How long do amorphous drug precipitates last in water, and which polymers make them last longer? Six dihydropyridines followed by synchrotron X-ray scattering

Background and motivation

An amorphous solid dispersion (ASD) improves the exposure of a poorly soluble drug because amorphous material dissolves faster and reaches higher concentrations than the crystal; the solution it produces is supersaturated, and a supersaturated solution holds the drug at a higher chemical potential, which raises membrane transport (Raina 2014, ref 7). The price is crystallization. The introduction argues that crystallization is the main limit of ASDs and that the risk has to be judged twice: in the solid during storage, and in solution during dissolution. Solid-state crystallization has been studied extensively; crystallization from the supersaturated solution much less so (refs 3, 4, 8). C1·C2

The paper builds on several earlier findings from the same group. The degree of supersaturation is the thermodynamic driving force for crystallization (Mullin 2001, ref 9). But a supersaturated solution does not have to deposit crystals: amorphous material can precipitate and persist for a while (refs 10–12). Raina et al. (ref 7) showed that a drug-rich phase forms once the concentration crosses a threshold, and that this threshold "has been shown to correlate" with the amorphous solubility, the binodal of a miscibility gap; the amorphous precipitate is therefore the product of liquid–liquid phase separation (LLPS) (refs 7, 13–15). Whether a given drug undergoes LLPS within the time of an experiment, or crystallizes first, depends on its inherent crystallization tendency. The authors note that large, flexible molecules with many rotatable bonds are "likely" slow crystallizers and rigid ones fast, but that predicting crystallization tendency from structure alone is "extremely difficult". C3·C4·C5·C6

Two tools for comparing crystallization tendency already existed. Baird et al. (ref 16) sorted more than 50 molecules into slow, moderate and fast crystallizers by a simple differential scanning calorimetry (DSC) test on undercooled melts, and this ranking carried over "to some extent" to solution (ref 10), where higher mobility, temperature and the solvent add further factors (ref 17). For solutions, nucleation is usually followed by turbidity or focused beam reflectance measurements (refs 20–22), which track scattering but cannot tell crystalline from amorphous particles. Van Eerdenbrugh et al. (ref 10) therefore proposed a high-throughput synchrotron method: supersaturation is made so high that precipitation is instantaneous, and X-ray diffraction shows how long the precipitate stays amorphous. Finally, "antinucleant" polymers had been shown to modify crystallization in supersaturated solutions in some instances and to stabilize systems that had undergone LLPS (refs 5, 7, 13, 14, 18, 19), but how they act, and how this depends on the drug, was not understood. C7·C8·C9

The paper sets out to do four things: rank the crystallization tendency of six structurally related dihydropyridines after a very high initial supersaturation; rank eight polymers as crystallization inhibitors by how much they delay the first detection of crystals; follow the solution concentration to see how much supersaturation survives; and measure by NMR how much polymer ends up in the precipitate. The six drugs (Figure 1) share the 1,4-dihydropyridine core and differ in their ester groups and ring substituents: cilnidipine, felodipine, nifedipine, nimodipine, nisoldipine and nitrendipine. A1·A2·A5

Background concepts

Supersaturation, amorphous solubility and LLPS. A solution above the crystalline solubility is supersaturated and can crystallize. Above a second, higher concentration, the amorphous solubility, it can also lower its free energy by splitting into two liquids: a drug-rich phase (seen as an amorphous precipitate or nanodroplets) and a water-rich phase that stays saturated with respect to the amorphous drug and therefore still supersaturated with respect to the crystal. The paper describes the two routes in its Discussion (Figure 11): a liquid–solid transition straight to crystals, which rapid crystallizers take, or a liquid–liquid transition followed by crystallization, which slower crystallizers take. The two-liquid state is lower in free energy than the homogeneous supersaturated solution but higher than the crystal in equilibrium with its saturated solution, so it is metastable. C22

Estimating the amorphous solubility. The amorphous solubility is not measured directly in this paper; it is calculated from the measured crystalline solubility. The first estimate uses the free-energy difference between the supercooled liquid and the crystal in the Hoffman approximation (ref 25, Eq. 1):

