Two additives at once: how polymer pairs and polymer–surfactant pairs change the crystal growth of ritonavir

Background and motivation

Many poorly water-soluble drugs are formulated as amorphous solids because an amorphous form dissolves to a concentration above the crystalline solubility, that is, it generates a supersaturated solution (Konno 2008, ref 10; Matteucci 2007, ref 11). C2 A supersaturated solution is unstable: the drug tends to nucleate and grow crystals until the concentration falls back to the crystalline solubility. Polymers are therefore added to delay crystallization (refs 12–15), and surfactants are added to help processing (ref 16) or dissolution (ref 17). C2 A real formulation therefore contains several additives at once, and these additives can interact with each other in solution. Polymer/polymer and polymer/surfactant pairs are known to form complexes whose properties differ from those of the components; examples cited by the authors include poly(ethylene glycol) (PEG) with polysorbate 80, hydroxypropyl methyl cellulose (HPMC) with poloxamer or with polysorbate 80, poly(vinylpyrrolidone) (PVP) with sodium lauryl sulfate, and poly(methacrylic acid) or poly(acrylic acid) with PEG and PVP (refs 4, 22–27). C3 Mixed carriers have also improved drug delivery: a felodipine dispersion made with HPMC, poloxamer and polyoxyethylene hydrogenated castor oil dissolved much faster and reached higher solubility (Won 2006, ref 22), and a PEG–polysorbate 80 dispersion improved the dissolution and bioavailability of a poorly soluble compound, LAB68 (Dannenfelser 2004, ref 23). C3 A hydrophilic additive used with a polymeric carrier "may help" prevent precipitation or keep a fine crystalline precipitate from agglomerating (Pouton 2006, ref 28). C4

The authors point out that the effect of such additive combinations on crystallization had not been studied systematically, although additives alone are well known to accelerate, inhibit or completely suppress nucleation and growth depending on how they change the solution and the crystallizing solute (refs 29, 30). C5 They argue that the interplay "could be very influential" on dosage-form performance, because a combination can change the system's properties in a favorable or an adverse direction. C5 The study builds directly on the group's earlier work: the same polymers had been tested one by one as growth inhibitors of ritonavir (Ilevbare 2012a, ref 31), where cellulose acetate propionate adipate 0.85 (CAP Adp 0.85) was the most effective cellulose ester (Ilevbare 2012b, ref 38), and polymers had been found to be less effective at higher supersaturation (refs 38, 39). C18·C26 Single additives had also struggled to stop desupersaturation when seed crystals were present at high supersaturation (refs 38, 39). C46

The paper sets out to measure how binary combinations of additives, polymer with polymer and polymer with surfactant, change the solution crystal growth of ritonavir, and to probe the dominant force between the additives by changing the ionic strength of the medium. A1·A4·A6 Ritonavir was chosen as a poorly water-soluble model compound; it is a very weak base (pKa 1.8 and 2.6) and is un-ionized at the test pH of 6.8. M1·C7 Its most stable polymorph, Form II, was used as seed material. R4·M1

Background concepts

Supersaturation. The driving force for crystal growth is expressed as the supersaturation ratio (Eq. 2):

Here $c$ is the drug concentration in the supersaturated solution and $c_0$ the equilibrium (crystalline) solubility at the same temperature. E2 In plain terms: $S$ says how many times more drug is dissolved than the crystal can hold in equilibrium. The Form II solubility of ritonavir at pH 6.8 and 37 °C was 1.3 ± 0.10 µg/mL in 100 mM sodium phosphate buffer and 2.3 ± 0.04 µg/mL in 50 mM buffer, so 10 µg/mL corresponds to $S$ of about 7.7 in 100 mM and 4.3 in 50 mM buffer. R1·R2·R3 The paper prints the 100 mM value both as 7.6 and as 7.7, and the 20 µg/mL value as 15.4 and 15.3; Eq. 2 gives 7.69 and 15.38, so these are the same conditions rounded differently. X1

