When a drug cation drags the polymer out of solution: drug charge, insoluble complexes and release from HPMCAS dispersions

Background and motivation

An amorphous solid dispersion (ASD) disperses a poorly soluble drug at the molecular level in a polymer. The amorphous drug reaches a higher transient concentration than the crystal, and the polymer slows crystallization (Section 1). The polymer also changes how fast the drug leaves the solid. When drug and polymer leave the dispersion at the same normalized rate, called congruent release, the polymer controls the drug's release, and because the usual ASD polymers dissolve much faster than poorly soluble drugs, the drug comes out faster than the neat amorphous drug would. Attention in the field has therefore shifted to how the polymer itself dissolves. C1

For copovidone (PVPVA) dispersions this picture has been worked out in detail by the same group. At low drug loading (DL) drug and polymer leave together and quickly; above a threshold called the limit of congruency they no longer do, and release drops abruptly, the "falling-off-a-cliff" pattern (Indulkar 2019, Saboo 2019 and others, refs 15–20). C2 Specific drug–polymer interactions, mostly hydrogen bonding, have been implicated in poor release at relatively low loading in copovidone (refs 17, 18, 21–27). C3·C22 Hydroxypropyl methylcellulose acetate succinate (HPMCAS) is different on two counts. Its dispersions often lose release more gradually with loading rather than falling off a cliff (refs 14, 19, 28, 29), and few studies address their release mechanism. C3 And HPMCAS is an enteric polymer: it carries succinate carboxylic acid groups that must ionize before it dissolves. The grade used here, HPMCAS-MF, becomes more soluble above pH 5 and dissolves above pH 6 (Section 1). M9

Ionizable groups open a second route of interaction besides hydrogen bonding: HPMCAS can form hydrogen bonds with drugs (ref 42) but also ionic interactions (ref 43). In an earlier study of lumefantrine dispersions, the group "suggested" that poor release with enteric polymers came from drug–polymer ionic interactions that reduced polymer hydration and dissolution (Hiew 2022, ref 14). C5 Enteric coatings have been studied extensively for how they rupture (refs 30–37), but how a molecularly dispersed drug changes the dissolution of HPMCAS had received much less scrutiny, although HPMCAS is often the polymer of choice for spray-dried dispersions (refs 38–41). The authors call the interplay between drug and polymer chemistry and release "relatively underexplored". C4

One more piece of background sets the conditions of the study. When an enteric polymer dissolves, the protons released by its ionizing acid groups lower the pH near the polymer–water interface, in the unstirred water layer, below the pH of the bulk medium (refs 34–37, 44, 45). C7 In the group's previous HPMCAS paper, the gel layer of a compact dissolving in 50 mM pH 6.8 phosphate buffer was estimated at pH 6.0, and in 50 mM pH 9.0 glycine/NaOH buffer no gel layer formed at all and the surface pH was below 5.5 (Bapat 2023, ref 46). R20 A weakly basic drug is therefore more ionized at the dissolving surface than the bulk pH suggests.

From these points the authors build a hypothesis (Section 1): lipophilic drug cations lower HPMCAS solubility by forming an ionic complex with the polymer, which gives poor ASD release even at low loading, whereas anionic or neutral drugs affect polymer solubility less and can be loaded higher with adequate release. Their stated aim is to show that the ionization state and charge of the drug play critical roles in release versus loading for HPMCAS dispersions. C6 To test it they chose weakly acidic, neutral and weakly basic drugs, measured release from compacts at a bulk pH of 6.8 (some systems also at pH 7.5 and 9.0), and measured insoluble complex formation at pH 6.0, the estimated gel-layer pH. C6·C9

Background concepts

Surface area normalized release. The release experiments use an intrinsic dissolution (Wood's) apparatus: a 100 mg compact sits in a die so that a single 0.5 cm² face meets 100 mL of stirred buffer at 37 °C, and drug and polymer are assayed in samples taken over one hour. P7 The paper turns each component's release into a normalized rate with eq 1:

Here $k$ is the slope of the regression line of the released component against time, $V$ the volume of medium (100 mL), $S$ the exposed area of the die (0.5 cm²) and $x$ the weight fraction of the component in the compact; $R$ is given in mg·min⁻¹·cm⁻². In plain terms: the release rate per unit area is divided by the share of that component in the solid, so that drug and polymer can be compared on one scale. If drug and polymer have the same $R$, they leave in the same proportion in which they sit in the dispersion, which is what congruent release means. E1 A check by the card's extractor recomputed eq 1 from the release profile of the 25% DL loratadine dispersion (Figure S2A) and obtained about 0.10 mg·min⁻¹·cm⁻², the value of Figures 4D and 6. E1·X1

