Written and maintained by CASRAI Editorial Board
Last updated
An out-of-specification (OOS) result is a test result that falls outside the specifications or acceptance criteria established in a drug application, compendial monograph, or internal specification. It is a narrower category than a deviation (any departure from an approved procedure) or an out-of-trend (OOT) result (a result that is within specification but statistically inconsistent with a product’s historical trend, most often flagged during stability testing). This guide covers the process a laboratory must follow once an OOS result actually occurs: FDA’s two-phase investigation framework, the specific points at which a result can be invalidated versus must be accepted as valid, and the 1993 federal court decision that shaped how the industry handles retesting and resampling before FDA ever published formal guidance on the subject.
The controlling FDA guidance is Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production (Center for Drug Evaluation and Research, finalized October 2006, after a 1998 draft). It applies to chemistry, manufacturing, and controls (CMC) testing under current Good Manufacturing Practice (CGMP), and its logic is widely applied by analogy in GLP and other regulated-testing settings even though it was written for drug production.
Phase I: The laboratory investigation
Every OOS result triggers a Phase I laboratory investigation before anything else happens. The purpose is narrow: determine whether the OOS result was caused by a documented, assignable laboratory error, as opposed to reflecting something real about the sample or the batch.
Initial assessment
The analyst and a supervisor first review, before any retesting, whether an obvious error occurred: a calculation mistake, a documented instrument malfunction, a sample-preparation error, an expired or out-of-calibration standard or reagent, or a deviation from the written test method. This initial assessment is meant to be fast — if a clear, documented cause is visible in the raw data, it can resolve the investigation without a formal escalation. If nothing obvious turns up, the investigation becomes formal.
Hypothesis testing and the retesting rule
Where the initial assessment does not identify an assignable cause, the guidance permits hypothesis testing: a second analyst, typically more experienced, retests using the same sample preparation, to help distinguish an analyst-specific error from a method or sample issue. This step must be governed by a written, pre-approved retesting SOP that fixes the number of retests and the decision rule in advance — the SOP has to exist before the OOS occurs, not be written or adjusted afterward to fit the result in hand. FDA’s guidance is explicit that retesting is not a mechanism for “testing into compliance”: a firm cannot keep retesting an unexplained result until a passing value appears and then discard the rest. Every retest performed, passing or failing, becomes part of the investigation record.
Outlier and statistical tests
A statistical outlier test can only be used to exclude an OOS chemistry result if that specific test was specified in the approved application or compendial method before testing began. It cannot be selected after the fact because it happens to explain away an inconvenient number. FDA’s guidance treats outlier tests as having a legitimate but narrow role in certain biological assays where inherent variability is well characterized; for standard chemical testing, an unexplained result cannot be discarded on statistical grounds alone.
The decision point: invalidate or accept
Phase I ends at one of two outcomes, and the dividing line is documentation, not preference:
- Invalidate the result only when the investigation has identified and documented a specific, assignable cause directly tied to a laboratory error — a named equipment fault, a confirmed calculation error, a documented deviation from the test method. The retest result (obtained under the pre-approved SOP) then becomes the reportable value, and the original result is retained in the record as invalid, with the rationale.
- Accept the result as valid whenever Phase I does not turn up a documented assignable cause — this is the default, not an exception. An unexplained OOS result is not something a lab is permitted to set aside; the absence of an explanation is itself the trigger that moves the investigation into Phase II.
Phase II: The full-scale (production) investigation
Phase II opens when Phase I fails to identify a laboratory cause. It shifts the inquiry from the analyst’s bench to the manufacturing process, and it is owned by the quality unit (QA), not by the laboratory alone.
Scope
A full-scale investigation typically reviews the complete batch production record, in-process control data, equipment qualification and maintenance history, environmental monitoring data, raw-material and component lot history, and whether other batches manufactured on the same line, with the same materials, or in the same timeframe show a related pattern. The question is no longer “did the lab make a mistake” but “does this batch, or this process, actually have a problem.”
