Why ledipasvir–copovidone dispersions stop releasing above 5% drug loading

Background and motivation

An amorphous solid dispersion (ASD) disperses a poorly soluble drug at the molecular level in a polymer. When it dissolves, it can create supersaturation and raise the free drug concentration well above that reached by crystalline drug, which can improve absorption (Section 1, refs 2–6) C1. The simplest description of dissolution, the Noyes–Whitney equation with a Nernst–Brunner diffusion layer (Eq. 1), sets a ceiling: the rate falls to zero when the bulk concentration reaches the solubility of the dissolving form, so an amorphous drug on its own can reach at most its amorphous solubility, where membrane transport is already maximal (Raina et al. 2014, ref 7) C2·E1.

Fast ASD release can go past that ceiling. The excess drug then forms a separate, water-saturated drug-rich phase by liquid–liquid or glass–liquid phase separation (Ilevbare and Taylor 2013, ref 10; Mosquera-Giraldo and Taylor 2015, ref 11) C3. If this phase is present as nano-sized colloids, it can act as a reservoir that keeps the membrane flux at its maximum (Indulkar et al. 2016, ref 12); coupled dissolution–absorption tests and in vivo studies point the same way (refs 13–15), and the paper calls the resulting view "an emerging opinion" C3. Colloids form only when dissolution is controlled by the polymer: when the amorphous drug controls dissolution, the concentration stops at the amorphous solubility and colloids are absent (Indulkar et al. 2019, ref 16; Saboo et al. 2019, ref 17) C4.

Whether release is polymer-controlled depends on drug loading (DL). For ritonavir–copovidone (PVPVA) ASDs, drug release switched from polymer- to drug-controlled above 25 wt% DL, with a dramatic drop in rate and extent; for cilnidipine and nilvadipine in PVPVA the switch came at 15 and 10 wt% R27. This boundary is the limit of congruency (LoC): below it drug and polymer release at the same rate as the pure polymer, above it the drug usually lags C6. The LoC "appears to be highly variable" for PVPVA dispersions, and which drug properties or drug–polymer interactions set it was unclear C6. Older work on polyphase mixtures by Higuchi and co-workers had linked each component's dissolution rate to composition, interactions and phase segregation at the dissolution front forming a barrier layer, but the interface composition of dissolving ASDs had not been studied widely (refs 18, 19) C5.

The authors chose ledipasvir, a hepatitis C drug sold as an ASD, as a contrast to the low-Tg drugs studied before. It has a high molecular weight (889 g/mol), is lipophilic (log P 6.8 in the introduction) and has a very high glass transition temperature (160 °C), and its structure carries many functional groups, among them two fluorine atoms on one carbon and four NH groups (Figure 1) M1·C7. They measured drug and polymer release from ledipasvir–PVPVA ASDs as a function of DL, tested small amounts of five surfactants, examined the dissolving tablet surface by fluorescence microscopy, micro-computed tomography and X-ray photoelectron spectroscopy, and probed drug–polymer interactions by NMR (Section 1) A1·A2·A6.

Background concepts

The dissolution ceiling. The paper starts from the classic dissolution equation (Eq. 1):

Here $dM/dt$ is the dissolution rate (mass per unit time), $D$ the diffusion coefficient of the solute, $A$ the surface area of the solid, $h$ the thickness of the diffusion layer, $C_s$ the solubility of the dissolving form and $C$ the bulk concentration E1. In plain terms: dissolution is driven by the gap between the solubility and the bulk concentration, so once the solution reaches the amorphous solubility an amorphous drug stops dissolving; with $D$, $A$ and $h$ fixed, the amorphous solubility sets the rate C2. The authors use the equation only as an argument: an ASD whose release is set by a water-soluble polymer is not bound by it C29.

Congruent release and the limit of congruency. Release is called congruent when drug and polymer leave the dosage form at the same fractional rate, set by the polymer, and the LoC is the highest DL at which this still holds C6·C9. Measuring it needs two assays from the same experiment, one for the drug and one for the polymer; that is why this paper analyses PVPVA by size-exclusion chromatography alongside ledipasvir by HPLC P4.

