What happens after an amorphous solid dispersion dissolves, and why rats notice: enzalutamide ASDs, amorphous aggregates and membrane mass flow

Background and motivation

An ASD raises the concentration of a poorly soluble drug by dispersing it, amorphous, in a hydrophilic polymer. How much this helps absorption depends on what the drug does after the polymer lets it go. Three outcomes compete. The drug can stay dissolved, supersaturated with respect to the crystal. It can crystallize, either in the wetted ASD particle (the "matrix") or from the supersaturated solution. Or, if enough drug dissolves to exceed the amorphous solubility and crystallization is slow, it can separate into a second, drug-rich amorphous phase that appears as colloidal aggregates. The introduction frames this third outcome as glass–liquid phase separation (GLPS) when the water-saturated drug-rich phase is below its glass transition temperature, and liquid–liquid phase separation (LLPS) when it is above (Ilevbare and Taylor 2013; Mosquera-Giraldo and Taylor 2015). C7

Earlier work from the same groups had shown why the aggregates might be useful. The amorphous solubility is the highest concentration of free drug that a solution can hold; once it is passed, extra drug does not raise the free concentration but forms more aggregates, which coexist with the solution in metastable equilibrium. Raina et al. (2014) proposed that the aggregates could act as a depot that replaces drug as it is absorbed and keeps supersaturation at an optimized level, and Indulkar et al. (2016) demonstrated such a reservoir effect in an in vitro membrane transport study, where nano-sized aggregates sustained the maximum flux and macroscopic amorphous particles did not. Several studies had also shown that ASDs, typically at low drug loading, dissolve past the amorphous solubility and form aggregates spontaneously. The authors note, however, that "there is little in vivo data to support this supposition". C8·C34·C39

A second problem motivates the experimental design. Conventional non-sink dissolution tests, run in a closed vessel without an absorption compartment, measure total dissolved drug over time. They cannot tell a solution that sits at the amorphous solubility with no reserve from one that sits at the same free concentration with a large pool of aggregates behind it, and they do not capture the coupling between dissolution and permeation that operates in the gut. The authors therefore pair dissolution with two further measurements: the speciation of the drug after dissolution (free drug, aggregates, residual solid) and the mass flow of the dissolved drug across a membrane. C38·C40

The model drug is enzalutamide, a treatment for castration-resistant prostate cancer that is marketed as a lipid-based soft-gel capsule (Xtandi). It is a Biopharmaceutics Classification System (BCS) class II drug whose oral absorption is limited by low aqueous solubility rather than permeability, it is unionized across the physiological pH range, it is not a P-glycoprotein substrate, and it is about 82% absorbed in humans from the lipid formulation. The authors chose it because these properties isolate the effect of supersaturation and speciation on absorption. They describe it as fast crystallizing. R19·C1·C30 The goal stated in the introduction is to extend the earlier in vitro studies and to determine whether ASD performance, in terms of the potential to form amorphous aggregates, matters for absorption in vivo: make ASDs with copovidone (PVPVA) and hydroxypropyl methylcellulose acetate succinate (HPMCAS), two of the most common commercial ASD polymers, characterize them with a side-by-side diffusion cell, probe their tendency to undergo GLPS, dose them to rats and look for correlations. C9

Background concepts

Supersaturation as an activity ratio. The authors start from the simple relation $J=pc$, where $J$ is the flux across a biological membrane, $p$ the permeability and $c$ the concentration available for permeation. For a high-permeability, low-solubility drug, $c$ limits the absorption rate, and a supersaturating formulation helps by raising $c$ above the crystal solubility. E6·C31 Because only free drug drives membrane transport, they define the supersaturation ratio $S$ through the thermodynamic activity rather than the concentration (Eq. 7):

In plain terms: $S$ is the activity $a$ of the dissolved drug divided by the activity $a^{*}$ of the drug in equilibrium with its crystal; written with concentrations $c$ and activity coefficients $\gamma$ (the asterisk marks the crystalline reference state). Setting $\gamma /{\gamma }^{*}=1$ reduces this to $S=c/c^{*}$, an approximation the authors say "may or may not be a valid assumption", depending on the medium and the concentration range. E7·C32

