Crystallization inhibitor or soluble carrier? Choosing a cellulose polymer for enzalutamide amorphous solid dispersions

Background and motivation

Most drug candidates are now poorly water soluble (Biopharmaceutics Classification System class II or IV). A drug must dissolve before it can cross the gut wall, so formulations that create a supersaturated solution, a dissolved concentration above the solubility of the crystal, can raise absorption. The catch is that supersaturated solutions are metastable and tend to crystallize, which removes the solubility advantage. Amorphous solid dispersions (ASDs), in which the drug is mixed with a polymer at the molecular level, are one of the most used supersaturating formulations and have outperformed crystalline forms in vivo (refs 3–6 of the paper).

The polymer in an ASD is "generally recognized" to have two jobs: to help release the drug from the amorphous matrix, and to delay the crystallization that follows (refs 7, 8). For a fast-crystallizing drug the second job decides whether any supersaturation survives. The field therefore holds that a good ASD polymer needs hydrophobic groups, which interact with the drug in water and prevent crystallization, and hydrophilic groups, which interact with water and let the polymer, and with it the drug, dissolve (refs 9–11). These requirements pull against each other. Earlier work found that a highly soluble polymer can release a drug quickly but then let it crystallize fast (Xie and Taylor 2016, ref 19), perhaps because very water-soluble polymers interact less with hydrophobic drugs (ref 20); a poorly soluble polymer can limit release so that little supersaturation forms (Li and Taylor 2018, ref 21). How to balance the two was, in the authors' words, "currently uncertain". C1·C2·C6

Two further problems motivate the paper. First, the polymers used in marketed ASDs (mainly HPMC, HPMCAS and PVPVA) are few, structurally similar and were not designed for ASDs, so they do not allow structure–activity studies. The Taylor and Edgar groups had synthesized many new polysaccharide derivatives for ASDs and tested them in vitro (refs 9, 11, 15, 24, 25), but none had been tested in vivo. Second, it is unclear which in vitro test predicts in vivo performance; recent studies suggested that membrane flux, which follows only the freely dissolved drug, "may provide greater insight than simple dissolution tests" (Wilson et al. 2018, ref 6; refs 22, 23). C3·C7·C8

The model drug is enzalutamide, a prostate-cancer drug in BCS class II. It is lipophilic (calculated log P 4.75) and does not ionize over the physiological pH range. Its crystalline solubility is 2.9 µg/mL, and its amorphous solubility, the most that can be molecularly dissolved before a drug-rich colloidal phase separates, is about 42 µg/mL, a ratio of about 14.5. The marketed product is a lipid-based formulation in soft gel capsules with low drug loading, so patients take four large capsules (9 mm × 20 mm) per dose. A high-loading ASD that absorbs as well would reduce this pill burden. M1·R2·R4

The study therefore had three aims: screen many polymers, most of them the groups' own designs, for how long they delay enzalutamide crystallization; pick two new cellulose derivatives with a different balance of hydrophobic and hydrophilic groups and compare their hydrophilicity; and test ASDs of these two polymers, including a high-loading one, by membrane flux and in rats, against a crystalline slurry and a lipid formulation similar to the marketed product. The abstract frames the comparison as an "extremely effective" crystallization inhibitor set against a "less effective, but more hydrophilic" one "that might afford better release". C9

Background concepts

Supersaturation, amorphous solubility and drug-rich colloids. The crystalline solubility is the equilibrium concentration over the stable crystal. Above it the solution is supersaturated and can crystallize. The amorphous solubility is higher; when more drug than this is added, the excess separates as drug-rich colloids (nanodroplets or amorphous aggregates) while the free drug stays at the amorphous solubility. In this paper the crystallization tests start at 70 µg/mL enzalutamide, of which about 42 µg/mL is free drug and the rest colloidal aggregates (from the authors' earlier study, ref 6). A polymer must then stop crystallization both from the free drug and from the aggregates. The authors describe 70 µg/mL as "approximately 1.5×" the amorphous solubility; the stated values give 1.67 (Note by the AI writer: a small rounding difference, recorded in the card as X6). R2·C11·X6

