1] [2].

Reviewers should assess whether users can alter date and time settings and whether changes require authorization, justification and a traceable record. Restrict access to these settings, control approved adjustments and verify time consistency across connected systems. An authorized clock correction is different from concealed backdating, but both the adjustment and its impact on affected records need to be understandable [1] [2].

User privileges and access rights

Analysts should have the permissions needed for their assigned work. Giving someone both routine testing access and unrestricted administrative control creates opportunities to change processing methods, system configurations or records without adequate oversight. The relevant question is what the role can do and how those actions are controlled [1].

FDA recommends that system administration, including permissions such as deleting files or changing configurations, be assigned independently of responsibility for record content. Routine analyst work should use an appropriately restricted account. Define roles, document authorized users' permissions, review access and ensure consequential changes receive appropriate authorization and review [1].

Human errors and deliberate misconduct

Incomplete or inaccurate documentation, improper sample handling and failure to record an activity at the time can compromise the reliability of laboratory evidence. Such problems may arise from unintentional mistakes; deliberate fabrication or concealment can also occur. Both need investigation, although their causes and corrective actions may differ [1] [2].

Assess training, procedures, workload, supervision and system design alongside the individual event. Correct the record transparently, evaluate affected results and address the cause. A conclusion of human error should be supported by the investigation and should not stop the review of underlying system weaknesses. ALCOA principles provide a practical basis for examining what went wrong [1] [2].

Data handling in LC-MS laboratories

In short: Keep the raw data, processing decisions and carryover assessment traceable to the result.

Liquid chromatography-mass spectrometry (LC-MS) and nuclear magnetic resonance (NMR) can generate complex datasets with several acquisition and processing steps. Their analytical capabilities do not remove the need for a complete record. Review must connect the raw data, processing decisions and reported result [2] [5] [6].

In LC-MS, baseline drift and signal interference can complicate automatic peak integration. Manual integration may be scientifically justified, but changing a baseline or peak boundary simply to obtain a passing value is not acceptable. Preserve the original data, processing settings, justification, revised result and review so that another scientist can assess the effect of the adjustment [1].

Carryover is another concern. Material from a concentrated sample can persist in parts of the injection system, column or MS source and contribute signal in a subsequent blank or sample. Sensitive LC-MS methods can detect small residual amounts, which may inflate a later result.

Blank responses should be assessed against method-specific acceptance criteria; an absolute zero signal is not a universal criterion. Investigate the source, assess affected injections and document corrective action and any justified reanalysis. Ignoring a known interference can make reporting unreliable, but observing carryover alone does not demonstrate manipulation [5].

Data handling in NMR laboratories

In short: Preserve the FID files and processing history so the displayed spectrum can be reconstructed.

Retaining only a processed NMR spectrum can leave the reviewer without the evidence needed to reconstruct processing. Preserve the original free induction decay (FID) files, acquisition metadata, processing parameters and reported spectra for regulated analyses. The raw record allows another scientist to examine how the displayed spectrum was produced [2] [6].

Unusual settings or changes confined to one spectral region warrant examination; they do not by themselves prove an attempt to conceal an impurity. Complete raw data, justified processing and documented review make it possible to distinguish an acceptable correction from a change that distorts the result. These controls preserve confidence in both LC-MS and NMR reporting [1] [2].

Five controls reviewers can test

1. Reconcile the complete testing history. Include failed, aborted, repeated and invalidated runs and check their explanations against the final report.
2. Examine relevant audit trail entries. Resolve unexplained changes to weights, integration parameters, methods, sample identities or sequence order.
3. Check individual attribution. Verify who acquired, changed and reviewed data, and whether their permissions fit their responsibilities.
4. Check chronology. Confirm that timestamps and authorized clock adjustments support the recorded sequence of events and applicable sample stability limits.
5. Reconstruct the reported result. Retain raw files, metadata, processing versions, investigation evidence and review records needed to explain the final value.

A practical release question

Before accepting an analytical result, ask whether an independent reviewer can reconstruct the full path from sample and instrument to reported value, including failed runs, changes and reasons for repeat testing. If the answer is no, resolve the record gap and assess its impact before relying on the result for a quality decision. Use this review framework with the site's approved procedures and applicable requirements.