Amorphous Solid Dispersions as Non-Equilibrium Materials

Abstract

An amorphous solid dispersion is a molecular glass that the pharmaceutical literature specifies as though it were a solution: by drug, polymer and drug loading, occasionally with a glass transition temperature. Composition constrains such a material without determining it. Dispersions of identical composition differ in physical stability and in dissolution, and the difference depends on the path taken onto the energy landscape. We treat the dispersion as a kinetically arrested material and describe it using seven state variables.

These variables capture the changes that occur during manufacturing, storage, downstream handling, and dissolution: thermal and relaxation state, mixing scale, residual order, water and residual solvent content, particle and surface structure, chemical integrity, and age. Against this description we grade every same-composition comparison we could identify, by the strength of the link between history and outcome. The comparisons do not agree on which preparation route yields the more stable or better-performing product. Where an outcome is explained, the explanation is a measured structural state—residual crystals, compositional heterogeneity, surface enrichment—rather than the route, which serves as a label for the states a process tends to produce. In single-component molecular glasses, preparation history survives in the configurational enthalpy and the relaxation spectrum at unchanged glass transition temperature and heat capacity, and predicts crystallisation.

We found no equivalent measurement for a drug–polymer dispersion: none of the graded comparisons reported a fictive temperature, an enthalpy recovery or a relaxation time. Whether an entangled polymer preserves or erases the drug’s relaxation memory is the field’s outstanding question. Because the variables that carry history are seldom reported, we propose a minimum set of state descriptors, each with its detection limit and the length scale over which it was tested.

1. Introduction

Roughly two-thirds of drug candidates entering development are too poorly soluble in water to be absorbed from the crystal, and rendering the drug amorphous is among the most effective ways to raise its apparent solubility and dissolution rate [1,2]. A neat amorphous drug is rarely durable enough to formulate, so it is dispersed in a polymer, giving an amorphous solid dispersion [3]. The approach is established commercially: the US Food and Drug Administration approved 48 such products between 2012 and 2023 [4].

The object so produced is a molecular glass, usually binary, in which a small rigid molecule with strong hydrogen-bonding capacity is mixed with a high-molecular-weight polymer whose segmental dynamics dominate the mixture. Its preparation routes differ qualitatively rather than in cooling rate alone: vitrification from the melt, removal of solvent from a solution, and mechanical amorphisation of the crystal. It is then stored for years, plasticised by atmospheric water, compressed into tablets and finally dissolved. Almost every process that glass science studies deliberately—quenching, ageing, rejuvenation, mechanical activation, plasticisation, devitrification—is applied to this material, usually without being named.

What is less familiar is how the material is specified. An amorphous solid dispersion is conventionally described by its composition: the drug, the polymer and the drug loading, sometimes with a glass transition temperature, Tg. That description would be sufficient for a solution. For a glass it is not, since structure and dynamics depend on the path by which arrest was reached and on what has happened since [5,6]. Two examples make the gap concrete:

• Same composition, three routes. Naproxen dispersions prepared by spray drying, hot-melt extrusion and cryo-milling differed in the highest drug loading that remained fully amorphous. In poly(vinylpyrrolidone) the ceiling was 30–35 wt% by spray drying and 55–60 wt% by extrusion [7].
• Same route, one setting changed. Two extrusion batches of a 40 wt% dispersion in a vinylpyrrolidone–vinyl acetate copolymer differed nominally in screw speed alone. Both showed a single Tg near 88 °C and were amorphous by X-ray diffraction. Raman mapping resolved regions well above 50 wt% drug in the slower batch, and only that batch crystallised within two months at ambient conditions [8].

Composition and Tg were the same in both cases, and the materials still behaved differently.

That amorphous pharmaceuticals are non-equilibrium solids is not a new observation [5,6], and recent structural work describes pharmaceutical glasses as heterogeneous, with preparation method and handling time among the determinants of that heterogeneity [9,10]. Reviews of manufacturing have compared production routes and discussed how thermal history, particle morphology and nucleation differ between them [11,12,13,14], including fictive temperature, structural relaxation and fragility [14]. Other reviews address physical stability and its prediction [15,16], supersaturation and phase behaviour during dissolution [17,18,19], and characterisation [20,21]. The present review connects these previously separated topics through a common, explicitly empirical state-variable framework that follows a dispersion from manufacture through storage and downstream handling to dissolution. It grades route and process comparisons according to whether the relevant state was measured, distinguishes composition-matched comparisons from adjacent or confounded evidence, and proposes a practical reporting set including detection limits and probed length scales.