In plain terms: the ratio of amorphous solubility ${\sigma }_a$ to crystalline solubility ${\sigma }_c$ grows with the free-energy gap $\Delta G$ between the amorphous and the crystalline drug, and the Hoffman form estimates that gap from the molar heat of fusion $\Delta H_f$, the melting temperature $T_m$ and the undercooling $\Delta T=T_m-T$ at the temperature of interest $T$; $R$ is the gas constant. Because an amorphous drug in water absorbs some water, which lowers its activity, the authors then apply the correction of Murdande et al. (ref 26, Eq. 2):

In plain terms: the solubility ratio is reduced by the factor $e^{-I(a_2)}$, the change in drug activity in the amorphous phase caused by dissolved water, which the authors obtained from moisture sorption measurements; Table 1 lists it as 0.74–0.91 for the six drugs. An ionization correction, also part of the Murdande approach, does not apply to these neutral compounds. E1·E2·C11·M1–M6

Crystallization tendency classes. In the Baird DSC test, crystalline drug is melted, cooled and reheated. A drug that recrystallizes during cooling is a fast crystallizer (Class I); one that recrystallizes only on reheating above its glass transition is moderate (Class II); one that does neither is slow (Class III). P6

Wide-angle X-ray scattering (WAXS). An amorphous solid gives a broad, featureless halo; a crystal gives sharp Bragg peaks. The data are plotted against the scattering vector, defined in the Methods as

In plain terms: $q$ (in Å⁻¹) converts the scattering angle $2\theta$ into a quantity independent of the X-ray wavelength $\lambda$, here 1.033 Å at 12 keV. The paper uses $q$ from 0.88 to 2.33 Å⁻¹. Synchrotron X-rays are intense enough to see a small crystalline fraction in a dilute suspension within a 1 s exposure. E3·P4

Polarized light microscopy (PLM). Between crossed polarizers, crystals are birefringent and appear bright or coloured; amorphous particles stay dark. The paper uses it as a second, independent check of the X-ray result. P5

Methods in detail

The study has one central experiment, repeated with and without polymers, and three supporting measurements that serve to interpret it. The flow below summarizes them.

Contribution 1: a crystallization clock for precipitates in water

Problem. Ranking how fast a drug crystallizes from solution needs a test in which every drug starts from the same state and in which crystalline and amorphous precipitates can be told apart.

What they did. Each drug was dissolved at 100 mg/mL in dimethyl sulfoxide (DMSO), and 1 mL of this stock was added to 9 mL of 50 mM phosphate buffer pH 6.8 in a polypropylene tube. The nominal concentration is therefore 10 mg/mL in 10% DMSO, and every sample showed a visible precipitate immediately. The suspensions were loaded into glass melting-point capillaries and measured at beamline 12-ID-B of the Advanced Photon Source (12 keV, sample-to-detector distance about 0.45 m for WAXS; the SAXS detector was not used). The beam was aimed at the precipitate with the help of a live video feed, and each pattern took 1 s. Measurements were made at 15 and 30 min and at 1, 3, 6, 12 and 24 h after mixing; 15 min was the earliest time the set-up allowed. Drops of the same suspensions were also examined between crossed polarizers at 10× magnification. The temperature of these experiments and the number of repeat capillaries are not stated. P1·P4·P5·R4·X16·X18

Why it works. At 10 mg/mL every drug is two to four orders of magnitude above its estimated amorphous solubility, so the precipitate forms at once and the clock starts at the same moment for all drugs. The X-ray signal then shows directly whether the precipitate is still amorphous. Its limit is time resolution: a precipitate first seen crystalline at 6 h may have crystallized at any time after the 3-h measurement. C23·R10·X18

Contribution 2: comparing the solution clock with the melt screen

The crystalline drugs were also run through the Baird DSC screen: heating at 10 °C/min in a pinholed aluminium pan, cooling at 20 °C/min and reheating at 20 °C/min, which gives the class, the glass transition temperature $T_g$, the melting point and the heat of fusion (Tables 1 and 2). This places the solution result next to the most common quick screen for crystallization tendency. P6·R1