Seeded desupersaturation and the growth rate. If crystals of the drug are added to a supersaturated solution, the excess solute can be used up in two ways: the seeds can grow, or new crystals can form by primary or secondary nucleation. C14 When nucleation is negligible, the rate at which the dissolved concentration falls is a measure of the bulk growth rate of the seeds. The authors took the slope of the concentration–time curve over the first 2 min as the initial crystal growth rate $R_g$, in µg mL−1 min−1. E4·P8

The effectiveness ratio. To compare additives, the paper uses the ratio of the growth rate without additive to the growth rate with it (defined in words in the caption of Figure 3):

Here $R_{g_0}$ is the growth rate of ritonavir without polymer and $R_{g_p}$ the growth rate with a polymer or polymer combination. E3 In plain terms: a ratio of 3 means growth is three times slower than without additive; above 1 the additive inhibits growth, and below 1 it accelerates growth. E3·C16 The ratios printed in Table 5 can be recomputed from its rates; for example 3.6/1.1 = 3.27, printed as 3.3. E3·R16

What "synergy" means in this paper. The authors call a pair synergistic when the pair is more effective than either of its polymers alone. C17 This is a comparison with each component, not with an additive or multiplicative prediction. In the screen, each polymer alone was used at 5 µg/mL, and each pair at 5 µg/mL of each polymer, so a pair carries twice the total polymer of a single-polymer run; the authors tested whether this matters for one pair (Figure 4). P8·R12·X7

The solubility parameter as a hydrophobicity scale. To rank the polymers by hydrophobicity, the authors estimated solubility parameters with the group-contribution method of Fedors (Eq. 1):

Here $\Delta e_i$ and $\Delta v_i$ are the contributions of atom or group $i$ to the energy of vaporization and to the molar volume, $\Delta E_v$ is the energy of vaporization and $V$ the molar volume. E1 In plain terms: the solubility parameter is the square root of the cohesive energy per volume, built up from the chemical groups of the repeat unit, so it needs only the structural formula. E1·P7 A high value means a polar, hydrophilic polymer. C9 The values ran from 20.05 MPa^1/2 for cellulose acetate butyrate adipate 0.25 (CAB Adp 0.25, the most hydrophobic, rank 14) to 32.24 MPa^1/2 for poly(4-vinylpyridine N-oxide) (PVPdn-O, the most hydrophilic, rank 1) (Table 3). M6·M15·C9 The group contributions themselves are not printed, so the values cannot be recomputed from the paper. E1

How additives slow crystal growth. If an additive does not change the supersaturation or the solution viscosity, it has to adsorb on the crystal surface to change growth (Mullin, ref 42). C32 For two additives, the authors list the possibilities: both adsorb independently, either at different sites or competing for the same sites; or they also form complexes in solution, which may adsorb as such or alongside the free additives, again with competition for sites (refs 43–45). C33 An adsorbed additive can also accelerate growth when its thermodynamic effect, a lower interfacial tension at the crystal surface, outweighs its kinetic effect of blocking kink sites (refs 29, 53–55). C51

Crystal habit and fast-growing faces. Ritonavir Form II grows as needles. A Bravais–Friedel–Donnay–Harker (BFDH) morphology prediction (Mercury 3.0.1, from ref 31) places the fastest growth along the a-axis, the length of the needle, with four fast-growing faces, (−1, 0, −1), (−1, −1, 0), (−1, 0, 1) and (−1, 1, 0), perpendicular to it. R5·P6 These faces expose carbonyl oxygen and thiazole nitrogen atoms, which in ref 31 were proposed to hydrogen-bond with the amide group of poly(N-isopropylacrylamide) (Pn-IPAAmd). R5·C13 The aspect ratio, length divided by width, is used as a measure of how needle-like the crystals are. E5

Critical micelle concentration. Above its critical micelle concentration (CMC) a surfactant forms micelles that can solubilize drug; below it the surfactant is present as single molecules. The CMCs used by the authors are 0.018 mM, about 24 µg/mL, for Tween 80 in water (ref 40) and 0.02 wt %, about 212 µg/mL, for TPGS (tocopherol polyethylene glycol 1000 succinate) at 37 °C (ref 41). M16·M17 At 5 µg/mL, both surfactants were well below the CMC. C52 Note by the AI writer (not discussed in the paper): 0.02 wt % corresponds to about 200 µg/mL at unit density rather than 212 µg/mL; either way the CMC is far above the concentration used. X16

Methods in detail

The study is one experiment, the seeded growth rate of ritonavir, repeated under many conditions. The flow below shows how the pieces fit together.

