On day 11 of a 28-day stability study at 40°C/75%RH, the chamber controller logged a high-temperature alarm at 03:47. The chamber had overshot to 44.2°C for 22 minutes before the controller corrected. The alarm was noted in the monitoring log. Testing continued. On day 28, the samples passed all acceptance criteria. The study appeared complete.
Then the regulatory submission reviewer asked for the chamber monitoring data. The 22-minute excursion was visible in the continuous log. The question — which the submission had not addressed — was whether the excursion affected the product's stability profile. The agency requested a scientific justification. It took six weeks to prepare a response that the agency accepted. The study was not repeated. But the submission was delayed by a quarter and the response cost more work than the deviation itself would have taken to handle correctly at the time it occurred.
Here is how to handle deviations and excursions correctly — before they appear in a submission or an audit.
The distinction that matters: planned deviation vs unplanned excursion
A planned deviation is a documented departure from the standard or test plan conditions that is identified and justified before testing begins. The chamber achieves 4.8°C/min loaded ramp rate against a required 5.0°C/min — this is a deviation. It goes in the test plan with a technical justification for why 4.8°C/min produces equivalent or more conservative stress than 5.0°C/min under the conditions of this test. An auditor reviewing the test record finds the deviation documented in advance, with a rationale, signed by the responsible engineer. This is a well-managed test programme.
An unplanned excursion is a departure from test conditions that was not anticipated in the test plan — a chamber overshoot, a controller fault, a power interruption, a sensor failure. The excursion may or may not affect the validity of the results. The critical question — which must be answered with scientific evidence, not opinion — is whether the excursion changed the stress experienced by the product in a way that either invalidates the test or makes it more or less severe than planned.
The four-step excursion response
Step 1: Document the excursion immediately. Before doing anything else, document what happened, when it happened, how long it lasted, what the chamber conditions were during the excursion, and what the DUT conditions were (if monitored). A continuous data log from the chamber controller is the primary evidence. If the chamber does not log data continuously, that is a programme gap — manual periodic logging creates gaps that an excursion can hide in. The chamber monitoring requirements are covered in Environmental Test Chamber Calibration.
Step 2: Assess the severity. Not all excursions are equal. A 2°C overshoot for 5 minutes in a 28-day stability study at 40°C is different from a 15°C overshoot for 4 hours in a 96-hour temperature cycling test. The assessment must consider: the magnitude of the excursion relative to the test conditions; the duration of the excursion relative to the total test duration; the failure mode that the test is designed to accelerate and whether the excursion increases or decreases the stress applied to that failure mode; and whether the DUT was monitored during the excursion for any functional or parametric response.
Step 3: Determine impact on results validity. There are four possible conclusions, and the one you reach determines the action you take. First, the excursion made the test more severe — the conditions were worse than planned, and the product passed anyway. This strengthens the result. Second, the excursion made the test less severe — the conditions were better than planned, but the product still passed and the severity difference was small enough that the result remains valid. This requires a quantified justification. Third, the excursion made the test less severe and the severity difference is large enough that the results cannot be relied upon — the test needs to be repeated. Fourth, the excursion's effect on test severity cannot be determined without additional data — additional analysis or repeat testing is required.
Step 4: Document the assessment and action taken. The excursion record must contain: a description of the excursion (conditions, duration, timing within the test); the assessment of impact on test severity; the conclusion on results validity; the action taken (no action, repeat test, additional analysis); and the name and signature of the engineer who made the assessment. In regulated industries — pharmaceutical, medical device, aerospace — this document becomes part of the test record and may be reviewed by a regulatory body. In non-regulated industries, it protects the programme if a customer questions the results years later.
The stability study case in detail
Pharmaceutical and cosmetic stability studies are the context where excursion management is most rigorously defined, because regulatory guidelines (ICH Q1A, FDA 21 CFR, EMA stability guidelines) explicitly address out-of-specification conditions. The general principle across all stability guidelines: short-term excursions within the Mean Kinetic Temperature (MKT) calculation may be acceptable; excursions that materially affect the MKT require scientific justification or repeat testing.
