Neutral versus acidic polymers for a basic, lipophilic drug: why lumefantrine dispersions trade release against stability

Background and motivation

Lumefantrine is the long-acting partner of artemether in the fixed-dose antimalarial Coartem (120 mg lumefantrine with 20 mg artemether). It is extremely poorly soluble: the paper cites a crystalline solubility of 2.6 ng/mL in fasted state simulated intestinal fluid. The marketed regimen is six doses of four tablets over three days, to be taken with food, because lumefantrine is absorbed slowly and erratically; its relative bioavailability rises 16-fold with a high-fat meal, and poor absorption on an empty stomach has led to treatment failure (Section 1, refs 2−8). Many formulation routes have been tried (self-nanoemulsifying systems, nanoparticles, solid lipid micro- and nanoparticles). A study in healthy volunteers found 24−48-fold higher bioavailability for hot-melt-extruded solid dispersions than for the marketed product, also in the fasted state, which the authors read as a sign that lumefantrine "could benefit" from being reformulated as an amorphous solid dispersion (ASD). M1·C6

An ASD has two jobs. In the solid state it must keep the amorphous drug from crystallizing; polymer selection, a miscible blend with a higher-Tg polymer and the resulting antiplasticization of a low-Tg drug are the usual tools. C5 On dissolution it must generate and hold a supersaturated solution. Above the amorphous solubility a drug can undergo liquid−liquid phase separation (LLPS) into colloidal drug-rich droplets, which act as a reservoir that keeps the free concentration at its maximum while drug is absorbed; such droplets are usually seen only at low drug loading (DL). C4

Work from the same group had framed ASD release in terms of congruent release (Indulkar 2019, ref 23; Saboo 2019, ref 24). At low DL the drug leaves the dispersion at the same normalized rate as the polymer, so release is polymer-controlled and fast; as DL rises, release becomes drug-controlled and the drug release rate drops abruptly. The highest DL that still releases congruently is the limit of congruency (LoC), and for drugs in polyvinylpyrrolidone−vinyl acetate copolymer (PVPVA) it had been found between 5 and 25% DL. C2·R28 Two further studies (Saboo 2020, ref 27; Yang 2021, ref 28) had suggested that hydrogen bonding between drug and PVPVA "may be detrimental" to release compared with systems without specific interactions. C3

The gap the authors identify is that almost all of this work used PVPVA. Whether dispersions in other polymers show the same "falling-off-the-cliff" pattern, and whether and how the drug changes the polymer's own release, was not established. C7 For a basic drug the obvious alternatives are enteric polymers: cellulose derivatives and methacrylate copolymers that carry carboxylic acids, dissolve only above a polymer-specific pH and can form acid−base interactions with basic drugs. Lumefantrine is lipophilic (log P 8.34) and weakly basic (tertiary amine pKa 8.73), with one hydrogen-bond donor and two acceptors. M1 An earlier XPS study of lumefantrine dispersions had found no protonation of the drug with HPMCAS at 20 and 40% DL but 50−75% protonation with HPMCP. R30

The paper therefore compares lumefantrine ASDs made with one neutral polymer, PVPVA, and four enteric polymers: hypromellose acetate succinate (HPMCAS), hypromellose phthalate (HPMCP), cellulose acetate phthalate (CAP) and the methacrylic acid−methyl methacrylate copolymer Eudragit L 100. For each it measures crystallization under accelerated storage, the release of both drug and polymer across DLs, the size of any drug-rich droplets, and the drug−polymer interactions in the solid and at the compact surface after partial dissolution. A1·A2·A3

Background concepts

Surface-normalized release of drug and polymer. A compact of the ASD is pressed into a die so that only one flat face of known area touches the medium (a Wood's intrinsic dissolution apparatus), and the amounts of drug and of polymer in solution are followed over time. Dividing the release rate by the exposed area gives a surface-normalized rate in mg·min⁻¹·cm⁻². To compare drug and polymer on one scale, each rate is also normalized by that component's weight fraction in the ASD; a polymer/drug ratio near 1 then means both leave in the proportion in which they are present, that is, congruent release. The paper does not print how the rates were calculated, but the ratios near 1 at very different DLs, and drug rates far above what the drug content alone could deliver (3.68 mg·min⁻¹·cm⁻² of lumefantrine from a 5% DL compact), imply this per-component normalization (note by the AI writer, from extractor note X7). R13·X7

Amorphous solubility. The amorphous solubility is the highest concentration of molecularly dissolved drug in equilibrium with the drug-rich amorphous phase; above it, further drug forms drug-rich droplets. The authors measured it in two ways: by adding a DMSO stock of lumefantrine to buffer with a syringe pump and noting the concentration at which light scattering at 450 nm starts (UV extinction), and by the change in the emission of the environment-sensitive probe pyrene as drug-rich droplets form. The two gave 6.6 ± 0.5 µg/mL and about 6 µg/mL in 50 mM pH 6.8 phosphate buffer at 37 °C, "in good agreement"; the crystalline solubility under the same conditions was below the 50 ng/mL quantification limit, so the amorphous form offers well over 100-fold supersaturation. R2·C8·R1 Drug-rich droplets from LLPS are typically 50−500 nm when they first form (ref 35), which is why dynamic light scattering (DLS) of the medium is used to detect them.

