How a crystalline co-drug lowers the supersaturation and membrane flux of an amorphous drug: atazanavir with darunavir

Background and motivation

Many chronic diseases, HIV infection among them, are treated with several drugs at once, either as separate dosage forms or as one combination product, and each drug may be crystalline or amorphous in that product (Section 1, Section 4). Combination therapy brings known clinical risks such as drug–drug interactions and altered pharmacokinetics. A less explored question is how co-administered drugs behave together in solution, before absorption (Dengale et al., 2016; Wilkins et al., 2024).

For amorphous drugs this question already has a partial answer. Studies of amorphous drug pairs found that combining or co-administering two amorphous drugs can lower the supersaturation and the membrane transport of both (Alhalaweh et al., 2016; Trasi and Taylor, 2015a; Nguyen and Van den Mooter, 2014). The explanation is thermodynamic: for poorly water-soluble drugs whose supercooled liquid or glassy forms mix with each other, the drug-rich phase that separates from water contains both drugs, and mixing lowers the maximum chemical potential of each below that of the neat amorphous drug (Trasi and Taylor, 2015b). C7·C32 When mixing is unfavourable in the presence of water, for example an ionized drug paired with an un-ionized one, each drug keeps its own amorphous solubility. C8 For ritonavir with darunavir, co-administration lowered the supersaturation and release of both drugs whatever polymer was used, and NMR found no direct drug–drug interaction in bulk solution; that study suggested the losses "could be attributable, at least in part" to solvent-mediated processes. R18·C35

Two gaps motivated the present study. First, these behaviours have been mapped for only a few binary combinations, and the effect of a crystalline drug given alongside an amorphous drug on the amorphous drug's solubility and achievable supersaturation "has not been thoroughly investigated" (Laitinen et al., 2017). C9 Second, earlier model drugs had low crystalline solubility, so the partner drug could hardly be studied in undersaturated solution. Darunavir hydrate dissolves to 265.6 µg/mL at pH 6.8 and 37 °C, which lets the authors hold darunavir below, at or above its saturation next to supersaturated atazanavir. C13·R3

The two model drugs are HIV protease inhibitors (Figure 1). Both are weak bases, atazanavir with pKa 4.3 and log P 4.5, darunavir with pKa 2.4 and log P 2.8 (Table 1, cited from Wuyts et al., 2013), so both are un-ionized at pH 6.8 and differ mainly in lipophilicity and solubility. M1·M2·C20·C18 Atazanavir was used as an amorphous solid dispersion (ASD) with polyvinylpyrrolidone (PVP K29/32), darunavir as its crystalline hydrate. The stated aim is to find how co-adding the two drugs in different solid forms affects their dissolution, supersaturation and membrane transport rate. P2·P1·A1·A2·A5

Background concepts

Supersaturation and the amorphous solubility. A crystalline drug in water dissolves up to its crystalline solubility. An amorphous form, or a solution made by adding a concentrated drug solution to water (antisolvent addition), can exceed that level: the solution is then supersaturated. There is an upper limit. Above the amorphous solubility, the excess drug separates as a second, drug-rich liquid or glassy phase, in the form of colloidal droplets; this is liquid–liquid phase separation (LLPS). The solution that remains is the drug-lean phase, and the amorphous solubility is the highest concentration of freely dissolved drug it can hold (Raina et al., 2014). C11 In this paper the drug-rich phase is also called the "colloidal phase" or "colloidal precipitate", and after ultracentrifugation only the drug-lean phase is analysed. P9·C3

Flux follows free drug. Drug crossing a membrane is described first by Fick's first law (Eq. 1):

Here $J$ is the flux (amount per unit membrane area per unit time), $D$ the diffusion coefficient of the drug in the medium and $dC/dx$ the concentration gradient across the membrane. In plain terms: flux grows with the concentration difference across the membrane. The authors note that this holds for ideal solutions at steady state, without interactions that change solution thermodynamics. E1 For non-ideal solutions they use a second form (Eq. 2, after Hubatsch et al., 2007):

