Introduction

Platform devices are often the preferred starting point for combination product development, and for good reason. A well-characterised platform offers a proven architecture, defined performance boundaries, and a clear path to market. However, an increasing number of therapies in development are pushing into territories where there is currently no existing platform to serve their needs.

Novel biologics are extending the boundaries of volumes and viscosities, cell and gene therapies are leading to ever-more tissue-targeted deliveries, suspensions and co-delivered (or sequentially delivered) therapies are among those demanding entirely new thinking about how, where, and whether a drug can be delivered. These formulations can all be gathered under the umbrella term “specialty delivery” and are united by a common challenge – where there is no existing platform device able to deliver the formulation, a delivery system becomes a core part of formulation development and testing. Where traditional developments are often able to rely on vial-and-syringe delivery to enable early clinical data to be gathered, these specialty delivery developments (especially those requiring tissue-targeted delivery) require more sophisticated devices to be available earlier in the therapy or formulation’s life cycle.

For these developments, feasibility work looks very different to what we expect when adopting a platform. Rather than asking “Can this platform device deliver our formulation?”, the question becomes “Is it possible to deliver this formulation at all?”, which could mean the difference between a few weeks of benchtop testing and months or even years of exploratory research. Answering this new feasibility question efficiently requires a structured, pragmatic approach that balances scientific rigour with the realities of early-stage development timelines and budgets.

This article explores how theoretical, synthetic and tissue models can be combined to de-risk novel drug delivery developments, and how to deploy them intelligently to turn an apparently impossible challenge into a feasible one. While the worked example we’ll draw on throughout – a novel device for drug delivery directly to the brain – sits at the more complex end of the spectrum, the principles apply equally to any development where an existing platform doesn’t fit, or where it isn’t yet clear whether a device is a good match for the therapy. The approach scales to the necessary level of complexity.

What Does Feasibility Really Mean?

Feasibility for specialty delivery means different things depending on where you sit on the development complexity spectrum. At one end, it might mean confirming that an existing device can deliver a formulation that sits just outside its stated performance envelope, e.g., pushing the viscosity boundary and testing whether a longer injection time is still acceptable to users.

At the other end, feasibility might mean starting from scratch with a therapy that has never been delivered before, to a target site that has never been accessed in this way, using a device that doesn’t yet exist. Here, the risks are profound, and the unknowns are numerous, spanning therapy efficacy, navigation and targeting (including positional accuracy), tissue tolerability, usability, and many more.

The first challenge is therefore to understand where your development sits on this spectrum, which can be done by identifying the key risks to be addressed. This starts by understanding what the needs of the formulation are as compared to existing formulations on the market or in clinic. If, for example, you are targeting subcutaneous delivery, chances are navigating the device to the right area is not going to be a huge risk. But if you’re targeting a specific structure of the brain or a nerve within the nasal cavity, suddenly how the formulation is delivered to the right location is an unknown factor that’s critical to the efficacy of the therapy. The same principles apply for other aspects – are you targeting a new user group? Or a new use environment? Each area where you’re looking to push a boundary is a potential risk to be investigated during feasibility work.

The Model Landscape: A Brief Orientation

Once the key risks have been identified, the next step is to select the right tools -often models- to investigate them. In the context of drug delivery feasibility, models broadly fall into three categories.

In silico models – computational and mathematical approaches – range from simple physics-based calculations through to complex simulations such as computational fluid dynamics (CFD) or finite element analysis (FEA). At their best, they are fast, flexible, and low cost to iterate, making them well suited to early-stage exploration. Their limitation is that they are only as good as the inputs you give them, and at feasibility stage those inputs are often hugely uncertain.

In vitro models – bench-based physical testing – provide empirical data that computational models simply cannot. Synthetic tissue analogues, flow rigs, and bench prototypes all fall into this category. They offer a tangible way to test device and formulation behaviour under controlled conditions and are generally more accessible and affordable than animal or human studies.

