A surfactant inside the dispersion: how TPGS lets felodipine–PVPVA amorphous solid dispersions release well at higher drug loading

Background and motivation

Many new drug candidates are poorly soluble in water, and the transient solubility gain of an amorphous drug alone is often too small for a high-dose drug. An amorphous solid dispersion (ASD), in which the drug is molecularly dispersed in a hydrophilic polymer, can do better because a hydrophilic polymer dissolves many times faster than a lipophilic drug. When release is controlled by the polymer, the drug is carried out with it and the solution becomes highly supersaturated, exceeding the drug's amorphous solubility; the excess drug then separates as amorphous drug-rich nanoparticles, a process called liquid–liquid phase separation. These sub-micrometre particles are thought to add dissolution and bioavailability benefits, although how they form and what they do in vivo is "still under debate" (Section 1; refs 3–15) C1.

This advantage holds only up to the limit of congruency (LoC), the highest drug loading (DL) at which drug and polymer are released at the same rate. For polyvinylpyrrolidone/vinyl acetate (PVPVA, Kollidon VA 64) dispersions the LoC is often low, so high-dose drugs end up in large or multiple dosage units C2. The group had shown earlier that changing the polymer shifts the LoC, but with a trade-off: relatively hydrophobic polymers gave higher LoCs at the cost of slower release than hydrophilic ones (Saboo et al. 2020, ref 4) C3. For felodipine with PVPVA, that study found an LoC of only 15% DL R12. The practical goal is therefore fast, complete release from a hydrophilic polymer at a reasonably high drug loading.

A third component is one route. Surfactants in ternary ASDs have been reported to improve wetting, inhibit crystallization in the solid and in solution, speed dissolution with or without solubilization, and promote and stabilize drug-rich nanoparticles; but they can also raise crystallization risk C4. Baghel et al. found that sodium dodecyl sulfate and poloxamer 188 reduced the physical stability and dissolution of polyvinylpyrrolidone- and HPMC-based ASDs by promoting crystallization (ref 25) R30. For itraconazole, TPGS gave synergistic precipitation inhibition in an HPMCAS-based ternary ASD but phase separation and crystallization in a PVPVA-based one (refs 26, 27) R29. The authors' position is that, provided crystallization is controlled by the choice of polymer–surfactant pair or the drug crystallizes slowly, surfactants "can provide release advantages" C5. The closest precedent is Que et al. (2019, ref 16): some, but not all, surfactants raised the LoC of ledipasvir–PVPVA ASDs, and the mechanism "remains elusive" R26·C6. Harmon et al. (ref 14) studied surfactant-containing ASDs at a single drug loading only.

The paper therefore had two aims: to show that a surfactant can raise the LoC of a model ASD and how this depends on surfactant level, and to probe the mechanism. The model system is felodipine, a biopharmaceutics classification system class II drug widely studied in ASDs, with PVPVA and D-α-tocopheryl polyethylene glycol 1000 succinate (TPGS), a surfactant present in the marketed ASD products Viekira Pak and Belsomra (Figure 1) M1·M2·M3. Felodipine has a molecular weight of 384.25 g/mol, a melting point of 147 °C, a Tg of 45 °C and a log P of 3.44 (Table S1, cited values) M1; its crystalline and amorphous solubilities in the pH 6.8, 100 mM phosphate buffer used throughout are 1.3 and 8 µg/mL (ref 4) M1.

Background concepts

Congruent release and the limit of congruency. In a surface-normalized dissolution test a compact of fixed area dissolves from one face, and the release of each component is measured separately. Release is congruent when drug and polymer leave at the same rate, so the drug's release is controlled by the fast-dissolving polymer; it is incongruent when the polymer leaves much faster than the drug, and drug release falls toward that of the neat amorphous drug. The authors explain the size of this switch by the difference between the two limits: neat PVPVA releases at about 3 mg/min/cm² and neat amorphous felodipine at about 0.001 mg/min/cm² (ref 4), roughly three orders of magnitude apart R13·C13. The LoC is the drug loading at which the switch happens; in this paper it is reported as the highest drug loading tested that still released congruently R12·X2.