Drug-rich colloids. When congruent release supplies drug faster than its amorphous solubility allows in solution, the excess forms sub-micron drug-rich droplets or particles that make the medium cloudy; here they were sized by dynamic light scattering (DLS) as a Z-average diameter C3·P9. For ledipasvir the amorphous solubility is given only as below 1 µg/mL (Section 4.2) or "<< 1 µg/mL" (Section 5), without a method, close to the HPLC detection limit of about 0.2 µg/mL M1·P4·X18.

Water-induced amorphous–amorphous phase separation (AAPS). A homogeneous ASD can split into drug-rich and polymer-rich amorphous domains when it takes up water, because a hydrophobic drug has little affinity for water and tends to self-associate; this has been seen at high relative humidity (Rumondor et al. 2009, ref 33), immediately on contact with water when thermodynamics allow (Purohit and Taylor 2015, ref 23), and with kinetics that depend on DL (ritonavir–PVP separated at once on immersion at 50% DL but not at 10%; Purohit and Taylor 2017, ref 24) C34·C37·R29. The tendency is thought to depend on the balance of drug, polymer and water interactions (Rumondor et al. 2011, ref 35) C37. Nile red, a dye that fluoresces more strongly in hydrophobic surroundings, collects in drug-rich domains and so marks them (refs 27, 28) C15.

Reading XPS binding energies. X-ray photoelectron spectroscopy (XPS) probes the top few nanometres of a solid. Two signals are used here. The N 1s peak differs between PVPVA and ledipasvir nitrogens, so fitting it with the spectra of the pure components gives the drug:polymer ratio at the surface, which is converted to weight percent from the chemical formulas (Figure 2S) P7·R18·E6. Fluorine occurs only in the drug; a change in its F 1s binding energy (BE) reports a change in its surroundings, and a lower BE means a more electron-rich fluorine R16·C18.

Reading the 19F NMR splitting. The two fluorines of ledipasvir sit on one prochiral carbon. When their surroundings are equivalent they give essentially one signal; when rotation is restricted, for example by contacts with other molecules, they become inequivalent and the spectrum splits into an inner pattern with outer "satellite" lines from their mutual (geminal) coupling (Figure 10) R19·C22. The size of this chemical shift difference is the authors' gauge of drug–drug association in solution C42. They used toluene because it hardly forms hydrogen bonds itself, so that drug–drug and drug–polymer hydrogen bonds can show C24.

Surfactants and their critical micelle concentration (CMC). Surfactants are common in ASD products (Kaletra contains Span 20, Kalydeco SDS) R30·C50. Above its CMC a surfactant forms micelles that can solubilize drug in the bulk solution; the authors chose amounts that stay below the reported CMCs after complete release, so that bulk solubilization should not explain any effect (refs 39, 40) C51. The paper describes the five surfactants as varying in hydrophilic–lipophilic balance (HLB) but prints no HLB or CMC values X19.

Methods in detail

The study has four strands that run in parallel and meet in one comparison: release as a function of DL and surfactant, the structure and composition of the dissolving tablet, drug–drug and drug–polymer interactions, and a common DL axis for the interaction data (Figure 16). The flow below summarizes them.

ASDs and their glass transitions

Ledipasvir and PVPVA (10 g in total) were dissolved in 200 mL ethanol and dried on a rotary evaporator with a 60 °C bath, then for 24 h in a vacuum oven at 40 °C; the powder was sieved and the 150–250 µm fraction kept (Section 3.1) P1. Binary ASDs were made at 2, 5, 7.5, 10, 20 and 50 wt% DL, and neat amorphous ledipasvir by the same route P1. For the surfactant ASDs, 500 mg surfactant was added per 10 g, that is 100 parts drug + polymer + 5 parts surfactant, so that the drug:polymer ratio stays that of the binary ASD; a "20% DL" ternary ASD is 20 parts drug, 80 parts polymer and 5 parts surfactant P1·C60. The surfactants were Span 85, Tween 80, vitamin E TPGS, poloxamer 407 and SDS M5·M6·M7·M8·M9. Glass transitions were measured by DSC on the third heating at 10 °C/min in pin-holed pans, n = 3 (Section 3.2) P2.