Resampling
Resampling — pulling additional samples from the same batch, as distinct from retesting the same sample — is permitted only where there is a documented, scientifically sound rationale specific to that batch, established under a predetermined procedure. It is not a routine second chance, and additional samples must come from the same batch or lot already implicated by the OOS result, never a different batch substituted in its place.
Averaging
Averaging a failing result together with passing results to arrive at a compliant mean is generally not an acceptable substitute for a documented investigation — it can mask a real variability problem rather than explain it, and FDA’s guidance treats reliance on averaging to dispose of an OOS with skepticism outside of test methods where averaging is a validated, specified part of the method itself (such as certain content-uniformity or dissolution testing schemes).
Disposition
The quality unit makes the final batch-disposition decision — release, reject, or rework — based on the complete Phase I and Phase II record, not on a single retest passing. Where a true OOS is confirmed, the investigation feeds directly into a formal CAPA and, where already-distributed lots are implicated, into a field-alert or recall assessment.
The Barr Laboratories precedent
Much of the structure above predates the 2006 guidance by more than a decade. United States v. Barr Laboratories, Inc., 812 F. Supp. 458 (D.N.J. 1993), arose from an FDA seizure action against a Barr manufacturing facility and is the case most credited with establishing the analytical framework FDA later formalized. The court, reviewing expert testimony from both sides, distinguished retesting from resampling, rejected the practice of testing repeatedly until a passing result appeared without a documented scientific basis, and laid out the same two-tier logic used today: first a laboratory-level inquiry into whether an assignable, documented error explains the result, and only then, absent one, a broader inquiry into the manufacturing process itself. The decision is why OOS investigation practice in the U.S. is often described as having been shaped by a federal court before it was ever codified in agency guidance — FDA’s 2006 document formalizes and extends the Barr framework rather than inventing it from scratch.
Timing and documentation expectations
FDA’s guidance does not set a fixed universal deadline for completing an OOS investigation, but it does expect a firm’s own SOP to specify target timeframes for each phase, and it expects investigations to be completed, documented, and closed before batch disposition — not after a batch has already shipped. In practice, most quality systems set an internal target (commonly in the range of 20–30 business days, set by the firm’s own SOP rather than by regulation) for closing the full record, with a documented rationale required for anything that runs longer. The full record — initial assessment, hypothesis testing and retest data, any resampling and its rationale, the Phase II review, and the final disposition decision — is what an FDA investigator or auditor will ask to see, not just the final conclusion.
Frequently asked questions
What is the difference between an OOS result, a deviation, and an out-of-trend (OOT) result?
An OOS result fails a specification or acceptance criterion outright. A deviation is any departure from an approved procedure, and may or may not produce an OOS result. An OOT result is within specification but inconsistent with the product’s established historical trend — most commonly flagged during stability testing — and triggers its own, generally lighter-weight investigation focused on trend and shelf-life impact rather than the full OOS framework.
Can a laboratory just retest a failing sample until it passes?
No. Retesting is only valid when performed under a written, pre-approved SOP that fixes the retest strategy in advance, and every result obtained — passing or failing — is part of the record. Retesting without a documented rationale, or discarding failing retests, is the exact practice both the Barr Laboratories decision and the FDA guidance were written to stop.
Who makes the final call on whether to invalidate an OOS result?
Invalidation requires a documented, assignable laboratory cause, and the decision to invalidate — as well as the final batch disposition if the result is not invalidated — sits with the quality unit (QA), not with the analyst or the lab supervisor acting alone.
What actually triggers a Phase II full-scale investigation?
The absence of a documented assignable cause at the end of Phase I. If the laboratory investigation cannot point to a specific, documented error, the result must be accepted as potentially reflecting a real process or product problem, which is what moves the investigation into the broader Phase II review of the manufacturing process itself.