Critical micelle concentration and micellar solubilization. Above its critical micelle concentration (CMC) a surfactant forms micelles, which can take up a hydrophobic drug and raise its apparent solubility. With a polymer in the solution the onset is called the critical aggregate concentration (CAC). The paper measures it two ways: by the TPGS concentration at which the solubility of crystalline felodipine starts to climb, and by the fluorescence probe pyrene, whose emission ratio I1/I3 (372/383 nm) falls when it moves from water into a less polar micellar environment P1·P2. If a surfactant helped only by solubilizing the drug, the effect should depend on whether the TPGS concentration in the medium is above the CMC.

Water-induced phase separation and morphology. Water entering an ASD is an anti-solvent for the drug, creating a thermodynamic driving force for the drug and polymer to separate (described by the delta chi effect, refs 38, 39, the different affinity of water for drug and polymer); water is also a plasticizer that raises molecular mobility and so speeds any favoured separation (ref 40) C28. The separated drug-rich phase can be present as discrete domains in a continuous polymer-rich phase, or as a connected, network-like phase. From the literature the authors argue that no separation, or discrete sub-micrometre drug-rich domains, goes with congruent, faster release, whereas semi-continuous or continuous drug-rich phases go with incongruent release C30. Two cited examples: in ritonavir–polyvinylpyrrolidone ASDs the switch from congruent to incongruent release coincided with discrete drug-rich domains becoming a continuous phase (Han et al., ref 36), and in lopinavir–HPMC film immersion only 15% DL formed discrete domains and dissolved completely, while higher loadings left contiguous drug-rich material (Li and Taylor, ref 35) R24·R25. Domain shape, size and composition, and the kinetics of separation, are proposed as the reason why earlier reports disagree on whether phase separation harms release C29.

Coarsening and interfacial energy. After phase separation, domains tend to grow over time and reduce the interfacial area between the phases, by coalescence of neighbouring domains or by Ostwald ripening (growth of large domains at the expense of small ones). The driving force is the interfacial energy; a surfactant that lowers it would be expected to slow coarsening. The paper uses this idea to explain its morphology results, but does not measure interfacial energy or coarsening rates C40·C51.

The transient gel layer. A dissolving amorphous polymer forms a hydrated, plasticized layer at its surface. Polymer chains in this layer must disentangle before they can leave into solution, and disentanglement starts only once enough solvent has entered to plasticize the matrix (refs 47, 49). The authors' release model, an extension of their earlier model for binary ASDs (ref 17), places phase separation inside this transient gel layer C37·C43.

AFM-nanoIR imaging. In atomic force microscopy (AFM) a sharp tip scans the surface and records height (topography) and cantilever deflection. In nanoIR imaging an infrared laser tuned to an absorption band heats the spot under the tip; the thermal expansion makes the cantilever ring, and the ringdown amplitude is proportional to the local absorption. Imaging at a band of the drug and a band of the polymer, and dividing the two images, maps drug-rich and polymer-rich regions with sub-micrometre resolution P16.

Methods in detail

The study has two parts. The first measures release across drug loading at four TPGS levels and locates the LoC for each. The second tests explanations: micellar solubilization (CMC measurements, the TPGS concentrations reached in the medium, and a control with TPGS dissolved in the medium), polymer dissolution (a drug-free control), molecular mobility (Tg), and the structure of the wetted material (tablet surfaces and humidified films). All work used the same pH 6.8, 100 mM phosphate buffer M6.