Release of drug and polymer from constant-area compacts

Each test used 100 mg of sieved powder pressed at 1500 psi for 60 s in an 8 mm intrinsic-dissolution die; the die turned on a paddle at 100 rpm in 100 mL of pH 6.8, 50 mM phosphate buffer at 37 °C (Section 3.3) P3. Fixing the exposed area removes particle-size and surface-area differences between formulations. Samples of 3 mL were taken at 10, 20, 30, 40, 50, 60 and 90 min and replaced with fresh medium P3. For the drug, 0.4 mL of sample was mixed with 0.6 mL ethanol so that any colloids dissolved, and analysed by HPLC at 205 nm with a detection limit of about 0.2 µg/mL; for the polymer, the rest of the sample was filtered (0.2 µm) and analysed by aqueous size-exclusion chromatography with UV detection at 205 nm (Section 3.4) P4. The pure-polymer reference was a 250 mg PVPVA compact, so its percent-release curve is normalized to a mass 2.5 times that of the ASD compacts P3·X13. The number of replicate runs is not stated X15.

Looking at the dissolving tablet

Micro-CT. Tablets were scanned before dissolution and after 10 min (5 and 10% DL) or 30 min (10% DL) at 8 µm per pixel; after dissolution they were dried overnight in a vacuum oven before the roughly 7 h scan, and the images are cross-sections through the centre (Section 3.5, Figure 5) P5·C58.

XPS. Compacts were ejected from the die before and after 10 min of dissolution and analysed with monochromatic Al Kα radiation at no fewer than three spots each; the C 1s C–H/C–C peak was set to 284.8 eV for charge correction, and spectra were fitted after a Tougaard background with model peaks from the pure compounds (Section 3.7) P7. The N 1s fit gave the surface weight fraction of drug at 20 and 50% DL; at lower DL the drug's nitrogen signal was too weak to separate R15. The F 1s signal was fitted with two components, at 687.1 and 684.0 eV R16.

Fluorescence of humidified films. Films at 5, 10 and 50% DL were spin-coated from ethanol with a trace of Nile red onto quartz slides and imaged before and after 3 days at 100% relative humidity (RH) and 37 °C, with green excitation (Section 3.6) P6. This tests whether water alone separates the drug from the polymer, but on a time scale of days rather than the minutes of a dissolution test P6·C16.

Probing hydrogen bonds by NMR

Ledipasvir's identity was checked by 1H, 13C and 19F NMR in DMSO-d6 (Figure 4S) R19·C21. All interaction work used toluene-d8 at 500 MHz (Section 3.8) P8. The 19F spectrum was recorded at 0.5, 2 and 4 mg/mL ledipasvir, and at 278, 298 and 318 K for 2 mg/mL P8. Polymers were then added to 2 mg/mL ledipasvir: 2 mg/mL PVPVA or 2 mg/mL polyvinyl acetate (PVA); PVP itself barely dissolves in toluene and could not be used R22. To isolate the vinylpyrrolidone unit, its monomer 1-vinyl-2-pyrrolidinone (VP) was titrated into eight portions of a 2 mg/mL ledipasvir stock at 0, 2, 4, 8, 16, 32, 64 and 80 mg/mL, and both 19F and 1H spectra were recorded (Figures 13, 14) P8·R23·R24.

Putting solution and solid data on one axis

To compare the NMR titration with the ASDs, each VP concentration was converted to the mass of PVPVA that would carry the same number of vinylpyrrolidone units, and from that to an equivalent DL (SI p.8). PVPVA 64 is 60 wt% VP and 40 wt% vinyl acetate (VA); with molecular weights of 111.14 and 86.09 g/mol this gives a molar ratio VP:VA of 5.4:4.6 and an average repeat unit of 99.62 g/mol E5. The PVPVA mass is then

where $c_{\mathrm{V}\mathrm{P}}$ is the VP concentration, $V$ = 0.6 mL the sample volume, $M_{\mathrm{V}\mathrm{P}}$ the molecular weight of VP and $M_{\mathrm{r}\mathrm{e}\mathrm{p}\mathrm{e}\mathrm{a}\mathrm{t}}$ = 99.62 g/mol E2. In plain terms: count the moles of VP, scale them up to moles of copolymer repeat units (VP is 54 mol% of them), and convert to mass. For 80 mg/mL VP the SI obtains 77 mg PVPVA E2. The equivalent DL is then

