Where does PVP slow crystal growth in amorphous felodipine? Two polymorphs as a test

Background and motivation

An amorphous form can lift the low dissolution rate, solubility and bioavailability of a poorly soluble crystalline drug, but the amorphous solid is physically unstable: once it crystallizes, the advantage is lost (Law 2004, ref 1; Kennedy 2008, ref 2). Predicting how fast an amorphous drug crystallizes at storage temperature is therefore a practical problem in formulation, and crystal growth kinetics, measured as the advance of a crystal front into the supercooled liquid or glass as a function of temperature, is one of the main tools for it.

Crystallization near room temperature is often too slow to follow, so growth is measured at higher temperatures and extrapolated down to near or below the glass transition temperature (Tg). That extrapolation assumes that growth slows in a predictable way and is controlled by bulk properties such as viscosity. Work by Yu and co-workers on indomethacin and nifedipine showed that this assumption can fail: those glasses show a sudden jump in growth rate near and below Tg, and crystals grow faster at a free surface than inside the material (Wu 2006, ref 3; Ishida 2007, ref 4; Zhu 2008, ref 5). If such modes are active, rates measured at high temperature may not predict low-temperature rates, and the stability of the amorphous formulation would be overestimated. C7

Polymers are the usual way to stabilize amorphous drugs against crystallization (Ford 1986, ref 6; Serajuddin 1999, ref 7). The Taylor group had studied felodipine, a calcium channel blocker, with poly(vinyl pyrrolidone) (PVP) in detail: PVP delays felodipine crystal growth, and the effect was "mainly attributed" to a specific interaction between PVP and felodipine (Kestur 2010, refs 8 and 9). What remained "not clear" was where that interaction acts: at the surface of the growing crystal, as polymers are thought to act in crystal growth from solution, or in the amorphous material from which the crystal grows. C8·C28

The paper sets out to answer that question by comparing the effect of temperature and of PVP on the growth rates of two felodipine polymorphs, form I and form II. A secondary aim is to see whether felodipine polymorphs also show the additional growth modes near Tg that had been reported for indomethacin and nifedipine. C1

Background concepts

Free-surface and bulk growth. In a thin amorphous film held between two cover glasses, crystals grow inside the material (bulk growth). If the top cover glass is removed, crystals can nucleate and grow at the exposed free surface. In several organic glasses surface crystals grow much faster than bulk crystals: for indomethacin, nifedipine and griseofulvin by one to two orders of magnitude, and indomethacin molecules diffuse about a million times faster at the surface than in the bulk (refs 3, 5, 11, 16). The usual explanation is that molecules at the surface are more mobile, the surface behaving more like a liquid and the interior more like a solid; coating surface crystals with a thin gold layer slows them to bulk-like rates (refs 5, 10, 12). R17·C16

Diffusion-controlled growth and the glass-crystal (GC) mode. In ordinary, diffusion-controlled growth, molecules must move to the crystal front before they can join it, so growth slows as the liquid becomes more viscous on cooling. A convenient yardstick is the number of molecular layers added to the crystal during one structural relaxation time of the liquid (the time for molecules to rearrange): for diffusion-controlled growth this is about 0.1–1 layer (Sun 2008, ref 13). Yu and co-workers found that some glasses, including o-terphenyl, indomethacin, nifedipine and the polymorphic compound ROY, show an abrupt increase in growth rate as they are cooled towards Tg, at a "transition temperature", and that the fast growth continues into the glass. This glass-crystal (GC) growth adds far more layers per relaxation time, 1700–2400 for different ROY polymorphs (Sun 2008, ref 14), too many for diffusion or structural relaxation to keep up, so it is called diffusionless; the addition of new layers has been related to fast oscillatory motions available in the glass and liquid. R16·C22 Above the transition temperature, the GC mode is thought to survive as fast-growing fibers: when the growth rates of these fibers and of GC growth are plotted against temperature they lie on one smooth line, so the "jump" belongs to the spherulite front, not to the individual fibers (refs 3, 4, 13, 14, 18). C21 For ROY, the polymorphs that show the fast mode are denser and more isotropically packed (ref 13). C24

Why two polymorphs make a test. Polymorphs are different crystal structures of the same compound. Form I and form II of felodipine grow from the same supercooled liquid but expose different crystal surfaces to it. If PVP slowed growth by adsorbing to crystal faces, its effect would be expected to depend on the face structure and so to differ between the polymorphs; if it slowed growth by changing the liquid, the same factor would apply to both. The paper's measure is the ratio of the growth rate without polymer to the growth rate with PVP, computed for each polymorph at each temperature: a ratio of 10 means PVP slows growth tenfold. C1·E2

Spherulites. Most of the crystals in this study are spherulites, roughly round aggregates of radiating crystallites that grow as one front; in this study their size increased linearly with time, and the slope is the growth rate. R5·P6

Methods in detail

The study is purely experimental: growth rates of the two polymorphs, at the free surface and in the bulk, with and without PVP, over a wide temperature range, followed by two derived quantities. The flow below summarizes it.

