Which polymer keeps which drug amorphous: a spin-coating screen of 8 fast-crystallizing drugs and 7 polymers

Background and motivation

Many new drug candidates dissolve poorly in water. One way to raise their apparent solubility or dissolution rate is a solid dispersion: the drug is dispersed at the molecular level in a carrier, usually a polymer, so that it is not crystalline. The higher energy of the noncrystalline drug can give a higher apparent solubility and, potentially, a higher oral exposure than the crystal (Leuner and Dressman 2000, ref 1; Serajuddin 1999, ref 2). C6 The same higher energy is the problem: the amorphous drug tends to recrystallize during storage, which the authors call one of the major drawbacks of the approach (Hancock and Zografi 1997, ref 3). C7 Polymers often improve physical stability, and poly(vinylpyrrolidone) (PVP), poly(vinylpyrrolidone-vinyl acetate) (PVPVA), hydroxypropylmethylcellulose (HPMC), hydroxypropylmethylcellulose acetate succinate (HPMCAS), polyacrylates and polymethacrylates are the usual carriers (refs 1, 4–6). C7

The authors point to a gap in how such studies are usually designed. A typical study combines one model drug with one or a few polymers. That makes two general questions hard to answer (Section 1). First, is the crystallization tendency of a drug inside a dispersion linked to the crystallization tendency of the pure amorphous drug? Second, can polymers be ranked by their general ability to prevent crystallization, so that one polymer could be called a better stabilizer overall? C9·C10 A practical issue adds to this: melt extrusion, spray drying and freeze drying, and newer routes such as mechanical disruption, consume relatively large amounts of drug, so methods that test the feasibility of a dispersion with little drug are of interest (Willart and Descamps 2008, ref 7). C8

The group had already built such a method for pure drugs. In two 2010 papers they screened 51 compounds for crystallization, either on cooling the melt inside DSC pans or after rapid solvent evaporation by spin coating, and sorted them into three classes (Baird et al. 2010, ref 8; Van Eerdenbrugh et al. 2010, ref 9). Class I compounds crystallize rapidly and nearly completely right after spin coating; class II compounds show little or no crystallinity right after spin coating but substantial crystallinity within 7 days of storage; class III compounds show little crystallinity even after 7 days. The two routes, from the melt and from solution, classified the compounds in a similar way. R11 The authors add that the compounds commonly used as models in solid-dispersion research are often class III, which are fairly stable even without a polymer. C17

This paper extends the spin-coating screen from pure drugs to drug–polymer films. It tests two hypotheses stated in Section 1: that more polymer is needed to stop crystallization of drugs with a high propensity to crystallize, and that the stabilizing ability of polymers varies and depends on the chemistry of both drug and polymer. C9·C10 To make the polymers' effects visible, only class I and class II drugs were used; class III drugs were excluded because they were expected to stay amorphous over the week even without polymer. C13 Finally, the authors checked whether the thin-film results carry over to bulk powders made by rotary evaporation, a slower and more conventional route. A5

Background concepts

Rapid solvent evaporation and spin coating

A solid dispersion made by spray drying or film casting forms as the solvent leaves. As the solution concentrates, the drug's solubility in the remaining solvent is exceeded at some point, and from then on crystals can nucleate and grow; crystallization can also start later in the dry solid during storage. C29 Spin coating reproduces the first step on a small scale: 200 µL of solution is spread on an 18 × 18 mm glass coverslip that then spins for 20 s at 8000 rpm, so the solvent leaves quickly and a thin film remains (Section 2.2). P1 Because each film uses only a small volume of solution, many compositions can be screened with little drug.