with $m_{\mathrm{d}\mathrm{r}\mathrm{u}\mathrm{g}}$ = 2 mg/mL × 0.6 mL = 1.2 mg, giving 1.5% for the 80 mg/mL sample E3. In plain terms: the titration sample is treated as if the monomer were the polymer, so its drug fraction can be placed on the DL axis. The SI expression uses 114.14 g/mol for VP although it states 111.14 g/mol just above; with 111.14 the 80 mg/mL sample corresponds to 79.7 mg and 1.48%, a negligible shift X4. For the XPS data the amount of fluorine at the lower BE in a 100 mg compact was computed as

where 889 g/mol is the molecular weight of ledipasvir, 2 the number of fluorines per molecule and $f_{\mathrm{l}\mathrm{o}\mathrm{w}}$ the share of F 1s at 684.0 eV E4. In plain terms: moles of drug in the compact, times two fluorines, times the fraction that sits in the electron-rich state. The SI example at 10% DL ($f_{\mathrm{l}\mathrm{o}\mathrm{w}}$ = 73.4%) prints 0.165 mmol, but its own inputs give 0.0165 mmol, which is also the value plotted in Figure 16; the printed result is a factor-10 slip E4·X3.

Experimental analysis

All ASDs are single-phase when dry

Every ASD from 2 to 50% DL showed one glass transition and no melting, which the authors take as a homogeneous dispersion of amorphous drug in polymer ("suggesting") R1·C8. The midpoint Tg rose from 107.1 °C at 2% DL to 135.5 °C at 50% DL, between PVPVA (107.5 °C) and amorphous ledipasvir (160.9 °C) (Table 1) R1·M1·M2. At 37 °C even the dry polymer-rich ASDs are about 70 °C below their Tg, so all are glasses before water enters; ledipasvir itself is a glass with a Tg about 120 °C above the test temperature R1·M1.

Release collapses between 5 and 7.5% DL

Neat amorphous ledipasvir released no detectable drug within 2 h R2. At 2 and 5% DL the drug and polymer curves almost coincide and reach about 100% by 60 min, the medium turns cloudy, and drug release quickly passes the amorphous solubility (Figure 2) R3·R4·R5. At 7.5 and 10% DL no drug is detected up to 60 min and polymer release creeps to about 17 and 23% by 90 min (Figure 2b) R3·R4. The binary 20% DL ASD released no drug and about 19% of its polymer in 90 min (Figure 7S) R11. Table A collects the values.

Table A. Percent of drug and polymer released from binary ledipasvir–PVPVA compacts, read from Figures 2 and 7S (approximate); 100 mg compacts, pH 6.8 buffer, 37 °C R3·R4·R11·R2.

The authors read the 2–5% DL behaviour as polymer-controlled and congruent ("This suggests" in the results; "clearly demonstrate" in the discussion) C9. They state the LoC as 5% DL in the results and as "between 5 and 7.5 % DL" in the abstract C10. Both describe the same bracket: 5% is the last congruent loading tested and 7.5% the first failing one C10·X7. Two details deserve care. The text says 7.5% DL gave "no detectable drug release", yet Figure 2a plots about 7% drug at 90 min for that ASD; the statement holds to 60 min R3·X6. And the comparison "similar to the dissolution rate of neat PVPVA" uses a 250 mg polymer compact whose curve lags the 2 and 5% DL ASDs at 50–60 min (about 71–81% against about 92–107%), so the two are not on the same mass basis R3·R4·X13.

Five surfactants: two help, three do not

At 7.5% DL, 5 parts Span 85, Tween 80 or vitamin E TPGS left release as poor as in the binary ASD: at most about 4% drug by 60 min and about 20–37% polymer by 90 min (Figure 1S) R6·C11. Poloxamer 407 and SDS, which the authors call the more hydrophilic surfactants, restored congruent release at 7.5% DL (Figure 3a) R7·C12. With SDS the 7.5% DL ASD released about 102% drug and about 113% polymer by 60 min; with poloxamer 407 release was congruent but slow, about 45% at 60 min and about 100% only at 110 min R7. At 10% DL SDS still gave full release (about 106% drug at 90 min), whereas poloxamer 407 failed (about 2% drug, about 17% polymer at 90 min) (Figure 3b) R8. SDS kept 20% DL congruent, slowed at 30% DL (about 60% at 60 min, about 97% at 90 min) and failed at 40% DL (about 7–8% drug) (Figure 3c) R9·C12. The conclusions sum this up as LoCs of 7.5% with poloxamer 407 and 30% with SDS, against 5% without surfactant C12·C56.

