WorkflowQuality Control Laboratory

Laboratory Incident Investigation

Laboratory incident investigation for API QC labs: agents rebuild each analytical session from its records and test whether the claimed cause predates the result, and the lab head approves or rejects it.

See one vial leak tested from sampler log to signature ↓
Watch with transcript →
The problem

The cause goes on the form after the analyst has seen a low number

A leaking headspace vial, an expired standard, two swapped vials, a tripped pump: most are not out of specification, so the OOS procedure never sees them. The incident lives on a one-page form, and the cause is a sentence the analyst wrote at the bench. The printouts, vial log, sampler log, chromatograms and audit trail that would prove it sit in five systems, each read on its own.

First review asks whether the explanation is plausible. An inspector asks about order: was the cause on record before anyone saw the result, or written after a disappointing number? A result within limits can be set aside on lighter review than a failure gets, and each leaking vial from one hand crimper closes on its own form.

Built forQC laboratory headsQC supervisorsQA compliance specialistsQC analysts

The Problem Today

An invalidation can rest on one sentence, while FDA allows it only for a test event observed and documented.

164
FDA citations of the investigations rule, 211.192, in FY2025 published

Each system read separately

  • Balance
  • Sampler
  • Injections
  • Audit trail
  • Passing results set aside quietlyA duplicate that disagreed draws less review than a failure
  • Leaks read as bad daysEach crimp leak closes alone on its own form

The Solution

  1. The session rebuilt on one lineBalance, vials, sampler, injections and audit trail by the minute
  2. The cause tested three waysOn record before the result, sized to the effect, confined
  3. An inspector challenges the draftThe agents' own included, which cited a retest as proof
  4. The lab head signs the decisionApprove or reject with a meaning; QA reviews passing results

Built-in controls

  • Nobody approves an incident they raised
  • Retest plans are approved before any retest
  • A ruling applies only after the lab head approves

What You Get

Any client auditor can open a set-aside result and read its cause, its records and the order they were made in.

  • Supervisors investigate like the lab headPast rulings applied to each new case
  • A crimper pattern at the third leakInvalidations joined by crimper, instrument and method
  • Evidence a client audit can followCause, timing and signatures in one pack
Features

Laboratory incident investigation of one vial leak, minute by minute

A residual solvents run whose two preparations disagreed: the analyst blamed a leaking vial, the agents tested the claim and the lab head signed off. Then a passing chiral stability result, set aside for an air bubble, goes the other way.

demoSample data throughout: the lab, staff, batches, instruments and readings.

  1. 01Session line

    Analytical session reconstruction with first-view and cause-written markers

    • Five record lanes on one clock. Balance printouts, vial crimping, the sampler's event log, each injection and the audit trail line up by the minute, and the supervisor reads the session as it ran.
    • Two markers taken from the audit trail. The first view of the result and the moment the cause was typed come from system records, whatever the analyst recalls a day later.
  2. 02Instrument log

    Headspace vial leak check failure read from the sampler's own event log

    • Any mark opens its record. The red sampler mark opens the instrument-written log at the failing line, so the evidence is the instrument's own.
    • Opened records counted. Each record is marked read after it stays open a moment, and signing stays locked until the key records are.
  3. 03Three tests

    Assignable cause tests: recorded first, explains the effect, confined

    • On record before the result. The cause must rest on a record made before anyone viewed the number; a note written afterward cannot stand alone.
    • Direction and size explained. A leak can only read low, and the diluent peak falling in that vial points to lost headspace, not a weighing slip.
    • Confined to what it touched. Standards, brackets and the other batch's vials are checked too, so one vial's fault cannot excuse a whole run.
  4. 04Inspector's challenge

    An inspector-style challenge that strikes the agents' own draft

    • Questions an inspector asks first. Why trust the other preparation, and could a weighing error explain it? Each point closes on a cited record, or waits for a person.
    • A retest is not proof of a cause. The draft that leaned on the retest is rewritten on methanol and the diluent peak falling together in one vial.
  5. 05Extent

