How does ICH Q1E determine shelf life?

Shelf life is the time at which the 95% confidence bound of the stability regression first reaches the acceptance criterion, not where the average line crosses it.

What is poolability testing in ICH Q1E?

A covariance analysis at the 0.25 significance level that decides whether multiple batches behave similarly enough to combine; if slope and intercept terms are not significant the data may be pooled.

How does accelerated stability predict shelf-life?

A moisture-modified Arrhenius model (ln k = ln A - Ea/RT + B*RH) is fitted to assay-loss rates measured at several temperature and humidity conditions, then the rate is extrapolated to a storage condition and the time to reach the specification limit is the projected shelf-life.

When is a humidity term needed in accelerated stability modelling?

When degradation is moisture-sensitive, such as hydrolysis in solid dosage forms. If the stress conditions all share one humidity, only a temperature-dependent Arrhenius model is fitted; varying the relative humidity across conditions resolves the humidity coefficient B.

Shelf Life Calculator - ICH Q1E Stability Analysis

How ICH Q1E shelf-life estimation works

Shelf life is not simply the last time point that still passed. Under ICH Q1E it is the time at which the 95% confidence bound of the stability trend — not the average line — first reaches the acceptance criterion, so that batches at the edge of the distribution still comply.

The confidence-bound rule

An attribute (e.g. assay) is regressed against time. The shelf life is where the one-sided 95% confidence limit of the regression crosses the specification limit:

Shelf life = time where the 95% lower (or upper) confidence bound meets the acceptance criterion

For a decreasing attribute the lower bound versus the lower specification limit is used; for an increasing one (e.g. a degradant), the upper bound versus the upper limit.

Poolability — can batches be combined?

Combining batches gives a tighter estimate, but only if they behave similarly. Q1E tests this with covariance analysis at a 0.25 significance level: if the batch-by-time (slope) and intercept terms are not significant (p > 0.25), the data may be pooled; otherwise the worst-case individual batch governs.

Worked example

Assay declines roughly linearly with a lower specification limit of 95.0%:

The mean regression line reaches 95.0% at ~30 months.
But the 95% lower confidence bound reaches 95.0% earlier, at ~26 months.
Rounded to the tested interval, the supported shelf life is 24 months.

Reference: ICH Q1E, Evaluation of Stability Data.

Accelerated prediction (moisture-modified Arrhenius)

The tool's second mode predicts shelf-life early, from short accelerated stress studies, instead of waiting for full real-time data. Degradation speeds up with temperature (Arrhenius) and, for many solid-state reactions, with humidity — fitting that dependence lets you extrapolate the rate down to a storage condition.

ln k = ln A − Ea/(R·T) + B·RH

The assay-loss rate k is measured at several temperature / humidity conditions; a regression gives the activation energy (Ea) and the humidity coefficient (B). The rate is then projected to a storage condition (e.g. 25 °C / 60% RH), and the time to reach the specification limit is the predicted shelf-life. This is the basis of accelerated predictive stability (ASAP).

Worked example (accelerated)

Assay loss is measured at 40, 50 and 60 °C:

Each condition yields a degradation rate; a plot of ln k vs 1/T is linear, its slope giving Ea ≈ 80–100 kJ/mol for a typical small molecule.
Extrapolating the fit to 25 °C / 60% RH gives the storage rate.
Dividing the allowed loss (initial − spec limit) by that rate gives the projected shelf-life — available in weeks, not years.

Reference: ICH Q1A(R2); Waterman, moisture-modified Arrhenius / accelerated predictive stability. Supportive of, not a replacement for, real-time confirmation.

Frequently asked questions

Why use the confidence bound instead of the average trend line?

The mean line describes the typical batch; the confidence bound protects against batch-to-batch and measurement variability, so that even a batch at the edge of the expected distribution still meets specification through the labelled shelf life.

What is the 0.25 poolability level and why so high?

It is the significance level for testing whether batches share a common slope/intercept. A conservative (high) 0.25 threshold makes it harder to wrongly pool genuinely different batches, protecting the estimate.

Can the shelf life be extrapolated beyond the data?

Only cautiously. Q1E permits limited extrapolation when the change is small, well-characterised and statistically supported; large extrapolations require real-time confirmation.

Can accelerated data really predict real-time shelf-life?

It gives a scientifically grounded estimate for risk assessment and study design, and works well when a single Arrhenius-type mechanism dominates. It does not replace ICH Q1A real-time and long-term confirmation, which remain the basis of the registered shelf-life.

When is the humidity term needed?

When degradation is moisture-sensitive (common for hydrolysis in solid dosage forms). If your stress conditions share one humidity, the tool fits a temperature-only Arrhenius model; vary the RH across conditions to resolve the humidity coefficient B.