Here $dM/dt$ is the mass crossing per unit time, $A$ the membrane area, $a$ the thermodynamic activity of the dissolved drug, $h$ the membrane thickness and ${\gamma }_m$ the activity coefficient of the drug in the membrane. In plain terms: with the membrane terms fixed, flux is proportional to the activity of freely dissolved drug, not to the total drug present. Drug held in micelles, colloids or a separated drug-rich phase does not add to flux. E2·C10

From this follows the picture in Scheme 1. As drug is added, flux rises linearly until the solution reaches the solubility of the solid in equilibrium with it: the crystalline solubility, or, if crystallization is avoided, the amorphous solubility. Beyond that point the added drug goes into the new phase and flux stays level. C11 A flux plateau therefore marks the free-drug ceiling, and the paper uses it twice: to measure darunavir's amorphous solubility, and to read the ceiling of atazanavir with and without darunavir. R5·R9

Why a second drug can lower the ceiling. If a second component dissolves in the drug-rich phase, the first drug is diluted there. Its chemical potential in that phase, and therefore in the solution that is in equilibrium with it, falls, by an amount that depends on the composition of the drug-rich phase (Alhalaweh et al., 2016). C12 The paper states this in words and prints no equation for it. The consequence is a lower maximum free concentration and a lower flux plateau, even though nothing has crystallized.

Measuring the amorphous solubility. The authors use two methods. In antisolvent titration, a 10 mg/mL methanol solution of the drug is infused into stirred buffer, and the concentration at which the extinction at 450 nm (a wavelength the drugs do not absorb) starts to rise marks the formation of scattering droplets, i.e. LLPS (Almeida e Sousa et al., 2015; Ilevbare and Taylor, 2013). P7·C16 In the diffusive-flux method, drug is added stepwise to the donor of a diffusion cell and the concentration at which flux stops rising is noted (Trasi and Taylor, 2015a). P10·C22

Miscibility in the dry state. Two amorphous components that mix fully give one glass transition temperature ($T_g$) in differential scanning calorimetry (DSC); two separate phases give two $T_g$ values. Nano thermal analysis (nanoTA) does the same locally: a heated atomic force microscope tip is placed on a micrometre-sized region, and the temperature at which the tip sinks into the softening surface is recorded as a local glass transition (Li and Taylor, 2016). C17·P12·P13

Methods in detail

The study asks one question in four set-ups of increasing biological relevance, and then checks whether the two drugs mix when dry. All solution work was done in 50 mM sodium phosphate buffer, pH 6.8, at 37 °C, except the Caco-2 study, which used Hank's balanced salt solution with HEPES (HBSS) at pH 6.5. P6·P8·P9·P10·P11

Solid forms and solubility

Darunavir, supplied amorphous, was turned into its crystalline hydrate by the method of Phull et al. (2016): about 500 mg was stirred in 5 mL water:methanol (1:1 v/v) for 1 h, the resulting cake was ground and stirred in water for 24 h, then filtered and dried under vacuum at room temperature. P1 The atazanavir ASD held 10% w/w drug in 90% w/w PVP and was made by dissolving 1 g of solids in 5 mL methanol and removing the solvent on a rotary evaporator (55 °C bath), followed by 24 h under vacuum; two batches were made and described as amorphous by DSC and powder X-ray diffraction (PXRD). P2·R2 Crystalline atazanavir matched the known free base crystalline form, darunavir hydrate showed its characteristic peaks at 7.1°, 13.8° and 16.7° 2θ, and the amorphous drugs gave halo patterns (Figures S1, S2, not available here). R1·C19 Note by the AI writer (not discussed in the paper): the materials section says atazanavir was obtained as atazanavir sulfate, while every later statement and the molecular weight in Table 1 (704.9 g/mol) refer to the free base; no conversion step is described. M1·X4

Crystalline solubility was measured by shaking 10 mg of drug in 15 mL buffer for 48 h at 37 °C, separating the solid by ultracentrifugation (35000 rpm, 30 min) and assaying the supernatant by HPLC, in triplicate. P6 Amorphous solubility came from antisolvent titration to the light-scattering onset, with at least four runs per drug and less than 1% methanol at the end, which did not change the solubility of either drug significantly. P7·R4 Darunavir's amorphous solubility was also checked by the diffusive-flux method. R5

