Temperature excursion (revision 62)

A temperature excursion is an episode in which a product is held outside the storage conditions stated on its label. The term is used in pharmaceutical distribution and manufacturing for a deviation whose consequences must be assessed before the product is released for further use, and the assessment is a documented judgement about degradation rather than a rule about temperature.^[1]

An excursion is defined relative to a label, not to an absolute temperature. Because compendial definitions of controlled cold and controlled room temperature build permitted excursion bands into the definitions themselves, a departure from the nominal range is not automatically an excursion in the regulatory sense; a product labelled for storage at 2–8 °C that reaches 15 °C for an hour may remain within its labelled conditions, whereas the same profile applied to a product labelled for frozen storage would not.^[2]

Assessment proceeds from three quantities: how far outside the range the product went, for how long, and how much such exposure it has already accumulated. Duration and magnitude are combined through the mean kinetic temperature, a degradation-weighted average that gives disproportionate weight to warm intervals and can exceed the permitted range even when the ordinary time-weighted mean does not. Cumulative accounting matters because a product that has spent three separate periods outside its range has consumed three portions of whatever stability margin the manufacturer's programme established.^[3][1]

The evidence that makes a disposition possible is the product's own stability data, generated under the conditions specified in ICH Q1A(R2) and supplemented where necessary by deliberate stress studies. A product with no such programme — which includes essentially all peptides distributed for research use — cannot have an excursion assessed in this sense, because there is nothing against which to compare the exposure.^[4]

Distribution and compendial guidance treat an excursion as a deviation requiring investigation. The European good distribution practice guidelines require that deviations be recorded and their effect on quality evaluated, and that product not be released for further distribution until that evaluation is complete. WHO model guidance takes the same approach and adds the concept of a stability budget: the total permitted exposure outside labelled conditions, allocated across the product's life, against which each event is charged.^[5][6]

Three framings of the same event coexist and are worth distinguishing.

The compliance framing

Did the product leave its labelled conditions? This is answered by comparing a record against a label, and is a question about documents.

The kinetic framing

How much additional degradation is expected? This is answered by applying a rate model to the exposure, and is a question about chemistry.

The analytical framing

Can any change be measured? This is answered by testing, and is a question about method capability.

The three can disagree, and the disagreement is informative rather than paradoxical. A brief excursion may be a compliance failure while producing degradation far below the detection limit of any available assay; conversely a profile within the permitted band may still cause measurable change in a product whose stability is marginal. Assessment practice therefore uses all three: compliance to determine whether an investigation is required, kinetics to predict, and testing to confirm.^[1][7]

An excursion is also not a single class of event. Warming above the labelled range and cooling below it have different mechanisms and are not interchangeable in an assessment, because freezing damage is not governed by the Arrhenius relationship on which the warming arithmetic depends. Guidance for products labelled 2–8 °C accordingly treats a freezing event as a separate category, often with a stricter disposition, and freeze indicators exist precisely because the mean kinetic temperature calculation cannot detect the event.^[8][9]

Modern labels for injectable incretin therapies do not state a single storage condition; they state a cold-storage condition for the unopened product and a permitted period at higher temperature, either before first use or after it. These allowances are excursion tolerances that have been established by stability testing and written into the label, which converts what would otherwise be a deviation into ordinary use.

Allowances are as stated in the respective prescribing information and differ between jurisdictions; the figures above follow United States labelling and are cited to the labels rather than to secondary summaries.^{[10][11][12][13][14]}

Three features of this table matter for interpretation. The allowances are single, non-renewable periods, not a budget that resets: a pen that has spent 21 days at room temperature has consumed the allowance whether or not it is subsequently refrigerated. They differ substantially between products with similar chemistry, which reflects differences in formulation, container and the stability data generated rather than differences in the peptide. And they are stated as maxima at a ceiling temperature, so a period at 35 °C is outside the allowance entirely rather than being a shorter permitted period.^[4]

The corresponding position for lyophilised research peptides is that no such allowance exists in a documented form. Supplier storage statements are recommendations, and the periods sometimes quoted for ambient tolerance are not supported by an identifiable stability study. The physical argument for the robustness of a dry cake is set out at Lyophilisation and is a reason to expect tolerance, not a substitute for having measured it.^[15]

Where an excursion consists of warming, its cumulative effect is summarised by the mean kinetic temperature. For intervals of duration t_i at absolute temperatures T_i:

T_MKT = (ΔH/R) ÷ [ −ln( Σ t_i e^{−ΔH/(R T_i)} ÷ Σ t_i ) ]

with ΔH/R taken as 10,000 K by convention.^[3][1]

A realistic worked case demonstrates the property that makes the statistic useful. A consignment labelled for storage at 2–8 °C is held for 30 days, of which 96 hours are spent at 22 °C — a courier depot over a long weekend, repeated — and the remaining 624 hours at 5 °C.

