A spray-dried inhalation particle is a few microns wide. It has to carry an amorphous API to the lungs, survive storage and handling without crystallizing, and release the drug quickly at the correct place. Three requirements that pull in opposite directions and are decided in the few seconds it takes a droplet to dry.
Bend Bioscience, one of the United States’ leading spray-dry CDMOs, partnered with amofor to develop a mechanistic understanding of exactly that kind of particle. This is a non-standard core-shell particle that carries 5-azacytidine (5-AZA), used in the treatment of lung cancer. 5-AZA degrades rapidly in water, over 20% breaks down within an hour. To avoid this, the team dissolved the drug in dimethyl sulfoxide (DMSO) and kept it separate from the aqueous excipient solution until just before spraying.
By combining structured process development with PC-SAFT-based thermodynamic modeling, the team made a hard-to-reproduce spray dryer process far more explainable and controllable. The work was published in Molecular Pharmaceutics in 2025.
Why are inhalation particles not “standard spray drying“?
Most spray drying targets uniform, homogeneous, amorphous particles. In this project we aimed for the opposite. The reason is a physical contradiction: bioavailability requires an amorphous, mobile, solvent-rich interior, while stability and handling require a hard, crystalline, low-mobility solid. You cannot have both in a homogeneous particle.
A core-shell architecture resolves the conflict by giving each requirement its own region of the particle:
- The core holds the amorphous API in a trehalose matrix that prevents crystallization. It is soft, mobile, and retains some solvent. This delivers bioavailability, but also makes the particle fragile.
- The shell is crystalline L‑leucine, an amino acid that crystallizes almost instantaneously during drying, forming a protective outer layer that gives the particle the robustness to survive handling, storage, and inhalation.
How critical is that shell? In this study, formulations without L-leucine yielded just 6% of usable powder. With L-leucine, yields ranged from 74 to 92%.
The core-shell concept, however, also makes development difficult. Success depends on controlled segregation and shell formation during drying, which is a fundamentally different engineering target than for standard ASDs.
Inside the droplet, the following events occur almost simultaneously: the solvent evaporates, the L-leucine becomes supersaturated and crystallizes on the surface, and the core solidifies into an amorphous matrix. The sequence must be correct; if the shell forms too late, the powder becomes sticky, and if the core dries too quickly, it crystallizes. While you can characterize the final powder, you cannot directly observe the pathway that produced it.
The core modeling challenge: crystallization changes the system while it dries
Producing a working core-shell particle requires optimizing several parameters simultaneously: solvent selection, excipient ratios (particularly the L‑leucine threshold below which no protecting continuous shell forms), drying temperature, and atomization mass flow. None of these can be tuned in isolation, and the operating window where all of them are satisfied at once is narrow.
The challenge was extending the spray-drying model to treat shell crystallization as an active process that changes the rest of the system in real time. Once L‑leucine crosses its supersaturation threshold, it stops being part of the liquid phase. The remaining core composition shifts because one component has just been pulled out of it, and the developing shell can influence how solvent continues to escape. Capturing this required defining, for this specific chemistry, at what supersaturation crystallization initiates, how quickly it propagates, what composition remains in the core afterward, and how DMSO partitions between core and shell.
This crystallization sub-model did not previously exist in amofor’s framework. It was developed for this project and integrated into the existing PC-SAFT core. To our knowledge, this is the first time a coupled crystallization-drying model of this kind has been applied to an inhalation core-shell particle.
What the model actually delivered
Bend Bioscience ran a structured design of experiments (DOE) across solvent composition, excipient ratios, and process conditions. The model explained why specific batches failed and why others succeeded, providing a predictive framework. Specifically, it could:
– Predict residual solvent partitioning, including DMSO retention trends and their implications for core mobility
– Link failed experiments to the unfavorable droplet’s trajectory through phase behavior space
– Identify when L-leucine reaches supersaturation and shell formation becomes likely
– Visualize the internal drying process, for the first time in this kind of particle architecture
– Help explain how higher 5-AZA loading (doubled from 10 to 20 wt%) changed the phase boundaries and drying behavior
One result illustrates just how far the process was pushed. The manufactured powders had glass transition temperatures as low as -6 °C, well below room temperature and around 80 °C lower than the spray dryer outlet temperature. Spray drying materials with such a low Tg was previously considered infeasible. However, the L-leucine shell made it possible, and the PC-SAFT model explained the reason why.
The phase diagrams for this system are complex: Solubility lines, crystallization boundaries, and drying trajectories on a multi-component composition triangle. They take effort to read. But once you can read them, they explain why a process condition produces a working or failing particle, not just that it does.
The model does not capture every aspect of the system. Some phenomena, for example fine-scale shell morphology, certain kinetic effects at very short timescales, remain outside its current scope. What it does provide is a first-principles view of why the system behaves the way it does, which is something end-state measurements alone cannot deliver.
What this means for drug formulators
If you work with inhalation products or any non-standard spray-dried architecture, three takeaways from this project are worth keeping in mind.
- Watch what crystallizes before the glass forms.
The assumption that all solutes stay dissolved until the final glass forms is rarely true in practice. The moment any component precipitates mid-drying, the trajectory you expect diverges from the trajectory the droplet actually takes. Recognizing this early is often the difference between a process you can scale and one you cannot.
- Threshold effects matter.
Shell formation requires a minimum L‑leucine concentration. Residual solvent has a maximum tolerable level. Core stability depends on a glass transition that shifts with retained DMSO. These are step changes, not gradients, and the PC-SAFT model captures those thresholds.
- Modular modeling scales to non-standard problems
The crystallization sub-model built here was integrated into amofor’s existing software framework. The broader idea is modular: a thermodynamic core framework plus targeted add-ons can address many highly specific formulation questions (co-crystals, lipid systems, multi-API particles, controlled-release architectures etc.).
Ready to de-risk your inhalation spray-drying process?
The hardest formulation problems are the ones where the mechanism is unclear.
If your formulation involves unusual particle architectures, difficult solvent systems, multi-phase drying, or coupled crystallization, this kind of modeling can help you explain failure modes, define robust process windows, and accelerate scale-up. Talk to Dr. Christian Lübbert to see how PC‑SAFT modeling in Solcalc can support robust spray-drying process development!