These strands have largely been pursued separately, and a shared description of state that follows a batch from manufacture to dissolution is still lacking. We therefore examine studies in which composition is held constantly, recognising that not every route comparison is strictly composition matched. We also propose reporting descriptors: in one survey of 40 performance studies, 86% did not state whether dissolution was run under sink or non-sink conditions [22].

This review treats a dispersion as a non-equilibrium material whose state depends on its history and uses seven empirically motivated variables as an organising framework: thermal and relaxation state; mixing scale; residual order, including detected crystals and possible nuclei; water and residual solvent; particle and surface structure; chemical integrity; and age with storage history. These variables are not exhaustive, mathematically derived, independent or universally sufficient to predict performance. Excipients, dosage-form composition and tableting can also modify behaviour; they are considered where they alter the dispersion’s state.

The scope is drug–polymer dispersions. Co-amorphous systems appear as a boundary case [23]; drugs in mesoporous carriers are excluded, since confinement rather than a bulk glass solution governs their behaviour. Dosage-form development is not reviewed separately, but milling, blending, compression and tablet-performance studies are discussed selectively where they show that downstream operations modify the manufactured state.

This is a narrative critical review. Literature was identified iteratively through database searching and backward and forward citation tracking. The search was question-led: (i) comparisons with constant composition and varied route or process settings; (ii) studies linking measured structural features to stability or performance; (iii) studies reporting relaxation descriptors such as fictive temperature, enthalpy recovery or relaxation time; and (iv) analytical-method papers defining detection limits and length scales. Because no systematic-review protocol was used, the corpus is not presented as exhaustive and no screening counts are claimed. The exact databases, search strings, final search date and record numbers were not retained and cannot be reconstructed reliably; they should be reported if contemporaneous records exist. Same-composition comparisons are concentrated in a few research groups, and we found none reporting relaxation measurements for a drug–polymer dispersion.

We assigned qualitative evidence grades. A-structure denotes a measured mixing, surface or crystalline state linked to stability or performance. A-relaxation denotes a measured relaxation state, such as enthalpy recovery, fictive temperature or relaxation time, linked to stability or performance. Grade B denotes different histories at nominally fixed composition, giving different outcomes without a measured linking state variable. Grade C denotes process conditions associated with outcomes only. Comparisons in which loading, excipients, particle size, residual solvent or downstream processing differed are flagged as not strictly composition-matched or as confounded. The grades are author-assigned, not validated risk-of-bias scores, and indicate the strength of the mechanistic link, not study quality. The record does not show whether both authors graded independently; the authors should confirm this before submission. The two A grades are kept separate so that drug–polymer evidence is not conflated with single-component glass evidence.

2. What Kinetic Arrest Implies for Specification

2.1. A Glass Is Not Specified by Tg

The physics needs no rehearsal here. A liquid cooled below its melting point without crystallising falls out of equilibrium when its structural relaxation time exceeds the observation time, and the temperature at which it does so depends on the rate of cooling and of subsequent reheating [24,25]. The resulting glass occupies one of many accessible minima on the potential energy landscape, and which one depends on the path [26,27]. Fragility measures how steeply the relaxation time rises on approach to arrest [25]; most small organic drugs are fragile [28]. Kauzmann’s observation that extrapolating the supercooled entropy below Tg would carry it beneath the crystal, so that arrest must intervene first, remains the cleanest statement of why the glass transition cannot be a thermodynamic boundary [24].

Two consequences matter for a formulation scientist. A glass is not at equilibrium and therefore continues to evolve. And Tg identifies where arrest occurred on one heating protocol, not which glass was made.

The second point has been demonstrated directly, though from outside pharmaceutics. Tsukushi and colleagues prepared tri-O-methyl-β-cyclodextrin as a glass by grinding the crystal in a vibrating mill at room temperature and, separately, by quenching the liquid. Adiabatic calorimetry from 13 to 375 K returned the same heat capacity for both and a glass transition at 345 K in both. On heating through that transition the ground glass released more than twice the configurational enthalpy of the quenched one [29]. Identical Tg, identical Cp, different position on the landscape.