The ceiling: binodal, spinodal and amorphous solubility. The discussion places the amorphous solubility on a phase diagram. The upper limit of supersaturation is the spinodal, where amorphous drug separates spontaneously; in practice the binodal, the boundary at which a drug-rich phase first becomes possible, is what is called the amorphous solubility. Between the two, separation has a driving force but may lag; beyond the spinodal it is instantaneous. The consequence that the whole paper rests on is that the free-drug concentration, and with it $S$ and the membrane flux, is at its maximum whenever an amorphous drug-rich phase is present: adding more drug only makes more of that phase. C33·C35

Glass or liquid. Whether the drug-rich phase is glassy or liquid depends on its glass transition temperature ($T_g$) when saturated with water. Enzalutamide melts at 470 K and has a $T_g$ of 362 K; it takes up an estimated 0.50 mol of water per mol at 100% relative humidity, and the Fox equation then gives a wet $T_g$ of 352 K (79 °C), well above the 310 K (37 °C) of the experiments. The authors conclude that amorphous enzalutamide is glassy even when water-saturated, and the precipitate indeed "appears glassy" under crossed polarizers, so the phase separation studied here is GLPS. The Fox equation is named, with its reference, but not printed. M1·R4·C14·C17·E8

Estimating the amorphous solubility. The amorphous solubility can be estimated from the crystal's properties. The free-energy difference between amorphous and crystalline drug is approximated with the Hoffman equation (Eq. 1):

In plain terms: the energy penalty of being amorphous grows with the heat of fusion $\Delta H_f$ and with the distance of the experimental temperature $T$ below the melting temperature $T_m$. The amorphous solubility then follows from the crystalline solubility $C_{eq}$ by the approach of Murdande and co-workers (Eq. 2; the paper prints square brackets where parentheses are used here):

In plain terms: the crystal solubility is multiplied by the free-energy advantage of the amorphous form, $\exp (\Delta G_c/RT)$ with $R$ the gas constant, and reduced by a term $\exp (-I(a_2))$ that accounts for the water absorbed by the amorphous drug, which lowers its activity; $I(a_2)$ is estimated from the water sorption isotherm (0.69 for enzalutamide). E1·E2·M1

Measuring activity with a membrane. In a side-by-side diffusion cell, a donor and a receiver compartment are separated by a membrane, here a regenerated-cellulose dialysis membrane with a 6–8 kD cutoff. The mass flow into the receiver, $dm/dt$, divided by the membrane area $A$ gives the flux (Eq. 3), and the flux is proportional to the activity of the drug in the donor (Eq. 4):

In plain terms: $D$ is the diffusivity of the drug, $h$ the membrane thickness and ${\gamma }_m$ the activity coefficient of the drug inside the membrane; all of these are taken as constant for a given set-up (Higuchi and Higuchi 1960; Raina et al. 2015), so the measured mass flow reports the activity $a$ in the donor. Combining the two gives Eq. 5:

In plain terms: whatever else is in the donor (aggregates, polymer, undissolved particles), only the free drug's activity sets how fast drug crosses; aggregates matter only through how well they keep that activity up while drug is removed. E3·E4·E5·C12·C43

Speciation. The paper uses "speciation" for the partition of the drug after an ASD has been in water for a while: molecularly dissolved drug, amorphous aggregates, and residual solid (undissolved ASD, agglomerated aggregates or crystals). Schematic 1 draws the possible conversions as rate constants: ASD dissolution ($k_{diss}^{ASD}$), transfer across the membrane ($k_{diff}$), aggregate formation ($k_{AA}$) and dissolution ($k_{diss}^{AA}$), crystallization from solution ($k_c$) and of aggregates ($k_c^{AA}$), and dissolution of crystals ($k_{diss}^c$). None of these rate constants was measured; the schematic organizes the argument. C40·C42

Methods in detail

The study has four strands that use the same four ASDs: characterizing the drug's solubility ceiling, following the ASDs through dissolution and speciation, measuring membrane mass flow, and dosing rats. The flow below shows how they connect.