Seeing crystals: polarized light microscopy

Crystals are birefringent: between crossed polarizers they appear bright or coloured against the dark or uniformly coloured background of an amorphous film. Figure 1 shows examples from 50/50 phenacetin films: a featureless film with PAA, scattered needle-shaped crystals with PVPVA at day 0, and a film birefringent over the whole field with HPMCAS by day 1. R1 The authors assume that in such thin films a crystal grows through the full height of the film, so the area covered by crystals, weighed against how much of the film is drug, gives a semiquantitative estimate of how much drug has crystallized ("it is reasonable to assume"). C11

The amorphicity index

Each film set was scored at four times on a five-step scale, each step written as four letters, A for amorphous and C for crystalline: AAAA (no crystallinity seen), AAAC (some crystals, drug mostly amorphous), AACC (comparable amounts), ACCC (mostly crystalline, but not completely) and CCCC (all drug crystallized). P2 The paper turns the four scores of a composition into one number, the amorphicity index (AI), defined in words in Section 2.3. Written as a formula:

In plain terms: $n_A$ is the number of A letters in the four scores of a composition (day 0, 1, 3 and 7), and 16 is the number of A letters a film set has when it is AAAA at every time point. A composition that stays fully amorphous all week scores 100%; one that is fully crystalline from the start scores 0%. E1 Note by the AI writer (not discussed in the paper): the index mixes two things, how fast and how far the film crystallized. A film that is slightly crystalline at day 0 and stays so (AAAC four times, 12 A letters) scores the same 75% as one that is amorphous until day 3 and then fully crystalline (AAAA, AAAA, AAAA, CCCC). Averages are then taken over the 7 drug/polymer ratios for each drug–polymer pair (Table 1), over all polymers for each drug (Figure 4) and over all drugs for each polymer (Figure 5). E2·E3·E4

Crystallization tendency classes

The class of each drug, I or II, comes from the group's earlier spin-coating study of the pure drugs (ref 9). It describes how readily the pure amorphous drug crystallizes after rapid solvent evaporation, not how it behaves with a polymer. R11 Five of the eight drugs here are class I (benzamide, phenacetin, flufenamic acid, chlorzoxazone and lidocaine), chlorpropamide and bifonazole are class II, and flurbiprofen is class I from a 1/1 (w/w) ethanol/dichloromethane solution but class II from ethanol or dichloromethane alone (Figure 2). M1·M2·M3·M4·M5·M6·M7·M8 This solvent dependence matters below, because HPMC and HPMCAS films were spun from the ethanol/dichloromethane mixture and all other films from ethanol. P1·X2·X12

Acids, bases and functional groups

The paper classifies drugs and polymers by their functional groups. The drugs include two amides, benzamide (BD) and phenacetin (PH); four acids, flurbiprofen (FB, computed pKa 4.37), flufenamic acid (FFA, 3.88), chlorpropamide (CP, 4.33) and the weak acid chlorzoxazone (CZ, 9.39); and two bases, bifonazole (BI, 6.69) and lidocaine (LI, 7.75). M1·M2·M3·M4·M5·M6·M7·M8 The pKa is the pH at which half of an acidic or basic group is ionized; the values here were computed, not measured (Figure 2, footnote a). P9 The polymers were described by the density of each functional group per gram, calculated from their structures (Figure 3): Eudragit E100 (E100) carries amine (5.0 mmol/g) and ester groups (10.0 mmol/g); poly(acrylic acid) (PAA) is the most acidic by count, with 13.9 mmol/g COOH; PVP carries 9.0 mmol/g of lactam amide; PVPVA 4.6 mmol/g amide and 5.5 mmol/g ester; poly(styrene sulfonic acid) (PSSA) 5.4 mmol/g of the strongly acidic sulfonic acid; HPMC ether (10.7) and alcohol (5.3 mmol/g) groups; and HPMCAS ether, alcohol and ester groups with only 1.0 mmol/g COOH. M9·M10·M11·M12·M13·M14·M15 PSSA is not a pharmaceutical excipient; it was included for the higher acidity of the sulfonic acid group, and its cross-linked analogues, ion-exchange resins, are used in formulations (ref 16). C14

