Salt formation in lumefantrine–copovidone dispersions: a higher glass transition buys stability and costs drug loading

Background and motivation

Many new drug candidates are large, lipophilic and poorly soluble, and amorphous solid dispersions (ASDs) are one of the main ways to deliver them: the amorphous drug, mixed at the molecular level with a polymer, dissolves to a supersaturated solution, and the polymer slows nucleation and crystal growth. The paper notes that the US FDA approved 16 ASD products since early 2010 and 19 between 2007 and 2017 (Jermain 2018, ref 2). Two properties decide whether an ASD works. It must stay amorphous on storage, and it must release its drug when it meets water.

For physical stability, the glass transition temperature (Tg) is the key solid-state property: the crystallization rate of an amorphous compound rises steeply above Tg because the supercooled liquid is far more mobile than the glass (refs 19–21). The usual remedy for a low-Tg drug is a high-Tg polymer, which lifts the Tg of the mixture (Ueda 2014, ref 22), and drug–polymer hydrogen bonds slow crystallization further (refs 12, 23–26). C1·C2 Lumefantrine is a difficult case on both counts. It is a weak base (tertiary amine pKa 8.73), very lipophilic (log P 8.34), has an amorphous Tg of 19.7 °C and no available hydrogen-bond donor, and in the authors' earlier study its PVPVA dispersions crystallized quickly at elevated temperature and humidity, whereas dispersions in which lumefantrine formed ionic interactions with an acidic polymer stayed stable (Hiew 2022, ref 10). M1·R2.a·C3·C4

Because its pKa is high, lumefantrine can in principle form salts with many acids; a pKa difference of more than 2 units between acid and base is the usual rule for salt formation (Tong 1998, ref 29). C5·E4 For amorphous drugs, salt formation is "a potential strategy" to reduce crystallization by raising Tg. Amorphous sodium indomethacin has a Tg about 75 °C above that of indomethacin free acid (Tong 1999, ref 30), the size of the alkali counterion tunes the rise (Tong 2002, ref 31), and amorphous propranolol salts crystallize at higher temperatures than the free base (Towler 2008, ref 19). C6 The authors point out two gaps. These limited studies say little about salts inside ASDs, and nothing about what a higher Tg does to release. The second gap is a real concern, because studies of PVPVA dispersions had given "some evidence" that systems with a Tg well above the dissolution temperature release poorly (Saboo 2020, ref 12; Yang 2021, ref 15; Que 2019, ref 32). C7

The paper therefore asks how salt formation, and the Tg rise that comes with it, affects the two ASD metrics at once. Lumefantrine was reacted with seven acids, the Tg of each amorphous salt was measured by differential scanning calorimetry (DSC), salts and their PVPVA dispersions were stored in open dishes at 40 °C/75% relative humidity (RH) and checked by powder X-ray diffraction (PXRD), and release was measured from compacts in a Wood's intrinsic dissolution apparatus. PVPVA was chosen as one of the most common neutral polymers in marketed ASDs (Wyttenbach 2017, ref 33).

Background concepts

Glass transition and mobility. Below its Tg an amorphous solid is a glass, with slow molecular motion; above it, it is a supercooled liquid in which molecules move and can find a crystal lattice. Water lowers Tg (it is a plasticizer). The authors use a rule of thumb from the literature (Tong 2004, ref 48): Tg falls by about 10 °C for every 1% w/w of absorbed water. E3 In this paper all Tg values are onset temperatures from modulated DSC. P5

Amorphous salts and disproportionation. An amorphous salt is a glass in which the drug is protonated and paired with a counterion. Ionic interactions usually raise Tg compared with the neutral free base. On storage a salt can either crystallize as the salt or disproportionate, that is, revert to the free base (and free acid). In PXRD the two are told apart by peak positions: peaks at the positions of crystalline lumefantrine or of the free acid indicate disproportionation. C11·C12·C13·C14

