Can a polymer keep a pH-shift precipitate amorphous? Four ionizable drugs, nine polymers and the concentration left in solution

Background and motivation

Many poorly water-soluble drugs are weak acids or weak bases, and for such compounds the ionized form is usually much more soluble than the neutral form. Precipitation therefore becomes possible whenever the drug meets a pH that favours the neutral species. A weak base dissolved in the acidic stomach can precipitate as it passes into the small intestine, and the salt of a weak acid can precipitate in the stomach after it dissolves (Stahl 2002, ref 1). Once the concentration of the neutral species exceeds the solubility of its crystalline form, the solution is supersaturated. A supersaturated solution carries excess free energy and lowers it by forming a new phase, through nucleation and growth or through spinodal decomposition, depending on how far it is supersaturated (Introduction). The paper's premise is that the nature of this new phase, crystalline or disordered, decides how far the concentration falls.

The study follows directly from the group's earlier work (Hsieh 2012, ref 2), in which 10 weakly basic compounds were taken through a low-to-high pH transition. Two behaviours appeared. Compounds that crystallized rapidly showed only short-lived supersaturation, because the neutral form precipitated as crystals. Compounds that were reluctant to crystallize stayed supersaturated for longer, because their precipitate was amorphous. The authors draw a practical consequence, conditional on whether "similar behavior is observed in vivo": the reluctant crystallizers exploit the amorphous solubility advantage and offer greater potential exposure, while for rapid crystallizers exposure depends on how quickly the crystal–solution equilibrium is reached. For the rapid crystallizers, they suggest that polymers acting as crystallization inhibitors "may be beneficial" when the brief supersaturation is not enough (Introduction).

Polymers are the obvious candidates for that role. They are used to inhibit crystallization in oral and transdermal dosage forms (Davis 1991, ref 3; Iervolino 2000, ref 4), and they extend supersaturation in aqueous solution, which can raise bioavailability (refs 5–11, among them Curatolo 2009, ref 8, on HPMCAS, and Alonzo 2010, ref 9). The authors point to two gaps. Polymers had been studied mainly in amorphous formulations, not for ionizable compounds whose supersaturation comes from a pH change. And precipitation studies often did not characterize the precipitate or link its nature to the final solution concentration.

The choice of model drugs adds a second question. Crystallization tendency had been classified for many compounds by cooling from the melt and by rapid solvent evaporation (Baird 2010, ref 25; Van Eerdenbrugh 2010, ref 26). By the melt-quench method, papaverine, warfarin and glyburide are slow crystallizers and dipyridamole is a rapid one. Yet all four had been seen to crystallize rapidly after pH-induced precipitation in water, and that is why they were chosen. If melt-based classes predicted behaviour in dilute aqueous solution, three of the four should have been easy to keep amorphous, and the paper tests this indirectly.

The stated aim is to evaluate how far polymer additives can change the precipitation outcome, by inhibiting formation of a crystalline phase, and what that does to the solution concentration after precipitation. Two weak bases (papaverine and dipyridamole) and two weak acids (glyburide and warfarin) were precipitated by a pH shift, alone and with each of nine polymers chosen to cover nonionic, anionic and cationic chemistry. The precipitate was read by polarized light microscopy (PLM) and by synchrotron wide-angle X-ray scattering (WAXS) from 15 min to 6 h, and the supernatant by HPLC, against the crystalline solubility measured with the same polymer present.

Background concepts

Supersaturation and the right reference. A solution is supersaturated when the concentration of dissolved drug exceeds the solubility of the crystalline form; the ratio of the two is the degree of supersaturation. In pH-shift precipitation the drug is first dissolved as its ionized form, and the pH is then moved so that the neutral form exceeds its solubility. The reference is therefore the solubility of the crystalline neutral form, which Table 2 labels "Intrinsic Solubility". Polymers can raise this reference value by solubilizing the crystal. A high concentration after precipitation is evidence of supersaturation only if it exceeds the crystalline solubility measured with the same polymer, a point the authors make explicitly (Solution Study).

Two routes out of supersaturation. The Discussion describes two ways a supersaturated solution can reach a lower free energy, and both give visible precipitate. In the first, the drug crystallizes; if the stable polymorph forms, the concentration falls toward the crystal solubility. In the second, a solution above a threshold supersaturation first undergoes liquid–liquid phase separation (LLPS) into a disordered, noncrystalline, drug-rich phase (Brick 2003, ref 23). This "often" happens when supersaturation is created quickly and nucleation is slow (Tung 2009, ref 24). Below the drug's melting point the disordered phase is a supercooled liquid, metastable relative to the crystal, so it can still crystallize later. While it lasts, the solution concentration stays above the crystalline solubility, at a level set by the amorphous phase; the ratio of amorphous to crystalline solubility is the amorphous solubility advantage. LLPS was not detected directly in this study; it is the framework for reading the results.

