How surfactants change the dissolving surface of ritonavir–PVPVA amorphous solid dispersions, and why that changes drug release

Background and motivation

Many drug candidates now reaching development dissolve poorly in water, and an amorphous solid dispersion (ASD), a molecular mixture of drug and polymer, is one of the main ways to raise their release rate and apparent concentration. For some ASDs, drug and polymer enter solution at the same normalized rate, which the field calls congruent release. The drug then exceeds its amorphous solubility because it forms a second phase of nanosized drug-rich droplets, called nanodroplets, which have improved absorption in vitro (Indulkar et al. 2016, ref 8) and in vivo (refs 9–11). This good behaviour is usually limited to low drug loading (DL). With the copolymer polyvinylpyrrolidone/vinyl acetate (PVPVA, Kollidon VA 64), a small rise in DL causes a sudden loss of release and of congruency, which the authors call the "falling-off-the-cliff" effect (Indulkar et al. 2019, ref 4; Yang et al. 2021, ref 12). The highest DL that still gives rapid congruent release is the limit of congruency (LoC). For the binary ritonavir (RTV)–PVPVA system studied here, ref 4 reported an LoC of 25% DL, only 30% of the RTV released after 90 min at 30% DL, and no release of either drug or polymer above 30% DL.

Earlier work tied this cliff to what happens at the ASD–water interface. Water can split an initially miscible ASD into two amorphous phases, a mostly insoluble drug-rich phase and a soluble phase rich in polymer and water. This is amorphous–amorphous phase separation (AAPS). Indulkar et al. (2019, ref 4) and Purohit and Taylor (2017, ref 14) linked poor release at high DL to extensive AAPS and showed that ASDs that resist drug enrichment at the surface release better. Li and Taylor (2019, ref 15) found that discrete drug-rich domains created before dissolution released faster than miscible films. Han et al. (2019, ref 16) and Yang et al. (2023, ref 17) then showed that continuous drug-rich domains at the dissolving interface harm release. For binary RTV–PVPVA, ref 17 described discrete drug-rich domains under 1 µm at low DL, pores filled with polymer and water holding discrete drug-rich domains at 30% DL, and a continuous drug-rich layer that blocked both drug and polymer at 40% DL. The authors conclude that both the kinetics and the morphology of AAPS "seem important".

The low LoC is a practical problem: the large amount of polymer needed for good release adds to pill burden (refs 18, 19). Surfactants are a promising way to raise DL without losing release. Que et al. (2019, ref 28) showed that sodium dodecyl sulfate (SDS) raised the LoC of ledipasvir–PVPVA ASDs from 7.5% to 30%. Saboo et al. (2021, ref 27) found that tocopherol polyethylene glycol succinate (TPGS) raised the LoC of felodipine–PVPVA ASDs and changed the shape of the drug-rich regions in films. Correa Soto et al. (2022, ref 24) found that SDS made clopidogrel–PVPVA ASDs more resistant to AAPS at high DL and raised their LoC. Harmon et al. (2016, ref 29) proposed that TPGS kept drug-rich nanoparticles from agglomerating.

The explanations offered for these gains vary: drug solubilization (refs 20, 33, 34), better wetting (refs 35, 36) and crystallization inhibition (refs 37, 38). The authors judge that direct experimental evidence for any of them is "generally sparse". Their own preliminary study (Yang et al. 2022, ref 39) exposed the same four RTV–PVPVA–surfactant ASDs to water vapour. The surfactants changed the kinetics and morphology of AAPS and lowered the glass transition temperature ($T_g$) of the ASDs; Span 20 and Span 85 partitioned into the RTV-rich phase after AAPS, and Tween 80 adsorbed on RTV-rich nanodroplets. But water vapour is not bulk water, and what surfactants do to the dissolving interface had not been seen. Two cautions apply: surfactants can also promote crystallization (refs 40, 41), and PVPVA ASDs with a low $T_g$ tend to have a higher LoC (Saboo et al. 2020, ref 71; ref 12), so a surfactant that plasticizes the ASD may help release for that reason alone.