The dispersions

Felodipine, PVPVA and TPGS were dissolved at the desired ratio in a 1:1 (v/v) methanol:dichloromethane mixture at 50 mg/mL total solids, rotary evaporated with the bath at 45 °C, vacuum-dried overnight, cryo-milled and sieved to 106–250 µm; all powders were crystal-free by polarized light microscopy and powder X-ray diffraction P4·R6. The TPGS level is expressed as a weight percent of the polymer portion, not of the whole ASD: a 30% DL ASD with 5% TPGS is felodipine:PVPVA:TPGS 30:66.5:3.5 by weight (Figure 2) P4. Table A reproduces the composition grid of Table S2, including the TPGS concentration that each ASD would give in the 100 mL medium if all of it dissolved.

Table A. ASD compositions studied or referenced (from Table S2): PVPVA wt% in the ASD, and in parentheses the TPGS concentration in µg/mL in 100 mL at 100% release. "—" = not prepared P4. The Table S2 caption prints the unit as "g/mL"; the values are µg/mL (3.5 mg TPGS in 100 mL = 35 µg/mL, as the text states for the 30% DL, 5% TPGS ASD) X9.

Release testing

Release was measured with a Wood's intrinsic dissolution apparatus. 100 mg of ASD powder was compressed in an 8 mm die (exposed area 0.5 cm²) at 1500 psi for 1 min, and the die was immersed in 100 mL of buffer stirred at 100 rpm; the DLS method states 37 °C as the dissolution temperature P6. At each time point 2 mL was withdrawn and replaced with fresh buffer; 100 µL was assayed unfiltered for drug and TPGS (TPGS by HPLC, UV 284 nm), and the rest was filtered through a 0.2 µm nylon filter for polymer analysis P6. The release rate of each component is the slope of the initial linear part of the percent-release-versus-time curve, normalized to area, from three replicates; the binary (0% TPGS) data and the rates of neat amorphous felodipine and neat PVPVA were taken from the earlier study (ref 4) P6.

Note by the AI writer (not discussed in the paper): the paper does not state the mass basis of the rates. Drug, polymer and TPGS of a congruent ASD give nearly the same rate in mg/min/cm², and the initial felodipine slope of the 30% DL, 5% TPGS ASD (about 14% in 10 min) times the 100 mg tablet over 0.5 cm² gives about 2.8 mg/min/cm², close to the plotted value; the rates therefore appear to be expressed per total ASD mass X4.

Two controls followed. A drug-free PVPVA matrix with 5% TPGS was prepared the same way and its intrinsic dissolution rate measured P7. And a 30% DL binary ASD was dissolved in medium containing 35 µg/mL of pre-dissolved TPGS, the amount a 30% DL ASD with 5% TPGS would release in full P8.

Measurements in solution

For the CMC by solubility, excess crystalline felodipine was equilibrated for 48 h at 37 °C in buffer containing TPGS (about 10–80 µg/mL), ultracentrifuged at 151,000 × g for 30 min at 37 °C and the supernatant assayed by HPLC (n = 3) P1. For the CMC by fluorescence, pyrene (0.01% w/v) was spiked into buffer with 0–120 µg/mL TPGS and excited at 332 nm P2. Both were repeated with 900 µg/mL PVPVA, chosen as a representative polymer concentration P1·P2. The amorphous solubility of felodipine was measured in buffer with 900 µg/mL PVPVA by the method of ref 4 (n = 3) P3. At the end of dissolution (60 min), the medium of the 5% TPGS ASDs and of the 10% and 20% DL binary ASDs was examined by nanoparticle tracking analysis (NanoSight LM-10, 532 nm laser, 37 °C, 100 × 80 µm field) and by dynamic light scattering (Malvern Nano-ZS, 173° backscatter, 37 °C) P9·P10.

