Stability chamber qualification

A stability study is only as good as the claim that the samples were held at the stated condition for the stated time. Qualification of the chamber is what supports that claim, and the thing people under-build is not the steady-state mapping — it is the evidence about what happens when the chamber is disturbed.

The conditions, and what the tolerance means

ICH Q1A(R2) defines the storage conditions for long-term, intermediate and accelerated testing. The commonly used ones:

Read the ± as a limit on the condition at the sample, not as the controller's deadband. A chamber whose control probe sits at 25.0 °C while the top shelf runs at 27.4 °C is outside the condition for the samples on that shelf, and the display will never say so. This is the whole reason mapping exists.

Mean kinetic temperature appears here too, and is often misapplied. MKT is the tool for summarising a varying temperature history against a labelled storage condition; it does not license a chamber that sits outside ± 2 °C to be called compliant because the kinetic average looks acceptable. Use it to assess an excursion, not to redefine the tolerance.

What the qualification has to cover

• Installation: utilities, calibration status of installed instrumentation, software and access control for the monitoring system, and the alarm path including who receives it out of hours.
• Operational: empty-chamber mapping at each qualified condition, control and alarm verification, door-opening recovery, and power-failure recovery.
• Performance: loaded mapping at the routine load pattern, run long enough to cover the chamber's full control cycle including any defrost or dehumidification cycle.

Sensor count follows the same volume logic as any mapping study, with the addition that shelf positions are the unit people care about: the report has to let a reviewer say whether a given shelf was inside the condition, so positions should map onto shelves rather than onto an abstract grid.

Humidity is where chambers fail

Temperature control in a modern stability chamber is usually excellent. Relative humidity is not, for three reasons. RH is a function of temperature, so a 1 °C gradient produces an RH gradient even with perfect vapour control. Humidity generation and dehumidification tend to cycle more coarsely than heating and cooling. And the measurement itself is worse: a good calibrated RH probe carries an expanded uncertainty of a couple of per cent RH against a ± 5 % RH tolerance, which leaves far less headroom than the temperature case.

The consequence for the protocol is that the RH acceptance criterion needs the measurement uncertainty applied to it explicitly. With U = 2 % RH and a limit of 60 ± 5 % RH, a reading of 64.5 % RH is not a demonstrated pass — a decision rule has to say what you do with it.

The two disturbance studies

A door-opening study records how far the condition departs when the chamber is opened for a routine sample pull, and how long it takes to come back. It produces a practical operating rule: how long the door may stay open, and how long after a pull the chamber should be left before the next one.

A power-failure study is the one that earns its keep. Run the chamber at condition, loaded, cut the power, and record the decay. The output is a number you can act on during a real outage: the time available before the samples leave the ICH condition. Without it, every power cut becomes a deviation investigation with no baseline to argue from.

Run the recovery studies loaded. An empty chamber decays far faster than a full one, so an empty-chamber power-failure study gives a pessimistic figure that nobody will believe, and a loaded study at a light load gives an optimistic one nobody should.

Requalification

Periodically, and on change. The periodic interval is a risk decision you document; the change triggers are not negotiable — repair affecting the air handling, refrigeration or humidity system, relocation, a change in load pattern, or a change to the monitoring system. A chamber that has had its humidity generator replaced has not been qualified in its current state, whatever the calendar says.