Why the solid form matters for absorption. Flux across a membrane depends on the concentration of dissolved drug, or more correctly on its thermodynamic activity. An amorphous precipitate leaves a solution that is still supersaturated and, in the authors' words, "can enhance bioavailability"; a crystalline precipitate gives short-lived supersaturation with "marginal impact" (Discussion). The paper does not measure flux or absorption; this argument motivates reading the solid form and is not a result.

Polymorphs. A polymorph is a different crystal packing of the same molecule. A metastable polymorph is more soluble than the stable one, and a solubility ratio of about two is commonly observed for polymorph pairs (Pudipeddi 2005, ref 22). A precipitate that is crystalline but metastable can therefore hold the solution near twice the stable-form solubility. This matters for the dipyridamole results.

Crystallization tendency and multiple molecular forms. Earlier studies suggested that compounds with greater molecular complexity, which gives several low-energy forms, tend to crystallize more slowly (Baird 2010, ref 25; Yu 2000, ref 27). Several forms can arise from conformational flexibility or from tautomers, isomers that differ in where a proton sits (Derdour 2011, ref 28). During crystallization only the right form joins a stable nucleus or a growing crystal surface, so when several forms coexist, the effective supersaturation of the right form is lower. Table 1 of the paper lists tautomer counts for glyburide (6) and warfarin (9) and rotatable-bond counts for all four drugs.

Reading the solid: birefringence and WAXS. Between crossed polarizers, crystals light up (birefringence) while amorphous solids stay dark. WAXS gives a more direct test: a crystalline solid gives sharp peaks at characteristic positions, an amorphous solid only a broad hump, the halo. Peaks at the positions of the reference crystal indicate the same crystal form; peaks at other positions indicate a different form. Positions are given as the scattering vector $q$, related to the scattering angle by the relation printed in the Solid-State Analysis section:

In plain terms: $q$, in Å⁻¹, measures how strongly the X-rays are deflected, scaled by their wavelength, so that peak positions can be compared between instruments and energies. Here $\theta$ is the angle variable as printed (the paper relates $q$ to the scattering angle $2\theta$), and $\lambda$ is the X-ray wavelength, 1.033 Å at the 12 keV energy used. A peak at larger $q$ corresponds to a shorter repeat distance in the crystal.

CheqSol titration. The pH-metric method used for the no-polymer profiles (CheqSol, "Chasing Equilibrium Solubility"; Stuart 2005, ref 19; Box 2006, ref 20) dissolves the drug as its ionized form and adds titrant until the neutral form precipitates, which a UV dip probe detects. A "pH swing" of small acid and base additions follows, and the instrument computes the concentration of the neutral form in real time from mass and charge balance. The paper states that the technique distinguishes crystalline from amorphous precipitation: a spike that collapses within minutes points to crystallization, a sustained plateau to a precipitate that holds the concentration up.

Methods in detail

The work splits into three contributions. The first is a no-polymer baseline: pH-metric titration and microscopy show how fast each drug crystallizes on its own after a pH shift. The second is a screen of 40 systems (four drugs, each without polymer and with nine polymers), in which the precipitate is read by PLM and by time-resolved synchrotron WAXS. The third is the solution side: the concentration left within 1 h of precipitation, compared with the crystalline solubility measured with the same polymer and, for warfarin, with an estimate of the amorphous solubility from the Hoffman equation. The flow below shows how the pieces connect.

Contribution 1: A no-polymer baseline from pH-metric titration

① Problem. To judge whether a polymer changes anything, one first needs to know what each drug does alone after a fast pH shift, and on what time scale. The melt-quench and solvent-evaporation classes (refs 25, 26) were available, but they come from conditions very different from a dilute aqueous solution. A direct measurement under the conditions of interest was needed.

② What they did. About 15 mg of powder was dispersed in ionic-strength-adjusted water (0.15 M KCl) to give 1 mg/mL overall, in a Sirius GLpKa titrator fitted with a D-PAS unit and a 1 cm path-length UV dip probe. The drug was dissolved as its ionized form with 0.5 N KOH (the acids) or 0.5 N HCl (the bases), back-titrated until the neutral form precipitated, and followed through the pH swing at 25 °C (Figure 2a–d). The precipitates were inspected under a polarized light microscope (Nikon Eclipse E600POL; Figure 2e–h), for glyburide at 15, 30, 45 and 60 min after precipitation (Figure 3). The titration was run without polymer only, and the number of replicates is not stated.

