Please confirm you are 21 or older to continue.
This page is an editorial and informational resource about stability and storage: what a date printed on a label actually promises, how that date is established from measurements taken over months and years, and the specific ways the conditions a material is kept in can quietly make the date untrue.
Nothing is sold on this page. It is a published editorial resource. Nothing here is an offer, no account can be opened and no order can be placed on this site. We store nothing, test nothing and assign no dates to anything.
Nothing on this page describes, recommends, compares or makes any claim about any material, item, treatment or substance, and no such claim is made or implied anywhere on this site. This resource is written for readers aged 21 and over.
A retest date and an expiry date are different promises, and both carry conditions in small print. Outside those conditions the date is not shorter. It is simply no longer about the material.
Heat, moisture, light and oxygen do almost all of the work. The container decides how much of each reaches the contents, which is why the container is part of the claim.
The date is where the lower confidence bound meets the limit, not where the average line does. The difference is often a year, and it is the year that gets quietly borrowed.
A warm afternoon on a loading dock is not a disaster and not nothing. Mean kinetic temperature turns it into a number, and a record turns the number into a decision.
Every date printed on a container is a prediction. It says that if the contents are kept in a particular way, they will still meet a particular description on that day. Both halves of the sentence matter, and only one of them is ever printed in large type. The date is easy to read. The conditions it depends on are usually a single line in small print, and the description it promises to meet is usually not on the label at all.
Behind a well-founded date sits a stability study: the same material held under fixed conditions for a long time, sampled at planned intervals, and measured each time with a method capable of seeing whether anything has changed. Behind a poorly founded date sits a guess, a habit, or a figure copied from a similar material. The two dates look identical once they are printed. The only way to tell them apart is to ask how the date was reached.
What follows is a reference guide to that reasoning: the difference between the kinds of date in common use, the four agents that change a material while it sits still, how long-term and accelerated studies are designed and why their results cannot simply be swapped, what a test method has to be able to do before its results mean anything over time, how a date is read off stability data honestly, what the container contributes, how to think about a temperature excursion without either panicking or shrugging, and how a small setting can store materials so that the dates on them remain true.
It is written for small settings, where there is one refrigerator, one or two people and no separate stability department. It is deliberately general. It does not describe, recommend, compare or make any claim about any material, item, treatment or substance, and it never tells a reader what to do with one.
In ordinary speech every printed date is an "expiry". In a record they must be kept apart, because they promise different things and they are handled differently when they pass.
| Date | What it promises | What happens when it passes |
|---|---|---|
| Retest date | The material is expected to remain within its description until this date, under the stated conditions. After it, the material must be examined again before it is relied on. | Nothing automatic. The material is tested again; if it still meets its description, a new retest date can be assigned. It is a checkpoint, not an end. |
| Expiry date | The material will remain within its description until this date, under the stated conditions, and is not to be relied on after it. | The material is finished. Retesting it does not extend the date, because the date was set to cover the whole population, not the one container in hand. |
| In-use period | After the container is first opened, or after the contents are prepared into another form, they remain within description for a stated period. | Whichever comes first, the in-use period or the printed date, governs. It starts on the day of opening, which therefore has to be written down. |
The third row is the one most often missing from a record. A sealed-container date says nothing at all about a container that has been opened, because opening changes exactly the things the date depended on: the headspace, the moisture, and the number of times the contents meet air.
Every one of the three carries its conditions with it. "Store at 2 to 8 degrees, protect from light" is not advice attached to a date. It is part of the date. A container kept at room temperature for its whole life does not have a shorter version of the printed date. It has no date at all, because nobody ever measured what happens under those conditions.
A material on a shelf looks inert. It is not. Almost all of the change that a stability study is designed to detect is driven by four agents, and it is worth knowing them separately because each one is controlled by a different part of the storage arrangement.
A fifth influence is less obvious and just as real: the container itself. Materials can adsorb onto container walls, react with them, draw components out of a stopper, or lose volatile parts through a closure that was never completely tight. A stability study is therefore always a study of the material in a particular container. Transferring the contents into something else, however clean, leaves the date behind.