Making the films

Crystalline felodipine form I, a gift from AstraZeneca, was melted onto an 18 × 18 mm cover glass for 3 min, covered with an 18 mm round cover glass and quenched by placing it on a cold surface. The amorphous films were 5–10 µm thick, and thickness did not affect the growth rate. Films cooled on a hot stage at 10 or 20 K/min grew crystals at the same rates as quenched films, which the authors read as no effect of cooling rate. M1·P1·R4·C9

For the PVP samples a two-step route was used: a 10 wt% dispersion of PVP K-29/32 in felodipine was made by rotary evaporation from a 1:1 (by weight) ethanol–dichloromethane solution and dried for 48 h in a vacuum oven, then diluted to the required level by blending with crystalline felodipine in a cryogenic mill and melt-quenched like the pure drug. The level used is given only in the figure captions, as "3% PVP"; the basis (weight or mole) is not stated. All samples were stored over phosphorous pentoxide to keep them dry. P4·P5·M4·X9

Surface, bulk and the choice of polymorph

Which polymorph grows depends on the geometry. Films without a top cover crystallized to form I, always faster than covered films, and faster still when the cover was removed below Tg than above it. Covered films that were not seeded crystallized to form II, first along the edges within 2 days and at the centre within 5 days (the storage temperature is not stated); Raman spectra and the melting point on the hot stage identified form II, as in the group's earlier work. R1·R2·X10 Form I in the bulk was therefore grown by seeding: either form I crystals were placed against the edge of the supercooled liquid, or the felodipine was only partly melted so that the remaining crystals acted as seeds. Both gave the same rates. P3·R4·C10 Surface growth of form II could not be measured, because crystals at the free surface always nucleated as form I. R3

Measuring growth rates

Growth was followed with a Nikon Eclipse E600 polarized light microscope. Temperature was held either by a hot stage (calibrated with the melting points of naphthalene and adipic acid) or by temperature-controlled ovens, from which samples were removed at intervals. A separate sample was prepared for each temperature and every measurement was made in triplicate; when growth took hours to days, at least three crystals were followed per sample. Crystal size rose linearly with time and the slope was taken as the growth rate; the figures show the mean of three readings and the standard deviation. P6·M8

New data cover form I at 18–100 °C (surface and bulk) and form II at 40–60 °C (pure) and 60 °C (with PVP). Form II and PVP data at 70 °C and above come from the group's earlier papers; the methods cite ref 8 for the range 70–110 °C, while the caption of Figure 2 cites ref 9, and the plotted data end at 105 °C. P6·R8·R13·X4 For form I in the bulk at 50–70 °C the crystals were fiber-like, and the reported rate is the advance of the thickened fiber front, not of the faster single fibers that grow out of it. R6·P6

Two derived quantities

The growth-rate ratio. For each polymorph and temperature, the authors divided the bulk growth rate without polymer by the bulk growth rate with 3% PVP (Figure 4). The relation is stated in words; the inputs are plotted, not tabulated. E2·R15

Molecular layers per relaxation time. To test whether the fast growth of form I near Tg can be diffusion-controlled, the authors estimated how many molecular layers are added during one structural relaxation time at 45 °C, just above the stated Tg of 43.5 °C. They used a bulk growth rate of 2.6×10⁻⁹ m/s, a relaxation time "taken as" 100 s and an assumed molecular radius of 9 Å, and obtained about 500 layers for form I and less than one layer for form II. No formula is printed; the estimate is the distance grown in one relaxation time divided by the molecular size. E1·R18·M3 Card check: 2.6×10⁻⁹ m/s × 100 s = 2.6×10⁻⁷ m, which divided by 9 Å gives about 290 layers (about 140 per 18 Å diameter), not 500; and the quoted rate matches the plotted bulk rate at 40 °C, while the plotted 45 °C point is about 4.2×10⁻⁹ m/s, which would give about 470 layers. X1·X2

Experimental analysis

Figure 2 holds the core data without polymer; Table A transcribes it. The values were read from the plotted points and are approximate.

