In Spec / Out of Spec

In spec/out of spec (OOS) is the MES-controlled decision that classifies a measured value against an approved specification limit, immediately determining whether a process step can proceed, a lot must be held, or an investigation must start. It applies to raw materials, in-process controls (IPCs), environmental conditions, and finished-product tests. In regulated contexts, the decision and its basis (limits, method, significant figures) must be preapproved, version-controlled, and applied consistently at data capture, not retrofitted post hoc.

ICH Q6A defines specifications as a list of tests, references to analytical procedures, and appropriate acceptance criteria for new drug substances and products. FDA’s OOS guidance clarifies that any test result outside specifications is OOS and demands a structured investigation; it also warns against practices like averaging to mask failure. In device and food/dietary supplement operations, analogous regimes require prompt segregation and nonconformance handling when results breach limits.

Specifications (acceptance criteria) are clinical/performance-justified, preapproved limits used for product disposition per ICH Q6A and 21 CFR requirements. Control limits are statistically derived (e.g., ±3σ) thresholds used to detect process instability (SPC) and trigger preventive actions but do not by themselves determine batch acceptance. Conflating the two causes either false rejections or missed failures.

• Specification limits: define pass/fail for product/material; linked to release.
• Control limits: define process stability; linked to investigation or adjustment.
• OOT: may breach control limits while still in spec; requires trending and risk-based assessment.
• OOS: explicitly outside specification; requires investigation and documented disposition before release.

MES should implement both checks concurrently: SPC monitors for drift (quality risk management per ICH Q9), whereas spec checks control release. When both fire, the release decision is governed by specifications, with SPC data enriching the root cause analysis.

ISA‑95 positions specification governance and conformance evaluation primarily at Level 3 (MES/LIMS), with enablement at Level 2 (SCADA/DCS/PLC) for in-process alarms and interlocks. Level 4 (ERP/QMS) holds master data approval and post-execution dispositions (e.g., release, NCR/CAPA), while Level 1/2 execute real-time measurements. A robust architecture ensures a single source of truth for limits, propagated deterministically to edge capture points and recorded with context.

Integration patterns should ensure versioned specification sets travel with the order/batch and are time-aligned to the measurement (e.g., by master data effective date or batch start timestamp), preventing misapplication of superseded limits.

21 CFR Part 211 requires scientifically sound laboratory controls and that each lot meet established specifications before release (e.g., §§ 211.160, 211.165). FDA’s OOS guidance mandates a phase-appropriate, thorough OOS investigation, prohibiting practices like retesting until passing without root cause or averaging failing and passing results to achieve compliance. For time-sensitive products (e.g., radiopharmaceuticals), real-time decisions still require documented, scientifically justified acceptance criteria and immediate segregation on failure.

• Establish and approve specifications via change control; align to ICH Q6A.
• Evaluate results using validated methods; preserve original data and audit trails.
• On OOS: initiate a two-phase investigation (laboratory assessment then full manufacturing review) with risk-based scope (ICH Q9).
• Disallow data manipulation to pass; only invalidate a test when a clear, assignable laboratory error is proven and documented.

21 CFR Part 820.90 requires identification, documentation, evaluation, segregation, and disposition of nonconforming product in medical devices, including process controls that detect and prevent unintended use. For dietary supplements, 21 CFR Part 111 mandates component and finished-product specifications (e.g., §§ 111.70–111.75) and requires that a batch meet specifications before release, or be rejected/approved under a documented material review and disposition process by quality personnel. In food CGMP/PC, failures of critical limits under a preventive control program necessitate corrections, corrective actions, and documentation.

Across these sectors, an OOS or nonconforming condition demands immediate status control (quarantine), documented evaluation, and traceable disposition; electronic systems must support role-based segregation and ensure records are attributable, legible, contemporaneous, original, and accurate (ALCOA+), with appropriate electronic signatures where required.

Specifications are master data that require formal lifecycle control: drafting, review, approval, effectiveness dating, and periodic review. They must include test identification, analytical procedure references, units, rounding/significant-figure rules, normal ranges, action limits (specifications), sampling plans, and any guard-banding policy. ICH Q6A emphasizes that acceptance criteria be clear and justified; ICH Q9 recommends risk-based evaluation of the impact of changes.

• Single source of truth for limits, referenced by site, product, material grade, and revision.
• Effective-dating and time-alignment to batch start/step start to prevent misapplication.
• Explicit rounding rules consistent with the validated test method; record unrounded raw values.
• Controlled propagation to MES/LIMS/SCADA with verification (checksum/dual-approval) and audit trails (GAMP 5, Part 11/Annex 11 principles).

Any change to specifications must trigger risk assessment (ICH Q9), impact analysis (procedures, training, equipment), and, where necessary, revalidation of methods or processes. Historical batches must retain evaluation under the limits in force at time of execution; re-interpretation requires formal rework or deviation with justification.

