Reference vs Retention Samples: Planning Storage and Future Testing

Reference vs retention samples: the two purposes Annex 19 separates

Most manufacturing and release sites hold back some material from every batch they release. What that material is for decides how much of it is needed, where it sits and how long it stays. The European Union's EudraLex Volume 4 Annex 19 names two distinct purposes: reference samples and retention samples. Treating them as one undifferentiated pile of "retains" tends to work until a complaint, an inspection or a site closure asks the samples to answer a specific question about a specific batch.

Under section 2.1 of Annex 19, a reference sample is a sample of a batch of starting material, packaging material or finished product that is stored for the purpose of being analysed should the need arise during the shelf life of the batch. The same section adds that, where stability permits, reference samples from critical intermediate stages (for example, those requiring analytical testing and release) and from intermediates transported outside the manufacturer's control should be kept. A retention sample is a sample of a fully packaged unit from a batch of finished product, stored for identification purposes: presentation, packaging, labeling, patient information leaflet, batch number and expiry date. Section 2.3 lists the situations the samples serve, including a dosage-form quality complaint, a query about compliance with the marketing authorization, a labeling or packaging query, and a pharmacovigilance report.

For finished products, section 2.1 notes that in many instances the two will be presented identically, as fully packaged units, and may then be regarded as interchangeable. The annex also recognizes exceptional circumstances where duplicate samples are not needed, such as small amounts of a batch packaged for different markets, or very expensive medicinal products. Interchangeable packs do not remove either set of rules. The packs held as the reference sample must still be enough, counted in unopened packs where necessary, to run the full approved analytical controls twice (section 4.1). And where a batch is packaged in two or more distinct packaging operations, at least one retention sample is needed from each operation unless an exception is justified to, and agreed with, the competent authority (section 4.3).

What changed in June 2026: the revised Annex 19 in operation

The revised Annex 19 is dated 23 June 2026 and carries the reference Commission Decision C(2026) 4135 final. Its cover page gives the reason for the change: the GMP/GDP Inspectors Working Group and the PIC/S Committee jointly recommended revising the 2006 version with respect to reference and retention samples for parallel imported, parallel distributed and parallel traded products. The document sets the deadline for coming into operation as three months from the date of publication by the European Commission; it does not print a calendar date. GMP Insiders reports that the text was issued on 24 June 2026 and became applicable on 24 September 2026, and the ECA Academy guideline index catalogues the 2026 revision alongside the 2006 text.

A side-by-side reading of the 2006 Annex 19 and the 2026 text shows how narrow the change is for most manufacturers. Duration of storage (section 3), sample size (section 4), storage conditions (section 5), written agreements (section 6), reference- and retention-sample location (sections 7 and 8) and closedown (section 10) carry over with only editorial changes. The legal basis now cites Directive 2017/1572/EU. The investigational-products cross-reference now points to the detailed Commission guideline on GMP for investigational medicinal products, a veterinary cross-reference in section 3.1 has been dropped, and section 3.2 now counts starting-material retention from release of the finished product rather than "release of product". For most site procedures the action is administrative: update the document references, not the sampling model.

The substantive change is in section 9. In 2006 it held two short rules for parallel imported and parallel distributed products: if the secondary packaging was not opened, only the packaging material used needed to be kept; if it was opened, one retention sample per packaging operation was needed. The 2026 text adds parallel traded products to the scope and becomes prescriptive. It covers re-packaging materials, says reference samples of the re-packaged product are not needed, sets the retention period and contents of the retention sample, and adds a new, conditioned photographic or digital alternative. Those rules are set out in the parallel-trade section below.

How long to keep each sample: EU, US, API and biologic rules side by side

EU GMP speaks of reference and retention samples. The US rule, 21 CFR 211.170, uses a single term, reserve sample, for material that serves both roles. For APIs, ICH Q7 section 11.7 uses "reserve/retention samples". The clocks also start from different events: expiry of the batch, expiry of the last drug product lot made with an ingredient, release of the finished product, or distribution. Mixing them up leads either to destroying samples too early or to holding them far longer than any rule requires.