Structure of the solid and the wetted material

Glass transitions were measured by modulated DSC (5 °C/min, modulation 1 °C every 60 s, second heating) on amorphous felodipine, PVPVA and 30% DL ASDs at each TPGS level P5. To see the dissolving surface, 30% DL compacts at 0, 2, 5 and 10% TPGS were taken out after 10 min of dissolution, vacuum-dried overnight and scanned by X-ray microcomputed tomography (micro-CT) at 3 µm/pixel P11. To see how the wetted material separates, 30% DL films were spin-coated onto ZnS under nitrogen below 20% RH (to avoid separation during preparation), dried, held for 12 h in a chamber saturated with water vapour (about 100% RH), dried again and imaged by AFM over 10 × 10 µm P12·P13·P14. For the binary film and the 10% TPGS film, nanoIR images were recorded at 1213 cm⁻¹ (a felodipine band) and 1243 cm⁻¹ (a PVPVA band), chosen from bulk IR spectra of the neat amorphous materials, and divided to give ratio images in which drug-rich regions appear red P15·P16·R22. The films are described as containing TPGS "by weight" without saying whether this is of the polymer portion, as for the powders X8.

Experimental analysis

The LoC moves with the TPGS level

Without TPGS (data from ref 4), the 10% DL ASD released PVPVA and felodipine together, at about 2.55 and 2.35 mg/min/cm², and about 59% and 51% of each was out after 60 min (Figures 2A and 3) R9.a·R7.a–b. At 20% DL release failed: PVPVA fell to about 0.1–0.2 mg/min/cm² and felodipine to about 0.009, and only about 1% of the drug was released in 60 min R9.c·R7.c–d. With 5% TPGS the picture changed (Figure 3). At 10, 20 and 30% DL, drug, polymer and TPGS left the compact at nearly equal rates of about 2.6–3.2 mg/min/cm²; at 35% DL release was still congruent but slower, about 1.7–1.8 mg/min/cm²; at 40% DL PVPVA release dropped to about 0.1–0.2 and felodipine to about 0.003 mg/min/cm², and TPGS was below its quantification limit R9.b·R9.d–g. The authors describe "A clear contrast" between the two series and place the LoC at 35% DL with 5% TPGS C12.

The same pattern held at the other TPGS levels (Figure S3). With 2% TPGS release was congruent at 10, 20 and 25% DL (about 2.2–3.2 mg/min/cm²) and failed at 30% DL (PVPVA about 0.2–0.3, felodipine about 0.008). With 10% TPGS it was congruent at 10, 20, 30 and 40% DL (about 2.65–3.25) and at 45% DL (about 1.0–1.1), and failed at 50% DL (PVPVA about 0.04, felodipine about 0.004) R10. Figure 5 summarizes the LoC as 15, 25, 35 and 45% DL for 0, 2, 5 and 10% TPGS R12. The authors write that the LoC "scaled with" the TPGS level and that the study is "clearly demonstrating an improved LoC" C16.

Table B. Surface-normalized release rates (mg/min/cm², n = 3) either side of each LoC, read from Figure 3 and Figure S3 (approximate; the axes break at 0.05 mg/min/cm², so values near the break are rough) R9·R10.

The release profiles (Figure 2B) show what "congruent but slower" means at 35% DL with 5% TPGS: almost nothing came out for the first 15 min (about 1.9% of the PVPVA at 15 min), after which drug, polymer and TPGS rose together to about 45–47% at 60 min. The 30% DL ASD released about 55% of its PVPVA, 49% of its felodipine and 48% of its TPGS by 60 min, and the 40% DL ASD only about 4.9% of its PVPVA and 0.4% of its felodipine R8.

Note by the AI writer (not discussed in the paper): three points qualify the LoC values. The binary LoC is stated as "only 15% DL" (Section 1), "between 10% and 20% DL" (Section 3.4) and "up to <20% DL" (Abstract); only the 10% and 20% DL binary results are shown X1. Each LoC is the last loading that passed on a grid of 5% DL steps, so the true switch lies somewhere in the next 5 points X2. And at each new LoC release was congruent but roughly half as fast as at lower loadings, with a lag at 35% DL, although the Abstract speaks of "rapid and congruent" release at higher drug loadings X3.

