Recalls and Field Alerts: Classification, Execution, and Effectiveness

A recall is the most visible thing a quality system ever does. Everything else happens behind the firewall of the site. A recall puts your decisions in front of distributors, pharmacies, hospitals, patients, and a regulator who is watching how fast you move and how well you account for every unit. By the time you are running one, the defect already exists. What you control is speed, completeness, and proof. This page covers how recalls and related field actions are classified, decided, executed, reconciled, and verified, for drugs and biologics under US FDA rules, with the cross-references to complaints, deviations, and CAPA that make the system actually work.

The vocabulary matters and people get it wrong in interviews. A recall is a firm’s removal or correction of a marketed product that the FDA considers to be in violation of the laws it administers. A market withdrawal is removal or correction of a product for a minor issue that would not be subject to legal action by FDA, or no violation at all (for example, normal stock rotation, or a tamper that is not a manufacturing defect). A stock recovery is retrieval of product that has not yet left the firm’s control. A field alert report (FAR) is a specific NDA/ANDA holder notification to FDA, separate from a recall, that can precede or accompany one. If you make a combination product (a drug plus a device), the device constituent carries a separate vocabulary: a device correction or removal is governed by 21 CFR Part 806 and a device “recall” is defined under 21 CFR Part 7 and Part 810. Use the right word for the right situation.

The four terms above get blurred together in practice, and the difference is not academic: it decides whether FDA legal action is a live exposure, who makes the call, and what triggers a report. Lined up side by side:

Read the table by two questions: is the product violative, and has it already left the firm’s control? Yes to the first makes it a recall. No to the first moves it to a market withdrawal. Product still in-house moves it to a stock recovery. Correction is a mechanism, not a category of its own; it can execute a recall or a market withdrawal depending on what triggered it.

The regulatory basis

21 CFR Part 7, Subpart C (Recalls)

The governing framework for FDA recalls is 21 CFR Part 7, Subpart C, “Recalls (Including Product Corrections), Guidelines on Policy, Procedures, and Industry Responsibilities.” It defines recall, market withdrawal, stock recovery, correction, and recall classification (Class I/II/III). It describes both FDA-requested recalls and voluntary recalls, the recall strategy, the elements of effectiveness checks, recall communications, and recall termination.

Key definitions to know cold, from 21 CFR 7.3:

Recall means a firm’s removal or correction of a marketed product that the Food and Drug Administration considers to be in violation of the laws it administers and against which the agency would initiate legal action, e.g., seizure. Recall does not include a market withdrawal or a stock recovery.

Correction means repair, modification, adjustment, relabeling, destruction, or inspection (including patient monitoring) of a product without its physical removal to some other location.

Most drug recalls are voluntary, initiated by the firm, sometimes after a conversation with FDA. FDA gained mandatory recall authority for drugs only in narrow circumstances; the general drug recall remains a voluntary action in the Part 7 sense, which does not mean optional. If a firm declines to recall a violative product, FDA can pursue seizure, injunction, and public warning. For certain product categories FDA holds explicit mandatory recall authority (for example, medical devices under section 518(e) of the Food, Drug, and Cosmetic Act, and human cells/tissues, infant formula, tobacco, and food under FSMA). Know which authority applies to your product.

21 CFR 314.81(b)(1) (NDA Field Alert Report)

Field alert reports for finished drug products are required by 21 CFR 314.81(b)(1), “Other postmarketing reports,” which applies to applicants holding an approved NDA or ANDA. The FAR is a fast early-warning mechanism. It is triggered by information concerning any incident that causes the drug product or its labeling to be mistaken for, or applied to, another article, and by information concerning any bacteriological contamination, any significant chemical, physical, or other change or deterioration in the distributed drug product, or any failure of one or more distributed batches to meet the specification established for it in the application.

The FAR is due to the local FDA district office within 3 working days of receipt of the information. It is a notification, not a conclusion. You file on a credible signal, not after the investigation closes.

21 CFR 600.14 (Biological Product Deviation Reports)

For licensed biologics, 21 CFR 600.14, “Reporting of biological product deviations by licensed manufacturers,” requires reporting of events that represent a deviation from cGMP, applicable regulations, standards, or established specifications that may affect safety, purity, or potency, for distributed product. The biological product deviation report (BPDR) is due within 45 calendar days of discovery. A parallel requirement for unlicensed registered manufacturers sits at 21 CFR 606.171 for blood and blood components.

Drug shortage notification, 21 CFR 314.81(b)(3)(iii)

A recall or a manufacturing interruption can trigger a separate obligation: notification of a permanent discontinuance or interruption in manufacturing of certain drugs under section 506C of the FD&C Act, implemented at 21 CFR 314.81(b)(3)(iii). If pulling product creates a shortage of a medically necessary drug, that notification clock runs in parallel and is frequently forgotten.

21 CFR Part 806 and Part 810 (the device constituent of a combination product)

A combination product pairs a drug or biologic constituent with a device constituent, for example a prefilled syringe, an autoinjector, or an on-body infusor. The drug or biologic constituent stays on the vocabulary and reporting path described above: Part 7 recall concepts, the FAR at 314.81(b)(1), and the BPDR at 600.14 where a licensed biologic is involved. The device constituent carries its own separate reporting and recall framework, and a defect in a combination product can trigger one path, the other, or both at once.

Device correction or removal, 21 CFR Part 806. Under 21 CFR 806.10, a device manufacturer or importer must submit a written report to FDA of any correction or removal of a device initiated to reduce a risk to health the device poses, or to remedy a violation of the FD&C Act caused by the device that may present a risk to health. The report is due within 10 working days of initiating the correction or removal, and extending the same correction or removal to additional lots is reported as an amendment within 10 working days of that extension decision. Where a correction or removal does not meet that reporting threshold, 21 CFR 806.20 still requires the manufacturer or importer to keep a record of it: the device identification, the event that gave rise to the action, the corrective or removal action taken, and a documented justification for not reporting it to FDA, reviewed by a designated person, retained for 2 years beyond the device’s expected life.