The matrix below sets out the minimum period in each source. Where a product is supplied to more than one market, the longest applicable period governs that sample.

Two readings deserve care. First, the "whichever is longer" structure belongs to ICH Q7 section 11.71 for APIs. It is not the 211.170 rule for drug products, which counts from the expiration date alone, subject to the radioactive and OTC variants. Second, the June 2015 ICH Q7 Questions and Answers define "completely distributed" as distribution of the entire API batch by the API manufacturer to the next party in the supply chain, and they apply that reading to every party that physically processes or repackages the API. The Q&A also states the intent: keep samples for as long as the API could be on the market. Its example is an API with a five-year retest date that is completely distributed immediately after manufacture; the reserve sample was never meant to be destroyed before the five-year retest date was reached.

The starting-material clock in Annex 19 section 3.2 runs from release of the finished product, not from receipt of the material. Take an API lot that is received, held for 18 months and then used in a finished batch that is released straight away. Its EU minimum ends two years after that release, 42 months after receipt; a conservative reading counts from the release of the last finished batch that used the lot. If the same lot also goes into product for the US, 211.170(a)(1) runs until one year after expiry of the last drug product lot containing it. With a 36-month shelf life, that is at least 48 months after the last lot is made, so the US period is the longer one and governs the sample.

Licensed biologics add their own floor. 21 CFR 600.13 requires manufacturers to keep, for at least 6 months after the expiration date unless additional standards specify a different period, a quantity of representative material from each lot that is sufficient for examination and testing for safety and potency. The samples must include at least one final container as a final package, or a package-equivalent, be stored under conditions that maintain the product's identity and integrity, and be held in addition to any samples that must be submitted to FDA.

Sizing samples for two rounds of testing — and keeping the lab able to run them

Annex 19 section 4.1 sets the EU rule: the reference sample should be large enough to permit, on at least two occasions, the full analytical controls on the batch in accordance with the approved Marketing Authorisation File. The US rule is written as arithmetic: 211.170 requires at least twice the quantity needed to perform all the required tests, except sterility and pyrogen tests, for both active ingredients and drug products. ICH Q7 section 11.72 asks API manufacturers to keep enough for at least two full compendial analyses or, where there is no pharmacopoeial monograph, two full specification analyses; FDA's Q7 guidance carries the same text. WHO's sampling guidance sizes a retention sample for at least two confirmatory analyses.

Turning "two full sets of controls" into a unit count is product-specific. A defensible plan works from the approved specification, test by test, and records each decision:

1. List every release test in the approved specification and the quantity each one consumes under the approved method, including replicate preparations.

2. Decide, and write down, how staged tests are counted. Some tests escalate to further units when an earlier stage does not meet its criteria. Whether one "full analytical control" includes the maximum escalation is a judgment the plan should record, not one an analyst should make on the day.

3. Note which tests can share units and which need their own. Units used in a non-destructive test may be reusable for a later test; units consumed by a destructive test are not.

4. Count packs, not only units. Section 4.1 says unopened packs should be used for each set of analytical controls where necessary, so two sets of controls means at least two separate groups of unopened packs.

5. For US supply, 211.170 lets sterility and pyrogen tests sit outside the doubled quantity. The Annex 19 text has no matching carve-out: any exception to the two-occasion rule has to be justified to, and agreed with, the competent authority.

6. Add national requirements. Section 4.2 says national requirements on the size of reference samples, and if necessary retention samples, should be followed where they apply.

As an illustration only: suppose one full run of the approved specification, counted this way, consumes 75 tablets. The reference sample then needs at least 150 tablets. If the product is marketed in 30-tablet packs and each round needs unopened packs, each round needs three packs (75 ÷ 30 = 2.5, rounded up), so the plan holds six packs, or 180 tablets, rather than the five packs that 150 ÷ 30 would suggest. Rounding happens per round, not across the total, because a pack opened in the first round cannot serve the second. The second round needs its own unopened packs because opening a container exposes the remaining units to moisture, oxygen and light in ways the marketed pack was designed to prevent.

