Out-of-Specification Peptide Test Results: Investigation and Retesting Limits

An out-of-specification result is any analytical value that falls outside the acceptance criteria established for that material, and it cannot be erased by testing a sample repeatedly until a passing number appears. Out-of-specification peptide test results must trigger a documented investigation before any retest is run, and the permitted number of retests has to be fixed in advance rather than decided once the data are in. A certificate of analysis is only as credible as the rule the laboratory followed on the day a result failed.

This is the least visible part of purity documentation and the part most likely to separate a rigorous supplier from a careless one. Two vendors can publish certificates showing the same chromatographic purity figure while operating under completely different rules about what happens when the first chromatographic run comes back low. The published number looks identical. The evidentiary weight behind it is not.

What Counts as an Out-of-Specification Result

The regulatory definition is broad. An out-of-specification result is any test result falling outside the specifications or acceptance criteria established in drug applications, drug master files, official compendia, or by the manufacturer itself, and the definition extends to in-process laboratory tests as well as final release testing. For a synthetic research peptide the relevant specifications typically include chromatographic purity by reversed-phase HPLC, identity confirmation by mass spectrometry, water content by Karl Fischer coulometry, counterion content, net peptide content, and bacterial endotoxin limits.

An out-of-specification result is distinct from an out-of-trend result. A batch assaying at 98.4 percent against a 98.0 percent lower limit is within specification, but if the preceding twelve batches all assayed above 99.2 percent, that value is out of trend and signals process drift even though nothing failed. Out-of-trend results do not block release. They should still open an inquiry, because they are frequently the early warning that precedes a genuine failure.

The 1993 federal decision that still governs industry practice sorted out-of-specification results into three categories: laboratory error, non-process-related or operator error, and process-related or manufacturing error. The distinction matters because only the first category can legitimately be resolved by testing again. If the synthesis or the purification produced a genuinely impure material, no amount of additional analysis will change what is in the vial.

The Barr Decision and the Limits of Retesting

United States v. Barr Laboratories, Inc., 812 F. Supp. 458 (D.N.J. 1993), decided 5 February 1993 and amended 30 March 1993, remains the reference point for how a failing result may be handled. The court set out findings that are unusually specific for a judicial opinion, and they translate directly to peptide analytical work.

Retesting

A retest means additional testing on the same sample, using a second aliquot drawn either from the source of the first aliquot or from the larger sample already collected for laboratory purposes. The court found that retesting is proper only after a failure investigation is already underway, because the outcome of that investigation determines whether a retest is even appropriate. Where the investigation points to analyst error, retesting is warranted and limited in scope, since the retest simply replaces a first round of testing that has been rejected for cause. Where the investigation points to a process-related problem, the court held that retesting is suspect, on the reasoning that the initial results are genuine and additional analysis alone cannot infuse the product with quality.

Critically, the court found that retesting cannot continue indefinitely. A firm’s predetermined testing procedure must contain a defined point at which testing ends and the material is evaluated, and if the results at that point are unsatisfactory, the batch must be rejected. The number of retests is a matter of scientific judgment rather than a fixed figure, but the judgment has to be made and written down in advance. A laboratory that decides how many replicate determinations to run after seeing the first failing value has inverted the entire logic.

Resampling

Resampling means leaving the laboratory and drawing a fresh sample from the batch. The court treated this as substantially more controversial than retesting, since it typically occurs only after both the initial test and the retests have failed, which indicates a more serious problem rather than a less serious one. Resampling is acceptable where a compendial method provides for it, and where a failure investigation supplies actual evidence, not mere suspicion, that the original sample was unrepresentative. Outside those narrow circumstances the court held that firms cannot rely on resampling to release material that has already failed testing and retesting.