How HPMCAS dissolves. The authors summarize glassy polymer dissolution in Section 5: water enters the solid, plasticizes it and raises molecular mobility, and hydrated chains then diffuse into the solution if the solvent quality is good enough. Protonated HPMCAS interacts poorly with water, so a critical extent of ionization of its succinate groups, and hence a threshold pH, is needed before it dissolves. A lipophilic drug mixed into the polymer lowers the rate and extent of hydration, so a higher drug loading is "expected" to slow polymer dissolution. C24·C25 The authors also invoke a model from their previous paper, in which polymer solubility at the gel–solution interface is postulated to be the rate-limiting step (eq 3):

Here $J_p$ is the polymer flux, $D_p$ the polymer diffusion coefficient, $C_p$ the polymer solubility at the polymer–solution interface and $\delta$ the thickness of the boundary layer. In plain terms: the polymer leaves the surface at a rate proportional to how much of it can dissolve right at the surface and how fast it diffuses away through the stagnant layer. Anything that lowers $C_p$, such as a drug that makes the polymer insoluble, should slow the polymer and, if drug release follows the polymer, slow the drug as well. E3·C31·C32

Ionization of weak bases and acids. The paper classes the weak bases by how much of each is ionized at the dissolution pH of 6.8, from literature pKa values (Table 2): ritonavir (RTV, pKa 2.6) is predominantly unionized, loratadine (LOR, pKa 5.2) and clotrimazole (CLO, pKa 5.9) are partially ionized, and cinnarizine (CNZ, pKa 8.4) is predominantly ionized at every physiologically relevant pH. LOR becomes predominantly unionized at pH 7.5 and CLO at pH 9.0. RTV could not be tested in its ionized form, below pH 2, because HPMCAS does not dissolve there. M3·M5·M6·M7·C12 Indomethacin (IND, pKa 4.5) is a weak acid and is anionic at pH 6.8. M1 At pH 6.0 the authors give HPMCAS as about 96% ionized (taking its pKa as 4.9), CNZ above 99%, CLO 46% and LOR 14%. R19 The relation behind these numbers is not printed. Note by the AI writer (not discussed in the paper): the usual Henderson–Hasselbalch relation reproduces the CNZ and LOR values but gives about 93% for HPMCAS and about 44% for CLO, so the printed 96% and 46% are best read as approximate. X4

Insoluble complexes and hydrophobic ion pairs. The mechanism the authors propose draws on colloid chemistry (Section 5). The solubility of a molecule in water can be reduced by pairing it with an oppositely charged amphiphile; such "hydrophobic ion pairs" are used for proteins, peptides and small drugs paired with surfactants (refs 57–62). Charged polymers, polyelectrolytes, also precipitate with oppositely charged amphiphiles: the anionic surfactant sodium dodecyl sulfate precipitates a cationic cellulose, and dextran sulfate or alginate precipitate with cationic amphiphilic drugs such as ciprofloxacin, chlorpromazine, amitriptyline and doxepin (refs 63–67). Two oppositely charged polyelectrolytes form a complex coacervate (refs 68–70). C33 Binding of this kind is driven by electrostatic attraction together with hydrophobic interaction; for a cationic amphiphilic drug and an anionic polysaccharide it depends on drug hydrophobicity, polymer charge density and chain flexibility, and on pH, ionic strength and the type of ions present (refs 67, 71). C34 A lipophilic drug cation and the ionized succinate groups of HPMCAS fit this pattern.

Percent complexation. To measure the tendency to form such a complex, the authors spike drug and polymer from an organic solvent into buffer and measure how much polymer stays dissolved (Section 3.7). The percent of polymer complexed follows from eq 2:

In plain terms: whatever polymer leaves the supernatant when the precipitate is spun down is counted as complexed with the drug. This rests on HPMCAS alone being fully soluble in the buffer, which the authors check with a control. E2·C8·R10 The drug-to-polymer ratio is set in moles of drug per mole of HPMCAS succinoyl (COOH) groups, using a molar substitution of HPMCAS-MF reported earlier (ref 14) and not printed here; a 1:1 ratio corresponds to about 30% DL for LOR, CNZ and IND. P9·R12

Methods in detail

The study combines three kinds of measurement that are made separately and brought together only at the end: release of drug and polymer from compacts, the share of polymer each drug takes out of solution in a test tube, and probes of the drug's ionization (pH series, XPS, a neat-drug control). The flow below shows how they connect.