Strand 1: the solubility ceiling

Crystalline enzalutamide was stirred in excess for 48 h in 50 mM phosphate buffer at pH 6.5 and 37 °C, in triplicate, also with 5 µg/mL HPMCAS; solids were removed by ultracentrifugation (35,000 rpm, 15 min) and the supernatant assayed by HPLC with UV detection at 237 nm. P1·P2 The amorphous solubility was found in two ways. In the fluorescence method, a methanol stock (10 mg/mL) was added stepwise to buffer containing 2 µg/mL of the environment-sensitive probe 4-Di-2-Asp to give 10–65 µg/mL; the probe's emission maximum shifts to shorter wavelength once a drug-rich phase forms. In the UV-extinction method, the stock was added continuously at 35 µL/min while the extinction at 445 nm was followed with a dip probe; a rise marks scattering from a new phase. The extinction runs were repeated with 900 µg/mL HPMCAS or PVPVA predissolved, to see whether the polymer moves the ceiling. Differential scanning calorimetry gave the melting point, $T_g$ and heat of fusion, and gravimetric water sorption from 5% to 95% relative humidity gave the water uptake used in Eq. 2. P3·P4·P5·P6

Strand 2: dissolution and speciation

ASDs at 10 and 50 wt% drug were prepared by rotary evaporation, PVPVA ones from 50:50 dichloromethane:methanol and HPMCAS-MF ones from 75:25 dichloromethane:acetone, dried 12 h under vacuum and ground. They are written as drug:polymer weight ratios: 10:90 and 50:50 Enz:PVPVA, 10:90 and 50:50 Enz:HPMCAS. XRPD and FTIR (SI Figures S1–S4) showed them amorphous after preparation and after up to 2 weeks at ambient conditions, and all experiments were done within that window. P10·P11·R5 For dissolution, enough ASD to give 100 µg/mL total drug was added to 12.5 mL of pH 6.5 buffer at 37 °C and 300 rpm, and the concentration was followed in situ by UV; for 10:90 Enz:HPMCAS small crystals disturbed the UV signal, so samples were filtered and assayed by HPLC. The pH of 6.5 was chosen to represent the small intestine and because HPMCAS does not dissolve at low pH. P12·C10

The amount of amorphous aggregates was measured with Nile Red, a dye that barely fluoresces in water and brightens in less polar surroundings, so that it reports the drug-rich phase. To calibrate it, solutions of 45–100 µg/mL enzalutamide, all above the amorphous solubility, were made from the methanol stock with 2 µg/mL Nile Red, and the intensity at 610 nm was compared with two independent measures of the aggregate mass: a mass balance that counts all drug above 42 µg/mL as aggregates, and the dried weight of the pellet after ultracentrifugation. P8·P9·C21·E10 ASD samples taken after 1 h of dissolution were filtered at 1.0 µm and read on this scale. P13 For the suspensions actually dosed to rats, ASDs were stirred for 1 h at room temperature (22 ± 2 °C), filtered at 5.0 µm through a filter pre-saturated with drug, and the filtrate diluted 200-fold into buffer that already held 40 µg/mL drug, 5 µg/mL HPMCAS and 2 µg/mL Nile Red, a diluent chosen so that the aggregates would neither dissolve nor crystallize before the reading. P14·C11 Aggregate size was measured by dynamic light scattering, without further detail. P20

Strand 3: membrane mass flow

A PermeGear side-by-side cell with a 30 mm orifice held 30 mL in each compartment. To map mass flow against concentration, the authors added a concentrated methanol solution of enzalutamide to the donor to reach 10–100 µg/mL (methanol below 1%, matched in the receiver), with 5 µg/mL HPMCAS on both sides to prevent crystallization, and sampled the receiver every 5 minutes; the mass flow is the slope of receiver concentration against time. P15 To test the ASDs, each was stirred for 1 h in about 30 mL of buffer and the whole suspension was placed in the donor, with plain buffer in the receiver; a crystalline slurry was measured the same way. P16 Temperature and stirring of the cell are not printed.

Strand 4: rats

Male Sprague-Dawley rats received the four ASD suspensions and a suspension of crystalline drug by oral gavage at 10 mL/kg, with free access to food and water. Blood was taken at 0.25–72 h, and plasma enzalutamide was measured by LC-MS/MS (transition m/z 465 > 209, diclofenac as internal standard). Cmax was read from each animal's data, and AUC was computed by non-compartmental analysis with the linear trapezoidal rule; Figure 8 uses AUC from 0 to 48 h. P17·P18·P19·E9 The paper does not print the dose in mg/kg, the number of animals per group or what the error bars show; Table 4, the speciation of the dosing suspensions, adds up to 50.00 mg of drug per formulation. X8·R12