Table A. Crystal growth rates of pure felodipine, read from Figure 2 (approximate; mean of three, error bars not transcribed). Form II at 70 °C and above is data reported earlier by the group. R8

Form I grows faster at the free surface, but only below about 65 °C

From 100 °C (Tg + 56.5 °C) to 18 °C (Tg − 25.5 °C) both form I rates fall by about 3–4 orders of magnitude, from about 1.4×10⁻⁶ to about 5.4×10⁻¹⁰ m/s at the surface and from about 1.2×10⁻⁶ to about 8.7×10⁻¹¹ m/s in the bulk. R9·R8.m·R8.a Above 70 °C the surface and bulk points coincide; below 65 °C the surface crystals grow faster, about 6 times faster at 18 °C. R9·C2 The discussion opens with the statement that the results "clearly show" faster growth from the free surface at lower temperatures, "approximately 5-6 times faster closer to and below Tg", and concludes that felodipine is another pharmaceutical glass with surface-enhanced growth. C2·C15 Card check: in the plotted means the surface/bulk factor is about 6.2 at 18 °C and about 5.7 at 30 °C, but about 3.5–3.7 at 40–45 °C, and about 1.5 at 70 and 80 °C. X13

In felodipine the surface factor is several-fold, whereas one to two orders of magnitude have been reported for indomethacin, nifedipine and griseofulvin; the paper cites both without comparing them. R9·R17 The crystal habit changes with temperature too (Figure 1): surface spherulites are compact and smooth-edged below 40 °C, less dense with a rough edge at 45–65 °C and zigzag-edged above 70 °C, while bulk crystals are dense compacts below 40 °C, fiber-like at 50–70 °C and faceted at higher temperature. R5·R6

Near Tg, form I growth jumps

Between 50 and 45 °C the form I rate rises on cooling instead of falling, for surface and bulk crystals alike, and the faster growth continues below Tg. R10 The text gives a 2–3-fold increase from 50 to 45 °C and places the transition temperature between 45 and 50 °C ("it apparent", as printed); the change from the thickened fiber-like crystals just above Tg to the new mode is abrupt. R10·C19 In the plotted means the bulk rate goes from about 1.7×10⁻⁹ to about 4.2×10⁻⁹ m/s (about 2.4-fold) and the surface rate from about 1.0×10⁻⁸ to about 1.6×10⁻⁸ m/s (about 1.6-fold); at 40 °C both rates are lower again (about 9.1×10⁻⁹ and 2.6×10⁻⁹ m/s) and they keep falling towards 18 °C. R8.c·R8.d·R8.e·X13

The authors call this "most likely" the GC mode described by Yu and co-workers. C3 The estimate of about 500 molecular layers per relaxation time at 45 °C, against 0.1–1 for diffusion-controlled growth, is "supporting the supposition" that form I grows in this diffusionless mode. C23·R18·R16 Fast-growing single fibers, like those Sun and co-workers saw above the transition temperature in other glasses, were also seen in form I at 60 °C (Figure 1e), but their curvature prevented measurement. C20·R6

Form II grows more slowly and smoothly

Form II slows smoothly with cooling, by about 5 orders of magnitude from 105 to 40 °C (about 1.7×10⁻⁷ to about 5.5×10⁻¹³ m/s), with no jump. R11·R8.n·R8.c At 45 °C fewer than one molecular layer is added per relaxation time, so in the authors' reading form II lacks the fast mode. R18·C23 The abstract and discussion put the bulk growth of form I "1-2 orders of magnitude" above that of form II; the conclusions call it "substantially higher". R12·C4 Card check: the ratio of the plotted bulk means is about 4700 at 40 °C, about 87 at 50 °C, about 18 at 60 °C and about 15 at 70 °C, but only about 7–9 at 80–100 °C. X5

Following the ROY work, the authors suggest that whether a polymorph shows the fast mode "is thought to be related" to its crystal density and the anisotropy of its molecular packing, and note that form I is denser than form II (1.45 vs 1.42). C24·M1·M2 Card check: the density unit is printed as "g/cm" with a superscript reference 19, a paper on crystal growth kinetics; ref 20 (Surov et al. 2012, on felodipine polymorphs) is listed but never cited, so the source is probably ref 20 and the unit g/cm³. X3

What 3% PVP does

Table B. Crystal growth rates from the 3% PVP felodipine dispersion, read from Figure 3 (approximate). Form II at 70 °C and above is data reported earlier (ref 8). R13

