Filling Technologies and the Sterile Product Path: How to Design an Aseptic Filling Process

Same line, same isolator that just passed requalification, same operators. Only the product changed: a protein formulated with a surfactant instead of the aqueous solution filled for eight years. At the first PQ the mean weight is on target, but dispersion doubles, the first vials of every start-up are underfilled, and a residual droplet appears on the neck. Nobody set a wrong parameter: a dosing system selected for a different product is operating outside the domain in which it was qualified.

This article is about the decision upstream of that failure: which dosing principle and which sterile product path suit a given product. It is process design, not procurement: taken when the URS is written, frozen with the qualification strategy, and from there it shapes the rest of the lifecycle — cleaning validation or single-use, interventions in Grade A, the representativeness of the APS. The rule that governs everything is broken constantly: filling technology is selected from the characteristics of the product and the path the product must travel, never the other way round.

Where the problem starts: a choice made downstream

In most projects filling technology arrives as an attribute of a machine already selected for other reasons — format, throughput, integration with the barrier — and the product path follows as supplied. The consequence shows up three steps later. At tender stage, bids based on different dosing principles are not comparable, because nobody stated which product characteristics the line must cover. In qualification, the OQ demonstrates accuracy on water or on a placebo that does not reproduce the real rheology. In operation the critical zone pays: every out-of-specification unit triggers an intervention, and every intervention is a risk to be justified in the CCS and simulated in the APS.

The regulatory picture

Dosing principles and their real trade-offs

No dosing technology is better than the others: there are families with different profiles, and the choice consists in matching a profile to the product. The comparison variables are always the same: accuracy and repeatability across the full range, shear, particulate, hold-up volume, priming, cleanability, sterilisability, compatibility with a single-use path.

Peristaltic

No contact between mechanism and product: the fluid stays inside the tubing. It is the natural candidate for a single-use path, for biologics that should not see metal, and where changeovers are frequent. The price is dependence on the tubing: precision derives from tube deformation and shifts with lot, material and temperature. The tube generates particulate through abrasion, requires a defined life limit, and repeatability must be demonstrated at the end of that declared life.

Time-pressure

Volume is set by tank pressure and valve opening time: no moving part in contact, low shear, a path simple to sterilise in place. It works well on low-viscosity fluids with stable properties, but is open loop on volume, sensitive to temperature, head and pneumatic integrity. Foaming products tolerate pressurisation poorly, viscous products cap the line rate.

Piston

Pure volumetric displacement: stable precision, insensitive to viscosity within the mechanical domain of the pump, suitable for wide ranges. The downside is the sliding fit in contact with the product, a potential particulate source, on a path harder to clean and sterilise. On fragile molecules, entry into the chamber is where the protein suffers.

Rolling diaphragm and hybrid solutions

Rolling diaphragm pumps aim at the compromise: volumetric accuracy without a sliding contact fit, hence lower particulate on sensitive products, with single-use versions available. In exchange the component costs more and has a finite service life, to be defined and managed under change control. Hybrid solutions raise the same questions, not the same answers.

The sterile product path, from vessel to needle

The pump is only one stretch of the path: tubing, manifold, valves, needles and connections weigh as much as it does.

Needles, splashing and dripping

Internal diameter, tip geometry and height above the neck decide whether the liquid falls as a continuous stream or breaks up. Too fast a jet splashes the wall and contaminates the sealing area; too slow a jet favours the terminal droplet. Dripping wets the neck, alters the sealing surface and generates rejects that become interventions; needle raising, suck-back and controlled deceleration are proven on the real product. Needles are also obstructions in unidirectional airflow: their effect on first air in the critical zone is a matter for airflow visualisation.

Connections: intrinsic or made in Grade A

Annex 1 8.14 is explicit: “Aseptic connections should be performed in grade A with a grade B background unless subsequently sterilised in place or conducted with intrinsic sterile connection devices” [REQUIREMENT]. There are three legitimate routes and they are decided at design stage. Reducing the number of connections made by hand under unidirectional airflow is the most effective lever: each one is an inherent intervention to be authorised, trained and simulated. Under 8.12, assembly of the sterilised set is itself aseptic processing [REQUIREMENT].

Single-use or reusable path

A stainless path is cleaned and sterilised in place with validated cycles: low variable cost, no batch limit, but cycle time, drainability, cleaning validation and requalification. A pre-sterilised single-use path removes cleaning validation and SIP and cuts critical connections, but shifts risk onto supply chain, set integrity on delivery, supplier qualification and extractables and leachables, assessed for the actual formulation and contact conditions. Annex 1 8.139 calls for verification of critical operations of single-use systems [REQUIREMENT].

IPC, calibration and hold-up

The in-process weight check is the primary sensor of the dosing system, and its calibration is a data integrity requirement as much as an accuracy one. Frequencies, tolerances and action limits are not transferable between plants: they are set through process development, measured capability and QRM. Hold-up volume must be measured on the real set: on small batches it can be the dominant selection criterion.

Selection matrix

Sterile filtration is not sterilisation, PUPSIT is not a post-use test

Sterile filtration is the physical removal of microorganisms from a fluid that cannot be sterilised in the final container: a separation whose outcome depends on filter, fluid, upstream bioburden and compatibility. Sterilisation is inactivation of microorganisms by a process with a demonstrated reduction capability on the treated item, such as SIP of the product path. The filter renders the fluid sterile; it is the sterilisation cycle that renders the downstream hardware sterile.