Experimental analysis

The ceiling is about 42 µg/mL, and polymer does not move it

Crystalline enzalutamide dissolved to 2.9 ± 0.1 µg/mL at 37 °C; with 5 µg/mL HPMCAS the value was 1.7 (0.8) µg/mL, which a t-test did not distinguish (p > 0.05). Both amorphous-solubility methods gave 42 µg/mL, and 900 µg/mL of either polymer left it at 42–43 µg/mL (Table 2). R1·C15 In the example of Figure 2, extinction stays at about 0.015–0.022 up to about 40 µg/mL and climbs from about 45 µg/mL; the arrow sits at about 42.5 µg/mL, and the probe's emission maximum falls from about 593–596 nm to below 590 nm over the same range. R2

Table A. Crystalline and amorphous solubility of enzalutamide, pH 6.5 phosphate buffer, 37 °C; standard deviations in parentheses, n = 3 (from Table 2). R1

The authors call the measured value "in good agreement" with their theoretical estimate of "around 39 µg/mL", a "13-fold enhancement". R3·C16 Note by the AI writer (not discussed in the paper): the estimate cannot be reproduced from the printed numbers. With $\Delta H_f$ = 33,820 J/mol, $T_m$ = 470 K and $T$ = 310 K, Eq. 1 gives 7.59 kJ/mol, and Eq. 2 with $C_{eq}$ = 2.9 µg/mL and $I(a_2)$ = 0.69 gives about 27.7 µg/mL; 39 µg/mL would need a free-energy difference slightly above the largest value the Hoffman form can give. The measured ratio 42/2.9 is 14.5, so the "S of approximately 13" used later in the discussion matches the theoretical 39 µg/mL, not Table 2. X1·X2

Dissolution: three ASDs reach the ceiling, one crystallizes

Figure 3 shows the four dissolution profiles at 100 µg/mL total drug. 10:90 Enz:PVPVA released fastest: about 38 µg/mL after 30 minutes and a peak of about 41.5 µg/mL near 1.1 h, followed by a fall to about 30.5 µg/mL at 2 h and about 22.5 µg/mL at 6 h. 50:50 Enz:PVPVA rose more slowly to about 40.5 µg/mL near 1.5 h and settled at about 27.5 µg/mL. 50:50 Enz:HPMCAS rose most slowly, about 35.5 µg/mL at 1 h, reached about 42 µg/mL after 3–4 h and showed no de-supersaturation for 7 hours and no crystals. 10:90 Enz:HPMCAS reached a maximum of only about 10 µg/mL, and crystals were seen in the medium. R6·C19 The authors attribute the fall of both PVPVA ASDs to crystallization of the drug; the precipitate itself is not characterized. C18·X12

Table B. Dissolved enzalutamide (µg/mL) read from Figure 3 (approximate), and the dissolution AUC (µg·h/mL) of Table 3. R6·R7

Integrated over 0–2 h, the ranking is 10:90 PVPVA > 50:50 PVPVA > 50:50 HPMCAS > 10:90 HPMCAS; over 0–4 and 0–7 h, 50:50 HPMCAS comes first. The authors' point is that "the quantitative ranking of AUC for ASDs changed with the integration time frame", so dissolution does not give one answer. R7·C20 Note by the AI writer: two entries of Table 3 do not match the curves of Figure 3. The 0–7 h value for 50:50 Enz:PVPVA (180) implies a mean of about 17 µg/mL between 4 and 7 h, while the curve sits at about 27.5–29 µg/mL; the 0–4 h value for 10:90 Enz:HPMCAS (24) implies about 4 µg/mL between 2 and 4 h, while the curve shows about 8–9 µg/mL. Figure 3 also ends at 6 h although 7 h values are reported. The change of ranking between 0–2 and 0–4 h does not depend on these entries. X5

Aggregates: only the 10% PVPVA dispersion sheds many

The Nile Red calibration (Figure 4) is linear over 45–100 µg/mL: the intensity at 610 nm rises from about 48 to about 605, the mass-balance aggregate mass from about 3 to about 58 µg/mL, and the weighed pellets give about 4.2 to about 58.3 µg/mL, with wide scatter at 70 µg/mL. The supernatant stayed at 42 µg/mL. The authors call this "indeed a good correlation" and note that the pellet weights, "although yielding somewhat variable results", support the mass balance. R8·C22