Materials

Ritonavir came from Attix Corporation and was used as received; the seeds were sieved below 250 µm. M1 Fourteen polymers were tested (Table 1). M2–M15 Five were synthetic: PVP K29/32, poly(vinylpyrrolidone vinyl acetate) K28 (PVPVA), poly(allylamine) (PAlAmn), Pn-IPAAmd and PVPdn-O. M2–M6 Three were commercial cellulose polymers: cellulose acetate phthalate (CAPh), HPMC 606 grade and hydroxypropyl methyl cellulose acetate succinate AS-MF grade (HPMCAS). M7–M9 Six were cellulose adipate esters made in-house: cellulose propionate adipate (CP Adp), cellulose acetate 320S adipate (CA 320S Adp), cellulose acetate propionate 504−0.2 adipate 0.85 and 0.33 (CAP Adp 0.85, CAP Adp 0.33) and cellulose acetate butyrate 553−0.4 adipate 0.81 and 0.25 (CAB Adp 0.81, CAB Adp 0.25). M10–M15 Figure 2 shows their structure: a cellulose backbone carrying acetate, propionate, butyrate and adipate groups, not regioselectively placed; the adipate ends in a carboxylic acid. M10 CAP Adp 0.85 and CAP Adp 0.33 differ only in the degree of substitution of the adipate. M14 The two surfactants were TPGS (Eastman) and Tween 80 (Sigma-Aldrich). M16·M17

Solubility and solid-state checks

Equilibrium solubility was measured by adding excess ritonavir to 100 mM sodium phosphate buffer, pH 6.8, in which the additive had been dissolved at 5 µg/mL (for pairs, 5 µg/mL of each), holding 48 h at 37 °C, separating the supernatant by ultracentrifugation at 40,000 rpm (274356g) and assaying it by HPLC with UV detection at 240 nm. P1·P2 The seed crystals were characterized by SEM (50 crystals measured per sample, aspect ratio from the average length and width), by DSC (melting point from a first heat at 5 °C/min, Tg from a second heat at 20 °C/min) and by PXRD (5–35° 2θ), the pattern being compared with that predicted for Form II from the Cambridge Structural Database entry YIGPIO01. P3·P4·P5

The seeded growth experiment

Each run used 50 mL of sodium phosphate buffer, pH 6.8 (100 mM unless stated), in a jacketed beaker at 37 °C, stirred at 400 rpm with an overhead axial-flow impeller. P8 The additive was dissolved in the buffer first, then 0.010 g of seed crystals was added and equilibrated at 37 °C. P8 Supersaturation was created by adding ritonavir from a 4 mg/mL methanol stock (0.25 mL into 50 mL, a 1:200 methanol-to-buffer ratio, which the authors state did not change the solubility) to give 5, 10 or 20 µg/mL. P8·C6 The concentration was followed in situ by a UV fiber-optic probe (5 mm path) every 5 s for 30 min, using second-derivative spectra to reduce scattering by the crystals; the calibration was linear from 1 to 30 µg/mL. P9 All runs were done in triplicate. P8 Note by the AI writer (not discussed in the paper): 0.25 mL of a 4 mg/mL stock in 50 mL gives 20 µg/mL, so the volumes or stock concentrations used for 5 and 10 µg/mL are not stated. X13