With PVP every rate is lower than without it, and the qualitative pattern of form I survives (Figure 3). R13·C5·C12 The fastest growth is at 100 °C for form I and 105 °C for form II, where the text describes the rate as "similar to pure felodipine". R14 Card check: at 100 °C the ratios in Figure 4 are still about 3.5 (form I) and about 6.5 (form II), and the abstract says PVP reduced growth "at all temperatures"; "similar" can only mean within a factor of a few on the log scale. X8 Form I slows by about 3–4 orders of magnitude down to 50 °C, its surface crystals grow 1–1.5 orders of magnitude faster than its bulk crystals (about 9–21 times at 30–50 °C in the plotted means), and it shows the sudden rise in the region of Tg once more. R14·X13 The authors place this rise "at around the same temperature" as without polymer; in the plotted means it lies between 45 and 40 °C, about 5 °C lower than without PVP, and is larger: surface from about 5.9×10⁻¹⁰ to about 3.3×10⁻⁹ m/s and bulk from about 6.8×10⁻¹¹ to about 3.2×10⁻¹⁰ m/s. C12·R13.b·R13.c·X14 Form II with PVP declined smoothly and grew too slowly below 60 °C to be measured; from the pure form II curve a jump "is not expected". R14·C13 The crystal habits of form I with PVP resembled those without it. R7

The ratio test

Table C. Ratio of the growth rate without polymer to the growth rate with 3% PVP (bulk), read from Figure 4 with the ends of its error bars (approximate). R15

Figure 4 is the paper's key test. Both ratios fall with rising temperature, from about 16–19 at 60 °C to about 3.5–6.5 at 100 °C, so within this range 3% PVP is a stronger brake the lower the temperature. R15 At 60–80 °C the two polymorphs' ratios lie within each other's error bars; at 90–100 °C the form II ratio is higher. R15·X6 The authors summarize: "within statistical error, there appears to be little difference" between the polymorphs (results), "little or no difference" at 60–100 °C (discussion), so that PVP reduces the growth rate "by approximately the same magnitude" for each polymorph. C14 Card check: dividing the plotted means of Figure 2 by those of Figure 3 reproduces Figure 4 for form II at every temperature and for form I at 90–100 °C, but not for form I at 70 and 80 °C, where it gives about 29 and 22 instead of about 17.5 and 13.2; the paper does not say how the ratios and their error bars were computed. X7

Discussion

Where PVP acts

The authors' argument runs as follows. Polymorphs have different crystal interfaces but grow from the same liquid; in crystal growth from solution, trace polymer is generally accepted to act by adsorbing to crystal faces, which would be expected to affect different faces differently. Since PVP lowered the growth rate of form I and form II by about the same factor at 60–100 °C, the authors infer that the PVP–felodipine "interacting environment" is similar for both polymorphs and that the polymer's effect has its origin in the properties of the liquid phase ("suggest", "suggesting"). The abstract states the same conclusion more firmly: the inhibitory effect "most likely arises from an effect on the amorphous material rather than at the crystal surface". C1·C28·C14·C6 Neither the adsorption of PVP on either crystal nor any property of the liquid, such as its mobility or viscosity, was measured. C6

The chain of reasoning in the discussion and conclusions; steps 1–2 are measured, step 3 and the branches are the authors' inference. C1·R15·C6·C26

What the authors draw from it for formulation

Because PVP removes about the same fraction of the growth rate whichever polymorph grows, a fast-growing crystal phase will need relatively more polymer to bring its growth down far enough to retard crystallization over the desired timeframe; the conclusions word this as "suggesting". C26 Since amorphous systems often crystallize to polymorph mixtures or to a metastable form, the authors advise that studies of crystallization from amorphous systems should "evaluate which polymorphic form is produced" and how additives affect its growth. C27

The surface results feed into stability prediction. Earlier work on amorphous felodipine powders, with and without PVP, found two stages of crystallization, an initial fast stage attributed to the surface and a slow stage attributed to the bulk (Kestur 2013, ref 17). In the temperature range where surface growth outpaces bulk growth, predictions from elevated-temperature data that ignore the surface mode "may" overpredict the stability of the amorphous solid. C17·R19·C18 The much faster growth of form I at the surface and especially below Tg, together with its tendency to nucleate at the surface, "most likely explains" why amorphous felodipine powders crystallize only to form I (refs 17, 21). C25·R19