How an enteric polymer dissolves. All four enteric polymers carry carboxylic acid groups (Figure 1): phthalyl groups on CAP and HPMCP, succinyl (with acetyl) groups on HPMCAS, and methacrylic acid units, in roughly a 1:1 ratio with methyl methacrylate, in Eudragit L 100. M3·M4·M5·M6 Like any polymer, an enteric polymer dissolves by solvent diffusion into the glassy solid, which swells into a gel, followed by disentanglement of the chains into solution, a step often described with the reptation model (de Gennes, ref 45). For an enteric polymer a third condition must be met: enough of the carboxylic acids must be ionized for the chain to become soluble. C46 The titrations in this paper give the acid content and pKa that set this condition (Table 5): 1.19, 1.69, 2.94 and 4.48 mmol/g of carboxylic acid with pKa 4.97, 4.32, 4.57 and 6.21 for HPMCAS, HPMCP, CAP and Eudragit L 100, so at pH 6.8 HPMCP and CAP are fully ionized, HPMCAS 99% and Eudragit L 100 only 80%. R22·C28 Each polymer therefore has a minimum pH for dissolution; in the SI the solubility of HPMCP and CAP rises sharply between pH 4 and 4.5, that of HPMCAS between 4.5 and 5, and that of Eudragit L 100 between 4.25 and 4.75 (Figure S2). R23

The authors borrow two models of ionizable polymer dissolution from photoresist research (Section 5). The percolation model of Reiser and co-workers treats the ionizable groups as occupied cells through which base and counterions must percolate:

In plain terms: the polymer dissolution rate $DR$ is zero until the concentration of solvent-accessible ionizable groups $X_i$ exceeds a critical concentration $X_{ic}$, and then rises steeply (as the square of the excess). An additive that blocks ionizable groups lowers $X_i$ and so the rate; the model assumes that diffusion of solvent species through an interfacial boundary is rate-limiting. E1·C50 The critical ionization model of Willson and co-workers instead treats dissolution as surface etching, with the acid−base reaction as the rate-limiting step:

In plain terms: the dissolution rate follows the ratio of ionized ($[A^{-}]$) to non-ionized ($[HA]$) acid groups on the dissolving surface, a ratio fixed by pKa and pH, and a chain dissolves only once a critical fraction of its groups is ionized. In Willson's picture, an additive that hydrogen bonds to the acid group raises its effective pKa and so slows dissolution. E2·C50 Both models predict a strongly non-linear dependence of the dissolution rate on the fraction of usable ionized groups, which the authors later use qualitatively. C58

Acid−base interaction seen by IR and XPS. When a basic amine takes a proton from a carboxylic acid, the acid becomes a carboxylate, which has characteristic asymmetric and symmetric stretching bands in the infrared (here near 1594−1560 and 1410−1343 cm⁻¹), and the protonated nitrogen appears in the X-ray photoelectron N 1s spectrum at a higher binding energy than the neutral amine. The area of the protonated component relative to the total gives the percentage of drug that is protonated within the top ≈10 nm of the sample. P16·R19·R20 Because only lumefantrine contains chlorine, the ratio of carbon to chlorine in XPS also tells how much of the surface carbon belongs to the drug; the authors measured C/Cl = 12 ± 1 for neat lumefantrine and used it to split the surface carbon of each compact into drug and polymer. R21

Glass-forming ability. In the classification of Baird and co-workers (ref 33), a Class III compound does not crystallize during a DSC heat−cool−heat cycle, so it forms a relatively stable glass. Lumefantrine is Class III, with an onset melting point of 127.9 °C and an onset Tg of 19.7 ± 0.2 °C. R3

Methods in detail

The study has no model of its own; it is a comparative experimental study with a mechanistic argument built from the authors' measurements and from literature data. Its parts are summarized in the flow below.

Part 1: Dispersions and their solid state

Lumefantrine (Gojira Fine Chemicals) was co-dissolved with each polymer at 20 mg/mL total solids and rotary-evaporated with a 45 °C water bath: PVPVA (Kollidon VA 64) from dichloromethane, HPMCP (HP-50) and CAP from 8:2 dichloromethane/methanol, HPMCAS (AQOAT AS-MF) from 1:1 dichloromethane/methanol and Eudragit L 100 from 1:1 dichloromethane/ethanol. The solids were dried under vacuum for at least 24 h, cryomilled and sieved below 250 µm. PVPVA dispersions were made at nine loadings (5, 10, 15, 20, 25, 35, 40, 45 and 50%), HPMCAS and HPMCP at 5, 10, 15, 20, 25 and 50%. The Methods list CAP and Eudragit L 100 at 5, 10, 25 and 50% only, but Tables 1 and 2 and Figures 3 and 4 also report 15% DL dispersions of both, so they were evidently made too. P5·X3