Contribution 1: release of drug and polymer from HPMCAS dispersions of eight drugs

Problem. To see whether a drug's charge matters, release has to be compared across drugs of different charge at matched loading, and the polymer's release has to be measured too, since the hypothesis acts through the polymer.

What they did. HPMCAS-MF (Shin-Etsu AQOAT-MF) dispersions were made by rotary evaporation. IND dispersions from 5 to 60% DL were cast from 1:2 methanol/dichloromethane with the water bath at 60 °C, indomethacin methyl ester (INDester) dispersions from 5 to 60% from tetrahydrofuran at 45 °C, and LOR, RTV, CNZ, phenolphthalein (PHPH) and CLO dispersions from 1:2 methanol/dichloromethane at 50 °C. They were dried under vacuum overnight, cryomilled and sieved to 106–250 µm, and their amorphous state was checked by powder X-ray diffraction (PXRD) and polarized light microscopy. P3·R18 The SI shows PXRD halos only for the 10% DL IND and INDester dispersions, and the paper does not say how the cilnidipine (CIL) dispersion used in Figure 12 was made. X16·X12 Neat amorphous LOR, RTV and CNZ were made by melting 2 g of drug 10 °C above its melting point, quenching in liquid nitrogen and cryomilling. P4

For release, 100 mg of powder was pressed at 1500 psi for 1 min into the die, which turned at 100 rpm in 100 mL of buffer at 37 °C; samples were taken at 5, 10, 15, 20, 25, 30, 40, 50 and 60 min and replaced with fresh buffer. Drugs were assayed by reversed-phase HPLC with UV detection (conditions per drug in Table S2), and HPMCAS by gradient HPLC with an evaporative light scattering detector (Shodex RS pak DS-413 column, standards 1–250 µg/mL). Each experiment was done in triplicate. P7·P11·P5 The standard medium was 50 mM pH 6.8 phosphate. The 25% DL LOR dispersion was also run in 50 mM pH 7.5 phosphate and in 100 mM pH 6.8 phosphate, the 25% DL CLO dispersion in 50 mM pH 9.0 glycine/NaOH, and neat amorphous LOR at pH 3.0, 6.8 and 7.5. A3·A4 The buffer recipes are given in Table S1; as discussed under Limitations, several of its rows do not match the stated pH and molarity. P10·X5

Why it works. Measuring both components on the eq 1 scale shows directly whether the polymer carries the drug out, and the neat HPMCAS compact in each medium gives a reference for how much a drug slows the polymer.

Contribution 2: a coprecipitation test for insoluble drug–polymer complexes

Problem. A complex inside a dissolving gel layer is hard to observe directly. The authors instead ask a simpler question: at the pH thought to prevail in the gel layer, does the drug take dissolved HPMCAS out of solution?

What they did. Stock solutions in tetrahydrofuran held either the drug alone, HPMCAS alone at 52.5 mg/mL, or drug plus HPMCAS at a set molar ratio of drug to succinoyl groups with the polymer fixed at 52.5 mg/mL. A small aliquot was added to agitated buffer, by default 50 mM phosphate at pH 6.0 (ionic strength 63 mM), so that the final organic solvent was 10–20 µL/mL and the polymer 0.5–1.0 mg/mL, far below the roughly 5 mg/mL that HPMCAS dissolves at pH 6 (ref 47). The precipitate was pelleted by ultracentrifugation and the polymer left in the supernatant was measured by HPLC or by a phenol–sulfuric acid colour reaction read at 490 nm; percent complexed follows from eq 2, with n = 3. P9·P6·E2·M9 The paper does not give the temperature, the time between addition and centrifugation, or the centrifugation speed for this test. X13

The controls make the reading of eq 2 possible. CNZ alone, added to 460 µg/mL, made the buffer turbid and the pellet was drug. HPMCAS alone at 1 mg/mL dissolved completely and gave no pellet. CNZ and HPMCAS added together gave a precipitate that could be pelleted and that contained HPMCAS; it was amorphous under polarized light. Since HPMCAS by itself is soluble at this pH, the authors take polymer in the pellet to "indicate" an insoluble complex with the drug. R10·C8 The test was run for all eight drugs at chosen ratios, for CNZ over ratios from 0.125:1 to 3:1, for LOR at 1:1 between pH 5.8 and 7.5, and for LOR at pH 6 and three ionic strengths. A5·A6·A7

Why pH 6.0. The authors chose pH 6.0 to mimic the gel layer–bulk solution interface of HPMCAS in 50 mM pH 6.8 phosphate, which their earlier paper estimated at pH 6.0; this is the pH at which complexation could affect release. C9·R20 The gel-layer pH of the dispersions studied here was not measured.