Figure 5 then reads the ASDs. After 1 h, the 10:90 Enz:PVPVA filtrate gave a strong Nile Red peak (about 480, against about 640 for the 100 µg/mL standard and about 50 for 45 µg/mL), which the authors convert to "approximately 50 µg/mL" of aggregates; with 42 µg/mL dissolved, 8 µg/mL is left for material removed by the filter, "most likely undissolved ASD or agglomerated nanoparticles". 50:50 Enz:PVPVA matched the 45 µg/mL standard, about 3 µg/mL of aggregates. Both HPMCAS ASDs gave a signal like the blank, "indicating that neither of those dispersions led to detectable GLPS": 10:90 Enz:HPMCAS because it crystallized, 50:50 Enz:HPMCAS because it never exceeded the amorphous solubility. Dynamic light scattering gave aggregates of about 100 nm. R9·R10·R11·C23·C24 Note by the AI writer: on the Figure 4 line, a peak of about 470–480 corresponds to about 45 rather than 50 µg/mL, and the calibration and the ASD spectra were recorded with different slit widths (3/10 nm and 5/5 nm); the legend of Figure 5 also gives sampling times for the two HPMCAS dispersions (5 h and 1 h) that differ from the caption (1 h, and 2 and 5 h). Neither changes the qualitative picture. X13·X4

The dosed suspensions tell the same story in milligrams (Table 4). For 10:90 Enz:PVPVA, 0.21 mg was free drug (assumed at the amorphous solubility), 37.65 mg was in aggregates and 12.14 mg remained as solid; for 50:50 Enz:PVPVA, 0.21 mg was free, none was in aggregates and 49.79 mg remained as undissolved ASD or crystals. R12 Note by the AI writer: these suspensions were characterized at 22 °C, and 0.21 mg at 42 µg/mL corresponds to 5 mL, so the drug was at about 10 mg/mL, roughly a hundred times the in vitro tests; the HPMCAS suspensions were not characterized. X7

Mass flow: a ceiling, and a clear ranking of the dispersions

For supersaturated solutions (Figure 6), mass flow rose from about 0.18 µg/min at 10 µg/mL to about 0.58 at 40 µg/mL and about 0.87 at 50 µg/mL, then stayed between about 0.81 and 0.91 µg/min up to 100 µg/mL; the text describes it as "fluctuating around 0.85 μg/min". The authors explain the plateau by GLPS: beyond the amorphous solubility added drug forms aggregates, the free drug stays at the amorphous solubility, and the activity, and hence the flux, is at its maximum ("can be explained"). R13·C25 Note by the AI writer: the rise continues from 40 to 50 µg/mL, so in this set-up the plateau begins near 50 rather than at 42 µg/mL. X10 A solution from 10:90 Enz:PVPVA and drug with 900 µg/mL PVPVA in the donor both gave about 0.76 µg/min; the caption notes that the two systems have the same mass flow. R14·C60

Table C. Mass flow of solutions after 1 h of dissolution (Table 5, triplicate; values in parentheses as printed) and the aggregate amounts of Figure 5. R15·R10

All ASD solutions gave more mass flow than the crystalline slurry; 10:90 Enz:PVPVA was highest, the two 50% dispersions were similar and 10:90 Enz:HPMCAS was lowest. C26 Note by the AI writer: the dispersion values lie below the plateau of Figure 6 even though three of them reached about the amorphous solubility in Figure 3. The two experiments differ (HPMCAS in both compartments and drug from methanol in Figure 6, plain buffer and real ASDs in Table 5), and at 1 h the 50% dispersions had not yet reached their peaks (about 38 and 35.5 µg/mL). The 30-minute 10:90 Enz:PVPVA point of Figure 6 (about 0.76) also differs from the 1 h value of Table 5 (0.632). X9·X3

Rats: aggregates add, crystallization subtracts

Figure 7 gives the mean plasma profiles. 10:90 Enz:PVPVA rose fastest (about 10.4 µg/mL at 0.5 h, 17.1 at 1 h) to about 30.8 µg/mL at 6 h and stayed highest to 24 h. 50:50 Enz:PVPVA reached about 21 µg/mL by 4 h and about 22 at 9 h; 50:50 Enz:HPMCAS rose more slowly (about 4.8 at 1 h, 10.9 at 2 h) to about 20 µg/mL at 9–12 h. 10:90 Enz:HPMCAS stayed below about 1.5 µg/mL, with a profile like the crystalline control. The authors see the slower absorption of the HPMCAS dispersion, and of 50:50 relative to 10:90 Enz:PVPVA, as "consistent with the in vitro dissolution profiles". R16·C27

Table D. In vitro mass flow (Table 5) against Cmax and AUC0-48 read from Figure 8 (approximate). R15·R17