Flufenamic acid needed a decision. It has a carboxylic acid and an aniline N–H that could donate a hydrogen bond. In the two crystal structures of FFA with published coordinates the aniline N–H forms an intramolecular hydrogen bond to the carboxyl oxygen, with N···O distances of 2.682 Å (form III) and 2.646 Å (form I), close to the 2.66–2.74 Å reported for resonance-assisted hydrogen bonds, a strong intramolecular bond that closes a six-membered ring with conjugated bonds (refs 11–13). R12 The authors therefore regard FFA as "most likely" able to interact with other molecules only through its acid group and treat it as purely acidic ("it is reasonable to ignore the aniline function"). C12

Methods in detail

The study has one main experiment and two checks. The main experiment is a full matrix: 8 drugs × 7 polymers × 7 drug/polymer ratios, 392 compositions, each spin-coated in triplicate and scored four times over a week. A1·A2·A3·A4·R2 The first check asks whether the acidic polymers act by changing drug solubility; the second asks whether the film results hold for bulk powders. A6·A5

Contribution 1: a drug × polymer × ratio matrix scored over a week

The problem. To compare polymers fairly across drugs, every drug has to meet every polymer at the same set of ratios, under the same preparation and storage, with a read-out quick enough for several hundred samples.

What they did. Drug and polymer were dissolved together in ethanol at drug/polymer weight ratios of 10/90, 25/75, 40/60, 50/50, 60/40, 75/25 and 90/10; for HPMC and HPMCAS a 1/1 (w/w) ethanol/dichloromethane mixture was used instead. P1 The concentration of the solutions is not given. X10 Some solutions of BI and LI with PAA or PSSA precipitated at some ratios; these were vortexed before coating. R6 For each film, 200 µL of solution was spread on an 18 × 18 mm coverslip and spun for 20 s at 8000 rpm on a KW-4A spin-coater; every solution was coated three times, and the films were stored in a desiccator over phosphorus pentoxide, at room temperature according to the abstract. P1 Within 30 min of coating (day 0) and after 1, 3 and 7 days, the three films of a composition were examined together under a Nikon Eclipse E600 POL polarizing microscope with 4× to 40× objectives and given one of the five scores. P2 The amorphicity index and its averages were then computed as described above. E1·E2·E3·E4

Why it works. Spin coating removes the solvent fast, as in spray drying, and the thin film lets the crystallized area stand in for the crystallized amount. C11 Scoring at day 0 separates crystals formed during coating from crystals formed during storage, a distinction the index then folds into one number. P2·E1

How it differs from earlier work. The group's earlier screen measured the crystallization tendency of pure drugs (ref 9). R11 Here the same rapid-evaporation route is applied to drug–polymer pairs, so each drug's behaviour with polymer can be set against its own class. A1

Contribution 2: testing whether the acidic polymers act through solubility

The problem. PAA and PSSA gave the most extreme results, best for some drugs and worst for others. One explanation is that a polymer changes how soluble the drug is in the drying solution, and so the supersaturation reached as the solvent evaporates ("might be potentially altered"). C29

What they did. Solutions of the polymers themselves were too viscous to separate from undissolved drug by centrifugation or filtration, so the authors used small-molecule stand-ins: acetic acid for PAA and benzenesulfonic acid for PSSA (the latter appears once as "benzylsulfonic acid" in Section 3.2, a misprint). C30·X4 They were added at 83.5 mg/mL acetic acid and 85.4 mg/mL benzenesulfonic acid, which the paper equates with the acid groups of a 10% (w/w) polymer solution. P4·E7 Excess drug was shaken in 1 mL of ethanol alone or with an additive for 24 h at room temperature, centrifuged at 16000 g for 10 min, and the supernatant diluted in methanol and assayed by HPLC-UV (LiChrosorb RP-8 column, 0.1% trifluoroacetic acid in water against methanol, 1 mL/min, detection at 230–291 nm depending on the drug), in triplicate. P4·P5