    Extent of a crimper fault across the vials it sealed since its last adjustment

    • Plain-word steering. The supervisor widens or narrows the search by typing, such as adding another GC or leaving out the standards.
    • Near-limit vials go to a person. A pass close to the limit is no proof of a good seal; whether kept samples are retested is the lab head's call.
  6. 06Retest plan

    Retest plan approved before any retest runs, against testing into compliance

    • Count, analyst, instrument and crimper set first. The supervisor approves how many preparations, by whom and on what before anyone retests.
    • Original data kept beside the retest. Original readings, chromatograms and sampler log stay on file; the retest mean becomes reportable once the lab head approves.
  7. 07Approval

    Invalidation approval with reason, optional note and signature meaning

    • Reason codes and a stated meaning. Name, time, reason and meaning are stored with the signature, controls designed for electronic records under EU GMP Annex 11 and 21 CFR Part 11.
    • The raiser cannot approve. An incident the lab head raised goes to the site QA head, so nobody signs off their own invalidation.
  8. 08Rejected claim

    Passing stability result set aside for an air bubble, tested and rejected

    • Same scrutiny for a passing result. A result inside its shelf-life limit but above the lab's trend alert gets the same three tests as a failure.
    • The original stands when the cause fails. With no bubble on record before the view and a steady pump, the lab head rejects it; both results are reported and an out-of-trend evaluation opens.
  9. 09Invalidation ledger

    Invalidation ledger by crimper, instrument, method and analyst

    • Patterns at the third leak. Three leaks on one crimper show as one pattern, with each of their incidents linked.
    • Out of use at once, CAPA on acceptance. Taking a crimper out of use takes effect immediately; the CAPA request waits until the site QA head accepts it.

Who raises an incident, who plans the retest and who signs the invalidation

QC lab heads buy it; these four jobs run it from bench to sign-off.

QC laboratory heads

Approve or reject each invalidation and own the laboratory's rulings

  • Each source of a session on one line
  • Signing locked until the key records are open
  • Rules proposed from their own past decisions
QC supervisors

First review of each incident within a working day, then the retest plan

  • The session rebuilt from records, minute by minute
  • Extent searches steered in plain words
  • Bulk first review with an audit line each
QA compliance specialists

Review each passing result set aside and a monthly sample of the rest

  • Passing-result invalidations routed to them after the lab head
  • Returning a decision clears the lab head's signature
  • An evidence pack for each incident
QC analysts

Raise incidents at the bench and run retests as the second analyst

  • The instrument's own record behind their claim
  • Retest results entered as numbers in fields
  • No approval of their own incidents

Five signatures stand between a leaking vial and the batch it touched

Nobody signs a record they have not opened first.

QC analyst
Sees
The sampler log and the duplicates that disagreed
Signs
The incident, as raised at the bench
QC supervisor
Sees
The rebuilt session and the kept vials
Signs
The first review and the retest plan, before any retest
QC lab head
Sees
The three tests, the challenge and their records
Signs
Approval or rejection with a reason and meaning; a crimper taken out of use
QA laboratory compliance specialist
Sees
Each passing result set aside, after the lab head
Signs
Agreement, or a return that reopens the decision
Site QA head
Sees
Batches waiting on an open incident or hold
Signs
The batch disposition, and acceptance of a CAPA request
More screens

A steady pump trace sank the bubble claim, and QA reviews that rejection too

Around the two incidents: the pump record, the queue, the rulings, the signatures, QA review and the batches waiting.

The pump trace at the injection in questionA bubble through the pump shows as a pressure dip and ripple; the trace at injection 4 shows neither, so the claim has nothing to rest on.
Open incidents with the clock each one runs onSeven open incidents timed against the first-review, lab head and QA clocks, each showing what it blocks.
Rulings the agents apply, and one waiting for approvalThree rulings in force, one being that a retest proves no cause; a fourth, raising an incident at the sampler itself, waits for the lab head.
Who signed what, why, and what the signature meantRaising, first review, retest plan, results, both decisions and the crimper withdrawal, each with signer and time; exported as a CSV or one incident's evidence pack.
QA laboratory compliance reviews each passing result set asideBoth decisions reach QA with their reason and the records read; the reviewer agrees, or returns one and the lab head decides again.
Batches whose disposition waits on the laboratoryThe site QA head sees which batch waits on which incident or hold, who owns it and what lifts the block.

demoSample data, from the working solution.