The state variable that carries this difference is the configurational enthalpy, conventionally expressed as the fictive temperature Tf, the temperature at which the glass would be in equilibrium with the supercooled liquid whose structure it retains [30,31,32]. A slowly cooled or aged glass has a low Tf; a quenched or mechanically activated one has a high Tf. Relaxation towards equilibrium is non-exponential and non-linear and is handled by the Tool–Narayanaswamy–Moynihan formalism [30,31,32] or measured as enthalpy recovery [33]. Vapour deposition makes the complementary point: depositing molecules onto a substrate held below Tg allows each to find a low-energy position before burial, producing glasses of a density and kinetic stability that no practical cooling rate can reach, as shown for indomethacin among other pharmaceutical molecules [34,35,36]. The preparation path does not perturb the glass; it selects which region of the landscape is reachable.

Motion does not cease below Tg. Cooperative α-relaxation slows steeply but remains measurable, and localised Johari–Goldstein β-relaxation persists far below [37,38]. Hancock, Shamblin and Zografi measured enthalpy recovery in indomethacin, poly(vinylpyrrolidone) and sucrose stored 16–47 K below their glass transitions and concluded that a glass must be held at least 50 K below its experimental Tg before calorimetrically detectable motion can be neglected over a pharmaceutical product’s shelf life [39]. The familiar ‘Tg − 50 K’ rule is therefore an empirical guideline derived from particular materials, measurements and timescales—not a universal stability criterion or thermodynamic limit. Its applicability should be assessed for each drug–polymer system, moisture level, storage duration and relevant relaxation or crystallisation mechanism.

Mobility is routinely connected to stability through configurational entropy, by way of the Adam–Gibbs relation and its Vogel–Tammann–Fulcher and Adam–Gibbs–Vogel forms [40,41]. The connection holds under conditions worth stating. Relaxation times measured above Tg and extrapolated below it predicted the crystallisation onset of amorphous indomethacin to within a factor of two, 473 h predicted against 624 +/− 158 h measured at 25 °C, but failed for other compounds in the same study [42]. Whether the extrapolation applies at all depends on whether molecular diffusion limits the rate; glasses with a strong crystallisation tendency deviate from it [43]. Mobility is necessary for crystallisation, though on this evidence it does not by itself predict it.

2.2. Preparation History Persists in Single-Component Glasses

The cyclodextrin result is not isolated. In neat drugs and excipients, history is retained in the relaxation state while Tg is not displaced:

Trehalose, four routes. Amorphous trehalose prepared by freeze-drying, spray drying, dehydration of the dihydrate and melt-quenching showed the same Tg and fragility across routes but different enthalpic relaxation, and stability followed the relaxation rather than the thermal parameters: the dehydrated material was least stable, the melt-quenched most stable, the two dried materials intermediate [44].

Trehalose, relaxation times. Three of those routes were later characterised dynamically. Tg (115.8–116.3 °C), the heat-capacity step and fragility were indistinguishable, yet the glassy α-relaxation times were not. Below Tg the dehydrated glass crystallised in 44 days at 120 °C and 59 days at 115 °C while the freeze-dried and spray-dried glasses had not crystallised after 170 days; crystallisation onset scaled with the α-relaxation time with log–log slopes of 1.13–1.23. Above Tg the distinction vanished [45].

Mechanical stress. In amorphous indomethacin, handling and compression altered the relaxation behaviour below about 35 °C, and the more stressed glasses crystallised sooner [46].

Grinding against quenching. Ursodeoxycholic acid amorphised by grinding and by melt-quenching gave the same X-ray halo but differed in heat of solution, thermal behaviour and infrared and NMR spectra, and crystallised differently on heating [47]. Cryo-ground indomethacin retained a memory of the polymorph it was ground from [48].

Surface area. Freeze-dried ritonavir, with a far larger specific surface area than melt-quenched material, began to crystallise much sooner at the same reduced temperature, departing from the correlation established on low-surface quenched glasses [49].

For single-component glasses, then, preparation history persists as a measurable state; that state predicts stability better than Tg, and it is lost once the material is returned to the liquid. Whether any of this survives the addition of an entangled polymer that dominates the mixture’s mobility has not been tested directly. Section 7.1 returns to it.

2.3. Glass-Forming Ability Is Defined by Its Protocol

The propensity of a drug to form and retain a glass is commonly classified by whether it crystallises on cooling from the melt, on reheating, or not at all, at a standard 20 °C/min [50]. The classification is useful, and it is defined by the protocol used to obtain it. Critical cooling rates span a wide range, about 118 °C/min for carbamazepine against about 588 °C/min for griseofulvin [50], so a drug near a class boundary crosses it when the rate changes.