Glass transitions were measured by modulated DSC (2 °C/min, modulation of 1 °C every 60 s, onset Tg from the reversing heat flow, n = 2). Crystallinity was checked by PXRD (Cu Kα, 4−40° 2θ) directly after preparation and after storage in open dishes at 40 °C/75% RH. The storage protocol is not described in the Methods: the condition appears in the Results and the sampling times only as labels in Figure 2 (from 1 to 76 weeks). P7·P8·X9

Part 2: Release of drug and polymer

For each DL, 50 mg of powder was pressed into an 8 mm die at 1500 psi for 1 min. The die was mounted on the shaft of a dissolution tester so that the single exposed face rotated at 100 rpm in 100 mL of deaerated 50 mM pH 6.8 phosphate buffer at 37 °C. Aliquots were taken over 2 h and replaced with fresh medium; at least three compacts were run per DL (n = 3 in Table 2). Lumefantrine was measured by reversed-phase HPLC (detection limit about 50 ng/mL). The polymers needed different assays: size-exclusion HPLC with UV detection at 210 nm for PVPVA and Eudragit L 100, and the phenol−sulfuric acid colorimetric assay at 490 nm for the three cellulose esters. As references, the five neat polymers and a melt-quenched compact of neat amorphous lumefantrine (melted at 180 °C and cast into the die) were run in the same way. P9·P2·P11·P12·P10

The use of compacts with one exposed face is what allows the authors to compare drug and polymer release rates directly, which is the basis of the congruent-release analysis; the trade-off is that a 50 mg compact in 100 mL can at most deliver 25 µg/mL of drug at 5% DL, so the extent of release is reported as a percentage of the drug in the compact as well as a concentration. DLS of the medium at 2 h (backscatter at 173°, triplicate, PDI below 0.2 for every reported size) was used to detect drug-rich droplets. P9·P15

Part 3: Interactions in the solid and at the surface

Attenuated total reflectance FTIR (ATR-FTIR) was recorded on ASD powders and, after 2 h of dissolution, on the compact surface, which was scraped off with a spatula after the compact had been removed and dried overnight under reduced pressure. Because ATR spectra of PVPVA respond to sorbed water, the authors also cast lumefantrine−PVPVA films on zinc selenide windows and dried them under a dry, CO₂-free purge until the PVPVA carbonyl bands (1736 and 1684 cm⁻¹) no longer moved, then recorded transmission spectra. For the 25% DL enteric dispersions, the spectrum of amorphous lumefantrine was subtracted from the ASD spectrum and the remainder compared with the neat polymer, to look for new carboxylate bands. P13·P14·R19

XPS was run on 25% DL compacts before and after partial dissolution, 10 min for PVPVA and 30 min for the enteric polymers. The N 1s spectrum was fitted with the reference spectra of neat lumefantrine and neat PVPVA plus a Gaussian/Lorentzian component for protonated lumefantrine, and the surface carbon was apportioned to drug and polymer with the C/Cl ratio of neat lumefantrine (12 ± 1). P16·R21 The different immersion times mean that surfaces are compared at different stages of release; comparisons are safest within one polymer, before against after. X14

Part 4: The enteric polymers' acid groups and the critical ionization argument

The acid content and pKa of each enteric polymer were found by dissolving the dried polymer with a known excess of NaOH and back-titrating with 0.1 M HCl; the two end points were located from the zero crossing of the second derivative of a polynomial fit, the acid amount came from the titrant volume between them, and the pKa was taken as the pH at 50% neutralization (n = 2). The solubility of the enteric polymers between pH 1.7 and 5.5 was measured by acidifying basified polymer solutions in 0.15 M KCl, centrifuging, filtering and assaying the supernatant. P17·P18

For the mechanistic argument the authors took literature dissolution rates of neat CAP and HPMCP as a function of pH (Spitael and Kinget 1979, ref 52) and re-expressed pH as the fraction of carboxylic acid groups ionized, using their measured pKa values (Figures S3 and S4). They then calculated, for a 25% DL dispersion, how many of the polymer's carboxylates would remain available for hydration if every protonated lumefantrine molecule blocked one carboxylate. For 100 g of ASD (25 g of lumefantrine, 528.94 g/mol, and 75 g of polymer), the SI gives:

In plain terms: the fraction of the polymer's acid groups $f_{\mathrm{C}\mathrm{O}{\mathrm{O}}^{-}}$ still free to ionize and hydrate is the number of acid groups $n_{\mathrm{C}\mathrm{O}\mathrm{O}\mathrm{H}}$ (75 g times the acid content of Table 5) minus the number of protonated drug molecules, which is the moles of drug $n_{\mathrm{L}\mathrm{M}\mathrm{F}}$ (0.047 mol) times the protonated fraction $x_{\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{t}}$ from XPS, all divided by the number of acid groups. For HPMCP, $(0.127-0.022)/0.127=0.83$; for CAP, $(0.221-0.027)/0.221=0.88$. Both results reproduce from the printed inputs. The CAP calculation in the SI uses 58% protonation, whereas Table 4 and the discussion give 55%; with 55% the result is still 0.88. E3·R25·X1