Checking that the rate is growth, not nucleation

The authors tested three things. C14·C15 First, the size distribution of seeds recovered after growth (no polymer, 10 µg/mL) was monomodal, with no second population of small crystals (Figure S4). R7 Second, they compared the increase in seed volume expected from the solute removed (end concentration 1.5 µg/mL, 0.010 g of seeds, no new crystals assumed), 3.17 × 10−17 m³, with the increase computed from the measured seed dimensions before and after growth, 7.19 × 10−17 m³, and called the two similar. R8·C14 Third, they measured nucleation without seeds: at 10 µg/mL the solution did not desupersaturate within 8 h (Figure S5), and at 20 µg/mL nucleation took about 2 h, both far longer than the 30-min growth runs. R9·C15 The polymer combinations did not shorten the induction time; CAP Adp 0.85/HPMC lengthened it to about 3 h. R9·C15

The variations

The screen compared 14 single polymers at 5 µg/mL with 13 pairs at 5 + 5 µg/mL, at 10 µg/mL ritonavir in 100 mM buffer (Figure 3). A1·R11 Four further experiments then changed one factor at a time. Dose: CA 320S Adp and CP Adp were also tested alone at 10 µg/mL, and CA 320S Adp was titrated from 0 to 7.5 µg/mL into 5 µg/mL CP Adp (Figures 4 and 5). A2·A3 Buffer: eight pairs were tested in 50 and 100 mM sodium phosphate (Figure 6). A4·P12 Supersaturation: ritonavir alone and four pairs containing CAP Adp 0.85 were tested at 5, 10 and 20 µg/mL (Figure 7), and at 20 µg/mL also CAP Adp 0.85, Pn-IPAAmd and PVP alone (Table 5). A5 Surfactants: TPGS and Tween 80 alone and Tween 80 with CAP Adp 0.33 or CAP Adp 0.85 (Figure 8). A6 The seed habit after growth was measured with and without CAP Adp 0.85, Pn-IPAAmd and their pair at $S$ 7.6 and 15.4 (Table 4). A7·R6

Experimental analysis

Baseline: the additives barely change solubility, and the seeds stay Form II

At 5 µg/mL, the additives tested left the ritonavir solubility at 1.3–1.8 µg/mL against 1.3 ± 0.10 µg/mL without additive (Table 2): CAP Adp 0.33 1.5 ± 0.10, CAP Adp 0.85 1.3 ± 0.02, Pn-IPAAmd 1.3 ± 0.01, CAP Adp 0.85/CAP Adp 0.33 1.5 ± 0.09, CAP Adp 0.85/Pn-IPAAmd 1.8 ± 0.20 and Tween 80 1.3 ± 0.02 µg/mL. R1 The authors conclude that the solubility "did not change significantly", so that the additives act on growth and not on the driving force. C8 Note by the AI writer (not discussed in the paper): no statistical test is reported, and the highest value belongs to the most effective pair; if 1.8 µg/mL were used as $c_0$, the initial $S$ at 10 µg/mL would be about 5.6 rather than 7.7 for that pair. No TPGS solubility value is printed, although the text states that neither surfactant changed the solubility. X6 The seeds were needle-like Form II, with a melting point of 121.0 °C and a glass transition of the cooled melt at 50 °C, in line with Law et al. (2001, ref 35), and they were still Form II after the growth experiments. R4·M1·C10

Finding 1: most polymer pairs inhibit growth more than either polymer

Without additive, the initial growth rate at 10 µg/mL in 100 mM buffer was 1.56 µg mL−1 min−1, and desupersaturation was complete after 30 min. R10 Figure 7 shows this point at about 1.4 µg mL−1 min−1, inside its error bar (about 1.08–1.71); the text value is the one stated. R10·X5 Table A lists all 13 pairs of Figure 3.

