From Visibility to Design: What Sweden Reveals About Building ATMP Systems

Most discussions around Advanced Therapy Medicinal Products (ATMP) still frame the challenge in familiar terms: better science, more efficient manufacturing, a larger and better trained workforce. None of these are irrelevant.

But they are no longer the primary constraint. We are increasingly able to build these therapies. What remains unresolved is how to deliver them reliably at scale.

That distinction is easy to miss in large fragmented markets, in which the delivery chain is too dispersed to observe clearly. Failures show up as local issues, a manufacturing delay, a site level deviation, or a logistics problem, rather than as symptoms of a system that does not operate as a coherent whole.

Sweden offers a different vantage point. Not because it has solved ATMP delivery—it has not. But because its ecosystem is structured in a way that makes the full chain visible, from patient to manufacturing to quality to logistics and back. When you can see the system end to end, patterns emerge that are harder to detect elsewhere.

Sweden’s ecosystem has developed in a way that makes coordination across the chain observable in practice. Facilities such as Karolinska’s Vecura, which produced the first CAR-T cells to enter a clinical trial in Europe in 2014, reflect decades of integration between research, manufacturing, and clinical care within the same institutional environment.

At the system level, university hospitals coordinate ATMP trials through a shared, publicly available national trial list. Treatment adoption is assessed through NT-rådet in a transparent process that has led to decisions that have been more difficult in other comparable systems.

These are not indicators of the largest or best-funded ecosystem, but of one that has built enough institutional alignment to make the behavior of the system visible. This piece uses Sweden in that way, not as a model to replicate, but as a diagnostic environment, a setting in which the underlying constraints of ATMP delivery become visible enough to examine directly.

What emerges from that view is that the core challenges are no longer in building the components, but in how the system around them is put together and made to function. The field is now limited less by what it can develop than by what it can reliably deliver.

System Visibility and What It Reveals

In most large markets, patient-specific manufacturing is too fragmented to understand as a coherent system. Clinical sites, manufacturing units, quality functions, and logistics providers operate across different institutions, geographies, and data environments.

Each part can be optimized locally, but the system as a whole remains difficult to see. The consequence is that failures are interpreted in isolation, a delay at a site, a deviation in production, a breakdown in transport. The root cause is often attributed to the most visible step, even when the issue originates elsewhere in the chain.

Sweden’s ecosystem is configured differently. Its life science infrastructure is relatively dense, with close links between academic hospitals, manufacturing, and clinical trial networks.

The same actors, or closely connected ones, often operate across multiple steps in the process. This proximity shortens feedback loops and makes failures visible as system issues rather than isolated events.

What emerges is not just operational friction, but structural constraints, features of the system that persist regardless of how well individual components perform. That is Sweden’s specific contribution, in that its value lies less in scale or capital, and more in the visibility it provides into how the system behaves in practice.

Three Structural Constraints

When the full chain is observable, three constraints appear consistently, not as isolated execution problems but as outcomes of how the system itself is designed.

1. Reliability is lost at handoffs, not within unit operations

Much of the technical focus in ATMP has been on unit operations, such as cell processing steps, expansion protocols, vector delivery, and release testing. These are critical and increasingly well controlled, but they are not where reliability typically breaks down.

In practice, failures occur at the interfaces between these steps, across the transitions from clinic to manufacturing, manufacturing to quality control, quality to logistics, and even between teams or shifts. Each handoff introduces the potential for misalignment in documentation, labelling, timing, and interpretation.

In a system where the effective batch size is one patient and where chain of identity and custody must be preserved without deviation, these interfaces become the primary points of fragility. This becomes clearer when the full delivery chain is viewed as a sequence of interdependent handoffs, in which each transition introduces a potential point of failure (Figure 1).

The implication is straightforward even if uncomfortable: improving individual steps is necessary, but reliability ultimately depends on how the interfaces between them are designed.

2. Variability is operational and institutional, not just biological

ATMP is often described as a field defined by biological variability. Starting material differs between patients and cellular behavior is inherently complex—this is true and unavoidable.