In the same SI section the authors tabulate the molar ratio of lumefantrine to polymer acid groups at 5, 10, 25 and 50% DL ("drug-to-polymer charge ratio"); at 50% DL it is 1.61 for HPMCAS, 1.12 for HPMCP, 0.65 for CAP and 0.42 for Eudragit L 100. R26

Experimental analysis

The fresh dispersions were single-phase glasses

All dispersions were X-ray amorphous after preparation and showed a single Tg without melting, except the 50% DL HPMCAS dispersion, which showed a melting endotherm at 126.3 ± 0.3 °C after heating above its Tg; the authors read this as some recrystallization during the DSC scan. R4·R6·C9 Onset Tg values reflect the very different polymers: at 5% DL, 98.6 °C (PVPVA), 110.0 °C (HPMCAS), 120.3 °C (HPMCP), 159.6 °C (CAP) and 192.7 °C (Eudragit L 100); at 50% DL, 52.7, 55.7, 85.7, 109.4 and 179.7 °C (Table 1). For PVPVA, HPMCAS and HPMCP the Tg falls steadily with DL, as expected for a low-Tg drug (19.7 °C). For CAP and Eudragit L 100 it does not (Eudragit L 100: 173.1 °C at 25% against 179.7 °C at 50% DL; CAP: 140.03 °C at 10% against 143 °C at 15% DL), a pattern the authors do not discuss and which, with n = 2, should not be fitted with a simple mixing rule (note by the AI writer, extractor note X16). R5·X16

PVPVA: fast, congruent release up to 35% DL, then a cliff

Neat amorphous lumefantrine released nothing detectable in 2 h, which puts its surface-normalized rate below 0.08 µg·min⁻¹·cm⁻². R8 In PVPVA the picture changes completely. From 5 to 35% DL the drug concentration rose quickly above the amorphous solubility and levelled off at about 25 min, with about 70% or more of the drug released; the 35% DL compact showed a lag of about 10 min and then reached about 20 times the amorphous solubility at 25 min (Figure 3a). The medium became cloudy with a bluish tinge, characteristic of drug-rich colloids, so the measured concentration includes colloidal as well as dissolved drug. R9·R10·C12 DLS found drug-rich particles of 193−244 nm at every DL from 5 to 35% (Table 3). R15

The normalized rates (Table 2, Figure 4a) make the congruence explicit. Lumefantrine left at 3.15−3.92 mg·min⁻¹·cm⁻² and PVPVA at 4.49−4.91, close to neat PVPVA (4.58 ± 0.06), giving polymer/drug ratios of 1.19−1.55. The authors read this as polymer-controlled dissolution ("suggests"). R13·C13·C14 At 40% DL the PVPVA rate dropped to 0.21 and at 45 and 50% DL to 0.04 and 0.004 mg·min⁻¹·cm⁻², and no drug release could be quantified. R13·R9 The authors conclude that the LoC is 35% DL: it is the last congruent loading tested, and the next one, 40%, already fails, so the threshold lies somewhere between 35 and 40% DL. C31·X17

The formation of drug-rich droplets is, for the authors, another marker of polymer-controlled release: a dispersion whose release is controlled by the drug has no thermodynamic driving force once the solution reaches the amorphous solubility, whereas a polymer-controlled one keeps delivering drug, which then separates as droplets. C31 The LoC of 35% DL is higher than the 5−25% DL reported for other drugs in PVPVA. The authors call this "consistent with" the absence of specific lumefantrine−PVPVA interactions, which earlier work had suggested favours release at higher DL, and add that ionization of lumefantrine on contact with the medium "may play a role". R28·C32

Table A. Surface-normalized release rates, lumefantrine / polymer, in mg·min⁻¹·cm⁻² (from Table 2; mean of n = 3, standard deviations in the paper). BQL: below the quantification limit. pH 6.8 phosphate buffer, 37 °C, 100 rpm. R13

Enteric polymers: no cliff, but a steep and continuous decline

With the enteric polymers there was no plateau of fast release followed by a sudden drop. HPMCAS and HPMCP compacts released drug linearly over the whole 2 h (Figure 3c, e), above the amorphous solubility at 5−20% DL and below it at 25 and 50% DL; at 50% DL release could be quantified only after 1 h. After 2 h, HPMCP had released about 78, 62 and 26% of the drug at 5, 10 and 15% DL, and HPMCAS about 64, 40 and 16%; the authors say HPMCP "consistently outperformed" HPMCAS at these loadings. R11·C15 At 20% DL the order reversed slightly: about 7% (HPMCP) against 10% (HPMCAS) after 2 h, read from Figure 3 as about 6.7 and 10 µg/mL. The Table 2 rates at 20% DL rank the two the other way round (0.071 against 0.06), and the printed pair "(∼10% vs ∼7%)" lists the HPMCAS value first while the sentence names HPMCP first; at 20% DL the two polymers are best treated as similar (note by the AI writer, extractor note X8). R10·R13·X8