Ex vivo models – testing in excised biological tissue – offer real tissue behaviour without the complexity and cost of full animal studies. For early feasibility work involving novel delivery routes or target tissues, ex vivo models are particularly valuable: they can provide rapid, biologically relevant data to characterise tissue properties and validate computational predictions at a stage where in vivo work would be premature.

In practice, the most effective feasibility programmes don’t rely on a single model type. The real skill is in knowing which combination to use, and in what order – and that starts with asking the right questions.

Framing the Right Questions: A Worked Example

Having the right models available is only half the challenge. The bigger risk in early feasibility work is asking the wrong questions of them, reaching for high-fidelity simulation before the problem is properly understood, or pursuing a level of detail that isn’t yet warranted. A multiphysics model of tissue mechanics or a full physiologically-based pharmacokinetic (PBPK) simulation might ultimately be the right tools but deploying them before the fundamental questions have been framed correctly is an expensive way to generate false confidence.

A more effective approach starts not with the model, but with the decision that needs to be made. At each stage of feasibility, it is worth asking: what do we need to demonstrate right now? What is stopping us from moving forward? Once that decision is clear, the next step is to identify the key drivers – what are the dependencies, and do we have reliable data for them? Only then can a model, or set of models, be properly selected, and the guiding principle should always be to choose the simplest approach that can answer the question with the necessary degree of confidence.

To illustrate how this plays out in practice, consider the development of a novel device for delivering drug directly to a target structure within the brain, a scenario that was once thought to sit firmly at the impossible end of the feasibility spectrum. There are many questions that could be addressed during feasibility, but we’ve selected three unrelated questions that have been posed to us in the past as examples of our approach to feasibility:

1. What is the delivery force?
2. Will the device achieve the required formulation distribution in tissue?
3. How will the device influence the clinical effect?

1.  What is the delivery force?

The initial instinct may be to model the delivery force in full using a high-fidelity simulation incorporating device mechanics, friction, fluid dynamics, and tissue backpressure simultaneously. However, at this early stage, information about the conditions of delivery e.g. tissue properties, fluid interactions, variability (patient, user, device); may not be fully known. As such, reframing the question reveals a more useful starting point: the real need wasn’t to predict the exact force profile, but to understand whether the force required to deliver the drug would exceed what a typical user could reasonably apply. With that decision in mind, a more pragmatic model strategy emerges; ex vivo tissue characterisation to measure backpressure and understand fluid-tissue interactions, early bench testing with a simple prototype to assess injection force, and a low-order physics-based mathematical model to combine these inputs. Fast to generate, easy to iterate, and sufficient to answer the question that actually mattered at this stage.

2.   Will the device achieve the required formulation distribution in the tissue?

The initial expectation might be a complex computational fluid dynamics (CFD) and fluid-structure interaction (FSI) simulation with a nonlinear, anisotropic tissue model and multiphase flow, an approach that would be computationally expensive and heavily dependent on tissue property inputs that aren’t yet fully known. Reframing shifts the question to something more tractable: can the device achieve the required distribution area (or volume) in tissue? This opens up a staged model strategy; simplified flow and porous media models for rapid screening, followed by testing in brain tissue analogues developed from published academic literature, allowing findings to be benchmarked against existing research. Targeted CFD is then reserved for where refinement is genuinely needed. Lower cost, easier to verify, and designed to work even when input data is uncertain.

3.  How will the device influence the clinical effect?

Building a full PBPK model from preclinical data is an understandable ambition, but difficult at feasibility stage, where the biological inputs required are often unavailable or unreliable. Reframing the question from predicting clinical efficacy to “can the device achieve sufficient therapy exposure for target absorption?” enables a more practical, modular approach. Mechanistic models (which describe system behaviour) can be used to characterise device-to-delivery behaviour. Simplified transport models and targeted experiments address delivery-to-distribution as described above. And for distribution-to-exposure, findings can be bridged to existing pharmacokinetic and pharmacodynamic (PK/PD) models (the mathematical frameworks that describe how a drug moves through and acts on the body) making it straightforward to hand off to specialist teams when the time comes. This keeps the focus on the decision at hand and avoids reliance on uncertain biology.

