Expiry date or retest date? What each one actually states

The difference is not a matter of terminology; it is a matter of who carries the burden of proof once the date passes. An expiry date is a full stop: after it the material is withdrawn, and the recipient cannot extend it by running an analysis of their own. A retest date is a checkpoint: after it the material is not useless — it is uncharacterised. It is re-examined against the specification and, if it conforms, it may continue to be used and receive a new retest date. ICH Q1A(R2) defines the second as the date after which samples of the drug substance should be examined to ensure that the material still complies with the specification and therefore remains suitable for use ^[2].

For anyone receiving research material the practical consequence is simple and unwelcome. While the retest date holds, the manufacturer has demonstrated with data that the material stands. The day after it, responsibility moves: if no analysis is performed, you do not hold material that has expired — you hold material of unknown characterisation, which in a serious record should be written down in exactly those words. The commonest failure is not that someone uses expired material; it is that someone copies a retest date into the "expiry" field of their register and loses the distinction permanently.

The two definitions, as the source texts write them

ICH Q7 — the text that in the European Union is incorporated as Part II of EudraLex Volume 4 and governs the manufacture of active substances ^[5] — devotes section 11.6 to the question, under the heading "Expiry and Retest Dating". Two of its points deserve particular attention. First, that any such date should rest on an evaluation of data from stability studies, not on a round number of months counted from manufacture. Second, that common practice for active substances is a retest date rather than an expiry date, while especially labile materials — the text names antibiotics and biotechnological products explicitly — may be required to carry an expiry date ^[1].

Why starting materials usually carry a retest date

The reasoning behind the distinction is more interesting than the distinction itself. A finished medicinal product is a defined formulation, in a defined container, intended for a defined purpose; so the question "how long does it hold?" has one answer, and that answer is written as an expiry date. An active substance, by contrast, is input material: it will enter processes its own manufacturer does not know about, and its fitness is defined only against the specification. A checkpoint that asks "are you still within specification?" carries more information than a date that says "stop". There is also the practical reason, which nobody states and everybody means: a retest date prevents the destruction of material that still meets its own specification.

What a retest actually is, and who can perform one

A retest is not a visual inspection and not an estimate. It is fresh analysis of the critical attributes against the same specification: identity, content, purity and related substances — where it is worth knowing in advance what a purity figure measures and what it leaves out — and water content where that matters. ICH Q7 provides that where the results conform, the batch may be used immediately, and that the new retest date is set on the basis of the data ^[1]. For a research laboratory the question is not whether a retest is permitted — it is whether one can genuinely be performed. An analysis is worth exactly as much as the laboratory that produced it: ELOT EN ISO/IEC 17025, the Greek adoption of the international standard, requires documented and validated methods and technical records sufficient for the test to be repeated under the same conditions ^[6]. In Greece accreditation against that standard is granted by ESYD, and the useful question to put to a laboratory is not whether it is accredited but whether the specific method sits inside its published scope of accreditation.

• The specification the retest is run against must be the one that applied originally; a retest against looser limits is not a retest.
• The sample is drawn from the same container that will be used, not from a "representative" container of another batch.
• The new retest date is set from stability data, not by adding an arbitrary interval to the previous one ^[1].
• The result is recorded alongside the original batch entry, under the same batch number, with the date it was performed ^[6].
• An out-of-specification result is neither deleted nor repeated until it comes out conveniently; it is recorded, and the procedure you defined for such results is followed ^[6].

The costly mistake: an expiry date cannot be extended

If the label says "expiry" and the certificate of analysis states the same, no analysis on the recipient's side brings it back. The distinction is not formalism: a manufacturer who chose an expiry date is stating that they know something about the degradation mechanism of the material which makes continued use indefensible, whatever a single measurement happens to show. Three further rules travel with the first and are rarely written down. An aliquot never acquires a longer life than the parent batch — dividing material does not manufacture time. Reconstitution or dilution voids the certificate's characterisation, because the contents are no longer in the form that was studied, and the certificate's dates do not transfer to the new solution. And a retest date nobody intends to honour is, in practice, an expiry date; the honest entry in the register says so.