Table A. Effectiveness ratio $R_{g_0}/R_{g_p}$ of single polymers (5 µg/mL) and 1:1 pairs (5 µg/mL of each) at 10 µg/mL ritonavir, 100 mM buffer, 37 °C; triplicate means read from Figure 3 (approximate). R11

Ten of the 13 pairs beat both of their polymers, which the abstract calls "surprisingly" the majority. C17·R11 The strongest single inhibitors, CAP Adp 0.85 (about 3.0) and Pn-IPAAmd (about 3.2), together reached about 12.1; the text puts it as a 12-fold decrease of the growth rate against about 3-fold for each polymer. R11.a·C19 Five of the ten effective pairs combined CAP Adp 0.85 with a synthetic polymer. C18 Polymers that did little alone still helped in a pair ("Interestingly"): PVP, with a ratio of about 1.0 and so no effect alone, raised CAP Adp 0.85 from about 3.0 to about 4.35, and PVPdn-O (about 0.85 alone) and PAlAmn (about 0.9 alone) gave about 3.9 and 4.1. R11.c–f·C20 Chemically similar cellulose esters also gave synergy in some cases ("Somewhat surprisingly"): CA 320S Adp/CP Adp reached about 4.8 and CAP Adp 0.85/CAP Adp 0.33 about 4.5. R11.b·R11.g·C21

Finding 2: for one pair, the gain is not just more polymer

Because each pair carried 10 µg/mL of polymer and each single run 5 µg/mL, the authors tested CA 320S Adp and CP Adp alone at both concentrations (Figure 4, Table B). R12·X7

Table B. Dose control for CA 320S Adp and CP Adp, 10 µg/mL ritonavir, 100 mM buffer; read from Figures 4 and 5 (approximate). R12·R13

Doubling either polymer alone changed its ratio only a little, which the authors call "relatively comparable", while the pair at the same total of 10 µg/mL was more than twice as effective as either polymer at 10 µg/mL. R12·C23·C24 In the titration, the first 0.5 µg/mL of CA 320S Adp raised the ratio about 2.5-fold, from about 1.68 to 4.31, and further CA 320S Adp did not raise it significantly (Figure 5). R13·C25 This control was run for this pair only. X7

Finding 3: some pairs add nothing

Three pairs did not beat their better polymer. CAP Adp 0.85 with HPMCAS gave about 2.7 and with CAPh about 2.85, against about 3.0 for CAP Adp 0.85 alone; the authors write that these pairs "did not inhibit crystal growth any further". R11.j·R11.k·C22 CAB Adp 0.81/CAB Adp 0.25 gave about 1.9, below CAB Adp 0.81 alone (about 2.7). R11.m The two pairs containing the most hydrophobic polymer, CAB Adp 0.25, were the least effective of all (about 2.25 and 1.9), although CAP Adp 0.33/CAB Adp 0.25 still beat both of its polymers (about 1.65 and 1.0). R11.l·R11.m·C22 The text and Figure 6 call the second of these pairs "CAB Adp 0.85/CAB Adp 0.25", a polymer that does not appear in Tables 1 or 3; Figure 3 labels it CAB Adp 0.81/CAB Adp 0.25. X4

Finding 4: the pairs work better in 100 mM than in 50 mM buffer

Changing the buffer from 100 to 50 mM sodium phosphate raised the ritonavir solubility from 1.3 to 2.3 µg/mL, so at 10 µg/mL the supersaturation fell from 7.7 to 4.3. R2·R3 Since polymers had been less effective at higher supersaturation in earlier work (refs 38, 39), the ratios were expected to be higher in 50 mM buffer. C26 Table C shows the opposite.

Table C. Effectiveness ratio of eight pairs in two buffers, 10 µg/mL ritonavir, 5 µg/mL of each polymer; read from Figure 6 by its legend (approximate). R14

For all eight pairs the ratio was higher in 100 mM buffer. R14 The authors state that growth was faster overall in 100 mM buffer because of the higher supersaturation, but that the pairs were more effective relative to growth without polymer, and that the largest relative difference occurred for the more effective pairs. C27 Two reading points matter here. The text on p. 6055 assigns 50 mM to the white and 100 mM to the red columns, but the legend of Figure 6 shows the reverse; the legend is right, because the white columns repeat the 100 mM values of Figure 3 for seven pairs. X3 And the last pair has a 100 mM ratio of about 1.70 in Figure 6 against about 1.9 in Figure 3, while the other seven agree within about 0.05. X4 Absolute growth rates in 50 mM buffer are not printed. X12