Nucleation induction time. The induction time is the delay between creating a supersaturated solution and the first detectable crystals. Here it was followed with a UV probe in the solution: crystallization shows as a fall of the absorbance at the drug's maximum (237 nm), as dissolved drug is lost, together with a rise of the signal at 446 nm, where enzalutamide does not absorb, caused by turbidity from crystals. A longer induction time means a better crystallization inhibitor. P5

Membrane flux and free drug. 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 kDa cutoff. It is "generally accepted" that only free (molecularly dissolved) drug crosses such a membrane and that flux is proportional to the free-drug concentration; colloids and polymer stay behind. The mass flow rate into the receiver (µg/min) is therefore a measure of how much free drug, that is how much supersaturation, a formulation produces. P6·C10·E3

Pharmacokinetic quantities. After an oral dose the plasma concentration rises and falls. The area under the concentration–time curve (AUC) measures total exposure, Cmax is the peak concentration and Tmax its time. The paper also reports the elimination half-life $t_{1/2}$ and the absorption half-life $t_{1/2,\mathrm{a}\mathrm{b}\mathrm{s}}$, both from a one-compartment model with first-order absorption and first-order elimination fitted with PKSolver; a short absorption half-life means fast absorption. The absolute bioavailability $F$ compares the oral AUC with the AUC after an intravenous (IV) dose, which by definition reaches the circulation completely. Because the oral dose was 50 mg/kg and the IV dose 10 mg/kg, the authors scale by five (eq 1):

In plain terms: the fraction of the oral dose that reached the blood is the oral AUC divided by the AUC that the same dose would have given if injected, estimated as five times the AUC of the 10 mg/kg injection. Here $AU{C}_{0-\mathrm{i}\mathrm{n}\mathrm{f}}$ is the area extrapolated to infinite time. The scaling assumes that exposure rises linearly with dose, which the authors support with dose-linear data for other compounds in rats (refs 28–30), not with enzalutamide data. E1·C21

Solubility parameters. A solubility parameter measures the cohesive energy of a material. Hoy's method estimates it by adding up group contributions of the molecule or polymer repeat unit, split into a polar part ${\delta }_p$, a hydrogen-bonding part ${\delta }_h$ and a dispersive part ${\delta }_d$. The total is (eq 2):

In plain terms: the total parameter is the length of a vector whose three components are the polar, hydrogen-bonding and dispersive contributions. Larger polar and hydrogen-bonding parts mean a more water-like, hydrophilic material. The authors use the values only "qualitatively", for the polymers in their unionized state. Table 1 prints the unit as J^1/2 cm^−2/3; the usual unit is J^1/2 cm^−3/2. E2·P3·X11

Degree of substitution. Cellulose has three hydroxyl groups per anhydroglucose unit (AHG). The degree of substitution (DS) is the average number of these that carry a given substituent, for example DS(Ac) for acetyl or DS(CP) for carboxypentyl. For hydroxypropyl chains, which can grow on themselves, the molar substitution (MS) counts all hydroxypropyl units per AHG. Table 3 of the paper lists these values for every polymer. M2·M3

Methods in detail

The study has three linked parts: a crystallization screen of many polymers in solution, the choice and characterization of two polymers for ASDs, and a test of those ASDs in vitro (membrane flux) and in vivo (rat pharmacokinetics). The flow below shows how they connect.

Contribution 1: A crystallization screen of 24 polymers

Problem. Enzalutamide crystallizes quickly from supersaturated solutions, so a useful ASD polymer must delay this. With many new polymers available, the first question was which of them inhibit crystallization at all, and whether structure predicts it.