Contribution 3: probing the role of ionization

Problem. If charge is what matters, changing a weak base's ionization should change both its complexation and the release of its dispersion, and the effect should differ from that of ionization on the drug alone.

What they did. LOR, with a pKa of 5.2, lets the extent of drug ionization be varied between pH 5.8 and 7.5 while HPMCAS stays ionized; the authors measured LOR complexation at 1:1 across this range. R13·C38 They ran LOR and CLO dispersions at a bulk pH where each is partially ionized (6.8) and one where it is predominantly unionized (7.5 for LOR, 9.0 for CLO), and ran neat amorphous LOR at pH 3.0, 6.8 and 7.5 as a control for the drug's own pH-dependent dissolution. A3·R4·R7 Finally, X-ray photoelectron spectroscopy (XPS) of the N 1s region of compacts (neat CNZ, and 25% DL CNZ dispersion before and after 40–45 min of dissolution at pH 6.8, dried under vacuum) estimated the share of protonated CNZ nitrogen in the top ∼10 nm by fitting a protonated component next to the neat CNZ peak shape. P8·R8

Experimental analysis

Result 1: drug and polymer leave together, and loading slows both

For IND dispersions in 50 mM pH 6.8 phosphate, drug and polymer released at almost the same normalized rate at every loading (Figure 2): the drug from 0.85 mg·min⁻¹·cm⁻² at 5% DL to 0.50 at 60% DL, the polymer from about 0.84 to about 0.51. Neat HPMCAS released at about 0.91. R2·R3 INDester dispersions behaved the same way at lower rates (Figure 3): drug 0.66 to 0.39, polymer about 0.67 to about 0.37. R2·R3 The 10% and 50% DL points of both series come from the group's earlier paper (ref 46). R2 The authors read the matching rates as release "controlled by the polymer", and the higher rates of IND than INDester at equal loading as IND having less impact on HPMCAS release. C10·C11 In the Discussion they state more generally that drug and polymer release "appear co-dependent" for the HPMCAS dispersions studied, which Figure 4 bears out for the weak bases as well. C23·R4

Table A. Normalized drug release rate (mg·min⁻¹·cm⁻², mean with SD in parentheses, n = 3) from HPMCAS-MF dispersions in 50 mM pH 6.8 phosphate at 37 °C (from Table S3; the same data are plotted in Figure 11). A dash means not reported. R2

Table A shows the central puzzle. All five drugs in Figure 11 have "somewhat comparable" log P values: 4.1 (IND), 4.5 (INDester), 5.7 (RTV), and 4.55 (LOR) and 5.9 (CNZ) calculated with Chemaxon software. M1·M2·M3·M5·M7 Yet IND still released at 0.5 at 60% DL, whereas CNZ had fallen to 0.01 at 25% DL and LOR to 0.1 at 30%. R2 If lipophilicity alone slowed the hydration of the dispersion, drugs of similar lipophilicity should give similar curves. The authors therefore conclude that the observations "strongly suggest" a critical role for drug–polymer interactions in addition to lipophilicity. C26 Note by the AI writer (not discussed in the paper): the printed log P values describe the neutral molecules, and IND is ionized at pH 6.8, so its effective lipophilicity in the gel is lower than 4.1 suggests; CNZ, the most detrimental drug, also has the highest log P. The comparison least open to a lipophilicity reading is RTV against LOR: the more lipophilic RTV released at 0.45 at 25% DL, the less lipophilic LOR at 0.17. X18·R2

Result 2: weak bases slow the polymer, most when ionized

Figure 4 compares 25% DL dispersions of the four weak bases with neat HPMCAS. In every case the polymer released more slowly from the dispersion than from the neat compact, and drug and polymer stayed close to each other. R4·C14 The size of the effect followed the drug's ionization. Figure 5 expresses it as the ratio of the neat polymer rate to the polymer rate in the dispersion: about 1.6 for RTV, about 7.4 for LOR and 11.2 for CLO at pH 6.8, and 30 for CNZ, for which the authors write that the neat polymer dissolved "30-fold" faster. R5·C14 Where LOR (pH 7.5) and CLO (pH 9.0) are predominantly unionized, the ratio fell to about 2.0 and 2.1: raising the pH to reduce their ionization "reduced their deleterious impact on HPMCAS release". R5·C14

