What is in Arc B
- Assignable cause: what counts, and what doesn't — the difference between a documented finding and a plausible-sounding guess
- The valid/invalid OOS distinction — what evidence actually invalidates a result, and what has to happen once a result stands as valid
- Retesting versus "testing into compliance" — the practice FDA calls unscientific and objectionable, and why accurate retests don't excuse it
- Outlier tests and averaging: the real rules — a narrow, specific tool, not a general-purpose way to make an inconvenient result disappear
This arc goes inside the Phase I laboratory investigation Arc A introduced. Arc C does the same for Phase II and the disposition decision that follows it. Arc D takes up conclusive root cause and CAPA impact assessment directly. Arc E closes with enforcement case studies, including one built around exactly the invalidation failure this arc describes.
Assignable cause: what counts, and what doesn't
Every distinction this arc draws — valid versus invalid, legitimate retest versus testing into compliance — comes down, in the end, to one question: is there an actual, documented assignable cause, or isn't there?
B1.1 Defining assignable cause
Requirement An assignable cause is a specific, documented, verifiable event or condition that can reasonably be determined to have caused an OOS result. It is not a category of explanation — it is a specific fact, established with evidence, tied to the specific result in question. FDA's guidance states this directly: in the case of a clearly identified laboratory error, the retest result substitutes for the original; where there is no such identified error, the original result stands.
B1.2 What qualifies, and what doesn't
- Qualifies: a maintenance log documenting an instrument fault during the specific run that produced the OOS result. Specific, contemporaneous, and independently verifiable.
- Qualifies: a documented transcription or calculation error, traced to the specific step where it occurred. Verifiable against the original raw data.
- Qualifies: confirmed reference standard or reagent failure, established against that material's own QC records. Tied to a specific, checkable fact, not an assumption.
- Does not qualify: "the analyst thinks the sample may have degraded," with no supporting data. A hypothesis is not a documented event.
- Does not qualify: a retest of the same sample that happened to pass. A different result is not an explanation for the first one.
- Does not qualify: the result's status as a statistical outlier relative to the rest of the data set. Outlier status describes the number; it says nothing about why it occurred. Module B4 returns to this specifically.
This is not an abstract standard. It is exactly what Arc A's Phase I checklist exists to produce: discussing the method with the analyst, examining raw chromatograms and spectra, verifying calculations, confirming instrument performance, and checking reference standards and reagents against their own specifications. An assignable cause is the documented output of that checklist — not a conclusion reached some other way and then supported by the checklist after the fact.
Module B1 — assignable cause
Five questions.
The valid/invalid OOS distinction
This module exists to answer, directly, the question this entire course was built around: what actually separates a result a firm may set aside from one it has to treat as real?
B2.1 Two outcomes, one investigation
Requirement Every OOS result starts in the same place: a test result outside specification, triggering the 211.192 investigation duty covered in Arc A. The Phase I investigation then resolves into exactly one of two outcomes.
| Invalid OOS result | Valid OOS result | |
|---|---|---|
| What the investigation found | A specific, documented, assignable test event that caused the result | No assignable cause, despite a genuine investigation |
| What it means | The result is an artifact of a testing error, not a true reflection of the batch | The result reflects an actual product or process condition |
| Evidence required | Observation and documentation of the specific event — not suspicion, not a passing retest alone | A documented, genuine attempt to find a cause that did not succeed |
| What happens next | The identified retest substitutes for the original result | Escalates to Phase II; scope extends to other batches/products per 211.192 |
B2.2 What evidence actually invalidates a result
Requirement FDA's OOS guidance states the standard plainly:
“Invalidation of a discrete test result may be done only upon the observation and documentation of a test event that can reasonably be determined to have caused the OOS result.”FDA, Guidance for Industry: Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production (Level 2 revision, May 2022), Section IV.B.
Requirement Notice what this standard does not accept: it does not accept suspicion, it does not accept a statistically improbable result, and — as Module B3 covers in detail — it does not accept a subsequent passing retest on its own. It requires the same thing Module B1 described: a specific test event, observed and documented.
B2.3 What a valid OOS result requires next
Requirement A valid OOS result — one where the Phase I investigation could not identify an assignable cause — is not the end of the process. It requires, in sequence: escalation to a formal Phase II investigation (Arc C); extending the investigation's scope to other batches of the same product and other products that may share the same cause, per 211.192; a documented disposition decision for the batch in question; and a CAPA that includes a genuine root cause determination and a properly scoped impact assessment — the subject Arc D takes up directly, including whether every CAPA needs a monitoring and confirmation component.