The paper tests one hypothesis: surfactants that improve release do so by hindering the formation or persistence of a continuous drug-rich barrier at the ASD–water interface. The design keeps one drug (RTV) and one polymer (PVPVA) and varies the surfactant across a wide range of hydrophilic–lipophilic balance (HLB): SDS (ionic, HLB 40.0), Tween 80 (polyoxyethylene sorbitan monooleate, HLB 15.0), Span 20 (sorbitan monolaurate, HLB 8.6) and Span 85 (sorbitan trioleate, HLB 1.8). Each is added at 10 parts per 100 parts of drug plus polymer, at loadings of 30–50% DL, which lie above the binary LoC. The authors measure release of both drug and polymer and image the dissolving interface by two-dye confocal microscopy, supported by several smaller measurements. Nothing is modelled; the phase diagrams in the discussion are schematics.

Background concepts

How drug loading is counted here. DL in this paper is the RTV fraction of RTV plus PVPVA only. A ternary ASD coded "RTV-PVPVA-SDS 30–70–10" contains 30 parts RTV, 70 parts PVPVA and 10 extra parts of surfactant, so the surfactant makes up 9.1 wt% of the solids (the paper's own theoretical value) and RTV makes up about 27 wt% of the solids, not 30 wt%. The results section says the surfactant level is "10% by weight", which is the parts notation rather than the solids fraction. The PVPVA–surfactant blends used to test polymer dissolution are 70:10 by weight, so they hold 12.5 wt% surfactant, more than the ternary ASDs.

Congruent release, LoC and nanodroplets. Release is congruent when drug and polymer, each as a percentage of its own amount in the compact, enter solution at the same rate; the drug then leaves with the dissolving polymer, often as nanodroplets, so polymer dissolution controls drug release. Above the LoC the polymer typically leaves faster than the drug. Because both are measured in the same samples here, matching curves in Figure 2 mean congruent release, and a polymer curve well above the drug curve means polymer is leaving without drug.

The ternary phase diagram, miscibility gap and lever rule. A drug–polymer–water system can be drawn as a triangle with one component at each corner. A dry ASD sits on the drug–polymer edge; as water enters, its composition moves toward the water corner. The miscibility gap is the two-phase region. Its edge, the binodal, gives the compositions of the two coexisting phases, and a tie line joins each coexisting pair. The spinodal lies inside the binodal and marks where a mixture becomes unstable to any small fluctuation. The lever rule then gives how much of each phase forms: the closer the overall composition lies to one end of the tie line, the more of that phase there is. A smaller miscibility gap means more water is needed before phase separation starts.

Glass transition, vitrification line and plasticization. Water and many small additives lower $T_g$; this is plasticization. On the phase diagram, the vitrification line separates glassy compositions (too immobile to separate) from rubbery ones. The authors note that phase separation only happens below this line, where mobility is high enough for mass transfer. A surfactant that lowers $T_g$ moves the vitrification line toward lower water content, so AAPS can start at a different point and follow different tie lines. A lower $T_g$ also makes a phase more liquid-like and, in the authors' argument, mechanically weaker. HLB is an empirical scale of surfactant hydrophilicity: higher numbers mean more hydrophilic.

The gel layer. When a glassy polymer matrix meets water, a hydrated rubbery gel layer forms between the dry core and the solution. Its thickness reflects two competing rates (refs 49, 50): water penetration, which builds the gel, and chain disentanglement at the gel–solution boundary, which removes polymer. While penetration is faster the gel grows; when the rates match, the thickness stays constant.

Reading the confocal images. Two fluorescent markers are added in trace amounts. Nile red is lipophilic and marks RTV-rich regions in red; Alexa Fluor 488 is hydrophilic and marks polymer/water-rich regions in green. A uniform orange colour means the two dyes are still mixed, that is, a miscible region. Water quenches Nile red fluorescence, so the water front in real-time images appears as the boundary between the orange dry compact and a greener, less intensely red hydrated region. Two surface methods are also used: X-ray photoelectron spectroscopy (XPS) samples only the top few nanometres, while attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) samples a few micrometres, so together they distinguish an outermost film from the composition just beneath it.

Methods in detail

The work splits into four contributions: release of drug and polymer from each ternary ASD, plus PVPVA dissolution from surfactant blends; imaging of the dissolving interface after partial dissolution and in real time; measurements that separate effects on the thermodynamics of phase separation from effects on mobility, water uptake and the surface; and a reading of all of this on a schematic ternary phase diagram. The figure below shows how the parts fit together.

Contribution 1: Measuring drug and polymer release, and polymer dissolution on its own

① Problem. Earlier surfactant studies showed that surfactants can raise the LoC of PVPVA ASDs (refs 24, 27, 28), but to tie a release gain to what happens at the interface one needs, for each surfactant and DL, both the amount of drug and the amount of polymer released, and whether they leave together. A surfactant might also raise drug release in the congruent regime simply by making the polymer dissolve faster, and that effect has to be measured apart from any effect on phase separation.