Where the numbers come from, and the limits on extrapolation

Both dates come out of stability studies, and here the standards become unusually specific. The WHO stability guideline for active pharmaceutical ingredients asks, for new substances, long-term data covering at least twelve months at the time of submission, continued for long enough to cover the proposed retest period, while for established substances it accepts a minimum of six months ^[4]. ICH Q1E sets the ceiling on extrapolation: where long-term data are amenable to statistical analysis, the proposed retest period or shelf life can be up to twice, but should not be more than twelve months beyond, the period covered by the long-term data; without statistical analysis, but with relevant supporting data, the limit drops to one and a half times and no more than six months beyond that period ^[3]. Translated into a practical check: a thirty-six month retest period cannot be built on six months of measurements, however attractive the curve looks.

It is worth noting that this framework is in transition. ICH is consolidating the Q1A to Q1F series together with Q5C into a single Q1 text; the draft went out for public consultation at Step 2b in April 2025 and had not reached Step 4 at the time of writing ^[10]. Until it does, the texts cited here remain the applicable ones.

The peptide as chemistry: why no date is a guarantee

A date on a label is a prediction under conditions. Peptides do not degrade by one route but by several degradation pathways running in parallel — deamidation of asparaginyl and glutaminyl residues, oxidation of methionine, cysteine and tryptophan, hydrolysis of peptide bonds, aggregation — and their relative speed depends on pH, moisture, temperature and on whether the material sits in the solid or the aqueous state; the standing review of protein pharmaceutical stability documents precisely this interplay of chemical and physical instability ^[7]. The scale of the dependence on sequence alone is striking: in a study that measured deamidation rates for 306 asparaginyl sequences in model peptides by mass spectrometry, the first-order half times ranged from roughly one day to 455 days ^[8]. Two orders of magnitude, decided by the neighbourhood of a single residue.

Honesty requires saying what that finding does not show. It is an in vitro experiment, in buffer at pH 7.4 and 37.0 °C, using small model peptides — conditions chosen to accelerate and isolate the mechanism, not to imitate a lyophilised solid in a freezer. The mechanism transfers; the numbers do not. And there is a gap no citation closes: no published stability dataset exists for research-grade lyophilised peptides as a class. Every expiry or retest date on such material is one manufacturer's claim about one batch, and it is worth exactly as much as the study behind it — a study you can, and should, ask to see.

The Greek and European frame, and the format of the date

In European law the expiry date is a labelling obligation on human medicines: Article 54 of Directive 2001/83/EC lists the particulars that must appear on the outer packaging and places the expiry date, stated in clear terms as month and year, at point (h), with the manufacturer's batch number at point (m) ^[9]. Research material is not a medicinal product and does not fall under those provisions — EOF, the Greek national medicines agency, supervises medicines, not research reagents. That is precisely why the wording matters: a research-use label that borrows pharmaceutical vocabulary invites the classification nobody wants. Describe what the supplier's data show, without implying a status the material does not hold.

Last, and dull until the moment it costs something: the format. A marking of "EXP 03/2027" does not identify a day, and two laboratories can quite reasonably read it differently — one as 1 March, the other as 31 March. ISO 8601-1:2019 gives the unambiguous form, largest element to smallest, and accepts reduced precision as YYYY-MM where the day is not known ^[11]. An internal register either holds a full date or states explicitly how a month without a day is interpreted. And it holds, next to the number, the type: a date without a note of whether it is an expiry or a retest date is half a piece of information.

• Ask in writing which of the two date types is stated; if the certificate says "retest" and the label says "expiry", the discrepancy is the finding.
• Check that the batch number on the certificate matches the label character for character before trusting any date on it.
• Record the date type in the register as a separate field, not as a note beside the number.
• If the date is a retest date and you do not intend to run an analysis, write that decision down explicitly: after the date the material is treated as uncharacterised.
• Record the storage conditions as a numeric temperature range; a date without the conditions under which it holds means nothing ^[2].