Device recall authority, 21 CFR Part 7 and Part 810. The general Part 7 recall framework applies to devices as it does to drugs and biologics, and 21 CFR Part 810 sets out FDA’s mandatory device recall authority under section 518(e) of the FD&C Act, the narrow circumstance where FDA can order a device recall (a cease distribution and notification order) rather than relying on the firm to act voluntarily, after giving the firm an opportunity to consult with the agency. Most device corrections and removals remain firm-initiated; Part 810 is the exception, not the default path.

Dual reporting. Where the defect sits in the device constituent, for example a delivery mechanism that under-doses or fails to activate, the correction or removal is assessed under Part 806 alongside whatever drug or biologic reporting the same event also triggers. A device malfunction that changes how much drug reaches the patient is not only a device problem; it is also a question about the delivered dose of the drug or biologic, which can independently trigger a FAR or BPDR assessment and a recall decision on the drug side. Regulatory affairs assesses both branches on every combination product event; treating the Part 806 clock and the drug-side clock as alternatives instead of parallel obligations is a common gap. See combination products and cGMP under Part 4 for how the base cGMP system is structured, and the complaint reportability decision matrix for how a single combination product complaint routes down multiple branches.

A worked example makes the split concrete. An autoinjector combination product has two possible failure modes. In the first, spring-mechanism wear causes a subset of units to deliver an incomplete dose while the drug fill itself meets every release specification: this is a device-constituent defect, reported under 806.10 within 10 working days of initiating the correction, with a Part 806 record built around the malfunction rate and the affected device lot codes; whether it also triggers a FAR or BPDR turns on whether the incomplete delivery counts as a failure of the drug product to be delivered at its established dose, a call QA and regulatory affairs make together, not a device-only decision. In the second, the drug fill volume in the same autoinjector platform is found low across a manufacturing lot with the device mechanism unaffected: this is a drug-constituent defect, assessed under 314.81(b)(1) or 600.14 as appropriate, with the Part 806 framework not in play because the device performed as designed. The two scenarios use identical hardware and produce two different reporting paths, which is exactly why the constituent the defect lives in, not the product family, decides which regulation applies.

A frequent inspection finding is a combination product recall record built entirely on the drug side with no Part 806 assessment on file, or the reverse: a device correction and removal report filed with no assessment of whether the same event also affects the delivered dose of the drug or biologic.

Other corners to know

• EU GMP: EudraLex Volume 4, Part I, Chapter 8 (“Complaints, Quality Defects and Product Recalls”) sets the EU expectation, supported by the GMDP rapid alert system and reporting to national competent authorities. The expectation that recalls can be initiated promptly and that a recall procedure is regularly evaluated by mock recall is explicit there.
• WHO publishes recall guidance in its GMP and distribution texts for markets that follow WHO standards.
• DSCSA (21 CFR and the Drug Supply Chain Security Act) changes how you trace and locate product at the unit and saleable-return level, which directly affects how fast you can find distributed lots in a recall.

Cross-link: see 21 CFR 210/211 walkthrough for the cGMP backbone, and what is GMP for the framing.

Recall classification: Class I, II, III

FDA, not the firm, assigns the recall class once the recall is reported, but you must be able to recommend a class and defend it. Classification is a health-hazard judgment about the consequences of using or being exposed to the violative product. It drives the recall depth, the urgency, the communication, and the effectiveness check coverage.

The mechanism FDA uses is a Health Hazard Evaluation (HHE), performed by an FDA committee, that weighs whether the population already exposed is at risk, the seriousness of the hazard, the probability of occurrence, the consequences (immediate or long-range), and vulnerable subpopulations (children, the elderly, the immunocompromised). Your internal HHE should mirror those factors so your recommended class survives FDA’s review.

How to reason about class, with examples

A worked example. A sterile ophthalmic solution batch shows a failed container-closure integrity result on a small number of stability units after distribution. Sterile route, eye, potential for serious eye infection, vulnerable use. That points to Class I or high Class II. Compare a film-coated tablet for a chronic non-critical condition that shows a single dissolution result slightly out of specification at the 12-month station with the drug still demonstrably absorbed. Reversible, limited clinical impact, that points to Class II or III. The driver is always clinical consequence to the exposed patient, not how embarrassing the defect is.

A second example with a numeric flavor. Suppose assay drifts and a batch of a narrow-therapeutic-index drug is found at 118 percent of label, where the specification is 90.0 to 110.0 percent. Superpotency of a narrow-therapeutic-index drug is a dosing hazard, which pushes toward Class I or strong Class II depending on the margin and the indication. The same 118 percent on a wide-margin vitamin is a different conversation. The number alone does not set the class; the number plus the clinical margin does.

Recall depth: how far down the chain

Classification tells you how bad; recall depth tells you how far to reach into the distribution chain. Part 7 frames three depths:

• Consumer or user level, the deepest, reaching the patient, the dispensing pharmacy, or the using clinician.
• Retail level, reaching pharmacies, hospitals, and dispensers but not individual patients.
• Wholesale level, reaching distributors and wholesalers only.

Depth is a function of class and of where the risk lives. A Class I sterile injectable almost always goes to the user level (hospital pharmacy, clinic, in some cases the patient). A Class III cosmetic defect on a tablet might stop at wholesale. Set depth in the recall strategy and be ready to justify it. Going too shallow is the classic finding; going deeper than necessary wastes effort and erodes credibility but is the safer error.