③ Why it works. The titrator reports the concentration continuously, so the profile shows how fast the supersaturation collapses without sampling or separation, and birefringence assigns the collapse to crystallization. Together they give, for each drug, the size of the transient supersaturation and how long it survives. The no-polymer precipitations also form the control rows of Table 2.

④ How it differs from earlier work. The titration approach is the one the group used for 10 weak bases in 2012 (ref 2). Here it is extended to two weak acids and used as a reference point rather than as the main experiment. Because three of the four drugs were classed as slow crystallizers from the melt, the baseline also tests whether melt-based classes carry over to a dilute aqueous environment.

Contribution 2: A polymer screen read by microscopy and time-resolved synchrotron WAXS

① Problem. A polymer might stop crystallization, delay it, or steer the drug into a different crystal form, and a concentration measurement alone cannot tell these apart. The solid must be read, and quickly, because the forms of interest can change within minutes.

② What they did. About 15 mg of drug was dispersed in 15 mL of 0.15 M KCl with or without 1 mg/mL polymer, so drug and polymer both started at 1 mg/mL. Each solution was titrated to dissolve the drug as its ionized form and then back-titrated to a final pH at least two units above (bases) or below (acids) the pKa: papaverine (pKa 6.4) to pH 8.4–9.0, dipyridamole (pKa 6.2 and 0.8) to 8.2–9.0, glyburide (pKa 4.9) to 1.9–2.9 and warfarin (pKa 5.3) to 2.2–3.3 (Table 2). Each sample was prepared in triplicate; the temperature of this step is not stated. The nine polymers were three nonionic ones, PVP (polyvinylpyrrolidone, Kollidon 30, MW 44–54 kDa), PVPVA (polyvinylpyrrolidone-vinyl acetate, Kollidon VA 64, MW 45–70 kDa) and HPMC (hydroxypropyl methylcellulose, Methocel E50, 20 kDa). Five were anionic: HPMCAS (hydroxypropylmethyl cellulose-acetate succinate) in the LF and HF grades (MW 18 kDa), PAA (polyacrylic acid, average MW 1800 g/mol), CMCAB (carboxymethylcellulose-acetate butyrate, MW 22 kDa) and CAP-adipate (cellulose acetate propionate-adipate, MW 97,000 g/mol, prepared by the Edgar laboratory, ref 18). The cationic polymer was Eudragit E100, an amino methacrylate copolymer of dimethylaminoethyl, butyl and methyl methacrylate units in a 2:1:1 ratio (Figure 1).

For PLM, the solid was separated by ultracentrifugation at 274,356 g (40,000 rpm) for 15 min and checked for birefringence; the observation time is not stated except for the glyburide series. For WAXS, samples were prepared at Argonne National Laboratory, centrifuged for 1 min at 15,000 g, and the precipitate was packed into Kimax melting-point capillaries. Patterns were collected at beamline 12-ID-B of the Advanced Photon Source at 12 keV (λ = 1.033 Å; $q$ calibrated with silver behenate), with a 1 s exposure over a $q$ range printed as 0.75–2.43.Card check: the paper prints the unit as Å, but $q$ has units of Å⁻¹ and the figure axes run from 0.75 to 2.35 Å⁻¹; read as a unit misprint (X16). Measurements were taken at 15 min, 30 min, 1 h, 3 h and 6 h after precipitation and compared with the reference pattern of the crystalline drug; the HPMC series for the two acids was also compared with amorphous and blank traces. Table 2 codes each system as C (crystalline), P (polymorph) or A (amorphous), read at 15 min unless noted.

③ Why it works. The two readouts are independent. Birefringence is fast but says only crystalline or not. WAXS separates the reference crystal from another form by peak position, and its time series shows whether an amorphous precipitate stays amorphous or one crystal form turns into another. A 1 s exposure on a small centrifuged sample makes five time points per system practical, which the authors call "rapid and effective" (Conclusions). The blank trace guards against reading polymer or capillary features as crystallinity.

④ How it differs from earlier work. The authors' criticism of earlier precipitation studies is that they rarely characterized the precipitate. Here each of the 40 systems receives a solid-state code from two methods, and the polymer set spans charge type and hydrophobicity, from very hydrophilic PAA to CAP-adipate, the most hydrophobic polymer tested. Note by the AI writer (not discussed in the paper): because the acids were precipitated at pH 2–3 and the bases at pH 8–9, the same polymer is in a different charge state in the two groups. The paper states that both HPMCAS grades were unionized at the acids' final pH but does not test charge state as a variable.