The time-weighted arithmetic mean is

(96 × 22 + 624 × 5) ÷ 720 = (2112 + 3120) ÷ 720 = 7.27 °C

which lies inside the labelled range. The mean kinetic temperature, however, weights by degradation. With T = 295.15 K and 278.15 K:

e^{−10000/295.15} = e^−33.8811 and e^{−10000/278.15} = e^−35.9518

The colder term is smaller by a factor of e^−2.0707 = 0.1261. The weighted sum is therefore

e^−33.8811 × (96 + 624 × 0.1261) ÷ 720 = e^−33.8811 × 174.7 ÷ 720 = e^−33.8811 × 0.2426

and

T_MKT = 10000 ÷ (33.8811 + 1.4164) = 10000 ÷ 35.2975 = 283.31 K = 10.2 °C

The mean kinetic temperature is 10.2 °C, above the 8 °C upper limit, while the arithmetic mean is 7.3 °C and comfortably inside it. The two statistics reach opposite conclusions about the same record. This is not an artefact: the arithmetic mean treats an hour at 22 °C as equivalent to an hour at 5 °C when for degradation purposes it is roughly eight times as costly.^[3]

The calculation's limits are as important as its result. It presumes a single degradation pathway obeying the Arrhenius relationship with the assumed activation energy; a real product has several pathways with different activation energies, and where the dominant pathway has an activation energy far from 83 kJ·mol⁻¹ the statistic misrepresents the exposure in a direction that cannot be signed without knowing which pathway dominates. It also compresses the profile, so it cannot answer whether a threshold — a freezing point, a glass transition — was crossed.^[1][7]

The divergence is a direct consequence of convexity. Degradation rate is a convex function of temperature, so the average of the rates at two temperatures exceeds the rate at their average temperature. Mean kinetic temperature is defined as the temperature whose rate equals the average rate, and therefore always lies at or above the time-weighted arithmetic mean, with equality only for a constant profile.

The gap widens with the spread of the profile. For the 30-day case above, a spread of 17 °C produced a gap of 2.9 °C. For the extreme cycle worked at Cold chain, a spread of 28 °C produced a gap of 8.1 °C. A record with a narrow spread can be summarised by its arithmetic mean with little error; a record with a wide spread cannot, and it is precisely the wide-spread records that arise from excursions.^[3]

The rate constant for a chemical degradation pathway is conventionally described by the Arrhenius equation, in which rate rises exponentially with the reciprocal of absolute temperature. With the compendial activation energy, the multiplier for a 10 K rise near 25 °C is

exp[10000 × (1/298.15 − 1/308.15)] = exp(1.0884) = 2.97

so approximately a threefold acceleration per 10 K. This is the origin of the informal expectation that degradation roughly triples for each 10 °C, and it is the reason a short warm interval can outweigh a long cold one.^[7]

Real products depart from the single-pathway model in documented ways. Peptide degradation proceeds by several routes with different temperature dependences — deamidation of asparagine, oxidation of methionine, backbone hydrolysis, disulfide exchange, and physical aggregation — so the apparent activation energy of total degradation changes with temperature as the dominant route changes. Extrapolation from accelerated conditions to real storage is therefore reliable only over a limited range, which is why ICH stability programmes require long-term data at the labelled condition and treat accelerated data as supporting rather than substituting.^[4][16]

Aggregation is the clearest departure. It is frequently nucleation-dependent, so it exhibits a lag phase and then accelerates, and its rate may be non-monotonic in temperature because the conformational states that aggregate are populated over a limited range. A short excursion that nucleates aggregation can produce consequences that continue to develop after the temperature has been restored, which no rate constant evaluated at the excursion temperature will capture.^[16]

Cooling a liquid formulation below its freezing point subjects the solute to several distinct stresses, none of which is described by the Arrhenius relationship. Ice formation concentrates the remaining solution — freeze concentration — raising solute and salt concentrations by as much as an order of magnitude and shifting pH where buffer components crystallise selectively. It creates a large ice-water interface at which peptides and proteins adsorb and may unfold. And it may bring the system through a temperature range in which cold denaturation occurs.^[9]

The practical consequences are that a freezing excursion may cause more damage than a warming excursion of far greater apparent severity; that the damage is often physical rather than chemical, so a purity assay that resolves degradation products may not detect it while a size-exclusion or subvisible-particle method does; and that repeated freeze-thaw cycles are more damaging than a single freezing event of the same total duration, because each cycle regenerates the interface.^[9][16]

For lyophilised material the picture reverses. A dry cake has no ice to form and no interface to create, and freezing a sealed lyophilised vial is generally regarded as harmless — which is why frozen storage is the recommended long-term condition for dry peptide while being contraindicated for the same peptide in solution.^[4]

Disposition of a product after an excursion follows a defined sequence, and its output is a documented decision rather than a measurement.