Non-sink dissolution

About 100 mg of ASD powder (10 mg atazanavir), 10 mg of darunavir hydrate, or both together (a 1:1 drug weight ratio) were placed in glass vials with 15 mL buffer, stirred at 300 rpm at 37 °C, without any polymer in the medium. The atazanavir dose corresponds to at least 7 times its amorphous solubility and the darunavir dose to about 2.5 times its equilibrium solubility. Samples were ultracentrifuged (35000 rpm, 45 min) and the supernatant assayed by HPLC; three independent experiments were run per time point. P8 The methods list sampling at 5, 15, 30, 60 and 120 min, but Figure 3 shows 15–120 min only. X12

The maximum achievable concentration with darunavir present

Here supersaturation was created by antisolvent addition in buffer containing 400 µg/mL of hydroxypropyl methylcellulose acetate succinate (HPMCAS, MF grade), which the authors added "to stabilize all systems against crystallization". P9·C14 200 µL of an 11 mg/mL atazanavir methanol stock was added to 15 mL buffer (a fixed 150 µg/mL atazanavir, per the Figure 5 caption), followed by 25–300 µL of a 30 mg/mL darunavir stock, giving darunavir additions up to about 579 µg/mL. The drug-rich phase was spun down (35000 rpm, 45 min) and both drugs were measured in the supernatant. P9·R11 The number of replicates and the time between addition and centrifugation are not printed; Figure 5 shows three points per level. P9

Side-by-side diffusion cell

The diffusion cell (PermeGear) had two 34 mL compartments joined by a 30 mm orifice and separated by a regenerated cellulose membrane with a 6–8 kDa molecular weight cut-off. The donor held 30 mL buffer with 400 µg/mL HPMCAS, stirred at 300 rpm "to minimize the thickness of the unstirred water layer", at 37 °C; the receiver held the same buffer with HPMCAS and was sampled for HPLC, in triplicate. P10·C15 Atazanavir (10 mg/mL) or darunavir (31 mg/mL) methanol stocks were added in varying volumes. For the combination, the atazanavir stock was added to a donor holding a darunavir hydrate suspension (0.5 mg/mL) that had been equilibrated for 36 h at 37 °C. P10 The sampling times, the way flux was calculated from receiver concentrations, and whether the suspension donor also contained HPMCAS are not printed. X13

Caco-2 monolayers

Three donor solutions in HBSS pH 6.5 were applied to the apical side: a darunavir suspension (excess hydrate at 0.5 mg/mL, stirred 48 h at 37 °C), the same suspension with 30 µL of a 10 mg/mL atazanavir stock added to 3 mL (about 100 µg/mL atazanavir plus saturated darunavir), and atazanavir alone at 100 µg/mL, "which is above the amorphous solubility of atazanavir". All donors contained about 1% methanol. The basolateral side was at pH 7.4, inserts were shaken at 500 rpm at 37 °C, and 600 µL receiver samples were taken at 15, 30 and 60 min and replaced with fresh buffer; n = 3, repeated twice. Monolayer integrity was checked with [¹⁴C]-mannitol. P11·R14 Note by the AI writer (not discussed in the paper): the 100 µg/mL donor is called supersaturated on the basis of the amorphous solubility measured in pH 6.8 phosphate buffer, not in HBSS, and no crystallization inhibitor is mentioned for these donors. X10

Miscibility of the dry amorphous drugs

Physical mixtures of amorphous atazanavir and amorphous darunavir at darunavir mole fractions of 0.3, 0.5, 0.7 and 0.9 were heated in the DSC to 108 °C, held for 2 min, cooled to −30 °C and reheated to 220 °C at 20 °C/min, and the $T_g$ values on the second heating were recorded. P12 For nanoTA, the two drugs were melted together between cover slides on a hot stage and quenched; the film was amorphous under polarized light. A heated probe (5 °C/s from room temperature) was placed at several points inside two 5 µm × 5 µm regions, at least 5 ramps per region, and the local maximum of each deflection curve was taken as the local softening temperature. P13·R16 The drug ratio of this film is not printed. X8

Statistics: unpaired two-tailed t-tests compared single- and two-drug arms; a one-way ANOVA compared diffusive fluxes; p < 0.05 counted as significant. P14