Contribution 3: Concentration after precipitation against the crystalline and amorphous references

① Problem. The quantity of practical interest is how much drug stays dissolved. Interpreting it needs two references. The first is the crystalline solubility with the same polymer, because a polymer that solubilizes the crystal raises the concentration without any supersaturation. The second, for an amorphous precipitate, is the expected amorphous solubility, which sets the ceiling the concentration could approach.

② What they did. Within 1 h of precipitation, the samples from Contribution 2 were ultracentrifuged and filtered through 0.45 µm hydrophilic polypropylene syringe filters (0.2 µm for glyburide), and the filtrate was assayed by HPLC-UV. The HPLC method was isocratic reversed-phase (Symmetry Shield RP8 column, acetonitrile with pH 2.25 phosphate buffer), with detection at 251, 284, 228 and 282 nm for papaverine, dipyridamole, glyburide and warfarin, linear calibration (correlation coefficient at least 0.999) and a detection limit of about 1 µg/mL. Crystalline solubility was measured by stirring excess crystalline drug for 72 h at 25 °C in pH 10 buffer (bases) or pH 2 buffer (acids), with or without 1 mg/mL polymer, followed by the same separation and assay. Glyburide was too insoluble for the HPLC method: its no-polymer solubility (0.2 µg/mL) was taken from the titration, and because the values with polymer were also below the detection limit, the same 0.2 µg/mL "was assumed" for every polymer (Solution Study).

To estimate the amorphous solubility of the two acids, the authors used the Hoffman equation (Eq. 1, Discussion) with melting data that they attribute to DSC:

In plain terms: the free-energy excess of the amorphous form over the crystal, $\Delta G_c$, grows with the heat of fusion $\Delta H_f$ and with how far the working temperature $T$ lies below the melting point $T_m$ (both in kelvin). The authors present it as an estimate "to a first approximation". The free-energy difference is then turned into a solubility ratio (Eq. 2):

In plain terms: the amorphous form is more soluble than the crystal by the same factor by which its thermodynamic activity $a$ is higher, and that factor is the exponential of the free-energy excess divided by $RT$, where $S$ is solubility and $R$ is the gas constant. With a melting point of 175.6 °C and a heat of fusion of 45.5 kJ/mol for glyburide, and 169.2 °C and 43.6 kJ/mol for warfarin, the authors obtain amorphous solubility advantages of about 60-fold and 48-fold. The temperature used is not printed. The card's extractor reproduced 59.9 and 47.7 assuming 25 °C (298.15 K), with a glyburide $\Delta G_c$ of 10.15 kJ/mol, so the printed values are reproducible. The estimate neglects water mixing into the amorphous phase, which can lower the amorphous solubility by a factor of up to 2 (Murdande 2010, ref 39). The DSC method itself is not described.

③ Why it works. Dividing each concentration by the crystalline solubility measured with the same polymer separates real supersaturation from solubilization. The Hoffman estimate supplies an upper bound: if the precipitate is amorphous and controls the concentration, the observed ratio should lie at or below the predicted advantage, and below it by up to a factor of 2 if the amorphous phase absorbs water. Because the solid form and the concentration come from the same samples, each concentration can be read against the solid that produced it.

④ How it differs from earlier work. As the authors describe earlier precipitation studies, the precipitate was often not linked to the final concentration. Here each concentration is tied to a solid-state code, the reference solubility is specific to the polymer, and the Hoffman comparison checks whether the warfarin values are of the size an amorphous precipitate would produce.

Experimental analysis

Setup in brief. Four drugs were tested: two weak bases (papaverine and dipyridamole) and two weak acids (glyburide and warfarin), each at 1 mg/mL in 0.15 M KCl. Each was tested without polymer and with nine polymers at 1 mg/mL, which gives 40 systems prepared in triplicate. The solid was read by PLM and WAXS, the concentration within about 1 h, and the crystalline solubility after 72 h with the same polymer. The standard deviation of the solubility and concentration values is stated only as "within 20%" (Table 2 footnote); error bars are drawn only for dipyridamole and warfarin (Figure 9). Table A summarizes the drug properties the paper uses in its argument.

Table A. Properties of the model drugs (from Table 1; pKa from refs 12–14, descriptors from PubChem, ref 16; MW printed without unit). Glyburide and warfarin T_m and heats of fusion are from the authors' DSC; the source of the other two T_m values is not stated. Melt-quench classes from refs 25 and 26 as cited in the Discussion.