Storage conditions, locations and who holds what

Annex 19 section 5 has two short rules. Reference samples of finished products and active substances are stored in accordance with the current version of the Note for Guidance on Declaration of Storage Conditions for Medicinal Products and Active Substances (5.1), and storage conditions follow the marketing authorization, for example refrigerated storage where relevant (5.2). For US drug products, 211.170(b) says the reserve sample is stored under conditions consistent with product labeling. In practice, the retained-sample store should be qualified and monitored to the labeled condition for the whole retention period, with excursions assessed as they would be for released stock.

Packaging is part of the storage condition. Under Annex 19 section 3.1, the finished-product reference sample is held in its finished primary packaging or in packaging of the same material as the marketed primary container; 211.170(b) asks for the marketed immediate container-closure system or one with essentially the same characteristics. Holding tablets that are marketed in high-barrier blisters as loose bulk in a polyethylene bag would meet neither text, because the sample would no longer have the barrier the patient's pack provides. APIs are treated differently. ICH Q7 section 11.72 allows storage in the same packaging system as the API or one that is equivalent to or more protective than the marketed system. The 2015 Q&A explains why: the reserve sample exists to allow future evaluation of the original API batch, for example of potential counterfeits, so it may be stored in packaging and conditions that better preserve the batch's original state.

Location and custody rules sit in Annex 19 sections 6 to 8:

• Reference samples belong near a validated laboratory. Section 7.1 says they should be conveniently available to a laboratory with validated methodology. For starting materials used in medicinal products made in the EEA, that is the original site of manufacture of the finished product; for finished products made in the EEA, it is the original site of manufacture. Reference samples of starting materials and packaging materials are kept at the site where they were used to make the medicinal product (7.2.3).

• Finished products made outside the EEA depend on MRA status. Where an operational MRA is in place, reference samples may be taken and stored at the non-EEA site of manufacture, covered by a written agreement between the importer or batch-release site and the manufacturer (7.2.1). Where no operational MRA is in place, reference samples of the finished product are taken and stored at an authorized manufacturer within the EEA, preferably where testing on importation was performed (7.2.2). Confirm the MRA's operational status and product scope for the specific country before relying on 7.2.1.

• Retention samples stay in the EEA. A retention sample represents the batch as distributed in the EEA and may be needed to confirm non-technical attributes, so retention samples are located within the EEA in all cases, preferably at the site of the certifying QP (8.1). They are stored at the premises of an authorized manufacturer to permit ready access by the competent authority (8.3). Where reference samples are kept outside the EEA under an MRA, separate retention samples are kept within the EEA (8.2).

• Written agreements and QP access. Where the MAH is not the same legal entity as the EEA batch-release site, or where any manufacturing or release activity happens at a site other than the one with overall responsibility for the batch, responsibility for taking and storing samples is defined in a written agreement under Chapter 7 (6.1, 8.4). The QP who certifies a batch should ensure that all relevant samples are accessible at all reasonable times (6.2). The same allocation questions come up in our review of CDMO quality agreement red flags.

Stability samples are not reserve samples: maintaining the purpose firewall

Stability samples and retained samples do different jobs, and the source texts say so. ICH Q7 section 11.70 is explicit for APIs: "The packaging and holding of reserve samples is for the purpose of potential future evaluation of the quality of batches of API and not for future stability testing purposes." The 2015 Q&A draws the contrast directly: unlike stability samples, the reserve sample is not meant to represent the quality of the batch in the market place but to allow future evaluation of the quality of the original batch.

Formal stability studies under ICH Q1A(R2) work differently. They follow a protocol on designated batches: at least three primary batches, two of which should be at least pilot scale for a drug product. Where the submission does not already include long-term data on three production batches covering the proposed period, the applicant commits to continuing or placing production batches on long-term study up to a total of at least three; with no production-batch data at all, that means the first three production batches. For a product with a proposed shelf life of at least 12 months, long-term testing normally runs every 3 months in the first year, every 6 months in the second year, and annually thereafter. Those pulls are scheduled, and the material is consumed by design.