What they did. Each polymer was pre-dissolved in 35 mL of pH 6.5, 50 mM phosphate buffer at 37 °C, stirred at 300 rpm, at 5, 25 or 50 µg/mL. Polymers that did not dissolve readily were added as 1–10 ppm solutions in an organic solvent (details in Supplementary Table S1, not available here). Then 245 µL of a 10 mg/mL methanolic enzalutamide solution was added, giving 70 µg/mL, and the UV probe recorded the induction time; each condition was run in triplicate. The 24 polymers (Table 3) comprise six commercial ones (HPMCAS MF grade, HPMC E3 grade, PVPVA VA 64, PVP K29/32, cellulose acetate phthalate and hydroxypropyl cellulose) and 18 cellulose derivatives synthesized in earlier work: carboxypentyl hydroxypropyl celluloses (CPHPC-106, -282, -069), cellulose acetate adipates and related esters (CAAd 0.67, CAAdP 0.85, CAAd PEG, CAAd HOEt, CAAd 3CES HE, CAUn 3CES TAE), cellulose acetate suberate (CA Sub), cellulose propionate trioxodecanoate (CPTod-202), a methyl cellulose adipate (MC-Ad) and methyl or ethyl carboxypentyl celluloses (MCCP-A/B, ECCP-A to D). Not every polymer was tested at every concentration: 12 at 5 µg/mL, 7 at 25 µg/mL and 20 at 50 µg/mL (Figure 1). P5·A1·A2·M2–M25

Contribution 2: Two polymers with opposite profiles

Problem. To test whether crystallization inhibition or release matters more in vivo, the authors needed two polymers that differ sharply in both.

What they did. From the screen they chose CA Sub, "an extremely effective crystallization inhibitor" (over 16 h) with a "low to moderate" aqueous solubility of 3 mg/mL (ref 10), and CPHPC-106, which delayed crystallization for only "~1 h" but has a much higher aqueous solubility of 43.5 mg/mL (ref 25). Both solubility values come from the groups' earlier papers. To compare hydrophilicity the authors calculated Hoy solubility parameters (Table 1). Structurally (Table 3), CPHPC-106 is a hydroxypropyl cellulose with DS(HP) 2.20 and MS(HP) 4.40, whose oligo(hydroxypropyl) chains are partly capped by a five-carbon tether ending in a carboxylic acid, DS(CP) 1.06. CA Sub is a cellulose acetate, DS(Ac) 1.82, carrying suberate half-esters, DS(Sub) 0.90, whose carboxylic acid sits at the end of a six-methylene chain. C16·M2·M3·R1

ASDs were made by dissolving drug and polymer in a common solvent, methanol for CPHPC-106 and tetrahydrofuran for CA Sub, removing the solvent by rotary evaporation (water bath 25 °C), drying under vacuum at 35 °C for 1 h and grinding. CPHPC-106 ASDs were made at 10% and 50% drug loading (written 10:90 and 50:50 Enz:polymer); CA Sub was made only at 50% "due to low polymer availability". P1·C20

Contribution 3: Flux and rat pharmacokinetics on the same formulations

Problem. The authors wanted to know both how the ASDs perform in vivo and whether a simple in vitro test ranks them correctly.

What they did in vitro. Following their earlier method (Wilson et al. 2018, ref 6), each ASD, or crystalline drug, was stirred for 1 h in 30 mL of pH 6.5 buffer at 37 °C and then placed in the donor compartment at a concentration equivalent to 100 µg/mL enzalutamide. The receiver held 30 mL of buffer behind the dialysis membrane and was sampled every 5 min for HPLC; runs were in triplicate. The SEDDS was not tested this way. A separate check looked at the CA Sub ASD suspension under a polarized light microscope for up to 5 h, and a release study compared the CA Sub ASD with crystalline drug (Supplementary Figure S1). P6·P7·P8

What they did in vivo. Male Sprague–Dawley rats with access to food and water received 50 mg/kg enzalutamide by mouth as one of five formulations: a crystalline slurry, the 50:50 CA Sub ASD, the 10:90 and 50:50 CPHPC-106 ASDs (suspensions at 5 mg/mL, prepared 1 h before dosing) and a SEDDS made by dissolving crystalline enzalutamide at 4.5 wt% in Labrasol, a non-ionic surfactant, to mimic the marketed product. Three rats received each formulation. Blood was taken from 0.25 h to 3 days. A separate group received 10 mg/kg intravenously in DMSO:PEG-400 (10:90) to give the reference AUC for $F$. Plasma was analysed by LC-MS/MS (m/z 465 > 209, diclofenac internal standard). AUC came from non-compartmental analysis, the half-lives from a one-compartment fit. P2·P9·P10·P11·P12

