Out of Specification Results in Peptide QC: Investigate First
A receiving lab runs its incoming purity check on a new bulk lot, and the number comes back below the acceptance limit. The instinct is to re-inject, and if the second number clears the limit, to move on. Handling out of specification results that way is the habit quality auditors criticise most often, because it quietly turns a measurement into a selection process. This article walks through what a failing result actually obliges a lab to do, how an investigation is structured, and what a retest is legitimately able to prove, with an eye to labs that check lots across many vials and keep records for years.
Why testing until it passes proves nothing
Think about a lot whose true purity sits right at its limit. Every determination has random scatter, so some results land above the line and some below. Run the test enough times and a passing value will appear, not because the lot improved but because chance eventually supplies one. Reporting only that value says nothing about the material; it only shows how many attempts were made.
The principle that follows is simple: once an instrument has produced a valid result, that result is part of the record. A later, more convenient number does not erase it. It can only be set aside if a specific, documented error shows it was invalid.
Out of specification results versus other surprises
Not every odd result is a specification failure, and it helps to name the categories before deciding what to do.
| Category | What it means | Example |
|---|---|---|
| Out of specification (OOS) | The result breaches a written acceptance criterion | Purity below the stated minimum |
| Out of trend (OOT) | The result passes but departs from the history of previous lots | A lot that passes comfortably yet sits well below every earlier lot you have logged |
| Atypical or unexpected | Something appears that no limit covers | An extra peak, a retention time that has moved, or an observed mass nobody predicted |
Only the first triggers a formal failure, but the other two deserve a note in the lot record. A buyer who holds results for many lots is often the only party able to see a trend at all, a point developed in lot-to-lot variation and trending.
Phase one: check the laboratory before the lot
A structured investigation starts by asking whether the measurement went wrong. The analyst and a supervisor review the run together, ideally before any solutions are thrown away. Typical questions:
- Did the system suitability injections meet their criteria for that sequence? The usual checks are described in HPLC system suitability.
- Was the reference standard weighed and diluted correctly, and was it within its expiry?
- Was the mobile phase made up to the method, and was the right column installed?
- Were calculations, dilution factors and transcriptions correct?
- Is there any instrument log entry (a pressure fault, a lamp warning, a failed injection) at the relevant time?
The key discipline is that a cause must be demonstrated, not supposed. “Probably a bad injection” is a guess. An injection log showing a short draw is evidence. If a real error is found and documented, the original result is invalidated and the test is repeated correctly. That repeat is a replacement, not an appeal.
Phase two: when the result belongs to the material
If no laboratory error turns up, the failing number stands as a genuine property of the sample that was tested. The next question is whether that sample fairly represented the lot.
For a multi-vial lot, this is where sampling matters. A result from one vial speaks for that vial. Testing additional vials chosen according to a plan that was written beforehand, as outlined in sampling plans and batch representation, shows whether the problem is isolated or lot-wide. It does not cancel the first result; it adds to the evidence about the lot.
Retests, resamples and averages: set the rules in advance
Three activities are often lumped together under “retesting”, and they answer different questions:
- Re-injecting the same solution checks the injection and the instrument.
- Re-preparing from the same sample checks the preparation step.
- Testing a new sample from the lot is the only one of the three that says anything new about the lot itself.
How many repeats are allowed, which of these they involve, and who can authorise them should be fixed in a written procedure before any result exists. A limit decided after a failure is really a negotiation with the data.
Averaging needs the same care. If a method defines its reportable result as the mean of, say, three preparations, then the mean is the result. If the method calls for a single determination, averaging a failing value with a later passing one hides the very thing the lab needs to see. Two results that differ by well over a percentage point point to a variable method, sample or material, and that spread is itself the finding, usually far larger than the measurement uncertainty the method claims.
What a certificate shows, and what it cannot
A certificate of analysis states the reportable result for a lot against its limits. It does not say how many determinations were run, whether some were thrown out as invalid, or how any investigation ended. A clean certificate is consistent both with a clean first run and with a thorough, well-documented investigation.
So the confidence a certificate carries depends heavily on the issuing laboratory having a sound procedure for failures. Accreditation is the ordinary outside check on that, within the limits discussed in what ISO/IEC 17025 accreditation covers. When your own incoming test disagrees with a supplier’s certificate, treat it the same way: investigate your own run first, then compare methods, before concluding that either number is wrong.
Honest outcomes for a lot that genuinely fails
When the investigation confirms that a lot does not meet its specification, there are only a few defensible paths, and none of them involves finding a better number:
- Reject the lot. Record the result and the investigation, and quarantine or return the material.
- Reprocess and retest as a new lot. Material that is further purified becomes a different lot with its own identifier, as described in what batch and lot numbers mean.
- Release against a lower, openly stated specification. This is legitimate if the report says so plainly and the buyer’s use tolerates it.
The third option is underused. A lot honestly described at its measured purity against a matching limit is more useful to a research lab than the same lot described more generously because a later determination happened to land higher. For teams keeping long-term inventory records, that honesty is also what makes year-over-year comparisons meaningful.
Bulk Peptides supplies research compounds for in-vitro and analytical laboratory use only; nothing sold here is intended for human or veterinary use.
