EU GMP Annex 17 of EudraLex Volume 4 is the GMP annex that governs Real Time Release Testing (RTRT) and parametric release: the authority-approved option of releasing a batch based on process data and material attributes measured during manufacture, instead of routine end-product testing. Rewritten in 2018, the annex turned a document that only covered parametric release of terminally sterilised products into a general framework for pharmaceutical manufacturing, aligned with ICH Q8, Q9, Q10 and Q11. In this article we look at what Annex 17 actually covers, the requirements it places on anyone adopting RTRT or parametric release, and what changes in practice for QA and the Qualified Person.
Annex 17: what it covers and why it was rewritten
The previous version of the annex dealt exclusively with parametric release of products sterilised in their final container. The revision — with a deadline for coming into operation of 26 December 2018, six months after publication — extended the scope to Real Time Release Testing in full: Annex 17 now applies to any stage of the manufacturing process and to any type of finished product, active substance or intermediate. The legal basis is Article 47 of Directive 2001/83/EC (and its veterinary counterpart), and the text should be read together with Chapter 1 of EudraLex Volume 4, Annexes 1 and 15, and the EMA guideline on Real Time Release Testing.
The logic behind the change is the Quality by Design paradigm: if you identified the critical quality attributes (CQAs) during development and built a risk-based control strategy, quality can be evaluated and ensured by a valid combination of in-line measured material attributes and process controls — the RTRT definition taken from ICH Q8 — with a level of assurance equivalent to or better than end-product testing.
The requirements for Real Time Release Testing
Section 3 of Annex 17 sets the conditions. Three deserve particular attention:
- Prior regulatory interaction: RTRT is authorised case by case by the competent authorities, before and during assessment of the dossier; the level of interaction grows with the complexity of the control strategy.
- Demonstrated predictive power: in-process measurements must be accurate predictors of the finished product attributes, and the combination of material attributes and process controls must rest on scientific evidence, typically parallel testing data comparing RTRT with the traditional test.
- Integration into the PQS: the RTRT strategy lives inside the Pharmaceutical Quality System — risk management, change control, training, validation, deviation and CAPA management — and must include a contingency plan for sensor or measurement equipment failure, to be handled as a deviation.
An operational detail that is often overlooked: attributes controlled indirectly through RTRT remain on the certificate of analysis, reported as "Complies if tested" with a footnote referring to the approved approach. And one non-negotiable principle: if an RTRT result fails or trends towards failure, you cannot "fall back" on end-product testing to release the batch — the failure must be investigated and the release decision follows from that investigation. Even a permanent return to end-product testing is not free: it requires a variation to the marketing authorisation.
RTRT, parametric release and control strategy are fast-moving regulatory topics. If you want a practical, no-frills weekly update on GMP and batch release, subscribe to The Pragmatic GMP, our free newsletter: ten minutes of reading to stay audit-ready.
Parametric release of terminally sterilised products
Section 4 of Annex 17 governs the historical and most common case: parametric release as an alternative to the sterility test, based on reviewing the sterilisation process monitoring records (temperature, pressure, time) rather than testing a sample. The rationale is statistical: the pharmacopoeial sterility test has inherently limited sensitivity, while the process data of a validated cycle say far more about the sterility of the batch. The main requirements:
| Requirement | What Annex 17 asks for |
|---|---|
| Permitted sterilisation methods | Only sterilisation in the final container: moist heat, dry heat or ionising radiation (dosimetric release), in line with European Pharmacopoeia requirements |
| Site prerequisites | A history of acceptable GMP compliance and a robust, documented sterility assurance programme |
| Bioburden | Pre-sterilisation bioburden testing on every batch |
| Record verification | Sterilisation records checked by at least two independent systems: two people, or a validated computer system plus a person |
| Approval | Explicit authorisation by the regulatory authorities; after approval, release is based on the approved specifications and review of the critical process data |
Here too the iron rule applies: a batch that fails the parametric release specification cannot be rescued by a passing sterility test. If the cycle is out of specification, the batch is out of specification.
What changes for QA and the Qualified Person
For the QP, RTRT and parametric release shift the centre of gravity of certification from the certificate of analysis to process data: the release decision rests on the review of critical parameters and on the strength of the control strategy, without prejudice to all other required tests and the site's GMP compliance. Approval is in fact site-specific, tied to the level of process understanding and to inspection outcomes: a manufacturing transfer does not "carry" the RTRT approval with it. The dedicated EMA guideline (EMA/CHMP/QWP/811210/2009-Rev1, effective since 1 October 2012, replacing the historical Note for Guidance on parametric release) lists what the dossier must demonstrate: CQAs identified during development, a scientifically grounded control strategy, a validated process, a justified relationship between the traditional test and RTRT with acceptance criteria, and a contingency plan.
The most common mistakes
- Treating RTRT as a laboratory shortcut: it is not a way to "skip" testing, but a more demanding control strategy that requires more process understanding, not less.
- A missing or theoretical contingency plan: a failed PAT probe without an approved plan of temporary alternative testing blocks batch release.
- Unauthorised "hybrid" release: using end-product testing to override an unfavourable RTRT result is exactly what Annex 17 prohibits.
- Forgetting the certificate of analysis: attributes covered by RTRT must still be reported, with the dedicated wording.
- Underestimating change control: every change to sensors, models or parameters of the control strategy must be assessed for its impact on the regulatory approval.
GuideGxP recommendation
If you are considering RTRT or parametric release, start from an honest gap assessment on three fronts: maturity of the control strategy (are the CQAs really linked to predictive in-process measurements?), robustness of the PQS (would deviations, CAPA and change control withstand the extra scrutiny?) and regulatory dialogue (engage the authority before investing in PAT instrumentation). For parametric release of terminally sterilised products, consolidate the sterility assurance programme first: cycle mapping, bioburden under statistical control and independent double-checking of records are the points inspectors look at first. And define in black and white, in a batch release SOP, who does what when a critical parameter goes out of limits.
Batch release under RTRT remains the responsibility of the Qualified Person: our guide Qualified Person (QP) — Operational Guide to Annex 16, Certification and Batch Release covers the entire certification process, including the cases where release is based on process data, with a toolkit of ready-to-use templates.