Experimental analysis

The two drugs: a large crystalline gap, a small amorphous one

Table A. Properties of the two drugs, from Table 1 (pKa and log P cited from Wuyts et al., 2013; $T_g$ by DSC) and Table 2 (pH 6.8, 37 °C, mean ± SD, n = 3). M1·M2·R3

Both solubilities of darunavir are higher than those of atazanavir, and the amorphous solubility of atazanavir is "in good agreement" with the earlier report (Alhalaweh et al., 2016). C21 The contrast that matters for the design is the gain from amorphization: Table 2 gives atazanavir 93.1 against 1.6 µg/mL, but darunavir only 363.8 against 265.6 µg/mL. R3·C34 The discussion states, from Table S3, that darunavir hydrate is 220 times more soluble than crystalline atazanavir on a molar basis, that the amorphous forms differ about 5-fold, and that amorphization raises solubility "approximately 1.30 times" for darunavir and "almost 82-fold" for atazanavir. R6 The 5-fold molar ratio follows from Table 2 and the molecular weights, but the other ratios do not: Table 2 gives 1.37 and 58, and a molar ratio of 214. Table S3 was not available, so this reading cannot tell whether it holds different values; the Table 2 values are used here. X1 The authors attribute the large crystalline solubility gap to differences in crystal packing and to the lower log P of darunavir ("can be attributed to"), and they did not measure either. C33 The Table 2 caption also describes both columns as "determined by the anti-solvent method", whereas the methods describe a shake-flask procedure for the crystalline values. X2

Darunavir's flux in the diffusion cell rose from about 2.8 × 10⁻⁴ µg·min⁻¹·cm⁻² at about 51 µg/mL to about 8.3 × 10⁻⁴ at about 310 µg/mL, and stayed at about 8.2 and 9.3 × 10⁻⁴ at about 410 and 513 µg/mL; the last three points did not differ (p > 0.05), which the authors read as "consistent with achieving saturation with respect to the amorphous form" (Figure 2). R5·C22 Note by the AI writer (not discussed in the paper): the plateau already includes the point at about 310 µg/mL, below the antisolvent value of 363.8 µg/mL, the two lowest points give the same flux, and no flux-based number is given, so the agreement of the two methods is qualitative. X11

Finding 1: darunavir hydrate lowers atazanavir release from the ASD, and atazanavir lowers darunavir

Table B. Dissolved drug (µg/mL) in non-sink dissolution, pH 6.8, 37 °C, no polymer in the medium; read from Figure 3 (approximate; means of three). R7·R8

The ASD alone released atazanavir at once and held about 81–88 µg/mL over 2 h, which the authors call "comparable" to the amorphous solubility of 93.1 µg/mL and, in the discussion, "matching" it, in line with earlier work. R7·R3·C23 With darunavir hydrate in the same vial, atazanavir was lower at every time point (p < 0.05) and 31% lower at 2 h. R7·C25 Darunavir hydrate alone dissolved slowly and stayed below its equilibrium solubility of 265.6 µg/mL for the whole 2 h, reaching about 227 µg/mL; from the mixture darunavir reached only about 119 µg/mL, about 48% lower. R8·R3·C24·C25 In the authors' words, the advantage of the ASD "was diminished" and, "Interestingly", the lower atazanavir concentration "was associated with" less darunavir dissolution. C25 The loss runs both ways; the authors' explanation for both comes in the discussion below.

Finding 2: with darunavir present, atazanavir flux levels off early and low

Table C. Atazanavir flux across the regenerated cellulose membrane (× 10⁻⁵ µg·min⁻¹·cm⁻²), pH 6.8 buffer with 400 µg/mL HPMCAS, 37 °C; read from Figure 4 (approximate; means of three). R9

Alone, atazanavir flux rose linearly and "appeared to reach a plateau" over 75–150 µg/mL added. Next to a darunavir hydrate suspension, the maximum flux was 30% of that without darunavir, and the plateau began at 37 µg/mL instead of about 75–100 µg/mL. R9·C26 The authors read this as LLPS of atazanavir occurring "at lower supersaturation" when darunavir is present ("indicating"). C2 Darunavir's own flux did not change significantly when atazanavir was added at different concentrations (p > 0.05; Figure S3). R10·C27 Note by the AI writer (not discussed in the paper): the discussion calls these "Flux studies of the atazanavir-PVP ASDs", but the methods and Figure 4 describe atazanavir added from a methanol stock; no membrane experiment in the paper used the ASD itself. X3