Congruent release and the limit of congruency. From PVPVA dispersions that release well, drug and polymer leave at the same normalized rate: the release is "congruent" and controlled by the polymer, which dissolves much faster than a neat amorphous lipophilic drug. C25 Above a drug loading (DL) that is specific to each drug–PVPVA pair, the drug release rate falls drastically ("falling off a cliff"). This has been ascribed to a drug-rich barrier layer on the compact surface, formed by amorphous–amorphous phase separation of drug and polymer after water enters, so that release becomes controlled by the drug. C27 The highest DL at which release is still congruent is the limit of congruency (LoC). Reported LoCs of PVPVA systems range from 5 to 35%, and lower drug Tg and a lack of specific drug–polymer interactions have been suggested to raise them. R22·C28 A difficulty the paper addresses directly is that Tg and hydrogen bonding usually move together: drugs with high Tg tend to carry hydrogen-bonding groups. For example, indomethacin, which hydrogen-bonds extensively to PVPVA and has the higher Tg, has an LoC of 10%, while its ester, which does not, has an LoC of 25% (Saboo 2020, ref 12). R23·C29

How a glassy polymer dissolves. Polymer dissolution is governed by three transport steps: water diffusion into the matrix, disentanglement of chains from the plasticized gel, and diffusion of the chains through the boundary layer (Narasimhan and Peppas, refs 43–45). C38 A polymer whose Tg lies above the experimental temperature must first absorb enough solvent to become rubbery. Narasimhan and Peppas expressed this as a critical polymer volume fraction $v_p^{*}$, which the paper prints as Eq. (1):

Here ${\rho }_p$ and ${\rho }_s$ are the densities of polymer and solvent, $T_g$ is the glass transition temperature, $T$ the experimental temperature, $\beta$ the expansion coefficient contribution of the solvent to the polymer and ${\alpha }_f$ the difference between the linear expansion coefficients of the polymer above and below its Tg; the ratio $\beta /{\alpha }_f$ was obtained by the method of Fujita and Kishimoto (ref 46). E1 In plain terms: the further the experimental temperature lies below Tg, the smaller $v_p^{*}$ becomes, so the more solvent the polymer must take up before its chains can start to disentangle. E1·C39 Once that point is passed, the disentanglement rate $k_d$ depends on hydration, Eq. (2):

where $B$ is a constant that depends on polymer molecular weight, solvent viscosity and temperature. In plain terms: above the critical point, the more the polymer is hydrated, the faster its chains disentangle. E2 The paper defines $v_p$ in Eq. (2) as the volume fraction of solvent in the polymer, while $v_p^{*}$ in Eq. (1) is a polymer volume fraction; the text later compares the two directly, writing that $v_p$ exceeds $v_p^{*}$, so the symbol is used in two senses. X18 For PVPVA (Tg 102.9 °C, true density 1.19 g/cm³ from ref 47) at 37 °C the authors estimate $v_p^{*}$ = 0.89, meaning that at least 11% water is needed before chain disentanglement can begin. R21·P17 The values of $\beta /{\alpha }_f$ and of the solvent density are not printed, so this number cannot be recomputed. E1

Methods in detail

The design is simple and its strength lies in the series: one drug, one polymer and a set of counterions chosen to spread the Tg of the amorphous drug as widely as possible. Every form then goes through the same tests. A1·A2

Making the salts and the dispersions

Lumefantrine and each acid were dissolved together at a 1:1 drug:acid mole ratio, except sulfuric acid, which was used at half the amount (2 lumefantrine : 1 sulfuric acid). Poly(ethylene) glycol 250 diacid (PEG 250 diacid), which carries two carboxylic acid groups, was used both at 1:1, which leaves free acid groups, and at 2:1. P1 The acids were benzoic acid (benzoate), PEG 250 diacid, benzenesulfonic acid (besylate), p-toluenesulfonic acid (tosylate), (1S)-(+)-10-camphorsulfonic acid (camsylate), hydrochloric acid (HCl) and sulfuric acid (sulfate) (Figure 1). M3–M9 Each salt was dissolved in a solvent chosen for it (Table 1: dichloromethane/methanol or dichloromethane/acetone 8:2 v/v, dichloromethane alone, or tetrahydrofuran for the tosylate), the solvent was removed by rotary evaporation at 45 °C, and the solid was dried overnight under vacuum. P1 Dispersions with PVPVA (Kollidon VA 64) were made the same way at 20 mg/mL solids, then cryo-milled and sieved below 250 µm. P2·M2 Loadings were matched to each salt's behaviour: 5–35% for the free base, 20–30% for benzoate and PEG 250 diacid, 10 and 15% for the sulfonates, 5–25% for HCl and 2–35% for the sulfate. A2 The paper does not say whether DL is calculated on the salt or on lumefantrine free base, which matters because the counterions differ widely in mass. X14