Contribution 3: a polymer screen across six drugs

The same experiment was repeated with polymer dissolved in the buffer before the drug was added. Eight polymers were used (Figure 2): poly(acrylic acid) (PAA, in two grades of about 450,000 and 1800 Da), poly(2-vinyl pyridine) (P2VP, about 40,000 Da), polyvinylpyrrolidone K-29/32 (PVP), Kollidon VA 64 (PVPVA, a vinylpyrrolidone–vinyl acetate copolymer), hydroxypropylmethyl cellulose (HPMC, Pharmacoat 606), HPMC acetate succinate (HPMCAS), carboxymethyl cellulose acetate butyrate (CMCAB, M_w 35,000 Da) and cellulose propionate adipate (CAPAdp, synthesized at Virginia Tech, ref 23). Most polymers were dissolved at 5 mg/mL in the buffer, which gives 4.5 mg/mL (0.45%) after the 1:9 dilution. CMCAB and CAPAdp, being hydrophobic and poorly water-soluble, were used at only 100 µg/mL. P1·M7–M15·C10

Note by the AI writer (not discussed in the paper): Figure 2 draws the second repeat unit of PVPVA with a methallyl-type side group and no acetate ester; the text identifies the material as Kollidon VA 64, a vinylpyrrolidone–vinyl acetate copolymer, so the drawing is a structural error. X13

Contribution 4: how much supersaturation survives

Crystalline solubility was measured by the shake-flask method at 25 °C in buffer and in buffer with 10% DMSO and 0.45% HPMCAS (ultracentrifugation at 274,356 g for 15 min, then HPLC with UV detection at 360 nm; n = 3). Amorphous solubilities were then calculated with Eqs. 1 and 2. The concentration actually present at 3 h was measured in the supernatant of the supersaturated suspensions, without polymer and with HPMCAS, after the same ultracentrifugation, by UV absorbance (calibration 0.5–50 µg/mL). HPMCAS was chosen as one of the most effective polymers. P2·P3·P7·R10

Contribution 5: where the polymer goes

To see where the polymer ends up, felodipine suspensions were made in deuterated solvents (felodipine 100 mg/mL in d6-DMSO; polymers 5 mg/mL in D2O; P2VP, because of its poor aqueous solubility, 2 mg/mL in d6-DMSO). After ultracentrifugation the supernatant was decanted and diluted with one volume of d6-DMSO, and the precipitate was dissolved in 1.25 mL of d6-DMSO. One-dimensional 1H spectra (Bruker Avance 500 MHz, room temperature) were integrated against reference solutions, using 1.28–2.05 ppm for PVPVA, 1.97–2.07 ppm for HPMCAS and 0.9–1.25 ppm for felodipine. The amounts were scaled to a 10 mL sample containing 100 mg felodipine and 45 mg polymer (18 mg for P2VP, which was present at 1.8 mg/mL). Four polymers were analysed: PVPVA and HPMCAS as two of the best inhibitors with very different chemistry, and PAA and P2VP as two ineffective ones. P8·R12

Note by the AI writer: the NMR method lists the D2O polymer solutions as "PAA, PVP, and HPMCAS", while Table 4 and the text report PVPVA and no PVP, and the integration region given is that of PVPVA; the methods also say all polymers were predissolved in buffer for the crystallization study but prepare P2VP in d6-DMSO for NMR without explaining how it entered the buffer. X11·X10

Experimental analysis

Result 1: six close analogues, crystallization from under 15 minutes to half a day

All WAXS patterns show a broad amorphous halo with its maximum near q ≈ 1.7–1.8 Å⁻¹ (read from Figure 4, approximate). Sharp Bragg peaks appear on the nifedipine pattern already at 15 min and stay; nimodipine and nitrendipine show peaks from 3 h, felodipine and nisoldipine from 6 h, and cilnidipine only at 12 h (Figure 4). Figure 5 condenses this into a grid, which Table A reproduces. PLM agrees: at 15 min only the nifedipine precipitate is birefringent (Figure 6), and at 6 h every precipitate except cilnidipine is (Figure 7). R5·R6·R7·R8

Table A. Crystallization of the precipitates without polymer (from Figure 5) next to the melt class (Table 2) and dry glass transition temperature (Table 1). Card rows R6, R1, M1–M6.