② What they did. RTV and PVPVA were combined at 100 parts and 10 parts of surfactant were added. A total of 1.10 g was dissolved in 10 ml of methanol (dichloromethane, DCM, for the Span 85 ASDs), rotary-evaporated, kept under vacuum overnight at room temperature, milled and sieved to the 149–250 µm fraction, and confirmed amorphous by polarized light microscopy (PLM). Ten ternary formulations were made: Span 85 and Span 20 at 30 and 40% DL, and Tween 80 and SDS at 30, 40 and 50% DL. For release, 100 mg of powder was compressed in an 8 mm die at 1500 psi for 60 s, and the die was rotated at 100 rpm in 100 ml of 50 mM phosphate buffer, pH 6.8, at 37 °C in a Wood's intrinsic dissolution apparatus (IDR). Samples were taken from 5 to 90 min, diluted with organic solvent to dissolve any nanodroplets, and assayed for RTV by HPLC and for PVPVA by size-exclusion HPLC, in triplicate. PVPVA–surfactant blends (70:10 by weight) and neat PVPVA were made and dissolved in the same way, and their normalized release rates were compared by an unpaired Student's t-test. The normalized release rate is eq (1):

Here $R$ is the normalized release rate (mg min⁻¹ cm⁻² on the Figure 3 axis), $k$ is the slope of a linear fit of total concentration against time, $V$ is the volume of medium (100 ml), $A$ is the exposed surface area of the compact (the 8 mm die face) and $x$ is the weight fraction of the component of interest in the compact. In plain terms: the mass released per minute is divided by the dissolving area and by the component's share of the compact, so drug and polymer rates can be compared on the same footing. The paper does not say whether $x$ is computed with or without the surfactant in the total, which matters for comparing these rates with those of binary ASDs (card note X10).

③ Why it works. The rotating die keeps the exposed area constant and the shear reproducible, so rates can be compared between formulations. Measuring both components in the same aliquots shows directly whether drug leaves with the polymer or is held back. The blends contain no drug, so any change in PVPVA rate there comes from the surfactant acting on the polymer.

④ How it differs from earlier work. Earlier surfactant studies mostly reported a new LoC or an improved release profile. Here the surfactant level is fixed, all four surfactants are tested on one drug–polymer pair, and every release profile is paired with images of the same formulation's interface. Note by the AI writer (not discussed in the paper): the surfactant-free RTV–PVPVA ASD was not released in this study. The binary LoC and the 30% DL binary profile are taken from ref 4, so every "surfactant versus binary" comparison is across studies (card note X5).

Contribution 2: Imaging the dissolving interface in two complementary modes

① Problem. Release curves cannot show where the drug-rich phase forms, whether it is discrete or continuous, or whether a layer, once formed, stays in place. The earlier study of the same ASDs (ref 39) used water vapour, not the bulk water in which release happens. The authors describe the real-time confocal approach in the abstract as "newly developed".

② What they did. ASDs were prepared with Nile red at 0.01% w/w and Alexa Fluor 488 at 0.001% w/w. In the first mode, dye-containing compacts were dissolved in the IDR exactly as above for 5, 10 and 20 min, pushed out of the die with a punch, dried by gently blowing dry air over them, and imaged from the top and in cross section after cutting them in half with a razor blade. In the second, in-situ mode, a freshly pressed compact was ejected and fixed to the cover-glass bottom of a 35 mm dish (10 mm micro-well) by heating at 180 °C for 30 s. A thin layer of vacuum grease on its top face prevented wetting from that side, the dish was placed in a chamber at 37 °C, 8 ml of the same buffer was added, and images were taken every 30 s. Agitation in this mode was applied by gently expelling liquid from a pipette. Both modes used a confocal microscope with excitation at 488 nm for Alexa Fluor 488 and 561 nm for Nile red. Gel-layer thickness was measured on cross sections in ImageJ (n = 3; Table I).

③ Why it works. Because the two dyes go to different phases, phase separation shows up as separation of red and green, and the cross sections show three layers: the dry glassy compact with visible particle boundaries, a homogeneous orange gel, and a top layer where red and green separate. The two modes cover each other's weaknesses: the partially dissolved compacts went through the same hydrodynamics as the release test but are dried snapshots, while the in-situ images show the order of events in time under different flow.

