How amorphous nifedipine crystallizes in water, and what polymers can and cannot change

Background and motivation

Amorphous solid dispersions (ASDs), in which a poorly soluble drug is mixed with a polymer at the molecular level, are used because the amorphous drug dissolves to a supersaturated solution, which can raise absorption. Most earlier work on how polymers stabilize ASDs looked at the solid during storage; far fewer studies looked at what happens to the amorphous drug while it dissolves (Section Introduction; refs 1–12). That gap matters because the solubility advantage only lasts as long as the drug stays amorphous or in solution.

The starting point is a framework from the same group (Alonzo et al. 2010, ref 11). When an amorphous solid meets water, it can crystallize by two routes: inside the wetted solid (matrix crystallization), or after dissolving, from the supersaturated solution (solution crystallization), or by a mixture of both. Matrix crystallization keeps the supersaturation low; fast crystallization from solution keeps it short. Either way the advantage of the amorphous form is lost. C1 Both routes need nucleation and then growth, and polymers can delay either step and so stabilize supersaturation; how much depends on the drug's own crystallization tendency, the polymer's chemistry, drug–polymer interactions and miscibility, the drug:polymer ratio and the temperature. C2

A further complication is polymorphism. The first crystals that form, in the matrix or from solution, may be a metastable crystal form that later converts to the stable one, and polymers are known to steer which form appears and how fast it converts: polymer heteronuclei selected two polymorphs of acetaminophen and all six forms of ROY (Price et al., ref 24), and some polymers greatly delayed the conversion of anhydrous carbamazepine to its hydrate in water (Gift et al. 2008, ref 26). C3·R19 The crystal form matters clinically: the authors recall that the stable form of chloramphenicol palmitate is biologically inactive (ref 28), that form II of sulfameter gave 1.4 times the bioavailability of form III, and that form II of sulphamethoxydiazine, twice as soluble as form III, was absorbed 40 times better (ref 29).

Figure 1 of the paper adapts the Alonzo scheme to include these steps: the amorphous solid in water may crystallize to a metastable form I in the matrix or, after dissolving, from solution; form II (stable) then grows at the expense of form I. If the matrix dissolves slowly, crystallization "likely" happens in the matrix; if it dissolves fast, the drug supersaturates and then crystallizes as form I, which converts to form II. C4

Nifedipine was chosen because it is a "relatively fast crystallizer" with several well-characterized polymorphs, so complex phase behaviour was expected (refs 31–36). M1 A second, methodological motive runs through the paper. Quantifying several solid forms usually requires calibration blends of physical mixtures, which is slow and difficult when metastable forms must be isolated and kept stable. Multivariate methods had been used for polymorph mixtures and degradation kinetics but rarely for crystallization in water. The authors therefore set out to use principal component analysis (PCA) and multivariate curve resolution (MCR) to get transformation kinetics directly from Raman spectra, without calibration sets. C5

The aims, in the authors' words, were to investigate the crystallization and phase-transition behaviour of amorphous nifedipine systems with a focus on kinetics, and the influence of the polymers PVP, HPMC and HPMCAS on these transitions. The polymers were tested in two positions: dissolved in the medium at 1 mg/mL, a level at which they do not raise nifedipine solubility (ref 17), and formulated with the drug as 50:50 ASDs. A1·A3·M2

Background concepts

Amorphous, matrix and solution crystallization. Amorphous nifedipine is called "glass" (g) in the paper. A wetted glass particle can crystallize where it is; under crossed polarizers the particle then becomes bright and coloured (birefringent) while keeping its size. In solution crystallization the particle shrinks as it dissolves and new, separate crystals appear in the liquid. Polarized light microscopy (PLM) can therefore tell the two routes apart by where the birefringence appears. P4

The crystal forms of nifedipine. Nifedipine has three main polymorphs, α, β and γ, and a fourth, form IV, has been reported; β and form IV are short-lived. α and γ are monotropically related, meaning α is the more stable form at all temperatures; α melts at 172 °C. M1 The authors estimate, from literature data, that the metastable form is about 2.4 times as soluble as α. M1 In this paper only g, γ and α were ever seen; neither β nor form IV was detected. R16