From this the authors derive the rank order, slow to fast: cilnidipine < felodipine = nisoldipine < nitrendipine = nimodipine < nifedipine. For five of the six drugs the first solid is therefore an amorphous precipitate; only nifedipine crystallizes directly. The abstract describes the behaviour as "highly variable ranging from immediate to several hours". The ties mean "first detected at the same sampling point"; with points at 1, 3, 6 and 12 h they cannot separate drugs that crystallize within the same interval. C13·C23·C46·X18

Can molecular structure explain the order? The authors compared it with molecular weight (346.33–492.52 g/mol), log P (1.82–4.10), hydrogen-bond donors (one for every drug) and acceptors (3–8), rotatable bonds (6–11) and polar surface area (65–120 Ų) and found no correlation. Only cilnidipine, the slowest, stands out as structurally more complex; the other five are very similar. Neither does the dry $T_g$ follow the order: nifedipine, the fastest, has the highest $T_g$ (47 °C). C31·M1–M6

Result 2: the melt screen hardly separates them

In the DSC screen, felodipine shows its $T_g$ at 45 °C on reheating and no recrystallization, only the original melting at 147 °C (Figure 3a). Nifedipine does not recrystallize on cooling either, but on reheating it shows its $T_g$ (47 °C), a crystallization exotherm peaking at about 115 °C (read from Figure 3b) and two melting endotherms, at 162 °C (T_m2) and 173 °C (T_m1). Nimodipine also recrystallizes on reheating and shows polymorphic transitions (T_m2 116 °C). Cilnidipine, nisoldipine and nitrendipine do not recrystallize at all. The screen therefore puts four drugs in Class III and two in Class II, and none in Class I. R1·R2·R3·C12

Figure 10 sets the two scales side by side. From the melt, four drugs are "slow" and two "moderate"; in solution the same six spread over four classes, from "slow" (cilnidipine) through "moderate–slow" (felodipine, nisoldipine) and "moderate–fast" (nimodipine, nitrendipine) to "fast" (nifedipine). The authors draw a one-sided rule from this: the drugs that crystallize in the DSC screen also crystallize rapidly from supersaturated solution, "but failure to crystallize in the DSC experiments does not guarantee that crystallization will be slow in a solution environment". Nitrendipine is the example: Class III in the melt, crystalline by 3 h in solution. They attribute the difference to the environment, with "temperature, solvent effects, and sample viscosity likely playing important roles". C28·C29·C30

Note by the AI writer: the text p.5 gives the nifedipine melting point as 172 °C, Table 1, Table 2 and Figure 3 give 173 °C; the Discussion calls cilnidipine a "very slow" crystallizer while Figure 10 places it in the "Slow" class and leaves "Very slow" empty. X14·X15

Result 3: three polymers delay crystallization for every drug

Figure 8 repeats the grid of Figure 5 for each drug with nine conditions (buffer and eight polymers; the low-molecular-weight PAA gave the same result as the high grade and is not shown). Table B gives, for each drug and polymer, the first sampled time at which crystallinity was detected.

Table B. First sampled time (h) at which crystallinity was detected, from Figure 8; "> 24" means still amorphous at the last point, 24 h. Polymers at 4.5 mg/mL unless stated. Card row R9.

Three polymers stand out and behave identically: with PVPVA, HPMC or HPMCAS, cilnidipine, felodipine, nisoldipine and nitrendipine stayed amorphous for the whole 24 h, nimodipine until 12 h, and even nifedipine gave an amorphous precipitate that lasted 1 h instead of crystals at 15 min. PAA did nothing for any drug, and P2VP did nothing except move the first detection of nifedipine crystals from 15 to 30 min. PVP was intermediate: it delayed cilnidipine, felodipine and nitrendipine to 24 h, nisoldipine to 12 h, nimodipine to 6 h and nifedipine to 3 h. CMCAB and CAPAdp, at their much lower concentration, were erratic: as good as the best polymers for cilnidipine, of some help for felodipine (CMCAB only) and nimodipine, a 15-minute delay for nifedipine, and no delay for nisoldipine and nitrendipine. R9·C14·C15

The authors summarize this as the rank PVPVA = HPMC = HPMCAS > PVP > CMCAB = CAPAdp > P2VP = PAA ("On the basis of the observed trends"), and state in the Discussion that HPMC, HPMCAS and PVPVA "were effective across the board", PVP "had some impact", and PAA, P2VP, CMCAB and CAPAdp "were largely ineffective". For nifedipine they write that the polymers slowed crystallization so much "that LLPS was observed and an amorphous precipitate could be sustained for 1 h". They also state that the order of the drugs was the same with every polymer, with nifedipine always fastest and cilnidipine slowest. C16·C38·C35·C36