Tg, water uptake and solubility

Onset Tg was measured by modulated DSC (heating at 3 °C/min with ±0.636 °C modulation every 40 s), taking the second heating after a first cycle that removed thermal history; each value is the mean of two samples, with error bars showing the standard error. P5 Water sorption of the amorphous salts was measured by dynamic vapour sorption from 5 to 95% RH in 5% steps, and the water uptake of selected dispersions was measured in a single step to 95% RH at 40 °C. P13·P14 The equilibrium solubility of the salts was measured by stirring an excess in 50 mM pH 6.8 phosphate buffer at 37 °C for 48 h, ultracentrifuging and analysing the supernatant by HPLC. P3·P4

Physical stability

Freshly prepared salts and dispersions were stored in open dishes at 40 °C/75% RH and analysed by PXRD (4–40° 2θ) at set times, for up to 32 weeks. P6·P7 The sampling schedule is not printed; the times reported are the first time point at which crystalline peaks appeared. R7·R8 Disproportionation was judged by comparing peak positions with those of crystalline lumefantrine, the crystalline acid and the crystallized salt (Figures 5 and 6). R7·R8

Release testing

Compacts were dissolved in a Wood's intrinsic dissolution apparatus following the method of the earlier lumefantrine study (ref 10), in 250 mL of 50 mM pH 6.8 phosphate buffer at 37 °C; two formulations were also tested at 10 °C. Aliquots were withdrawn, diluted and analysed for lumefantrine by HPLC (n = 3). P8·P4 The surface-normalized release rate was taken from the linear part of each concentration–time profile. R12 For two formulations that showed a lag phase, PVPVA in the medium was also measured by the method of refs 34 and 35. P9·R13·R14 Partially dissolved compacts of these two formulations were stopped at 0, 10, 60 and 90 min, dried with air, coated with platinum and imaged by scanning electron microscopy (SEM) at 1000×. P11·P12 Compact mass, exposed area and stirring speed are given only by reference to ref 10. P8

Two further tests probe the interpretation. FTIR spectra of spin-coated dispersion films in the carbonyl region (1800–1550 cm⁻¹) look for hydrogen bonding between the salts and the PVPVA carbonyl groups. P10·R18 Films stained with Nile red were imaged by confocal microscopy at 37 °C and high RH to see whether drug and polymer phase-separate; these images are in the supplementary information, which was not available for this reading. P15·P16·R19

Experimental analysis

Salt formation raised the Tg, by counterion

No salt dissolved measurably in pH 6.8 buffer: the equilibrium solubility of every salt was below the HPLC detection limit of 0.002 µg/mL at 37 °C. R1 Every amorphous form showed a single glass transition, and every salt had a higher Tg than the free base (Figure 2). R2·C8 The rise depended strongly on the counterion. The two PEG 250 diacid salts and the benzoate, all carboxylates, reached only about 28–42 °C; the three sulfonates about 70–85 °C; the HCl salt about 96 °C; and the sulfate about 101 °C, close to PVPVA itself (102.9 °C). R2.b–j The text describes the benzoate rise as a "marginal elevation of ~10 °C", whereas Figure 2 places it about 13 °C above the free base. R2.c·X5 The excess-acid PEG 250 diacid salt (1:1) had a lower Tg than the 2:1 salt, about 28 against 42 °C. R2.b·R2.d

Table A. Tg of each neat amorphous form, the highest loading with good release and the surface-normalized release rates (read from Figures 2, 13 and 8, approximate; free base Tg from the text). R2·R20·R12

Dispersions of high-Tg salts keep the polymer's Tg

The Tg of the dispersions followed the Tg of the drug form (Figure 3). Free base dispersions fell from about 99 °C at 5% DL to about 73 °C at 35% DL, roughly 5 °C for each 5% of drug, and the carboxylate salt dispersions lay between about 62 and 79 °C at 20–30% DL. R3.a–n·C9 Dispersions of the sulfonate, HCl and sulfate salts stayed between about 101 and 105.5 °C at every loading tested, essentially the Tg of PVPVA. R3.o–ad·C9 In the discussion the authors write that all the dispersions have a lower Tg than neat PVPVA; that holds for the low-Tg forms, not for the high-Tg salts, some of whose dispersions lie slightly above the polymer. C43·X12

Release fails at lower loading as the salt Tg rises

Figure 7 contains the main data: release profiles of every form at every loading. Table B condenses them to the amount released after 90 min.