Raman marker bands. The forms are told apart by their Raman spectra. In the C–C–O stretching region the band sits at 1214 cm⁻¹ for the glass, 1224 cm⁻¹ for α, 1216 cm⁻¹ for β and 1205 cm⁻¹ for γ; in the C=C region at 1647, 1648, 1651 and 1644 cm⁻¹ (values from ref 36). M1 Note by the AI writer (not discussed in the paper): the Results section gives the α band in the C=C/C=O region as 1680 cm⁻¹ instead of 1648 cm⁻¹; the C–C–O positions, which the paper uses for its discussion, agree on both pages. X1

Solution-mediated phase transformation (SMPT). A metastable solid in contact with its solution can convert to a more stable form through the liquid. The authors describe three stages: the metastable phase dissolves and creates a supersaturated solution; the new phase nucleates, driven by that supersaturation; the new phase grows, depletes the solution and so lets more metastable phase dissolve. C34 If nucleation and growth are fast, dissolution sets the overall rate. C35

Multivariate curve resolution. MCR treats a series of spectra recorded over time as a mixture problem. The matrix of measured spectra is written as

In plain terms: each measured spectrum is a weighted sum of the spectra of the pure forms. $𝐃$ holds the measured spectra (intensity at each wavenumber for each time point); $𝐂$ holds the amount of each pure form at each time point (the kinetic curves); ${𝐒}^T$ holds the spectra of the pure forms; $𝐄$ is the residual that the decomposition cannot explain, with the same size as $𝐃$ (Eq. 1). E1 Such a decomposition is not unique ("rotational ambiguity"), so the authors constrained it: spectra and amounts must be non-negative, and the amounts of all forms must add up to one at every time point (closure, because mass balance is expected). P10·C7 The amounts that come out are therefore relative amounts between 0 and 1, not calibrated concentrations. X15

Principal component analysis. Before MCR the authors used PCA to check how many independent components the spectra contain; close to 96% of the variability in all data sets was explained by three components, matching the three forms g, γ and α. P9·R3

Methods in detail

The paper has three contributions: an in situ way of following several solid forms of a crystallizing amorphous drug in water without calibration samples; a direct comparison of polymers placed in the medium with the same polymers placed in the solid; and a set of side experiments (microscopy, a dissolution test, a low-pH test) that identify which crystallization route operates. The method flow is:

Contribution 1: following three solid forms in a slurry without calibration

① The problem. To see how polymers change crystallization and polymorph conversion, one needs the amount of glass, γ and α at each moment in a wet, crystallizing slurry. Conventional quantification would need calibration mixtures of all three forms in several ratios, including the unstable glass and the metastable γ form, which the authors describe as tedious and challenging. C5

② What they did. Amorphous nifedipine was made by heating the crystalline drug 10 °C above its melting point and quenching it in liquid nitrogen, then cryomilling (45 s at 5 impacts per second, 3 cycles) to particles of 1–300 µm; it was used immediately. P1 For each slurry, about 500 mg of this glass was added to 2 mL of 50 mM pH 6.8 phosphate buffer, either plain or containing 1 mg/mL PVP K29/32, HPMC (Pharmacoat 606) or HPMCAS (MF grade). ASD slurries contained 1 g of ASD in 4 mL buffer with no added polymer. Slurries were run at 25 and 37 °C. P6·P7 A non-contact fiber-optic Raman probe recorded a spectrum every minute for the first 15 min and then every 5 min up to 3 h. P6 The spectra were analysed in two ways. First, by hand: the bands at 1205 cm⁻¹ (γ) and 1224 cm⁻¹ (α), normalized, give a qualitative estimate of how much of each form is present; Table 1 summarizes which forms are present and which dominates over time, and Table 2 gives the time at which each form was first detected. P8·C8 Second, by PCA and MCR over 1000–1800 cm⁻¹: baseline offset and 20-point smoothing, normalized pure spectra as references, non-negativity and closure constraints, and at most 200 iterations of the alternating least squares fit. P9·P10

③ A key choice. No pure γ sample was available. The authors took as the γ reference a slurry time point at which only γ was present, and checked against literature spectra that no other form was present there. C6 The resolved spectra agreed with measured reference spectra (Figure 4): γ at about 1205 cm⁻¹, α at about 1224 cm⁻¹ and a broad amorphous band near 1214 cm⁻¹. R4·C9

