Mean kinetic temperature (MKT): what it is and when to use it

Mean kinetic temperature answers one question: if this product had been held at a single constant temperature for the whole storage period, what temperature would have caused the same amount of chemical degradation as the profile it actually experienced? It comes from the Arrhenius relation, which says reaction rate rises exponentially with temperature, and it is defined in ICH Q1A for exactly this purpose.

That exponential is the whole point. An hour at 25 °C does not do twice the damage of an hour at 12.5 °C; it does far more. An arithmetic mean treats every reading as equally weighted, which is the wrong model for a process whose rate doubles every few degrees.

The difference, in numbers

Four 30-day profiles, each computed with ΔH/R = 10 000 K as ICH Q1A specifies.

Read the second row carefully. A single day above limit in a month moves the arithmetic mean by 0.02 °C, which is invisible in any report. It moves MKT by 0.83 °C. That is the entire argument for using MKT on stability and storage data: the mean is structurally incapable of showing you a short, hot event.

The fourth row is the other half of the lesson. Symmetrical variation around a setpoint, with no excursion at all, still produces an MKT above the mean. Warm periods contribute more than cold periods take away, always.

A room-temperature example

The effect scales with the absolute temperatures involved. A 25 °C controlled room that spends 51 days at 25 °C and 7 days at 40 °C has an arithmetic mean of 26.8 °C and an MKT of 28.5 °C. For a product labelled "store below 30 °C", the mean says comfortable and the MKT says close to the edge.

When MKT applies, and when it does not

• Use it for long-term storage and stability data, where cumulative chemical degradation is the concern and the product has stability data behind it.
• Use it for a transport lane assessed over many shipments, where the question is total thermal stress rather than any single leg.
• Do not use it to dismiss an excursion. A brief excursion above a critical threshold can cause damage that no cumulative average captures: freezing a vaccine, melting a suppository, denaturing a protein.
• Do not use it for products whose failure mode is not chemical degradation. MKT models an Arrhenius reaction rate, and nothing else.
• Do not use it to paper over a gap in the data. MKT computed over a dataset with a missing day is MKT of a different profile.

The common misuse is reporting an in-limit MKT as evidence that an out-of-limit period was acceptable. It is not. MKT and excursion handling answer different questions, and a regulator will expect both.

To compute one rather than read about it, the MKT and excursion assessor takes a pasted logger export, reports MKT alongside the arithmetic mean, and says which of the two the journey length makes the meaningful summary — because over a short record with one sharp excursion, MKT is the wrong answer.

Computing it

The definition weights each reading by exp(−ΔH/RT) with the temperatures in kelvin, averages those weights, then converts back. Two practical consequences: every reading must be in kelvin before the exponential is applied, and the intervals must be equal, or each reading needs weighting by the time it represents. Averaging a 1-minute logger against a 30-minute one without weighting gives a number that means nothing.

ICH Q1A fixes ΔH/R at 10 000 K, equivalent to an activation energy of about 83.1 kJ/mol. That is a convention chosen so results are comparable between organisations, not a measurement of your product. If you have real stability data giving a different activation energy, say which value you used and why.