A compressed-air report may show compliant particles, moisture and oil while leaving a product-contact use insufficiently assessed. The oil method may measure aerosol only, the dew point may be reported at an unspecified pressure, and the sample may have been collected before the final hose. None of these gaps can be repaired by placing “GMP compliant” on the certificate.
The task is to connect a defined process risk to a representative sample, a suitable analytical method and an approved acceptance decision. This article covers compressed air and relevant process gases at source, distribution and point of use. It complements the Critical Utilities Systems area; aseptic process interfaces also belong with Aseptic Fill-Finish & Barrier Systems.
Start with the decision, not the test catalogue
[GUIDEGXP RECOMMENDATION] For every test, identify the question it will answer. Are you accepting a new source, qualifying a branch, checking a final assembly, monitoring a known system or investigating an excursion? The same analyser can produce valid measurements that are irrelevant to the intended decision if the sample boundary is wrong.
Record gas identity, intended use, contact pathway, quality attributes, acceptance criteria, location and operating conditions in one sampling specification. Distinguish direct product or primary-container contact, indirect exposure and uses with no credible product pathway. Instrument air is a functional description, not proof that a failure cannot affect a process or create contamination through a connected device.
Separate the supplier certificate from the site result. A certificate may characterise a supplied batch or production lot under the supplier's sampling arrangement. It does not establish the condition of local storage, regulators, hoses, filters or newly installed branches. Define receipt controls and point-of-use verification as related but distinct activities.
What the references do and do not establish
[REGULATORY REQUIREMENT / EU GMP] For sterile manufacture in scope, Annex 1, 2022, sections 6.18–6.20 addresses product-contact gas quality, aseptic gas filtration and backflow. Apply those provisions to their stated context, not automatically to every pneumatic actuator.
[TECHNICAL STANDARD] ISO 8573-1:2010 supplies a compressed-air classification framework. Test methods and classification do not select a universal pharmaceutical acceptance class. [QRM] ICH Q9(R1), EMA Corr.2, 23 January 2025 supports the site rationale; risk assessment must consider the actual contamination mechanism and uncertainty.
[COMPENDIAL REQUIREMENT] Apply a pharmacopoeial gas monograph when the gas and its regulated use genuinely require it, using the applicable current licensed text. A medicinal gas requirement cannot simply be assigned to every process-gas application, and a supplier grade cannot replace the site's user requirements. The practical sampling choices below are original GEP recommendations.
Choose locations that represent distinct questions
Source sampling evaluates generation or incoming supply. Distribution sampling evaluates what happens during storage and transport within the site. Point-of-use sampling evaluates the approved final assembly. Compare these boundaries when diagnosing deterioration; do not use their names interchangeably. A sample downstream of a local filter cannot establish the condition upstream of it.
Select locations by risk and operating knowledge: product contact, long or newly modified branches, intermittent users, local pressure reduction, final filtration and demanding consumption profiles. One distant point rarely represents all failure modes. A nearby hose exposed during changeover may be more relevant microbiologically than the longest continuously flowing line.
Define whether the sample includes the production hose and final nozzle. A dedicated sample valve is convenient, but its history, geometry and flow may differ from the process connection. Where surrogate locations are used, establish what they represent and periodically revisit that justification after changes. Mark locations unambiguously on current drawings and sample labels.
Control the sampling assembly as measurement equipment
The sample can gain or lose contamination before reaching the detector. Tubing may adsorb vapours, release particles or admit ambient moisture. Regulators may contain lubricants or create condensation during expansion. A reducer intended for pressure control is not automatically suitable for particle, oil or microbial sampling. Assess materials, cleanliness, compatibility and flow arrangement for each method.
Determine conditioning and stabilisation from the method and assembly characteristics. Avoid inventing one purge time for every point. Record how stable measurement conditions are recognised and preserve evidence when startup itself is the condition under investigation. A prolonged purge may remove the very event that the sampling programme was intended to detect.
Specify actual sampling pressure, flow and collected volume, together with reference conditions used for reporting. Account for dilution or pressure reduction explicitly. Leak checks, blanks and handling controls should be appropriate to the method. Include safe gas discharge and compatibility, especially for oxygen, inert gases in occupied spaces and carbon dioxide expansion; involve competent site safety personnel.
Particles: measurement is not identification
ISO 8573-4:2019 addresses particle measurement. An optical counter does not establish whether particles are viable organisms, lubricant droplets or solid debris. Interpret its result against the method's measurement capability and the site's defined size ranges, reporting basis and acceptance criteria.
Review pressure reduction, transport losses, tubing length, detector suitability and possible condensation. A high reading produced by droplets after expansion calls for investigation of the sampling conditions, not automatic dismissal as harmless. Conversely, particle deposition in a long sample line may conceal contamination. Compare like configurations when trending or changing contractors.
Moisture and pressure dew point
ISO 8573-3:1999 addresses humidity measurement in compressed air. Pressure dew point describes the condensation threshold at a stated pressure; it is not a pressure-independent synonym for all forms of water. Free liquid, water vapour and moisture entering through an exposed hose require different diagnostic thinking.
Record the pressure at the sensor and whether the instrument converts to another pressure basis. Do not compare atmospheric dew point directly with an acceptance limit defined as pressure dew point. Check sensor range, contamination history, response time and sample-line equilibration. A stable display immediately after connecting a wet sampling hose may reflect the sensor's earlier condition rather than the utility.