④ How it differs from earlier work. Multivariate analysis had been applied to dry polymorph mixtures and to degradation, but rarely to crystallization in aqueous media; here it replaces calibration entirely. C5 Note by the AI writer (not discussed in the paper): because closure forces the three amounts to add up to one, MCR gives fractions of the solid forms, not their absolute amounts or the dissolved drug; and the only check of the MCR amounts is the qualitative peak-ratio analysis. X15

Contribution 2: polymer in the medium versus polymer in the solid

① The problem. Polymers dissolved in the medium had been shown to delay crystallization of amorphous felodipine, a structural analogue of nifedipine, and PVP and HPMC could inhibit felodipine matrix crystallization. R17 Would the same hold for the faster-crystallizing nifedipine, and does it matter whether the polymer is in the liquid or mixed into the solid?

② What they did. The same three polymers were used both ways: 1 mg/mL dissolved in the buffer around the neat glass, and as 50:50 w/w ASDs made by dissolving drug and polymer in 1:1 ethanol:dichloromethane, removing the solvent by rotary evaporation, drying overnight and cryomilling to 1–300 µm. A1·A3·P2 PXRD and PLM confirmed that the starting materials were amorphous (the patterns are not shown). R1

③ A caution on the comparison. Note by the AI writer (not discussed in the paper): the neat-glass slurries held about 250 mg/mL of drug with 1 mg/mL of polymer, while the ASD slurries held 125 mg/mL of drug and, once the ASD dissolved, up to 125 mg/mL of polymer. The polymer concentration reached in the ASD slurries is not reported, so differences between the two kinds of slurry are not due to the crystallization route alone. X9

Contribution 3: identifying the route

① Microscopy. Particles were sprinkled on a glass slide, wetted with 2–3 drops of buffer (with or without dissolved polymer), covered, and imaged under crossed polarizers at 10× and 40×. P4 Birefringence inside the particle means matrix crystallization; new crystals in the liquid mean solution crystallization.

② Dissolution. To check whether the ASDs really pass through a supersaturated solution, neat glass and the 50:50 HPMC ASD were dissolved in the same buffer at 37 °C with stirring at 500 rpm, and the concentration was followed by in situ fiber-optic probes (n = 3). The solids loading had to be 5 mg/mL, far below that of the Raman slurries, because dense suspensions scatter too much light. P11

③ The pH 2 test. To separate "the polymer stops the matrix from crystallizing" from "the matrix simply dissolves before it can crystallize", the HPMCAS ASD was put in pH 2 buffer, where HPMCAS, an ionizable polymer, is poorly soluble, so the matrix stays in place and can be watched for crystallization for 3 h. P5·C25

Experimental analysis

The neat glass crystallizes in its own particles within minutes

Under the microscope, a dark (amorphous) particle of the neat glass (Figure 3A) was covered with birefringent crystals 15 min after it was wetted with pH 6.8 buffer at 25 °C (Figure 3B). The picture was the same when PVP, HPMC or HPMCAS had been dissolved in the buffer at 1 mg/mL (Figure 3C–E). R2 The crystallized material kept the size of the original particles, 1–300 µm. R2 The authors conclude that neat amorphous nifedipine crystallizes rapidly, "primarily through the matrix route", and that the dissolved polymers changed neither the onset nor the route; even with polymer in the buffer, matrix crystallization was "instantaneous". C10·C11

The Raman data give the timing. Immediately after buffer was added, the spectrum already showed both the glass band (1214 cm⁻¹) and the γ band (1205 cm⁻¹); the α band (1224 cm⁻¹) appeared after 5 min by peak height and then grew while the γ band shrank; at 40 min most of the crystals were α, and small traces of γ were still seen at 3 h (Figure 5A). R5·C12 MCR resolves this into a clean sequence at 25 °C (Figure 5B): the glass amount stayed at 1 for about 2 min, fell to about 0.61 at 4 min and about 0.35 at 5 min, and was gone at 6 min; γ rose over the same minutes, stayed at 1 from 6 to 13 min, then fell (about 0.47 at 20 min) and reached 0 at 40 min; α appeared at 14 min, passed about 0.53 at 20 min and reached 1 at 40 min. R6 The authors read this as a crystallization step "reaching completion within about 10 min", followed shortly by the conversion of γ to α. C13 In the Discussion they give the half-life of γ in pure buffer as about 20 min. R7