Note by the AI writer: three of these summary statements are wider than the grid. The drug order is not preserved in full: with PVP nitrendipine (24 h) outlasts nisoldipine (12 h) and nimodipine (6 h), and with CMCAB felodipine (12 h) separates from nisoldipine (6 h); only "nifedipine fastest, cilnidipine slowest or tied" holds throughout. CMCAB and CAPAdp, called "largely ineffective", matched the best polymers for cilnidipine at about one forty-fifth of their concentration, which also makes the rank PVP > CMCAB = CAPAdp a comparison at unequal concentrations. "LLPS was observed" for nifedipine rests on the absence of Bragg peaks; no droplets or second liquid phase were imaged with polymer present. X6·X7·X8·X19

Result 4: how much supersaturation survives at 3 h

Table C reproduces Table 3 of the paper. Without polymer, the 3-h concentrations were measured in the 10% DMSO of the supersaturated mixture, while the crystalline and estimated amorphous solubilities in the same columns refer to buffer; the columns with HPMCAS all refer to 10% DMSO with 0.45% HPMCAS.

Table C. Solubilities and 3-h concentrations in µg/mL, mean ± 1 SD, n = 3 (from Table 3; "+ HPMCAS" = 10% DMSO + 0.45% HPMCAS in buffer). Card row R10.

The authors pick out felodipine and nifedipine (Figure 9). For felodipine, a slow crystallizer, the 3-h concentration without polymer (30.3 µg/mL) is far above its crystalline solubility and, in their words, "in good agreement with the estimated amorphous solubility (in the presence of DMSO)", the 25.3 µg/mL of the HPMCAS column: the solution is "consistent with an amorphous precipitate". For nifedipine, which is already crystalline, the 3-h concentration (21.8 µg/mL) is "close to the values of the equilibrium crystalline solubility in water with 10% DMSO" (28.2 µg/mL). With HPMCAS both rise well above the estimated amorphous solubility: 119 µg/mL for felodipine and 641 µg/mL for nifedipine. Cilnidipine and nisoldipine reach concentrations near or above their amorphous estimate without polymer; nimodipine and nitrendipine reach high concentrations only with HPMCAS. In the Discussion the authors add that an effective polymer raises the 3-h concentration "in particular for compounds that crystallize rapidly"; in Table C the rise is from 21.8 to 641 µg/mL for nifedipine and from 8.7 to 158 µg/mL for nitrendipine, against 30.3 to 119 µg/mL for felodipine. The Conclusions turn this into a correlation: "slower crystallizers exhibited higher solution concentrations, whereas fast crystallizer such as nifedipine exhibited lower concentrations". C18·C19·C37·C49·R11

Note by the AI writer: four points about Table C. (1) The amorphous solubilities cannot be reproduced from Eq. 2 and the printed inputs: for felodipine, 0.74 µg/mL × 0.74 × e^ΔG/RT with ΔH_f = 74.6 J/g × 384.25 g/mol, T_m = 420.15 K and T = 298.15 K gives 5.93 µg/mL, not 6.68; and the ratio of amorphous to crystalline solubility, which Eq. 2 fixes for each drug, differs between the two media (felodipine 9.03 versus 11.0, nimodipine 23.0 versus 43.9). (2) Figure 9 draws the buffer amorphous solubilities of felodipine and nifedipine at about 7.9 and 65 µg/mL, not 6.68 and 48. (3) The Results text reads columns 5 and 4 as "in the presence of DMSO" and "in water with 10% DMSO", while the table header, the Methods and the Figure 9 legend assign them to DMSO plus HPMCAS. (4) The concluding correlation holds relative to each drug's amorphous solubility, not in absolute numbers: cilnidipine, the slowest crystallizer, has the lowest 3-h concentration (1.9 µg/mL), and with HPMCAS nimodipine (116 ± 18) stays below its estimate (131.8) and nitrendipine (158 ± 24) only reaches it (153). X2·X1·X3·X4·X5

Result 5: effective polymers split between the phases, ineffective ones do not

Table D. Mass in each phase from 1H NMR, scaled to a 10 mL sample with felodipine 10 mg/mL and polymer 4.5 mg/mL (P2VP 1.8 mg/mL), all with 10% DMSO (from Table 4). Card row R12.

