Skip to main content
v2026.11,986 entries · CC-BY 4.0

ICH Q1E: Evaluation of Stability Data and Shelf-Life Extrapolation

ICH Q1E explained: how stability data set a retest period or shelf life, the extrapolation limits by storage class, and the 95% confidence limit method.

Written and maintained by CASRAI Editorial Board

Published · Last updated

On this page

ICH Q1E, titled “Evaluation of Stability Data”, is the ICH quality guideline that tells a sponsor how to turn stability results into a proposed retest period for a drug substance or a shelf life for a drug product, and how far that period may be extended past the long-term data actually in hand. Its core rule is that each quality attribute is assessed separately, the proposed period can never exceed what any single attribute supports, and any extrapolation beyond the long-term data is capped by the storage class, the behaviour of the accelerated and intermediate data, and whether a statistical analysis backs the proposal. The Step 4 text is dated 6 February 2003.

What ICH Q1E covers

Q1E is a companion to ICH Q1A(R2), which it calls the “parent guideline”. Q1A(R2) sets out how to design and run formal stability studies; Q1E explains how to evaluate the resulting data when proposing a retest period or shelf life in a registration application. According to its own introduction, Q1A’s evaluation guidance is “brief in nature and limited in scope”, so Q1E expands it, particularly for studies with several factors (for example strength and container size) and for reduced designs. The overview of the wider guideline family is on the ICH Q1 stability testing entry.

The scope statement limits the guideline to new molecular entities and their drug products, for storage at or below “room temperature”, across single-factor and multi-factor designs and across full and reduced designs. The guideline adds a footnote that “room temperature” describes the general customary environment and must not be read as the storage statement for the label. Acceptance criteria themselves are outside Q1E: it sends the reader to ICH Q6A and Q6B for setting and justifying them, and to ICH Q1D for choosing between full and reduced designs. Q1E was developed by the ICH Expert Working Group and issued as a harmonised tripartite guideline, recommended for adoption by the regulators of the European Union, Japan and the United States.

How the evaluation works, step by step

1. Start from at least three batches

Q1E states that a stability study establishes a retest period or shelf life, and label storage instructions, “based on testing a minimum of three batches” that apply to all future batches made and packaged under similar circumstances. The more the batches differ from one another, the less confidence there is that a future production batch will stay inside its acceptance criteria.

2. Assess every attribute on its own

The data to be presented include physical, chemical, biological and microbiological results, plus dosage-form attributes such as dissolution for solid oral forms, and the adequacy of the mass balance should be assessed. The guideline then states the governing rule: the proposed retest period or shelf life “should not exceed that predicted for any single attribute”. A product with an excellent assay profile and a degradant that creeps toward its limit is therefore dated by the degradant.

3. Know which attributes suit statistics

Q1E says certain quantitative chemical attributes, namely assay, degradation products and preservative content, can generally be assumed to follow zero-order kinetics in long-term storage, which makes them amenable to linear regression and poolability testing. Other quantitative attributes such as pH and dissolution have generally unknown kinetics, but the same analysis can be applied if appropriate. Qualitative and microbiological attributes “are not amenable” to this kind of analysis. The guideline also says statistics are not preferred where they can be justified as unnecessary.

4. Walk the accelerated, intermediate and long-term data in order

For room-temperature products the assessment begins with whether there was significant change at the accelerated condition and, if appropriate, at the intermediate condition, and then moves on to the trends and variability of the long-term data. Appendix A of Q1E is a decision tree for this sequence, covering everything except frozen products. Throughout it, X is the period covered by long-term data and Y is the proposed retest period or shelf life.

5. Watch the release assay

Q1E warns that a drug product should be formulated to deliver 100 percent of the labelled amount at release. If the batches in the application have release assay values above 100 percent of label claim, after allowing for manufacturing and analytical variation, the proposed shelf life “can be overestimated”; if a batch releases below 100 percent it may reach the lower limit before the end of the proposed shelf life.

Extrapolation limits by storage class

Extrapolation is defined in Q1E as using a known data set to infer information about future data. It may be proposed “particularly if no significant change is observed at the accelerated condition”, and whether it is appropriate depends on knowledge of the change pattern, the goodness of fit of any model, and the existence of relevant supporting data. The data table below lists every limit stated in Sections 2.4 and 2.5, with its scenario and the section where it appears. This page does not repeat the table in the body; it explains how to read it.

Four patterns recur across the room-temperature and refrigerated limits.