Supersaturation and nanoparticles follow congruency

Figure 4 follows the felodipine concentration in the medium. Congruently releasing ASDs drove it well above the amorphous solubility (9.5 ± 1.1 µg/mL measured here with 900 µg/mL PVPVA, within error of the 8 µg/mL without polymer) R4·C9: at 60 min about 68, 116, 165 and 158 µg/mL for 10, 20, 30 and 35% DL with 5% TPGS and about 52 µg/mL for the 10% DL binary ASD R11. Failing ASDs stayed below the amorphous solubility: about 2 µg/mL at 60 min for the 20% DL binary and the 40% DL ternary ASD, against about 0.15 µg/mL for neat amorphous felodipine R11. The authors conclude that congruent ASDs exceed the amorphous solubility while incongruent ones release "negligible amounts of drug" C14, and that TPGS "significantly improves" the loading range for polymer-controlled release C15.

Nanoparticle tracking analysis (Figure 8) showed scattering particles in the media of the 10% DL binary ASD and of the 10, 20, 30 and 35% DL ASDs with 5% TPGS, and no particles for the 20% DL binary and the 40% DL ternary ASD; nanoparticles therefore formed up to the LoC R16·C19. Their Z-average diameters at 60 min were about 276–280 nm for 10% and 20% DL and about 237 and 270 nm for 30% and 35% DL with 5% TPGS, and about 278 nm for the 10% DL binary ASD (Figure 9), with unimodal distributions (Figure S4; volume peaks near 295 nm without and 345 nm with TPGS) R17·R18. The authors report sizes stable over 60 min without agglomeration (data not shown) and no trend with drug loading or TPGS C20. This differs from an anacetrapib study in which nanoparticles became smaller as the TPGS level rose (ref 14) R28·C21.

Table C. Solution-side outcome of dissolution, read from Figures 4, 8 and 9 (approximate; the concentration axis of Figure 4 is broken between 1 and 20 µg/mL) R11·R16·R17. The paper does not say whether the binary curves of Figure 4 were measured anew or taken from ref 4 X12.

Is it micellar solubilization?

Both CMC methods agreed (Table 1): by fluorescence about 50 µg/mL in buffer and about 40 µg/mL with 900 µg/mL PVPVA, and by solubility 49 ± 4 and 41 ± 3 µg/mL R1. Below the CMC, crystalline felodipine solubility stayed near that in plain buffer (about 0.6–1.9 µg/mL; about 1.6–2.4 µg/mL with PVPVA), and above it rose to about 20 µg/mL at 80 µg/mL TPGS (Figure S1) R2. The authors read the small CMC–CAC difference as "little-to-no interaction" between TPGS and PVPVA affecting micellization, and the flat pre-CMC solubility as showing no significant solubilization by TPGS monomers C7·C8; they assume the same holds for the amorphous solubility, which was not measured with TPGS C10.

The key argument is a comparison with Table A C33. At 2% TPGS, the TPGS concentration at full release is only 14–18 µg/mL, far below the CMC, yet the LoC rose from 15% to 25% DL. At 30% DL, the 5% TPGS ASD would give 35 µg/mL (below the CMC) and the 10% TPGS ASD 70 µg/mL (above it), and both released congruently. The authors conclude that solution solubilization by micelles "plays a negligible role" in the LoC C33·P4·R10·R12.

Two controls support this. With the same 35 µg/mL TPGS dissolved in the medium, a 30% DL binary ASD still released only about 4.9% of its PVPVA and 0.4% of its felodipine in 60 min (Figure 7), against about 55% and 49% for the 30% DL ASD with TPGS inside (Figure 2B) R15·R8.a–b. And 5% TPGS did not make a drug-free PVPVA matrix dissolve faster: about 3.2 against about 3.4 mg/min/cm² for PVPVA alone (Figure 6) R14·C17. The authors take the first control as "indicating" that the surfactant must be part of the ASD matrix, as showing an effect "beyond any benefit incurred from wetting improvement", and as suggesting that the improvement belongs to the hydrated matrix or the ASD–solution interface C18·C34·C36. Note that the text quotes the neat PVPVA rate as about 3 mg/min/cm² (Section 3.4), while Figure 6 plots about 3.4 for the same cited data X11.