Finding 3: every step of added darunavir lowers the atazanavir ceiling

Table D. Drug left in the drug-lean phase after ultracentrifugation, with atazanavir fixed at 150 µg/mL, in pH 6.8 buffer with 400 µg/mL HPMCAS at 37 °C; ranges of the individual points read from Figure 5 (approximate). R11

Without darunavir, atazanavir in the supernatant was similar to its amorphous solubility. Each addition of darunavir lowered it, down to about 22–25 µg/mL at the highest darunavir level. R11·C1 The first three additions (up to about 196 µg/mL) are below darunavir's hydrate solubility (dashed box in Figure 5), and atazanavir had already fallen to about 61–63 µg/mL there: darunavir did not need to be saturated, let alone to form its own drug-rich phase, to lower the atazanavir ceiling. R11·C1 Darunavir in the supernatant was always below the amount added, also below its own amorphous solubility; the authors take the missing darunavir to have left solution with the atazanavir drug-rich phase. R11·R3·C3 The authors' reading is that "there is some miscibility between the two drugs in the drug-rich phase in the aqueous environment studied" ("suggest"). C4

Figure 9 replots the same data against the "mole fraction of darunavir in drug-rich phase": atazanavir falls from about 91–96 µg/mL at a mole fraction of 0 to about 22–24 µg/mL at about 0.84, which the authors present as the amorphous solubility of atazanavir decreasing as darunavir in the drug-rich phase increases. R12·C39 Note by the AI writer (not discussed in the paper): the paper does not say how this mole fraction was obtained. Recomputed from the Figure 5 readings and the molecular weights, the plotted values (about 0.33 at the lowest darunavir level to about 0.84 at the highest) are close to the darunavir share of all drug added (about 0.30 to 0.83), while a mass balance of the drug that left solution gives a lower darunavir share (about 0.13 to 0.75). The figure therefore shows the trend, but its composition axis should not be used for quantitative mixing calculations without re-derivation. X5

Finding 4: in Caco-2 cells atazanavir flux drops by about a third; darunavir's does not

Table E. Apical-to-basolateral flux across Caco-2 monolayers over 60 min at 37 °C; read from Figure 6 (approximate; means, n = 3 repeated twice). R13

Across Caco-2 monolayers, atazanavir from the 100 µg/mL donor crossed about 36% more slowly when darunavir suspension was present. In the results the authors write that darunavir "may negatively impact" atazanavir transport; in the conclusions the cell data are "confirming that darunavir negatively impacted the membrane transport rate of atazanavir". R13·C28 Darunavir flux was only slightly higher from the pure suspension than from the mixture, and the authors conclude that atazanavir "has little to no effect on darunavir transport". R13·C29 Note by the AI writer (not discussed in the paper): the Caco-2 fluxes are about a thousand times the diffusion-cell fluxes although both are printed in µg·min⁻¹·cm⁻², and the media differ, so the two set-ups can be compared only through the relative changes they show. X9

Finding 5: dry, the two amorphous drugs stay apart

Table F. Glass transitions of dry amorphous atazanavir–darunavir physical mixtures by DSC (20 °C/min); read from Figure 7 (approximate). R15

Every mixture showed two glass transitions, which the authors describe as corresponding to those of atazanavir (105 ± 1.0 °C) and darunavir (62 ± 0.8 °C): "This suggests the drugs are phase-separated in the dry state". R15·C30 A melt-quenched film of both drugs looked heterogeneous under the optical microscope (Figure 8A). nanoTA gave softening at 118 ± 4 °C in one region, assigned to atazanavir, and at 92 ± 2 °C in another, assigned to darunavir, each region uniform within itself (Figure 8B; Table S2). R16·R17 The authors conclude that the visual heterogeneity is phase separation and that the amorphous forms "do not mix in the binary system" ("further suggest"). C31 Note by the AI writer (not discussed in the paper): the lower $T_g$ in the mixtures sits about 5–9 °C above that of neat darunavir, which leaves room for a small amount of atazanavir in the darunavir-rich phase; and the nanoTA temperatures, taken at 5 °C/s, lie 13 and 30 °C above the DSC values and identify each region only by its softening temperature. Both observations still support two phases. X7·X8