PUPSIT and the post-use integrity test answer different questions. Annex 1 8.87 concerns verification of the sterilised filter's integrity before use, to detect damage caused by its own preparation and sterilisation [REQUIREMENT]; the same clause accepts that the test may not be feasible because of process constraints, with an alternative approach supported by risk assessment [QRM]. The post-use test instead confirms downstream that the filter was integral during filtration: a CCS citing only one of the two leaves a whole failure mode uncovered. PUPSIT also weighs on the plant, because it requires wetting and venting downstream of the filter.

QRM: how the choice is justified

Selecting the dosing technology is a textbook ICH Q9(R1) object. The risk assessment starts from product quality attributes and the failure modes of the path: underfill, particulate, shear degradation, contamination at a connection, loss of set integrity, dripping. Detectability carries the most design leverage: a 100% weight check changes the underfill risk profile, not the contamination risk.

The outcome feeds the CCS, which under Annex 1 2.3 must define all critical control points and assess the effectiveness of the controls [REQUIREMENT]. The product path must be described in full: where the product is exposed, how many connections exist and of what type, which interventions are covered by the APS. Annex 1 9.33 requires the APS to closely imitate routine production [REQUIREMENT]: a viscous product simulated with liquid medium is a simplification that must be justified.

Worked example: Tucana Site

"Tucana Site" is a realistic but entirely fictitious example. It fills an aqueous solution in vials and is introducing a monoclonal antibody with polysorbate in nested prefilled syringes, on the existing time-pressure line with a stainless path and SIP.

The reasoning error looks prudent: we already have a qualified technology, we extend it to the new product. The assessment is framed around the machine's ability to reach the nominal volume, not around the product. At FAT, with an aqueous placebo, everything works. At PQ three things appear together: foam in the pressurised vessel, weight drift as the head changes, and dripping on the syringe flange.

The correction is not a finer calibration. The team rebuilds the path upstream: it defines rheology and foaming sensitivity as inputs to process design, evaluates dosing principles against those characteristics, and selects a pre-sterilised single-use path with volumetric diaphragm dosing, cutting at the same time the connections made in Grade A by using intrinsic sterile connection devices under 8.14. It redesigns the needle and builds the weight-check strategy on measured capability. Cost per batch rises, interventions in the critical zone fall. The point is not that single-use is superior: it is that the choice now derives from the product.

Levels of prescriptiveness

Checklist

• State viscosity, foaming tendency, shear sensitivity and volume range in the URS.
• Define the operating domain of the dosing system, not only the nominal volume.
• Map the product path, classifying every connection: intrinsic, sterilised in place, or made in Grade A.
• Qualify accuracy and repeatability on a rheologically representative placebo.
• Demonstrate the absence of splashing and dripping at operating speed.
• Fix the service life of product-contact components and verify performance to its end.
• Distinguish sterile filtration from sterilisation, and PUPSIT from the post-use test, in the documents.
• Assess extractables and leachables for the actual formulation and contact conditions.
• Freeze the product path configuration before PQ and manage variants under change control.

Recurring mistakes and red flags

The first warning sign is a URS that specifies fill volume and says nothing about viscosity, foam and shear: the supplier will decide. The second is qualification run on water alone and presented as accuracy on the product: the number is real, the domain is wrong. The third is a product path documented only as a mechanical schematic, with no classification of aseptic connections: without that count no credible CCS is possible and the APS cannot be sized. The fourth is dripping treated as a visual inspection issue rather than a process design defect. The fifth is single-use justified purely by the saving on cleaning validation. The sixth is confusion between PUPSIT and the post-use test in procedures. The seventh is dosing technology chosen before knowing which formats the portfolio needs, when it also depends on line design and container type.

The thread running through these mistakes is the same: a technical decision taken by inertia and then defended with documentation. Whoever designs from the product does more work at the start and needs far fewer interventions later. If that is the reasoning you are after, The Pragmatic GMP is where we publish it consistently.

Key points

• No filling technology is universally better: peristaltic, time-pressure, piston and rolling diaphragm have different domains and different limits.
• The choice starts from viscosity, foam, shear and volume range; without those data it has not been made.
• Sterile filtration removes microorganisms from the fluid; sterilisation inactivates them on the treated item.
• PUPSIT prevents one failure mode, the post-use integrity test establishes another.
• Single-use and reusable paths shift risk rather than remove it: they are compared over the lifecycle.
• IPC frequencies, tolerances, hold-up and component service life are site data.

References

• EU GMP Annex 1 — Manufacture of Sterile Medicinal Products, C(2022) 5938 final
• FDA — Sterile Drug Products Produced by Aseptic Processing, cGMP (2004)
• EudraLex Volume 4 — EU GMP, including Annex 15 Qualification and Validation
• ICH Q9(R1) — Quality Risk Management
• 21 CFR Part 211 — Current Good Manufacturing Practice for Finished Pharmaceuticals
• ISO 13408-2:2018 — Aseptic processing of health care products: Sterilizing filtration
• PIC/S — GMP Guide PE 009 and related publications