The normalized rates fall steeply with DL for both drug and polymer: from 5 to 20% DL the drug rate dropped about 10-fold for both HPMCP (0.72 to 0.071) and HPMCAS (0.61 to 0.06). The authors contrast this with drops of about 2.5-fold for felodipine−HPMCAS dispersions from 10 to 40% DL and 4.2-fold for loratadine−HPMCAS from 5 to 50% DL in earlier studies. R13·C33·R29 Drug and polymer nevertheless fell together at low DL. The polymer/drug ratio was 1−1.04 for HPMCP from 5 to 20% DL and 1−1.11 for HPMCAS from 5 to 15% DL, then rose (HPMCAS 1.67 at 20% and 1.45 at 25% DL; HPMCP 1.24 at 25% DL; 5 and 3.5 at 50% DL), which the results section reads as similar normalized release at lower DL ("suggests") and polymer release outpacing the drug at higher DL ("indicating"). R13·C16 The discussion later states that drug and polymer released at a similar normalized rate up to 25% DL and that drug release at 50% DL was undetectable; the Table 2 ratios for HPMCAS and the quantified 50% DL drug rates (0.002) do not support either statement, so the table values should be used (extractor note X5). C34·X5

CAP and Eudragit L 100 dispersions did worst: no loading reached the amorphous solubility in 2 h (concentrations at 2 h of about 1.3−3.4 µg/mL), and at 25 and 50% DL neither drug nor polymer release could be quantified. The authors describe the polymer rates of these two as "consistently higher" than the drug rates, although at 15% DL CAP the two are equal (0.021 and 0.021), and the Eudragit L 100 ratios (1.15−1.29) lie in the range called congruent for HPMCAS and HPMCP (extractor note X6). R10·R11·R13·C17·X6 DLS of CAP and Eudragit L 100 media always gave a high polydispersity, read as no drug-rich droplets, consistent with release below the amorphous solubility. HPMCP and HPMCAS media held droplets of 80−92 nm and 122−141 nm at 5−20% DL, smaller than those from PVPVA. R15·C18

The neat polymer sets the ceiling

Comparing polymers at the same loading shows how much the carrier matters. At 5% DL lumefantrine left PVPVA compacts at 3.68 mg·min⁻¹·cm⁻², about 5 times faster than HPMCP compacts (0.72), and neat PVPVA dissolves about 5 times faster than neat HPMCP (4.58 against 0.87); the drug rates at 5% DL follow the order of the neat polymer rates for all five polymers (Figure 12). R13·C44 The authors conclude that the neat polymer release rate "is important in dictating the achievable release rate of the drug at a given DL" and that a faster-dissolving polymer is "advantageous to achieve higher DL systems with adequate release rates". C35·C38 Figure 12 plots the PVPVA bar at about 3.78 with a small error bar, which matches the 20% DL entry of Table 2 (3.79 (0.07)) rather than the 5% DL entry (3.68 (0.39)); the 5-fold comparison holds with either value (extractor note X2). R14·X2

The drug also changes the polymer's own release, but only for the enteric polymers. In the Meyerhofer plot (Figure 11: log polymer rate against DL, a format from photoresist research) PVPVA stays flat at 4.49−4.91 up to 35% DL and then drops by orders of magnitude; the enteric polymers decline from the first increment of drug: HPMCAS from 0.79 (neat) to 0.67 at 5% DL, HPMCP from 0.87 to 0.72, CAP from 0.56 to 0.25 and Eudragit L 100 from 0.23 to 0.09. R13·C36 CAP and Eudragit L 100 fall more steeply, and lumefantrine "appears" to slow them more. C37 By extrapolation the authors expect CAP and Eudragit L 100 at 25% DL to release about as slowly as HPMCP and HPMCAS at 50% DL, which "likely explains" why no drug release was detected there. C39 Near 50% DL the polymer rates of PVPVA, HPMCAS and HPMCP converge (0.004, 0.010 and 0.007) despite neat rates that differ more than five-fold, which "suggests" that drug-controlled dissolution dominates at high DL. R13·C40 The authors add, citing water contact angles of 30.6° for PVPVA and 58.6° for HPMCAS from another study, that the slower release from enteric polymers "could be due, at least in part," to their greater hydrophobicity. M2·M3·C45

How the drug interacts with each polymer

For PVPVA, ATR-FTIR of the powders showed small blue shifts of the carbonyl bands with DL (2 cm⁻¹ for the vinyl acetate band at 25% DL and up to 6 cm⁻¹ for the vinylpyrrolidone band at 50% DL), which the authors attribute "most likely" to less sorbed water at higher DL. In the rigorously dried films neither band (1736 and 1684 cm⁻¹) moved at any loading (Figure S1), so lumefantrine does not hydrogen bond to the PVPVA carbonyls. R16·C19·R17·C20 The SI figure labels the films 10, 25 and 50% DL, whereas the Methods list 10, 20 and 50%. X15