How the authors connect the findings

The chain starts from the frame of Scheme 1: the free concentration of a drug, and so its flux, stops rising where a drug-rich phase separates. C10·C11 Earlier work showed that drugs miscible in the solid state co-precipitate into one drug-rich phase and lower each other's ceiling by co-dilution. C32 Here the authors propose that darunavir, although it is crystalline in the dosage and does not mix with atazanavir when dry, partitions into the atazanavir drug-rich phase in water. C4 The reduced supersaturation and flux of atazanavir "can be attributed to dilution of atazanavir in the colloidal precipitate", the drug-rich phase, caused by darunavir partitioning, which lowers atazanavir's chemical potential; this mixing "is key" to the lower maximum concentration and flux; the conclusions put it more cautiously, "likely due to the partitioning of darunavir into the colloidal phase of atazanavir". C37 The same step explains the lower darunavir concentrations: darunavir that partitions into the colloid is removed with it by centrifugation, so less is measured in solution, and the apparently slower darunavir dissolution from the ASD–hydrate mixture "can be explained by" this partitioning. C3·R8·R11 Darunavir's own membrane flux did not change, which the authors note was measured with a crystalline darunavir suspension present. C29·R10·R13

The authors rule out a role of the polymers. HPMCAS and PVP have both been reported not to lower supersaturation at the grades and concentrations used, so they are "not expected" to reduce atazanavir supersaturation, and because they do not take part in the drug-rich phase they are "unlikely" to help the drugs mix there (citing Alhalaweh et al., 2016; El Sayed et al., 2024). C38 Note by the AI writer (not discussed in the paper): the data give indirect support; with HPMCAS present and no darunavir, atazanavir in the supernatant was about 91–96 µg/mL (Figure 5), close to its amorphous solubility of 93.1 µg/mL measured without polymer. R11·R3 Polymer content of the drug-rich phase was not measured. C38

The last link is the role of water. Because DSC and nanoTA showed the dry amorphous drugs to be phase separated, while in water the drugs leave solution together, "it is apparent that water facilitates miscibility between atazanavir and darunavir in the colloidal phase"; the abstract puts it as "This suggests", the conclusions as "water appeared to promote the mixing". C5·C30·C31 The proposed reason is "likely due to an enhanced entropy contribution" from having three components: atazanavir, darunavir and water. C40 No measurement in the paper tests this entropy argument, and the composition and water content of the hydrated drug-rich phase were not determined. X6

Discussion

What the paper establishes. Across four independent set-ups, darunavir lowered atazanavir in solution or atazanavir flux: by 31% at 2 h in ASD dissolution, to 30% of the maximum flux in the diffusion cell, by about 36% across Caco-2 cells, and step by step in the supersaturation test, including at darunavir levels below its own saturation. R7·R9·R13·R11 The direction is the same in every set-up, and darunavir's transport was not affected. R10·R13·C29 What the paper does not establish directly is the mechanism: the partitioning of darunavir into the drug-rich phase and the water-promoted mixing are inferred from the depletion of both drugs from solution and from the flux plateaus. X6

Relation to earlier work. The results extend the amorphous–amorphous picture (Trasi and Taylor, 2015a, 2015b; Alhalaweh et al., 2016) to a case where only one drug is amorphous. For ritonavir with darunavir, Nguyen and Van den Mooter (2014) also saw mutual losses and, having excluded direct interactions in bulk solution by NMR, pointed to solvent-mediated processes; the present paper names a specific route, mixing in the drug-rich phase. R18·C35·C37 The authors present the contrast between dry-state immiscibility (the drugs "were found to remain phase separated in the dry state") and partitioning in water as a new observation. C4·C30

Implications the authors draw. Co-administration of drugs, even in different solid forms, "must be carefully considered", because lower supersaturation "may compromise bioavailability" and therapeutic outcomes. C42 They add that the slower apparent dissolution of darunavir "could conceivably impact darunavir absorption" given its time in the gastrointestinal tract. C36 No in vivo or biorelevant-media data were collected to test either statement. C42