Finding 5: the best pair stops growth at low supersaturation and still helps at high

Table D. Growth rate $R_g$ (µg mL−1 min−1) against supersaturation, 100 mM buffer, 5 µg/mL of each polymer; values with "about" read from Figure 7, values with ± from Table 5 (mean ± SD, triplicate). R15·R16·R10

Growth was reduced at every supersaturation, and the largest relative reduction was at the lowest $S$ of 3.8, where CAP Adp 0.85/Pn-IPAAmd "completely inhibited" growth. R15·C28 At $S$ 7.6 the three other CAP Adp 0.85 pairs were, according to the caption of Figure 7, statistically equally effective (the test is not named). R15·C29 At 20 µg/mL, about the amorphous "solubility" of ritonavir (stated as about 20 µg/mL without a reference), CAP Adp 0.85/Pn-IPAAmd cut the growth rate 3.3-fold, against 1.4-fold and 1.6-fold for its polymers alone, a "significant decrease" in the authors' words. R16·C30·C47·M1 PVP, ineffective alone (ratio 1.0), still raised CAP Adp 0.85 from 1.4 to 1.8 at this supersaturation. R16·C49 The pair advantage therefore shrinks as $S$ rises but does not disappear. R15·R16·C28

Finding 6: surfactants speed growth and cancel the cellulose inhibitors

Table E. Surfactants and surfactant/polymer pairs at 10 µg/mL ritonavir (S 7.6), 100 mM buffer; read from Figure 8 (approximate), which is also the graphical abstract. R17

Both surfactants, at a concentration well below their CMC and without changing the solubility of ritonavir (Tween 80: 1.3 ± 0.02 µg/mL), increased the growth rate by a factor of about 2. R17·R1.g·C31 With Tween 80 present, CAP Adp 0.33 and CAP Adp 0.85, which alone slowed growth about 1.6- and 3-fold, no longer slowed it at all: the pairs grew faster than ritonavir without any additive. R17·C31 The authors write that Tween 80 "significantly decreased" the effectiveness of the polymers. C31 Only Tween 80 was tested in pairs, and only with the two CAP Adp esters. R17

Seed habit after growth

Table F. Seed crystal size (mean ± SD of 50 crystals) before and after 30 min of growth, 5 µg/mL of each polymer (Table 4; SEM images in Figures S1–S3). Rows marked with a reference are values from the group's earlier papers. R6

Seeds grown at the higher supersaturation were more needle-like, and the polymers lowered the aspect ratio at both supersaturations; the authors read this as inhibition of growth along the fastest, lengthwise direction. R6·C11·C12 The text gives the aspect ratio without polymer at $S$ 7.6 as 7.7, but Table 4 gives 8.8 for that case; 7.7 is the value before growth. X2 Recomputing length over width reproduces the printed ratios except for CAP Adp 0.85 at $S$ 7.6 (54.0/10.5 = 5.14, printed 5.2). E5·X14 Note by the AI writer (not discussed in the paper): the lengths with polymer are mostly below the length of the seeds before growth, the standard deviations are 10–19 µm, and at $S$ 7.6 the lower aspect ratio comes mainly from wider crystals, so the habit data support only a qualitative reading. X9

Discussion

How the authors explain the pairs

The authors start from the requirement that an additive must adsorb on the crystal to change growth, since these additives barely change the solubility. C32·C8 They then "speculate" on which of the adsorption scenarios applies to each kind of pair. C33·C34 The chain below summarizes their reading; none of its middle steps was measured in this paper.