For a 28-day study at 40°C/75%RH, a 22-minute excursion to 44.2°C contributes approximately 0.03°C to the MKT — well within any reasonable acceptance criterion. The justification for accepting the study results is straightforward to construct. But it must be constructed — it cannot simply be asserted without the calculation. The calculation, with the excursion temperature and duration, belongs in the deviation record. The manufacturer profiles for chambers used in ICH-regulated stability studies — Binder GmbH and Memmert — both describe the data logging and alarm systems that make this documentation possible.
The worst excursion: the one nobody logged
Every experienced laboratory engineer has encountered this. The test ran over a weekend. The chamber alarmed at 02:00 Saturday. The alarm reset itself. The data logger was not configured for continuous logging — it sampled every four hours. The excursion is not in the record. On Monday, the samples looked fine. Testing continued.
This scenario is unresolvable if it comes to light during an audit. The correct answer is: you cannot demonstrate the test was run under the specified conditions during that interval. The conservative action is to restart the test. The data logging requirements that prevent this — minimum sampling interval, alarm configuration, alarm notification to a responsible person — should be defined in the test plan. If the chamber cannot provide continuous data logging, that is a capability gap that should be assessed before the test begins, not discovered when the audit happens. The documentation requirements that flow from this are covered in What Environmental Test Records Must Contain.
The MKT calculation, worked through once so you never have to guess at it again
Mean Kinetic Temperature converts a fluctuating temperature record into a single equivalent temperature that would produce the same overall degradation — it is the standard tool for arguing that a temperature excursion did not meaningfully change a stability study's severity. The calculation: MKT = (Ea/R) / ln[ (Σ e^(-Ea/RTi) ) / n ], where Ea is the activation energy (a default value of 83.144 kJ/mol is commonly used per ICH guidance when a product-specific value is not available), R is the gas constant, Ti is the absolute temperature at each recorded interval, and n is the number of intervals.
In practice, nobody does this by hand — the calculation is run across the full continuous temperature log for the entire study duration, with the excursion interval included at its actual recorded values. The output is one number: a temperature in Kelvin, convertible to Celsius, representing the effective constant temperature the product experienced. If the 28-day study's specified condition was 40°C and the calculated MKT comes out to 40.03°C, the excursion is defensible on its face. The justification document needs the inputs (the full temperature log, the Ea value used, and its source) alongside the output — not just the final number, because an auditor or reviewer needs to be able to reproduce the calculation, not just trust it.
The excursion that looks fine and isn't: gradual drift instead of a single spike
Everything above addresses the easy case — a single, bounded excursion with a clear start and end time. The harder case is a chamber that drifts slowly out of tolerance over days, with no single moment that looks alarming on its own. A controller with a failing sensor might read 0.3°C high on day 3, 0.6°C high on day 10, and 1.2°C high on day 20 — never triggering a high-temperature alarm threshold set for sudden excursions, because nothing about it is sudden. This pattern is invisible to monitoring systems built to catch spikes, and it is exactly the pattern continuous data logging with periodic engineering review — not just automated alarms — is designed to catch. A monthly look at the full temperature trend, not just a check for alarm events, is the practical control that finds drift before it accumulates into a result nobody can defend.
Writing the justification so the next reader does not have to trust you
A deviation or excursion justification that says "impact assessed as negligible, no action required" is an assertion, not a justification — and it reads as one to anyone reviewing it later. A justification that survives scrutiny shows the work: the magnitude and duration of the deviation, the specific calculation or reasoning connecting that magnitude to the test's purpose, and the conclusion that follows from the calculation rather than preceding it. The difference is whether a stranger reading the document six months later, with no access to the original engineer, can follow the reasoning and reach the same conclusion independently. If they cannot, the justification has not actually justified anything — it has just stated a comforting conclusion.