Note by the AI writer (not discussed in the paper): the paper names three AUC intervals. Table 2 reports $AU{C}_{0-\mathrm{i}\mathrm{n}\mathrm{f}}$; the methods describe $AU{C}_{0-48}$ by the linear trapezoidal rule; and the y-axis of Figure 3 reads "AUC_{0-72 hours}" while its points sit at the Table 2 values. The vehicle of the ASD suspensions, the way the crystalline slurry was prepared, the size of the IV group and the IV AUC itself are not given, and no solid-state characterization of the prepared ASDs is reported. X1·X14

Experimental analysis

Result 1: Only a few polymers kept enzalutamide in supersaturation for long

Without polymer, enzalutamide crystallized in about 15 min (Figure 1). At 50 µg/mL, four polymers held it free of crystals beyond 16 h, the end of the plotted range: HPMCAS and three of the new derivatives, CAAd 3CES HE, CA Sub and ECCP-B. HPMC reached about 4 h, CAAd 0.67 and MCCP-B about 2.5 h, and the remaining 13 polymers, including PVP and PVPVA (about 1.5 h) and all three CPHPC polymers (about 0.6–0.7 h), stayed under 2 h. CAUn 3CES TAE showed no measurable delay. Table A gives the readings for the seven polymers that were tested at all three concentrations. R1

Table A. Time to crystallization (h) of 70 µg/mL enzalutamide in pH 6.5 buffer at 37 °C with pre-dissolved polymer, read from Figure 1 (approximate; average of triplicates). ">16" marks bars that run to the end of the plot. CPHPC-106 was tested only at 50 µg/mL, where it read about 0.7 h; the text gives "~1 h". R1

The authors draw three conclusions from the screen. First, inhibition depended on concentration for some polymers, "in particular CA Sub", which delayed crystallization for about 5 h at 5 µg/mL and over 16 h at the higher concentrations. They describe CAAd 0.67, CA Ph, PVPVA, PVP, HPMCAS and HPMC as independent of concentration. Note by the AI writer: Table A shows that several of these fall at 50 µg/mL (CAAd 0.67 from about 5.0 to about 2.45 h, CA Ph from about 3.6 to about 1.45 h), and HPMC peaks at 25 µg/mL; the error bars are not defined, so "no dependence" should be read loosely. C13·X7

Second, the four best inhibitors at 50 µg/mL share four features in the authors' account: a cellulose backbone, branched structures in the tether carbon chain, a carboxylate at the end of the tether and six or fewer tether carbons. These features did not guarantee success, since MC-Ad, the MCCP polymers and three of the ECCP polymers performed about like PVP and PVPVA. The authors conclude that the observations "point to the subtle differences in structure" that affect inhibition and are not visible from "simple summing of chemical functional groups", and in the discussion that "there were no obvious chemical or structural features that correlated with crystallization inhibition", so screening is essential. Note by the AI writer: in Table 3 the suberate tether of CA Sub and the carboxypentyl tether of ECCP-B are drawn as unbranched chains, and the discussion itself calls CA Sub's tether "octamethylene", while Table 3 draws six methylene groups; the shared-feature list is best read as descriptive. C14·C15·X9

Third, polymers that delayed crystallization for more than 2 h "can therefore theoretically have a large impact" in vivo, because transit through the small intestine takes about 2–4 h (refs 26, 27). The discussion counts 4 of the 24 polymers above 5 h; that count holds at 50 µg/mL, while HPMC passed 5 h at the two lower concentrations. C12·C32·X8

Result 2: CPHPC-106 is more hydrophilic than CA Sub

Table B. Hoy solubility parameters (unit as printed in Table 1, J^1/2 cm^−2/3) and aqueous solubilities, from Table 1. The polymer solubilities of CA Sub and CPHPC-106 are cited from refs 10 and 25; the source of the HPMCAS-MF value and the conditions of all three are not given. M1–M4