④ How it differs from earlier work. The authors' earlier work applied this imaging to binary RTV–PVPVA ASDs (ref 17); here it is extended to ternary ASDs in bulk water. The authors state the main caveat themselves: in some cases the morphologies of the IDR compacts and the in-situ samples "were not identical", which they say "can be explained by" differences in hydrodynamics (lower and less consistent shear in the chamber) and by the glass substrate. They also note that released nanodroplets stick strongly to glass, which accounts for a large red disc in the last in-situ image of the Span 20 30% DL compact (Figure 4k).

Contribution 3: Separating what surfactants do to miscibility, mobility, water uptake and the surface

① Problem. A surfactant could change the interface in several ways at once: by shrinking or enlarging the miscibility gap (a thermodynamic effect), by lowering $T_g$ and raising mobility (a kinetic effect), by speeding water ingress, by accumulating at the surface, or by causing the drug to crystallize. Each needs its own measurement before the images can be interpreted.

② What they did. Water needed for AAPS. Films of the binary ASD and of the four ternary ASDs at 30% DL were spin-coated from 100 mg/ml solutions containing 0.1% w/w Nile red, heated at 100 °C for 5 min to remove heterogeneity, and dried under vacuum overnight. They were stored for 12 h in chambers over saturated salt solutions and then examined by confocal microscopy. The lowest relative humidity (RH) that produced phase separation was recorded and converted into a water content (as a percentage of the dry solid mass) using 37 °C sorption isotherms measured by dynamic vapor sorption (DVS) in 5% RH steps (Figure S5; Table II). No crystallization was seen in the films over 12 h. Water uptake rate. About 20 mg of each 30% DL ASD was melted in a DSC pan at 110 °C for 2 min to form a smooth surface of equal area, dried at 0% RH and exposed to 95% RH at 37 °C (n = 3; Figure 14). Plasticization. Amorphous RTV and RTV–Span 20 and RTV–Span 85 mixtures at 80–20 and 60–40 by weight were made by rotary evaporation from DCM, and onset $T_g$ was measured by DSC, heating from −30 to 100 °C at 5 °C/min (n = 3; Table III). The $T_g$ values of the ternary ASDs themselves were reported in ref 39. Surface composition. The SDS 40% DL compact was analysed before and after 30 min of dissolution by XPS (C 1s peak fitted with reference spectra of RTV, PVPVA and the surfactant; n = 5 on one compact; Figure 12) and by ATR-FTIR (n = 3; Figure 13). The ATR-FTIR method uses the ratio of an RTV band at 1522 cm⁻¹ to a PVPVA band at 1369 cm⁻¹, calibrated on binary ASDs from 0 to 100% DL and fitted with a second-order polynomial whose coefficients are not printed. Crystallization. Material scraped from compact surfaces after dissolution was examined by PLM in silicon oil, and dried compacts by scanning electron microscopy (SEM) at 5 kV (Figure 11, Figure S3).

③ Why it works. A film of fixed composition held at a series of humidities brackets the water content at which the composition path crosses into the two-phase region; if a surfactant raises that water content, it has shrunk the miscibility gap at that composition. The melted DVS samples all expose the same area, which the authors consider "a reasonable comparison" of water uptake rate, similar to a compact in the IDR. Because the Spans partition into the RTV-rich phase (ref 39), RTV–Span mixtures stand in for the drug-rich phase that forms during AAPS. SDS has no IR bands near 1522 and 1369 cm⁻¹ (ref 39), so the binary calibration "should apply" to the SDS ASD; the measured SDS ASD before dissolution (peak ratio about 0.79) fell on the calibration curve, which supports this.

④ How it differs from earlier work. Ref 39 observed AAPS of the same ASDs under vapour; here the onset is expressed as a water content and set beside bulk-water images. Note by the AI writer (not discussed in the paper): each threshold in Table II is a single value without a standard deviation, bracketed by the humidities tested (75, 84 and 92% RH appear), and the salts and storage temperature are not stated (card note X15). The $T_g$ data cover only the two Spans in binary mixtures with RTV at 20–40% Span, not Tween 80 or SDS, and not the drug-rich layer itself.

Contribution 4: Reading the morphologies on a schematic ternary phase diagram

① Problem. The images show discrete droplets in some formulations and continuous layers in others. To explain why DL and surfactant type move a formulation between these two outcomes, the authors need a framework that links composition to the amount and shape of the drug-rich phase.