The recall decision: from signal to launch

A recall does not start with the recall. It starts with a signal, usually a complaint, a deviation, an out-of-specification result, a stability failure, or a regulatory finding. The discipline is in how fast that signal is escalated and how the decision is made and documented.

The path from that first signal to a closed file runs through the same sequence every time. Each step below gets its own subsection or section further down this page; the diagram is the map you can hold in your head.

Step 1: Capture and escalate the signal

Signals arrive through several doors:

• A product complaint (see product complaint handling) reporting a defect, foreign matter, lack of effect, or adverse event.
• A deviation or investigation (see deviation management) that, on review, casts doubt on distributed product.
• An OOS or OOT result on a retain or stability sample (see OOS investigation process and out-of-trend investigations).
• A stability program signal at a pull point (see stability programs ICH).
• A regulatory or supplier notification, including a supplier recall of an API or component that you used.

The non-negotiable: any signal that questions the safety, identity, strength, quality, or purity of distributed product must be escalated to QA and to the recall decision body immediately, not after the investigation finishes. The FAR clock (3 working days) and BPDR clock (45 days) start at receipt of the information, not at the end of root cause analysis.

Step 2: Stand up the recall committee and run the health hazard evaluation

Convene the standing recall committee. Perform an internal HHE that mirrors the FDA factors: severity of the hazard, probability of occurrence, exposed population and vulnerable subgroups, immediate and long-range consequences, and whether the hazard is detectable or self-limiting. Document a recommended class. Pull the data needed to bound the problem: which batches, how many units, where they went, what is still in your control, and whether the defect is confined or systemic.

Step 3: Decide the field action and its scope

Decide, on the record:

• Is this a recall, a market withdrawal, a stock recovery, or no action? Apply the Part 7 definitions, not convenience.
• Lot scope. Is it one batch, a campaign, or all lots within expiry? Define the inclusion rule (for example, “all lots of product X manufactured between date A and date B on line L”). Resist the urge to under-scope; expanding a recall later is far worse than scoping correctly the first time.
• Recall depth. Consumer, retail, or wholesale.
• Reason statement. One or two plain sentences a pharmacist can act on, no jargon.

Step 4: Notify FDA and file the required reports

Contact the FDA recall coordinator at the responsible district and the relevant Center (CDER for drugs, CBER for biologics). Provide the recall information that Part 7 expects: identity of the product, codes/lot numbers, reason, health hazard assessment, quantity produced and distributed and the time span, distribution pattern, your recall strategy, and proposed communications. In parallel, file the FAR within 3 working days if an NDA/ANDA product is involved, and the BPDR within 45 days if a licensed biologic is involved. If a shortage of a medically necessary drug may result, trigger the 506C notification.

Step 5: Issue recall communications

Issue the recall notification to your direct accounts (the firms you shipped to) per the recall strategy. A Class I recall is normally accompanied by a press release and may use Urgent Drug Recall letters and, where appropriate, public notification.

The recall strategy and the elements that go in it

Part 7 expects a documented recall strategy. Whether FDA-reviewed or firm-prepared, it should state every operational element. Build it as a template so you are not authoring it from scratch under pressure.

Contents of a recall strategy

Effectiveness check levels

The recall communication and effectiveness checks have defined coverage levels in Part 7:

A Class I recall typically requires Level A or B. The level is part of the strategy and FDA may push you up a level. The purpose of effectiveness checks is to confirm that consignees received the notice, understood it, and took the requested action. It is not enough to mail letters; you must verify action.

Recall communications: the recall letter

The recall notification letter to consignees is a controlled document. It should be conspicuous (Class I letters are often on red-bordered paper or marked “URGENT DRUG RECALL”), brief, and unambiguous.

What goes in a recall letter

• Clear identification: product name, strength, NDC, affected lot/control numbers, expiry, package size.
• A plain statement that it is a recall and the recall class if known.
• The reason for the recall and the health hazard in language the recipient can act on.
• Specific instructions: stop distribution and use immediately, quarantine affected stock, and (for distributors) notify their own customers (the sub-recall).
• A response/business reply form so the consignee can report quantity on hand, quantity already dispensed, and the action taken, even a “zero on hand” response.
• Disposition instructions: return for credit, or destroy with a destruction certificate.
• A contact name, phone, and email for questions, and a deadline to respond.
• A note that the recall has been or is being reported to FDA.

A worked recall-letter snippet

URGENT DRUG RECALL, CLASS I

Product: [Drug] Injection, 50 mg/mL, NDC 00000-000-00 Lots affected: AB1234, AB1235, AB1240. Expiry: 2027-08. Reason: Potential microbial contamination identified during stability testing. Use of a contaminated sterile injectable may cause serious infection. Action required: Immediately quarantine all units of the listed lots. Do not dispense or administer. If you distributed these lots to other facilities, notify them at once (sub-recall). Complete and return the enclosed response form within 5 business days, including units on hand and units already dispensed, even if your count is zero. Disposition: Return units using the enclosed prepaid label, or destroy and return the destruction certificate. Contact: Recall Coordinator, [phone], [email]. This recall has been reported to the U.S. Food and Drug Administration.

Reconciliation: accounting for every unit

Reconciliation is the arithmetic that proves the recall worked. It is the single most-cited weakness in recall execution because firms run the action but cannot show where the units went.

How reconciliation works

Build a quantitative ledger per lot:

• Quantity manufactured for the lot.
• Quantity distributed, from your shipping records, broken down by consignee.
• Quantity not yet distributed (stock recovery, your own quarantine).
• Quantity returned by consignees.
• Quantity destroyed in place, supported by destruction certificates.
• Quantity accounted for as already used/dispensed/administered, from consignee responses.
• Quantity unaccounted for, the residual.

