Stability

Run this study#

Your question: how long do results stay within your allowable change under each storage condition?

Bring: materials representative of the intended use, with enough volume for every planned measurement.

Suggested starting plan for paired specimen stability: 10 independent specimens, tested in duplicate at baseline and at 3 later times under each of 2 conditions. With one shared baseline, that is 20 baseline results + 60 per condition = 140 results. For an independent design, use 10 different specimens at baseline and 10 new specimens at each later condition and time, all in duplicate: 70 specimens and 140 results. Choose times before, at and beyond your intended duration. The final count depends on variability and the allowable change.

1. Define the claim first. Choose Create study. On Set up, record what is stored, each condition, the intended duration, and the allowable change in the study unit, as a percentage, or both. To require a minimum duration, choose Enter your own limit under Acceptance limits. For reagent or control stability, decide what counts as an independent material or preparation.
2. Choose the design that matches your collection. For Paired, prepare all aliquots at baseline and keep one specimen ID across times. For Independent, use different specimens at baseline and each later condition and time. Give each specimen a new ID, keeping that ID only for its replicates.
3. Measure at the planned times. Record every repeat: specimen ID, condition, elapsed time, result and replicate. Leave the condition blank for a shared baseline.
4. Import and calculate. On Data, choose Import a file and choose your file (start from Blank template (CSV) or stability.csv). Answer any question the dialog asks, then choose Import into this study. Choose Calculate results. Compare each condition's changes and intervals with your limit. On Report, choose Download PDF.

You get: for each condition, the last tested time in an unbroken run of within-limit decisions. An inconclusive time ends support, even if a later time passes. An interval needs two paired materials, or two different specimens at baseline and at each later time. Such small groups often leave times inconclusive. Only tested times can be supported, so test the conditions and durations you intend to claim, as in the FDA specimen-handling principle.

Purpose#

Compare results after storage with baseline, usually time 0, condition by condition. A paired design remeasures the same materials; an independent design uses different specimens from the same population at each time.

When to use it#

Use it for specimens waiting before testing (room temperature, refrigerated, frozen, freeze-thaw), for reagent, calibrator or control material over on-board, open-vial or stored time, or to compare several storage conditions in one study. Keep specimen stability and reagent, calibrator or control stability in separate studies.

For an added substance, use Interference. For a high sample affecting the next, use Carryover. For two reagent lots, use Reagent lot comparison.

Study setup#

1. In your project choose New study. Enter Analyte and Unit, and select Stability under Sample and material effects.
2. Enter Time unit. Choose What is stored, and answer Same material measured at every time? with Yes for the paired design or No for the independent design. Largest acceptable change from baseline (%) is optional. Choose Create study, or Create and import data if you already have results.
3. On Set up, under Stability settings, check the Stability type, Design, Time unit and Baseline time. Choose Allowable change set as: In the study unit, Percentage of the baseline mean, Either (the larger allowance) or Both must be met, then enter the Allowable change. For Either or Both, enter the amount in the study unit and the percentage. The change is compared with the amount and the relative change with the percentage. Either passes a time when one of them is met; Both only when both are. Without an allowable change, the changes are calculated but get no decision.
4. Under Storage and handling conditions, choose Add all conditions found in the data (N) or Add condition. Complete Condition name and optionally Claimed duration. Names must be unique and match the data. Case doesn't matter.
5. Percent limits need the Measurement scale under Calculation settings to be Ratio scale · CV usable, which is the default.
6. Optionally, under Acceptance limits, choose Enter your own limit, then Tested duration within the allowable change, and enter the shortest acceptable duration in the time unit, such as 72. Keep All conditions under Applies to, or choose one condition and add a separate limit for each condition that needs a different duration. A limit whose condition matches no condition in the data is not evaluated, and the study can't pass. See Acceptance limits.

Paired means each material's change from its own baseline, then the mean across materials. Independent means the mean at each time minus the mean of the baseline specimens.

Entering data#

One row per result:

Choose Import a file or type into the grid. To add later times, import a file with only the new readings and leave Replace the current rows unticked. For paired data, keep the same material IDs; for independent data, use new IDs for each later condition and time. If the dialog says rows are already in this study, choose Import an intentional repeat only if they are new measurements. See Importing and mapping.

Fix a row with no result or a qualified result such as <0.01 or >1000: correct it, or exclude it with a reason. See Missing and censored values.