Table B. Release rates (mg·min⁻¹·cm⁻²) of 25% DL weak-base dispersions and neat HPMCAS in the same medium, read from Figure 4 (approximate), with the ratio of neat to dispersion polymer rate read from Figure 5. pH 9.0 is 50 mM glycine/NaOH; the other media are 50 mM phosphate. R4·R5

Two discrepancies sit in Table B. The CNZ ratio of 30 does not follow from the Figure 4 bars, which give about 0.90 divided by about 0.05, or about 18; 30 is instead what neat HPMCAS divided by the CNZ drug rate gives. The RTV ratio (about 1.6) is also lower than the bars imply (about 2.1). The order of the drugs is not affected. X2 And the release values for the same 25% DL dispersions differ between Figure 4 and Table S3: LOR about 0.10 against 0.17, CNZ about 0.03 against 0.01, RTV about 0.40 against 0.45. Recomputing eq 1 from the LOR release profile in Figure S2 supports the Figure 4 value. X1·E1

The neat polymer itself dissolved more slowly in the pH 9.0 glycine buffer (about 0.63) than in pH 6.8 phosphate (about 0.90). The authors explain this by the low buffer capacity of a buffer with pKa 9.0 at a dissolving surface whose pH is far lower; in their earlier work no gel layer formed in this buffer, placing the surface pH below 5.5. R4.g·R20·C13

Result 3: the opposite pH response of the dispersion and of the neat drug

For LOR, the dispersion released about six times faster at pH 7.5 (about 0.62) than at pH 6.8 (about 0.10). R4.j·R4.l The release profiles in Figure S2 show the same: about 2.5% of the LOR was released in 60 min at pH 6.8, about 17.8% at pH 7.5. R17 Neat amorphous LOR did the opposite (Figure 7): about 0.22 at pH 3.0, where it is ionized, about 0.05 at pH 6.8 and about 0.01 at pH 7.5. R7·C16 The neat drug behaves as a weak base should, dissolving faster where it carries a charge; in the dispersion the charge works against release. The abstract calls this "somewhat counterintuitive". C50 For reference, the amorphous solubility of LOR fell from 9.0 µg/mL at pH 6.8 to 7.6 µg/mL at pH 7.5, and that of CLO from 9.00 at pH 6.8 to 5.2 µg/mL at pH 9.0 (Table 1). R1.d·R1.e·R1.f Note by the AI writer (not discussed in the paper): at pH 7.5 the dispersion released about 17.8% of 25 mg of LOR into 100 mL within an hour, several times the amount the amorphous solubility would allow in solution, and the dashed "amorphous solubility" line in Figure S2B (about 1.3% release) does not match Table 1 (about 3.0%). The paper does not discuss either point. X14

Result 4: which drugs take HPMCAS out of solution

Table 4 is the paper's key complexation result. At pH 6.0 and a 1:1 ratio of drug to succinoyl groups, the anionic and neutral drugs took only 5–17% of the polymer into the precipitate, the weak bases LOR and CNZ 70% and 90%, and CLO 80% at 0.8:1. R12·C19 The neutral PHPH, which the authors expected to bind weakly, took 60%. R12.l·C19

Table C. Percent of HPMCAS-MF taken into an insoluble precipitate at pH 6.0, 50 mM phosphate (from Table 4, mean ± SD, n = 3); ionization at pH 6.0 as stated in Section 5. R12·R19

Complexation rose with the ratio. For CNZ (Figure 9) it went from about 41% at 0.25:1 to about 90% at 1:1 and about 99–100% at 1.5:1 and above, a plateau the authors place near 1.5 mol drug per mol COOH. R11·C18 For LOR it rose from 30 ± 5% at 0.25:1 to 70 ± 5% at 1:1, while RTV stayed between 6 and 15% over a tenfold range of ratios. R12.d–f·R12.h–k Two small source points: the text says ">90%" for CNZ at 1:1 where Table 4 gives 90 ± 5, and it counts CLO among drugs with "≥70% for a 1:1 ratio" although CLO was tested only at 0.8:1. X8 Note by the AI writer (not discussed in the paper): the IND 2:1 entry is matched to 60% DL, but doubling the drug:polymer mass ratio of the 30% DL mixture gives about 46% DL; 60% DL would correspond to about 3.5:1. The other rows of Table 4 scale consistently. X9