The identity that must close:

Distributed = Returned + Destroyed-in-place + Used/Dispensed + Still-unaccounted

Worked reconciliation example

For AB1234: accounted = 6,100 + 1,500 + 1,400 = 9,000 of 9,200 distributed, leaving 200 unaccounted, 97.8 percent accounted. The unaccounted residual is what you investigate and what FDA asks about. For each unaccounted unit you need a credible explanation (administered before notice reached the site, normal patient consumption of a dispensed drug, lost in a closed pharmacy) and you document it.

Acceptance criteria for reconciliation

• The ledger closes per lot, with every category supported by a record (shipping log, response form, destruction certificate).
• Used/dispensed and unaccounted quantities are explained, not just stated.
• Returned product is received into a controlled quarantine, segregated, and not commingled with saleable stock.
• The total accounted percentage and the residual are reported in the recall closure package.
• For DSCSA-covered product, returned and quarantined units are handled consistent with verification and traceability requirements (see DSCSA serialization).

Effectiveness checks: proving consignees acted

Effectiveness checks are distinct from reconciliation. Reconciliation accounts for units; effectiveness checks verify that consignees received, understood, and acted on the recall.

How to run them

1. Build the consignee list from distribution records (your direct accounts). This is your denominator.
2. Determine the level (A through E) from the strategy.
3. Contact the required percentage by the chosen method (visit, phone call, letter, or response-form return).
4. For each contact, confirm three things: the notice was received, it was understood, and the requested action was taken (quarantine, sub-recall, return/destroy).
5. Record date, method, person contacted, and result. Re-contact non-responders; a consignee that does not respond is not an effective check.
6. Compute the effectiveness check completion rate and document it.

Acceptance criteria

• The percentage of consignees verified meets or exceeds the level (Level A means 100 percent verified, not 100 percent contacted).
• Non-responders are escalated and pursued, not dropped.
• Sub-recall obligations to downstream customers were communicated and confirmed for distributors.
• Findings of failed or partial action trigger corrective contact.

A worked target: for a Class I recall at Level A with 320 direct consignees, you must obtain a verified action response from all 320. If 11 do not respond after letter, call, and follow-up, those 11 are open items in the recall file with a documented escalation path, and FDA will ask about every one.

Recall termination

A recall is not over when product stops coming back. Part 7 describes recall termination, normally requested from FDA in writing, when the firm and FDA agree that:

• The firm has made reasonable efforts to remove or correct the violative product per the recall strategy.
• It is reasonable to assume the product has been removed and proper disposition or correction has been made commensurate with the degree of hazard.
• Effectiveness checks and reconciliation support that conclusion.

The termination package summarizes the action: classification, scope, communications issued, effectiveness check results, reconciliation per lot, disposition of returned product, root cause and CAPA status, and the request to terminate. FDA confirms termination in writing. Until then, the recall is open and is tracked in FDA’s weekly Enforcement Report.

Field Alert Reports in detail

The FAR deserves its own treatment because it is fast, frequently triggered, and routinely mishandled.

What it is and why

Per 21 CFR 314.81(b)(1), an NDA/ANDA holder must alert the local FDA district office within 3 working days of receiving information about:

• A mix-up where the product or its labeling could be mistaken for, or applied to, another article.
• Bacteriological contamination of the distributed product.
• A significant chemical, physical, or other change or deterioration in the distributed product.
• Any failure of one or more distributed batches to meet the specification established in the application.

The rationale is early warning. FDA wants to know about a potential field problem while it is still a signal, before the investigation concludes and before any recall decision is made. Filing a FAR does not commit you to a recall, and many FARs close without one.

What goes in a FAR

FARs are commonly submitted on Form FDA 3331 (initial, follow-up, and final). Contents:

• Applicant and application number (NDA/ANDA).
• Product name, strength, NDC, lot/control numbers, expiry.
• A description of the problem and how it was discovered.
• The distribution status of affected lots.
• Immediate actions taken (quarantine, investigation opened).
• Whether the report is initial, follow-up, or final.

How to handle the FAR

1. The moment a triggering signal lands (a complaint of contamination, a confirmed OOS on a distributed lot, a label mix-up report), start the 3-working-day clock at receipt, not at confirmation.
2. File an initial FAR even if the investigation is incomplete. It is acceptable, and expected, to file on a credible unconfirmed signal and to update later.
3. File follow-up reports as the investigation develops.
4. File a final FAR when the investigation closes, stating the conclusion and whether a recall was initiated.

Common FAR mistakes

• Treating the 3-day clock as starting when the OOS is confirmed rather than when the information was received. The clock starts at receipt of the information.
• Waiting to file until the firm “knows for sure,” missing the window. File initial, update later.
• Filing for the contamination but forgetting the parallel BPDR for a licensed biologic (45 days) or the recall reporting.
• Not closing out with a final report, leaving the FAR open in FDA’s system.

A practical tip: maintain a simple FAR log keyed to complaints and OOS events with the receipt date stamped, so the 3-day clock is never ambiguous. The most defensible position is “we filed initial within 3 working days of receipt and updated as we learned more.”

Recalling investigational (clinical trial) product

Everything above assumes marketed product already in commercial distribution. Investigational product distributed under an IND runs through a different framework, because it has not been introduced into commercial distribution and the population exposed is a defined set of trial subjects at named sites rather than the open market. The objective is the same, get a known defect out of use fast and prove you did, but the mechanics change.

Why the commercial recall framework does not simply carry over

Part 7’s recall and classification concepts were built for marketed product moving through wholesalers, pharmacies, and patients who bought or were dispensed the product independent of a research relationship. Investigational product moves through a closed chain: sponsor to site to enrolled subject, under a protocol, with an investigator and an IRB or IEC already in the loop for every subject. There is ordinarily no press release, no retail-level notification, and no FDA-assigned recall class in the Part 7 sense. What replaces that machinery is IND safety reporting, direct sponsor-to-investigator communication, and site-level drug accountability, all of which already exist as GCP obligations before a defect is ever found; a defect just activates them urgently.