The three dispersions that reached the amorphous solubility gave higher AUC and Cmax than the crystalline control (t-test, p < 0.05), which in the authors' words "clearly translates" into exposure, as simple relationships between membrane transport and free drug concentration predict. R18·C36 But since all three reach the same free-drug ceiling in vitro, supersaturation "is not adequate to explain" the higher Cmax and absorption of 10:90 Enz:PVPVA; its Cmax differed significantly from that of 50:50 Enz:HPMCAS (p < 0.05). C37·R18 The authors describe "a trend of approximately 30% more absorption and a higher Cmax for an equivalent dose" for the dispersion that forms aggregates; from Figure 8 the AUC ratio is about 770/575, about 1.34. C46·R17 The crystallizing dispersion gave no more exposure than crystalline drug (AUC about 61 against about 91 µg·h/mL): "Clearly the ASD that crystallizes leads to a dramatic reduction in absorption", and the type of precipitate formed matters. C47·R17

Which in vitro test predicts the rats?

Plotting Cmax and AUC0-48 against the mass flow (Figure 8) puts 10:90 Enz:PVPVA at the top, the two 50% dispersions together in the middle, and 10:90 Enz:HPMCAS and crystalline drug at the bottom. The authors write that "Somewhat remarkably" the in vivo data show "good correlation" with the in vitro mass flow, and that although "we do not have a great deal of data points", this "preliminary observation is extremely encouraging". Dissolution data alone "would not lead to adequate discrimination between three of the ASDs". C51·C52·C54 The exception is 10:90 Enz:HPMCAS, whose mass flow is more than twice that of the slurry while its exposure is not higher. The authors' explanation is "most likely" the difference in time frames: the in vitro measurement covers the first hour, when some supersaturation remains and little drug has to be replaced, whereas absorption in vivo lasts hours and depends on how fast the crystallized drug redissolves. C29·R15·R17

Discussion

How the authors connect the results

The argument runs through Schematic 1. Absorption needs dissolution, diffusion to the membrane and permeation, in series; if the solution is saturated with respect to the form that is dissolving, more drug can dissolve only as drug is removed by permeation. For a BCS class II drug, permeation is fast, so dissolution, "even from an amorphous formulation, may be the rate limiting step". C38·C39 What follows depends on what the drug has become:

In the authors' optimal scenario for a BCS class II drug, the ASD dissolves rapidly, passes the amorphous solubility and forms nanosized aggregates; the free drug is then held at the amorphous solubility by "the reservoir effect of amorphous aggregates". The route of aggregate formation drawn in Schematic 1 was verified for nifedipine and "is invoked" for enzalutamide. C40·C41 If crystallization is fast, "which appears to be the case" for 10:90 Enz:HPMCAS, absorption becomes dissolution limited and depends on how fast the crystalline precipitate dissolves. C42 Table 4 and Schematic 1 together "highlight", in the authors' words, that the 50% and 10% PVPVA dispersions differ not only in the initial supersaturation but in whether, and how much, they form amorphous aggregates. C44

Why should aggregates and crystals differ so much, when both are "precipitation events"? The authors expect $k_{diss}^{AA}$ and $k_{diss}^c$ "to be quite disparate", because the aggregates are more soluble than crystals and, at about 100 nm, have a high surface area; they say the in vivo data support this. C48 For the benefit of aggregates they list two mechanisms that "have been suggested": a reservoir effect, in which the small, prewetted aggregates equilibrate quickly with the solution and sustain a high flux, as shown in vitro by Indulkar et al.; and drift of the concentrated aggregates into the unstirred water layer next to the membrane, which "would provide an additional mechanism for transport of drug to the membrane", as proposed by Stewart et al. for itraconazole, although the drug must still dissolve before it permeates. They do not decide between the two ("Regardless of the exact mechanism"). C49

What the paper adds and where it stops

The paper's contribution is an in vivo comparison that earlier work lacked: two dispersions that reach the same free-drug ceiling, one with aggregates and one without, dosed side by side. The authors say plainly that very little in vivo data compare systems that do and do not form aggregates, that "it is impossible to know the generality of this observation", and that recent itraconazole work points the same way. C50 They also note a result that runs against common expectation: lower drug loading usually performs better in dissolution and in vivo, yet here the 10% HPMCAS dispersion was the worst ASD. They attribute this to "rapid matrix crystallization of enzalutamide upon contact with water", absent at 50% loading, and leave "the underlying reason for susceptibility of the particular dispersion to crystallization" to be elucidated. C55·C56