For the enteric polymers the carbonyl band also shifted to higher wavenumber: by 2−6 cm⁻¹ for HPMCAS and HPMCP, already 2 cm⁻¹ at 5% DL, and by 8 and 21 cm⁻¹ (CAP) and 13 and 21 cm⁻¹ (Eudragit L 100) at 25 and 50% DL, with no shift at 5 and 10% DL. Because these polymers self-associate through their carboxylic acids, the authors read the shift as "suggestive of" a change in hydrogen bonding within the polymer as lumefantrine is incorporated. R16·C22 The decisive evidence comes from two other measurements. After subtraction of the drug spectrum, all four 25% DL enteric dispersions showed extra bands between 1594−1560 and 1410−1343 cm⁻¹, the asymmetric and symmetric carboxylate stretches, which "provides evidence of an acid−base interaction" (Figure 7). R19·C25 And XPS found a protonated nitrogen component in every enteric dispersion and none with PVPVA (Table 4, Figure 8).

Table B. XPS of 25% DL compacts before and after partial dissolution: percentage of lumefantrine protonated (from Table 4; mean with SD, n = 3) and the share of surface carbon belonging to the drug (read from Figure 9, approximate). R20·R21

Protonation was largest with CAP and smallest with HPMCAS. Earlier work had found none with HPMCAS at 20 and 40% DL, so the authors note that here all enteric polymers protonated lumefantrine at 25% DL. R20·C26·R30 The solvent-evaporated dispersions thus already held some drug−polymer salt before contact with water, "in agreement with previous studies". C48

What happens at the compact surface

After 2 h of dissolution, the ATR spectra of 40, 45 and 50% DL PVPVA compact surfaces showed much weaker polymer carbonyl bands and much stronger lumefantrine bands at 1600−1550 cm⁻¹ than the initial dispersions, which "strongly suggest" a drug-enriched, polymer-depleted surface (Figure 6a). HPMCAS, HPMCP and CAP surfaces changed little, whereas Eudragit L 100 surfaces gained drug bands. R18·C23·C24 XPS is more sensitive (it probes about 10 nm) and showed more drug carbon at every surface after immersion (Table B). Protonation fell after dissolution for HPMCAS (19 to 3%), HPMCP (46 to 21%) and Eudragit L 100 (49 to 12%), "consistent with" drug enrichment, but not for CAP (55 to 54%). R21·R20·C26·C27 The authors use these data to attribute the cliff of high-DL PVPVA dispersions to drug enrichment of the surface, and to conclude that HPMCP and CAP "appeared effective in minimizing" amorphous−amorphous phase separation during dissolution. C67·C43 Two points weaken both readings (extractor notes X11 and X12): phase separation itself was not measured, and the XPS surface of the 25% DL PVPVA compact, which releases congruently, is the most drug-enriched of all, while no below-LoC PVPVA surface was examined by IR. X11·X12

Stability runs the other way

Under open-dish storage at 40 °C/75% RH the order of the polymers reversed. PVPVA dispersions showed small peaks of crystalline lumefantrine after 1, 2, 4 and 8 weeks at 50, 25, 10 and 5% DL (Figure 2b). Among the enteric polymers, the text reports drug crystallization only at 50% DL: HPMCAS after 1 week, Eudragit L 100 after 8 weeks, HPMCP after 32 weeks and CAP after 40 weeks. R6·C10 The CAP dispersions at 5, 10 and 25% DL did show peaks after 1, 4 and 16 weeks, but at positions that do not match lumefantrine; cryomilled neat CAP contained crystalline material with the same peaks, so these are not drug crystals, and CAP dispersions stored at room temperature with desiccant stayed amorphous for 36 weeks. R6·R7·C11

Table C. Crystallization on storage, from the text of the Results and Figure 2. R6

Two details of Figure 2 deserve care (extractor notes X9 and X10). The storage times reached by dispersions that did not crystallize are not given anywhere, so "more stable" is a relative statement. And Figure 2f, captioned as formulations that "showed crystalline drug peaks", includes HPMCP 5% DL after 64 weeks and 10% DL after 76 weeks, each with a small sharp peak near 27° 2θ, close to a reflection of crystalline lumefantrine; the text does not mention them. X9·X10

Free carboxylates against the critical ionization fraction

The literature data replotted in Figure S3 show that neat HPMCP dissolves faster than CAP at any fraction ionized, the gap widening as ionization falls: HPMCP still dissolves at about 78 µg·min⁻¹·cm⁻² at a fraction ionized of about 0.923 (pH 5.4), whereas CAP is down to about 15 at about 0.944 (pH 5.8). R24·C54 From these curves the authors take CAP's critical ionization fraction to be about 0.94 and, extrapolating Figure S4 to a minimum dissolution pH of about 5, HPMCP's to be "slightly less than 0.83". R27

Table D. The authors' explanation of why the 25% DL HPMCP dispersion still releases and the CAP dispersion does not (from the SI calculation, Figures S3 and S4, Table 2). R25·R27·R13