Main result 1: without polymer, all four drugs crystallize, glyburide after a delay. The titration profiles in Figure 2a–c show three sharp spikes. Dipyridamole, papaverine and warfarin rose to brief peaks of about 84, 240 and 165 µg/mL and fell within minutes to low levels, and their precipitates were birefringent: needles and laths for dipyridamole and papaverine, particle aggregates for warfarin. The authors read this as "indicating rapid crystallization" (Results). Glyburide behaved differently (Figure 2d). It peaked at only about 6.5 µg/mL and then held about 4.5–5.3 µg/mL for about 18 min (the text says about 20 min) before dropping to about 0.6 µg/mL and then 0.2–0.3 µg/mL. Its precipitate was round, non-birefringent aggregates at 15 min, and birefringence appeared at 30 min and grew at 45 and 60 min as elongated particles formed (Figure 3). The authors conclude that glyburide first formed an amorphous precipitate, which they call "confirmed" by the lack of birefringence, and that its crystallization then brought the concentration down to the equilibrium solubility.

Table B. Behaviour without polymer. Titration values read from Figure 2a–d (approximate; times are on the titration clock); PLM from Figures 2e–h and 3; WAXS peak positions read from Figures 5, 6a and 8a–b (approximate); concentrations and solubilities from Table 2.

Three points follow from Table B. First, every drug gave at least weak crystalline WAXS peaks by 15 min, and the peaks grew at later time points: papaverine reached its full reference pattern by 30 min with no further growth, which the authors read as "suggesting" the sample was close to fully crystallized; dipyridamole kept growing within the first 60 min (Figures 5 and 6a). Warfarin looked more crystalline than glyburide at 15 min (Figure 8a–b). Second, without polymer the concentration after precipitation was at or below the crystalline solubility for all four drugs, so the transient supersaturation had gone within the 1 h sampling window. Third, glyburide shows how the onset of crystallinity depends on the method: one weak WAXS peak at 15 min, no birefringence until 30 min, and a titration plateau of about 18 min. Table 2 nevertheless lists its no-polymer precipitate as birefringent, at an unstated observation time (card extractor note X3).

Papaverine, warfarin and glyburide were slow crystallizers by melt-quench, yet all three crystallized after a rapid pH shift. The authors say melt-quench classes "may not necessarily" predict crystallization tendency in dilute aqueous solution, and that this "appears to be somewhat true" here (Discussion).

Main result 2: no polymer prevented the two bases from crystallizing. All 20 papaverine and dipyridamole precipitates were birefringent and gave crystalline WAXS peaks within 15 min (Table 2). For papaverine, every polymer system showed the reference peaks within 15 min (text; patterns with polymer not shown). For dipyridamole, PVP, PAA, CAP-adipate and Eudragit E100 gave patterns like the no-polymer one. HPMC was different (Figure 6b): at 15 min the pattern carried peaks at q ∼1.45 and ∼1.215 Å⁻¹ that do not match the crystalline reference. The ∼1.45 Å⁻¹ peak was present at all four time points, and smaller peaks below it grew from 30 min. Both HPMCAS grades and CMCAB gave similar patterns (not shown), and Table 2 codes these four systems P. The authors say the extra peaks "may be attributed" to an additional polymorph; the abstract states it more firmly, as a metastable polymorph stabilized by several cellulosic polymers.

Table C. The two weak bases: crystalline solubility with the same polymer, concentration within about 1 h after precipitation, and their ratio (values from Table 2; ratios computed from them). All 20 precipitates were birefringent; all papaverine systems are coded C. Final pH 8.4–9.0 (papaverine) and 8.2–9.0 (dipyridamole).

The solution data in Table C match the solid-state picture in broad terms. For dipyridamole, six systems stayed close to the crystalline solubility, which the authors read as "indicating rapid and complete crystallization to the stable form within 1 h" (Solution Study). The four P-coded systems with HPMC, both HPMCAS grades and CMCAB held 12.4–14.9 µg/mL against solubilities of 4.9–7.3 µg/mL. The authors attribute this "most likely" to a metastable polymorph, since a ratio of about two is common for polymorph pairs, and offer an alternative: these polymers "may" extend the time before the precipitate is fully crystalline. The PVPVA row does not fit the polymorph argument well. That system showed the same ∼1.45 Å⁻¹ peak up to 1 h, yet its ratio at about 1 h was only 1.19 (card extractor note X6). For papaverine, the authors state that the excess over crystalline solubility was "generally" below a factor of 2.Card check: recomputed from Table 2, the ratio exceeds 2 for PVPVA (2.52) and CMCAB (2.18), two of nine polymer systems (X5).