Note by the AI writer (not discussed in the paper): the concentrations in Table A assume all TPGS is released, but only about 45–48% of it was out after 60 min in the 5% TPGS profiles (Figure 2B). If the 10% TPGS ASD at 30% DL released a similar fraction, its bulk TPGS would have been about 35 µg/mL, also below the CMC, so the "above CMC" case was probably never above the CMC in the bulk during the test. This strengthens the conclusion that bulk micelles are not needed, but the comparison it rests on is nominal X14. Wetting itself was not measured, and the location control used one concentration below the CMC X13.

TPGS lowers the glass transition

Each 30% DL ASD showed a single Tg and no melting endotherm, "consistent with" a single-phase dispersion (Table 2) R5·C11. The Tg fell from 89.2 ± 1.7 °C without TPGS to 86.0 ± 0.5, 80.8 ± 0.8 and 71.1 ± 0.8 °C with 2, 5 and 10% TPGS, between the values for amorphous felodipine (44.7 ± 1.8 °C) and PVPVA (105.4 ± 1.9 °C) R5. All four lie well above the 37 °C test temperature in the dry state; the replicate number is not stated X17.

Tablet surfaces after partial dissolution

Micro-CT of 30% DL tablets after 10 min of dissolution (Figure 10) showed "A clear difference": tablets with 0% and 2% TPGS had a porous surface layer at the dissolving face, full of dark voids, while tablets with 5% and 10% TPGS had a relatively homogeneous morphology through their thickness R19·C22. At 30% DL the 0% and 2% TPGS ASDs are above their LoCs and the 5% and 10% ASDs below theirs, so the porous layer appears exactly where release was incongruent R9.e·R10.d·R10.g. The authors read the porous layer as what remains when the polymer dissolves faster than the drug, "leaving behind a porous drug-rich interface on the tablet surface", as seen earlier for other incongruent systems (refs 4, 7) C23; its composition was not measured.

Wet films: network versus discrete domains

Freshly prepared binary and 10% TPGS films were featureless by AFM (Figure S5), so the films started homogeneous R20. After 12 h at about 100% RH every 30% DL film showed domains (Figure 11). The authors describe "irregularly shaped semi-continuous domains" at 0% and 2% TPGS and "circular, discrete domains" at 5% and 10% TPGS, and report no significant difference in domain height between the films, judged from comparable height scales (±30 to ±50 nm) R21·C24. Bulk IR spectra (Figure 12) identified 1213 cm⁻¹ and 1243 cm⁻¹ as bands that discriminate felodipine from PVPVA R22. nanoIR at these bands (Figure 13) showed that the raised features are drug-rich and the continuous phase around them is polymer-rich or drug-lean: connected, elongated drug-rich regions in the binary film, and round, well-separated drug-rich domains in the 10% TPGS film R23·C25.

Note by the AI writer (not discussed in the paper): nanoIR covered only 0% and 10% TPGS, so the drug-rich assignment at 2% and 5% rests on topography alone X6. The height comparison has no statistic; Figure 13 uses a ±90 nm scale for the binary film against ±35 nm for the 10% TPGS film X18. The ratio images are shown on a −1.5 to 1.5 scale whose normalization is not stated X19. The IR method states a range of 1200–1800 cm⁻¹, while Figure 12 is plotted to 1000 cm⁻¹; both marker bands lie inside either range X10.

Discussion

The authors' release model

From the morphology differences, seen both in films where no mass leaves and at the tablet surface after partial dissolution, the authors propose a release model for the 30% DL ASDs (Figure 14). It extends their earlier model for binary ASDs (ref 17) and rests on a transient gel layer at the dissolving face, where water plasticizes the polymer chains, which disentangle before they leave C37. "We propose" that at high drug loading, whether release is drug- or polymer-controlled depends largely on the architecture of the drug-rich phase that forms when the hydrated matrix phase-separates, "assuming that phase separation occurs faster than polymer dissolution" C37·C38.

