How FDA Cites 21 CFR 211.192: Inadequate Deviation, OOS, and Yield-Discrepancy Investigations

How FDA reads 21 CFR 211.192: inadequate deviation, OOS, and yield-discrepancy investigations

21 CFR 211.192 is one of the most frequently cited CGMP provisions in FDA drug enforcement. It requires that any unexplained discrepancy, or any failure of a batch or any of its components to meet a specification, be thoroughly investigated whether or not the batch has already been distributed, that the investigation extend to other batches of the same product and other products that may have been associated with the failure, and that a written record of the investigation include the conclusions and follow-up. In warning letters and Form 483 observations, FDA cites both outright failures to investigate and investigations that were opened but were superficial, scientifically unsupported, too narrowly scoped, or resolved by retesting rather than root-cause determination.

This overview maps the specific failure patterns FDA has cited, organized by the four problem areas an RA or quality reviewer is most likely to encounter, and then reconstructs FDA's own description of what an adequate investigation looks like from its remediation language.

Out-of-specification (OOS) investigations: the largest cluster

FDA's central position is consistent across letters: a passing retest, a suspected analyst mistake, an assumed contamination event, or a hypothetical glassware or equipment problem does not by itself constitute an assignable cause. A firm must preserve and evaluate the original evidence, document a timely and scientifically sound investigation, establish laboratory error conclusively before invalidating an OOS, and investigate manufacturing causes when laboratory error remains inconclusive ^100101105116.

Root cause asserted without scientific support. FDA repeatedly rejects root-cause conclusions that the firm's own data do not support. Zhuhai United Laboratories invalidated low-assay OOS results after only stating it was possible that sample glassware was not thoroughly cleaned; the investigations found no anomalies, did not identify laboratory error, and lacked scientific justification, yet the firm released the batches on passing retests ¹⁰². Signa SA de CV concluded an impurity OOS was caused by an inadequate reaction, but FDA found the conclusion unsupported and the firm was already using the maximum relevant parameter ¹⁰⁴. Izeen Pharma invalidated confirmed thyroid-API OOS results on a "possible assignable cause" (bound water) without hypothesis testing, supplier evaluation, or an explanation of why prior batches had not required extended incubation ⁶⁸.

Invalidating failing results without conclusive evidence of laboratory error. FDA's rule is that any OOS represents lot quality unless a thorough investigation conclusively supports invalidation, and it is unacceptable to invalidate an OOS without a documented, scientifically supported investigation ¹¹⁶. Cosmaceutical Research Lab assumed a dilution error and retested new sample solutions without testing the original sample or performing hypothesis testing; the firm acknowledged it had "no justification for invalidating the original sample" ¹⁰⁰. Professional Disposables International attributed OOS results to calculation or analytical error "without adequate scientific justification" and released on passing retests ¹²⁰. Apotex Research attributed a low individual assay result to "unknown lab error" and treated it as an outlier, and FDA noted that outlier testing does not apply where the test assesses product variability such as content uniformity ¹¹². Lupin invalidated 134 of 139 initial OOS results over two years, generally attributing them to laboratory error even where the investigations did not clearly establish that laboratory error had occurred ¹⁰⁹.

Retesting into compliance. FDA treats testing repeatedly until a passing result appears as a distinct violation. Cetylite Industries obtained OOS potency results, including by a USP compendial method, then retested with two additional methods until it obtained passing results and released the product; FDA said the firm "repeatedly tested the sample until" it passed, without scientific justification ⁷². Seatex had a practice of testing new samples after a failure until it obtained a specified number of failures or a passing result ¹⁰⁷. Brigham and Women's Hospital resampled and repeated an endotoxin OOS until it passed, under an SOP that allowed OOS endotoxin retesting to continue without limiting the number of retests ¹¹⁴. Missouri Analytical Laboratories used freshly prepared, passing samples to invalidate failing original-preparation retests without ever determining a root cause ¹¹⁹.

Stopping at the laboratory phase. FDA repeatedly states that an inconclusive laboratory investigation must extend to potential manufacturing causes (the "Phase II" production investigation). CTX Lifesciences stopped at the laboratory phase and did not extend root-cause analysis to manufacturing operations ¹⁰¹. Fresenius Kabi Oncology did not proceed to a Phase 2 manufacturing/product-quality investigation after its laboratory hypothesis about heat degradation lacked supporting evidence, and separately concluded contamination was the "most probable cause" without identifying the source or confirming it with a hypothesis study ¹⁰⁵. Claris Injectables invalidated OOS results without a Phase II production investigation and failed to timely address potential manufacturing causes ¹⁰⁶.