IND safety reporting

Under 21 CFR 312.32, a sponsor that determines information from a clinical, epidemiological, animal, or in vitro finding suggests a significant risk to trial subjects must notify FDA and every participating investigator. The general clock is 15 calendar days from the sponsor’s determination that the information qualifies for reporting. Where the finding is an unexpected fatal or life-threatening suspected adverse reaction, the clock shortens to 7 calendar days from the sponsor’s initial receipt of the information. A product-quality defect that creates a plausible safety signal, a potency drift in a biologic, a sterility concern in an injectable investigational product, a labeling mix-up that could lead to a wrong dose, is assessed against these same clocks the moment it is identified, in parallel with the quality investigation, not after it concludes.

Investigator notification and site-level action

The sponsor notifies every participating investigator using the affected product, not only the site or sites where the defect was first found, because the same lot is frequently shipped across the study. The notification states what is known, what action the site must take (quarantine remaining supply, cease dispensing, or continue under specific instructions pending further information), and what to tell subjects already dosed. Each site’s principal investigator carries GCP responsibilities to protect subject safety, which includes acting on the sponsor’s instruction promptly and notifying the site’s IRB or IEC where the finding affects the risk picture described in the informed consent. See ICH E6 good clinical practice for the investigator and sponsor obligations this sits inside, and GMP for investigational products for how phase-appropriate quality control produces the lots now in question.

Drug accountability reconciliation at sites

Where a commercial recall reconciles a distribution chain of consignees, an investigational product action reconciles drug accountability at each site. Under 21 CFR 312.62, the investigator maintains records of the disposition of the investigational drug, dates, quantity, and use by subjects, and returns unused supply to the sponsor or otherwise disposes of it under 21 CFR 312.59 at the sponsor’s direction. When a lot is pulled, the reconciliation question at each site takes the same shape as the commercial ledger: quantity shipped to the site, quantity dispensed to subjects, quantity administered, quantity remaining on site, quantity returned or destroyed, and any residual gap explained. Roll the site-level ledgers up to a study-level total the same way lot-level reconciliation rolls up in a commercial recall. The IMP accountability and reconciliation log is the working record this reconciliation runs on.

Acceptance criteria

• The safety reporting clock (7 or 15 calendar days, as applicable) is met, with the determination date and the filing date both on record.
• Every participating investigator using the affected lot is notified, with confirmation of receipt tracked the way a commercial recall tracks effectiveness checks.
• Each affected site’s drug accountability reconciles, with a documented explanation for any residual.
• IRB/EC notification is assessed and completed at every affected site where the finding changes the risk information subjects were given.
• Where subjects were dosed with the affected lot, a medical assessment of the exposure is documented, independent of whether an adverse event was reported.
• Resupply, protocol continuity, and any amendment needed to keep the study compliant are tracked to closure alongside the safety and accountability items.

A worked example

A sponsor’s cell and gene therapy IND ships a manufacturing lot to three clinical sites. A potency assay result on a retained sample, read after two of the three sites have already dosed subjects, comes in below the pre-specified acceptance range. The sponsor determines this is a significant risk finding: subjects at the two dosed sites may have received a sub-potent product. The IND safety report is filed within the 15-day clock from the determination date (the finding is not an unexpected fatal or life-threatening reaction, so the 7-day clock does not apply). All three participating investigators are notified the same day the determination is made, not held for the formal safety report. The third site, which had not yet dosed, quarantines its remaining supply on sponsor instruction. Each site’s investigator confirms the quantity shipped, dispensed, administered, and remaining, closing the reconciliation for that lot at each site. The two sites that dosed subjects report the exposure to their IRBs and document a medical assessment of each affected subject. The root cause traces to a potency assay drift rather than the drug substance itself, and the CAPA revises the in-process potency control point before the next lot ships.

Common finding

Treating a defect on investigational product as a routine internal deviation, run to closure inside the manufacturing quality system alone, with no IND safety reporting assessment, no site-level accountability reconciliation, and no evidence any investigator outside the discovering site was ever told. The commercial recall machinery does not apply, but the absence of an equivalent trail is exactly what an inspector reads as a gap, not as a reason the requirement did not exist.

Counterfeit and suspect product: a different trigger under DSCSA

Not every field action starts with a manufacturing defect. A unit can look wrong, or arrive through a channel it should not have, without any failure in your own process at all: a counterfeit, a diverted lot, a stolen shipment, or product whose chain of custody cannot be verified. The Drug Supply Chain Security Act, section 582 of the FD&C Act, builds a separate trigger and a separate response path for this category, distinct from the quality-defect recall this page otherwise describes.

Why this is a different animal

A quality-defect recall starts from evidence that your own genuine product failed a specification or presents a hazard, and the investigation runs through your batch records, your process, and your reserve samples. A suspect-product event starts from a doubt about whether the unit in front of you is authentic, was handled by authorized trading partners the whole way, or has been tampered with, and the investigation runs through transaction history, product identifier verification, and physical inspection instead. Running a suspect-product report through the standard complaint-and-recall investigation, batch record review first, defect testing first, misses the actual question, and running every ordinary quality complaint through a DSCSA suspect-product workflow buries the real illegitimate-product signals in noise.