Two further features of Table C matter. Some polymers raised the crystalline solubility itself: papaverine reached 20.9, 21.9 and 19.0 µg/mL with HPMC, HPMCAS-LF and HPMCAS-HF, against 12.2 µg/mL without polymer, which is why the authors call the polymer-specific value "important" as the reference (Solution Study). Several concentrations also lie below the crystalline solubility, for example dipyridamole without polymer (3.4 against 5.4 µg/mL), which the paper does not discuss. Solubility was measured in pH 10 or pH 2 buffer after 72 h and concentration in 0.15 M KCl within about 1 h, so ratios near 1 cannot be separated from these differences (X8). Figure 9a also draws four dipyridamole solubility bars slightly differently from Table 2, for example about 6.2 rather than 6.7 µg/mL for HPMC, within the stated 20% (X7); the table values are used here.

Main result 3: most polymers kept the two acids amorphous, and their solutions stayed supersaturated. For glyburide, eight of nine polymers gave non-birefringent precipitates coded amorphous: PVP, PVPVA, HPMC, both HPMCAS grades, PAA, CMCAB and Eudragit E100. Only CAP-adipate gave a birefringent, crystalline precipitate, as without polymer. For warfarin, six polymers gave amorphous precipitates (PVP, PVPVA, HPMC, both HPMCAS grades and Eudragit E100), while PAA, CMCAB and CAP-adipate gave birefringent crystals; Figure 4 shows the warfarin PLM images. With HPMC, both drugs showed only a broad halo from 15 min to 6 h (Figure 8c–d); slight peaks also present in the blank "can be regarded" as artifacts. For PVP, PVPVA, HPMCAS and Eudragit E100 the text reports the same, without showing patterns.

Table D. The two weak acids (values from Table 2; ratios computed from them). Glyburide crystalline solubility is 0.2 µg/mL from titration and was assumed for every polymer, so no glyburide ratio is given. "Dark" means no birefringence. Final pH 1.9–2.9 (glyburide) and 2.2–3.3 (warfarin).

Warfarin gives the clearest result of the paper. With the six polymers that kept the precipitate amorphous, the concentration within 1 h was 103–224 µg/mL, against crystalline solubilities of 5.1–10.7 µg/mL with the same polymers. With the three polymers that did not, it was 4.9–12.8 µg/mL. The error bars in Figure 9b are far smaller than this gap; the widest, for PVPVA, spans about 82–123 µg/mL. The authors state that the concentration was 15–25 times the equilibrium solubility "for some, but not all" of the polymer solutions.Card check: against the polymer-specific solubility the six ratios are 13.1–24.9, two of them below 15; against the no-polymer solubility (3.7 µg/mL) they would be 27.8–60.5 (X4). They attribute the elevated values to amorphous warfarin ("can be attributed") and note that only polymers shown by WAXS and PLM to stabilize the amorphous form sustained supersaturation. They describe the ineffective polymers as giving "no significant increase", without a statistical test; with CAP-adipate the concentration was 2.2 times the solubility (X15).

Glyburide follows the same pattern in direction: with the eight effective polymers its concentration reached 2.0–37.0 µg/mL, against 0.2 µg/mL without polymer and with CAP-adipate. The numbers carry more uncertainty than the warfarin ones. The reference solubility comes from titration rather than shake-flask HPLC, it was assumed equal for every polymer, and both the reference and the 0.2 µg/mL concentrations lie below the stated HPLC detection limit of about 1 µg/mL (X9, X10). The effective polymers also split into two groups. PVPVA, PAA and CMCAB gave only 2.8, 2.0 and 2.2 µg/mL, whereas PVP, HPMC, both HPMCAS grades and Eudragit E100 gave 20.0–37.0 µg/mL, although all eight precipitates were coded amorphous. The authors report this difference without explaining it.

Main result 4: the time course of the solid. WAXS patterns are shown for eight of the 40 systems, and Table E collects what they show over time. They carry the time-resolved part of the argument: the HPMC halo for the two acids up to 6 h, and the conversion of the dipyridamole form with PVPVA.

Table E. WAXS time courses of the eight systems whose patterns the paper shows (read from Figures 5–8, approximate). The other 32 Table 2 codes rest on statements in the text.

The PVPVA–dipyridamole series is the only direct observation of a form change. The authors read it as a polymorphic transition with "complete conversion to the stable polymorphic form" by 6 h, and Table 2 codes it P → C (3 h). Since no pattern after 6 h is shown, "complete" applies within the observed window (X13). The HPMC series for the two acids is the direct basis for the claim that the amorphous precipitate persisted "for at least 6 h". The abstract says "more than 6 h", which goes beyond the last time point (X1).

Comparison across systems. The clearest division in the data is by drug, not by polymer. PVP, PVPVA, HPMC, both HPMCAS grades and Eudragit E100 kept both acids amorphous but did nothing to stop papaverine or dipyridamole crystallizing. Among the acids the polymer still matters: PAA and CMCAB worked for glyburide but not for warfarin, and CAP-adipate worked for neither. The polymers that worked for both acids were nonionic (PVP, PVPVA, HPMC), unionized at the low final pH (both HPMCAS grades, as the authors note), or the cationic Eudragit E100. For the bases, the cellulosic polymers changed which crystal appeared and roughly doubled the concentration.