A stability study holds samples from a defined batch under defined conditions and pulls portions for testing at planned intervals. The intervals are fixed before the study starts, typically at zero, three, six, nine, twelve, eighteen and twenty-four months, then yearly. The zero point matters as much as any other, because every later result is read as a change from it.
Two families of condition are used, and they answer different questions.
| Study | Typical condition | Question it answers |
|---|---|---|
| Long-term | The intended storage condition, for example 25 °C and 60 % relative humidity, 5 ± 3 °C, or −20 ± 5 °C. | What actually happens under the conditions the label will state. This is the only study that can support a date directly. |
| Intermediate | A moderately raised condition, for example 30 °C and 65 % relative humidity. | Whether a change seen under accelerated conditions also appears nearer to real storage, or only at the extreme. |
| Accelerated | A deliberately harsh condition, for example 40 °C and 75 % relative humidity, usually for six months. | Which changes are possible and how quickly they begin. It is an early warning and a way of ranking risks, not a substitute for time. |
For materials stored cold, the "accelerated" condition is usually room temperature, and the long-term condition is the refrigerator or freezer. The logic is the same: the harsher condition shows what can happen, the intended condition shows what does.
The temptation is obvious. Six months at forty degrees is cheap and quick; three years at five degrees is slow and expensive. If the harsh condition could stand in for the gentle one, every date could be set in half a year. Sometimes it roughly can. Often it cannot, and the next section is about why.
The reason accelerated studies are tempting is a well-established relationship between temperature and reaction rate, usually named after Arrhenius. For a single, simple reaction, the rate rises exponentially with temperature in a predictable way, and a common rule of thumb says that each ten-degree rise roughly doubles or triples the rate. On that basis, six months at forty degrees might be read as the equivalent of several years at twenty-five.
The rule is sound for the situation it describes, and the situation is narrower than it looks. It holds when one reaction, and the same reaction, dominates at every temperature in the range. It breaks in at least four common ways.
The honest use of accelerated data is therefore narrow. It shows which changes are possible, it ranks candidate containers and conditions against each other, and it supports a short provisional date that is confirmed or withdrawn as long-term data arrives. A date that rests on accelerated data alone should say so, and it should be treated as provisional until the long-term study has reached it.
Here is the failure that makes every other part of a stability study meaningless. The samples are pulled on schedule, the measurements are made, and every result is the same as the zero point. The material is declared stable. The true explanation is that the method could not tell the original material apart from what it had turned into.
A test method used for stability has to be stability-indicating: able to separate the intact material from the products of its own change, and to measure each of them. A method that reports a single total, or that lets a changed form appear in the same place as the original, will report perfect stability for a material that is quietly becoming something else.
The way this is demonstrated is forced degradation, sometimes called stress testing. Before the study starts, portions of the material are deliberately damaged: heated, exposed to strong light, to acid and to base, to an oxidizing agent, to high humidity. The point is not to learn how the material behaves in storage. The point is to produce the changed forms on purpose and to confirm that the method can see them, separate them and measure them.
This is the stability version of a principle that applies to every negative result: a finding of "no change" is only worth something if the method has been shown able to detect a change when one is present. Without the stress test, a flat line on a stability chart is not evidence of stability. It is evidence of a flat method.
Once long-term results exist at several pull points, the date is estimated from them. The careless way to do it is to fit a straight line through the results and read off where the line meets the specification limit. The careful way is the same, with one important addition.
The fitted line is an estimate of the average trend. Individual containers, and future batches, will scatter around it. The date is therefore set where the one-sided 95 % confidence bound of the line, rather than the line itself, meets the limit. For a property that falls over time, that means the lower bound. Because the bound widens the further it is taken from the data, it reaches the limit earlier than the line does, often considerably earlier.
| Reading | Where the date is set | What it implies |
|---|---|---|
| Average line | Where the fitted trend crosses the limit. | About half of containers will be outside the limit on that day. It is the date by which the typical container has already failed. |
| Confidence bound | Where the one-sided 95 % bound of the trend crosses the limit. | The average is expected to remain within the limit with high confidence. This is the defensible reading. |
| Last good pull point | The latest time at which every result was within the limit. | Conservative and simple, and appropriate when there are too few points to fit anything at all. |
The gap between the first two rows is the year that gets quietly borrowed. Nobody writes it down as a decision. It simply happens when a line is read instead of a bound.