CAP has more acid groups than HPMCP, so more of them remain free (0.88 against 0.83), yet CAP needs a larger fraction ionized to dissolve at all; the free fraction falls below that need, and the 25% DL CAP dispersion releases nothing, while HPMCP sits just at its threshold and still releases, slowly. CAP, needing "a greater fraction of −COOH groups to ionize before dissolution can commence", is therefore "more susceptible" to an additive that ties up those groups, and its release falls faster with DL, "consistent with" the non-linear dependence in eqs 1 and 2. C57·C58 The authors consider and set aside a molecular-weight explanation: chain disentanglement is "unlikely" to limit the release of the neat polymers under these conditions. C55

Note by the AI writer (not discussed in the paper): the literature rates used in Figures S3 and S4 sit close to this paper's own neat-polymer measurements, about 568 µg·min⁻¹·cm⁻² for CAP at pH 6.8 against 0.56 mg·min⁻¹·cm⁻² measured here, which lends some support to using them; but the thresholds that decide the argument lie outside the range of those data, especially for HPMCP, whose lowest data point is at a fraction ionized of about 0.92. R24·R13·X13

Discussion

The authors' mechanism

The discussion assembles these observations into one picture for release and one for stability. Release from an enteric dispersion is limited by the dissolution of the polymer, and that dissolution requires carboxylates that are both ionized and hydrated. "It is important to consider that it is not only the ionization per se that leads to dissolution but also the accompanying hydration" of the carboxylate: if the pH is too low, or if the ionized groups are blocked by an additive, the polymer does not dissolve (Figure 13). C51 At pH 6.8, lumefantrine (pKa 8.73) "will ionize" at the polymer−solvent interface and competes with the sodium and potassium ions of the buffer for the polymer's carboxylates; XPS shows that some of these ion pairs already exist in the dry dispersion. C47·C48 The authors "believe" that the ion pair removes carboxylates from hydration, and they expect that less hydrated chains in the gel layer take longer to disentangle, so both polymer and drug release slow down. In the Conclusions this becomes: ionic interactions "are postulated to be responsible" for the diminished release, and the hydrophobic ion pair "is thought to reduce the hydration of the polymer carboxylate ions". C49·C53 They note that neither literature model of additives, percolation-site blocking (Reiser) or a hydrogen-bond-induced rise in effective pKa (Willson), "readily explains" their observations, because here the interaction is ionic. C50·C52

The authors' proposed chain for enteric polymer dispersions (Discussion, Figure 13); steps 2 and 3 were not measured. C53·C56·C57

PVPVA behaves differently because its dissolution does not depend on ionization. Its release stays at the neat-polymer rate as DL rises, "consistent with" the absence of drug−polymer interactions, until a critical DL is reached; there, dilution by the hydrophobic drug "likely prevents a sufficient volume of water from being absorbed to enable chain disentanglement", and a drug-rich layer at the compact surface "may also" block polymer release. C59 For the PVPVA cliff the authors also recall the earlier explanation from their group: water induces amorphous−amorphous phase separation in the hydrating compact, the PVPVA-rich phase dissolves first, and a drug-enriched surface remains as a barrier. C41

For stability the reasoning runs in parallel. Favourable drug−polymer interactions are generally considered important for miscibility and for inhibiting crystallization, and besides hydrogen bonding, halogen bonding and acid−base interactions can contribute. C60 Lumefantrine's hydroxyl forms an intramolecular hydrogen bond with its own nitrogen in the crystal structure; "if this interaction persists in the ASD", only very weak or no specific interactions with PVPVA remain. With a very hygroscopic polymer and a lipophilic drug, the authors argue, "these factors predispose" lumefantrine−PVPVA dispersions to amorphous−amorphous phase separation and then crystallization ("This can be rationalized ..."). C61·R33 The enteric polymers can hydrogen bond to the acceptor groups of lumefantrine and, in addition, form an acid−base interaction with its tertiary amine; this is offered as "a potential explanation for the improved physical stability". C62 The exception, the 50% DL HPMCAS dispersion that crystallized within a week, holds 1.61 molecules of drug per polymer acid group, and its crystallization "could be attributed to a low molar concentration of carboxylic acid groups". C63·R26

What it suggests for formulation

The authors draw three practical lessons. Release should ideally be polymer-controlled, because the neat polymer dissolves far faster than a lipophilic amorphous drug. C29 The neat polymer's dissolution rate sets the ceiling for drug release, so a faster-dissolving polymer allows a higher DL with adequate release. C35·C38 And for a basic drug, replacing a neutral polymer with an enteric one "may be a worthwhile trade-off from the improved physical stability point of view": PVPVA released lumefantrine rapidly but crystallized, while HPMCP and HPMCAS released more slowly but were more stable, "especially HPMCP ASDs". C65 The abstract frames this as a balance between physical stability and dissolution that "must be achieved" through polymer selection. C69

Note by the AI writer (not discussed in the paper): for a weakly basic drug the acid groups of an enteric polymer work in both directions. They protonate the drug and appear to help keep it amorphous, and the same ion pairs are blamed for slowing the polymer's dissolution, so in this set no polymer gave both PVPVA-like release and enteric-like stability. Within the enteric group, the polymer with the fastest neat dissolution and the lowest pKa (HPMCP) gave the best release at low DL and, in the text, the latest crystallization at 50% DL. The one dispersion with more drug than acid groups (HPMCAS, 50% DL) crystallized within a week, which makes the molar ratio of drug to polymer acid groups a quantity worth tabulating early in development, although here it rests on a single dispersion.