Note by the AI writer (not discussed in the paper): the additive concentrations match the acid content of 100 mg polymer per mL. Acetic acid at 83.5 mg/mL is 1.39 mmol/mL, the same as 13.9 mmol/g COOH × 0.100 g/mL of PAA; benzenesulfonic acid at 85.4 mg/mL is 0.540 mmol/mL, the same as 5.4 mmol/g × 0.100 g/mL of PSSA. E7 So the level corresponds to a 10% weight-per-volume solution, not weight-per-weight as printed. X11

Contribution 3: checking the films against bulk powders

The problem. A screen is only useful if it predicts what happens in a real preparation. Rotary evaporation is much slower than spin coating, which gives nuclei more time to form and grow in solution, and its 35 °C water bath may itself promote crystallization ("can also potentially promote"). C39

What they did. Two polymers that behaved very differently in the films were chosen: PAA, good for the amides and bases and poor for the acids, and PVP, relatively good for every drug except the bases. C37 For every drug except BI, 1 g of drug and 1 g of polymer (50/50 w/w) were dissolved in 80 mL of ethanol and the solvent removed on a Büchi Rotavapor R with a 35 °C bath; LI with PAA precipitated again and was homogenized with a rotor-stator mixer first. P6·R6 BI was left out because its high film AI values suggested that differences in bulk would be small. C38 The powders were dried under vacuum overnight, stored over phosphorus pentoxide, and examined after 1, 3 and 7 days by polarized light microscopy and by DSC (TA Q2000, nitrogen 50 mL/min, pinhole pans, heated at 5 °C/min from 20 °C to above the drug's melting point, in duplicate). P6·P7·P8 Crystallinity was taken as the drug's melting enthalpy in the powder, corrected for drug content, relative to that of the pure crystalline drug, following ref 10. E5 No crystallization exotherm appeared on heating, so no correction for crystallization during the scan was needed. R10

Why DSC was needed. Polarized light could tell a crystalline powder from an amorphous one (CP with PAA against CP with PVP, Figure 7A and B) but could not grade the extent of crystallinity (BD with PAA against BD with PVP, Figure 7C and D), so the five-step film scale could not be applied to powders. R8·C40 The authors warn that the DSC values "cannot be taken as exact": crystal disorder can lower melting enthalpies, some drugs are polymorphic and may melt, recrystallize and melt again, and mixing of drug and polymer in the melt can reduce the apparent enthalpy (refs 21, 22). C41

Experimental analysis

The pure drug's crystallization tendency carries over into the films

Averaged over all seven polymers and all seven ratios, the five class I drugs kept the lowest amorphicity, 47–59%, while flurbiprofen (class I/II), chlorpropamide and bifonazole (class II) kept 68, 73 and 97% (Figure 4, Table A). R4 In the authors' words, "even upon polymer addition, class I compounds still show a more pronounced crystallization tendency". C15

Table A. Average amorphicity index of each drug over 7 polymers and 7 ratios (from Figure 4 and Section 3.2), with chemistry and class from Figure 2. R4·M1·M2·M3·M4·M5·M6·M7·M8

The same pattern shows composition by composition in Table S2. At 75/25 drug, BD, PH, CZ and LI films were fully crystalline at every time point (AI 0) with every polymer except PSSA with PH (AI 56). R2.a–g·R2.v–ab·R2.aq–aw·R2.ax–bd Bifonazole films, in contrast, stayed largely amorphous even at 90/10 drug, with AI between 69 (PVP) and 100 (PAA). R2.h–n Chlorpropamide stayed fully amorphous up to 75/25 with E100, PVP, PVPVA and HPMC. R2.o–u Flufenamic acid is a class I drug but behaved more like the other acidic drugs: with E100, PVP and PVPVA it stayed fully amorphous up to at least 60/40. R2.aj·R2.al·R2.am