Table B. Drug released after 90 min (% of the drug in the compact) at each loading, read from Figure 7 (approximate, n = 3); bold marks the loading plotted in Figure 13, "–" means not tested, "lag" means release started only after about 20 min. R10

The free base released fast at every loading tested, reaching about 63–84% by 25–30 min and 67–81% after 90 min, with surface-normalized rates of about 3.2–3.9 mg·min⁻¹·cm⁻². R10.a–f·R12.a–f The carboxylate salts went to 20–25%: the benzoate released about 60% at 20% DL but only about 14% at 25%, and at 30% DL drug appeared only after about 20 min; both PEG 250 diacid salts also showed a lag at 30% DL. R10.g–n The text says the benzoate 20% DL dispersion had released "~60%" at 30 min; Figure 7c shows about 48% at 30 min and about 58–61% from 45 min on, so "~60%" is the plateau. R10.i·X5 The three sulfonates released well at 10% DL and almost nothing at 15%. R10.o–t Among them the tosylate is the weakest "good" formulation, plateauing at about 58% after 30 min. R10.q The HCl salt released about 78–93% at 5–15% DL, about 26% at 20% and nothing at 25%. R10.u–y The sulfate released about 94% at 2% DL, but at 5% DL only about 3.5%, about 0.1 mg after 90 min, and nothing at higher loadings. R10.z–ae·R11

The release rates draw the same line (Figure 8). Every formulation the authors class as near-complete release had a rate above 3 mg·min⁻¹·cm⁻². C17·R12 At the next loading, rates fell to about 0.1–0.84 mg·min⁻¹·cm⁻². R12 The highest rates, about 4.3–4.5 mg·min⁻¹·cm⁻² for the HCl salt at 5–10% DL and the sulfate at 2% DL, are, in the authors' words, similar to that of neat PVPVA; the neat PVPVA rate itself is not reported in this paper. R12.s·R12.t·R12.w·C45·X13 The besylate and camsylate dispersions, which plateau near 80%, are called "near complete" release in the text. R10.o·R10.s·X5

Figure 13 plots the highest loading with good release against the Tg of the neat drug form. The authors describe the upper limit as declining "precipitously" as Tg increases, and they single out the sulfate and the free base: neither shows convincing FTIR evidence of specific interactions with PVPVA, yet their LoCs are 2 and 35% with Tg of about 101 and 20 °C, so "Tg appears to play an important role in impacting release". R20·C32·C31 From this they conclude that drug Tg "may be an important factor impacting the achievable DL limit for PVPVA ASDs, even for an ASD without drug–polymer specific interactions". C33

Note by the AI writer (not discussed in the paper). Three points limit how far Figure 13 can be read. Each value is the last good loading of a coarse grid, so each LoC is an interval (for example 2–5% for the sulfate, 10–15% for the sulfonates), and the free base value of 35% is only a lower bound because no higher loading was tested. X7 No numerical criterion for "good release" is given; formulations that plateau near 60% or have rates of about 1.9–2.8 mg·min⁻¹·cm⁻² count as good, and drug and polymer release were compared for only two formulations, so these are drug-release limits rather than measured limits of congruency. X6 The order is not strictly monotonic: the PEG 250 diacid (2:1) salt (about 42 °C) reached 25% against 20% for the benzoate (about 33 °C), and the HCl salt (about 96 °C) reached 15% against 10% for the cooler sulfonates. X8 The trend is clear between the two groups, low-Tg carboxylates and high-Tg strong-acid salts, but the groups also differ in counterion chemistry. X9

