Almost every design review of a sterile facility features a table of room pressure setpoints that nobody can justify. Months later, during qualification, smoke drifts back through the airlock as the door opens, and a return grille placed high for site convenience short-circuits part of the supply before it reaches the critical zone.
Those numbers were values chosen instead of an airflow strategy. A pressure differential is the effect of a flow imbalance across the leakage paths of a room: without envelope tightness, door regime and the actual air path, the figure on the mimic panel describes nothing verifiable. Airflow and pressure cascade are a single design object, part of the Cleanrooms & HVAC Systems track.
Why the choice reaches beyond the design phase
The airflow strategy sets the design air volume, and with it AHU size, duct sections, terminal filter area and installed power, but also the acceptance criteria of qualification: Annex 1 §4.25 lists air flow and velocity, differential, airflow direction and visualisation. Upstream sit the HVAC architecture and the zoning and classification decisions.
Regulatory framework: what actually binds
Binding text, expectations explicitly qualified as guidance and voluntary method are worth separating from the outset.
| Source | Status | What it actually binds |
|---|---|---|
| Annex 1 §4.14–4.16 | Applicable since 25 August 2023 | Positive pressure and/or airflow relative to the lower-grade background in all operational conditions; 10 Pa between adjacent grades stated as "(guidance value)"; indicators installed; critical differentials continuously recorded with alarms; airflow patterns visualised |
| Annex 1 §4.12–4.13, §4.30, §4.32 | As above | PAL and MAL separated, interlocking for grade A/B; 0.36–0.54 m/s at the working position as guidance; requalification including verification of the differential |
| EudraLex Chapter 3 §3.12 and Chapter 5 §5.21 | Operational since 1 March 2015 | Effective ventilation appropriate to products and operations; airlocks and pressure cascades among the technical measures against cross-contamination |
| FDA, Aseptic Processing Guidance | September 2004, nonbinding | Illustrative recommendations: at least 10–15 Pa between adjacent rooms of different classification with doors closed, at least 12.5 Pa towards an unclassified room, 0.45 m/s ±20 % as typical value |
| 21 CFR 211.42(c)(10) and 211.46 | eCFR as of 27 August 2026 | HEPA-filtered air under positive pressure for aseptic processing, temperature and humidity control, environmental monitoring. No numbers |
One verified absence matters as much as a requirement: Annex 1 contains no air change rate requirement. The number of air changes is derived to achieve the classification and the clean up period established in qualification, for which §4.29(iii) gives less than 20 minutes as a guidance value [GUIDANCE]. The only citable ACH figure is FDA's: at least 20 air changes per hour, "typically acceptable" for ISO 8 supporting rooms only [GUIDANCE].
How air actually moves inside the room
Three regimes, three test criteria
Unidirectional airflow protects by sweeping particles away from the point of exposure: performance is measured by velocity at the working position, not by total air volume. Annex 1 §4.30 requires velocity and measurement location to be justified in the protocol and gives a guidance range of 0.36–0.54 m/s at the working position [GUIDANCE], unless scientifically justified otherwise in the CCS. FDA reports 0.45 m/s ±20 % as a typical value in a footnote, historically referred to the filter face [GUIDANCE]: it is not a requirement, and it is not the same measurement location.
Non-unidirectional airflow protects by dilution, and its effectiveness depends on the uniformity of mixing more than on volume. Mixed airflow superimposes localised unidirectional protection on a non-unidirectional background: the weak point is the transition, where the absence of background entrainment must be demonstrated.
Supplies, returns and disturbances
Supply and return positions decide the real air path: a high return close to the supply short-circuits the room, and clean air returns to the system unused. Three disturbances are design inputs, not qualification surprises. Obstructions: any horizontal surface above the critical point creates an aerodynamic shadow and a recirculation, which makes equipment layout an HVAC input. Thermal plumes: hot surfaces, motors and people generate rising currents that at low velocity oppose a downward flow and carry particles upward, the scenario §4.15 requires to be excluded through visualisation [REQUIREMENT]. Personnel are both a mobile source and a mobile obstacle, and they define the working position. None of this shows in a pressure table: it is seen through the airflow visualisation study under dynamic conditions.