Controls and what they rule out. The no-polymer rows show that, alone, all four drugs end up crystalline and near their crystalline solubility within about 1 h, so a polymer effect is a change from that baseline. The polymer-specific solubility rules out solubilization as the source of the warfarin values: the highest crystalline solubility with any effective polymer was 10.7 µg/mL, against 103–224 µg/mL measured. The blank and amorphous reference traces rule out reading polymer or capillary features as crystallinity, and PLM and WAXS codes agree in all 40 systems. The controls do not rule out effects of the different media and times for solubility and concentration, which bear on ratios near 1, or colloidal glyburide passing the filter, which the authors raise themselves.

Discussion

Where the findings apply. The results hold for dilute conditions: drug and polymer each at 1 mg/mL in 0.15 M KCl, a titrator-driven pH shift to at least two pH units beyond the pKa, and observation up to 6 h. Crystalline solubility was measured at 25 °C; the precipitation temperature is not stated. Within these limits, a polymer can switch the precipitate of a weak acid from crystalline to amorphous, and the concentration then stays well above crystalline solubility. When the precipitate crystallizes, the concentration ends near crystalline solubility, or near twice it if a different crystal form appears. The authors summarize this in the Conclusions as solution concentration being "highly correlated" with precipitate structure, although no correlation was computed.

Why the drugs differ. The authors offer a drug-side explanation, clearly hedged. Glyburide and warfarin can form several tautomers (refs 29, 30), and multiple low-energy forms lower the effective supersaturation of the form that can crystallize; this "may explain" why most polymers kept them amorphous. Polymers may also hinder the molecular rearrangement needed to build an ordered nucleus, a point the authors introduce with "It has been speculated" (ref 31). Papaverine and dipyridamole, on the other hand, self-assemble in an organized way that polymers could not disrupt, which "may reflect" reduced conformational freedom; papaverine is reported to adopt a conformation in solution similar to that in its crystal structure (ref 32). The authors close the argument with "Clearly more work is needed". Note by the AI writer (not discussed in the paper): Table 1 does not support the conformational-freedom part of this argument. Dipyridamole has the most rotatable bonds (12) and warfarin the fewest (4), so the counts printed in the paper run against the ranking (X12). Tautomer counts are given only for the two acids.

Why the polymers differ. The authors note that polymer inhibition of crystallization is not well understood and is thought to involve adsorption on the drug solid or disruption of nucleus growth (refs 33, 34), so effectiveness "likely" depends on the drug–polymer interaction. They suggest hydrogen bonds between the drugs' donors and the strong acceptors of PVP and PVPVA or the moderate acceptors of the cellulosics. Alternatively, the donors of the cellulosics and Eudragit E100 could bond to drug acceptors. They add that such bonds "may not be favorable" in water, where solvation competes (ref 36). The failures of PAA, CMCAB and CAP-adipate suggest that hydrogen bonding is not the only factor, and polymer hydrophobicity relative to the drug "may also be a factor" (refs 31, 37, 38). None of these properties was measured. Note by the AI writer: because the acids and bases were precipitated at opposite ends of the pH scale, the design cannot separate a drug effect from a polymer-charge effect.

What remains open. The dipyridamole form seen with cellulosic polymers was not identified (no reference pattern, thermal data or solubility), and the authors leave two explanations for the twofold concentrations open. Amorphous solubility and the water content of the precipitates were not measured, so the warfarin comparison rests on an estimate. There are no data after 6 h, on membrane flux or in vivo, which the authors themselves frame as conditional.

What it suggests for formulation and further work. The authors recommend measuring supersaturation against the crystalline solubility with the same polymer, and call it "critical" to understand the crystallization kinetics of ionizable compounds and how additives change them. AI reading, not stated in the paper: a polymer screen for a weak acid or base should record the solid form of the precipitate as well as the concentration, because a polymer that holds one drug amorphous may do nothing for another. With HPMCAS-LF the warfarin crystalline solubility was almost three times the no-polymer value (10.7 against 3.7 µg/mL), so a ratio quoted against the polymer-free solubility can overstate the gain. For rapid crystallizers such as papaverine and dipyridamole, 1 mg/mL of these polymers was not enough; higher polymer levels or other excipients were not tested.