A vial leak logged before the number, and a bubble written after it

Two sessions, read minute by minute.

The leak was logged first

The sampler wrote the vial 9 failure at 13:02, 113 minutes before the result was first seen at 14:55; the cause note followed at 15:20.

The diluent fell with the solvent

Methanol read 42.0% lower in vial 9 and the DMSO diluent peak 40.6% lower in the same vial, which a weighing slip cannot do.

The first draft cited the retest

The agents' own draft used the 448 and 455 ppm retest as proof of the leak; the inspector's challenge got it corrected under the lab head's ruling.

No bubble anywhere in the record

The bubble was written at 11:48, after the result was viewed at 11:20, with the pump at 182 bar and no alarm; the 0.11% stands.

demoSample data, from the working solution.

One agent argues the inspector's side against the case the others built

Each agent hands on by name and stops where a person decides.

Session rebuilder

Puts each timed record on one line and marks the first view and the cause note.

  • Edits, reprocesses or reintegrates no data
  • Guesses no time a record does not carry
Cause tester

Tests the claimed cause three ways and writes supported or unsupported, with the record behind each test.

  • Counts no retest as evidence of a cause
  • Invalidates no result
Inspector

Challenges the incident, and the draft the agents wrote, as an investigator on site might.

  • Leaves its points for others to settle
  • Closes no point
Extent checker

Finds what else the cause could have touched since the last good check.

  • Clears no result by itself
  • Sends near-limit items to a person
Retest planner

Drafts the retest plan from the procedure before any retest runs.

  • Changes no retest count after results
  • Runs no retest
Ledger keeper

Joins closed incidents by instrument, method, consumable, crimper and analyst.

  • Proposes rules that wait for the lab head's approval
  • Keeps each analyst's view between that analyst and the lab head

Questions QC labs bring to laboratory incident investigation

What is laboratory incident investigation?

Laboratory incident investigation is how a QC laboratory examines a bench event, such as a leaking vial or a pump fault, before a result is set aside. Here agents rebuild the session from its records and test the claimed cause; the QC lab head approves or rejects the invalidation.

Is an assignable cause the same as a root cause?

An assignable cause is the test event that justifies setting one result aside: on record before the result was seen, sized to the effect and confined. A root cause asks why the event happened, such as a hand crimper that seals unevenly, and goes to a CAPA request.

How does it guard against testing into compliance?

The supervisor approves the retest plan before any retest, with the preparations, the analyst and the instrument fixed in advance, and the count holds once results arrive. A retest is never evidence of the cause, and all original data stays on file.

Do OOS invalidation and a passing result set aside get the same tests?

Yes. A result within limits that is set aside gets the three tests a failing one gets, and QA laboratory compliance reviews each such decision after the lab head signs. FDA's OOS guidance finds no scientific basis for invalidating an initial result in favor of a passing retest.

What goes into the laboratory investigation report?

The analyst's claim, the session line, the three tests with their records, the challenge, the retest plan and results, and the signatures, downloadable as one evidence pack.

How does the invalidated OOS rate fit in?

FDA's draft quality metrics guidance named the invalidated OOS rate as a site metric. The ledger tracks invalidations against reportable results each quarter and by crimper, instrument, method and analyst.

Who releases a batch that waits on a laboratory incident?

Disposition belongs to the site QA head, after the incident is decided and QA has reviewed it; release to market belongs to the client's quality unit. The agents make no release decision, and a question about a held batch is pointed to the person who decides.

Bring one set-aside result
and its instrument logs.

Send one incident with its printouts, instrument log, chromatograms and audit trail, and we rebuild the session and test the cause with your lab head.