The cascade is an air balance, not a table of Pascals
Room tightness and doors as design variables
The differential is the effect of the air displaced through joints, service penetrations, ceilings, pass-boxes and door undercuts: for the same target, a leaky room demands far more of it. Tightness is therefore a contractual specification, verified before TAB, because every later penetration shifts the balance; the construction criteria are covered in the article on finishes, airlocks and hygienic construction.
Door opening is equally predictable, with a frequency, a duration and a swing direction: it determines how long the cascade sits outside setpoint. FDA qualifies its recommendation of at least 10–15 Pa as applying "with doors closed" [GUIDANCE], because with a door open the protecting parameter is not the differential but the direction of flow through the opening. A mature design sizes that flow and verifies it during a real opening.
Product protection and containment: when the sign has to flip
Protecting the product means moving air from the cleaner room towards the less clean one, with the critical zone at positive pressure. Containment means the opposite: the processing area is held at negative pressure. With potent products or biological material both objectives apply and the cascade cannot have a single sign.
The separation is built with airlock geometry: an airlock at positive pressure between two rooms under negative pressure preserves the intended direction on both sides, while a negative one acts as a sink. The cascade, sink and bubble taxonomy is engineering terminology and must be declared as such, because Annex 1 distinguishes only PAL and MAL. §4.14 foresees containment arrangements, but for a sterile product any departure from positive pressure must be justified in the CCS with documented compensating measures. HBEL values under the EMA guideline feed the risk assessment but do not address zoning or air handling.
§4.12 requires airlocks flushed with filtered air and PAL separated from MAL; §4.13 prohibits simultaneous opening of airlock and pass-box doors and requires interlocking for grade A/B [REQUIREMENT]. The interlock needs an emergency override, a bypass alarm and event recording, and the minimum interval between openings follows from the clean up period established in qualification.
Instrumentation, alarms and failure scenarios
§4.16 requires installed differential indicators and, for those identified as critical, continuous monitoring and recording with alarms [REQUIREMENT], with setpoints and criticality considered in the CCS. Three choices follow: pressure taps away from diffusers and doors, an unambiguous reference, and instrument class consistent with the alarm band. Every alarm delay must be justified with site qualification data, as discussed in the article on operating ranges and alarms.
| Failure scenario | Effect on the cascade | What the system must do | What the procedure must do |
|---|---|---|---|
| Door held open beyond the expected duration | Local collapse of the differential | Direction sustained by the sized through-door flow; abnormal duration signalled | Maximum duration defined; clean up restored before resuming |
| Supply fan trip or changeover to the standby unit | Loss of differential across several interfaces at once | Defined duty/standby sequence, immediate alarm, no reversal | Suspend aseptic activity, assess batch impact, restart from the qualified state |
| Sensor drift or failure | False reading with a correct physical cascade | Detect out-of-range values and inconsistencies between correlated points | Verify with a calibrated instrument and assess suspect data |
| Local exhaust started without compensation | Sign reversal at an adjacent interface | Interlock exhaust with supply compensation; alarm on lost direction | Use prohibited outside the qualified configuration |
| Power loss with disorderly restart | Transient with multiple reversals | Restart sequence stabilising the critical interfaces first | Qualified clean up and documented return to the qualified state |
CFD: a predictive design tool, not evidence of compliance
CFD earns its place early, when the room does not exist and alternatives are cheap: comparing return positions, assessing obstructions and heat sources, finding stagnation and recirculation. Its limits are structural: the result depends on boundary conditions, turbulence model, mesh and assumptions about heat sources and people, and the worst-represented phenomena — doors and moving personnel — dominate real risk. CFD produces hypotheses to be verified: it does not replace qualification, nor the visualisation study required by §4.15, nor field testing with the methods of ISO 14644-3 [GUIDEGXP]. Every critical prediction must be confirmed in the field.
Justifying the choices with QRM
ICH Q9(R1), Step 4 on 18 January 2023, is the methodological frame: Annex II.4 applies QRM to facilities, equipment and utilities, including zoning and qualification scope. It remains methodology and contains no HVAC design parameter. §5.1 treats formality as a continuum; §5.3 addresses bias, and here the dominant bias is anchoring on the Pa value of the previous project. The output belongs in the CCS under §2.3–2.5: interfaces with sign and criticality, failure scenarios, compensating measures and alarm criteria.
Worked example: Site Vega
Site Vega is a realistic but entirely fictitious example: aseptic filling of a potent biological product in a barrier system with a grade B background and, next to it, a preparation room whose HBEL value calls for containment. The two requirements conflict along one grade C corridor.