In lag-phase dispersions the polymer leaves first

To understand the lag phases, PVPVA release was measured for the 30% DL benzoate and PEG 250 diacid (2:1) dispersions (Figure 9). R13·R14 From the benzoate dispersion about 20% of the PVPVA had dissolved after 30 min while almost no drug had appeared; polymer release then slowed (about 24% at 45 min) and rose again to about 50% at 90 min, as drug release began and reached only about 8.5%. R13.a–b The authors call this behaviour "indicative of incongruent release" and suggest that the mismatch "could account for" the low overall drug release of about 10%. C18·C19 For the PEG 250 diacid (2:1) dispersion about 9% of the polymer left in the first 30 min before any drug, after which both rose roughly linearly to about 51% (polymer) and 34% (drug) at 90 min. R14.a–b

The authors also estimate how much the early polymer loss enriches the compact in drug: to about 33% DL for the PEG 250 diacid (2:1) dispersion, which "could explain" its steadier release, and to "~57%" for the benzoate dispersion. R15·C20 The second number does not follow from the data given. With 20% of the polymer gone, a 30% DL compact holds 30 parts drug and 56 parts polymer, an effective DL of about 35%; counting 20% of the whole compact as lost polymer, as the PEG 250 diacid calculation appears to do, gives about 37.5%. X3 The direction of the argument stands, but the size of the enrichment is far smaller than printed. X3 The PEG 250 diacid (2:1) 30% DL formulation also released more drug in the run of Figure 7d (about 36% at 60 min and 44.5% at 90 min) than in the run of Figure 9b (about 12.5 and 34%), so lag-phase values vary between runs. X16

SEM of the partially dissolved compacts (Figure 11) shows the surfaces behind these curves. The benzoate compact had large pits after 10 min and a rough, irregular surface, more pitted at 60 min; the authors "postulated" that the pits form when a discrete PVPVA-rich phase separates and dissolves first, since no drug was released at that stage. R17.b–c·C22 Agglomerated structures on its surface appeared at the time drug release picked up, after about 45 min. R17.d·C23 The PEG 250 diacid (2:1) compact was smoother, with pits 10–30 µm across at 10 min that became fewer at 60 and 90 min. R17.f–h

A cold medium stops release

To test Tg relative to the dissolution temperature directly, two formulations that released well at 37 °C were dissolved at 10 °C (Figure 10). A4·R16 At 37 °C the benzoate 20% DL dispersion reached about 59 µg/mL (about 60% of its drug) and the besylate 10% DL dispersion about 38–39 µg/mL (about 80%); at 10 °C both stayed at about 0–1 µg/mL over 120 min. R16.a–d The authors conclude that the dissolution temperature relative to Tg "appears to play a role" in release, in line with refs 12 and 15. C21

Note by the AI writer (not discussed in the paper). Cooling slows the polymer as well as the drug: at 10 °C PVPVA lies even further below its Tg, and by Eq. (1) it needs more water before its chains can disentangle. No polymer release, no neat PVPVA and no free base dispersion was measured at 10 °C, so this experiment cannot separate a drug-Tg effect from a slower polymer. X10·E1

Physical stability runs the other way

All freshly prepared salts and dispersions were X-ray amorphous. R7·R8 Table C collects the first appearance of crystalline peaks in open-dish storage at 40 °C/75% RH.

Table C. First crystalline peaks at 40 °C/75% RH in open dishes (text of Sections 4.4 and 5.3, Figures 5 and 6; Tg from Figure 2, approximate). R7·R8·R9·R2

Broadly, the high-Tg forms resisted crystallization longer. The neat carboxylate and besylate salts crystallized within 4 weeks, the camsylate only after 32 weeks, and the HCl and sulfate salts not within 16 weeks; the recrystallized neat salts had peak positions different from crystalline lumefantrine, which the authors read as no disproportionation. R7·C11 Among the dispersions, the carboxylate salts crystallized within a week at 20–30% DL, while the camsylate held out for 12 weeks at 10% DL, about 10 weeks longer than the free base at the same loading. R8.a–c·R8.f·R9·C37 Several dispersions crystallized with free-base peaks: the 2:1 PEG 250 diacid salt and the besylate disproportionated, and the benzoate did so to some extent. C12·C13·C14 The authors conclude that the drug or drug-salt Tg "appears to be an important predictor" of physical stability, especially at higher loading. C37