Note by the AI writer (not discussed in the paper): the two analyses do not see the minor forms at the same time. By peak height α appears after 5 min and γ persists in traces at 3 h; by MCR α is zero until about 13 min and γ is zero from 40 min. X13 The early part of the same run also differs between figures: Figure 6A puts γ at about 0.38 already at 2 min, Figures 5B and 7A at 4 min. X12

Dissolved polymers do not prevent crystallization, but they hold the metastable form

With polymer in the buffer, the glass still disappeared within a few minutes and γ was the majority form at the first measurement at both temperatures (Table 1). R7·C16 What changed was the second step. At 25 °C the last MCR point at which γ was still 1 moved from about 13 min in plain buffer to about 20 min with PVP, about 60 min with HPMC and about 110 min with HPMCAS (Figure 7A); γ reached zero at about 40, 55, 110 and 145 min. R9 The α curves (Figure 8A) mirror this: α reached 1 at 40, 60, 105 and 145 min. R10 The authors rank the polymers HPMCAS > HPMC > PVP for stabilizing γ, "similar at each temperature", and describe HPMCAS as increasing the lag before conversion "from about 10 min to nearly 2 h". C17 Table A collects the numbers.

Table A. Neat amorphous nifedipine slurried in pH 6.8 buffer with or without dissolved polymer. MCR times read from Figures 7 and 8 (approximate); "α majority" is the time at which α is printed in red in Table 1. R7·R9·R10

The authors also draw a mechanistic inference from the shape of the curves. Polymers could change a phase transformation by adsorbing to the parent crystals and reducing heterogeneous nucleation of the new form on them, by inhibiting nucleation of the new form from solution, or by inhibiting its growth. C31 Because the lag before α appears is extended while, "particularly at 25 °C", the slopes of the curves after onset are similar with and without polymer, they consider it "most likely" that the polymers interfere with the nucleation of α; slower growth would have flattened the curves, as polymers do in anhydrate-to-hydrate transformations (refs 26, 61). C32·R19 Figure 8A supports the similar-slope reading at 25 °C: each α curve rises from near zero to 1 over about 25–35 min. R10 Note by the AI writer (not discussed in the paper): at 25 °C the γ curves for buffer, PVP and HPMCAS in Figure 7A take the same sequence of values on the way up (about 0.05, 0.38, 0.65, 1.0) and on the way down (about 0.88, 0.69, 0.47, 0.30, 0.16), only shifted in time; the paper states no replicates and does not comment on this. At 37 °C the HPMC and HPMCAS curves fall over about 70–95 min against about 20 min in plain buffer, so the similar-slope argument holds only at 25 °C, as the authors' own wording allows. X11

Note by the AI writer (not discussed in the paper): the peak-ratio table and the MCR curves agree on the order of events but not exactly on their times. The point where γ and α are about equal is 45 min for PVP at 25 °C in Table 1 but about 35–40 min in the MCR curves, and 110 against about 125–130 min for HPMCAS; the authors call the two "in good agreement". X5·C15 With closure, γ and α should add up to one once the glass has gone; they do for plain buffer at 25 °C, but with PVP at 37 °C γ is about 0.43 and α about 0.02 at 20 min, and at 25 °C the α rise with each polymer starts about 5 min after the γ fall. X6

Temperature: the conversion starts sooner at 37 °C

At 37 °C the neat glass also crystallized at once, and γ converted to α sooner: in plain buffer the γ plateau ended at about 9 min instead of about 13 min, and α was complete at 30 min instead of 40 min (Figure 6). R8·C14 With every polymer the γ plateau also ended earlier at 37 °C (Table A). R9 The authors summarize that transformation kinetics were "faster at a higher temperature". C14 Note by the AI writer (not discussed in the paper): this holds for the onset of conversion but not always for its completion: with HPMC α was complete at 135 min at 37 °C against 105 min at 25 °C, and with HPMCAS at 135 against 145 min; Table 1 also puts the α majority later at 37 °C for HPMC and HPMCAS. X7