CPHPC-106 has higher polar (14.4 against 12.1) and hydrogen-bonding (11.4 against 8.2) parts than CA Sub, which the authors take as "indicating" that it is the more hydrophilic polymer, in line with its far higher solubility. The total parameters can be recomputed from their components with eq 2 (for enzalutamide $\sqrt{{4.3}^2+{15.2}^2+{27.1}^2}=31.37$, printed 31.4). The authors also call enzalutamide more hydrophobic than any polymer "based on the total solubility parameter, as well as the calculated Log P". Note by the AI writer: the drug's high total parameter (31.4) comes from its dispersive part; its polar part (4.3) is the lowest in the table, so the polar part and log P, not the total, carry this reading. C17·C18·E2·X10

The discussion explains the solubility difference from structure. CPHPC-106 carries "oligo(hydroxypropyl) substituents, themselves somewhat hydrophilic", some capped with a C5 chain ending in COOH; CA Sub "lacks polar, neutral hydroxyalkyl groups" and carries its carboxyl at the end of a long hydrophobic tether, so "the lower water solubility of CA Sub is unsurprising". The degree of substitution of the ionizable COOH differs only slightly, 1.06 against 0.9. C36·M2·M3

Result 3: The soluble polymer won in rats

Figure 2 shows the plasma profiles. Both CPHPC-106 ASDs rose quickly to a peak at about 3 h (about 15 µg/mL for 10:90 and about 12 µg/mL for 50:50) and stayed above the SEDDS (peak about 9 µg/mL) for most of the 72 h. The CA Sub ASD and the crystalline slurry stayed near 1 µg/mL throughout. Table C summarizes the parameters. R5·R6

Table C. Pharmacokinetics after a 50 mg/kg oral dose in rats, mean (SD), n = 3, from Table 2. F against an IV dose of 10 mg/kg. R6

The 10:90 CPHPC-106 ASD gave the highest AUC and Cmax. The 50:50 CPHPC-106 ASD gave "a very similar profile to the lipid-based formulation, in spite of the tenfold increase in drug loading". The CA Sub ASD, made with the far better inhibitor, gave an exposure "similar to that of the crystalline slurry", with a later Tmax (12.0 against 5.0 h) and slower absorption. Because no crystals appeared in the CA Sub suspension for up to 5 h, the authors state that the low plasma levels "cannot be due to crystallization of enzalutamide from the formulation". C22·C23·C24·R7

The authors summarize the ranking as 10:90 CPHPC-106 > 50:50 CPHPC-106 > SEDDS > 50:50 CA Sub = crystalline slurry for Cmax, F and AUC, with about 4% bioavailability for the crystalline slurry and CA Sub ASD and "approximately 9×" higher values for the SEDDS and CPHPC-106 ASDs. Note by the AI writer: AUC and Cmax follow this order, but F does not, since the SEDDS (31.5%) lies above the 50:50 CPHPC-106 ASD (28.4%). F is also not proportional to AUC, which eq 1 with one IV reference would require: dividing each AUC by 5F gives implied IV AUCs from about 180 to about 239 µg h/mL. The paper does not say how F was averaged, so the F column is best treated as approximate and the AUC column used for ranking. The fold changes are also stated loosely: Table C gives AUC ratios to crystals of 7.1 (50:50 CPHPC-106) and 12.0 (10:90), F ratios of 7.9 and 12.8, while the conclusions speak of "a five-fold improvement". C25·C26·C53·X2·X3·X4

Result 4: Flux ranked the formulations as AUC did

Figure 3 plots the mass flow of the four formulations tested in the diffusion cell against their AUC. The points fall into two groups: the crystalline slurry and the CA Sub ASD at low flux and low AUC, and the two CPHPC-106 ASDs at high flux and high AUC. R8

Table D. Mass flow across the dialysis membrane, read from Figure 3 (approximate; 100 µg/mL drug equivalent in the donor, pH 6.5 buffer, 37 °C), with AUC from Table 2. R8·R6