The record has inconsistencies worth knowing. The sulfate 5% DL dispersion is said to remain X-ray amorphous up to 16 weeks in Section 4.4 and to show crystalline peaks after 12 weeks in Section 5.3, and Figure 6f shows a weak peak at the main lumefantrine position. R8.h·X1 The free base 5% DL dispersion (from ref 10) is said to crystallize after 8 weeks in one place and at 4 weeks in another. R9·X2 The Figure 5 caption lists the HCl salt among salts that crystallized within 32 weeks, but its pattern in Figure 5 shows only an amorphous halo. R7·X4 Note by the AI writer: the besylate is a clear exception to the Tg ranking, a salt with a Tg of about 70 °C whose 10% DL dispersion crystallized in 3 days, faster than the free base, with disproportionation; the text's remark that crystalline peaks appeared "as soon as 3 days" for low-Tg salts refers in fact to this high-Tg salt. X11·C36

Little evidence of drug–polymer interactions

FTIR of the dispersion films (Figure 12) showed the PVPVA carbonyl bands at 1736 cm⁻¹ (vinyl acetate) and 1684 cm⁻¹ (vinylpyrrolidone) unshifted at every loading for every salt except the tosylate. R18 Peak-height ratios changed with loading for several dispersions (Figure S4), most visibly for the excess-acid PEG 250 diacid (1:1) salt, which the authors read as at most "a minor extent" of interaction. R18·C24 They conclude that the stoichiometric salts interact with PVPVA less than drugs studied before, and that neither the free base nor the sulfate dispersion shows compelling evidence of specific interactions. C30·C31 Section 5.2 names the "PEG 250 diacid/sulfate salts with an unreacted acid group" as showing some hydrogen bonding, which sits awkwardly with the 2:1 sulfate stoichiometry, with the same paragraph's statement about the sulfate and with the tosylate being the only salt with a peak shift. X17

Water uptake and phase separation

The neat amorphous salts took up little water (Figure 4): at 95% RH, about 0.8% for the benzoate and 1.4–2.0% for the sulfonates and HCl, 3.3% for the sulfate, about 3.9% for PEG 250 diacid (2:1) and 6.6% for PEG 250 diacid (1:1), which rose steeply above 85% RH and liquefied. R4·C10 The dispersions, by contrast, took up 35–55% water at 40 °C/95% RH, more at lower loading (Table S1). R5 PVPVA is quoted at about 45% water at 95% RH. R6 Note by the AI writer: the upper dispersion value exceeds the PVPVA value, although lower loading means more uptake; the temperature of the PVPVA value is not given, so the two cannot be compared directly. X19 In films at high RH (SI Figures S6–S8), the free base and benzoate dispersions phase-separated, while the sulfate dispersion stayed visually homogeneous. R19 The authors attribute the free base's behaviour to the lack of specific interactions and its high lipophilicity, which "predisposed" it to water-induced phase separation. C35

Discussion

Why would a high-Tg salt stop release?

The authors first rule out the simplest explanation. Adding drug lowers the Tg of the low-Tg dispersions below that of PVPVA, which by Eq. (1) should make them easier to plasticize and faster to dissolve, although less water uptake caused by the lipophilic drug "may" offset this. C43 Taking the rule that each 1% w/w of water lowers Tg by about 10 °C, and water uptakes of 35–55% at 95% RH, every dispersion should be rubbery at 37 °C once wet, so the release behaviour is "difficult to rationalize" from the wet Tg of the dispersion and the disentanglement model alone. E3·R5·C44 The authors also note an open point of their own: how drug additives change the overall Tg and the disentanglement kinetics of PVPVA "has not been evaluated and requires further consideration". C41

Their explanation instead looks at the two phases that form when drug and polymer separate in the gel layer. The chain below follows Sections 5.1 and 5.5.