  • Two ceilings on every extension. Each limit is a multiple of X and an absolute number of months beyond X, and the lower of the two governs. “Up to twice, but not more than 12 months beyond” means the smaller of 2X and X + 12.
  • Statistics buy a bigger extension. For room-temperature products with change or variability in the data, an unanalysed proposal is limited to 1.5X (not beyond X + 6), while a proposal backed by a statistical analysis and supporting data can reach 2X (not beyond X + 12).
  • Supporting data are defined. Q1E counts satisfactory long-term data from development batches that have a closely related formulation, a smaller manufacturing scale, or a similar container closure system as relevant supporting data.
  • Significant change at the accelerated condition cuts the ceiling. If the intermediate condition then shows no significant change, the unanalysed ceiling falls to X + 3 months and the analysed ceiling to 1.5X (not beyond X + 6). If the intermediate condition also shows significant change, Y should not exceed X, and a shorter period could be called for.

For refrigerated products Q1E applies the same principles “except that the extent of extrapolation should be more limited”, and a significant change at the accelerated condition ends extrapolation: the period is based on long-term data, and if the change appears within the first 3 months a discussion of short-term excursions such as shipping or handling should be added, which may be supported by testing one batch at the accelerated condition for under 3 months. For freezer storage the period is based on long-term data, with one batch tested at an elevated temperature (the guideline gives 5 °C ± 3 °C or 25 °C ± 2 °C as examples) to address excursions; below -20 °C the period rests on long-term data and is assessed case by case.

Worked arithmetic: which bound governs

The figures below are computed here from the rules in Section 2.4 of Q1E. They are an illustration of the arithmetic only, not examples taken from the guideline, and they assume room-temperature storage and that the stated conditions for each row are met.

Long-term data (X) Scenario Multiplier cap Month cap Maximum Y
12 months No significant change at accelerated; little change and variability 2X = 24 X + 12 = 24 24 months
12 months Change or variability; no statistics; supporting data 1.5X = 18 X + 6 = 18 18 months
12 months Significant change at accelerated, none at intermediate; no statistics not applicable X + 3 = 15 15 months
6 months No significant change at accelerated; little change and variability 2X = 12 X + 12 = 18 12 months (2X governs)
24 months No significant change at accelerated; little change and variability 2X = 48 X + 12 = 36 36 months (X + 12 governs)

The pattern is that the multiplier governs for short data sets and the month cap governs once the long-term data exceed 12 months. The two ceilings coincide at X = 12 months for the 2X rule and at X = 12 months for the 1.5X rule too, which is why a 12-month data package maps so neatly onto 24-, 18- and 15-month proposals. These are upper limits for what “can be” proposed; Q1E does not say a sponsor is entitled to the maximum, and every proposal still has to be supported by the attribute-by-attribute evaluation described above.

The statistical approach: 95% confidence limits and poolability

Regression and the confidence limit

Q1E treats regression analysis as an appropriate method for a quantitative attribute. The recommended approach is to find the earliest time at which the 95 percent confidence limit for the mean intersects the acceptance criterion. Which limit to use depends on direction: the lower one-sided 95 percent limit for an attribute known to decrease, the upper one-sided limit for one known to increase, and two-sided limits where the direction is not known. If the approach is followed, the guideline says the mean can be expected to remain within acceptance criteria through the end of the period at a 95 percent confidence level. The general idea behind such limits is covered in Confidence Interval Explained.

Appendix B gives two single-batch examples with 12 months of long-term data and a proposed shelf life of 24 months. For an assay with limits of 95 to 105 percent of label claim, two-sided limits were used because the direction was unknown, and the lower confidence limit crossed the lower criterion at 30 months, supporting 24 months. For a degradation product with a limit of not more than 1.4 percent, the upper one-sided limit crossed the criterion at 31 months, again supporting 24 months. In both, the guideline qualifies the conclusion as holding “provided the recommendations in Sections 2.4 and 2.5 are followed”, so the extrapolation limits still apply.

Poolability of batches

Before data from several batches are combined, Q1E expects a preliminary test of whether the regression lines share a common slope and a common time-zero intercept. Analysis of covariance, with time as the covariate, can be used, and each test is run at a significance level of 0.25 to compensate for the expected low power of a typical stability study. The same 0.25 level appears in Q1A(R2). The order and the outcomes are specified.