② What they did. In the Discussion, Figure 15a draws a schematic RTV–PVPVA–water triangle with a miscible region and a two-phase region bounded by the binodal, tie lines (green dashed) and a vitrification line (yellow). A dry low-DL ASD sits at A1 on the drug–polymer edge. As water enters, its composition moves toward the water corner (red arrow). Once the path is inside the two-phase region and below the vitrification line, AAPS starts at X1, and the mixture splits into P1, a polymer/water-rich phase with very little drug, and D1, a drug-rich phase. The amounts follow the lever rule:

Here X1 is the overall composition where AAPS starts, and P1 and D1 are the two coexisting compositions at the ends of the tie line through X1. In plain terms: the phase whose end of the tie line lies closer to X1 is the more abundant one. For a low-DL ASD, X1 lies close to P1, so the drug-rich phase is a small minority and is expected to form discrete domains, as ref 17 found for the binary ASD at 20% DL. For a high-DL ASD starting at A2, the drug-rich phase is more abundant, a continuous morphology "is expected", and, if it is mechanically strong enough to resist the flow, it grows into a barrier, as ref 17 found at 40% DL. Figure 15b redraws the diagram with SDS, with a smaller two-phase region: the path from A1 may not enter it at all, and from A2 the lever rule gives a smaller drug-rich fraction. Figure 15c follows a high-DL SDS ASD from A3: AAPS at X3 gives similar amounts of D3 and P3, hence a continuous drug-rich phase, but P3 keeps taking up water and separates again into P3′ and a small amount of D3′, which forms discrete droplets inside the pores. For the kinetic route, a surfactant that lowers $T_g$ moves the vitrification line toward lower water content, so AAPS starts at a different point and follows different tie lines. The resulting causal chain is shown below.

③ Why it works. The lever rule turns a position on the diagram into a phase amount, and a phase amount into a likely morphology: a small minority phase tends to form droplets, comparable amounts tend to form a continuous network. The vitrification line puts the thermodynamic route (a smaller miscibility gap) and the kinetic route (a lower $T_g$) on one picture. The second separation in Figure 15c explains the porous layers with droplets in their pores at 50% DL, which the authors suggest "may be a transition state" between low and high DL, also seen for the binary ASD at 30% DL (ref 17).

④ How it differs from earlier work. Krummnow et al. (2022, ref 59) predicted the RTV–PVPVA–water diagram by calculation; Figure 15 is drawn by hand, and no binodal, tie line or vitrification line in it is measured or calculated. Note by the AI writer (not discussed in the paper): the diagram organizes the observations but is not independent evidence for them. Its only measured input is the water content needed for AAPS of one 30% DL composition per formulation (Table II); whether the miscible region with SDS is larger at 40 and 50% DL, or how the Spans shift the vitrification line, was not measured.

Experimental analysis

Experimental setup. All release and imaging experiments used RTV, PVPVA and 10 parts of surfactant, with compacts rotating at 100 rpm in pH 6.8 buffer at 37 °C; release profiles are means of three runs. The images are qualitative; apart from gel thickness and the 50–100 µm layer quoted for Span 20 at 40% DL, the authors did not quantify domain sizes or areas. Table A collects the release data and the interface morphology for every formulation.

Table A. Drug and polymer release and interface morphology of the ten ternary ASDs (release read from Figure 2, approximate; n = 3). The binary row is cited from ref 4; for binary ASDs above 30% DL, ref 4 reports no release of drug or polymer.

Main result 1: every surfactant helped at 30% DL, but only Tween 80 and SDS still gave partial release at 40–50% DL. Against the 30% of RTV that the binary ASD released at 30% DL in ref 4, all four ternary ASDs did better at that loading: about 85–87% from 45 min on with Span 85, about 92–97% between 30 and 60 min with Span 20, about 70% with SDS and about 50% with Tween 80. The authors' conclusion here is stated without hedge (Abstract): "All surfactants improved drug release at 30% DL". At 40% DL the Spans failed: the Span 85 compact released almost nothing of either component, and the Span 20 compact released about 8% of its RTV while about 78% of its PVPVA left. Tween 80 at 40% DL behaved like Tween 80 at 30% DL, and the normalized release rates at the two loadings were not significantly different by t-test (rate values not printed). SDS gave the fastest early release at 30% DL and then a plateau after about 30 min at every loading, at levels that fell with DL. At 50% DL, both Tween 80 and SDS released polymer much faster than drug. Several statements in the text are stronger than the figure.Card check: the text calls Span 85 at 30% DL "completely released" after 45 min, but Figure 2a levels at about 85–87% RTV (note X1); Tween 80 at 30% DL is described on p.2827 as "congruent and complete" release, while p.2822 and Figure 2e give about half the drug and about 71% of the polymer at 90 min (note X2); Span 20 at 40% DL is said to dissolve its polymer "completely", while Figure 2d gives about 78% (note X3); SDS at 30 and 40% DL is called congruent for the first 30 min, but at 40% DL Figure 2i shows RTV about 38% against PVPVA about 52% at 15 min (note X4). The figure values are used here.

