Trial Injections, System Suitability, and Audit Trails in FDA 483s and Warning Letters

Trial injections, system suitability, and audit trails: how FDA 483s and warning letters treat unreported chromatography data

FDA has not tolerated "trial," "test," "unofficial," or otherwise unreported chromatography injections of product samples. In more than a decade of warning letters and Form 483 observations, investigators have treated sample injections outside the reportable sequence as incomplete laboratory records at best and as testing into compliance at worst. This holds even where a firm's own SOP allowed the practice. FDA's accepted alternative is a validated method with predefined, standard-based system suitability, complete retention of every injection (including aborted and reprocessed data), and quality-unit review of electronic data and audit trails before batch release.

The regulatory basis FDA relies on

The CGMP provisions FDA cites most often for trial-injection findings are:

• 21 CFR 211.194(a) (complete laboratory records). Cited at Wockhardt (2013) ⁵³, Micro Labs (2015) ⁷¹, Hospira SpA (2015) ⁷⁷, Cosmaceutical Research Lab (2020) ⁶⁷, Winder Laboratories (2019) ⁵⁴, and Grupo Quimico (2024) ⁷⁸.
• 21 CFR 211.68(b) (computerized system controls). Cited at Apotex Research (2015) ⁷⁰, Sri Krishna Pharmaceuticals (2016) ⁷⁴, Hospira SpA (2015) ⁷⁷, and Ultra Seal (2022) ¹⁰¹.
• 21 CFR 211.192 (production record review and investigation of discrepancies). Cited at Aspire Pharmaceuticals (2022), where FDA also pointed to 211.22, 211.165, and 211.194 for its position that all product-sample analytical data must be retained and reviewed ⁷².
• 21 CFR 211.160 (scientifically sound laboratory controls), cited at Micro Labs ⁷¹, Grupo Quimico ⁷⁸, and Analytical Food Laboratories (2024) ²³⁵.
• 21 CFR 211.22 (quality unit responsibilities), cited at IDT Australia (2018), where analysts did not print test injections and the quality unit made disposition decisions on incomplete data ⁸².

The interpretive anchor is FDA's final guidance Data Integrity and Compliance With Drug CGMP: Questions and Answers. It prohibits "sampling and testing with the goal of achieving a specific result or to overcome an unacceptable result," which FDA calls testing into compliance. It also states that using an actual sample in "test, prep, or equilibration" runs to disguise testing into compliance is a violative practice ¹². The same guidance requires CGMP records to contain all data, "including obvious errors and failing, passing, and suspect data." Data may be excluded from a conformance decision only after a documented, scientifically sound investigation ¹².

How FDA has characterized trial and test injections

Common patterns in warning letters

The warning letters describe the same practice under many names and storage locations. The common thread is sample data generated outside the official sequence and kept out of the record the quality unit reviews.

Form 483 observations

Form 483s show the same issues at the observation stage, often with detailed lot-level examples:

• Impax Laboratories (March 2012). Investigators documented repeated "trial" and "official" injections with no evidence the same sample was used ¹⁰³. For one acarbose lot, three trial injections were followed by an official injection, with impurity profiles that differed across all four ¹⁰⁴. A QC supervisor said the highest-weight dantrolene capsule was chosen for a trial "to see 'if it would cause us problems'" ¹⁰³.
• Morton Grove Pharmaceuticals (March 2014). The HPLC SOP allowed trial injections but did not say whether they were standards or samples, what purpose they served, or how the quality unit would review them. It also did not require audit-trail review for them ¹¹¹. Trial chromatograms could not be traced to documented sample preparations, and common analyst logins allowed raw-data deletion ¹⁰².
• Akorn (August 2018). The firm's investigation looked only at injections named "trial." FDA's limited review found many other anomalously named injections the firm had not reviewed ("Trail," "test," "S1," "Sample-Assay-1" through "-6," and others). FDA concluded there was limited assurance in data submitted to the agency ¹⁰⁶.
• Fujifilm Diosynth Biotechnologies Texas (March 2025). Audit-trail reviews left out in-process results and "Trial Injections" datasets ²⁰³.

Lessons from these cases

• An SOP that allows trial injections is not a defense. At Wockhardt, the SOP permitted sample injections to equilibrate the system. FDA noted that neither ICH Q2(R) nor USP <1058> tells firms to perform trial injections for a validated method ⁵³. Morton Grove's SOP was itself part of the observation ¹¹¹.
• Firms have to prove what was injected. Winder said its trial injections used diluent, mobile phase, or standards. FDA did not accept this because the firm offered no evidence that samples had not been injected ⁵⁴. Aspire said its SOP permitted trial injections of standards only. Its own investigation, along with audit trails showing standalone injections outside the sequence, showed that test samples had been used ²¹⁶.
• Keyword-only investigations fall short. Akorn's search for the word "trial" missed differently named injections ¹⁰⁶. Apotex was criticized for not extending its investigation to electronic systems across its laboratories ⁷⁰.