But biological variability is only one layer. Operational variability is equally significant, differences in how procedures are interpreted across teams, how documentation is completed, how deviations are assessed and escalated.

Institutional variability adds another dimension. Hospitals, manufacturing units, and quality systems operate under different constraints, cultures, and levels of experience. Even when following the same protocols, execution diverges in ways that matter.

In fragmented systems, these layers are difficult to disentangle. Outcomes are often attributed to biology because it is the most visible factor and the rest remains implicit.

In a more observable environment, the picture becomes clearer. Reliability depends on managing variability across biological, operational, and institutional layers simultaneously, rather than treating it as something that can be reduced within isolated steps.

3. Ownership is fragmented where the process requires continuity

The ATMP chain behaves as a single, continuous process, in which each step is tightly linked and failure at any point compromises the entire outcome. Ownership, however, is distributed across multiple actors:

• Hospitals manage patient interaction and early-stage processing.
• CDMOs or internal manufacturing units handle production.
• Quality functions carry regulatory accountability.
• Logistics providers manage transport under strict conditions.
• Suppliers enable storage, monitoring, and traceability.

Each of these actors operates under its own incentives, budgets, and accountability structures. No single entity owns the reliability of the system end to end.

This is not accidental; it reflects how the ecosystem has evolved. But it creates a structural tension that cannot be resolved through better execution alone. The process requires continuity, yet ownership of that continuity is distributed.

In Sweden, where relationships between actors are closer and more transparent, this tension becomes visible. In larger markets, it often remains hidden until it manifests as failure.

What Other Systems Reveal

Every ATMP system resolves one constraint while leaving others exposed. This is not a function of execution, but of design.

Each configuration addresses a specific bottleneck in the chain, while shifting pressure elsewhere. This becomes visible when comparing how different systems are structured, each addressing a specific constraint while shifting pressure elsewhere (Figure 2).

Spain offers a clear example of a system structured around co-location. At Hospital Clínic de Barcelona, ARI-0001, an academic CAR-T cell therapy, is manufactured on-site under the hospital exemption pathway.

By placing manufacturing and clinical care within the same institution, the handoff problem is largely resolved, but that same structure limits scale. The therapy remains tied to the institution where it was developed, and fragmentation reappears at the system level.

Spain’s broader access landscape, operating across 17 autonomous regions without a unified national reimbursement mechanism, makes this visible. The United States represents the opposite configuration, a system structured around centralized, industrial-scale manufacturing.

This enables standardization and production at scale, addressing the constraints of capacity and process control. But it extends the delivery chain.

Coordination becomes the limiting factor and accountability diffuses across multiple actors. The result is a persistent gap between what is clinically possible and what institutions can reliably deliver.

The United Kingdom takes a more deliberate approach, structuring the system around dedicated infrastructure. Through the Cell and Gene Therapy Catapult, it has built an institutional layer designed to connect clinical development with commercial manufacturing, addressing fragmentation through coordination.

This begins to resolve the interface problem at system level, but it does so by absorbing it. The infrastructure manages complexity within the system, making it more operationally reliable while also making underlying points of failure less visible.

Sweden sits in a different position. Its system remains structured around hospital-based manufacturing and a relatively dense, closely connected ecosystem.

This reduces distance between actors, but does not yet introduce the same level of coordinating infrastructure as the UK. As a result, the underlying constraints remain exposed.

Handoffs, variability, and fragmentation are not fully managed within the system, and therefore surface more directly as system-level issues. Initiatives such as CCRM Nordic reflect a system that has identified these constraints and is beginning to respond to them.

Where the UK manages fragmentation, Sweden still exposes it, providing a clearer view of what determines whether the system can scale. As ATMP moves toward wider adoption, that distinction becomes critical.

Systems that scale without understanding their points of failure risk embedding them.

From Observation to Design

Seeing these constraints is only useful if it changes what gets built. If reliability is lost at interfaces, then interfaces need to be treated as core design elements, not administrative steps, but rather engineered points of control. Standardized handoff protocols, systems that reduce ambiguity, tools that make deviations visible in real time.