Limitations

Limitations stated by the authors. The authors say that the extent of ionization of drug and polymer in the solid state and in solution "warrants more in-depth studies" to understand their role in both stability and dissolution, and that the effect of additives on the dissolution of ionizable polymers "is not well understood". C64·C52 They present the hydrophobicity argument as partial ("at least in part"), the PVPVA ionization effect as possible ("may play a role") and the mechanism of diminished enteric release as postulated. C45·C32·C53

The AI writer's own judgement (1): what was and was not measured. The release patterns, the protonation and the stability ranking are measured directly and are consistent with each other; the printed polymer/drug ratios all reproduce from the printed rates, and the polymer rates in the Meyerhofer plot match Table 2. The mechanism linking them is not measured. Hydration of the carboxylates in the gel layer, the ionization state of lumefantrine at the dissolving surface and phase separation during dissolution or storage were all inferred. Protonation was measured by XPS on dry compacts at one loading (25% DL), and applying it to the hydrated interface is an assumption. The critical ionization fractions that decide the CAP versus HPMCP comparison come from 1979 literature data on neat polymers of unspecified lots, extrapolated well beyond the data for HPMCP. The authors' own word for the explanation, "can be accounted for", fits a plausibility argument rather than a test. X13·X14·C57

The AI writer's own judgement (2): where the text and the data part ways. Several summary statements are stronger than the tables: the discussion's congruence "up to 25%" for HPMCAS (Table 2 gives ratios of 1.67 and 1.45 at 20 and 25% DL), the "consistently higher" polymer rates for CAP (equal at 15% DL), and the "undetectable" drug release at 50% DL (quantified at 0.002). X5·X6 Smaller inconsistencies do not change the conclusions but matter for anyone reusing the numbers: the CAP protonation is 55% in Table 4 and 58% in the SI calculation; the PVPVA bar in Figure 12 matches the 20% rather than the 5% DL entry of Table 2; the Methods omit the 15% DL CAP and Eudragit L 100 dispersions that were tested; and the HPMCP solubility plotted at pH 4.5 in Figure S2 (about 55 000 µg/mL) exceeds the concentration of the about 1 mg/mL stock it was made from. X1·X2·X3·X4 How the normalized rates were calculated is not printed, and at 20% DL the ranking of HPMCP and HPMCAS depends on whether one uses the rate or the 2 h extent. X7·X8

The AI writer's own judgement (3): stability and surface data. The stability comparison is clear for PVPVA but open-ended for the enteric polymers, because the last storage times of dispersions without crystals are not stated and Figure 2f shows small, undiscussed peaks for HPMCP at 5 and 10% DL after 64 and 76 weeks. The surface data support drug enrichment during dissolution but not its role as the cause of the PVPVA cliff: the 25% DL PVPVA compact, below the LoC, shows the strongest enrichment by XPS, and no below-LoC PVPVA surface was examined by IR. Finally, the study uses one drug, one medium (pH 6.8 phosphate buffer, 2 h) and compacts rather than powders, so how far the ion-pair picture carries to other basic drugs or to biorelevant media is not addressed. X9·X10·X12·X11

Conclusion

The paper compares one neutral and four enteric polymers for the same lipophilic base and measures both halves of the formulation problem on the same dispersions. With PVPVA, lumefantrine and polymer leave the compact together at nearly the neat-polymer rate up to 35% DL, higher than the 5−25% DL reported for other drugs in PVPVA, and the dispersions then fail abruptly; but they crystallize within 1−8 weeks at 40 °C/75% RH. With HPMCAS, HPMCP, CAP and Eudragit L 100, lumefantrine is 19−55% protonated, release falls continuously and steeply from the lowest loading, and the order of release follows the order of the neat polymers' dissolution rates, while drug crystallization in the text appears only at 50% DL. R13·R6·R20·C31·C33·C35

The authors' explanation, that a hydrophobic drug−polymer ion pair takes carboxylates out of hydration and pushes an enteric polymer towards or below its critical ionization fraction, accounts for the ranking of the enteric polymers and for the failure of CAP at 25% DL, but it remains "postulated": hydration, interfacial ionization and the critical fractions themselves were not measured on these materials. The next tests the paper points to are direct measurements of drug and polymer ionization in the solid and in solution, and, beyond the paper, measurements of polymer hydration in the gel layer and of the true critical ionization fractions with and without drug. For formulation, the study shows in one data set that for basic drugs the choice between neutral and enteric polymers is a trade-off between release and stability rather than a choice of the better polymer. C53·C57·C64·C65