Main result 2: release followed the interface. The images in Figures 4–10 line up with Table A formulation by formulation. In the Span 20 30% DL compact, a homogeneous gel (the authors' reading of no phase separation yet is "suggesting") lay under a top layer of discrete red droplets in a continuous green phase; in situ, droplets formed at the gel–solution interface and left into the bulk together with the green polymer dye, which the authors say "explains" the rapid congruent release. With Span 20 at 40% DL, a red layer about 50–100 µm thick, crossed by green pores, covered the interface from 5 min; the authors conclude that it kept the drug from being released beyond its amorphous solubility while polymer left through the pores. For Span 85 at 40% DL, the text describes (for an SI figure) a continuous network that grew into a thick layer, "accounting for" the lack of release of drug and polymer. Tween 80 at 30% DL formed discrete droplets that stuck together and entered the bulk as chunks; agglomeration "could potentially explain" the slower release. SDS at 40% DL showed a homogeneous gel with droplets only at the gel–solution boundary. At 50% DL, Tween 80 and SDS formed porous drug-rich layers whose pores held droplets; the authors explain the drug lag by drug held in the layer while droplets in the pores left with the polymer, and for SDS they infer that AAPS "appeared to" start at lower water content than at 40% DL.

Main result 3: a continuous layer that does not persist does not block release. The most informative single sequence is Span 85 at 30% DL (Figure 6). Compacts taken from the rotating IDR showed only small discrete domains at the interface and intense green polymer/water-rich domains deeper in the gel. In situ, a continuous red layer formed at the interface immediately after buffer was added; at 4 min pipette agitation removed it and released the droplets embedded beneath, and a new layer then formed. The ASD released about 86% of its RTV and essentially all of its PVPVA in the IDR. The authors attribute the difference between the two modes to shear: the rotating disk removes the layer "at a rate similar to or faster than its formation", while pipette agitation is weaker and uneven. They argue that the layer has a very low $T_g$ because Span 85 partitions into the drug-rich phase (ref 39), so it "is expected to be more liquid-like" and weaker. They also suggest that an orange background deeper in this gel marks a third, mixed phase that "appeared to" coexist with the other two (refs 46–48).Card check: the text cites Figure 6g–l, but the caption lists in-situ panels g–k only; the sequence ends at 5 min (note X7).

Table B. Probes of miscibility, water uptake and polymer dissolution. RH and water content for AAPS from Table II (single values, 12 h storage of films); water uptake is the mass gain of melted samples at 95% RH and 37 °C, read from Figure 14 (approximate, n = 3); PVPVA release rate from PVPVA–surfactant 70:10 blends without drug, read from Figure 3 (approximate, n = 3, t-test against neat PVPVA). HLB values are cited (refs 42–44).

Main result 4: SDS shifted the onset of AAPS; the other surfactants did not raise it. In Table B, the 30% DL film with SDS needed 24.5% water before it phase separated, against 14.0% for the binary film. By the card's assessment (note X15), this is the only shift larger than the spacing of the humidities tested. The authors read it as SDS increasing the miscibility of the RTV–PVPVA–water system (Abstract: "Importantly"; Results: "suggesting"). It fits the homogeneous SDS 40% DL gel, and so do the ATR-FTIR data: on an RTV plus PVPVA basis, that compact's surface held about 41% RTV before and about 40% after 30 min, so no drug enrichment built up. The authors conclude that AAPS was limited to the very surface of the gel and that this change in the miscibility gap is "likely responsible" for the better SDS release; they describe the other formulations' water contents as "similar" to the binary value.Card check: Span 20 and Tween 80 films separated at a lower humidity (75% RH) and lower water content (10.5 and 12.5%) than the binary film (84% RH, 14.0%), so these surfactants did not raise, and may have lowered, the water needed for AAPS (note X16). On this basis the authors judge that the Spans "most likely" act kinetically. The DSC data support a kinetic role: amorphous RTV had an onset $T_g$ of 46.8 °C, which fell to 14.8 and 7.7 °C with Span 20 at 80–20 and 60–40, and to 27.8 and 12.7 °C with Span 85 at the same ratios (Table III).