  1. Test slopes before intercepts.
  2. If slopes differ significantly, do not pool. Estimate each batch with its own intercept and slope, using the pooled mean square error, and take the shortest batch estimate as the period for all batches.
  3. If slopes are equal but intercepts differ, combine data for a common slope, estimate with individual intercepts, and again take the shortest estimate.
  4. If neither test rejects, combine all batches. The pooled estimate is usually longer because the confidence limit narrows as data accumulate.

There is a shortcut. If every batch’s own estimated period already exceeds the proposed one, Q1E says there is generally no need for poolability tests. For multi-factor designs, the guideline says data from different factor combinations should not be combined unless supported by statistical tests for poolability; analysing batch poolability only, separately for each combination, tends to give shorter shelf lives than testing all factors together.

Commitment batches and verification

Extrapolation assumes the observed change pattern continues. Within the long-term data, the data check that assumption; beyond them, no internal check is possible. For that reason Q1E states that a period granted on the basis of extrapolation “should always be verified by additional long-term stability data” as it becomes available, and the protocol for commitment batches should include a time point that corresponds to the end of the extrapolated period. Where a statistical method was used to evaluate the long-term data, the same method should be used on commitment batch data to verify or extend the approved period. Trend monitoring after approval is a separate discipline; see out-of-trend investigation for how statistical trend limits are used on ongoing stability results.

Common mistakes

  • Dating the product by its best attribute. The proposal cannot exceed what any single attribute predicts, so one trending degradant limits the entire shelf life.
  • Applying regression to attributes it does not suit. Qualitative and microbiological attributes are not amenable to this analysis, and pH or dissolution are only analysed this way “if appropriate”.
  • Treating the multiplier as the answer. The month cap (X + 12, X + 6 or X + 3) often binds first, as the worked arithmetic shows.
  • Skipping the justification for omitting statistics. Where the data show little change and little variability, statistics are normally unnecessary, but Q1E expects a justification of the omission.
  • Pooling in the wrong order or at the wrong level. Slopes are tested before intercepts, at the 0.25 level, and unequal slopes mean no pooling.
  • Ignoring release assay position. Release values above 100 percent of label claim can inflate the projected shelf life.
  • Leaving out the end-of-period time point. A commitment-batch protocol without a time point at the end of the extrapolated period cannot verify the extension.
Guideline Relationship to Q1E, as stated in the guidelines
ICH Q1A(R2) Stability Testing of New Drug Substances and Products The parent guideline. It sets the long-term, intermediate and accelerated conditions (for the general case, long-term 25 °C/60% RH or 30 °C/65% RH, intermediate 30 °C/65% RH, accelerated 40 °C/75% RH), requires long-term data of at least 12 months on at least three primary batches at submission, and carries the brief evaluation text that Q1E expands.
ICH Q1D Bracketing and Matrixing Designs Referenced by Q1E for recommendations on full versus reduced design studies.
ICH Q6A and Q6B Specifications Referenced by Q1E for setting and justifying acceptance criteria.
Consolidated ICH Q1 draft A single revised Q1 text intended to bring the Q1A to Q1E series and Q5C into one guideline was a draft in consultation and had not reached Step 4 when last recorded on 2026-07-17, so the Step 4 text of Q1E described here remains the published version.

For post-approval changes to what was approved, see ICH Q12 lifecycle management. The body that issues these guidelines is described at ICH (International Council for Harmonisation).

FAQ

What does ICH Q1E cover?

Q1E explains how to evaluate stability data when proposing a retest period for a drug substance or a shelf life for a drug product in a registration application. It expands the brief evaluation text in Q1A(R2) and applies to new molecular entities and their drug products stored at or below room temperature, across single-factor and multi-factor designs and across full and reduced designs.

How far can a shelf life be extrapolated beyond the long-term data?

For a room-temperature product with no significant change at the accelerated condition and little change and variability in the long-term data, up to twice the period covered by long-term data (2X), but not more than 12 months beyond it (X + 12). Where there is change or variability, an unanalysed proposal is limited to 1.5X (not beyond X + 6), while a statistically analysed proposal with supporting data can reach 2X (not beyond X + 12). The lower of the multiple and the month cap governs. Refrigerated products receive more limited extrapolation, and frozen products are based on long-term data.

When is statistical analysis needed?

Q1E treats regression as appropriate for quantitative attributes such as assay, degradation products and preservative content, which can generally be assumed to follow zero-order kinetics. Statistics are not preferred where they can be justified as unnecessary, for example when the data show little change and little variability, but the justification for omitting them is expected. Qualitative and microbiological attributes are not amenable to this analysis.