The calculation stops when:

• a row is marked, for example for a condition not set on Set up, a time before baseline, or a replicate label used twice
• a specimen has results at more than one time in the independent design (use the paired design instead)
• every included row is a baseline, with no storage condition set or found in the data

In a paired design, a material without a baseline is left out of the change. A condition without a baseline gets no change or duration.

Statistics#

Paired design (m materials with baseline and time results):
  d_i        = mean(material i at time) − mean(material i at baseline)
  change     = mean(d_i)
  SD         = sample SD of d_i (n − 1)
  SE         = SD / sqrt(m)
  interval   = change ± t(1 − α/2, m − 1) × SE
  relative % = mean and Student t interval of 100 × d_i / baseline mean of material i

Independent design:
  change     = mean(specimen means at time) − mean(specimen means at baseline)
  interval   = Welch interval, SE = sqrt(s_t²/n_t + s_0²/n_0), Welch–Satterthwaite df
  relative % = 100 × change / baseline mean (interval scaled the same way)

Percent limit decisions use the relative change. Study-unit limit decisions use the change.
Either / Both: the change is judged against A and the relative change against P%.
  Either: within if one part is within; exceeds if both exceed.
  Both:   within if both parts are within; exceeds if either exceeds.
  Without a relative change, Either uses A alone and Both is inconclusive.
Supported duration = last time of the unbroken run of "within" decisions from the first tested time.

Relative intervals treat the baseline mean as fixed.

The limit Tested duration within the allowable change compares each condition's supported duration with your required duration. A condition with fewer than 2 materials or specimens is Undecided. A required duration longer than the longest tested time can't be met.

Sources: NIST/SEMATECH e-Handbook §1.3.5.2, §1.3.5.3 (Welch) and §1.3.5.6. See Methods and sources.

Worked example#

Download stability.csv: synthetic glucose (mg/dL), five specimens S01 to S05 in duplicate. Baseline at 0 h with a blank condition (shared). Room temperature, unspun at 2, 4, 8 and 24 h. Refrigerated, separated at 4, 24 and 72 h.

Setup: Specimen (sample) stability, paired, time unit h, baseline 0, Allowable change set as Percentage of the baseline mean, Allowable change (%) 5, 95 % confidence. Conditions: "Room temperature, unspun" (claimed 8 h) and "Refrigerated, separated" (claimed 72 h). Acceptance limit: Tested duration within the allowable change at least 72 h, All conditions.

Shared baseline: 5 materials, 10 results, mean 165.4 mg/dL.

Room temperature, unspun

Supported: 4 h, because 8 h is inconclusive. Claimed 8 h: not supported.

Refrigerated, separated

Supported: 72 h. Claimed 72 h: supported.

On the required duration of at least 72 h, room temperature is Not met (4 h supported, tested only to 24 h) and refrigerated is Met (72 h). Study status: Criteria not met.

Two variations:

• Absolute limit of 6 mg/dL: every room-temperature time became inconclusive (at 2 h, −6.9 to 1.1 crosses −6), so nothing was supported. Refrigerated stayed at 72 h.
• S05 baseline excluded: S05 was left out of every change, for example −1.4% at 2 h at room temperature. Supported durations stayed at 4 h and 72 h.

Reading the results#

Pick a condition with the Condition selector.

The Change from baseline plot draws time to scale, at the times you entered. With a baseline time of 0 the axis reads time since baseline; otherwise it names the baseline time. Diamonds are the mean change with its interval. Dots are each material's change. Select one to find its row. Lines at ±limit show the allowable change, and a vertical line marks the supported duration. With a percent limit, the plot is in % of baseline.

Above the results, each condition has a line with its supported duration and claim. The condition panel shows which baseline was used: Condition-specific baseline, Shared baseline (blank condition) or Missing. Timepoints · condition gives each time's counts, changes, intervals and decision.

When it fails:

• Early exceeds: check the time, result and condition name. If they are right, the material changed. Shorten the claim or retest under tighter conditions.
• Inconclusive everywhere: the intervals cross the limit. Use the paired design for materials measured repeatedly. More materials narrow the intervals.
• Point estimate only everywhere: a time has fewer than 2 materials, or every material changed by the same amount.
• Not met: the supported duration is shorter than required. Extend testing or accept the shorter duration.