Limitations

Limitations stated by the authors. The authors say the glyburide results are "harder to interpret" for three reasons. Its crystalline solubility is so low that the value, especially with polymer, is "somewhat uncertain". Glyburide could crystallize during filtration. Very fine colloidal particles could pass the filter, and some filtrates stayed turbid after filtration. They conclude that more sensitive analytical methods "would thus be required" to confirm the relationship between amorphous solubility and precipitate structure for glyburide. They present the Hoffman estimate as a first approximation that neglects water mixing into the amorphous phase. They describe polymer inhibition as not well understood, say more work is needed on why crystallization tendency differs between molecules, and offer delayed crystallization as an alternative to the dipyridamole polymorph.

The AI writer's own judgement (1): what the solid-state evidence covers. The card rates the evidence "moderate". Birefringence and WAXS agree in all 40 systems, but WAXS patterns are shown for only 8 of them (X14). The "at least 6 h" persistence is shown in figures only for HPMC; for the other polymers it rests on text, and for glyburide with PAA and CMCAB only the 15 min code is given. The abstract's "more than 6 h" goes beyond the last time point (X1). The Conclusions' claim of a crystalline precipitate within about 15 min "in all cases" does not hold for glyburide by PLM or titration (X2), and the birefringence column has no stated observation time (X3).

The AI writer's own judgement (2): stated ratios and the numbers behind them. Several summary statements do not match the table they summarize. The warfarin "15–25 times" is 13.1–24.9 against the polymer-specific solubility and 27.8–60.5 against the no-polymer value (X4). This matters for the Hoffman comparison. A 48-fold advantage reduced by up to 2-fold for water leaves at least about 24-fold, which is near the top of the observed range rather than around it. Against the no-polymer solubility, the Eudragit E100 value (60.5) exceeds even the uncorrected 48-fold estimate. The "consistent with" conclusion therefore holds only for the polymer-specific reference, and loosely. The papaverine "generally less than a factor of 2" excludes two of nine polymer systems (X5). The dipyridamole polymorph rests on two unassigned peaks, and the PVPVA row (same peak, ratio 1.19) does not follow the twofold argument (X6). "No significant increase" has no statistical test behind it (X15).

The AI writer's own judgement (3): the reference values. Solubility and concentration were measured in different media and at different times, and several concentrations fall below the solubility (X8). The glyburide reference comes from titration rather than shake-flask HPLC, and for papaverine and warfarin the titration and shake-flask values differ (about 15 against 12.2 and about 8 against 3.7 µg/mL), so the glyburide reference is not directly comparable with the others (X9). Note by the AI writer (not discussed in the paper): multiplying the assumed 0.2 µg/mL by the Hoffman advantage of about 60 gives about 12 µg/mL, yet five polymers gave 20.0–37.0 µg/mL. Either the crystalline solubility with these polymers is higher than assumed, or part of the measured glyburide was colloidal. Both possibilities match the authors' own caveats, and the glyburide numbers are better read as qualitative.

The AI writer's own judgement (4): scope and unstated details. The explanations based on tautomers, conformational freedom, hydrogen bonding and hydrophobicity were not tested, and the printed rotatable-bond counts run against one of them (X12). The design used one drug and one polymer concentration, one non-biorelevant medium, no time point after 6 h, and titration without polymer only. The DSC method and the preparation of the amorphous reference samples are not described (X11), and the papaverine starting form (hydrochloride or prepared free base) is not stated (X17). None of this overturns the main contrast, because the warfarin difference between amorphous and crystalline precipitates is large and well outside the error bars. It does make the finer quantitative claims, about fold-supersaturation and the twofold polymorph, weaker than the text suggests.

Conclusion

The paper shows, for four ionizable drugs that crystallize readily after a pH shift, that the outcome of adding a polymer depends first on the drug. For the two weak acids, glyburide and warfarin, most of nine polymers at 1 mg/mL kept the precipitate amorphous for at least 6 h. The solution then stayed far above crystalline solubility; for warfarin it was 103–224 µg/mL against 5.1–10.7 µg/mL. For the two weak bases, papaverine and dipyridamole, no polymer prevented crystallization within 15 min. Their concentrations stayed within about 2.7-fold of crystalline solubility, and the cellulosic polymers appear to have steered dipyridamole into a different crystal form. Synchrotron WAXS with 1 s exposures made it possible to read the solid of each precipitate at five time points.

The explanations offered for these differences, tautomerism and conformational freedom on the drug side and hydrogen bonding and hydrophobicity on the polymer side, are untested hypotheses, and some summary ratios differ from the table values. What remains solid is that, in this data set, the concentration followed the form of the precipitate. A precipitation-inhibitor screen should therefore read the precipitate as well as the solution, and measure supersaturation against the crystalline solubility with the same polymer.