For formulation practice the authors draw two lessons. First, "it is desirable" that ASD particles dissolve in vivo to the maximum yield of amorphous drug-rich aggregates, and "we need to ensure that crystallization is inhibited over absorption-relevant timeframes". Second, tests that combine dissolution with a mass-transport measurement "may be useful to screen formulations", whereas non-sink closed-compartment dissolution "was not particularly predictive". C59·C53·C57

Note by the AI writer (not discussed in the paper): the polymer, not the loading alone, decided what 10% loading meant for enzalutamide. In PVPVA it gave fast release and the largest yield of aggregates; in HPMCAS-MF it let the drug crystallize. A screening programme that varies loading in one polymer would have missed this. The mass-flow measurement after 1 h also has a blind spot for late crystallization, as the 10:90 Enz:HPMCAS case shows; a longer or repeated measurement would be needed for drugs that crystallize after the first hour.

Limitations

Stated by the authors. The in vitro–in vivo comparison has few data points and is a "preliminary observation". C52 The generality of the aggregate benefit cannot be known from one drug. C50 The mechanism of the benefit, reservoir effect or drift through the unstirred water layer, is left open. C49 The reason why 10:90 Enz:HPMCAS crystallizes "remains to be elucidated". C56 The mass-flow test misses the slow redissolution of crystals that matters over hours in vivo. C29

The AI writer's own judgement. The direct results are solid: the amorphous solubility is measured twice and agrees, the mass flows are tabulated with their spread, and the advantage of the three dispersions that reach the ceiling over crystalline drug is statistically supported. R1·R15·R18 The headline in vivo claim is weaker than the abstract's wording suggests:

• No statistical test compares the aggregate-forming 10:90 Enz:PVPVA with 50:50 Enz:PVPVA, the comparator with "a similar maximum supersaturation" named in the conclusions, and no AUC test between dispersions is reported; the dose, the number of rats per group and the meaning of the error bars are not printed. The plotted error bars of 10:90 Enz:PVPVA do not overlap those of the 50% dispersions, so the difference is probably real for this study, but its size is uncertain. X6·X8
• Aggregate formation is not the only difference. The aggregate-forming dispersion also dissolved fastest, left the least undissolved solid in the dosed suspension (12.14 against 49.79 mg) and carried nine times more polymer per mg drug. Faster dissolution alone could raise Cmax and AUC; the authors fold fast dissolution into their "optimal scenario" rather than separating it. X14
• The aggregate amounts come from different conditions in vitro (37 °C, 100 µg/mL) and in the dosed suspensions (22 °C, about 10 mg/mL), and the HPMCAS suspensions were not characterized. X7
• Crystallization is inferred rather than shown, both for the 10% HPMCAS matrix and for the de-supersaturation of the PVPVA dispersions. X11·X12
• Several numbers do not reconcile: the theoretical amorphous solubility (39 against 27.7 µg/mL from the printed inputs), the supersaturation ratio (13 against 14.5), the 30-minute mass-flow point against Table 5, the Figure 5 legend against its caption, and two Table 3 entries against Figure 3. None of them changes the qualitative conclusions. X1·X2·X3·X4·X5
• The mass-flow plateau begins near 50 µg/mL rather than 42 µg/mL, and the supersaturated-solution and ASD mass flows were measured in different media, so the ASD values should be compared with each other, not with the plateau. X10·X9

Conclusion

For enzalutamide, an ASD's value in vivo depended on what the drug became after dissolution. Reaching the amorphous solubility gave a clear gain over crystalline drug; forming nanosized amorphous aggregates on top of that gave about 30% more absorption and a higher Cmax in rats; crystallizing during dissolution wiped out the gain. R17·R18·C46·C47 A side-by-side diffusion cell, which measures the activity of the dissolved drug rather than its total concentration, ranked the four dispersions in the order the rats did, while closed-vessel dissolution could not separate three of them. C51·C54·C57 The authors read this as support for "the role of drug-rich aggregates" in the oral absorption of solubility-limited drugs and recommend designing ASDs to dissolve to the maximum yield of aggregates while holding off crystallization. C58·C59 The evidence rests on one drug, one rat study and one aggregate-forming dispersion, so the next tests are a statistically powered comparison against a dispersion that reaches the same ceiling without aggregates, a way to separate aggregate formation from dissolution rate, and other drugs. X6·X14