From this the authors draw two conclusions. Class I compounds "can be considered much more challenging" during both film formation and storage, and are for that reason "ideal model compounds" for telling polymers apart. C16 And knowledge of the pure compound's crystallization tendency is "extremely important" in formulation development; data like those of their pure-drug screen "could serve as an initial guidance" on whether a solid dispersion is feasible ("these results suggest"). C18 The abstract states the first hypothesis as confirmed in a hedged form: compounds with a higher crystallization tendency "tended to be more difficult to stabilize", and "more polymer was required". C1 The paper supports this with the averaged index; it does not tabulate the polymer fraction each drug needs, although Table S2 allows it to be read off. X13

Polymer averages, and why they mislead

Averaged over all eight drugs, the polymers rank PVP = PVPVA (70%) > E100 (67%) > HPMC (66%) > HPMCAS (62%) > PAA (49%) > PSSA (48%) (Figure 5). R5·C19 The authors then ask whether a higher average means a better polymer regardless of the drug, and answer: "The answer is clearly no". C20 Table B, the paper's Table 1, shows why.

Table B. Average amorphicity index (%) of each drug–polymer pair over the 7 ratios (from Table 1). R3

Every polymer spans a wide range across the drugs: PSSA from 8 (CZ) to 99 (BI), PAA from 13 (FFA) to 100 (BI), E100 from 13 (PH) to 99 (BI) and PVP from 29 (LI) to 94 (CP). R3 The two polymers with the lowest averages, PAA and PSSA, gave either the highest values among all polymers (for BD, PH, BI and LI) or the lowest (for FB, FFA, CP and CZ). C21 Their low averages therefore combine extreme successes with extreme failures rather than middling performance. R3

The averages can be reproduced from Table S2, and Table S2 can be reproduced from the day-by-day scores of Table S1 for 391 of the 392 compositions. E2·E3·E4·R2 The exception is FFA with PSSA at 60/40: Table S1 shows ACCC at day 0 and CCCC afterwards, an AI of 6, but Table S2 prints 94, and the averages in Table 1, Figure 4 and Figure 5 use 94. X1 Note by the AI writer (from the card's extractor note): with 6 instead of 94, the FFA–PSSA average would be about 3 rather than 15, FFA about 57% rather than 59% and PSSA about 46% rather than 48%; no rank changes. X1

A reversal set by drug chemistry

Amides. For benzamide and phenacetin the polymers fall in the same order, PSSA > PAA > PVP > PVPVA > HPMC > HPMCAS > E100, so "the most acidic polymers are best at inhibiting crystallization for these compounds". C23·R3.a–b With PSSA, benzamide films stayed fully amorphous up to 50/50 and phenacetin films up to 40/60 drug. R2.e·R2.bb With E100, at the other end, phenacetin was already partly crystalline at day 0 even at 10/90 drug (AI 56). R2.ax

Bases. The authors write that for bifonazole and lidocaine "the acidic polymers also yield the highest amorphous stability". C24 For bifonazole, PAA gave 100 and PSSA 99, but E100 and PVPVA also gave 99. For lidocaine, PSSA gave the highest value (71) while PAA (61) came just below HPMC (62), so the statement holds clearly for PSSA only. R3.g–h·X7 PVP was the weakest polymer for both bases (93 and 29), "indicating that it is not a very effective crystallization inhibitor for these compounds". C26 Lidocaine with PVP crystallized completely within 30 min of coating from 40/60 drug upward. R2.as The bases also behaved differently in solution: with PAA and PSSA some solutions precipitated, yet the films showed no crystalline material, and the authors suggest that "it appears that these drugs tend to complex with the polymers in solution forming amorphous precipitates", as reported for acidic polymer mixtures (ref 17). R6·C25