Figure S1 adds the effect of ionic strength on LOR at 1:1 and pH 6: about 70% complexed in 50 mM phosphate (ionic strength 63 mM), about 65% with NaCl added to 150 mM, and about 48% in 100 mM phosphate (130 mM). R16 The authors use these data to argue that the complex "appears unlikely" to be disrupted at biologically relevant ionic strength. C37

Result 5: LOR complexes only when it is ionized

Between pH 5.8 and 7.5, HPMCAS complexation with LOR at 1:1 fell from about 85% to about 70% (pH 6.0), 60% (6.2), 50% (6.3), 25% (6.5), 5% (6.8) and 2% (7.5) (Figure 10). R13 HPMCAS remains ionized across this range while LOR loses its charge, so the authors conclude that "it is clear" that ionization of LOR is necessary for insoluble complexation; LOR has no hydrogen bond donors, and they conclude that it binds ionized HPMCAS mainly through electrostatic interactions. C38·C20 The curve is steeper than the reported pKa predicts: at pH 6.0 only 14% of LOR should be ionized, yet 70% of the polymer was complexed. The authors call this "not unreasonable", since pKa values shift in local microenvironments, and propose that the apparent pKa of LOR bound to anionic HPMCAS may be raised, or that hydrophobic interactions make binding cooperative ("could result"). C21·C39·R19

XPS adds a solid-state view for CNZ (Figure 8, Table 3). Neat CNZ showed no protonated nitrogen (0 ± 0%). The dry 25% DL dispersion showed a second N 1s component at higher binding energy, about 2 eV above the main peak near 399 eV, worth 20 ± 1% protonation; after partial dissolution it fell to 11 ± 1%. R8·R9 The authors infer that ionization "largely occurs upon hydration within the gel layer", since only about a fifth of the CNZ is ionized in the dry dispersion while both CNZ and HPMCAS are nearly fully ionized at pH 6.0. C35 They link the drop after dissolution to surface drug enrichment relative to polymer; the paper reports no XPS surface composition to show that enrichment. C17·X7

Result 6: release falls along two lines with complexation

Figure 12 brings the two kinds of measurement together: each drug's release rate at a given loading is plotted against the percent complexation measured at the matching drug:COOH ratio. The drugs fall "roughly onto two trendlines". C27·R14 On the left are the drugs with 17% complexation or less: IND (about 0.68 at 30% DL, 0.50 at 60%), INDester (about 0.54 at 30%), RTV (about 0.66, 0.43 and 0.15 at 10, 25 and 47%) and CIL (about 0.25 at 40%). On the right, along a falling line, are LOR (about 0.47, 0.27, 0.17 and 0.11 at 10, 20, 25 and 30% DL, with 30–70% complexation), CNZ at 10, 25 and 30% DL (about 0.40, 0.03 and 0.00, with 41–90%), PHPH (about 0.22 at 60%) and CLO (about 0.07 at 80%). R14 CNZ at 5% DL, with 12% complexation, sits among the left group at about 0.60. R14.h Within the left group release still declines as complexation rises, which the authors find "interesting". C42 Some values in Figure 12 differ from Table S3 for the same dispersions, and the CIL, CLO and CNZ 30% release points appear only in this figure. X1·X12

The release of the LOR dispersion also tracks the gel-layer pH in Figure 13: about 0.12 at a gel-layer pH of about 6.0 (50 mM pH 6.8 phosphate), about 0.29 at about 6.2 (100 mM pH 6.8) and about 0.68 at about 6.3 (50 mM pH 7.5). R15 The authors match the steep rise over pH 6.0–6.3 to a steep part of the complexation–pH curve of Figure 10. C40 The figure does not say which component's rate is plotted; its values match the polymer rather than the drug bars of Figures 4D and 6, and only the pH 6.0 value has a stated source. X6 The buffer-strength experiment behind the middle point (Figure 6) raised the release of both components of the LOR dispersion from about 0.10 and 0.12 to about 0.25 and 0.29, while neat HPMCAS rose only from about 0.91 to about 1.00; the text gives factors of "∼2.9" and "∼1.2", the bars about 2.5 and 1.1. R6·C15·X3