1. The record is retrieved, its instrument's calibration status confirmed, and the profile reconstructed with custody points identified.
2. The exposure is characterised: maximum and minimum reached, duration outside the labelled range, mean kinetic temperature over the relevant period, and whether any threshold such as freezing was crossed.
3. The exposure is charged against the product's stability budget, cumulatively with any previous excursions recorded for that consignment.
4. The manufacturer's stability data are consulted, including accelerated and stress data, to establish whether the exposure lies within a range for which behaviour is known.
5. Where the exposure falls outside what the data support, confirmatory testing is performed on a sample, using methods capable of detecting the degradation expected.
6. A disposition is recorded: release unchanged, release with shortened dating, quarantine pending further data, or rejection.^[6][5]

Two structural features of this sequence are worth noting. It is asymmetric in favour of rejection, because the burden is to show that quality is unaffected rather than to show that it is affected. And it depends at step 4 on data that exist only for products with a stability programme, which is why the sequence cannot be applied outside regulated distribution however carefully the temperature was recorded.^[4]

Cumulative accounting is the step most often omitted. A consignment may pass three separate excursion assessments individually and fail the same criteria when the exposures are summed, and guidance is explicit that the accounting is cumulative across the product's life rather than per event.^[6]

A stability-indicating method is one shown to resolve the intact substance from its degradation products, so that a change in the assay result can be attributed to degradation rather than to variability. Demonstrating that property is a validation exercise, and the demonstration is conventionally made by forced degradation: the substance is deliberately stressed by heat, humidity, acid, base, oxidant and light, and the method is shown to separate and quantify what results, with mass balance accounted for.^[4][16]

After an excursion, such a method is asked to answer a narrow set of questions:

• has the assay value for the intact peptide fallen relative to the pre-excursion result for the same lot;
• have known degradation products increased, and specifically those expected from the stress applied — deamidation products after warm aqueous exposure, oxidation products after oxygen exposure, aggregate after freezing;
• has any new peak appeared that was not present before and is not attributable to the method;
• do orthogonal methods agree, since a single chromatographic method may be blind to a physical change that a size-exclusion or particulate method detects.^[16]

Two limitations are structural. Without a pre-excursion result for the same lot, a purity figure establishes only the current state, not a change; the retained sample is therefore the critical asset, and its absence is what most often makes a post-hoc assessment impossible. And a method's limit of detection sets a floor below which no statement can be made: an assay resolving 0.1% of a degradation product cannot demonstrate the absence of 0.05%, and a report of no detected change means no change above that floor rather than no change.^[4]

Where material is distributed without a stability programme, records or retained samples, the assessment machinery described above has no inputs. What remains is inference from general chemistry and observation by recipients, and the two are of very different value.

Inference from chemistry supports statements of the form that a dry lyophilised peptide sealed in a vial is expected to tolerate transient warming far better than the same peptide in aqueous solution. That is a well-supported general proposition. It does not support a statement about a particular parcel, because the parameters that determine the outcome — the glass transition temperature of that formulation, the moisture content, the integrity of the closure, the actual exposure — are all unmeasured.^[4][15]

Recipient observation is weaker than it appears. Community reports that material used after warm transit produced no noticeable difference in effect suffer from a specific set of defects: no pre-exposure analytical baseline, so any deficit cannot be attributed to transit; no measurement of the exposure, so the reported condition is a guess; no blinding, and a subjective endpoint; and survivorship, since material that was visibly ruined is discarded rather than reported. The absence of a noticed difference is also a low-sensitivity endpoint — a 15% loss of active peptide is unlikely to be perceptible and would be a serious analytical finding.^[15]

Analytical testing of a post-transit sample by a third-party laboratory is a genuine improvement, and gives a purity figure with a real detection limit. It nonetheless cannot attribute a deficit to transit without a matched pre-transit result on the same lot, because manufacturing variability and transit degradation are indistinguishable in a single measurement. Where a vendor's own certificate of analysis exists for the lot, the comparison is closer to informative, subject to the certificate being genuine and the sample being from the same lot.^[15]

• Cold chain
• Peptide stability table
• Beyond-use date
• Deamidation
• Methionine oxidation
• Peptide aggregation
• Lyophilisation

• Seevers RH, Hofer J, Harber P, Ulrich DA, Bishara R. "The use of mean kinetic temperature (MKT) in the handling, storage and distribution of temperature-sensitive pharmaceuticals." American Pharmaceutical Review (2009).
• Parenteral Drug Association. Technical Report No. 39: Guidance for Temperature-Controlled Medicinal Products, revised edition.

• ICH Q1A(R2) — Stability testing guideline; freely available and the basis of every labelled storage statement.