In dispersions the matrix stays amorphous and the drug crystallizes from solution

Because dissolved polymer could not protect the neat glass, the authors asked what happens when the polymer is mixed into the solid. Under the microscope, particles of the 50:50 PVP dispersion in pH 6.8 buffer dissolved without becoming birefringent, while small plate-shaped crystals appeared in the surrounding liquid within minutes and multiplied up to 10 min (Figure 10). R13 These crystals were smaller than 2 µm, against 1–300 µm for crystallized neat particles; the authors report the same for the HPMC and HPMCAS dispersions, without showing images. R13·C22·X18 They write that the images "strongly suggest" that the route of crystallization changed from the matrix to the solution. C21

The dissolution test supports the supersaturation step. At 37 °C and 5 mg/mL solids, the HPMC dispersion rose to about 100 µg/mL (SD about ±13) at about 22 min and then fell to about 72 µg/mL at 50 min and about 41 µg/mL at 150 min, while the neat glass stayed at about 6 µg/mL throughout, on the crystalline solubility line (Figure 11). R14·M1 The authors read this as the dispersion "indeed" dissolving to a supersaturated solution that then crystallizes, while the neat glass reaches only about the solubility of α, "consistent with" crystallization in the matrix. C23 Note by the AI writer (not discussed in the paper): this is the only measurement of supersaturation in the paper; it was made for the HPMC dispersion only, at a 50-fold lower solids loading than in the Raman slurries, so the supersaturation inside the slurries is inferred, not measured. X10

The pH 2 test then separates the two possible reasons why the dispersion matrix does not crystallize. In pH 2 buffer the HPMCAS dispersion did not dissolve quickly and stayed non-birefringent; a few small crystals appeared in the liquid after about 1 min and were visible at 10 and 20 min, but they did not grow substantially over 3 h; at 3 h the matrix was still free of birefringence and Raman detected only amorphous nifedipine (Figure 12; the text refers to these images as Figure 11 B–D). R15·X3 The authors conclude that the polymers "do prevent" crystallization of the hydrated ASD matrix, while crystallization from the solution is extremely rapid. C24

Dispersions crystallize and convert faster than the neat glass

The most surprising result, in the authors' word, concerns speed. In all three dispersions γ was detected at 1 min. α was first detected at 3, 25 and 30 min at 25 °C and at 4, 7 and 9 min at 37 °C for PVP, HPMC and HPMCAS (Table 2). R11 By MCR at 25 °C (Figure 9), the PVP dispersion had converted from γ to α between about 3 and 5 min, and the HPMC and HPMCAS dispersions reached full α at 40 min, with α rising from about 7–8 min. R12 The authors find, "surprisingly", that both the amorphous-to-γ and the γ-to-α conversions were "much faster" for the ASDs than for the neat glass, and that HPMC and HPMCAS "appeared to be superior" to PVP at impeding the γ-to-α conversion, with none of the polymers delaying the onset. C20·C18

Table B. Dispersions slurried in pH 6.8 buffer without added polymer. First-detection times from Table 2 (peak heights); MCR times read from Figure 9 (approximate, 25 °C only); the last column from Figure 8A for comparison. The neat glass in plain buffer reached all α by 40 min. R11·R12·R10

Note by the AI writer (not discussed in the paper): three points qualify this result. First, Table 2 and Figure 9 disagree for the cellulose dispersions: Table 2 gives first α at 25 and 30 min at 25 °C, while Figure 9B shows α from about 7 and 8 min and above 0.5 by about 15–20 min; the authors call the two "in good agreement". X4·C19 Second, the text says Figures 9A and 9B show 25 and 37 °C, but the figure shows 25 °C only, so the dispersion kinetics at 37 °C rest on Table 2 alone. X3 Third, against the neat glass in plain buffer (all α by 40 min) the HPMC and HPMCAS dispersions are about as fast; the large difference is against the neat glass in polymer solutions, and the slurries also differ in drug and polymer amount. X8·X9

Discussion

The authors' chain of reasoning

The Discussion ties the results together with one idea: whether amorphous nifedipine crystallizes in the particle or in the solution, and how fast it then converts to α, depends on how fast the solid dissolves compared with how fast the drug nucleates and grows. C35·C33 The chain, as the authors give it, is:

Diagram by the AI writer from the authors' Discussion. "Good" marks the branch that reaches supersaturation, "bad" the one that does not; it does not mean the dispersion avoids crystallization.