Averaging and the Outlier Test

Averaging is the subtlest of the three practices and the easiest to defend badly. The court accepted that averaging can be rational in principle while cautioning that averages hide the variability among individual results. The specific example recorded in the findings is worth stating exactly: three values of 89, 89, and 92 measured against an acceptance range of 90 to 110 produce an average of 90, which passes, while two of the three individual results fail. Expert testimony in that case held such a pattern should prompt further testing rather than release. Relying on an average without examining and explaining the individual failing results was described in the findings as highly misleading and unacceptable.

The outlier test fared no better. The United States Pharmacopeia expressly permits outlier tests for biological and antibiotic assays, where innate assay variability is high, but is silent on their use with chemical assays. The court construed that silence as prohibitory. Chromatographic purity determination is a chemical assay, considerably more precise than a microbiological one, which means a failing HPLC purity value generally cannot be statistically discarded as an outlier. That single finding forecloses the most common shortcut available to a laboratory under commercial pressure.

Why Two Competent Laboratories Report Different Numbers

None of this implies that a single number carries no uncertainty. It carries a great deal, which is precisely why the decision rule has to be set before the data arrive. The clearest available evidence comes from the international key comparison CCQM-K115.b, coordinated by the Bureau International des Poids et Mesures with the National Institute of Metrology of China, in which national metrology institutes were sent aliquots of the same homogeneous synthetic oxytocin material and asked to assign its mass fraction.

The reported values were as follows: BIPM 799.8 mg/g, NIM China 796.5 mg/g, NRC Canada 786.6 mg/g, INMETRO Brazil 781 mg/g, UME Turkey 773.09 mg/g, LGC United Kingdom 767 mg/g, and NMIJ Japan 766.3 mg/g, later revised to 773.2 mg/g after an identification error involving an oxytocin succinimide impurity at the asparagine residue was confirmed. Expanded uncertainties at a coverage factor of 2 ranged from 6.5 mg/g to 52.84 mg/g depending on the approach used, with mass balance methods generally yielding smaller uncertainties than amino acid analysis approaches.

Set aside the one participant whose value could not be used because it lacked correction for peptide-related impurities, and the remaining spread runs from 766.3 to 799.8 mg/g. That is a range of roughly 33 mg/g, about 4 percent in relative terms, generated by some of the most capable measurement laboratories in existence working on identical material. A commercial laboratory operating under time pressure will not do better. Any specification limit applied to a single reported value therefore has a genuine probability of misclassifying material in either direction, which is the reasoning behind the guard band approach set out in the Eurachem and CITAC guide on the use of uncertainty information in compliance assessment, second edition, 2021. A guard band shifts the practical decision threshold inward from the specification limit by an amount derived from the measurement uncertainty, so that a result declared compliant is compliant at a stated confidence level rather than merely on paper.

The same comparison produced a second figure with direct relevance to how purity is reported. Trifluoroacetate counterion in that oxytocin material was measured at 104.3 plus or minus 0.5 mg/g by fluorine-19 quantitative NMR and 103.5 plus or minus 1.6 mg/g by ion chromatography, while water content was 62.2 plus or minus 13.8 mg/g by Karl Fischer coulometry. Counterion and water together accounted for over 16 percent of the material by mass, in a sample whose chromatographic purity would have been reported as very high. This is the gap between chromatographic purity and net peptide content, and it is why a certificate reporting a single percentage without the accompanying content determination leaves the most important question unanswered.