Different sites. For CAP Adp 0.85/Pn-IPAAmd, the authors argue that the two polymers are chemically so different that "it is quite likely that they adsorb, at least to some extent, to different sites", which would explain the strong synergy. C34 Part of the effect "may be partly attributed" to hydrogen bonding of the Pn-IPAAmd amide to the fast-growing faces, as proposed in their earlier paper. C35·C13

Hydrophobic interpolymer complexes. The synergy between cellulose esters of similar chemistry, such as CA 320S Adp/CP Adp and CAP Adp 0.85/CAP Adp 0.33, is harder to explain by different sites. C21·C36 For these the authors propose that hydrophobic interactions, which are driven by an entropy gain and stabilize interpolymer complexes in water, "may be important for determining whether synergistic interactions occur". C36 They exclude the two other usual forces. Hydrogen bonding between polymers is weak in water, which competes for hydrogen bonds, and the adipate group (pKa about 4.43) is nearly fully ionized at pH 6.8, which removes it as a hydrogen-bond donor; electrostatic attraction is excluded because the ionized carboxylates repel each other. C41 For CAP Adp 0.85 with CAP Adp 0.33, which differ only in adipate substitution, the synergy is read as "the possibility of complex formation that enhances effectiveness". C40

The buffer test as support. Salts that structure water strengthen the association of hydrophobic groups (refs 47, 52), so stronger interpolymer complexes, and more adsorption of these complexes on the crystal, were expected at higher ionic strength. C42 The higher ratios in 100 mM buffer (Table C) are read as support: the lower effectiveness at lower ionic strength is described as "suggesting that hydrophobic interactions are promoted at higher ionic strengths", and the expectations as "in general supported". C43·R14 Hydrophobic forces are also "likely" to contribute to the adsorption of the polymer on the drug crystal itself. C44

The hydrophilic/hydrophobic balance. As for single polymers in their earlier work, the authors argue that a complex must have a hydrophilic/hydrophobic balance to adsorb: a very hydrophobic polymer is more likely to form a very hydrophobic complex in solution, "which may not adsorb or, if it does, will not form very extensive interactions" with the crystal surface; Yu and Somasundaran (1996, ref 6) proposed the same. C45 This is offered as the reason why pairs with the most hydrophobic polymer, CAB Adp 0.25, were the weakest, and why pairing two moderately hydrophobic polymers generally worked better. C37·C45 For CAP Adp 0.85 with HPMCAS or CAPh, which are about as effective alone as CAP Adp 0.85, the lack of synergy "suggests that these polymers may compete for the same adsorption sites". C39

Surfactants. Surfactants can promote growth by solubilizing solute in micelles or by lowering the interfacial tension and/or forming surface clusters (refs 53, 58). C51 Because the concentrations were well below the CMC, the authors conclude that "it appears that they promote crystal growth via adsorption" at the crystal–solution interface. C52 For the polymer/surfactant pairs, they cite the common observation that a surfactant reduces polymer adsorption on solid surfaces, attributed to polymer–surfactant aggregates in solution (refs 8, 20), and write: "This most likely explains the experimental results shown in Figure 8". C53 Interactions between polymers and nonionic surfactants are usually weak, but "could occur if a sufficiently hydrophobic polymer is used", as here. C54

What the paper means for formulation

The authors draw four practical points. Careful evaluation of additive combinations on crystallization is recommended for supersaturating dosage forms, and with judicious selection "it may be possible to substantially delay crystal growth" at the supersaturation generated by an amorphous solid while improving other formulation properties. C50 Because PVP, ineffective alone, still helped CAP Adp 0.85, more hydrophilic polymers "can be potentially combined" with moderately hydrophobic ones in amorphous matrices to improve drug release. C49 The complete inhibition at $S$ 3.8 "may be of relevance" for preventing Ostwald ripening of drug nanoparticles, and polymer combinations should be studied more widely. C48 Finally, the loss of polymer inhibition with Tween 80 is "clearly important" for amorphous formulations that contain surfactants as plasticizers (ref 60). C55

Note by the AI writer (not discussed in the paper): two of these results cut across common practice. The single-polymer ranking did not predict which pairs would work: HPMCAS and CAPh were fair inhibitors alone yet added nothing to CAP Adp 0.85, while PVP and PVPdn-O, ineffective alone, helped (Table A). R11 And Tween 80 at 5 µg/mL, about a fifth of its CMC, reversed the inhibition by both cellulose esters tested (Table E). R17 A growth screen that tests the polymer without the surfactant of the final formulation could therefore overstate the protection against crystal growth.