Each step carries a hedge. Dispersions dissolve faster than the pure amorphous material because the polymeric carriers are highly soluble and dissolve fast (ref 62), and the faster conversion of the dispersions is "most likely explained" by this faster dissolution. C33·R21 For nifedipine nucleation and growth are fast, "as they appear to be", so dissolution controls the solution-mediated transformation. C35 The dispersion "may" reach a higher supersaturation before crystallizing; higher supersaturation "would" give faster nucleation and many small crystals, which were indeed seen; these small crystals have a larger surface area than the original powder, which speeds their dissolution and offers more surface for heterogeneous nucleation, "which may explain" the faster γ-to-α conversion. C36·R13 When dissolution is very slow, as for the HPMCAS dispersion at pH 2, supersaturation stays low and nucleation and growth from solution are slow. C37·R15

None of the quantities in this chain (dissolution rate in the slurries, supersaturation at the moment of nucleation, number or surface area of the crystals, nucleation rate of α) was measured. X10·C36 The chain is therefore the authors' explanation of the observations rather than a tested mechanism, and the paper presents it with the hedges quoted above.

Why nifedipine behaves differently from felodipine

Earlier studies found nifedipine to crystallize faster than its structural analogue felodipine, although amorphous powders of both crystallize immediately on contact with buffer; dissolved polymers delay felodipine but, in this paper, not nifedipine. R17·R2 The authors conclude that the differences between the two drugs in aqueous environments "only manifest themselves in the presence of polymers", and, since the two drugs have similar glass transition temperatures and moisture sorption, that "it appears" the faster inherent crystallization of nifedipine carries over to these experiments. C27·C28 All felodipine data are cited from earlier work; no felodipine experiment is part of this paper. X16

The metastable γ form as the common first step

In every system studied, crystallization started with γ and ended with α, which the authors call "the common pattern" for the amorphous systems studied; their Conclusions state it without hedge. C29 This contrasts with crystallization of nifedipine from organic solvents, where α forms directly or β appears briefly (ref 36). R18 The authors note that metastable forms are common in melt crystallization because conversion is slow in a viscous melt, but "may be difficult" to see in solvents because conversion is fast; the short sampling interval of the in-line Raman probe let them catch γ, whose half-life in plain buffer was about 20 min. C30·R7

What it suggests for formulation

The authors draw two practical points. First, protecting the amorphous form from any crystallization is the goal, but for molecules whose crystallization polymers cannot hold back this "may not always be feasible"; then stabilizing a metastable polymorph "can be advantageous". C38 Metastable forms that persist for a biologically relevant time "can lead to" higher residual solution concentrations, and both the free energy difference and the persistence matter: for mefenamic acid the solubility ratio of the forms is only 1.2 and the plasma levels are about the same, whereas for nifedipine the metastable form is estimated at 2.4 times the solubility of α, so the conversion rate "could be of biological significance". C39·C40·R20·M1 Second, they state that the polymers in the nifedipine dispersions, while not preventing crystallization from the highly supersaturated solution, stabilized the metastable polymorph and gave a well-dispersed, fine precipitate, and that this complex behaviour "can only be understood" with high-resolution, real-time monitoring of the precipitating solid. C41·C42

Note by the AI writer (not discussed in the paper): three cautions apply when using these points. The "stabilized the metastable polymorph" statement for dispersions holds relative to PVP only; in Figure 9 γ is gone within about 35 min in the HPMC and HPMCAS dispersions, no later than for the neat glass in plain buffer. X8 The polymer ranking HPMCAS > HPMC > PVP for holding γ is consistent across both polymer positions, which makes the cellulose polymers the more promising choice when a more soluble metastable form is worth keeping. R9·R12 And all kinetics come from very concentrated slurries (125–250 mg/mL drug), so the minutes reported here are not dissolution or in vivo times. P6·P7·X9