Acids. For flurbiprofen, flufenamic acid, chlorpropamide and chlorzoxazone the order reverses. PAA (13–31) and PSSA (8–30) gave the lowest averages; the paper calls them "very poor crystallization inhibitors" for these drugs and, in the abstract, "extremely poor stabilizers for acidic drugs". R3.c–f·C27·C3 Many of these films failed during coating itself: FFA with PSSA was fully crystalline at day 0 from 10/90 to 50/50 drug, CP with PSSA from 25/75 to 75/25, and CZ with PSSA from 25/75 upward, and FFA with PAA was already semicrystalline at day 0 at 10/90. R2.an·R2.s·R2.z·R2.ak HPMCAS, with only 1.0 mmol/g of COOH, was "not quite such a poor crystallization inhibitor" but the next worse (70, 62, 88 and 46); HPMC did somewhat better than HPMCAS but, except for CP, worse than E100, PVP and PVPVA, which "tend to show the best performance" (63–96). C27·R3.c–f

Table C shows the time courses behind these numbers for a few compositions. They illustrate the two ways a film can fail, at once during coating or gradually during storage.

Table C. Selected film scores (read from the grey shades of Table S1; AI as printed in Table S2). R2.e·R2.ba·R2.j·R2.ak·R2.an·R2.as

The authors read the whole pattern as "an interesting relationship between the drug–polymer chemistry and the resultant crystallization tendency" ("there appears to be"), and conclude that "both the drug and polymer properties are key" in deciding whether an amorphous dispersion can be formed and maintained ("suggest"). C22·C28

Drug solubility does not explain the reversal

Table D shows how the stand-in acids changed drug solubility in ethanol (Figure 6). For most drugs the change was small: with acetic acid solubility stayed at about 77–118% of the ethanol value, and with benzenesulfonic acid at about 82–141%. R7 Bifonazole was the exception, rising to about 252% with acetic acid and about 958% with benzenesulfonic acid; chlorpropamide rose to about 197% with benzenesulfonic acid only. R7.g·R7.e

Table D. Drug solubility in pure ethanol (mean ± SD, n = 3, printed below the axis of Figure 6) and relative solubility with the stand-in acids (read from the log-scale bars of Figure 6, approximate). R7

The authors call the influence of the acids "in general minor", with a significant increase only for BI. C31 Lidocaine is the telling case: it is a base that the acidic polymers stabilize well, yet its solubility did not change. The authors allow that its very high solubility in ethanol could mask a rise, but note that an absolute rise like bifonazole's would still be visible. C31 They conclude that the large difference between basic or amide drugs and acidic drugs "cannot be explained by changes in solubility" that would alter supersaturation during evaporation. C32 Two qualifications apply. No statistical test is given for "significant", and the near doubling for chlorpropamide with benzenesulfonic acid and the rise to about 141% for benzamide are not discussed. X6 And the test used small-molecule stand-ins in ethanol, not the polymers themselves, nor the ethanol/dichloromethane mixture used for HPMC and HPMCAS. C30·X12

Bulk powders versus films

DSC crystallinity of the 50/50 powders changed little between day 1 and day 7 (Figure 8). R9 Table E sets the day-7 values next to the 50/50 film results for the same pairs.

Table E. Film AI at 50/50 (Table S2) and DSC crystallinity of 50/50 rotary-evaporated powders (read from Figure 8, approximate, n = 2). A high film AI should pair with a low bulk crystallinity. R2.b·R2.c·R2.ay·R2.az·R2.ad·R2.ae·R2.ak·R2.al·R2.p·R2.q·R2.w·R2.x·R2.ar·R2.as·R9

For the four acidic drugs the two preparations agree on which polymer is better: with PAA the powders were about 64–117% crystalline, with PVP at most about 3%. R9.e–l·C42 For benzamide and lidocaine the order of the two polymers is also the same as in the films, although PAA powders of both were somewhat crystalline (about 7–12%). R9.a–b·R9.m–n·C43 The authors write that "none could be observed in the spin coated samples" for these two drugs with PAA; Table S1 shows the LI–PAA 50/50 film fully amorphous but the BD–PAA 50/50 film slightly crystalline at day 7. C43·X8 "The only real deviation" is phenacetin with PAA: the 50/50 film was only slightly crystalline at day 7, but the powder was about 80–95% crystalline. C44·R9.c·R2.ay