Main result 5: polymer dissolution and water uptake. SDS and Tween 80 sped up PVPVA dissolution from blends (about 6.9 and 5.7 against 3.6 mg min⁻¹ cm⁻²); the Spans did not. The authors suggest this "could result from" faster water uptake, plasticization and, for SDS, a soluble SDS–PVPVA complex, none of which was measured on the blends. Faster polymer release "is expected" to speed congruent drug release, and the SDS 30% DL ASD did have the highest initial drug release; Tween 80 did not follow this rule, which the authors relate to droplet agglomeration. All surfactants raised the early water uptake at 95% RH, but by 30 min the Span samples were level with the binary sample, so the claim that all surfactants increased the water ingress rate holds for the initial rate only (card note X13).

Gel-layer thickness. Table I gives the thickness at 5, 10 and 20 min (unit not printed; the images carry µm scale bars, card note X8). For Span 20 the gel stayed at 114–125 at 30% DL but thinned from 90.7 to 50.7 at 40% DL, which the authors read as "likely due to" slower water ingress behind a quickly formed barrier plus polymer loss through pores. With Span 85 at 40% DL the gel grew from 84.0 to 131.1, which the authors explain by almost no polymer leaving. Tween 80 gels grew at every DL (from 147.6 to 279.6 at 30% DL), consistent with fast water ingress and slow polymer release; SDS gels at 30 and 40% DL stayed at 93.1–113.6.Card check: the gel-layer discussion links the Span 20 30% DL steady state to Figure 2a, which shows Span 85; the Span 20 30% DL profile is Figure 2c (note X6).

Main result 6: SDS covered the surface and RTV crystallized there. XPS of the SDS 40% DL compact gave about 60 wt% SDS at the surface before dissolution (text 60, figure about 61), against 9.1% in the bulk, with RTV and PVPVA close to their 4:6 ratio; after 30 min, SDS was about 98 wt% of the surface and RTV was no longer detectable. The authors explain this by surfactant migration during preparation (ref 51) and then to the solid–liquid interface in the mobile gel. PLM and SEM of the SDS 30% DL compact surface after 30 min showed needle-shaped birefringent crystals (Figure 11); the text states that none formed with the other surfactants (Figure S3). The authors write that the plateau "can be attributed" to RTV crystallization at the surface, and that SDS accumulation "conceivably" accounts for it, "likely" through higher surface mobility (ref 85). Crystals were shown only at 30% DL and were not identified by diffraction or spectroscopy (card note X12).

Comparisons across systems and controls. The strongest internal comparison is within each Span: both release at 30% DL and fail at 40% DL, and both failures coincide with a continuous layer that stays in place. High-DL performance (SDS and Tween 80 above the Spans) does not follow HLB alone; Tween 80 and SDS also speed polymer dissolution, which the Spans do not. Controls narrow the explanations: all ASDs were amorphous by PLM before release, films showed no crystals over 12 h, and the reported absence of crystals with the other surfactants (Figure S3, text only) argues against crystallization as the cause of the Span failures at 40% DL. The blends separate polymer dissolution from phase separation. The binary ASD itself is not controlled: its release (ref 4) and gel morphology (ref 17) come from earlier studies.

Discussion

Where the picture applies. The study extends the binary framework of refs 16, 17 and 60 to ternary ASDs: release depends on whether the drug-rich phase at the gel–solution interface is absent, discrete, or continuous and persistent. The authors find that "for all cases where drug release was observed, no persistent continuous drug-rich barrier was formed", so preventing or disrupting it "appears to be an important mechanism", and they "propose" it is the most critical surfactant effect for release. The evidence covers one drug, one polymer, one surfactant level, one buffer and one rotation speed, at loadings just above the binary LoC.