Note by the AI writer (not discussed in the paper): a rough check of the size of the effect. When darunavir crystals are present and in equilibrium, darunavir's activity relative to its amorphous form is fixed at about its crystalline-to-amorphous solubility ratio, 265.6/363.8, about 0.73. R3 If the two drugs mixed ideally in the drug-rich phase and that phase held little water, atazanavir would make up roughly the remaining quarter of it, and its ceiling would fall to roughly a quarter of 93.1 µg/mL. R3 The observed plateau with the darunavir suspension, 30% of the control flux from 37 µg/mL added, is of the same size. R9 This estimate is the AI writer's, uses assumptions the paper does not make, and is offered only to show that the authors' dilution mechanism can account for the magnitude of the loss.

Note by the AI writer (not discussed in the paper): what a formulator can take from it. First, dry-state miscibility screening (DSC, nanoTA) did not predict the solution behaviour of this pair; for drug combinations the maximum achievable concentration of each drug measured in the presence of the other is the direct test. R15·R17·R11 Second, a crystalline partner is not inert when it is soluble: the part of it that dissolves can lower the ceiling of the amorphous drug, and a more soluble partner brings more drug into solution. R11·R3 Third, which drug gains from amorphization matters: in this pair amorphization raises atazanavir's solubility far more than darunavir's (Table 2), and that is the advantage the crystalline partner erodes. R3·C34

Limitations

Limits the authors state. The authors do not list limitations of their own study. In the introduction they note that such solution behaviour has been studied for "a limited number of binary drug combinations", which applies to this single pair as well. C9

The AI writer's own judgement (1): the mechanism is inferred. The drug-rich phase was never isolated or analysed, so its darunavir, atazanavir and water content are not known; the LLPS onset of atazanavir with darunavir present was not measured by light scattering, although the antisolvent method used for Table 2 could have done so. X6 The composition axis of Figure 9 is not explained and lies closer to the darunavir share of the drugs added than to a mass balance of what left solution. X5 The entropy explanation for water-promoted mixing is not tested. C40

The AI writer's own judgement (2): what was tested where. All membrane data come from supersaturated solutions made from methanol stocks; only the dissolution test used the ASD, and in that test the medium contained no polymer. X3·P8 The Caco-2 donor was judged supersaturated against an amorphous solubility measured in a different buffer, with no stated crystallization inhibitor, and the darunavir concentration in the pre-equilibrated suspensions is not printed. X10 Diffusion-cell and Caco-2 fluxes differ by about three orders of magnitude in the same units, so only the relative changes compare across set-ups. X9

The AI writer's own judgement (3): smaller inconsistencies. Two solubility ratios in the discussion (about 1.30 and almost 82-fold) do not follow from Table 2 (1.37 and 58), and the Table 2 caption names one method for both columns. X1·X2 Atazanavir was bought as the sulfate, but no conversion to the free base described in the results is reported. X4 The darunavir-rich $T_g$ in the dry mixtures sits above neat darunavir without comment, and the nanoTA temperatures are not DSC $T_g$ values. X7·X8 The 5 min dissolution samples are not shown, and the Supplementary data (Figs. S1–S4, Tables S1–S3) were not available for this reading. X12 None of these changes the direction of the main findings, which rests on several independent measurements.

Conclusion

The paper shows, for atazanavir with darunavir hydrate, that the benefit of the amorphous form can be partly lost when a crystalline drug is present: atazanavir release from a PVP ASD fell by 31%, its flux plateau in a diffusion cell fell to 30% of control and started at 37 instead of about 75–100 µg/mL, and its Caco-2 flux fell by about 36%, while darunavir's transport did not change. R7·R9·R13·R10 Added darunavir lowered the atazanavir ceiling step by step, already below darunavir's own saturation, and darunavir itself left solution with atazanavir. R11·C1 The authors explain this by darunavir partitioning into and diluting atazanavir's drug-rich colloidal phase, with water promoting a mixing that does not occur in the dry state ("likely"). C37·C5·C40 In the AI writer's view, the losses are measured and the mechanism is inferred: its direct test would be to analyse the drug-rich phase, and its practical test whether the loss persists in biorelevant media and in vivo. X6·C42