The text says the low pair had mass flows "almost 5 times lower" than the 10:90 ASD; the plotted values differ by about 8 times (Note by the AI writer, card X5). The authors read the mass flow as showing that the free drug from the CA Sub ASD is "similar to that obtained from dissolution of the crystalline form", in line with the similar AUCs. A release study (Supplementary Figure S1, not available here) "confirmed" this: the CA Sub ASD reached only about 3 µg/mL, very close to crystalline drug, and released drug more slowly than the crystals dissolved. Their summary is that flux measurements "appear to be a good in vitro surrogate for rank ordering in vivo performance"; the abstract and conclusions state the correlation without a hedge. C27·C40·C42·C44·R9·X5

Result 5: Drug from the CPHPC-106 ASDs was absorbed fast

The absorption half-life was 0.9 h for the 50:50 and 1.0 h for the 10:90 CPHPC-106 ASD, close to the 1.1 h of the SEDDS, in which the drug is already dissolved; it was 1.9 h for the crystalline slurry and 3.6 h for the CA Sub ASD (Table C). The authors write that the ~1 h absorption half-life "confirms that a substantial portion of the drug is absorbed prior to crystallization", that the match with the SEDDS "suggests that dissolution is not the rate limiting step" for the CPHPC-106 ASDs, and that the slower absorption from crystals and from CA Sub "is consistent with lower luminal concentrations and slower dissolution rates". R6·C47·C48·C49

Discussion

The central result is a reversal of the intuitive ranking. In the authors' words, "a polymer which is not the most effective crystallization inhibitor, CPHPC-106, leads to substantially improved absorption relative to the extremely effective crystallization inhibitor, CA Sub". They read this as showing that "there is indeed an interplay between effectiveness as a crystallization inhibitor and other key polymer properties, most notably polymer solubility in this instance". The chain of reasoning they offer, step by step, is shown below. C34·C35

Why the CA Sub ASD failed. Since no crystals were seen, the authors attribute the low free-drug concentration to "the low polymer solubility combined with strong drug-polymer interactions in the matrix" (citing Li and Taylor 2018, ref 21). They add that polymer solubility alone cannot explain it, because the measured solubility of CA Sub "would indicate that it can completely dissolve" in the volume of the flux experiment; "thus it is likely that the presence of the drug strongly suppresses the polymer dissolution". The difference in polymer solubility "presumably accounts, in part, for the different levels of molecularly dissolved drug from the two ASDs". These are interpretations: the paper did not measure polymer release from the ASDs, the polymer concentration in solution, or drug–polymer interactions. C37·C41·C43

Why a short inhibition time was enough for CPHPC-106. The authors give two reasons why the in vitro screen understates CPHPC-106. First, the gut differs from the beaker in hydrodynamics and in fluid composition, notably bile salts, both known to influence crystallization (refs 34, 35), so "crystallization may well occur over a longer timeframe in vivo". Second, absorption removes drug from the gut, so "the rate of absorption is likely to impact crystallization rate and extent"; if absorption is as fast as or faster than crystallization, a supersaturating formulation delivers more drug for a drug whose absorption is limited by solubility, assuming no extensive first-pass metabolism or efflux. The roughly 1 h absorption half-life of the CPHPC-106 ASDs is their evidence that most drug is absorbed before it crystallizes. C45·C46·C47

How long must inhibition last? The authors call this "one unresolved question". Gastrointestinal transit suggests several hours, but "for many drugs, the window of absorption is actually quite short". Their practical conclusion is that screening for crystallization inhibition "should clearly be combined with other in vitro tests", with in vivo correlation as feedback for choosing surrogate tests; flux is their candidate. C31·C33·C44

