Cleaning validation

Cleaning validation is the documented programme that proves a written cleaning procedure, executed as written, reduces three things on shared product-contact surfaces to acceptable, justified limits: residual active ingredient from the previous product (chemical), residue of the cleaning agent itself (chemical), and bioburden and endotoxin (microbiological). The proof is empirical — three consecutive successful runs on the worst-case product, worst-case equipment, worst-case soiling condition — not a single 'we cleaned it and it looked clean'.

It sits inside 21 CFR 211.67 ('equipment shall be cleaned, maintained, and… sanitized at appropriate intervals'), is mandated explicitly by EU GMP Annex 15 §10, and since 2014 EMA's PDE/HBEL guideline replaced the legacy 1/1000-of-minimum-dose and 10 ppm limits with health-based exposure limits derived toxicologically.

Before 2014 most sites used one of two empirical limits — 1/1000 of the minimum therapeutic dose of the previous product carried into the maximum daily dose of the next, or 10 ppm of the previous product in the next. Both are now considered insufficient on their own. EMA's 2014 guideline (EMA/CHMP/CVMP/SWP/169430/2012) requires every shared product to have a Permitted Daily Exposure (PDE), now usually called Health-Based Exposure Limit (HBEL), derived from a No-Observed-Adverse-Effect-Level (NOAEL) divided by uncertainty factors F1–F5.

The HBEL is then converted into a Maximum Allowable Carryover (MACO) for each equipment train: MACO = HBEL_next × MinBatchSize_next ÷ MaxDailyDose_next. MACO is then divided by the total shared surface area to give a swab/rinse limit in µg/cm². The 1/1000 and 10 ppm limits are still allowed as a sanity cross-check, but the HBEL-derived limit governs — and if HBEL is tighter (as it usually is for cytotoxics, hormones, β-lactams), HBEL wins.

Sites do not validate every product/equipment combination — they bracket. Product grouping ranks each product by a combined score of solubility, toxicity (HBEL) and difficulty-to-clean visual rating, picks the worst case in each group, and validates the worst case as a proxy for the rest. Equipment grouping does the same: a family of identical fluid-bed dryers is validated on the largest (highest surface area) unit. Worst-case soiling means dirty hold time (maximum time between end-of-process and start-of-cleaning), maximum batch size, maximum campaign length.

• Product worst case: lowest solubility, lowest HBEL, hardest to remove (visually, or by rinse recovery).
• Equipment worst case: largest shared surface, hardest-to-access geometry (filter housings, dip tubes, valve clusters), most porous material (silicone, PTFE).
• Process worst case: longest dirty hold time, longest clean hold time, lowest cleaning-agent concentration in the approved range, lowest temperature, lowest contact time.
• Operator worst case: least-experienced trained operator (where manual cleaning is used).

Three sampling methods, each with a known recovery factor that must be established before the validation protocol runs. Swab sampling targets a 25 cm² (or 100 cm²) coupon on a representative location — including the hardest-to-clean sites identified during the cleaning DQ. Rinse sampling collects the final rinse water and measures residue per unit volume, then back-calculates to surface concentration using the equipment train wetted area. Visual inspection is mandatory under EMA Q&A on HBEL — 'visually clean' is a necessary but not sufficient acceptance criterion.

Recovery studies are non-negotiable. A swab applied to a stainless coupon spiked with known product residue must recover ≥ 70 % of the spike (industry-typical floor) and the recovery factor (RF) is then used to back-correct field swab results: True residue = Measured residue ÷ RF. A site that reports swab residue without recovery correction is reporting the wrong number — and it's one of the top cleaning-validation 483 observations.

Most modern cleaning programmes use TOC as the workhorse for rinse sampling (broad, fast, sensitive to sub-ppm) and reserve HPLC for product-specific verification at protocol qualification or when investigating an excursion.

Three time parameters must be validated alongside the cleaning procedure itself. Dirty Hold Time (DHT) is the maximum interval between end-of-process and start-of-cleaning before the cleaning procedure can no longer achieve the limits — typically 24–72 h. Clean Hold Time (CHT) is the maximum interval between end-of-cleaning and start-of-next-batch before bioburden regrowth makes the equipment unfit — typically 7–14 days, sometimes 30. Campaign length is the maximum number of consecutive batches of the same product allowed between full cleanings (vs minor intermediate rinses); each end-of-campaign clean must still meet the chemical and microbiological limits.

ASTM E3106 and the FDA's 2011 Process Validation guidance both pushed cleaning validation into the same three-stage lifecycle as process validation: Stage 1 cleaning-process design (DoE on detergent concentration, temperature, time, action — the four CIP variables); Stage 2 cleaning performance qualification (the three consecutive runs against the limits); Stage 3 continued cleaning-process verification (ongoing monitoring with periodic re-swab, trend on rinse data, change-triggered re-qualification). 'Validate once, never look again' is no longer defensible.

Monitoring in Stage 3 is typically a reduced-sampling regime — every nth campaign-end clean is re-swabbed, every rinse is conductivity-monitored, and any HBEL change, new product introduction, equipment modification, or detergent reformulation triggers re-evaluation under change control.

1. HBEL/PDE report missing for one or more shared products — limits still set on 1/1000 and 10 ppm only.
2. Swab recovery factor never established, or established once five years ago and not re-verified after method or coupon change.
3. Worst-case rationale documented but the worst-case product changed when the portfolio changed — and the rationale wasn't updated.
4. DHT and CHT validated for 48 h / 7 d, batch records routinely show 72 h DHT.
5. Visual-clean inspection performed by the cleaning operator (self-inspection) with no independent verification.
6. Cleaning agent residue (detergent) limit set but never measured — conductivity used as a proxy without a correlation study.
7. Final-rinse TOC results trending upward over 18 months — no action taken because each result is still inside spec.
8. Equipment train change (added a transfer hose) processed under change control but cleaning validation impact assessment marked 'no impact' without justification.

• Every shared piece of equipment carries an asset record with surface area, geometry, material and the swab locations defined during DQ — the cleaning protocol generates the swab map automatically.
• Product master holds HBEL/PDE, solubility class and difficulty-to-clean rating; MACO and swab limits recalculate automatically when a product is added, retired or HBEL revised.
• Recovery factors live in the analytical method record with their own expiry; an out-of-window RF blocks acceptance of any swab using that method until re-verified.
• Dirty hold time and clean hold time are timestamps on the equipment log — countdown on the kiosk; exceeding either creates a deviation automatically and blocks the next batch from release.
• Campaign counter on the work-order chain — when the validated limit is reached, the next work order is blocked from release until the end-of-campaign clean is recorded and accepted.
• Stage 3 dashboard trends rinse TOC, conductivity and swab residue with rolling Cpk against the limit; trend toward the limit opens a CAPA before the OOS.