Changing the preparation route does exactly that. Of 26 drugs classified from the melt, 22 changed class when the glass was formed by spray drying instead: of the 16 in the most stable class, nine moved to the middle class and six to the least stable, and seven of the ten in the middle class moved to the least stable [51]. Classifications from rapid solvent evaporation and from melt cooling agreed in 68% of 93 compound–solvent cases [52]. The ranking between drugs survives the change in protocol even where the class does not: drugs requiring a low critical cooling rate also required only short milling times to amorphise [53]. Glass-forming ability is best read as a relative molecular property measured under a stated protocol. Attempts to build a general stability criterion by relating crystallisation on cooling to isothermal storage through the reduced temperature Tg/T [54,55] inherit the same caveat, and as noted above, high-surface material departs from the correlation [49].

2.4. What the Polymer Changes

Adding a polymer raises the mixture’s Tg, reduces mobility and introduces specific interactions, typically hydrogen bonds [56]. Miscibility is usually treated by Flory–Huggins theory with the interaction parameter obtained from melting-point depression or solubility parameters [57,58], and increasingly by equation-of-state approaches such as PC-SAFT [59], yielding temperature–composition phase diagrams that identify the loading thermodynamically stable at a given temperature [60,61].

Two distinctions carry through the rest of this review. The first is between solubility and miscibility: a drug may be miscible with a polymer as a supercooled liquid across a wide composition range while its equilibrium crystalline solubility in that polymer is far lower [62]. The second is between what is thermodynamically permitted and what a given trajectory actually produces. Phase diagrams describe equilibrium; they manufacture traps for a kinetic state. The maximum loading that remains detectably amorphous is therefore not an intrinsic property of manufacturing route alone: it depends jointly on the drug–polymer system, the process conditions and the manufacturing route, as well as the analytical method and its detection limit [7].

Water is a third component and an unusually consequential one. It plasticises, lowering Tg, and it competes for the drug–polymer interactions that stabilise the mixture, so that humidity can drive an initially homogeneous dispersion to separate into drug-rich and polymer-rich amorphous phases [63,64]. For a physicist the point is that the composition of a pharmaceutical glass is not fixed by the formulator; it is set jointly by the formulator and the atmosphere.

2.5. Boundary Cases

In co-amorphous systems the drug is stabilised by a second small molecule—an amino acid, an organic acid, a second drug—rather than by a polymer [65,66]. They share the physics described here and have been produced by melt extrusion as well as by quenching and milling [67], but they lack the high molecular weight and entanglement that dominate mobility in polymer dispersions, which is precisely the variable at issue in Section 7.1. They are invoked here only where they illuminate the general argument. Drugs confined in mesoporous silica and related carriers are excluded: pore size, surface chemistry and degree of filling govern their behaviour, not the properties of a bulk glass. The distinct history-dependent glass states are illustrated schematically in Figure 1.

Figure 1. Enthalpy–temperature diagram for a single composition. The equilibrium and supercooled liquid are drawn as one line; below it, separate glass lines correspond to slow cooling, fast quenching, solvent removal and mechanical activation, each with its own fictive temperature. Arrows show ageing towards the supercooled-liquid line, plasticisation by water, and reheating above Tg, where the distinct histories converge. Schematic conceptual illustration prepared by the authors; the underlying glass-history concepts are discussed in [24,25,26,27,28,29,30,31,32,33,34,35,36,39].

Download the full article as PDF here Amorphous Solid Dispersions as Non-Equilibrium Materials

or continue reading here

Ebrahimi, F.; Major, I. Amorphous Solid Dispersions as Non-Equilibrium Materials. Entropic Disord. Matter 2026, 1, 6. https://doi.org/10.3390/edm1020006

Read more interesting articles on Solid Dispersion here:

• The solid dispersion of a novel curcumin analogue (CACN136) prepared using PVP/Mannitol composite carrier significantly improved oral bioavailability and depression-like behavior in mice
• Novel Hybrid Polymer-Lipid-Based Formulations for Enhanced Dissolution Rates of Poorly Water-Soluble Drugs: A Quality by Design (QbD) Approach via Hot Melt Extrusion
• Amorphous Solid Dispersions – Part 1