The DSCSA path

DSCSA requires a trading partner to quarantine and investigate a suspect product, one that is reasonably suspected of being counterfeit, diverted, stolen, intentionally adulterated, the subject of a fraudulent transaction, or otherwise unfit for distribution such that it could present a risk to health. The investigation verifies the product identifier through the Verification Router Service, reviews the transaction information and history, inspects the physical unit, and reaches out to the upstream trading partner and the manufacturer. Where the investigation confirms the product is illegitimate, or where a manufacturer determines a suspect product poses a high risk of illegitimacy, the trading partner in possession must notify FDA and, as applicable, immediate trading partners, within 24 hours of that determination, under section 582(c)(4)(B)(ii) and (d)(4)(B)(ii) of the FD&C Act. Notification is normally submitted on Form FDA 3911. See DSCSA and serialization for the verification and transaction-data mechanics this investigation draws on, the suspect and illegitimate product SOP for the full investigation and notification procedure, and the DSCSA exception and data discrepancy log for keeping routine data noise, a failed scan, a late transmission, separate from a genuine suspect-product trigger.

How the two paths connect

The two triggers are not mutually exclusive, and the investigation itself is what tells you which one you are in, or both. A suspect-product investigation can uncover that the product in question is, in fact, your genuine product, diverted from the legitimate supply chain and re-introduced through an unauthorized channel; that finding closes the DSCSA notification question but can open a real question about whether the diverted product was stored and handled outside your controlled conditions, which is a quality and potentially a recall question in its own right. In the other direction, a quality investigation that starts as an ordinary complaint can turn up a lot number or product identifier that does not match your records at all, which converts it into a suspect-product trigger rather than a quality-defect one. The triage decision, and the point at which one workflow hands off to the other, belongs on the record either way.

Acceptance criteria

• Every suspect-product trigger is quarantined and documented immediately, before the investigation determines legitimacy.
• The investigation addresses product identifier verification, transaction history, physical inspection, and upstream partner contact, not only one of these.
• The 24-hour notification clock is met from the determination date, with Form FDA 3911 (or the applicable notification) on file, whenever the product is determined illegitimate or high-risk.
• Where the investigation clears the product, the rationale is documented and retained even though no report is filed.
• Where a suspect-product investigation reveals the affected units are genuine product that left controlled conditions, or a quality investigation surfaces a suspect-product identifier mismatch, the hand-off between the two workflows is recorded, not silently absorbed into one or the other.

A worked example

A pharmacy contacts the company about a vial with a label that looks slightly off, seal color visibly different from the reference sample the pharmacy keeps on file. The report is logged and the unit is requested back. On physical inspection, the lot number does not match any lot the company ever produced, and the product identifier fails Verification Router Service lookup. This is a suspect product, quarantined and investigated under the DSCSA path, not the quality complaint path: no batch record exists to review because the company never made this unit. The investigation confirms the product is illegitimate, likely counterfeit, and FDA is notified within 24 hours of that determination, along with the immediate trading partner that supplied the unit to the pharmacy. Contrast this with a second pharmacy report the same week describing similar hazy appearance on a unit whose lot number and product identifier verify cleanly as genuine: this second report is a quality complaint, not a suspect-product trigger, and runs through the standard investigation path this page otherwise describes.

Common finding

A suspect-product report handled entirely as a quality complaint, with testing performed on a unit before anyone confirmed it was genuine company product, or a genuine defect signal absorbed into a DSCSA investigation and never assessed for the ordinary FAR, BPDR, or recall obligations that a real quality defect on genuine distributed product would trigger.

International and multi-country recall coordination

A firm distributing the same product into several markets can find itself running one underlying defect through several different national recall systems at once. The substance of the problem does not change by crossing a border; the process to close it out does.

What changes market to market

• Classification systems differ in structure even where the underlying health-hazard logic is similar. The US three-class system (Class I, II, III) is one implementation of a health-hazard-based classification; other regulatory systems apply their own terminology and thresholds for what counts as a serious versus a lesser risk. Do not assume a Class II conclusion reached for the US market automatically maps onto an equivalent category elsewhere; each authority’s classification is assessed against its own framework using the same underlying hazard evidence.
• Notification timelines are set independently by each authority. Some markets expect notification of a quality defect that may lead to a recall within a short window measured in hours or a few days; others work to a different cadence. There is no single global clock, and assuming the US FAR or BPDR timeline applies everywhere is a common and avoidable error. Regulatory affairs confirms the notification expectation for each market the affected lot reached before the clock in that market is treated as met or missed.
• Product presentation is not always identical across markets. The same drug substance can be marketed under a different trade name, strength, pack size, or labeling in different countries, sometimes manufactured at a different site for a given market. Confirm the specific presentation, and whether the same defect actually applies to the market-specific version, before assuming a global recall scope; a defect confirmed for one market’s presentation is not automatically present in another market’s differently formulated or packaged version of the “same” product.

Keeping the message consistent while adapting the format

The reason-for-recall statement is the piece most at risk of drifting between markets, because each national authority may specify its own submission format, language, and level of detail. The discipline is to author one master reason statement and one master health hazard rationale, translate and adapt the format for each authority, and never let a local affiliate or a local regulatory team independently draft a reason statement that says something substantively different from the master version. Two regulators comparing notes on the same recall, which happens more than firms expect, should read the same explanation of what went wrong, not two versions that appear to describe different problems.

Coordinating execution across markets

• Name a single global recall coordinator who owns the overall action, with a named local contact in each affected market who executes the local notification, consignee communication, and effectiveness checks under local requirements.
• Use one master recall reference that every market-specific action ties back to, even where a market’s own authority assigns its own local case number.
• Set depth and effectiveness-check expectations per market against that market’s own consignee structure; do not assume the depth appropriate for one country’s distribution model (for example, direct-to-pharmacy) fits another’s (for example, a wholesaler-heavy chain).
• Reconcile quantities per market the same way a single-market recall reconciles per lot, then roll the market-level reconciliations up to a total distributed-versus-accounted-for picture across every market the lot reached.
• Close each market’s action on that market’s own authority confirmation; a US termination letter under 21 CFR 7.55 does not close the obligation to any other market’s authority, and the reverse is equally true.