In each case, the pattern is the same: resist the pull towards complexity, reframe the question to be asked around the decisions to be made, and choose the simplest model that can answer it. Models don’t just answer questions, used well, they help reveal the questions worth asking in the first place.

Conclusion

Even the most complex drug delivery challenges can be broken down into manageable pieces. The key is phase-appropriate pragmatism; understanding where your development sits on the feasibility spectrum, selecting models that are fit for the question rather than fit for the complexity, and knowing when you have enough information to make the next decision.

The worked example above illustrates how this plays out in practice. What was once considered an impossible development became feasible through the efficient combination of analytical, synthetic and tissue models, each chosen not for its sophistication, but for its ability to answer a specific, well-framed question at the right stage of the programme. It is an approach we have refined across many such projects, and one that we continue to apply wherever a development pushes beyond the boundaries of what existing platforms and precedent can answer.

To accelerate innovation, we need prompt decision making. Obtaining feasibility answers in weeks rather than years allows redirection of resource toward where it’s most needed (e.g. reframing of therapy or administration route), ensuring effective treatments reach patients faster. The framework works precisely because it is designed to find the answers that matter efficiently.

The tools are available. The challenge, and our experience, is in deploying them wisely.

Featured in ONdrugDelivery News, Jessica Alzamora, Dr Karla Sanchez and Emily Chang discuss the necessity for precision when delivering cell and gene therapies, explore how this precision can be designed and demonstrated, then go on to describe how a minimum viable product approach to device development can act as a strong predictor of a successful drug delivery device.

Cell and gene therapies (CGTs) are at the forefront of precision medicine, with the potential to repair or replace faulty genes and cells to treat disease at its biological source. Despite this promise, the success of CGTs depends on one defining factor: precision. Every stage, from designing a vector to delivering it in the body, demands careful control to ensure that the treatment reaches the targeted region and/or cells, at the right dose and with minimal off-target effects (Figure 1).

A clear example of this reliance on precision comes from a currently available gene therapy to help improve functional vision in patients with an inherited retinal disease due to a genetic mutation. The approved adeno-associated virus 2 (AAV2) gene therapy Luxturna^® (voretigene neparvovec, Spark Therapeutics, Philadelphia, PA, US) must be delivered via a highly targeted subretinal injection to ensure that the therapy reaches and acts on the exact layer of cells needed for vision. Even small variations in injection depth or placement can change how effectively it restores function, and incorrect placement can increase the risk of inflammation.¹ This shows that the success of a therapy depends as much on how it is delivered as what it delivers – the therapeutic effect is dependent on the accuracy of the delivery modality.

To unlock the full potential of CGTs, the industry must not only consider molecular innovation but also focus equally on the method of precision delivery to expand the pivotal link between discovery and patient benefit. Achieving reproducible precision will determine how effectively these breakthroughs translate from rare success stories into accessible, scalable therapies.

Where Precision Matters Most In CGTs

CGTs are not produced in the same way as small molecules or standard biologics. Many programmes are patient-specific or produced in small, labour-intensive batches, with customised biomanufacturing and strict cold chain to preserve vector integrity or cell viability. These constraints make products extremely costly: Luxturna®, for example, is priced at around US$850,000 (£650,000) per patient.² Given the resource-intensive nature of producing usable material, development teams must prioritise process efficiency and precision from the earliest stages of production.

Potency and safety are also tightly linked. Small deviations in target delivery or poor biodistribution control can provoke serious immune-mediated toxicities,³ among other serious side effects, which is particularly true in gene therapies.⁴ For instance, intrathecal delivery (administration into the cerebrospinal fluid, e.g. via lumbar injection, allowing direct access to the central nervous system) can have a biodistribution-associated risk that results in dorsal root ganglion inflammation and neuronal degeneration, particularly with higher doses, where neither the therapy’s tropism (affinity with specific cells) nor cerebrospinal fluid dynamics have been fully characterised.⁴