For the 0% and 2% TPGS ASDs, the contiguous network seen in films "likely also forms" in the gel layer during dissolution; the pores of the micro-CT surface layer would then be where the polymer-rich phase has left, and release becomes incongruent because the drug-rich and polymer-rich phases differ in solubility C39. For the 5% and 10% TPGS ASDs, the persistent discrete domains "can be rationalized" by TPGS lowering the interfacial energy between the phases, which reduces the coarsening rate; such domains "can release" intact as the gel layer recedes C40. Later, from the AFM images, the surfactant is called "clearly effective" at keeping the drug-rich phase discrete, "suggesting that surfactant reduces coarsening kinetics"; the micro-CT results are "supportive" of continuous drug-rich domains without surfactant, so rapid coarsening is "clearly a feasible explanation" for the loss of release C51·C50. An alternative is offered alongside: the surfactant "may reduce the kinetics of phase separation", and if separation in the gel layer is slower than polymer dissolution, congruent release "is anticipated" C41. The authors themselves point out that the AFM images are a snapshot after 12 h at 100% RH, "a very different timescale" from that of a dissolving compact C42.

A second route: molecular mobility

TPGS acts as a plasticizer: it lowered the Tg of the 30% DL ASD by up to 18 °C and so raises molecular mobility R5·C43. The authors propose that the higher LoC "could be related to this increase in molecular mobility". Polymer disentanglement starts only once enough solvent has entered to plasticize the matrix (refs 47, 49); a lower Tg "should theoretically" let it start earlier, with less water, so that polymer dissolution "may compete favorably" against phase separation and prevent the drug-rich surface layer C43. They support the relevance of mobility with their earlier finding that indomethacin–PVPVA ASDs had a lower LoC when the dissolution temperature was below the Tg (ref 17) R27·C45. Enhanced mobility "and/or" the favourable domain morphology "could help drive" the higher LoC C44; in the Conclusions, mobility "may also play a role" C43.

Note by the AI writer (not discussed in the paper): the drug-free control does not support this route as stated. If a lower Tg favoured polymer disentanglement and dissolution, PVPVA with 5% TPGS should dissolve faster than PVPVA alone, but it did not (about 3.2 against 3.4 mg/min/cm²); the Tg of that matrix was not measured, so the test is indirect X15.

Where do the nanoparticles come from?

The origin of the drug-rich nanoparticles, and their link to water-induced amorphous–amorphous phase separation (AAPS) in the matrix, is described as unresolved. Harmon et al. suggested that drug-rich domains form in the hydrated matrix and are released into solution as the polymer diffuses out (ref 14); Indulkar et al. suggested, from an isotope-scrambling study, that the particles form by phase separation of drug above its amorphous solubility in the aqueous phase after release (ref 13). The authors note that the two need not be mutually exclusive, since several zones of different hydration exist between the dry core and the bulk medium (ref 47) C46.

Here, comparing the domains in the AFM images with the 200–300 nm particles measured by DLS, "it becomes clear" to the authors that the film domains are larger than the particles in solution C47. The mismatch "suggests that the local environment at the time of formation and thereafter has a major impact" on morphology: separation likely occurs in different regions of the phase diagram for films and for dissolving compacts, depending on how fast the water content rises (faster in a compact than in a film exposed to vapour), and coarsening by Ostwald ripening and coalescence runs at different rates in each C48. The AFM experiments, they add, show the susceptibility of the system to phase separation and the tendency of drug-rich regions to minimize their surface area, but say nothing about separation kinetics or how they relate to release C49. It remains "unclear" whether, with surfactant, the drug first releases molecularly and then forms particles in solution, or whether particles form in the hydrated matrix and are kept from ripening by the surfactant; methods that capture the evolving ASD–water interface "are likely to be useful" C52. The lateral size of the film domains was not measured, so the size comparison is visual X7.