Felodipine itself sits almost entirely in the precipitate. The polymers differ sharply. PVPVA and HPMCAS are found in both phases: the precipitate holds "around 7% by weight" of polymer (6.8 mg with 94.2 mg felodipine; 7.1 mg with 101.4 mg), and most of the polymer remains in solution. PAA is almost entirely in the supernatant (0.9 versus 46.6 mg), and P2VP could not be detected in the supernatant at all, the measured 17.7 mg having precipitated with the drug. R12·C20

Note by the AI writer: Table D has no replicates or uncertainty, and the totals deviate from the stated amounts (felodipine 94.8–102.7 mg against 100 mg; PAA 47.5, PVPVA 43.5 and HPMCAS 41.3 mg against 45 mg). The contrast between the polymers is large compared with these deviations. X12

Discussion

The authors' account: two phases, both supersaturated

The Discussion ties the results together in a short chain. Supersaturation in vivo can come from dissolving ASDs, a change of pH, or the digestion of lipids that solubilize a drug (refs 28–30). At low supersaturation a solution stays homogeneous until it crystallizes; above the amorphous solubility an amorphous precipitate can form first, "unless crystallization is very rapid". In this study the nominal 10 mg/mL "far exceeds" every amorphous solubility, so every drug separates at once. Even with the precipitate present the system remains supersaturated with respect to the crystal, and crystallization will follow, from the precipitate or from the solution around it. C21·C22·C23

The causal chain as the authors give it in the Discussion (pp.8–11). Steps 2 and 3 and the polymer branches are their interpretation; the partitioning was measured for four polymers with felodipine only. Card rows C22–C25, C39–C43.

Two pieces of this chain carry hedges that matter. First, the physical state of the precipitates: "On the basis of the $T_g$ values shown in Table 1 and the experimentally estimated water content, most, if not all of the amorphous precipitates are likely to be above their $T_g$s", that is, viscous supercooled liquids. The dry $T_g$ values are 17–47 °C, and water lowers them further, but the water contents are not shown in the paper. Second, the role of the two phases: since crystals can form in either phase, "it is intuitive that effective polymers should interact with the drug in both phases". The site where crystals actually nucleate was not determined. C24·C39·X17

The polymer argument then runs as follows. PVPVA and HPMCAS, two of the best inhibitors with very different chemistry, are "indeed distributed" between precipitate and solution. This distribution "most likely arises from the moderately hydrophobic nature" of both polymers, which gives affinity for the drug-rich phase, while their affinity for water "can be inferred from the adequate water solubility". For the poor inhibitors, the more hydrophilic PVP and PAA and the more hydrophobic P2VP, CMCAB and CAPAdp, "We propose that this situation may arise because they have a much greater affinity for one phase relative to another." The NMR data support the supposition for PAA, which stays in the supernatant, and P2VP, which precipitates with the drug. The authors connect this with earlier nucleation and growth studies in which only moderately hydrophobic polymers were effective (Ilevbare 2012, ref 24), and they add a caution of their own: hydrophobicity "is only one important polymer property", other structural features are "likely" to matter, and "specific drug–polymer interactions such as hydrogen bonding may also play a role". C40·C41·C42·C43·C44·C45

In the abstract this becomes: "Polymers with intermediate hydrophilicity/hydrophobicity were found to substantially delay crystallization, whereas strongly hydrophilic or hydrophobic polymers were largely ineffective", and NMR "supported the supposition that polymers need to have affinity for both the drug-rich precipitate and the aqueous phase". The Conclusions put it more firmly: the affinity of the polymer for the drug-rich versus the water-rich phase "was found to correlate well with the crystallization inhibition ability". C47·C48

What the paper says about prediction

The second theme of the Discussion is prediction. The authors argue that assessing crystallization tendency in each environment, the glass or supercooled liquid, the hydrated amorphous matrix and the supersaturated solution, "enables prediction of the relative risk of crystallization during storage and delivery". Their own data show why the solution test cannot be skipped: the melt screen identifies the fast crystallizers but cannot clear the slow ones, and no molecular descriptor in Table 1 tracks the solution ranking. The variability among "seemingly similar compounds highlights the difficulty in predicting a priori" whether a compound can be formulated for sustained supersaturation. Several drugs stay amorphous for only hours or less, short compared with in vivo transit times, so to use the membrane-transport benefit of supersaturation "it is desirable to prolong the lifetime of the amorphous precipitates", which is what the effective polymers do. C27·C29·C31·C32·C33