Table B. Percent released at 60 min from surfactant ASDs, read from Figures 2, 3, 1S and 6S (approximate) R3·R4·R6·R7·R8·R9·R10·C11·C12.

Three points from the discussion qualify the table. The surfactant amount was chosen so that, after full release, the medium stays below the reported CMCs, which the authors take to rule out bulk solubilization C51. From the methods, a 100 mg compact holds at most 4.8 mg surfactant, or about 0.048 mg/mL in 100 mL (a calculation by the AI writer; the CMCs themselves are not printed) X19. SDS ASDs released at about the rate of pure PVPVA, "suggesting" that SDS did not raise the LoC by speeding polymer dissolution C52. And the slow but congruent cases (poloxamer 407 at 7.5%, SDS at 30%) suggest a third regime between fast congruent and failed release C52.

Smaller colloids with surfactant

DLS sized the drug-rich colloids for three formulations that all reached 100% release: 5% DL without surfactant, 5% DL with poloxamer 407 and 7.5% DL with SDS (Figure 4) R12·C13. Over 10–60 min the Z-average was about 225–257 nm without surfactant, about 96–130 nm with poloxamer 407 and about 60–63 nm with SDS; the text rounds these to 250, 120 and 60 nm R12. The authors describe the colloids as relatively stable, with no sign of agglomeration C13. Two observations temper this: the binary and poloxamer 407 colloids grew by about 15–35% over the hour, and the three formulations did not release at the same rate (the poloxamer 407 ASD reached only about 54% drug at 60 min, against about 100% for the binary 5% DL ASD), although the text calls their dissolution rates comparable R12·R10·R3·X11·X12.

The surface turns drug-rich

Before dissolution the compact surface matched the nominal composition: 17 wt% ledipasvir for the 20% DL compact and 40 wt% for the 50% DL compact (Figure 7) R15·C17. After 10 min of dissolution the surface held 54 and 61 wt% ledipasvir R15. Over the same 10 min the 20% DL ASD released only about 2% of its polymer, which the authors take as a sign that the drug-rich layer "appears to evolve rapidly" R11·C36. (The discussion quotes the rise as "from 20% to 54%", using the nominal rather than the measured starting value X20.)

Micro-CT shows the same contrast in shape (Figure 5) R13. The 10% DL compact kept its thickness after 10 min and showed a top layer distinct from the bulk, which the authors read as hydrated; after 30 min its top face bulged and was full of pores and cracks R13·C14. For the 5% DL compact the text reports a thickness decrease after 10 min C14. Note by the AI writer (not discussed in the paper): measured against the scale bars, the 5% DL tablet in Figure 5b is about 1.3–1.8 mm thick against about 1.6 mm for the tablet before dissolution in Figure 5a, whose DL is not stated, so the thinning cannot be confirmed from the images shown R13·X14.

Water separates the drug in films above 5% DL

Before humid exposure all Nile red films were homogeneous at the microscope's resolution (Figure 6a) R14·C16. After 3 days at 100% RH and 37 °C the 5% DL film stayed mostly homogeneous, the 10% DL film showed small domains of different intensity, and the 50% DL film showed large domains of about 10–20 µm R14. The authors read this as minimal or no, some, and more extensive phase separation ("suggesting") and as support for water-induced separation of ledipasvir from PVPVA C16·C33.

PVPVA changes the environment of ledipasvir's fluorine

Neat amorphous ledipasvir shows one F 1s peak at 687.1 ± 0.1 eV; in the ASDs a second peak appears at 684.0 ± 0.2 eV, about 3 eV lower (Figure 3S) R16. Its share of the fluorine signal was 77, 73, 42 and 18% at 5, 10, 20 and 50% DL (Figure 8) R16. The authors read the shift as fluorine that has become more electron-rich through different interactions in the polymer ("indicates"), and the falling share as drug molecules that are isolated at low DL and closer to one another at high DL ("suggests") C18·C19. After 10 min of dissolution the share at 684.0 eV dropped to about 9, 3 and 2% at 10, 20 and 50% DL (Figure 9), read as drug that has left the polymer for a drug-rich environment ("suggests") R17·C20. The text gives the starting values for this comparison as 77, 73 and 42%, but Figure 9 and the paragraph before give 73, 42 and 18% for these loadings; the text list is shifted by one loading R17·X2.