Arc E examines a 2026 FDA warning letter to Medical Products Laboratories, Inc., built around exactly the failure this module is warning against: a firm invalidating OOS results — including stability failures — on the basis of resampling data alone, with no identified laboratory error and no documented assignable cause. The result was a required retrospective review of every invalidated result the firm had on file. The standard in this module is not academic; it is the standard regulators actually apply after the fact.
Module B2 — valid vs. invalid OOS
Six questions.
Retesting versus "testing into compliance"
Retesting is a normal, proper part of an OOS investigation. It is also the part of the process most often abused — which is why FDA singles it out by name.
B3.1 "Testing into compliance"
Requirement FDA's guidance describes, and names, a specific abuse of the retesting process:
“FDA inspections have revealed that some firms use a strategy of repeated testing until a passing result is obtained, then disregarding the OOS results without scientific justification. This practice of ‘testing into compliance’ is unscientific and objectionable under CGMP.”FDA, Guidance for Industry: Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production (Level 2 revision, May 2022), Section IV.B.1.
Practice Notice precisely what makes this objectionable: not that a retest was performed, and not that the retest happened to pass. What makes it "testing into compliance" is disregarding the earlier failing results without scientific justification — treating the number of attempts as a mechanism for producing the answer you wanted, rather than as part of a genuine investigation.
B3.2 What legitimate retesting requires
Requirement The maximum number of retests to be performed on a sample must be specified in advance, in a written SOP — decided before any particular result is known, not chosen after the fact based on which attempt happens to pass.
Requirement And critically:
“If no laboratory or calculation errors are identified in the first test, there is no scientific basis for invalidating initial OOS results in favor of passing retest results. All test results, both passing and suspect, should be reported and considered in batch release decisions.”FDA, Guidance for Industry: Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production (Level 2 revision, May 2022), Section IV.B.1.
B3.3 Retesting versus resampling, revisited
Requirement Arc A introduced the distinction Barr Laboratories drew between retesting (working the original, homogeneous sample again) and resampling (drawing an entirely new specimen). FDA's guidance adopts the same line: resampling is appropriate only where there is evidence — not suspicion — that the original sample was prepared improperly and was therefore not representative of the batch. A firm that resamples because it hopes a new specimen will behave differently, without that evidence, has not resampled for a legitimate reason; it has simply found a new way to test into compliance.
Module B3 — retesting vs. testing into compliance
Six questions.
Outlier tests and averaging: the real rules
Two more tools appear constantly in OOS investigations, and both are narrower than they are often treated as being: the outlier test, and the average.
B4.1 What an outlier test actually does
Requirement FDA's guidance is direct about the limits of an outlier test:
“An outlier test is only a statistical analysis of the data obtained from testing and retesting. It will not identify the cause of an extreme observation and, therefore, should not be used to invalidate the suspect result.”FDA, Guidance for Industry: Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production (Level 2 revision, May 2022).
Requirement An outlier test is also not applicable where variability itself is the thing being assessed — content uniformity, dissolution, and release-rate testing, for example, are built specifically to characterize how much individual units vary, so excluding a result as a statistical outlier would remove the exact signal those tests exist to capture.
Requirement Consistent with Barr Laboratories' holding in Arc A, outlier testing may be appropriate for biological or antibiotic assays, given their substantial innate variability, per USP <111> (Design and Analysis of Biological Assays) — not for chemical assays generally.
B4.2 Averaging: appropriate use and the line it cannot cross
Requirement Averaging is appropriate for a genuinely homogeneous sample or a properly defined-replicate test. It is not appropriate for content-uniformity or blend-uniformity type determinations, for the same reason outlier testing is not: those tests exist to measure variability, and an average erases exactly what they are measuring.
Requirement The more consequential rule concerns the OOS investigation itself: averaging the original result that prompted the investigation together with additional retest or resample results obtained during that investigation is not appropriate, because it hides the variability among the individual results — the same concealment Barr Laboratories condemned when it called this kind of averaging highly misleading and unacceptable.
Requirement Where some individual results are OOS and others pass, but the average of all of them falls within specification, the guidance's rule is deliberately conservative: a firm should err on the side of caution and treat the average itself as an OOS result, even though the average, taken alone, is within specification.
Arc C goes inside the Phase II full-scale investigation and the disposition decision it produces — including both versions of "Phase III" flagged in Arc A. Arc D is where this course directly takes up conclusive root cause, CAPA impact assessment, and whether every CAPA needs a monitoring and confirmation component. Arc E closes with enforcement case studies, including the Medical Products Laboratories, Inc. letter referenced in Module B2.
Module B4 — outlier tests and averaging
Five questions.