As cell and gene therapies (CGTs) continue their transition from experimental interventions to established treatment modalities, one of the most persistent strategic debates in the industry concerns manufacturing model selection. Should CGT manufacturing be centralised in large, industrial facilities, or decentralised closer to the patient in hospital-based or regional sites?
The debate is often framed as a stark contrast. Centralised manufacturing is associated with scale, efficiency, and commercial discipline, while decentralised manufacturing is positioned as agile, patient-centric, and clinically integrated. Yet when examined through the lenses of Quality by Design (QbD), LEAN manufacturing principles, and ICH-aligned quality systems, a different conclusion emerges.
Centralised and decentralised approaches to CGT manufacturing are not really that different.
They are governed by the same manufacturing science, constrained by the same physical realities, and ultimately succeed - or fail - based on the same principles.
The common constraint: Throughput in limited space
Regardless of geography, all CGT manufacturing must contend with a fundamental challenge: delivering sufficient throughput within constrained cleanroom space, labour availability, and capital investment.
For centralised manufacturing, this requirement is self-evident. A single facility may be responsible for producing drug product for entire countries or regions. Under these conditions, low throughput density quickly translates into unsustainable cost of goods, long lead times, and supply risk.
What is less obvious, but equally true, is that decentralised manufacturing faces the same constraint, often in more compressed form. Hospital-based and regional CGT facilities typically operate within much smaller footprints, frequently embedded within existing academic or clinical infrastructure. Cleanroom space is scarce, expansion is costly, and staffing models are tightly constrained. Yet many of these facilities have demonstrated clinical success, treating patients locally and reliably. As confidence grows, demand follows. Decentralisation does not eliminate the need for throughput - it concentrates it.
A QbD perspective: Design space applies everywhere
Quality by Design, as articulated across the ICH Q-series guidelines, begins with a simple but demanding premise: quality must be built into the process through a deep understanding of how inputs, process parameters, and unit operations influence critical quality attributes.
From a QbD standpoint, throughput is not merely an operational metric - it is a process parameter that directly affects quality risk. As utilisation approaches capacity limits, variability increases, deviations become more frequent, and control strategies are stressed
This applies equally to centralised and decentralised facilities. The difference lies only in where the constraint manifests. In centralised sites, it appears as scale pressure. In decentralised sites, it appears as space pressure. In both cases, the outcome is the same: a narrowing of the effective design space.
When demand grows faster than manufacturing flexibility, quality risk accumulates regardless of facility size.
Clinical expansion exposes manufacturing fragility
Early decentralised CGT programmes were often justified on the basis of modest patient volumes and narrow indications, most commonly CD19-positive haematologic malignancies. Under these conditions, relatively low throughput appeared sufficient.
That assumption no longer holds.
Local CAR-T programmes are expanding into BCMA for multiple myeloma and are expected to move rapidly into additional targets and disease settings. Each new indication increases manufacturing demand without proportionally increasing available space or staffing.
From a QbD perspective, this represents an unmanaged shift in operating conditions. From a LEAN perspective, it represents overburden – muri - a well-known precursor to variability, waste, and failure.
Clinical success, while desirable, is often the first stress test of manufacturing architecture.
The throughput limitation of one-product-at-a-time systems
Most decentralised CGT manufacturing today relies on highly integrated, all-in-one systems designed to take an apheresis input and deliver a finished drug product within a single enclosed unit. These systems offer clear advantages in early deployment: simplified training, reduced open handling, and procedural consistency.
However, they also introduce a structural limitation that becomes increasingly problematic at scale: each unit can manufacture only one drug product at a time.
Consider a hospital treating 300–400 patients annually for large B-cell lymphoma or multiple myeloma. To accommodate enrollment variability, multi-day expansion timelines, scheduling conflicts, and surge demand, such a programme may need to support three to five concurrent manufacturing runs.
When downtime, redundancy, and maintenance are factored in, achieving this capacity requires a surprisingly large number of individual units.
Over time, the facility accumulates equipment, expands cleanroom space, and adds operational complexity, often approaching the scale of a centralised manufacturer.
At that point, the distinction between centralised and decentralised becomes largely semantic.
Most decentralised CGT manufacturing today relies on highly integrated, all-in-one systems
LEAN insights: Complexity is a form of waste
LEAN manufacturing principles emphasise flow, simplicity, and continuous improvement. From this perspective, scaling throughput by multiplying large, capital-intensive systems introduces several forms of waste:
- Underutilised equipment during non-expansion phases
- Increased motion, maintenance, and coordination overhead
- Longer deviation investigations across parallel units
- Reduced visibility into true process drivers
LEAN would characterise this as system-level inefficiency masked by local optimisation.
The process may appear efficient at the unit level while becoming fragile at the system level.
Importantly, these dynamics are not unique to decentralised facilities. Centralised manufacturers have encountered the same issues when scaling through monolithic equipment rather than modular capacity.
A converging conclusion across both models
When QbD and LEAN principles are applied rigorously, a common conclusion emerges: sustainable CGT manufacturing depends on throughput per square foot, not throughput per machine.
This insight is increasingly shaping manufacturing strategy across the industry. Rather than relying exclusively on all-in-one architectures, both centralised and decentralised programmes are moving toward modular approaches that decouple critical unit operations.
In practice, this often means separating closed, automated upstream processing from high-density expansion capacity, allowing throughput to scale linearly without proportional increases in capital equipment or cleanroom size.
From a QbD standpoint, this expands the design space. From a LEAN standpoint, it restores flow.
Alignment with ICH expectations
ICH Q8, Q9, and Q10 collectively emphasise process understanding, risk-based decision-making, and continual improvement. Modular manufacturing architectures align naturally with these expectations by:
- Improving process visibility and control
- Reducing the scope of re-validation when changes occur
- Simplifying root-cause analysis during deviations
- Enabling incremental, data-driven improvement
Crucially, these benefits apply regardless of whether manufacturing is centralised or decentralised. The regulatory framework does not favour geography; it favours robust, well-understood processes.
Moving beyond a false dichotomy
The question of centralised versus decentralised CGT manufacturing is often treated as a strategic fork in the road. In reality, it is a false dichotomy. Both models operate under the same constraints, respond to the same quality risks, and benefit from the same manufacturing principles.
Facilities that succeed will be those that recognise early that quality, throughput, and space are inseparable, and design their processes accordingly. Those that do not will encounter the same scaling challenges, regardless of where they are located.
So are centralised and decentralised approaches to CGT manufacturing really that different?
Not really.
The same manufacturing science applies. The same constraints exist. And increasingly, the same solutions are emerging.