System suitability: what FDA rejected and what it accepts

Practices FDA has rejected

• Actual batch samples used as system suitability, equilibration, or "prep" injections. At International Trading Pharm Lab (2020), FDA said an actual sample from the batch under test may not be used in test, preparation, or equilibration runs. It required a procedure that prohibits trial injections and allows only qualified reference standards to verify equipment suitability ⁴⁹. Grupo Quimico analysts routinely used samples for system-suitability injections ⁷⁸.
• "Readiness" injections as a substitute for validated suitability criteria. Hospira said its "Test" folders were used to equilibrate columns. FDA rejected this, noting that validated methods must contain specific suitability procedures. It pointed to USP <621> replicate injections of a standard preparation ²¹⁹.
• Extra suitability injections followed by deletion. A Chongqing Pharma Research Institute analyst said it was practice to perform extra system-suitability injections and delete undesirable results ⁸⁵.
• Inadequate suitability designs. FDA found a single standard injection unacceptable for HPLC system suitability (Jilin Shulan, 2010) ⁵⁷. It criticized standard injections run only at the end of a sequence because they do not show the system worked before samples were analyzed ⁵¹. At Analytical Food Laboratories (2024), the firm checked whether a peak appeared in the expected retention window and did not show this was equivalent to USP <621> ²³⁵.

Practices FDA has described as acceptable

• Standard-based replicate injections under a written procedure. The Data Integrity Q&A quotes USP: system suitability includes replicate injections of a standard preparation or other standard solutions. It also says the procedure should identify the preparation injected and the rationale for choosing it ¹².
• A characterized secondary standard from a different batch. If an actual product sample is used, FDA says it should be a properly characterized secondary standard, used under written procedures, and taken from a batch different from the one being tested ¹².
• Suitability established before samples and maintained across the sequence. System suitability should be built into the method with acceptance limits and should show the system is adequate before samples are analyzed ⁵¹. For long sequences, FDA expects suitability testing that shows the system stays suitable throughout the run ⁶⁰. Parameters FDA has named include resolution, peak symmetry or tailing factor, theoretical plates, and replicate-standard reproducibility with an acceptance criterion ⁶²⁶⁴. FDA's PET CGMP guidance recommends at least one injection of a reference or internal standard before test samples ⁷.
• Suitability failures handled as incidents. Form 483s have cited firms that invalidated sequences for suitability failures without first calculating sample results to check for OOS (Divi's Laboratories, 2017) ¹⁸¹. Firms were also cited for not documenting SST failures through laboratory incident forms (MSN Life Sciences, 2024) ¹⁸² and for having worksheet acceptance criteria that did not match the method (Dr. Reddy's, 2021) ¹⁸⁵. After a power interruption at Shilpa Medicare (2020), the firm reinjected only the affected sample and standard, and FDA noted that no new system suitability or bracketing standard was run ²²¹²²⁶.

Aborted, interrupted, and reprocessed runs

FDA's guidance treats aborted and reprocessed data as part of the complete record. Chromatographic data should be saved at completion of each injection, including "finished, incomplete, or aborted injections," and "aborted or incomplete injections should be captured in audit trails and should be investigated and justified" ²⁷. Saving only the final result of reprocessed chromatography is not acceptable. Each result should be retained for review ¹². The OOS guidance applies the same logic: all original data must be kept, and without an identified laboratory error there is no basis to invalidate an initial OOS result in favor of a passing retest ³⁸.

Enforcement has followed this position:

• Chongqing Lummy (2016) aborted HPLC stability runs so that partial results would be deleted automatically, reset the PC clock, and reported only the reinjection ⁷⁹.
• Ipca Laboratories (2016) aborted a GC injection without justification, and it was deleted automatically on reinjection ⁷³⁸¹.
• Intas (2023) aborted hundreds of QC sequences over nearly three years. FDA faulted the lack of trending and CAPA even though each incident had been investigated ²³⁶.
• Form 483s at Aurobindo Unit IX (2022) and Biocon Biologics (2022) cited interrupted sequences that were logged as "no chromatogram" or "Data Incomplete" without adequate investigation. At Biocon, staff did not know the software could show whether the sample had actually run ¹⁸⁴¹⁹⁹.
• Regeneron Ireland (2023) did not record integration attempts. Its manual-integration comments were generic, for example 1,185 CE-SDS results with the same justification ¹⁹⁶.