The answer is not a compromise on setpoints but a change of geometry: an airlock held at positive pressure relative to both sits between preparation room and corridor, with MAL separated from PAL. Before TAB the team measures tightness, finds a leakage area larger than assumed because of penetrations above the ceiling, and seals rather than raising the differential air volume. Door openings are simulated at worst case during visualisation, giving the maximum duration written into the procedure and the alarm delay.
Evaluation matrix for cascade options
The matrix compares options at a single critical interface; the team assigns the weights, because no weighting transfers between sites [QRM].
| Criterion | Weight | Positive cascade | Negative cascade | Positive airlock between opposite signs |
|---|---|---|---|---|
| Protection of the exposed product | Direct and verifiable | Requires barrier or separative system | Preserved on both sides | |
| Operator and environment protection | Not provided by the cascade | Primary objective | Provided towards the outside | |
| Robustness to a single failure | Reversal on loss of supply | Reversal on loss of exhaust | More points to monitor | |
| Qualification effort | Fewer interfaces | Containment verification | More scenarios to demonstrate |
Levels of prescriptiveness
| Statement | Level | Source |
|---|---|---|
| Airflow patterns visualised, indicators installed, critical differentials continuously recorded with alarms, PAL and MAL separated, interlocking in A/B | [REQUIREMENT] | Annex 1 §4.12–4.16 |
| Minimum 10 Pa between adjacent grades and 0.36–0.54 m/s at the working position; FDA cites 10–15 Pa with doors closed, 12.5 Pa towards an unclassified room and 0.45 m/s ±20 % in a footnote | [GUIDANCE] | Annex 1 §4.14 and §4.30, "(guidance value)"; FDA 2004, nonbinding |
| Pressure differential, airflow direction and recovery have standardised methods that set no acceptance criteria | [STANDARD] | ISO 14644-3:2019 and -4:2022 |
| Differential air volume is sized on measured tightness and the door regime | [GEP] | Good engineering practice |
| Interface signs and differential criticality justified by documented risk analysis | [QRM] | ICH Q9(R1) Annex II.4; CCS §2.3–2.5 |
| CFD enters the dossier with declared assumptions and predictions confirmed in the field | [GUIDEGXP] | GuideGxP recommendation |
Operational checklist
- State the airflow regime and the protective function required for each room.
- Assign a sign to every interface and justify the choice between protection and containment.
- Specify envelope tightness in the contract as a verifiable requirement.
- Recalculate the differential air volume on measured tightness before TAB.
- Count door openings per shift and size the flow through the opening.
- Include obstructions, heat sources and personnel in the visualisation scenarios.
- Classify critical differentials in the CCS and define their monitoring and alarms.
- Document each failure scenario with the system response and the procedural response.
- Confirm every critical CFD prediction in the field and archive the comparison.
- Plan periodic requalification including differential verification within the §4.32 intervals.
Recurring mistakes and red flags
- Presenting an air change rate as a regulatory requirement.
- Attributing the 0.36–0.54 m/s range to ISO 14644, or treating 0.45 m/s as an obligation.
- Declaring every differential critical, generating alarms that erode response.
- Using a CFD report in place of visualisation or field testing.
- Using the cascade, sink and bubble taxonomy as if it were regulatory.
If this way of separating requirement, guidance and good practice is useful in daily work, the discussion continues in our technical newsletter The Pragmatic GMP.
Key takeaways
- The differential is an effect of air volume and tightness: design starts there, not from a Pa value.
- Annex 1 §4.14 gives 10 Pa as a guidance value; 10–15 Pa with doors closed and 12.5 Pa towards an unclassified room are illustrative FDA recommendations.
- Each airflow regime has its own test criterion, from working-position velocity to the absence of entrainment.
- Product protection and containment may require opposite signs: airlock geometry resolves them.
- CFD anticipates problems but never demonstrates compliance: it is confirmed in the field.
References
- EudraLex Volume 4, Annex 1, applicable since 25 August 2023.
- EudraLex Volume 4, Annex 15, operational since 1 October 2015.
- FDA, Sterile Drug Products Produced by Aseptic Processing, September 2004.
- 21 CFR Part 211, 211.42 and 211.46.
- ISO 14644-3:2019 and ISO 14644-4:2022.
- ICH Q9(R1), Step 4 of 18 January 2023.