What significance level applies to poolability testing?

Q1E specifies a significance level of 0.25 for the preliminary tests of common slope and common time-zero intercept, to compensate for the expected low power of a typical stability study. Slopes are tested before intercepts. If slopes differ significantly, batches are not pooled and the shortest batch estimate sets the period.

Must an extrapolated period be verified?

Yes. Q1E states that a period granted on the basis of extrapolation should always be verified by additional long-term stability data as it becomes available, and the commitment-batch protocol should include a time point at the end of the extrapolated period.

Sources

  • ICH, Q1E Evaluation for Stability Data, Step 4 version, 6 February 2003: https://database.ich.org/sites/default/files/Q1E%20Guideline.pdf (read 2026-11-14).
  • ICH, Q1A(R2) Stability Testing of New Drug Substances and Products, Step 4 version, 6 February 2003: https://database.ich.org/sites/default/files/Q1A%28R2%29%20Guideline.pdf (read 2026-11-14).
  • FDA, Q1E Evaluation of Stability Data (FDA guidance page for the ICH text): https://www.fda.gov/regulatory-information/search-fda-guidance-documents/q1e-evaluation-stability-data (listed 2026-11-14).

Data: ICH Q1E extrapolation limits for retest period or shelf life, by storage class and data scenario

Every maximum extrapolation limit that ICH Q1E (Step 4, 6 February 2003) states in Sections 2.4 and 2.5 for proposing a retest period or shelf life beyond long-term data, arranged by storage class and by the behaviour of accelerated, intermediate and long-term data. X is the period covered by long-term data and Y is the proposed period. Compiled from the guideline text and its Appendix A decision tree.

Period covered: ICH Q1E Step 4 text dated 2003-02-06.

ICH Q1E extrapolation limits for retest period or shelf life, by storage class and data scenario
Storage classData scenarioStatistical analysisMaximum proposed period YQ1E section
Room temperatureNo significant change at accelerated; long-term and accelerated show little or no change and little or no variabilityNormally unnecessary (omission needs justification)Up to 2X, not more than X + 12 months2.4.1.1
Room temperatureNo significant change at accelerated; change or variability; data not amenable, or amenable but not analysedNot performed; relevant supporting dataUp to 1.5X, not more than X + 6 months2.4.1.2
Room temperatureNo significant change at accelerated; change or variability; data amenable and analysedPerformed, backing the proposal with supporting dataUp to 2X, not more than X + 12 months2.4.1.2
Room temperatureSignificant change at accelerated, none at intermediate; data not amenable, or amenable but not analysedNot performed; relevant supporting dataUp to X + 3 months2.4.2.1
Room temperatureSignificant change at accelerated, none at intermediate; data amenable and analysedPerformed, backing the proposal with supporting dataUp to 1.5X, not more than X + 6 months2.4.2.1
Room temperatureSignificant change at the intermediate conditionStatistics can verify if long-term data varyNot more than X; a shorter period can be called for2.4.2.2
RefrigeratorNo significant change at accelerated; little change and little variabilityNormally without statistical analysisUp to 1.5X, not more than X + 6 months2.5.1.1
RefrigeratorNo significant change at accelerated; change or variability; not analysed or not amenable, with supporting dataNot performedUp to X + 3 months2.5.1.1
RefrigeratorNo significant change at accelerated; change or variability; amenable and analysedPerformed, backing the proposal with supporting dataUp to 1.5X, not more than X + 6 months2.5.1.1
RefrigeratorSignificant change at accelerated between 3 and 6 monthsVerification by statistics can be appropriate if long-term data varyBased on long-term data; no extrapolation; shorter period can be called for2.5.1.2
RefrigeratorSignificant change at accelerated within the first 3 monthsVerification by statistics can be appropriate if long-term data varyBased on long-term data; no extrapolation; excursion discussion required2.5.1.2
FreezerIntended for freezer storageNot statedBased on long-term data; one batch tested at an elevated temperature for excursions2.5.2
Below -20 CIntended for storage below -20 CNot statedBased on long-term data; assessed case by case2.5.3

Sources

Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · full access

Ask about ICH Q1E: Evaluation of Stability Data and Shelf-Life Extrapolation

Full access: large monthly usage, up to 10 source passages behind each answer, and MCP and API access for Claude Code, Codex and other CLIs. Cancel any time.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →