A quality-control analyst runs a purity assay on a peptide lot with an acceptance criterion of not less than 98.0%. The result comes back at 96.8%. The analyst is fairly sure the material is fine, the previous three lots all ran above 99%, and the instrument has been temperamental this week. The tempting response is to run it again and see. What happens next, and whether it follows a written rule or a hunch, determines whether the eventual certificate means anything.
What “out of specification” means
An out-of-specification (OOS) result is a test result that falls outside an established acceptance criterion: purity below its lower limit, an impurity above its upper limit, a mass outside its tolerance. The term describes the number, not a conclusion about the material. An OOS result could reflect a real problem with the lot, or it could reflect an error in how the measurement was made. Distinguishing between those two possibilities is the entire purpose of an investigation.
In regulated pharmaceutical production, FDA guidance sets out how such results should be investigated. Research-grade supply is not bound by that framework, but its logic is sound general laboratory practice and it is the standard against which any quality system can be judged.
The founding rule: a result cannot be unrecorded
Once an instrument has produced a valid-looking number, that number is data. It cannot be erased because a later number is more agreeable. A laboratory that repeats a test and discards failures until a pass appears is no longer measuring the material; it is selecting an answer.
The statistics make the danger concrete. For a lot that genuinely sits close to its limit, measurement variability alone means repeated testing will eventually produce a passing value. Reporting only that value turns random scatter into a claim of compliance. This practice is often called “testing into compliance,” and it is exactly what a written OOS procedure is meant to prevent.
Investigating in the right order
Stage one: the laboratory
The first question is whether the measurement itself went wrong. Typical checks include:
- Were standards and samples weighed and diluted correctly, with calculations verified?
- Was the mobile phase prepared as the method specifies?
- Did system suitability pass before and during the run?
- Were the correct column, wavelength and integration parameters used?
- Was the result transcribed accurately into the record?
Any cause identified must be demonstrated, not assumed. “The detector probably drifted” is speculation. A calibration or suitability record showing the drift is evidence. Only a documented, assignable error allows the original result to be invalidated.
Stage two: the material and the sample
If the laboratory review finds no error, the result stands as a property of the sample. The investigation then turns to whether the sample fairly represented the lot, how it was drawn, and whether the lot itself is the problem.
What a retest can and cannot do
| Legitimate use | Not legitimate |
|---|---|
| Replacing a result invalidated by a documented laboratory error | Overruling a valid result simply because a second run passed |
| Following a retest plan written into the procedure beforehand | Deciding how many retests to allow after seeing the failure |
| Testing fresh material from the lot when representativeness is in question | Repeating sample introduction from the same vial and calling it a lot retest |
The distinction in the last row is easy to miss. Re-running the same prepared solution tests the instrument and the sample introduction. Preparing a fresh solution from the same portion tests the preparation. Only newly sampled material tests the lot. When no laboratory error is found, the original result and any retests are all data, and the material is judged on the full set.
The averaging trap
Suppose a lot gives 96.9% on the first determination and 99.3% on a second. The mean is 98.1%, just above a 98.0% limit. Reporting the mean conceals a spread of more than two percentage points, which is itself the most important finding: something in the sample, the preparation or the method is behaving inconsistently.
Averaging is appropriate when the method was designed around it, for example a procedure that specifies the mean of three independent preparations as the reportable result. It is not appropriate when a method specifies a single determination and a mean is computed only after one of the results failed.
OOS, out of trend and atypical results
Not every worrying number breaches a limit. An out-of-trend result passes its specification but departs from the history of previous lots, such as a lot at 98.3% when earlier lots clustered near 99.5%. An atypical result is an anomaly without a limit attached, such as an unexpected peak or a shifted retention time. Both deserve follow-up, and neither appears on a certificate, because a certificate compares one lot with its own limits. They are visible only to whoever keeps the historical data.
The acceptable outcomes for a failed lot
- Reject it.
- Reprocess it, then test it as a new lot with its own identity.
- Release it against a different, lower specification that is stated openly on the report.
The third option is perfectly honest and underused. A result of 95% reported against a stated 95% limit gives a researcher accurate information. The same material reported at 98% because the third attempt happened to pass does not.
What this means when reading a report
A certificate shows the final reported value. It cannot show how many determinations were made, whether any were invalidated, or why. Confidence in the number therefore rests largely on whether the testing laboratory follows a defined procedure for failures. Questions to ask a supplier or laboratory are collected in HPLC method questions for a peptide supplier, and the structure of the document itself is covered in reading a peptide certificate of analysis. For why results differ legitimately between laboratories, see why suppliers report different peptide purity.
Questions
Is a single retest ever acceptable?
Yes, when the procedure defines it in advance or when a documented laboratory error invalidated the first result. It is not acceptable as an informal second chance.
Who decides whether a result is invalid?
The written procedure should name who authorizes invalidation, and that decision should rest on documented evidence of an assignable cause.
Why not just average all the results?
Averaging hides variability. Unless the method was designed to report a mean, averaging after a failure conceals the inconsistency the failure revealed.
Can a passing certificate follow an OOS result?
Yes, if the original result was properly invalidated for a documented laboratory error, or the lot was reprocessed and retested as a new lot.
Research use only. All products supplied by Battle Born Peptides are laboratory reference materials for in-vitro research and analytical use by qualified professionals. They are not drugs, foods, dietary supplements, cosmetics or medical devices; they are not approved by the FDA or any other regulator for use in humans or animals; and they are not intended to diagnose, treat, cure, mitigate or prevent any disease, or to affect the structure or any function of the body of humans or animals. Nothing in this article is preparation, handling or dosing guidance. See our full research-use terms.