Key Research Findings

• CCQM-K115.b reported synthetic oxytocin mass fractions spanning 766.3 to 799.8 mg/g across seven national metrology institutes and the BIPM analysing the same homogeneous material, a relative spread of approximately 4 percent.
• Expanded uncertainties in that comparison ranged from 6.5 mg/g to 52.84 mg/g at a coverage factor of 2, an eightfold difference driven by choice of analytical approach.
• Trifluoroacetate counterion measured 104.3 plus or minus 0.5 mg/g by fluorine-19 qNMR and water measured 62.2 plus or minus 13.8 mg/g by Karl Fischer coulometry in the same oxytocin material, together exceeding 16 percent of total mass.
• United States v. Barr Laboratories, 812 F. Supp. 458 (D.N.J. 1993) established that retesting is proper only after a failure investigation is underway, and that a predetermined stopping point must exist beyond which the batch is rejected.
• The Barr findings record that three values of 89, 89, and 92 against an acceptance range of 90 to 110 average to a passing 90 while two individual results fail, and that release on the average alone is unacceptable.
• The court construed United States Pharmacopeia silence on outlier tests for chemical assays as prohibitory, meaning a failing chromatographic purity value generally cannot be statistically excluded as an outlier.
• The FDA guidance on investigating out-of-specification test results was issued in October 2006 and updated by a Level 2 revision on 16 May 2022.

What the Current Guidance Requires

The FDA guidance for industry on investigating out-of-specification test results, originally issued in October 2006 and revised in May 2022, structures the response in phases. The first phase is a laboratory investigation conducted by the analyst and supervisor to determine whether an assignable analytical cause exists, carried out before any material is discarded and documented as it proceeds. If no laboratory cause is identified, the inquiry expands beyond the laboratory into the production record, where sampling representativeness and process performance come into scope. Only after that structure is in place does additional testing have any evidentiary meaning, and the original failing result is never deleted from the record. It is retained, explained, and reported alongside whatever follows.

For research materials that are not manufactured under a drug application, none of this is legally mandatory. That is exactly why it functions as a differentiator. A supplier operating outside a regulatory obligation who nevertheless adopts a predetermined retest limit, a documented investigation trail, and a policy of retaining failing data has chosen a standard that nothing external compels. A supplier who has not made that choice will produce certificates that look the same and mean less.

Reading a Certificate With This in Mind

Several practical checks follow. A certificate that reports a purity figure without stating the analytical method, column conditions, and detection wavelength cannot be assessed for whether the method was capable of resolving the impurities that matter. A certificate reporting only chromatographic purity, with no water determination, counterion content, or net peptide content, describes the ratio of peak areas rather than how much peptide is present. Reporting thresholds are equally load-bearing, since impurities below the reporting threshold do not appear at all and the choice of that threshold shapes the apparent profile, a topic covered in more detail in our discussion of peptide impurity thresholds and COA reporting.

Method validation status is the next question, since a purity figure produced by an unvalidated method has no established precision or specificity, as set out in our review of analytical method validation under ICH Q2. Where a result sits close to a specification limit, comparability between laboratories becomes the deciding factor, which is the subject of our analysis of peptide reference standards and interlaboratory comparability. Maple Research Labs publishes independent third-party laboratory reports for its catalogue on the certificates of analysis page, where the reported figures can be examined directly rather than taken on description.

Limitations of the Present Evidence

Two caveats apply. The Barr findings are a United States federal district court decision interpreting current good manufacturing practice regulations for finished pharmaceuticals, and they do not bind laboratories analysing research chemicals in Canada or anywhere else. Their value is as a detailed articulation of what constitutes scientifically defensible handling of a failing result, which is a question independent of jurisdiction. Second, the CCQM-K115.b comparison examined a single peptide of approximately 1 kDa with one disulfide cross-link, and interlaboratory spread for larger, more hydrophobic, or more heavily modified sequences may differ substantially in either direction. Extrapolating a 4 percent figure to every peptide would overstate what that study supports.

What survives both caveats is the structural point. A specification limit is a decision rule applied to an uncertain measurement, and the credibility of the resulting certificate depends entirely on whether that rule was fixed before the measurement or adjusted afterwards to accommodate it.

An out-of-specification investigation is only as trustworthy as the records supporting it, which is the subject of peptide COA data integrity and ALCOA+ audit trails. Readers assessing a supplier certificate directly will find the relevant fields covered in how to read a certificate of analysis, and the separate question of whether a batch result describes an individual vial is addressed in peptide vial content uniformity.

For research purposes only. Not for human consumption. Not for diagnostic or therapeutic use.