Acceptance criteria

• Every market the affected lot was distributed into is identified from distribution records, not assumed from the product’s intended market footprint.
• Each market’s classification, notification clock, and authority format are assessed against that market’s own current requirements, confirmed with regulatory affairs or local counsel rather than carried over from the US assessment.
• One master reason-for-recall statement and health hazard rationale governs every market’s local version.
• Reconciliation is reported both per market and as a combined total.
• Each market’s action is tracked to its own closure, independent of when any other market closes.

This section stays deliberately general because notification formats, classification terminology, and authority-specific submission requirements vary by market and change over time. Confirm the current requirement for each market you distribute in with regulatory affairs or local counsel before relying on any specific figure; the EU GMP Chapter 8 and WHO references noted earlier on this page are starting points, not a substitute for the current national guidance in each market.

Mock recalls: testing the system before you need it

You do not want the first full test of your recall system to be a real recall. EU GMP Chapter 8 explicitly expects the recall procedure to be evaluated periodically, and a mock recall is the standard tool. FDA inspectors routinely ask to see your last mock recall and its timing results.

What a mock recall tests

• Can you, from a lot number, reconstruct the full distribution list (who got it, how much, when)? This is the traceability test.
• How fast can you do it? You should be timing it.
• Are your contact lists for consignees and your recall committee current?
• Does your notification and response-form machinery work end to end?

How to run a mock recall

1. Pick a real distributed lot (and ideally a “hard” one, like a lot that went through multiple distributors).
2. Start a clock. Task the team with producing the complete distribution list to the chosen depth.
3. Record the elapsed time to retrieve the distribution data.
4. Optionally exercise the notification step with a test letter to a sample of accounts (clearly marked as a test) or a tabletop walkthrough of communications.
5. Reconcile what you could and could not retrieve.
6. Document the result, the time taken, gaps found, and corrective actions, and feed those into the procedure.

Acceptance criteria and a worked target

• The full distribution list is retrievable to the required depth within a pre-defined target time. Many firms set a target such as retrieving the distribution list within 24 hours, with the actual time recorded.
• 100 percent of the lot is traceable to consignees (no “we lost track” gaps).
• Contact information is verified current.
• Gaps generate CAPA.

Worked example of a mock recall summary entry:

A common finding is mock recalls that are too easy (always the same simple lot) or that test retrieval but never test the notification and effectiveness-check mechanics. Rotate scenarios and occasionally exercise the full chain.

Roles and responsibilities

The recall coordinator role matters: FDA wants one named, reachable person. Diffuse ownership is a finding waiting to happen.

Cross-link: see GxP roles and responsibilities and pharmaceutical quality system for how these seats fit the broader org.

How recalls connect to the rest of the quality system

A recall is an output of the quality system, not a standalone event. The connections are what inspectors probe.

• Complaints feed recalls. A pattern of complaints on a lot or product is a classic recall trigger. See product complaint handling.
• Deviations and investigations can implicate distributed product. The deviation system must escalate to the recall decision body. See deviation management.
• OOS and OOT results on retains and stability can be the originating signal. See OOS investigation process and out-of-trend investigations.
• CAPA closes the loop. A recall without a root cause analysis and an effective CAPA is incomplete; FDA expects to see the systemic fix, not just product removal. See what is a CAPA, root cause analysis techniques, and CAPA effectiveness verification.
• Quality event triage decides early whether a signal is recall-relevant. See quality event classification triage.
• Management review sees recall metrics and trends. See management review Q10 and quality metrics and KPIs.

A useful mental model: complaint or deviation, then investigation and OOS, then field alert and health hazard evaluation, then recall decision and strategy, then execution (communications, returns), then reconciliation and effectiveness checks, then root cause and CAPA, then termination and management review.

Common mistakes and inspection-finding patterns

These are the generic patterns that show up in inspectional observations and warning letters on recall and field-action topics. No company names, just the patterns.

• Slow or missing FAR. The OOS or contamination signal was received but the field alert was not filed within 3 working days, or the 3-day clock was wrongly started at investigation closure.
• Failure to extend to distributed product. An investigation concluded on the manufacturing problem but the firm did not assess or act on product already in the field.
• Under-scoped recall. Only the obviously affected lot was recalled when the root cause (a contaminated component, a process drift, an equipment failure) plausibly affected other lots, forcing a later expansion.
• No reconciliation, or it does not close. The firm sent letters but cannot show how many units came back, were destroyed, or remain unaccounted, and cannot explain the residual.
• Effectiveness checks confused with contact. Letters were mailed but action was never verified; non-responders were not pursued.
• No mock recall, or a token one. No periodic test of the recall system, or a mock recall that only ever uses an easy lot and never times retrieval.
• Stale contact lists. Consignee and committee contact information out of date, slowing the action.
• Misclassification by convenience. Calling a recall a “market withdrawal” to avoid Part 7 obligations and FDA reporting when the product is genuinely violative.
• Disposition gaps. Returned product commingled with saleable stock, or destruction without certificates.
• Weak root cause and CAPA. Product removed but no systemic corrective action, so the defect recurs. The recall fixed the symptom, not the cause.
• Missed parallel obligations. Forgetting the biologic BPDR (45 days), the 21 CFR 806 correction/removal report for a combination product’s device constituent, or the drug shortage notification.

Cross-link: see FDA warning letters patterns and 483 warning letter response for how these surface and how to respond.