Note by the AI writer (not discussed in the paper). Two features of the data deserve attention for a formulator. First, the ratio falls from about 16–19 at 60 °C to about 3.5–6.5 at 100 °C, so 3% PVP is a much weaker brake at high temperature; accelerated tests run well above Tg would understate the benefit of the polymer, just as they understate surface and GC growth. R15 Second, at 60 °C form I in the bulk still grows with 3% PVP at about 3.9×10⁻¹⁰ m/s, close to form II without any polymer (about 2.8×10⁻¹⁰ m/s): the polymer buys about one order of magnitude, while the choice of polymorph spans one to several. R13.e·R8.g

Limitations

Limits the authors state. The results are "complicated" because form I shows an enhanced growth rate closer to Tg, and because the slow growth of form II precluded measurements with PVP at lower temperatures in a reasonable time. C29 They suggest that it "would thus be of interest" to study a polymer's effect on a different, faster-growing polymorphic system in which the fast growth mode does not occur. C30 For form II with PVP, the absence of a jump below 60 °C is an expectation ("is not expected"), not a measurement. C13

The AI writer's own judgement (1): the ratio test is narrow. One PVP level was used, given only in the captions as "3% PVP" without a basis, so neither the equal-ratio result nor the claim that faster-growing crystals need more polymer was tested at other polymer levels. X9·C26 The ratios cover 60–100 °C, above Tg, while the behaviour that matters for storage, including the GC mode of form I, lies below 60 °C. R15·C29 Within the range, the two polymorphs agree within error at 60–80 °C, but at 90–100 °C the form II ratio is about 1.2–1.9 times the form I ratio, with error bars apart. X6 And for form I at 70 and 80 °C the values in Figure 4 do not follow from the plotted rates of Figures 2 and 3 (about 17.5 and 13.2 in Figure 4 against about 29 and 22 from the plotted means). X7 With the ratio of plotted means, the form I ratio at 70–80 °C would exceed the form II ratio by nearly two-fold, the opposite of the gap at 90–100 °C. X7·X6

The AI writer's own judgement (2): what equal ratios can and cannot show. Equal factors for two polymorphs are what a liquid-phase mechanism predicts, but they would also follow if PVP adsorbed to both crystal surfaces to a similar extent; the paper offers no independent measure of either mechanism, so the conclusion rests on the expectation that different faces would respond differently. The authors' own hedges ("most likely", "suggest") fit this. C6·C1

The AI writer's own judgement (3): the diffusionless-growth estimate. The estimate of about 500 layers per relaxation time uses an assumed relaxation time of 100 s, a molecular radius of 9 Å and a growth rate of 2.6×10⁻⁹ m/s that matches the plotted 40 °C point rather than the 45 °C point; the printed inputs give about 290 layers. X1·X2 Any of these versions is two orders of magnitude above the 0.1–1 layers of diffusion-controlled growth, so the qualitative inference stands, but the number itself should not be quoted as measured. R16·X2

The AI writer's own judgement (4): smaller points. Several factors quoted in the text (2–3 times from 50 to 45 °C, 5–6 times surface over bulk near Tg, 1–2 orders between the polymorphs) hold only at some temperatures in the plotted means. X13·X5 Form II data at 70 °C and above, with and without PVP, come from earlier papers, and the caption and the methods cite different references for them. X4 The growth front that is measured changes habit with temperature, from compact spherulites to a thickened fiber layer to faceted crystals, so rates at different temperatures do not refer to the same kind of front. X12 The crystal densities carry a misprinted unit and probably the wrong reference number. X3 The temperatures of the unseeded crystallization observations are not given. X10

Conclusion

The paper makes three experimental points about amorphous felodipine. Form I grows faster at a free surface than in the bulk below about 65 °C, about 6 times at 18 °C. R9·C2 Form I also shows a jump in growth rate between 50 and 45 °C, near the Tg of 43.5 °C, which the authors "most likely" assign to the diffusionless glass-crystal mode, while form II grows more slowly and its rate falls smoothly by about 5 orders of magnitude from 105 to 40 °C. R10·C3·R11·C11 With 3% PVP both polymorphs grow more slowly, and the ratio of rates without and with polymer is similar for the two at 60–100 °C, which the authors read as PVP acting through the amorphous phase rather than at the crystal surface, so that faster-growing polymorphs would need more polymer. R15·C14·C6·C26

For formulation the lasting messages are methodological: know which polymorph crystallizes, measure its growth near and below Tg and at a free surface, and do not expect a polymer to rescue a fast-growing form by the same margin it gives a slow one. The mechanistic conclusion about where PVP acts rests on a single polymer level and on ratios measured above Tg, and the authors hedge it ("most likely"). C27·C18·C26·C6·X9