Two further points belong in any honest reading. First, extrapolation beyond the last real data point should be limited and stated. A date twice as long as the longest observation is a forecast, not a finding. Second, where several batches have been studied, their results should only be pooled into one line if a statistical test shows their slopes and starting points do not differ meaningfully. If they do differ, the date comes from the worst batch, not the average of all of them.
A date established for one container and closure does not transfer to another. The glass, the stopper, the seal and the fill level all control how much of each of the four agents reaches the contents, and a change to any of them is a change to the study.
For materials that are opened and used repeatedly, it is often better to divide them once, at the start, into several smaller sealed containers. Each is then opened only when it is needed, and the contents of the others are never exposed. This practice is sometimes called aliquoting. It turns one in-use period into several, each starting only when that portion is first opened.
Sooner or later every storage arrangement has an excursion: a delivery left on a step, a refrigerator door not quite closed, a power cut overnight. The two instinctive responses are both wrong. One treats any excursion as ruinous and discards everything. The other treats every excursion as trivial and records nothing. The useful response is to measure it.
The tool for this is mean kinetic temperature, or MKT. It is a single temperature that expresses the cumulative effect of a varying temperature history on the rate of change. Because rates rise exponentially with temperature, a warm period counts for more than an equally long cool period, and MKT weights accordingly. It is always higher than the simple average of the same readings.
MKT has one important limitation. It assumes that heat is the only thing that changed. A freeze, a period of high humidity, or exposure to light is not captured by a temperature average of any kind, and each has to be judged on its own. A refrigerated material that froze overnight has not had a "cold excursion". It has had a different event entirely, and freezing is excluded by many storage statements for good reason.
In a large operation, storage is a qualified room with alarms and a mapped temperature profile. In a small setting it is a refrigerator, a freezer and a cupboard. The principles are the same, and most of them cost nothing.
Two conventions govern what gets used first. First in, first out uses the oldest stock first. First expiring, first out uses whatever has the nearest date first. They agree most of the time. They disagree when a newer delivery carries an earlier date, which happens more often than anyone expects, and when they disagree the second rule is the right one.
A date is only as good as the record that shows the conditions were met. In a small setting, a handful of entries is enough, provided they are made at the moment rather than reconstructed at the end of the week.
None of these entries takes longer than half a minute. Together they turn a printed date from something the reader has to trust into something the reader can check.
Read any printed date against these ten questions. Each one can be answered from the label, the documentation and the storage record alone. Any answer that depends on somebody remembering something is a failed answer.
A date that answers all ten is not thereby guaranteed. It is checkable, which is a different and more useful property. A date that cannot answer them depends entirely on the storage history being what everybody assumes it was, and that assumption does not appear anywhere on the label.
The argument of this guide is narrow and it is not a criticism of anybody's measurements. It is that a printed date is a conditional statement, and that the conditions are the part most often lost between the study that produced the date and the shelf where it is read. How the date was estimated, what container it belongs to, what temperature the material actually saw, and whether anybody wrote down the day it was opened. Those facts are cheap to record at the moment and impossible to recover later.
Which is why the date on the label is only half of the claim. The other half is how the container was kept.
Who publishes it, why it exists, and how to reach the editor with a correction or a question about the guide.
This is an editorial resource on stability and storage: what a printed date actually promises, how long-term and accelerated studies establish it, why a method has to be shown able to see change before its results mean anything, how a date is read off data honestly, and how a small setting can keep materials so that the dates on them stay true. It exists because the conditions behind a date are the part most often lost on the way to the shelf. The position behind the page is a plain one: a date describes a material kept in a stated way, and nothing else. Everything that keeps it true happens after the label is printed.