What it suggests for formulation

The authors' practical message is that adding an appropriate surfactant to an ASD "is a potential strategy to design high drug loading ASDs without loss in dissolution performance", and that the surfactant should be "an ASD component rather than an external excipient" C54·C35. More generally, a microstructure with discrete sub-micrometre drug-rich domains in the hydrated high-loading matrix "could potentially overcome" the loading limit of ASDs C32, a view framed by their statement that the LoC of a given drug–polymer pair "can be identified but not altered without changing the polymer" C31.

Note by the AI writer (not discussed in the paper): for formulation work three points follow from the data. The gain is large, 15% to 45% DL for felodipine–PVPVA, but at each new LoC release was about half as fast and, at 35% DL with 5% TPGS, delayed by about 15 min, so a working loading below the measured LoC is safer R12·X3. TPGS lowered the Tg by up to 18 °C, and surfactants are known to raise crystallization risk in some ASDs, but storage stability was not tested R5·C4. And because TPGS in the medium did nothing, the benefit depends on the surfactant being mixed into the dispersion itself, which favours co-processing over blending R15.

Limitations

Stated by the authors. The AFM images are a 12-hour snapshot of films exposed to water vapour, a very different timescale from a dissolving compact C42; they show that the system can phase separate and coarsen, but not the kinetics of separation or how it relates to release C49. Whether TPGS acts by slowing separation or by stopping matrix-formed particles from ripening is "unclear" C52. Most evidence on congruent release concerns PVPVA ASDs, and its generality to other polymers "is not fully established" C26.

The AI writer's own judgement.

• All structural evidence (micro-CT, AFM, nanoIR, Tg) comes from one drug loading, 30% DL. No film or tablet was examined at the new LoCs (25, 35, 45% DL) or at a failing loading with 5% or 10% TPGS, although the Abstract links discrete domains to congruent release and a contiguous phase without surfactant "at higher DLs" X5.
• The binary baseline is cited from earlier work and stated three ways; each LoC is the last passing loading on a 5% DL grid; release at the LoC is congruent but slower X1·X2·X3.
• The micelle argument uses TPGS concentrations at full release, which were not reached within the 60 min tests, and the location control used one below-CMC concentration; wetting was not measured X14·X13.
• The proposed causes, lower interfacial energy, slower coarsening, slower phase separation and higher mobility, were not measured, and the drug-free control does not support the mobility route C40·C41·C43·X15.
• Film domain sizes and heights are compared by eye; nanoIR covered 0% and 10% TPGS only X7·X18·X6.
• Replicate numbers are not stated for Tg, DLS, NTA and imaging; the mass basis of the release rates is not stated; whether the binary concentration curves were re-measured is not stated X17·X4·X12.
• One drug, one polymer and one surfactant were studied, without storage stability or in vivo data C54.

Conclusion

The paper shows, for felodipine–PVPVA, that a few percent of TPGS built into the dispersion raises the LoC in step with the TPGS level, from 15% DL without surfactant (earlier data) to 25, 35 and 45% DL with 2, 5 and 10% TPGS, with drug, polymer and TPGS released at matched rates up to each LoC and drug release collapsing one step above it R12·R9·R10·C16. Congruent release brought supersaturation and nanoparticles of about 200–300 nm; failing ASDs gave neither R11·R16·R17. The gain does not come from bulk micelles: the LoC rose even with TPGS far below its CMC of about 40–50 µg/mL, TPGS in the medium did not rescue a binary ASD, and TPGS did not speed up PVPVA alone R1·C33·R15·R14. At 30% DL, the switch from incongruent to congruent release between 2% and 5% TPGS coincides with the wet drug-rich phase changing from a contiguous network to discrete domains and with the disappearance of a porous surface layer on partly dissolved tablets R21·R23·R19. The authors' reading is that TPGS "is thought to" act by preventing a continuous drug-rich layer at the compact–water interface, probably through slower coarsening, with higher molecular mobility that "may also play a role"; these mechanisms were proposed, not measured, and the structural evidence covers one loading C53·C40·C43·X5.