What it suggests for formulation

Stated by the authors: test crystallization from the supersaturated solution, not only from the melt; use a polymer that prolongs the amorphous precipitate; and choose polymers of intermediate hydrophobicity with affinity for both phases. C27·C33·C47·C48

Note by the AI writer (not discussed in the paper). Three practical readings follow from the data. For this drug class, PVPVA, HPMC and HPMCAS at 4.5 mg/mL were interchangeable in delaying crystallization, so the choice between them would rest on other grounds such as processing or solid-state stability. A partitioning measurement (polymer in precipitate versus supernatant) is simple and, if the rule holds beyond felodipine, could rank polymers before longer crystallization studies. And a drug with a low absolute concentration after precipitation, such as cilnidipine (1.9 µg/mL at 3 h), can still be the slowest crystallizer; the useful quantity is the concentration relative to the drug's own amorphous solubility.

Limitations

Limits the authors state

The authors call crystallization in these systems "extremely complex", with two supersaturated phases and kinetic factors that "may differ" between them. They note that hydrophobicity is only one relevant polymer property and that hydrogen bonding "may also play a role". They state that predicting crystallization tendency from structure is currently "extremely difficult", and that the variation between solid- and solution-state behaviour arises from environmental factors that are "likely" important. The first measurement was possible only at 15 min "because of experimental constraints", so nifedipine's actual crystallization time is unknown. C39·C45·C6·C30·R6

The AI writer's own judgement

Assessment by the AI writer, based on the card's extractor notes, not a statement made in the paper. The central observation, the crystallization ranking of the six precipitates and the effect of each polymer, is direct X-ray evidence, confirmed by microscopy at two time points and complete for every drug–polymer pair; it is solid within the limits of a seven-point time grid with unstated repeats and temperature. The mechanistic conclusion is much thinner. It rests on partitioning data for four polymers with one drug, single determinations without uncertainty, in which P2VP was used at 1.8 instead of 4.5 mg/mL and prepared in a way the methods leave unclear. The polymers whose failure the argument explains by their hydrophilicity or hydrophobicity (PVP, CMCAB, CAPAdp) were not analysed at all, and no measure of hydrophobicity is given. The cellulose esters were tested at about 45-fold lower concentration than the other polymers, which confounds their place in the ranking. X18·X16·X9·X12·X8·X10

The supersaturation analysis has its own weaknesses. The estimated amorphous solubilities cannot be reproduced from Eq. 2 and the printed inputs, they imply different amorphous-to-crystalline ratios in the two media although Eq. 2 allows only one per drug, and Figure 9 draws two of them at values that differ from Table 3. The 3-h concentrations without polymer were measured in 10% DMSO but are compared with solubilities in buffer, and the Results text describes the medium of two columns differently from the table. Two summary statements are wider than the data: the correlation between crystallization rate and concentration fails in absolute terms, and the HPMCAS concentrations do not "far exceed" the amorphous estimate for nimodipine or nitrendipine. Smaller points: the PVPVA structure in Figure 2 is drawn wrongly, the nifedipine melting point appears as both 172 and 173 °C, the Discussion cites a schematic (Figure 11) for timing data that are in Figure 5, and references 4 and 18 are the same article. None of these changes the main ranking. X2·X1·X3·X4·X5·X13·X14·X15·X21

Conclusion

Six dihydropyridines with nearly the same structure, precipitated from the same highly supersaturated solution, stayed amorphous for very different times: nifedipine crystallized within 15 minutes, nimodipine and nitrendipine by 3 h, felodipine and nisoldipine by 6 h, and cilnidipine only between 6 and 12 h. The melt DSC screen, which placed four of them in the same "slow" class, did not reveal this spread. Polymers did not stop precipitation but delayed crystallization, and three of them, PVPVA, HPMC and HPMCAS, did so for every drug, keeping four drugs amorphous for at least 24 h and turning nifedipine's immediate crystallization into an amorphous precipitate that lasted 1 h. PAA and P2VP had no effect. With felodipine, the two effective polymers analysed were found in both the drug-rich precipitate and the water, PAA only in the water and P2VP only in the precipitate, which the authors read as support for the proposal that an inhibitor must have affinity for both phases. That proposal is plausible and consistent with earlier growth-inhibition work, but in this paper it rests on four polymers and one drug.