Audit-trail and access-control expectations

What the guidance says

• The people responsible for reviewing a CGMP record should review the associated audit trails along with the rest of the record. Because 211.22 requires data review before batch release, audit trails should be reviewed at that same frequency. Where no frequency is set, firms should use a risk-based frequency ³¹.
• For an HPLC run, the audit trail should capture the user name, run date and time, integration parameters, and details and justification of any reprocessing ²⁸.
• Chromatographic records are dynamic because baselines and integration can be changed. A static printout therefore does not preserve the complete original record ¹⁵³³⁷. Laboratory records are subject to second-person review ³⁷.

What investigators cite

Warning letters have cited audit trails that were disabled on HPLC systems (Aarti Drugs, Zhejiang Hisun, Qinhuangdao Zizhu, Grace Analytical) ^88638984. Hisun ran 80 injections with the audit trail off and then re-enabled it before repeating the analyses ⁶³. Letters have also cited audit trails that were never reviewed or were reviewed only at random ⁹⁰⁹⁵⁹⁶, shared logins ^86889875, and analysts with administrator rights who could delete data or switch off audit trails ⁷⁴⁹³⁹⁹. FDA told Sekisui Medical that upgrading software and enabling features would not be enough unless the quality unit actually reviewed data and audit trails during batch release ⁹⁴.

Recent Form 483s focus on review practice rather than configuration:

• Staska Pharmaceuticals (2026): QC reviewed printed reports only, not electronic chromatograms, manual integrations, deleted or repeated injections, or audit trails ¹⁹⁴.
• Asteria Health (2025): the laboratory manager reviewed printouts rather than raw electronic data and ran potency testing under an unrestricted "Admin" profile ¹⁹⁷.
• Amman Pharmaceutical Industries (2023) and Biotika (2018): SOPs required audit-trail review without saying how to do it or document it ²⁰⁰²⁰¹.
• Bio-Thera (2025) and Fujifilm Diosynth Texas (2025): permissions let analysts see area counts and results before deciding whether to save or reintegrate a chromatogram ²⁰⁷²⁰³.
• Gland Chemicals (2025): external method-transfer users held analyst-level Empower accounts that stayed active after the transfer ended ¹⁹⁵.

What FDA asks for in remediation

When trial injections or related data-integrity lapses lead to a warning letter, FDA's requests follow a consistent pattern:

1. A full, retrospective review of trial data. Micro Labs had to review all trial data, standards as well as samples, to find and investigate every OOS result. It also had to build a chronology from Chromeleon audit trails of all single manual injections that replaced HPLC data ⁷¹⁷⁶. Sun Pharma had to classify HPLC and GC file types, investigate why analysts hid or deleted runs, and justify each invalidation of a failing result ⁵².
2. An investigation of the extent of inaccurate records, with a documented protocol, the systems and sites covered, and interviews of current and former staff, preferably by a qualified third party ^85122123.
3. A product-quality and patient-risk assessment of lots released on potentially affected data ⁸¹¹³⁹.
4. A management strategy and CAPA plan that covers the reliability and completeness of all data. The plan should state whether the individuals responsible can still influence CGMP or application data ¹³⁰¹²⁷.
5. Specific laboratory controls: a procedure prohibiting trial injections that allows only qualified reference standards for suitability checks ⁴⁹, procedures for reinjections ⁷⁷, defined user roles and restricted privileges ⁸⁶⁸⁹, and implemented audit-trail review ⁸⁴. Intas was also asked for an independent retrospective review of three years of invalidated OOS results ²³⁶.
6. Ongoing independent verification, such as consultant audits for at least two years to assess CAPA effectiveness. FDA has stressed that hiring a consultant does not relieve management of its CGMP responsibilities ¹²⁶¹³¹.

Practical implications for QC laboratories

Based on the enforcement record and guidance above:

• Do not allow sample injections outside the defined, method-driven sequence. If an SOP allows any pre-run injections, limit them to blanks, mobile phase, or qualified standards. Record them in the sequence, and make sure audit trails can prove what was injected ^4954216.
• Base system suitability on replicate injections of reference standards, with predefined acceptance criteria, run before samples, and bracketed across long sequences. Use a product sample only as a characterized secondary standard from a different batch ¹²⁵¹⁶⁰.
• Treat every aborted, interrupted, or failed-suitability injection as data that must be documented, investigated, and trended ^27182236.
• Review electronic data and audit trails, not printouts, before batch release. Cover single injections, aborted runs, reprocessing, and deleted result sets ^3178194.
• When investigating, search by audit-trail event type and injection metadata, not only by injection names ¹⁰⁶.

Topics a reader may want to explore further include how FDA has handled trial injections in ANDA pre-approval inspections, how investigators treat method-development or R&D folders on QC systems, and whether expectations differ for biologics chromatography platforms such as CE-SDS.