NMR: drug–drug association that PVPVA breaks

In DMSO-d6 ledipasvir gives one main 19F peak, which the authors put down to free rotation ("presumably"); in toluene the signal splits (Figure 10) R19·C22. The splitting widened as the drug concentration rose from 0.5 to 4 mg/mL (the outer lines move from about 1.31 to about 1.46 ppm apart, Figure 11) and narrowed as temperature rose from 278 to 318 K (from about 1.59 to about 1.28 ppm apart, Figure 5S) R20·R21. Because an intramolecular effect would not depend on concentration, the authors take this as proof that the splitting comes from intermolecular drug–drug interactions ("confirmed"), and the temperature effect as more bond rotation C23.

Adding 2 mg/mL PVPVA reduced the splitting; 2 mg/mL PVA did not change it (Figure 12) R22. VP monomer from 2 to 80 mg/mL merged the fluorine peaks step by step, the same trend as with PVPVA (Figure 13) R23. The authors conclude that PVPVA "potentially can break" the ledipasvir–ledipasvir interaction and that the vinylpyrrolidone unit rather than vinyl acetate "may be responsible" C25. In the 1H spectrum, four NH peaks lie above 10 ppm (Figure 14) R24. With increasing VP the two smaller ones moved downfield by about 0.4 ppm (from about 10.74 and 10.56 ppm to about 11.16 and 10.97 ppm), while the two larger ones moved by about 0.1 ppm or less R24. The authors assign the two larger peaks to imidazole NH groups held in intramolecular hydrogen bonds to a carbonyl, forming a seven-membered ring (Figure 15a), and the other two to NH groups that can donate an intermolecular hydrogen bond to fluorine (Figure 15b) C26. VP, they argue, breaks these intermolecular bonds by hydrogen bonding to NH with its carbonyl, which they find "more likely" than a halogen bond or dipolar contact with fluorine C27·C39.

Table C. The interaction evidence, compiled from Figures 8 and 10–14, Figure 5S and Table 2; shifts read approximately from the spectra R16·R20·R21·R22·R23·R24·R26.

The paper's Table 2 summarizes this picture R26·C41. It also lists 1H NMR shifts with drug concentration (downfield) and temperature (upfield) for the drug–drug bond, but the paper and SI show only 19F spectra for those two variables; no 1H data at different concentrations or temperatures appear X10.

Saturation near the LoC

Figure 16 puts both interaction measures on one DL axis R25. The NMR chemical shift difference rises steeply from about 0.09 ppm at about 1.5% equivalent DL to about 0.16 ppm at about 7%, and then flattens to about 0.19 ppm at 37.5% R25. A small difference means drug–drug bonds are broken, so the authors conclude that the polymer disrupts drug–drug interaction effectively only below 7% DL C42·C43. The XPS measure, moles of fluorine at the lower BE in a 100 mg compact, climbs from about 0.0087 mmol at 5% DL to about 0.0165 mmol at 10% and then levels off (about 0.0188 at 20% and 0.0197 at 50%) R25. Both, the authors say, "suggest" that ledipasvir–PVPVA interactions saturate below about 10% DL, after which drug–drug interactions predominate C43. They note that this boundary lies close to the DL at which release switched, adding "While this may be coincidental" C45.

Note by the AI writer (not discussed in the paper): the NMR points in Figure 16 sit at about 1.5, 1.8, 2.4, 3.6, 7.0, 13 and 37.5% equivalent DL, with the line ending at 50%. Converting the VP series of Figure 13 with the SI formula gives about 38, 24, 13, 7.2, 3.7, 1.9 and 1.5%, so one plotted point (about 2.4%) has no listed counterpart, none appears at about 24%, and the 50% point may be the 2 mg/mL PVPVA sample of Figure 12 R25·X9. The shaded "congruent release" band in Figure 16 spans about 3–10% DL, which leaves out the congruent 2% DL ASD and takes in the failing 7.5% DL ASD R25·X7. And there is no XPS point between 5 and 10% DL X8.