Documentation, retention, and timeliness. Zhuhai United Laboratories was cited expressly for failure to "adequately investigate and document" OOS results per procedure ¹⁰². Seatex did not record or retain OOS results for review, which FDA said compromised the quality unit's ability to assure CGMP conformance ¹⁰⁷. Eco Lips conducted no investigation of initial OOS results, retested and released batches even though all assay results for one batch failed, and failed to initiate timely investigations or implement CAPA ¹⁰⁸. Keshava Organics handled numerous OOS events as "incidents" under a procedure that did not require substantive OOS investigation, and acknowledged its OOS decisions lacked sufficient inquiry and scientific rationale ¹¹⁰.

The same deficiencies appear verbatim in Form 483 observations. Aurolife Pharma closed OOS investigations after passing retests despite no identified laboratory error, failed to initiate the required Phase 2 investigation, and in one case could not locate the original sample and released the lot on new-sample results; another investigation opened more than three weeks after testing, by which point the original sample solution had expired ⁴²⁵¹. Mylan Laboratories invalidated OOS results without sound scientific justification, including reports containing contradictory conclusions such as "no laboratory error" followed by a conclusion of laboratory error ⁵⁰⁴⁵. Ohm Laboratories selectively discarded OOS results after unproven explanations (air bubble, glassware contamination, column carryover, a system-generated peak) without testing to prove those causes ⁴⁶. Spectrum Laboratory Products' OOS files lacked supporting documents, analyst interviews, adequate root-cause analysis, product-impact review, and CAPA, and all OOS investigations concluded without CAPA ⁴³. Ipca Laboratories reached Phase IA/IB with no assignable cause, used an inconclusive sample-preparation hypothesis, did not extend to Phase II manufacturing, then retested and invalidated the initial OOS on passing results ⁴⁹.

Deviation and unexplained-discrepancy investigations

For non-laboratory deviations, FDA faults investigations that reach a conclusion without examining the relevant records, samples, or process conditions, and CAPA that does not prevent recurrence.

Superficial or absent investigation. Westwood concluded an SPF 30 lot was "not properly emulsified" without reviewing manufacturing records or processing parameters, assessing other lots, or taking appropriate corrective action, and then reprocessed and released it ⁶². AG Hair's investigation of an irritation complaint consisted only of a four-person "panel test" of its own employees, with no further lot testing ⁶⁹. Several firms did not investigate at all: Hubei Kangzheng did not investigate an API IR spectrum that differed from the standard despite using that API in U.S.-distributed batches ⁶⁰; EyePoint did not investigate an atypically high individual drug-core release rate and released the batch, which later generated two lack-of-effect complaints ⁷³; US Pharmaceuticals did not investigate multiple excessive microbial and TOC failures in water used in distributed OTC products ⁷⁷.

Root cause asserted without support, or contradicted by the firm's own data. Sun Pharmaceutical first attributed an unknown impurity to "dirty glassware," then to another product's contamination, and FDA found neither conclusion scientifically justified because, among other things, the proposed contaminant's retention time did not match the unknown peak ⁶⁷. Strides identified "old reagent" as the cause of an API impurity OOS despite being unable to reproduce the impurity and without evaluating the original failing sample ⁷⁴.

CAPA that was absent, inadequate, or not demonstrably effective. Spartan's investigations lacked appropriate CAPA and its corrective-action procedure lacked provisions for root-cause determination, effectiveness checks, and preventive action ⁵⁹. Takeda's recurring black-particle investigation had no CAPA plan and concluded there was no product impact because particles settled at the vial bottom, while FDA found the firm allowed operations to continue without a risk assessment for released U.S. product ⁶¹. Toyobo concluded it was "highly possible" that washing/sterilization could not remove adhered particles but did not address upstream causes or implement timely CAPA, and a prior inadequate CAPA had not resolved recurring stopper-stain defects ⁷⁶.