Limitations

Limits the authors state. The paper has no separate limitations section. Its explanations are worded as possibilities: "We can speculate", "quite likely", "may be important", "suggesting the possibility", "most likely". C34·C36·C40·C53 The authors note that the polymer combinations were not more effective than the single polymers at extending nucleation induction times, unlike their effect on growth. C15

The AI writer's own judgement (1): what was measured and what was inferred. The growth-rate data are direct, replicated and large where it matters: the best pair's ratio of about 12 against about 3 for each polymer far exceeds the scatter. R11 The mechanisms are not measured: there are no data on interpolymer complexes, polymer adsorption on ritonavir crystals, or polymer–surfactant aggregates. C56·C53 For the surfactant pairs, a simpler reading fits the numbers too: the pair ratios (about 0.70 and 0.83) lie between Tween 80 alone (about 0.53) and the polymers alone (about 1.64 and 2.97), which two opposing effects acting independently would also give. X15

The AI writer's own judgement (2): comparisons that are not like for like. Each pair carries twice the total polymer of a single-polymer run, and an equal-dose check exists only for CA 320S Adp/CP Adp. X7 The buffer comparison changes the supersaturation together with the buffer concentration (4.3 against 7.7), the ionic strength itself is not stated, and absolute growth rates in 50 mM buffer are not given. X12 The most effective pair also gave the highest solubility (1.8 ± 0.20 µg/mL), which, if real, would lower its supersaturation. X6

The AI writer's own judgement (3): the hydrophobicity rule has exceptions. CAP Adp 0.33/CAB Adp 0.25, both among the most hydrophobic polymers (solubility parameters 20.56 and 20.05), still beat both of its polymers, although the authors say the more hydrophobic combinations did not show synergy. X8 PAlAmn (20.70) is more hydrophobic than CAP Adp 0.85 (21.27) by this scale yet helped it, and HPMCAS and CAPh, which did not help, sit in the same range as HPMC, which did. X8

The AI writer's own judgement (4): internal inconsistencies. Several labels and values disagree within the paper: $S$ at 10 µg/mL is printed as 7.6 and 7.7; the text swaps the colors of Figure 6; a "CAB Adp 0.85" appears that is not in the materials list; the text gives an aspect ratio of 7.7 where Table 4 gives 8.8; and the baseline growth rate is 1.56 µg mL−1 min−1 in the text but about 1.4 in Figure 7. X1·X3·X4·X2·X5 The growth-not-nucleation argument is qualitative: the two seed-volume estimates differ by a factor of about 2.3, and the size distributions are on pixel axes with different ranges before and after growth. X10 The induction-time comparison rests on one printed value. X11

Conclusion

The paper makes three points with data. First, in seeded growth at pH 6.8 and 37 °C, 10 of 13 polymer pairs at 5 µg/mL each inhibited ritonavir growth more than either polymer, and for one pair this held at equal total polymer. R11·R12·C17·C24 Second, the best pair, CAP Adp 0.85/Pn-IPAAmd, stopped growth at $S$ 3.8 and still cut it 3.3-fold at 20 µg/mL, close to the amorphous solubility, where its polymers alone managed 1.4- and 1.6-fold. R15·R16·C28·C30 Third, TPGS and Tween 80 below their CMC doubled the growth rate, and Tween 80 turned both cellulose esters tested into net growth promoters. R17·C31 The explanations, different adsorption sites, hydrophobic interpolymer complexes favored by salt, competition for sites, and surfactants displacing the polymer, are plausible but untested in this paper. C56·C43·C39·C53 The work is limited to one drug, initial rates over 2 min with seeds, and µg/mL additive levels. P8·E4