Some therapies may only succeed when they are placed with millimetre-scale accuracy. For a rare neurological disorder called aromatic L-amino acid decarboxylase deficiency, the AAV2-based therapy Upstaza™ (eladocagene exuparvovec, PTC Therapeutics, Warren, NJ, US), is delivered through stereotactic neurosurgery, which delivers four small infusions into the putamen in a single session (two per hemisphere).5 The product label specifies the route, infusion sites and dosing parameters, as the efficacy of the therapy depends on reaching the correct brain region while avoiding wider systemic exposure. This is precision delivery built directly into the treatment’s design. Furthermore, for one-off or single-administration gene therapies, re-delivery may not be possible (e.g. due to pre-existing antibodies to AAV) or may be considered too risky to conduct (e.g. direct-to-brain administration).

Precision in where and how therapies are delivered determines how safely it can be dosed, how consistently it can be scaled and how much product is needed to achieve a therapeutic effect.

When Precision Becomes A Moving Target

Precision is easy to define, in theory, but difficult to achieve in practice. For many CGTs, location, distribution and dose must be defined long before clinical trials begin, yet each is influenced by complex and patient-specific variables (Figure 2). Precision is less critical for ex vivo approaches, such as chimeric antigen receptor T-cell therapies, where cells are modified outside of the body prior to intravenous administration. These treatments have demonstrated success, as seen with Kymriah® (tisagenlecleucel, Novartis) and Yescarta® (axicabtagene ciloleucel, Kite Pharma, Santa Monica, CA, US) in haematological malignancies. In contrast, precision becomes far more consequential for in vivo gene and stem cell therapies. What seems simple – such as targeting a specific organ for a rare disease – quickly becomes challenging when teams must decide what level of precision is sufficient in terms of which part of the organ and its diverse cell populations to target for the therapy to be effective.

Location

This challenge is clearly visible in liver-directed AAV therapies, where defining location goes beyond reaching the organ itself. The liver’s intricate vasculature and cell diversity means that vector access and expression vary widely, while efficacy depends on transducing enough hepatocytes without excessive uptake by other cells that may trigger immune responses or reduce potency.⁶ Achieving this balance relies on optimising the route of administration, delivery site and dose flow control.

Distribution

Parameters such as vector concentration, infusion rate and device (e.g. cannula) geometry determine how the therapy is distributed through the tissue and how reliably it reaches target cells. To manage these interdependencies, computational and experimental modelling are integral throughout development of the therapy and delivery device. By modelling vector flow, convection and uptake in patient-specific anatomy, device developers can predict how a formulation or delivery approach will behave before starting animal studies, or they can refine it alongside these studies. These models enable the integrated team (composed of formulation/modality specialists, device developers and more) to optimise distribution patterns, reduce experimental uncertainty and accelerate iteration, allowing precise delivery to be engineered rather than inferred.

Dose

A clearer understanding of anatomical location and distribution also improves how the team defines and manages dose precision, which ultimately determines efficacy and safety. Dosing CGTs is about far more than volume; it reflects how much active vector or number/type of cells are needed to ensure the desired effect within the target tissue. Achieving precise dosages means controlling both potency and delivery conditions so that the administered quantity can translate into a safe and effective treatment. Advances in data analytics (e.g. vector analysis), flow-controlled infusion and real-time delivery monitoring are helping to define this relationship more accurately, enabling teams to move from empirical dose escalation to evidence-based dose design.

Although device design cannot completely negate biological variability, it can stabilise the physical conditions of delivery in terms of location flow and distribution, reducing the influence of external factors on therapeutic performance. In this sense, delivery systems are an integral and essential part of the therapy’s design; the therapeutic without the device is useless. A minimum viable product (MVP) delivery device is essential even in early-stage therapy development, as it underpins both the predictability and scalability of clinical outcomes, as well as reducing risk to both the patient and therapy programme.

How To Demonstrate Precision

If defining precision is difficult, demonstrating it under clinical conditions is even harder. Many CGTs show encouraging results in modelling and in vitro studies, only to encounter unexpected variability once tested in animals or humans. Translating a theoretical understanding of location, dose and delivery pattern into reproducible, in vivo performance remains one of the toughest challenges in the field.

