What pressure differential, air change rate and clean-up time does a GMP cleanroom need?

1The numbers, and how binding each one is.

For a GMP manufacturer, none of these numbers is a legal limit.

• The US regulations set no figure. Aseptic processing needs HEPA-filtered air “under positive pressure”, and equipment for “adequate control over air pressure” when appropriate.⁵⁶ FDA’s guidance is labelled “Contains Nonbinding Recommendations” and offers its 10 to 15 Pa and 12.5 Pa figures as examples.²
• Annex 1 calls its three numbers guidance values. It gives 10 Pa, 20 minutes and 0.36 to 0.54 m/s each as a “guidance value”, and for air speed adds “unless otherwise scientifically justified in the CCS”.¹ PIC/S PI 058-1, in force since 28 October 2026, accepts other speeds justified in the CCS, for powders where the range itself risks contamination and for barrier systems whose airflow studies show first air protection.⁷
• The UK and PIC/S use the same annex. MHRA points manufacturers to EudraLex Volume 4, which holds it,⁸ and PIC/S, the EU and WHO agreed to interpret it the same way.⁷
• Non-sterile EU sites get no figure. The EU GMP guide asks for production areas to be “effectively ventilated”, and lists “appropriate use of air-locks and pressure cascade” among measures against cross-contamination.⁹¹⁰ For a number, WHO recommends a minimum of 5 Pa for containment.⁴
• Containment can reverse the cascade. For pathogenic, highly toxic, radioactive or live viral or bacterial materials, Annex 1 allows positively or negatively pressurised airlocks, but air that must flow into a critical zone should come from the same or a higher grade.¹
• Pharmacies compounding under section 503A follow USP <797> instead. Section 503A exempts their drugs from CGMP on conditions that include compounding to USP’s chapters, which states, not USP, enforce.¹¹¹² See 503A vs 503B.

1.1Air changes per hour for ISO 8, ISO 7 and ISO 5.

Air changes per hour is the air supplied to a room in m³ per hour divided by the room volume in m³.³ Only FDA’s guidance ties a figure to a cleanroom class: at least 20 for ISO 8 support rooms is “typically acceptable”, and “significantly higher” rates are normally needed for ISO 7 and ISO 5, with no number.² Annex 1 never uses the term; it judges the air supply by classification, clean-up time and airflow studies.¹ Any per-class range you find for ISO 7 or ISO 5 is a design rule of thumb, not a requirement. ISO 14644-1 defines its classes by particle concentration, not airflow.¹³

Sizing it from the room. A textbook estimate, not a regulatory method: with perfect mixing and clean supply air, a room settles where concentration equals particle emission ÷ supply air, so supply air = emission ÷ target concentration. ISO 14644-4:2022 recommends using the air supply rate to calculate contaminant dilution, and calls the rate at which a person, machine or object emits particles its source strength.¹⁴ Example, invented: a 90 m³ ISO 7 room whose operators and filling machine are estimated to emit 288 million particles ≥0.5 µm an hour. To hold 80,000 per m³, under a quarter of the 352,000 limit to allow for imperfect mixing and busy moments,² it needs 288,000,000 ÷ 80,000 = 3,600 m³/h, or 40 air changes an hour. Halve the emission with better gowning or fewer people and half the air holds the same level. Then prove the figure by classification in operation and a measured clean-up time.¹

Under unidirectional airflow the measure is air speed: 0.45 m/s from a full filter ceiling 3 m high supplies 540 air changes an hour (0.45 × 3,600 ÷ 3), so an ISO 5 air change figure only restates the speed.

2Example: a cascade and air change check for a filling suite.

The suite and every figure below are invented. The filling line runs in a RABS, so its background should be at least grade B, and the final stage of an airlock takes the grade of the room it leads into.¹

2.1The alarm-band trap.

Suppose each room alarms only on its own pressure against the corridor, at ±5 Pa from its set point. The airlock can read 40 Pa and preparation 35 Pa with no alarm: a 5 Pa step. If each transmitter is accurate to ±1 Pa, the true step could be 3 Pa. That passes WHO’s minimum test, that cumulative instrument tolerances “should not cause a situation where an undetected reversal of airflow is possible”.⁴ It still lets the step fall far below Annex 1’s 10 Pa guidance value unseen.¹

2.2Across each door or to a common reference.

Annex 1 wants pressure indicators “fitted between cleanrooms”, which can be wired two ways.¹

• Across each door. One transmitter reads the step that protects the door, with one instrument’s tolerance.
• To a common reference. Every room is read against one point, such as the corridor, which WHO names as one way to keep cumulative tolerances from hiding a reversal.⁴ A step between two rooms is then the difference of two readings and carries both tolerances, so alarm on that calculated step, or the trap above returns.

Either way, WHO asks for room pressure to be traced back to ambient “by summation of the room pressure differentials”, and for gauges whose range and graduation can be read to an appropriate accuracy.³ Choose a range that runs below zero, such as −10 to +30 Pa for a 15 Pa step, so a reversal is recorded as a negative value rather than clipped at zero. The suite measures across each door, with transmitters accurate to ±1 Pa.