Retained samples are held for questions nobody can schedule: a complaint, a query about marketing-authorization compliance, a labeling query, a pharmacovigilance report (Annex 19 section 2.3). Using them to fill a missed or damaged stability pull mixes the two purposes. The stability record then contains a unit that was not stored as the protocol required, a lost time point that should have been handled as a deviation is hidden, and the quantity sized for two future rounds of testing shrinks. A lost stability time point belongs in the stability deviation process, not in the retained-sample store. Where a stability signal is already under investigation, our article on stability out-of-trend signals covers how those investigations affect launch supply.

Parallel trade and digital samples: the revised section 9 rules

Section 9 is where the 2026 text changes obligations. It applies to parallel imported, parallel distributed and, newly named, parallel traded products: products authorized in one market that are re-packaged for another, for example with new labels, a new carton, or a patient information leaflet in another language. The 2006 text already required one retention sample per packaging operation when the secondary packaging was opened. The revision sets out a fuller set of rules:

• Re-packaging materials (9.1): physical samples of the packaging materials used in re-packaging, such as labels, carton, patient information leaflet and other package inserts, are retained for the shelf life of the re-packaged finished product.

• No reference samples (9.2): reference samples of the re-packaged product are not required.

• One retention sample per operation (9.3 and 9.4): a retention sample of the re-packaged finished product is taken for each re-packaging operation and kept for at least one year after the expiry date. It should represent the product as released for the market and include both the primary and the secondary package. Where the secondary package is not opened, only the packaging material used needs to be retained.

• Photographic or digital sample (9.5): permitted only where it is duly justified that a physical retention sample of the re-packaged product cannot reasonably be retained, and the approach has been agreed in advance with the competent authority.

A worked sample plan record you can adapt

The workflow and record below apply the rules to the scenario at the top of this article. They are illustrative: the quantities, storage conditions and site roles are assumptions for a hypothetical product, not values from any real file. The figures in a real plan come from the approved specification, the marketing authorization and the written agreements.

flowchart TD
    A["Batch manufactured or re-packaged"] --> B{"What is being sampled?"}
    B -->|"Starting material lot"| C["Reference sample, 3.2"]
    C --> C1["Keep at least 2 years after finished product release"]
    C1 --> C2["Hold at the site where the material was used, 7.2.3"]
    B -->|"Critical intermediate"| I["Reference sample where stability permits, 2.1"]
    B -->|"Finished product"| D["Reference sample sized for full controls twice, 4.1"]
    D --> D1["Unopened packs for each set where necessary"]
    D1 --> D2["Keep at least 1 year past expiry, 3.1"]
    B -->|"Each packaging operation"| E["Retention sample, 4.3"]
    E --> E1["Fully packaged unit held in the EEA, 8.1"]
    E1 --> E2["Keep at least 1 year past expiry, 3.1"]
    B -->|"Packaging material batch"| K["Reference sample at the packaging site, 2.2"]
    K --> K1["Keep for the finished product shelf life, 3.2"]
    B -->|"Parallel trade re-packaging"| F["Section 9 rules"]
    F --> F1["Keep re-packaging materials for shelf life"]
    F1 --> F2["No reference sample of re-packaged product"]
    F2 --> F3{"Physical retention sample feasible?"}
    F3 -->|"Yes"| F4["One per re-packaging operation, 1 year past expiry"]
    F3 -->|"No, and agreed in advance"| F5["Photographic or digital sample, 9.5"]

The record below sets out the plan line by line for the 36-month tablet. Section references are to Annex 19 unless marked as US.

Annex 19 section 2.4 asks for records of traceability of samples that competent authorities can review. For each line, the record should carry at least the batch or lot number, the sampling point and date, the quantity and number of packs taken, the storage location, the condition, the retention end date and the rule that set it, and every withdrawal with its purpose and remaining balance. The sampling instructions belong in the master record and the evidence that they were followed belongs in the executed one; our explainer on master versus executed batch records covers that distinction. Sample plans are easiest to settle before the first commercial batches. If those batches are also process performance qualification batches, see our article on PPQ protocols versus commercial batch release records.