Discussion

The authors' mechanism

The Discussion assembles the results into one causal chain. Drug release from HPMCAS dispersions depends on polymer release (refs 3, 11, 14, 28, 46), so anything that slows the polymer slows the drug. C30 HPMCAS dissolves to about 5 mg/mL at pH 6.0 because it is extensively ionized; with certain lipophilic drugs, its solubility at that pH fell through complexation and it precipitated. C29 Because the rate-limiting step is postulated to be polymer solubility at the gel–solution interface (eq 3), a lower $C_p$ means a slower polymer flux. C31·E3 The authors state the link with graded confidence: eq 3 "can be used to partially rationalize" the lower release with complexation; reduced solubility through complexation "likely largely accounts" for the drugs on the right-hand trendline of Figure 12; and drug lipophilicity is "likely the dominant factor" on the left. C32 The abstract states the same chain without a hedge: electrostatic interactions between HPMCAS and lipophilic cationic drugs "led to" insoluble complexes, which "resulted in" poor release. C32

The authors' chain as given in Section 5; the gel-layer pH is cited from their earlier paper, and the complex inside a dissolving gel layer was not observed. C9·C28·C32·C47·R12

For CNZ, both drug and polymer are nearly fully ionized at pH 6.0, and the authors argue from the XPS result (only about 20% protonated in the dry dispersion) that the ionization "largely occurs upon hydration within the gel layer". C35·R8 The excess of drug cations beyond the 1:1 ratio expected for electrostatic complexation, visible as a plateau only from about 1.5:1, "likely" reflects shielding by buffer ions at an ionic strength of 63 mM. C36·R11 For the conclusions they write that if such low-solubility complexes form in the gel or boundary layer, "a slow or nonexistent dissolution rate will be observed". C47

The phenolphthalein exception

PHPH is neutral at the test pH, yet it took 60% of the polymer into a precipitate and its 27% DL dispersion released at 0.22 mg·min⁻¹·cm⁻², on the same trendline as the cations. R12.l·R2·C41 The authors rationalize this by "hydrophobic hydrogen bonding", a term from work on lipophilic complexes formed by hydrogen bonding between a hydrophilic macromolecular drug and a nonionic surfactant (ref 74): the ionized carboxylate of HPMCAS offers a hydrogen-bond site for the phenolic OH groups of PHPH. PHPH also forms an insoluble complex with copovidone in water, likely through hydrogen bonding (ref 18), and phenol forms hydrogen-bonded complexes with carboxylate ions, including succinate, in water (ref 75). The authors present this explicitly as a "conjecture". C41 It was not tested, for example by spectroscopy of the precipitate.

Where the result applies

The data cover one polymer grade (HPMCAS-MF), eight drugs, compacts in a rotating die, bulk pH 6.8 (with LOR and CLO also at 7.5 and 9.0), and complexation in dilute buffer at pH 6.0. A1·A2·A5·P7·P9 Within those limits the authors draw practical conclusions. Drug chemistry, pKa and log P, together with the buffer capacity, pH and ionic strength of the medium (through their effect on gel or boundary layer pH) and the drug loading, are "all likely determinants" of release from HPMCAS dispersions. C44 Weakly basic lipophilic drugs with pKa values above 5 "may lead to poor release" from HPMCAS-MF dispersions, particularly at loadings above a 1:1 molar ratio of drug cations to polymer anionic groups. C45 With anionic or neutral drugs, "much higher drug loadings can be achieved" before release is impaired. C43 When formulating, the drug's ability to form interactions with the polymer that persist in water "should be considered". C46 The boundary-layer pH is "likely" lower than the bulk pH, which matters whenever ionization is predicted from pKa. C48

Note by the AI writer (not discussed in the paper): two practical consequences follow if the authors' reading holds. First, a quick coprecipitation test at the expected gel-layer pH could flag weak bases at risk before any dissolution work, and the 1:1 drug:COOH ratio, computable from the grade's succinoyl content, gives an upper bound on loading rather than a safe one; LOR already lost much of its release at 10–20% DL, well below its 1:1 point of about 30% DL. R2·R12 Second, the result in a strong or high-pH buffer can mislead: the LOR dispersion released about six times faster at pH 7.5 than at 6.8, and faster in 100 mM than in 50 mM buffer. R4.j·R4.l·R6 A strong drug–polymer interaction is often sought for physical stability; for a weak base in HPMCAS the ionic interaction appears to cost release, so the two goals can conflict.