Recurrence treated as isolated events. Spartan routinely lacked investigations into recurring water-system microbial failures ⁵⁹. US Pharmaceuticals had recurring excessive microbial levels in purified water dating to 2016 without acting to ensure the water met minimum quality standards ⁷⁷.

The corresponding 483 language is often quoted directly from the regulation. Auro Pharmacies was cited that "written records are not always made of investigations into unexplained discrepancies and the failure of a batch or any of its components to meet specifications," having failed to initiate OOS investigations for environmental-monitoring failures ¹⁵⁶. Reliance Life Sciences was cited that written records "do not always include the conclusions and follow-up," with critical-equipment work orders lacking deviations, quality-impact assessment, or CAPA ¹³⁸. Impax was cited on a repeat basis for failure to "thoroughly review and document unexplained discrepancies," where unknown HPLC peaks had "no deviation report, investigation, or root cause assessment" ¹⁵⁴. California Pharmacy & Compounding Center had no investigation reports for positive personnel and environmental microbial results and no microbial identification ¹⁵³.

Yield and reconciliation discrepancies: the 211.192 / 211.103 nexus

Yield-discrepancy citations sit at the intersection of 211.103 (calculation of actual and theoretical yields) and 211.192 (investigation of discrepancies). FDA's approach is not merely to require the calculation: firms must reconcile expected and actual output, explain discrepancies in contemporaneous batch records, set scientifically supportable yield-deviation limits, and investigate deviations that cross those limits.

Cytosol Laboratories was cited under 211.103 for failing to determine actual yields and percentages of theoretical yield "at the conclusion of each appropriate phase," where batch records listed a theoretical yield and a different number of units filled and "the discrepancy in formulated versus filled volume is not explained in the batch production records" ²². Pocono Coated Products proposed a yield-deviation threshold "as the investigational limit," but FDA found the response inadequate because the firm had not explained the "scientific rationale" for the limit, and required the firm to explain how it would calculate theoretical and actual yield and justify the yield limits "that would trigger an investigation under 21 CFR 211.192" ²⁷.

Form 483 observations show the more common pattern: batches that failed yield limits were released without any investigation. Care-Tech Labs was cited because "No investigation was performed" when multiple batches failed theoretical-yield limits, and the batches were "released and distributed" ⁹⁵. LEESAR "failed to investigate failures of the compounding batch yields," where nonconforming batches (for example, an 89% yield) were approved and released with "No investigations… opened" and no potency testing ⁹⁷. Denver Solutions (Leiters Health) initiated "No deviation" for reconciliation inconsistencies across three Vancomycin for Injection batches, including one with more and two with fewer units than expected ⁸⁶. Cadila Healthcare had capsule exhibit batches with theoretical yields outside established parameters where "A deviation investigation was not initiated and there was no documentation in the batch records" ⁹³. Neeyaan LLC had a capsule batch at 86.47% actual yield with "no documentation" in either the OOS or Deviation logbook, despite the batch being approved and distributed ⁹².

Where firms did investigate, FDA challenged the scope. Aurobindo Pharma identified material loss during processing as a probable OOS root cause and initiated a CAPA to calculate yields at additional stages, but identified 24 similar low-API products and did not extend the corrective action, with "no justification" for not extending stage-wise yield calculations to similar products ⁸⁴. Laboratorios Farmaceuticos Rovi attributed a low-yield/defect failure to inadequate cleaning but did not re-evaluate the adequacy of its visual-inspection approach or reinspect the lots ⁹⁶.

Scope: extending the investigation beyond the first lot

The requirement in 21 CFR 211.192 to extend an investigation to "other batches of the same product and other products" is a recurring standalone citation. FDA faults firms that confine an investigation to the initially identified lot, product, equipment, or apparent cause.