Annual visual examination, closedown and inspection exposure

Retained samples are not a set-and-forget archive. Two obligations recur after the samples are put away: a periodic look at the samples themselves, and a plan for what happens to them if a site closes.

For US drug products, 211.170(b) adds a recurring duty that the EU annex does not set out in the same form: "reserve samples from representative sample lots or batches selected by acceptable statistical procedures shall be examined visually at least once a year for evidence of deterioration unless visual examination would affect the integrity of the reserve sample." Radioactive drug products described in 211.170(b)(2) are excepted. Any evidence of deterioration is investigated in accordance with 211.192, and the results of the examination are recorded and maintained with the other stability data on the drug product. The plan should therefore name the statistical basis for choosing lots, what the examiner looks for in each dosage form, and how a pack is examined without compromising it, or why it cannot be.

Section 10 of Annex 19 is unchanged from 2006 but tends to be read only when it is needed. When a manufacturer closes and its manufacturing authorization is surrendered, revoked or ceases to exist, unexpired batches are likely to remain on the market. For them to stay there, the manufacturer should arrange transfer of reference and retention samples, with the relevant GMP documentation, to an authorized storage site, and satisfy the competent authority that the samples can be readily accessed and analyzed (10.1). If the manufacturer cannot make those arrangements, they may be delegated to another manufacturer. The MAH is responsible for that delegation and for giving the competent authority the information it needs, and should consult the competent authority of each Member State where an unexpired batch has been placed on the market (10.2). The same applies when a manufacturer outside the EEA closes; there, the importer has a particular responsibility to make sure satisfactory arrangements are in place and the authorities are consulted (10.3). In a contract manufacturing relationship, these arrangements belong in the quality agreement before they are needed.

Review triggers that should reopen the sample plan:

• Any further revision of Annex 19 or of the Note for Guidance on Declaration of Storage Conditions, amendments to 21 CFR part 211, or updates to ICH Q7 or Q1A(R2).

• Guidance or Q&As from EU authorities on photographic or digital retention samples under section 9.5.

• Transfer of manufacturing, packaging or release testing to another site, or retirement of an analytical method, instrument platform or reference standard.

• National law in the Member State of manufacture that extends starting-material retention beyond the two-year minimum.

• Deterioration found during the annual 211.170(b) examination, or a temperature or integrity problem in the sample store.

• A manufacturer, packager or importer in the supply chain announcing a closure.

Sources

This analysis is based on the following regulatory texts, harmonized guidelines and secondary reporting:

• European Commission — EudraLex Volume 4, EU Guidelines for GMP, Annex 19 Reference and Retention Samples (revised), Commission Decision C(2026) 4135 final, Brussels, 23 June 2026.

• European Commission — EudraLex Volume 4, Annex 19 Reference and Retention Samples (2006 version, 14 December 2005, in operation 1 June 2006).

• Electronic Code of Federal Regulations — 21 CFR 211.170 Reserve samples (current text).

• International Council for Harmonisation — ICH Q7 Good Manufacturing Practice Guide for Active Pharmaceutical Ingredients, section 11.7.

• US Food and Drug Administration — Q7 Good Manufacturing Practice Guidance for Active Pharmaceutical Ingredients, Guidance for Industry.

• International Council for Harmonisation — ICH Q1A(R2) Stability Testing of New Drug Substances and Products (February 2003).

• US Government Publishing Office — 21 CFR 600.13 Retention samples (April 2013 annual edition).

• World Health Organization — Guidelines for sampling of pharmaceutical products and related materials, WHO Technical Report Series No. 929, Annex 4.

• International Council for Harmonisation — ICH Q7 Questions and Answers (June 2015).

• ECA Academy — EU GMP Annex 19: Reference and Retention Samples (2026 Revision), guideline entry.

• GMP Insiders — Revised EU GMP Annex 19: What Changes for Reference and Retention Samples?