The authors present this as "a potential explanation" for a phase separation that is "thermodynamically favorable but is arrested by asymmetry in the dynamics of the two phases", a concept established for polymer–solvent blends and other systems (refs 49, 50). C47 Two observations, they write, support it: dispersions that released well at 37 °C released poorly at 10 °C, and the low-Tg free base and benzoate dispersions phase-separated in humid films while no obvious separation was seen in the high-Tg sulfate dispersion. C48·R16·R19 For higher-Tg salts, any phase separation after water uptake "is expected to become increasingly detrimental to release", and the sulfate 5% DL dispersion illustrates the end point: physically stable, but releasing almost nothing. C49·C50

The stability side of the trade-off

On storage the same mobility works the other way. Without specific interactions and with a log P of 8.34, lumefantrine is "predisposed" to water-induced phase separation from the hygroscopic polymer in humid conditions; once separated, a low-Tg drug-rich phase has lost the polymer's protection and crystallizes, which is worst at higher loading. C35·C36 A high-Tg salt keeps the drug-rich phase glassy at 40 °C/75% RH and delays crystallization, as the camsylate and sulfate results show. C37·R8.f·R8.h Put together, Tg elevation by salt formation improved physical stability but lowered the LoC, which "potentially limits" the loading with rapid release; the authors' summary is that "it appears that there is a trade-off" between stability through Tg elevation and good release at higher loading. C52·C53

What it suggests for formulation

The authors' recommendation is short: when a salt is used in an ASD, "attention should be paid to the Tg of the salt, since it may show opposing effects on physical stability and drug release, at least for PVPVA-based ASDs". C54 Note by the AI writer (not discussed in the paper). For a low-Tg weak base like lumefantrine, choosing a high-Tg counterion to cure crystallization can cut the workable loading several-fold, from 35% to 2–15% in this series, so release should be tested at the intended loading before a counterion is chosen. R20 The data also show that Tg is not the only lever: the HCl salt released well at 15% DL while the cooler sulfonates failed at 15%, and the besylate dispersion crystallized faster than any other, with disproportionation. R20.h·R8.d·X8·X11 A low-Tg carboxylate salt keeps loadings of 20–25% but crystallizes within a week in open-dish stress storage, so it would need protection from humidity rather than a different counterion. R8.a–c

Limitations

Stated by the authors. The conclusions are limited to PVPVA ("at least for PVPVA-based ASDs"), and the effect of drug additives on the Tg and disentanglement kinetics of PVPVA, and so on the dissolution rate, "has not been evaluated". C54·C41 They describe their mechanism as "a potential explanation" and the role of Tg in the release cliff as "not well understood". C47·C28

The AI writer's own judgement. The release data are direct and cover nine drug forms, and the split between a low-Tg group releasing to 20–35% DL and a high-Tg group releasing only to 2–15% is clear. R10·R20 The weaknesses lie in how that split is turned into a law and a mechanism:

• The loading limits are last-good points of a coarse grid, with no numerical criterion for good release and with congruency checked for only two formulations; the free base value is a lower bound. X6·X7
• Within each group the order is not monotonic, and Tg is confounded with counterion class (all low-Tg salts are carboxylates, all high-Tg salts strong-acid salts), with hygroscopicity and with disproportionation. X8·X9
• The mechanism rests on the Tg of hydrated drug-rich phases and on a barrier at the dissolving surface, neither of which was measured; the phase-separation images are in SI that was not available, with loading and RH not stated in the main text. X20
• The 10 °C experiment lacks a polymer or free base control. X10
• Several printed numbers conflict or cannot be reproduced: the sulfate and free base 5% DL crystallization times, the HCl salt in Figure 5, and the "~57%" effective loading. X1·X2·X4·X3
• The basis of DL (salt or free base) is not stated, and release was measured without stated filtration while the salts' solubility is below 0.002 µg/mL, so "released" means total drug in the medium, including colloidal drug. X14·X15

Conclusion

This paper uses one low-Tg weak base and a series of counterions to vary the Tg of an amorphous drug from about 20 to about 101 °C while keeping the polymer and the tests fixed. R2·C55 In PVPVA dispersions, the higher-Tg salts generally crystallized later, but they also released well only at much lower loadings: 2–15% against 20–35% for the low-Tg forms. R7·R8·R20 Polymer leaving before drug in lag-phase dispersions and the loss of release at 10 °C fit the authors' picture of a poorly hydrated, glassy drug phase that cannot separate and leave with the polymer, but that picture is an untested "potential explanation", and Tg in this series moves together with counterion chemistry. R13·R16·C47·X9·X20 The practical message is the authors' own: in PVPVA, a salt's Tg may help stability and hurt release at the same time. C54