Limitations

Limitations stated by the authors. The authors state one limitation directly: future studies are needed to establish whether these in vitro observations translate into poor in vivo performance. C51 Several of their explanations are hedged in their own words: the role of lipophilicity for the low-complexation drugs ("likely"), the apparent pKa shift or cooperative binding for LOR ("could result"), the shielding explanation for the plateau ("likely"), the survival of the complex at physiological ionic strength ("appears unlikely" to be disrupted), and hydrogen bonding for PHPH ("conjecture"). C28·C39·C36·C37·C41 They also note that the steep pH dependence of LOR complexation is "not perfectly predicted" by the reported pKa. C21

The AI writer's own judgement (1): what is measured and what is inferred. Complexation and release were measured separately, under different conditions: complexation in dilute buffer at pH 6.0, with 0.5–1.0 mg/mL polymer spiked from tetrahydrofuran; release from compacts at a bulk pH of 6.8. The step between them, an insoluble complex forming in the gel layer and lowering the polymer's interfacial solubility, was not observed: no complex was isolated from a dissolving compact, no interfacial polymer solubility was measured, and the gel-layer pH of these dispersions comes from the group's earlier paper. Figure 12 therefore shows a correlation across drugs, which the authors themselves describe with hedges. X10·C32 The coprecipitation conditions (temperature, holding time, centrifugation) are not given, so the percentages are comparable within this paper only. X13 One observation points to a second route for the buffer-strength effect: in 100 mM phosphate at pH 6, LOR complexation fell to about 48%, against about 70% in 50 mM, so the faster release in stronger buffer may come partly from weaker complexation at the same pH, not only from a higher gel-layer pH as Figure 13 implies. X11·R16·R6

The AI writer's own judgement (2): internal consistency of the numbers. Release values for the same dispersions differ between sources: the 25% DL LOR dispersion is 0.17 mg·min⁻¹·cm⁻² in Table S3 and Figure 12 but about 0.10 in Figures 4D and 6 and in a recomputation from Figure S2A; CNZ at 25% DL is 0.01 in Table S3 against about 0.03 in Figures 4B and 12; RTV at 25% DL is 0.45, about 0.40 and about 0.43 in the three places. X1 The CNZ polymer-slowdown ratio of 30 does not follow from the Figure 4 bars (about 18), and the buffer-strength factor of ∼2.9 is about 2.5 in Figure 6. X2·X3 Table S1 lists identical salt masses for the 50 mM pH 6.8 and pH 7.5 phosphate buffers, and its pH 6.0, 6.2 and 6.3 rows, all labelled 50 mM, contain one tenth, one fifth and one half of the salts of the pH 6.5 row at the same ratio; the pH 3.0 medium is missing. X5 The ionization percentages for HPMCAS and CLO are not reproduced by the standard relation, and one Table 4 ratio (IND 2:1 at 60% DL) does not match its loading. X4·X9 Figure 13 does not label the plotted component, and the CNZ amorphous solubility is cited to an analytical-method paper. X6·X15 None of these changes the two-group pattern, which is large compared with the differences, but they matter for anyone reusing individual numbers.

The AI writer's own judgement (3): scope of the comparison. The lipophilicity argument uses log P values of neutral molecules, although IND is ionized at pH 6.8; the cleanest contrast is RTV against LOR, where the more lipophilic drug complexes less and releases faster. X18 Only three weak bases with pKa above 5 were tested, all in one polymer grade, and the neat-drug control for the pH reversal was run for LOR only. M5·M6·M7·R7 The amorphous state was shown by PXRD for two 10% DL dispersions only, the CIL dispersion's preparation is not described, and release above the amorphous solubility of LOR at pH 7.5 was not examined. X16·X12·X14

Conclusion

The paper shows, for eight drugs in HPMCAS-MF, that drug and polymer leave the dispersion together, that weak bases slow the polymer most when they are ionized, and that the weak bases ionized at pH 6.0 take most of the polymer out of solution as an insoluble complex while anionic and neutral drugs take little. R2·R4·R5·R12 LOR complexes only in its ionized form, and an LOR dispersion releases faster at the higher pH where the neat drug releases slower. R13·R4.j·R4.l·R7 Release falls roughly along two lines against complexation, and the authors read the right-hand line as complex-lowered polymer solubility at the gel surface, "likely largely" the cause. R14·C27·C32 The link is a well-supported correlation rather than a demonstrated mechanism, and the gel-layer pH that ties it together is cited, not measured. X10 For formulation the message is the authors' recommendation that lipophilic weak bases with pKa above 5 "may" release poorly from HPMCAS-MF dispersions, especially above a 1:1 ratio of drug cations to polymer acid groups. C45