High drug loading. The 50:50 CPHPC-106 ASD performed, by the authors' account, "slightly better" than the SEDDS, which contains only 4.5% drug, so "considerably less of the CPHPC-106 excipient" gives exposure comparable to the commercial-type formulation. They see "a significant potential advantage in terms of patient compliance". Note by the AI writer: AUC (305 against 283 µg h/mL) and Cmax (11.7 against 9.34 µg/mL) favour the ASD, F (28.4 against 31.5%) favours the SEDDS, and the standard deviations overlap, so "comparable" is the safer word. C50·C51·R6·X3

Note by the AI writer (not discussed in the paper): the result concerns release of the polymer from the dispersion, not the solubility of the neat polymer. CA Sub dissolves to 3 mg/mL on its own, which by the authors' own calculation would be enough to dissolve fully in the flux experiment, yet the 50:50 dispersion behaved like crystals. A practical screen therefore needs a test on the ASD itself, such as the flux test used here or a release test that follows both drug and polymer. The comparison between the two polymers is also limited to one loading (50%): the 10:90 CPHPC-106 ASD differs from the 50:50 one both in drug loading and in the amount of polymer per dose (nine times more polymer per mg of drug), so the higher exposure of the 10:90 ASD cannot be assigned to loading alone. X15

Limitations

Limits the authors state. CA Sub was tested at only one drug loading, 50%, "due to low polymer availability". The authors point out that the in vitro crystallization test differs from the gut in hydrodynamics and fluid composition, and that the timeframe over which inhibition must last is "unresolved". Their mechanistic statements carry hedges ("presumably", "likely", "can be attributed", "appear to be a good in vitro surrogate"). C20·C31·C45·C37·C43·C44

The AI writer's own judgement.

• Three rats per formulation, with large variability for the 10:90 CPHPC-106 ASD (AUC SD 137 µg h/mL, F SD 18.6%). The difference between the CPHPC-106 and CA Sub ASDs is large enough to stand; finer distinctions (50:50 CPHPC-106 against SEDDS) are not. R6
• The F values are not proportional to the AUC values as eq 1 requires, the stated F ranking conflicts with Table 2, and three AUC intervals are named (0–inf, 0–48 h, 0–72 h). Use AUC_0–inf from Table 2 for comparisons. X1·X2·X3
• Several summary numbers in the text do not match the tables or figures: "five-fold" and "9×" against AUC ratios of 7.1–12.0, "almost 5 times lower" flux against a plotted ratio of about 8, and "approximately 1.5×" the amorphous solubility against 1.67. X4·X5·X6
• The mechanism is inferred, not measured: no polymer release data, no drug–polymer interaction data for the CA Sub ASD, no crystallization data in vivo or beyond 5 h in the CA Sub suspension (whose Tmax was 12 h). The polymer solubilities are cited, with unstated conditions. X16·X11
• Flux was measured at 100 µg/mL drug equivalent after 1 h of pre-stirring, while the rats received a 5 mg/mL suspension; the four flux points fall into two clusters, so the correlation rests on a two-level comparison, and the SEDDS was not tested. X15
• The suspension vehicle, the preparation of the crystalline slurry, the IV group and its AUC, and any solid-state check of the prepared ASDs are not reported. Induction times are quoted in the text without their polymer concentration, and some structural statements and reference numbers in Table 3 do not match the text. X14·X8·X9·X12·X13

Conclusion

The paper screened 24 polymers, most of them new cellulose derivatives, for how long they keep supersaturated enzalutamide from crystallizing, and found several new derivatives that work as well as HPMCAS in solution. It then tested ASDs of two of them in rats. The strongest inhibitor, CA Sub, gave an ASD that delivered no more drug than crystals (AUC 45 against 43.1 µg h/mL), while the weak but highly soluble CPHPC-106 gave 305 µg h/mL at 50% loading and 516 µg h/mL at 10% loading, matching or exceeding a lipid formulation like the marketed product. Membrane flux put the formulations in the same order. The authors explain the result by polymer solubility and by the drug holding back polymer dissolution, and conclude that ASD performance is "a complex interplay of drug and polymer properties", so a crystallization screen must be combined with tests of release such as flux. The measured part of this argument, the in vivo ranking and its agreement with flux, is clear; the mechanism remains an interpretation. R1·R6·R8·C35·C33·C52