MES must evaluate in-spec/OOS at the point of capture for in-process checks (e.g., blend uniformity, compression force, fill weight, torque) and environmental/utility parameters where they are release-relevant. Interlocks at Level 2 should prevent continuation on critical failures; for noncritical parameters, soft holds and supervisor reviews may apply. Automated status changes (quarantine/reject) must be deterministic and recorded with the value, limit set, timestamp, equipment ID, and operator/e-signature.

• Hard stops on critical quality attributes (CQAs) or critical process parameters (CPPs); prevent proceeding until resolved.
• Checkweigher/vision rejects logged as OOS events or as nonconformances based on disposition rules.
• Sampling plans linked to lot genealogy; each unit/aliquot tied to its result and decision.
• Real-time dashboards that clearly distinguish spec failures from SPC alerts to avoid misinterpretation.

OOS decisions hinge on consistent application of method-defined significant figures and rounding. The system should evaluate pass/fail using unrounded raw values and display rounded results per the validated method to avoid pass-by-rounding artifacts. Where measurement uncertainty is material, risk-based guard bands around specification limits can reduce false accepts; if used, they must be preapproved and consistently applied.

1. Initial evaluation: compare result to specification using predefined rounding rules and units; classify immediately.
2. If OOS: initiate laboratory assessment to determine assignable error; preserve original data and audit trail.
3. If lab error is proven: invalidate the test with justification and metadata; perform a full retest.
4. If no lab error: treat confirmatory tests as part of the investigation but do not average results to mask failure; proceed to full manufacturing review and disposition.

"It is inappropriate to discard results merely because they do not meet specifications; an investigation is always required to determine the cause of the OOS result."

A compliant OOS workflow begins with rapid containment (quarantine), followed by a structured investigation: laboratory assessment (method, analyst, instrument, reagents) and a full manufacturing review (materials, equipment status, environmental conditions, deviations, SPC trends). Integrate evidence from LIMS (raw data), MES (execution history), calibration/maintenance systems, and supplier quality records. Link root cause analysis (e.g., 5 Whys/Fishbone) to CAPA with defined effectiveness checks.

• Documented decision tree aligned with FDA OOS guidance; quality unit oversight and approval.
• Traceable disposition (release, rework, reject) with rationale; retain all data per record retention rules.
• Lot and sublot genealogy update to propagate status and prevent unintended use in downstream orders.
• Metrics: OOS rate by test/material, mean time to containment, investigation cycle time, CAPA recurrence rate.

Because in-spec/OOS gates determine product disposition, electronic records and signatures must satisfy data integrity expectations (ALCOA+) and be validated proportionate to risk (GAMP 5). Ensure secure, computer-generated, time-stamped audit trails capturing creation, modification, and evaluation of results and limits. Role-based access and segregation-of-duties should prevent unauthorized specification edits and ensure independent quality review. Test method interfaces (e.g., instrument data capture) should avoid manual transcription where feasible.

• Requirements traceability: specification data elements and decision logic mapped to test protocols (CSV/CSA).
• Negative/edge-case testing: boundary values at the limit, unit/decimal mismatches, time-zone and daylight-saving edge cases.
• Audit trail review workflows focused on OOS-critical events and overrides.
• Backup/restore and disaster recovery validated to preserve both raw data and evaluation outcomes.

Beyond compliance, in-spec/OOS execution data informs continuous improvement. Differentiate special-cause failures (true OOS) from common-cause variability trending toward limits (OOT) and address via control strategies. Use risk-based dashboards: OOS Pareto by failure mode, capability indices for key attributes (Cp, Cpk), and timeliness metrics (containment within X minutes). Tie learnings back to control strategies and specifications, updating only via controlled change.

• Top OOS contributors by material lot and equipment state to target maintenance or calibration actions.
• Correlation of OOT signals to subsequent OOS to refine preventive thresholds.
• Effectiveness checks on CAPA closing the loop on recurrence within a defined window.
• Supplier scorecards including incoming OOS rate and corrective action responsiveness.

V5 Ultimate keeps specifications as governed master data, effective-dated and versioned, consumed natively by MES, LIMS, and eBMR/eDHR. Results are evaluated at capture with method-defined rounding and units; failures drive automatic holds and open linked OOS/deviation records in QMS. Lot statuses synchronize to WMS to physically quarantine. Investigations draw evidence from execution history, lab raw data, equipment maintenance, and supplier files—closing CAPA and release decisions without reconciliation across systems.

• Using SPC control limits as acceptance criteria: keep distinct; govern specs via quality approval.
• Pass-by-rounding: evaluate unrounded raw values, display per method rounding; document rules in the specification.
• Averaging to mask failures: prohibited by FDA OOS guidance; never average failing and passing tests to pass.
• Specification drift without change control: enforce effective-dated versions and independent quality approval.
• Manual transcription at critical gates: integrate instruments to reduce error; if unavoidable, require witnessed entry and enforce forced-signature steps.
• Delayed containment: automate quarantine on OOS to limit exposure; measure containment time as a KPI.

An effective program separates monitoring from disposition, prevents ad hoc overrides, and ensures that every OOS has a transparent, risk-based investigation trail to a justified disposition.