Interview-ready: questions and how to answer

“Walk me through the difference between a recall, a market withdrawal, and a stock recovery.” A recall is removal or correction of marketed product FDA considers violative and would take legal action against. A market withdrawal addresses a minor issue not subject to FDA legal action, or no violation at all, like normal stock rotation. A stock recovery is retrieval of product that never left the firm’s control. The dividing line is whether the product is violative and whether it left your control.

“What classes of recall are there and who assigns the class?” Class I (reasonable probability of serious harm or death), Class II (temporary or reversible harm, or remote probability of serious harm), Class III (not likely to cause harm). The firm recommends a class, but FDA assigns it through a health hazard evaluation. I’d give an example of each tied to clinical consequence.

“A complaint comes in that a distributed sterile injectable lot may be contaminated. What happens in the first 72 hours?” Log and escalate the complaint immediately to QA and the recall committee. Open the investigation and quarantine retains and any undistributed stock. Run a health hazard evaluation. File the FAR within 3 working days of receipt for the NDA/ANDA product, and prepare the BPDR if it’s a licensed biologic. Pull the distribution list and bound the scope. Contact the FDA recall coordinator and prepare a recall strategy. For a contaminated sterile injectable, I’d expect Class I, user-level depth, Level A effectiveness checks, and a press release.

“When is a field alert report due and what triggers it?” Within 3 working days of receiving the information, per 21 CFR 314.81(b)(1). Triggers are a product/label mix-up, bacteriological contamination of distributed product, a significant chemical/physical change or deterioration in distributed product, or any failure of a distributed batch to meet its application specification. The clock starts at receipt of the information, and you file an initial report even before the investigation closes.

“What’s the difference between reconciliation and effectiveness checks?” Reconciliation is the unit-level arithmetic: distributed equals returned plus destroyed plus used plus unaccounted, closed per lot with records and an explanation for the residual. Effectiveness checks are verification that consignees received the notice, understood it, and took the action, at the coverage level (A through E) set in the strategy. One accounts for product; the other verifies behavior.

“What is a mock recall and why do you run one?” A periodic test of the recall system, usually starting from a lot number to reconstruct the full distribution list to the required depth, timed against a target. It tests traceability, contact-list currency, and the notification machinery before a real recall does. EU GMP Chapter 8 expects the recall procedure to be evaluated periodically; FDA inspectors ask to see the last mock recall and its timing.

“What are the recall effectiveness check levels?” Level A is 100 percent of consignees, Level B is a percentage between A and C, Level C is 10 percent, Level D is 2 percent, Level E is no checks. A Class I recall is typically Level A or B.

“What additional reports apply to a biologic versus a small-molecule drug?” For a licensed biologic, a biological product deviation report under 21 CFR 600.14 within 45 days of discovery, in addition to any FAR-equivalent handling and the recall reporting. For a combination product, the device constituent can also pull in a correction/removal report under 21 CFR 806. Knowing which obligation attaches to which product type is the point.

“How do you decide recall depth?” By class and where the risk lives. Class I sterile injectables go to user level; a low-risk Class III defect might stop at wholesale. Going too shallow is the common finding, so when in doubt I go deeper and document the rationale.

“FDA asks why 200 units of a recalled lot are unaccounted for. How do you answer?” With the reconciliation ledger. For those 200 units I’d show the distributed-to-account breakdown, the consignee responses, and a documented explanation for each residual: administered before the notice arrived, dispensed and consumed by patients, or genuinely lost in a closed account. The answer is records plus rationale, never a shrug.

A condensed execution checklist

Use this as the operational spine. Each line maps to a section above.

1. Capture the field signal (complaint, deviation, OOS, stability, supplier notice) and escalate to QA and the recall committee at receipt.
2. Start the FAR clock (3 working days) and, for biologics, the BPDR clock (45 days).
3. Convene the recall committee; run the health hazard evaluation; recommend a class.
4. Decide the action (recall vs withdrawal vs stock recovery), lot scope, and depth.
5. Notify the FDA recall coordinator and the relevant Center; file FAR/BPDR; check for shortage notification.
6. Issue recall communications to direct accounts with response forms; add a press release for Class I.
7. Execute returns and disposition into controlled quarantine; collect destruction certificates.
8. Run reconciliation per lot until the identity closes and the residual is explained.
9. Run effectiveness checks to the required level; pursue non-responders.
10. Complete root cause analysis and CAPA; verify CAPA effectiveness.
11. Assemble the termination package and request FDA termination.
12. Report recall metrics into management review; feed lessons into the procedure and the next mock recall.

Practical tips

• Keep the receipt date stamped on every signal. Most FAR and reporting failures are timing failures that trace back to an ambiguous start date.
• Pre-build templates: recall strategy, recall letter, response form, reconciliation ledger, effectiveness-check tracker, termination package. Authoring under pressure produces errors.
• Maintain a current, named recall coordinator and a standing committee with verified contacts. FDA wants one reachable human.
• Scope on root cause, not on the obvious lot. Ask “what else could this defect have touched?” before you draw the recall boundary.
• Treat the residual (unaccounted units) as the thing to explain, not the thing to hide. A clean reconciliation with explained residuals beats a vague “all recovered.”
• Rotate mock-recall scenarios and time every run. A mock recall that never fails is not testing anything.
• Do not let the recall close on product removal alone. The CAPA is what stops the next one, and the inspector will look for it.
• For DSCSA-covered product, your serialization and traceability data is now part of how fast you can find and verify recalled units; keep that data path tested.

Related reading

• Product complaint handling
• Deviation management
• OOS investigation process
• Out-of-trend investigations
• What is a CAPA
• CAPA effectiveness verification
• Root cause analysis techniques
• Quality event classification triage
• Stability programs ICH
• DSCSA serialization and track and trace
• Combination products and cGMP under Part 4
• ICH E6 good clinical practice
• GMP for investigational products
• Management review Q10
• FDA warning letters patterns