Annex 1 wants any alarm delay “assessed and justified within the CCS”; WHO likewise wants “justified dead times”.¹³ FDA calls it critical that “the time a door can remain ajar be strictly controlled”, so separate door-open alarms sound after 20 seconds on personnel doors and 45 seconds on the trolley door.²

2.3Air changes and clean-up time.

On the same textbook model, particle concentration falls by a factor R in ln(R) ÷ N hours, where N is the air change rate. The factor comes from Annex 1 Table 1, set out in cleanroom grades: grade B should fall from its in-operation limit of 352,000 particles ≥0.5 µm/m³ to its at-rest limit of 3,520, a factor of 100; grade C from 3,520,000 to 352,000, a factor of 10.¹

Both measured times beat the 20-minute guidance value and run longer than the estimates, because no real room mixes perfectly. The measured period goes into the procedure that restores the rooms after a disruption, as Annex 1 4.29 asks.¹

3Records an inspector will ask for.

4When a pressure alarm sounds.

1. Acknowledge and record it. Annex 1 says the warning “should not be overridden without assessment”; FDA says all alarms should be documented and deviations from established limits investigated.¹²
2. Read the trace, not only the alarm. How low the differential fell, whether it reversed and for how long, against the delay you justified.
3. Find the cause. A held door, a failed interlock, a machine extract starting, a transient WHO tells designers to allow for, or a faltering air handling unit.³
4. Assess the exposure. Was product or sterile equipment open in grade A? Check particle counts and viable results near the door for that window, against the critical differences your CCS names.¹
5. Restore and confirm. When an aseptic room’s overpressure to an unclassified neighbour falls below its minimum, FDA says its environmental quality should be “restored and confirmed”; Annex 1 points to the clean-up period you qualified.²¹
6. Escalate by rule. An action-level breach or a reversal into a higher grade opens a deviation.
7. Trend the alarms. Annex 1 asks for alarm events to be evaluated for trends, with critical alarms reviewed immediately.¹ A door that trips every shift needs a design or procedure fix, not a longer delay.

If the air stops. Annex 1 also wants a warning of “any failure in the air supply”, and lists an interruption of air movement among the changes that can call for requalification.¹ FDA wants deviation or change control to cover air handling shutdowns and a written procedure for returning to operating conditions, and says “Any disruption in power supply, however momentary, that could affect product quality is a manufacturing deviation and must be included in batch records”.² Record when the air stopped and restarted, what was open, the decision on product in process and on breaking down the line, and the cleaning, clean-up period and monitoring before restart. Keep the monitoring system recording through the outage, or there is nothing to assess.

5What qualification and requalification test.

Qualification includes airflow volume and velocity, the pressure difference, airflow visualisation and a recovery test, referenced to ISO 14644.¹ Requalification, at least every 6 months for grades A and B and every 12 for C and D, needs at least five tests: classification, final filter integrity, airflow volume, the pressure difference between rooms, and air velocity, which a recovery test replaces where airflow is not unidirectional.¹ Measure air speed “at the working position”, justify the location in the protocol, and make the airflow visualisation agree with it; see smoke studies.¹

6What inspectors cite.

FDA’s letter to a Greek sterile manufacturer named both an inadequate cascade and the missing record:

“Furthermore, you failed to routinely record basic facility monitoring data, including differential pressure, temperature, and humidity. Your monitoring devices displayed real-time data only and lack data storage capability. As a result, parameter excursions that occur when operators are not actively monitoring pressures went undocumented. You also had no alarm history or recorded data.” Pharmathen International S.A., warning letter, 27 May 2026¹⁵

The same letter found the aseptic filling rooms “had lower pressure than the adjacent rooms”.¹⁵ A Form 483 issued in August 2026 to a Texas producer of sterile drugs recorded, as its second observation:

“A pattern of frequent or acute pressure reversals was observed from areas of less clean air to areas of higher air cleanliness.” Houston Rx LLC, Form FDA 483, 28 August 2026¹⁶

Its records showed every reading for six weeks below zero between the ISO 7 buffer room and the ISO 8 ante room, and every lot made in that time had been released.¹⁶ FDA’s 2020 guidance on insanitary conditions at compounding facilities lists that pattern among conditions so serious that it “strongly recommends” an immediate recall of purportedly sterile drugs and a halt to sterile operations.¹⁷

European inspectors cite the same failure. Portugal’s medicines authority found this at an Indian maker of sterile cephalosporin powder for injection:

“Critical failures were observed in facility design and pressure cascade maintenance, including inadequate differential pressures between classified and non classified areas, reversed pressure gradients, and layout constraints that resulted in sterile API being exposed in Grade B.” Ananta Medicare Limited, statement of non-compliance FT094/MH/001/2024/NCR, EudraGMDP, 6 August 2026¹⁸

7How it runs in Seal.

In Seal, the contamination control strategy connects rooms, grades, pressure cascades and personnel and material flows to approved layouts, qualification and routine monitoring, so a door alarm or a pressure failure identifies which risks and product-exposure windows need assessment. Pressure instability can be combined with other signals, such as repeated low recoveries or a flora shift, into one CCS signal, and a room or HVAC change cannot close until the affected CCS links and narrative are reviewed.

In commissioning and qualification, an air handler is qualified together with its rooms, controls, alarms, sensors and monitoring, and a critical alarm challenge records the stimulus, the expected alarm and the response.