The difficulty often emerges during the transition from therapeutic discovery to device-specific preclinical testing. Early studies may demonstrate vector bioavailability or device function separately, focusing on establishing foundational performance characteristics; however, this separation can limit understanding of how the two interact under physiological conditions. As a result, the first time the full system is tested, typically in animal models, teams may struggle to interpret poor outcomes. The question being: is the issue with the therapy itself or with how it was delivered?

If the delivery device or route is not well characterised before entering in vivo preclinical work, study design, surgical procedures and even success criteria can become ambiguous or have a lack of reproducibility.

Study Design

Preclinical study design therefore becomes the first true test of precision. The chosen route of administration determines not only how the therapy will be delivered, but also which model is appropriate for advancing an MVP approach to device design that supports overall therapy development. For example, a device that matches the therapy development stage and its requirements allows for evidence gathering on the control of delivery – isolating results for therapeutic effectiveness.

Anatomical and physiological differences, particularly in vascular structure, tissue density or organ size, mean that delivery parameters optimised in animals may not translate directly to humans. Building these constraints into the study design early on can help teams interpret results with greater confidence.

Procedural Control

Demonstrating precision also depends on procedural control. Every step, from therapy preparation and handling to administration and post-delivery care, can influence efficacy. For cell therapies, cell sedimentation during preparation or delays between thawing and delivery can alter dose consistency and viability. For gene therapies, infusion rate, device placement and user variability can all shift distribution patterns. Integrating human factors engineering into device and protocol design using procedural expertise helps to standardise these steps, thus improving reproducibility and safety.

Regulatory Scrutiny

Ultimately, preclinical and clinical studies are where precision delivery meets regulatory scrutiny. Demonstrating that a therapy and its delivery system consistently achieve targeted exposure is essential for proving both safety and efficacy. Without an early integrated approach to development of the device, formulation and route of administration, teams risk employing complex and expensive animal models or clinical studies only to discover that the delivery method itself limits their ability to assess therapeutic potential.

Incorporating delivery design and evaluation early in development is therefore not just good engineering – it is a strategic safeguard. Precision that is defined, engineered and tested in parallel with the therapy dramatically increases the chances of reproducible success in the clinic.

Conclusion: Precision Delivery Is The Next Frontier

The future of CGTs will not be defined solely by novel vectors or manufacturing breakthroughs, but by the industry’s ability to deliver these therapies with accuracy and consistency at scale. As CGTs move towards broader indications, the need for predictable, accessible delivery will only intensify. Achieving precision demands earlier integration of biological, engineering and human factors design, alongside continued investment in modelling and device innovation. Precision delivery bridges the gap between discovery and patient impact, turning theoretical efficacy into real-world benefit.

The lesson is clear: precision delivery is not a supporting technology, but the missing link that will connect scientific ingenuity with clinical and commercial success. Those who master it will define the next era of CGTs.

References

1. Patel MJ et al, “Surgical Approaches to Retinal Gene Therapy: 2025 Update”. Bioengineering, 2025, Vol 12(10), art 1122.
2. “Spark’s gene therapy price tag: $850,000”. News Article, Nature Biotech, Feb 6, 2018.
3. Morris EC, Neelapu SS, Giavridis T & Sadelain M, “Cytokine release syndrome and associated neurotoxicity in cancer immunotherapy”. Nature Rev Immunol, 2022, Vol 22(2), pp 85–96.
4. Perez BA et al, “Management of Neuroinflammatory Responses to AAV-Mediated Gene Therapies for Neurodegenerative Diseases”. Brain Sci, 2020, Vol 10(2), art 119.
5. “Upstaza (eladocagene exuparvovec)”. Web Page, EU EMA, accessed November 2025.
6. Cao D et al, “Innate Immune Sensing of Adeno-Associated Virus Vectors”. Hum Gene Ther, 2024, Vol 35(13–14), pp 451–463.