Two routes, with different dependencies. The authors' Table IV lists positive and negative factors for each surfactant: barrier disruption by plasticization and faster water ingress for both Spans, plus a weaker drug-rich layer for Span 85; faster polymer dissolution and water ingress but droplet agglomeration for Tween 80; higher miscibility, faster polymer dissolution and water ingress, but crystallization, for SDS. The SDS route acts before the layer forms, by reducing how much drug-rich phase appears. The Span route acts after it forms, by making it easier to remove. Note by the AI writer (not discussed in the paper): the second route depends on the shear at the surface, as the Span 85 30% DL images show, so its benefit may differ between a rotating disk, other dissolution tests and the gut.

What remains open. The authors state that further studies are needed to separate "the relative importance of each of the myriad of effects" of surfactants. Beyond that, the softness of the layer was inferred, the water needed for AAPS was measured only at 30% DL, and the three-phase coexistence suggested for Span 85 rests on colour alone.

Implications for formulation. The authors' own recommendations are that surfactants can raise the DL at which a PVPVA ASD still releases well by making a continuous barrier less likely, and that surfactants must be chosen carefully because they can promote crystallization. The AI writer's reading (not stated in the paper): the gain from 10 parts of surfactant is modest here, since the Spans rescued 30% but not 40% DL, and SDS and Tween 80 gave only partial release at 40–50% DL. Part of the 30% DL gain may come from dilution, since a ternary "30% DL" ASD holds 30 of 110 parts RTV (about 27 wt% of solids), close to the binary LoC of 25%. SDS, the surfactant that kept release at the highest loadings, also brought surface crystallization, so pairing it with a crystallization inhibitor is the obvious next question. Imaging the interface under realistic flow is a useful screen, because the same Span 85 ASD formed a barrier in a quiet chamber and none under rotating-disk shear.

Limitations

Limitations stated by the authors. The authors note that the IDR and in-situ morphologies were not always identical because of differences in hydrodynamics and the glass substrate, that in the in-situ images "other pores might still exist outside this field of view", and that the water contents for AAPS are estimates read from sorption profiles ("Based on this estimation"). They also state that further work is needed to separate the many surfactant effects, that surfactant effects on crystallization are system specific, and that the physical stability of surfactant-containing ASDs during storage must be considered.

The AI writer's own judgement. The card's overall assessment is "moderate". The link between a persistent continuous layer and failed release is consistent across ten formulations and two imaging modes, but several points weaken the conclusions. First, the surfactant-free comparator is cited (ref 4 for release, ref 17 for gel morphology), so every comparison with the binary ASD is across studies (card notes X5, X18). Second, the text overstates the release of four formulations relative to Figure 2 (notes X1–X4). Third, plasticization of the barrier is an inference: it rests on $T_g$ of RTV–Span mixtures at 20–40% Span and on partitioning cited from ref 39, while the layer formed in situ was not characterized, and Tween 80 and SDS were not tested in Table III (note X17). Fourth, each AAPS threshold is a single value bracketed by humidity steps; only the SDS shift exceeds that resolution, and Span 20 and Tween 80 separated at less water than the binary film (notes X15, X16). Fifth, crystallization was shown only at 30% DL with SDS and without identification of the crystals, so its role in the 40 and 50% DL plateaus is assumed (note X12). Sixth, the Span 85 and Tween 80 40% DL morphologies and the crystal check for the other surfactants rest on text descriptions of SI figures that were not available, and XPS covers only the SDS 40% DL compact (note X11). Finally, the evidence is correlational: no experiment changed the layer while holding everything else fixed.

Conclusion

The paper's release data and confocal images of the dissolving interface support the view that surfactants improve release from ritonavir–PVPVA ASDs mainly when they keep a continuous drug-rich layer from forming or from staying at the surface. All four surfactants helped at 30% DL; only SDS and Tween 80 still gave partial release at 40–50% DL. The authors identify two routes: SDS raised the water needed for AAPS (24.5 against 14.0% for the binary film), so the gel dissolved largely without phase separating; the Spans, which lowered the $T_g$ of RTV to 7.7–27.8 °C in mixtures, are thought to make the layer soft enough for flow to remove it. Against these gains, Tween 80 was linked to droplet agglomeration, and SDS to RTV crystals at the surface, to which the authors attribute the release plateau after about 30 min.

The barrier framework is well supported as a correlation within this system; the two mechanisms behind it are plausible but rest on indirect, single-point measurements and a comparator from earlier studies. Tests against a binary ASD run in parallel, at other surfactant levels and under other flow would show how far it can guide surfactant selection.