Consider transient events. A dryer can maintain an acceptable average while a switchover briefly releases moisture. A laboratory snapshot may miss the event; an online sensor may capture it but sit upstream of a vulnerable branch. Use the two evidence streams according to their strengths instead of assuming one automatically replaces the other.
Oil aerosol, oil vapour and total oil
The current public references verified for this article are ISO 8573-2:2018 for oil aerosol and liquid oil methodology and ISO 8573-5:2025 for oil vapour. The older 2007 and 2001 editions respectively are withdrawn. Method versions matter when comparing contracts and historic results.
Define which oil fractions the specification addresses and which the selected methods actually recover. An aerosol result alone cannot be relabelled total oil. Where a combined assessment is required, align sample conditions, units, reporting limits and treatment of values below quantification. Do not add incompatible numbers or assume that “not detected” means zero contamination.
An oil-free compressor describes part of the generation technology; it does not exclude intake hydrocarbons, contaminated distribution or service-related residues. Investigate the whole pathway. Adsorbent media and coalescing elements perform different functions, and differential pressure alone does not prove removal of vapour. Base testing and maintenance decisions on the specific mechanism being controlled.
Microbiological sampling and recovery
ISO 8573-7:2003 concerns viable microbiological sampling. A particle count is not a microbial count, and a culture method does not measure endotoxin. Identify separately any endotoxin concern arising from intended use, rather than treating “microbiology passed” as evidence for every biological hazard.
Assess the effects of gas expansion, impaction, desiccation, collection medium and incubation on recovery. The method must suit the gas and sample conditions. Inert gas does not mean sterile gas. Nitrogen, carbon dioxide and oxygen can behave differently in a sampling device or recovery method, so compressed-air suitability should not be assumed to transfer unchanged.
Control aseptic handling, sample-port preparation, sterile assemblies where required and transport to incubation. Include relevant blanks and traceability of media and equipment. When an excursion occurs, review organisms, location, history and handling evidence with microbiology expertise. A negative repeat is useful information, but it does not erase the original finding or establish its cause.
Gas identity and purity need targeted methods
A nitrogen purity measurement based on residual oxygen does not by itself quantify every potential impurity. Define what the analyser measures directly, what is calculated and which contaminants remain outside its scope. Supplier purity, water, particles, oil and microbial quality are separate attributes unless a justified specification explicitly relates them.
Review method suitability for the actual gas matrix, interferences, calibration gases and measurement range. Carbon dioxide, oxygen and argon applications may need different analytes and materials. For generated gases, include startup and changing demand when relevant. For delivered gases, retain supply traceability and evaluate the transition between deliveries or cylinder banks.
Original sampling decision matrix
| Question | Useful evidence | Decision limitation |
|---|---|---|
| Did the dryer lose control? | Pressure-based dew-point trend aligned with load and switching events | Does not establish microbial quality or every downstream hose condition |
| Is oil contamination controlled? | Methods covering the specified aerosol, liquid and vapour fractions | A single fraction cannot support an unqualified total-oil claim |
| Does the final connection add particles? | Controlled comparison before and through the operating assembly | Different sampling reducers can confound the comparison |
| Is a microbial finding local? | Location-based investigation, organism information and handling controls | Negative repeats alone do not identify the cause |
| Can online data replace a laboratory test? | Demonstrated equivalence for the specific attribute and decision | Coverage, detection capability and sampling boundary may differ |
Frequency, trending and report review
Qualification sampling establishes capability under justified conditions; routine sampling checks continued control. Set frequency from contact risk, variability, detectability, failure history and available online evidence. Increase attention after relevant maintenance, source changes or excursions. Reduce frequency only through an approved reassessment with adequate supporting data, not because recent certificates look repetitive.
Trend results with operating state, source, location and method version. Keep below-quantification results distinguishable from numerical zero. Check unit conversions and whether a contractor changed equipment or reference conditions. A graph is not comparable evidence if its values describe different fractions, pressure bases or assemblies.
Review reports beyond the pass/fail box. Confirm sample identity, collection conditions, method, calibration, quantification capability, laboratory responsibility and deviations. Where the method cannot distinguish acceptance from failure at the required limit, improve the method or decision strategy before interpreting the result as compliant.
Practical case: an apparent oil-free air failure
In a hypothetical installation, a new product-contact branch shows an oil result above its approved criterion while the source remains acceptable. The site uses an oil-free compressor. Investigation therefore includes the recently installed regulator, assembly lubricants, sampling tubing and the test's oil fraction instead of excluding contamination on the basis of compressor type.
Controlled paired samples and handling blanks identify whether the anomaly belongs to the operating assembly or the sampling train. The team preserves the first result, defines the potentially affected production window and separates immediate containment from final cause confirmation. Corrective work is followed by targeted verification and a review of installation and contractor instructions.
Checklist before accepting the sampling package
- The sample answers a defined source, network or point-of-use question.
- The actual assembly, operating state and pressure basis are recorded.
- Methods cover the specified attributes and oil fractions with suitable sensitivity.
- Particle, microbial and endotoxin evidence remain distinct.
- Gas-matrix effects, recovery and sample handling are assessed.
- Online and laboratory evidence have explicit, justified roles.
- Excursions retain original data, impact assessment and a documented follow-up decision.
A defensible report allows another competent person to reconstruct what was measured, where, under which conditions and why it supports the process decision. That is the useful standard of evidence for pharmaceutical gas quality, far beyond a generic certificate heading.
