Ask anyone who has run a manual sterile filtration line where the real risk sits, and they will point to the same places. The hose changes. The connections. The integrity test. Every one of those is a moment when a person is close to the sterile boundary, and where a procedure is the only thing standing between routine and contamination.
That instinct is now written into the rules. The revised EU GMP Annex 1 pushes manufacturers to take manual intervention out of aseptic processing wherever they can. Most of the attention has gone to cleanroom design, environmental monitoring and contamination control strategies. Sterile filtration has had less, which is odd, because it is full of exactly the manual steps the guidance wants gone.
There is a useful example to learn from, and it came before the guidance arrived.
The problem GSK chose to solve properly
Several years before the revised Annex 1, GSK was running a manual sterile filtration system at its Ulverston site in the UK. It transferred product into a Grade C cleanroom and it worked. But it leaned on manual hose changes, manual integrity testing and connections that exposed the sterile boundary while the process was live.
The easy route would have been to upgrade the equipment and move on. GSK did something harder. The team rebuilt the filtration architecture from the ground up, and set itself a brief that looks ambitious even now: keep the sterile boundary closed during integrity testing, take equipment out of the cleanroom, and remove as many manual steps as the process would allow.
What they built
The finished system used permanently hard-piped sterile filters, automated integrity testing, dual filtration, and automated cleaning and sterilisation. Most of the valves and controls were moved into a technical area outside the cleanroom, behind a hygienic barrier. That single decision cut equipment density in the classified zone and made maintenance far simpler.
In total the system ran 55 separate operating sequences, covering production, integrity testing, cleaning and sterilisation. After all of it, one manual task remained: changing the filter element. Everything else was automated.
It is worth sitting with that number. One manual intervention, on a system that previously depended on several. That is the gap between managing a risk and designing it out.
Why it reads like an Annex 1 case study
The project was not built to satisfy Annex 1, because Annex 1 in its current form did not yet exist. Yet the design lines up almost point for point with what the guidance now asks for. Less human contact. Fewer aseptic manipulations. Contamination control justified through engineering rather than through a thicker procedure file.
This is the part worth holding onto. Procedures depend on people doing the right thing every time, under pressure, on a night shift, at the end of a long run. Engineering does not. When you connect filters permanently, automate the integrity test and close off the open ends, you are not asking an operator to be careful. You have removed the chance to be careless.
The quiet lesson: how the work was run
The design matters, but so does how GSK got there. Rather than treating suppliers as separate vendors handing over parts, the team brought process specialists, automation engineers, valve manufacturers, design engineers and site staff together from the start.
That meant the awkward questions surfaced early, while they were still cheap to answer. Dead-leg minimisation. Line slopes. How the skid would actually integrate with the cleanroom. The design went through several iterations before anyone built anything, which is usually the sign of a project that will work rather than one that will need fixing later. The result was recognised internationally with the IChemE Pharma Award for its fully integrated sterile filtration unit.
What this means if you are modernising now
Most facilities upgrading for Annex 1 are arriving at the same conclusions GSK reached early: integrated filtration, automated integrity testing, fewer manual steps, and contamination control built in from the first sketch rather than bolted on at the end.
Every site is different, and no one is suggesting you copy a skid built for one product into a plant making something else. But the principle travels. The most reliable way to meet Annex 1 is not to write more about what your people must do. It is to design a process where there is less for them to do at the sterile boundary in the first place.
That is the shift worth making, and it is an engineering decision long before it is a compliance one.