The flufenamic acid–PAA powder reads about 116–117% crystalline at all three times. R9.g Note by the AI writer (from the card's extractor note): a crystallinity above 100% cannot be real, so this value shows the size of the DSC errors the authors warn about, and their list of error sources does not address values above 100%. C41·X9 Fig. 8 also has no bars for several PVP powders (CP and CZ at all times, FB at days 3 and 7, FFA at day 7); the paper does not say whether these mean zero crystallinity or missing data. X14 Taking the differences between the techniques into account, the authors judge the agreement "reasonable" and conclude that spin coating "can be potentially used as a screening tool to assess polymer performance, with minimal drug consumption". C4·C5

Discussion

How the authors explain the pattern

The argument runs by elimination. Crystals can form during solvent evaporation, once the drug's solubility in the drying solution is exceeded, or during storage. C29 A polymer could act by changing that solubility; the stand-in acids showed it does not, at least not in a way that separates bases and amides from acids. C31·C32 What remains is the direct interaction between drug and polymer molecules. Because polymers of similar chemistry group together in how well they inhibit crystallization, and the grouping depends on the drug's chemistry, the authors find it "reasonable to speculate that the formation of drug–polymer intermolecular interactions underlie the observed differences". C33 Such interactions are known to affect drug–polymer miscibility (Marsac et al. 2009, ref 18) and have been proposed to inhibit crystallization in solid dispersions (Taylor and Zografi 1997, ref 19; Kestur and Taylor 2010, ref 20). C34 The authors add that the patterns across drug chemistries "strongly suggests that these interactions are extremely important in the prevention of crystallization", and the conclusions ascribe the compound dependence to "the synergy or lack thereof between the drug and polymer chemistry". C35·C33

Note by the AI writer (from the card's How solid assessment): two parts of this chain are supported by data and two are not. The solubility step rests on measurements (Figure 6), and the drug-dependent ordering of polymers rests on the full matrix (Table 1). R7·R3 The interaction step and the complex formation of the bases are interpretations: the paper contains no spectroscopic or other direct measurement of drug–polymer interactions, and the precipitates were only seen to be non-crystalline. C33·C25·R6 The paper also gives no specific interaction for why E100, PVP and PVPVA serve the acidic drugs best. C27 The authors themselves state that "the number of compounds studied is still too small to enable a comprehensive rationalization", and announce a larger study aiming at formulation rules based on drug and polymer structures. C36

Where the screen applies

The films are made by rapid solvent evaporation and stored dry for a week at room temperature, so the results speak most directly to solvent-based processes such as spray drying and to early, dry physical stability. P1·P2 The authors suggest that the screen could be made more predictive for spray drying by spin coating at the temperatures met in that process ("can be further optimized"). C45 The bulk check shows that a slower evaporation route mostly preserves the ranking of two very different polymers, with phenacetin–PAA as the exception. C42·C43·C44

What it suggests for formulation

The authors' own advice is twofold: know the crystallization tendency of the pure drug early, since it sets how hard the task is, and use class I compounds when the aim is to discriminate between polymers. C18·C16 Spin coating, in their view, "can be potentially used as a screening tool to assess polymer performance, with minimal drug consumption". C5

Note by the AI writer (not discussed in the paper): the clearest practical lesson is that a polymer ranking averaged over many drugs is a poor guide for one drug. In this set PAA and PSSA moved from the top to the bottom of the ranking depending only on whether the drug was an amide or base or an acid, and PVP, a common default, was the weakest polymer for both bases (average AI 93 for bifonazole and 29 for lidocaine). R3 A first polymer choice can start from the drug's acid–base chemistry and then be checked with a small screen of this kind. The complexation of the bases with the acidic polymers in solution, which the authors observed as precipitates, could also affect processing, which a dry-film screen would not show. R6