Discussion

The authors' explanation of the cliff

The discussion joins the strands into one chain (Section 5). Ledipasvir is hydrophobic; when much water enters the ASD during dissolution, hydrophobic drug molecules tend to self-associate and separate from the polymer C37. In the dry ASD, PVPVA's vinylpyrrolidone carbonyl takes over the NH groups that otherwise bind ledipasvir molecules to one another through NH···F hydrogen bonds; this frees the fluorine (the low-BE peak) and keeps drug molecules apart C38·C39·C40. The polymer's capacity to do this saturates at a low loading C43. From there the outcome splits.

In the authors' words, faster amorphous–amorphous phase separation at higher DL "can result" in a drug-rich barrier at the dissolution front; once the surface is covered, polymer in the interior cannot release, which provides "a reasonable explanation" for the 7.5 and 10% DL results C35. The surface layer then makes dissolution drug-controlled, which for amorphous ledipasvir is too slow to detect C31. From the micro-CT images "it appears" that the layer acts as a physical barrier, even to the water-soluble polymer, and leads to swelling; the abstract and conclusions state the barrier role without hedging C32. The weak drug–polymer and strong drug–drug interactions "may contribute" to the surface drug enrichment seen at DL ≥ 10% C44. The overall reading, given in the conclusions, is that release performance "may depend on the relative kinetics between dissolution and amorphous-amorphous phase separation" and that drug–polymer interactions at different DLs "may play a role" C35·C45.

The strength of that last statement varies across the paper. The abstract says that "Evidence is provided" that drug–polymer interactions as a function of DL "play a central role in dictating the observed release behavior"; the discussion says the coincidence of the saturation boundary with the LoC "may be coincidental" and "clearly deserve[s] further investigation"; the conclusions say "may play a role" C45·X16. The structural complexity of ledipasvir, the authors add, makes it difficult to study these interactions further, but "it is reasonable to speculate" that competition between drug–drug and drug–polymer interactions, "in particular in the presence of water", "may be a key factor" in ASD dissolution C46·C47.

Why ledipasvir and not ritonavir

Other PVPVA ASDs reach much higher LoCs; ritonavir–PVPVA releases congruently up to 25% DL R27·C48. Ritonavir has a log P of 4.3, "slightly higher than for ledipasvir", and both drugs have a high molecular weight, so the authors conclude that the LoC "does not appear to depend" on hydrophobicity or molecular weight R28·C48. They point instead to the glass transition: ritonavir (Tg 50 °C, wet Tg around 0 °C) is a supercooled liquid under release conditions, whereas ledipasvir (Tg 160 °C) is a glass ("One interesting difference"), and they call for further work on how compound properties and interactions set release from copovidone ASDs R28·C49. Ledipasvir's log P is given as 6.8 in the introduction and 3.8 in the discussion, from different references; only 3.8 fits the comparison with ritonavir's 4.3 M1·X1.

How surfactants might help

Only poloxamer 407 and SDS raised the LoC C12. If release depends on a race between dissolution and phase separation, "it is possible" that these two surfactants reduced the rate or tendency of phase separation, SDS much more effectively than poloxamer 407; with 5 wt% SDS the congruency boundary moved from about 5% to about 30% DL C53. The authors say the mechanism "certainly needs more investigation" C53. They also note that the surfactant colloids were smaller, so their larger specific surface area "can potentially" improve absorption through faster replenishment from the nanodroplet reservoir (Kesisoglou et al. 2019, ref 41), and that colloid size is "definitely another factor" when designing and differentiating ASD formulations with comparable in vitro release C54.

What it means for formulation

The authors draw three practical points. A high LoC is beneficial so that dosage forms stay of reasonable size C30. Incorporating certain surfactants is "a formulation strategy to increase the LoC" C56. And the optimal release profile "may be one" that exceeds the amorphous solubility with drug-rich colloids, which requires polymer-controlled dissolution, so supersaturation alone should not be the only goal C29.