• Same component or material. Luitpold attributed particles to glass delamination in a specific 5 mL glass lot but "failed to extend the investigation to other unexpired lots" made with the same glass; two more lots were later found to contain particles, and FDA required analysis of the effect on marketed lots ¹²³.
• Shared equipment or sterilization cycle. Cadila concluded other products using the same equipment were unaffected after a failed requalification cycle and did not extend CAPA; a later failure showed biological-indicator growth and a batch recall, and FDA requested a retrospective review of all U.S.-distributed, unexpired batches processed with the equipment ¹²². Denison rejected a single lot contaminated with non-food-grade lubricant without considering other drug products, including infant products, made in the same mixing tank over roughly the same period ¹²⁸.
• Retrospective review of invalidated OOS. Kolmar Korea invalidated OOS results without scientific justification and did not extend the investigation to a retrospective evaluation of all OOS results invalidated without justification; FDA required review of invalidated results for U.S.-distributed products with notification or recall as needed ¹²⁴.
• Stability failures and market impact. Laboratorio Magnachem's stability-OOS investigation lacked "sufficient rigor and scope" as to root cause and "the extent of impact to other batches and products," omitting a risk assessment for U.S.-market product; five U.S. batches were recalled ¹²⁵.
• Container-closure and cross-contamination. Cipla recalled six Albuterol batches tied to a defective valve lot, but complaints continued across batches using valve lots beyond the four initially identified, and FDA said the firm failed to conduct a comprehensive risk assessment and determine recall scope ¹³⁴. Nephron's cross-contamination investigation relied on limited retain-sample testing that assumed "not detected" results represented an entire lot, which FDA said was insufficient to determine contamination scope ¹²⁷.

The 483 record mirrors this. Brands International was cited that investigations "did not extend to other batches of the same drug product or other drug products" ¹³⁶. Teva Parenteral Medicines was cited that unexplained-discrepancy investigations of water inside vials did not extend to other batches and products ¹⁵². Zhejiang Huahai did not assess whether equipment-area contamination affected other batches after finding foreign matter in a rejected tablet batch ¹⁴⁸. Zydus Lifesciences released other batches with an unknown impurity peak consistent with cross-contamination to the U.S. market without justification ¹⁴⁹.

What FDA describes as an adequate investigation

FDA's remediation language in 211.192 letters reconstructs a positive standard. An adequate investigation program, in FDA's own framing, does the following:

1. Independently and comprehensively assesses the whole investigation system, not just the individual event, covering deviations, discrepancies, complaints, OOS results, atypical events, and failures, and strengthening scope determination, root-cause evaluation, procedures, and quality-unit oversight so that all phases of investigations are properly conducted ⁷⁹¹⁵.
2. Preserves the complete record and routes all testing data to the quality unit for batch-record review; where an OOS is invalidated, the batch record must contain the investigation, and only a scientifically sound and conclusive investigation can justify excluding an OOS result from the certificate of analysis ³.
3. Does not invalidate an OOS without conclusive scientific evidence of laboratory error. For every invalidated OOS the firm must determine whether the evidence conclusively or inconclusively demonstrates causative laboratory error; unsupported explanations (contamination, analyst error) are inadequate where the firm cannot explain the evidence, the scope, or why other samples in the sequence were unaffected. A firm may not disregard failing results, retest until passing, or use unvalidated methods to obtain passing results ¹⁷¹⁵.
4. Conducts a retrospective, independent review of all invalidated OOS results, including in-process, release, and stability results, often regardless of whether the batch was distributed in the U.S., producing a report addressing each OOS, and examines any statistical-outlier use that may have improperly invalidated results ^1615211.
5. Investigates manufacturing causes when the laboratory cause is inconclusive, through a full production review: batch records, adequacy of manufacturing steps, equipment and facility suitability, raw-material variability, process capability, and deviation, complaint, and batch-failure history ¹⁶¹⁵.
6. Assesses batch disposition, product impact, and market action, determining whether nonconforming product was released and, for distributed product, providing an action plan that addresses product-quality and patient-safety risks including customer notification and recall where warranted ⁸¹⁵¹¹.
7. Implements CAPA that addresses the true root cause and systemic causes, not superficial retraining, and builds in effectiveness checks so the firm verifies that the CAPA actually works ^9162510713.
8. Places the quality unit in active oversight of investigations and disposition decisions, including identifying adverse laboratory-control trends, resolving causes of laboratory variation, and initiating manufacturing investigations whenever a laboratory cause cannot be conclusively identified ¹⁹¹⁶.

In short, FDA's model is not an isolated laboratory retest. It is an independent, evidence-based, fully scoped investigation that determines a scientifically supported root cause, evaluates both laboratory and manufacturing causes, retrospectively reassesses invalidated OOS results, evaluates affected and distributed batches, implements systemic CAPA, and verifies that the CAPA works ¹⁹¹⁵.

Quick reference: recurring failure archetypes