Limitations

Limitations stated by the authors. The authors note that MCR is affected by rotational ambiguities and constrained it with non-negativity and closure for that reason. C7 Their peak-ratio summary (Table 1) is qualitative, and small amounts of a form printed in black "may still be present". C8·R7 The dissolution test had to use 5 mg/mL solids, much lower than the Raman slurries, because dense suspensions scatter too much light for absorbance readings. P11 In the pH 2 test, Raman could not detect the small amount of crystalline material that formed in solution. R15

The AI writer's own judgement (1): what the kinetic data can carry. No replicate number is given for any Raman slurry, so each kinetic curve should be read as a single run. X14 Several timings disagree between the two analyses or between figures of the same run: Table 1 against MCR by up to about 15 min, Table 2 against Figure 9 by about 20 min for first α in the cellulose dispersions, and the early points of the neat-glass run by about 2 min between Figures 5B, 6A and 7A. X5·X4·X12 The γ and α curves do not always add up to one despite the closure constraint. X6 At 25 °C, three of the four γ curves in Figure 7A repeat the same sequence of values, shifted in time, which is the basis of the "similar slopes" argument for nucleation inhibition. X11 None of this changes the order of the polymers, whose lags differ several-fold, but it limits any use of the curves for rate constants or slopes.

The AI writer's own judgement (2): what the mechanism rests on. The explanation of the faster dispersion conversion (faster dissolution, higher supersaturation, many small crystals) is not measured in the system where the conversion was measured: supersaturation was determined only for the HPMC dispersion at 5 mg/mL and 37 °C, and the dispersion and neat-glass slurries differ in drug and polymer amount as well as in route. X10·X9 The nucleation-inhibition reading of the polymer effect rests on curve shapes at 25 °C; at 37 °C the cellulose curves are clearly shallower. X11 "Faster at higher temperature" holds for the onset of conversion but not for its completion with HPMC. X7 The MCR amounts are relative fractions under closure and were checked only against the qualitative peak ratios, so "quantitative" here means relative timing, not calibrated amounts. X15

The AI writer's own judgement (3): editorial errors. The caption of Figure 5 names the α form where the plot shows γ forming first; the text says Figure 9 covers 25 and 37 °C although it shows 25 °C only; the faster dispersion conversion is referred to Figures 7A and 8B, which show neat glass; and the pH 2 images are cited as Figure 11 instead of Figure 12. The α band in the C=C region is given as 1648 cm⁻¹ in the Methods and 1680 cm⁻¹ in the Results. X2·X3·X1 None of these changes the conclusions.

Conclusion

The paper shows, with in-line Raman spectroscopy and microscopy, that amorphous nifedipine in pH 6.8 buffer crystallizes within minutes and always passes through the metastable γ form before the stable α form. R6·C29 Where the polymer sits decides what it can do. Dissolved at 1 mg/mL, PVP, HPMC and HPMCAS cannot stop the neat glass from crystallizing in its own particles, but they delay the γ-to-α conversion, from about 13 min in plain buffer to about 110 min with HPMCAS at 25 °C, which the authors attribute "most likely" to inhibited nucleation of α. C11·R9·C17·C32 Mixed into the solid as 50:50 dispersions, the polymers keep the wetted matrix amorphous, but the drug dissolves, supersaturates (about 100 µg/mL for the HPMC dispersion against about 6 µg/mL for the neat glass) and crystallizes from solution as fine crystals, with both conversions faster than for the neat glass in polymer solutions, which the authors attribute "most likely" to faster dissolution. C24·R14·C20·C33 The authors call MCR "indispensable" for obtaining these kinetics without calibration curves; the AI writer reads it as a reliable tool for the order and relative timing of the forms, given the timing differences and the lack of stated replicates. C43·X15·X14

For formulation, the study is an early, clear demonstration that for a fast crystallizer the polymer must be in the solid to protect the glass, and that even then a fast-dissolving dispersion may only move crystallization from the particle into the solution. The open questions are the ones the authors' own hedges point to: which polymer mechanism (nucleation, growth or adsorption) holds γ, how much supersaturation the dispersions actually reach in the slurries, and whether a metastable form about 2.4 times more soluble than α persists long enough in vivo to matter. C31·C32·X10·C40