If variability operates across biological, operational, and institutional layers, then control strategies need to reflect that. Not just tighter protocols, but feedback systems that capture how processes actually behave in practice.

If ownership is fragmented, then coordination mechanisms need to function as infrastructure, including shared platforms for traceability and documentation, models that align incentives across actors, and structures that reflect how the system actually operates.

This is not about incremental optimization. It is about recognizing that the field is building a system, not just a set of technologies.

Where Capital Goes, and Why

Capital in ATMP has largely flowed into therapies, platform technologies, and manufacturing capacity. This is not accidental.

These areas have clear ownership, defined boundaries, and a direct link between investment and value capture, which makes them fundable within conventional models. The constraint, however, is increasingly elsewhere.

The capabilities that determine whether therapies can be delivered reliably—such as coordination across institutions, structured knowledge transfer, workforce capability, and chain of identity and custody—sit across organizational boundaries. They lack a single owner and the value they create is distributed, which places them outside conventional investment categories and often frames them as overhead rather than infrastructure.

This leads to a predictable outcome. Investment continues to optimize individual components, while the limiting factor sits in how those components connect and operate together, with progress in science and process outpacing the system’s ability to deliver reliably.

The more relevant question is therefore not how much capital is available, but where in the ecosystem it can flow, and why. Capital follows ownership and value capture, which means the most fundable parts of the system are not necessarily the ones that determine whether it works.

This becomes clearer when mapping investment categories against fundability and system criticality (Figure 3). Therapeutic assets and manufacturing capacity continue to attract private capital.

By contrast, coordination infrastructure, workforce capability and knowledge transfer, and chain of identity and custody sit in the opposite corner, critical for delivery but difficult to fund because they span institutional boundaries and lack clear ownership. As these constraints become more visible, capital allocation begins to shape not just capacity, but how the system itself is organized.

Recent EU funding decisions in 2023-2024 to establish a limited number of ATMP centers of excellence illustrate this dynamic. By concentrating resources into specific institutional configurations, these initiatives embed coordination within defined structures rather than leaving it to emerge across independent actors, influencing how the system is organized as it scales.

The capabilities required to make these systems work span multiple actors, yet the value they create is not captured by any single one, placing this layer outside typical investment incentives. Public funding begins to address parts of this gap but does so by embedding coordination within specific structures.

More broadly, resolving it requires models that operate across institutional boundaries, combining public funding, industrial participation, and shared infrastructure. Without this, the system will continue to optimize what is fundable rather than what is necessary.

Conclusion

What Sweden offers is not a template, but a way of making the system visible enough to understand where it breaks. The field has spent a decade learning how to develop these therapies and will continue to improve the science.

The constraint now sits in whether delivery systems evolve with the same level of intent.

About the Author

Bita Sehat is a Partner at Trill Impact advising on life science investments within Ventures. She writes on capital allocation, health care infrastructure, and the system dynamics that determine whether scientific progress translates into clinical impact.

Sources and data notes

Global ATMP clinical trial and developer figures: HiTech Health / Alliance for Regenerative Medicine (ARM), 2024. Global investment figures 2020–2024: ARM, as reported by BioSpace (October 2024). US vs Europe funding split Q1–Q3 2024: labiotech.eu citing ARM sector report. Sweden clinical trial activity and CAR-T infusion volumes: ATMP2030 Annual Report 2024, ATMP Sweden. Swedish SME EU funding: ATMP Sweden (November 2024). Vein-to-vein timelines and manufacturing failure rates: Journal of Experimental Medicine (February 2024). Multiple myeloma waitlist mortality: Journal of Experimental Medicine (February 2024), citing Al Hadidi et al. (2023). Spain ARI-0001: Bone Marrow Transplantation (January 2022); SOHO Insider (May 2025). CCRM Nordic: AFRY press release (April 2025); CCRM Nordic website; Vinnova announcement (May 2023). UK CGT Catapult: GMP Manufacturing Report 2024 (November 2024); BIA annual review 2024.