Note by the AI writer (not discussed in the paper): for a drug that is glassy at 37 °C and associates with itself through hydrogen bonds, PVPVA alone may support only a few percent of drug, and this paper shows that a small addition of SDS can change that several-fold. Because the cause of the SDS effect is untested, the result is a lead to screen for each drug rather than a rule. The study's method, assaying the polymer as well as the drug from constant-area compacts over a fine loading series, is what exposes the switch; a drug assay alone would record only "no release" at 7.5% DL.

Limitations

Limits the authors state. The match between interaction saturation and the LoC "may be coincidental" C45. The structure of ledipasvir makes further interaction studies difficult C46. The mechanism of the SDS effect "certainly needs more investigation" C53. More drug–PVPVA ASDs are needed to understand the role of drug–drug versus drug–polymer interactions C47. And factors dictating the release rate of drugs from ASDs "are not currently well understood" C57.

The AI writer's own judgement (1): proxies for the key link. The argument that interaction saturation sets the LoC rests on systems that differ from the dissolving compact: NMR of ledipasvir with VP monomer (up to 80 mg/mL) or 2 mg/mL polymer in toluene, and XPS of dry compacts X17. Neither probes the hydrated surface where phase separation and the barrier form. The XPS series has no point between 5 and 10% DL, the NMR series cannot be fully mapped to its samples, and nothing except release was measured at 7.5% DL, the first failing loading X8·X9·X21. Phase separation in the dissolving compact itself was not observed; the humid films show it over 3 days, not over the 10 minutes in which the surface enriches R14·C35. The link is therefore a coincidence of two thresholds measured in proxy systems, as the authors themselves allow C45·X16.

The AI writer's own judgement (2): the barrier. Surface enrichment is measured directly at 20 and 50% DL, but its barrier role is inferred: the CT images are of tablets dried after dissolution, one tablet per condition, and the thinning reported for the 5% DL compact cannot be seen against the before-image, whose DL is not stated R15·R13·X22·X14. The barrier picture fits the release data well, but permeability or transport through the layer was not measured C32.

The AI writer's own judgement (3): text and figure slips. Several printed values or references disagree with the data: the pre-dissolution F 1s shares in the text are shifted by one loading (Figure 9 gives 73, 42 and 18%) X2; the SI calculation for Figure 16 prints 0.165 instead of 0.0165 mmol and uses 114.14 g/mol for VP after stating 111.14 X3·X4; figure cross-references point to Figure 3b instead of 3a, Figure 2S instead of 6S, Figure 5 instead of 6, and Figure 15 instead of 4 X5; log P is 6.8 or 3.8 X1; and "no detectable drug release" at 7.5% DL sits beside about 7% plotted at 90 min X6. Table 2 lists 1H NMR trends with concentration and temperature that are not shown X10. None of these changes the main release result, but they matter for anyone reusing the numbers.

The AI writer's own judgement (4): comparisons on different bases. The pure-PVPVA reference was a 250 mg compact against 100 mg ASD compacts X13. The colloid sizes come from formulations that differ in loading and released at different rates, and two of the three size series drift upward over the hour X11·X12. Replicate numbers and the meaning of the error bars are not given for release, DLS or imaging X15. HLB and CMC values of the surfactants are not printed, so the "more hydrophilic" ranking and the below-CMC argument cannot be checked within the paper X19.

Conclusion

For ledipasvir–PVPVA compacts in pH 6.8 buffer at 37 °C, drug and polymer release congruently, with drug-rich colloids, only at 2 and 5% DL; at 7.5% and above release nearly stops, so the LoC lies between 5 and 7.5% DL R3·R4·C10. Above the LoC the compact surface becomes drug-rich within 10 minutes, which the authors read as a barrier that makes release drug-controlled R15·C31·C32. XPS of the dry ASDs and NMR in toluene show that PVPVA's vinylpyrrolidone carbonyl breaks ledipasvir's NH···F drug–drug hydrogen bonds, but that this effect saturates below about 10% DL, close to the LoC, a match the authors call possibly coincidental R16·R25·C39·C43·C45. Five parts SDS raised the LoC to 30% DL and poloxamer 407 to 7.5%, while Span 85, Tween 80 and TPGS did not; why SDS works so well is left open R7·R9·C12·C53.