Limitations

Limits the authors state

• The thin-film scoring rests on the assumption that crystals span the full film height ("it is reasonable to assume"), and the scores are semiquantitative. C11
• The number of compounds is "still too small" for a comprehensive rationalization of polymer performance by drug chemistry. C36
• Rotary evaporation is much slower than spin coating and uses a 35 °C bath, both of which can promote crystallization in the bulk samples ("can also potentially promote"). C39
• DSC crystallinity values "cannot be taken as exact", because of crystal disorder, polymorphism and drug–polymer mixing in the melt. C41

The AI writer's own judgement

The study is broad and internally consistent, and its main contrasts are large: flufenamic acid kept an average AI of 13 with PAA against 87 with PVP, and benzamide 68 with PSSA against 30 with E100. R3.d·R3.a The day-by-day scores reproduce 391 of the 392 printed AI values, and the averages reproduce from those values. R2·E1·E2·E3·E4 Several points limit how far the results can be pushed.

• A visual, pooled score. Crystallinity was judged by eye from birefringent area; the three films of each composition were scored together, and no statistics are given. P2·C11
• One inconsistent value. Table S2 prints 94 for FFA–PSSA at 60/40 where Table S1 gives 6; the averages that use it are slightly high, though no rank changes. X1
• Mechanism not measured. Drug–polymer interactions, the proposed cause, were not measured, and the solubility test used small-molecule stand-ins in ethanol rather than the polymers. The rises for chlorpropamide (about 197%) and benzamide (about 141%) with benzenesulfonic acid are not discussed, and the additive level corresponds to 10% w/v rather than the stated w/w. C33·C30·X6·X11
• Generalizations a little wider than the data. For lidocaine, PAA is not among the best polymers (61, below HPMC at 62), so the acidic-polymer advantage for bases holds clearly only for PSSA. The benzamide–PAA 50/50 film was slightly crystalline at day 7, although the text says no crystallization was seen in the spin-coated samples. X7·X8
• A narrow bulk check. Only PAA and PVP at 50/50 were compared, bifonazole was left out, one DSC value exceeds 100%, and several bars are missing without explanation. C38·X9·X14
• A solvent confound. HPMC and HPMCAS films were spun from ethanol/dichloromethane and the others from ethanol, and for flurbiprofen the pure-drug class differs between these solvents. The text and Figure 2 also disagree on flurbiprofen's class in the mixed solvent. X12·X2
• Unstated conditions. The concentration of the coating solutions, film thickness, storage humidity and the ratios at which precipitation occurred are not given. X10
• Small labelling slips. Lidocaine is "LD" in Figure 2, benzenesulfonic acid appears once as "benzylsulfonic acid", and Table S2 repeats two PAA rows at the end of each half. X3·X4·X5

Conclusion

This paper turned the group's spin-coating test for pure drugs into a screen for drug–polymer pairs and applied it to a full matrix of 8 drugs, 7 polymers and 7 ratios. A1·A2·A3 Its two hypotheses held for this set. Drugs whose pure amorphous form crystallizes rapidly stayed the hardest to keep amorphous with any polymer, and polymer performance was "highly compound dependent", so no polymer was universally best. C1·C15·C20 The acidic polymers PAA and PSSA were the best inhibitors for the amide drugs and, mostly, for the bases, and the worst for the acidic drugs, for which E100, PVP and PVPVA did best. C3·C23·C24·C27 Because acid stand-ins barely changed drug solubility, the authors ascribe this reversal to drug–polymer interactions, an interpretation that the paper does not test directly. C32·C33 Bulk powders made by rotary evaporation ranked PAA and PVP the same way as the films for six of seven drugs, supporting spin coating as a low-consumption screen for polymer selection. C42·C44·C5