Polymer choice and storage humidity: how fast amorphous felodipine crystallizes from PVP and HPMCAS dispersions

Background and motivation

An amorphous solid dispersion (ASD) is a mixture of an amorphous drug, the active pharmaceutical ingredient (API), with a second amorphous component, usually a polymer. Because the amorphous form has a higher free energy than the crystal, it has a higher aqueous solubility, and this can raise the bioavailability of poorly soluble drugs, as shown for ritonavir (Law et al.) C1. The approach only works if the drug stays amorphous during storage, and the paper starts from the observation that polymers slow the crystallization of many drugs, which is why ASDs are "commonly considered" a viable route to amorphous products with adequate physical stability C1.

The polymers that do this job, among them poly(vinyl pyrrolidone) (PVP), poly(vinyl pyrrolidone-co-vinyl acetate) (PVPVA), hypromellose and hypromellose acetate succinate (HPMCAS), are often hygroscopic, so a dispersion can take up water during manufacture or storage C2. Water is known to harm physical stability: it promoted crystallization of amorphous indomethacin (Andronis et al.) and of amorphous sucrose–PVP (Shamblin and Zografi) C3. For felodipine, the drug of this paper, earlier work from the same group had shown that the nucleation rate in various dispersions rose with moisture content (Konno and Taylor 2008), and that the crystal growth rate, with or without polymer, in general rose with moisture (Rumondor, PhD thesis 2009) C3·R22·R23. A third strand of earlier work matters for the discussion: felodipine–PVP dispersions exposed to around 75% relative humidity (RH) and above were reported to split into drug-rich and polymer-rich amorphous phases (Konno and Taylor 2008; Marsac PhD thesis 2007; Marsac et al. 2009) R24.

What was missing, according to the authors, was a systematic study of how polymer type and storage RH together affect the bulk crystallization rate of a dispersion, the quantity a formulator sees in a stability study. Nucleation and growth had been measured separately; the overall outcome had not. The paper fills this gap with one drug, felodipine, chosen because its nucleation and growth under moisture were already known, and two polymers chosen because they differ in hygroscopicity: PVP, which takes up much water, and HPMCAS, which takes up less C4·R25. The read-out is the percentage of felodipine that has crystallized, measured by powder X-ray diffractometry (PXRD) over up to two years P2·P5.

Background concepts

Why amorphous drugs crystallize, and why polymers slow it. The discussion of the paper frames crystallization from the amorphous state in two parts C17·C18. Thermodynamically, a glass or supercooled liquid crystallizes because the crystal has the lower free energy; the difference is the driving force for crystallization, written ΔG_c and also called the "degree of undercooling". For a pure amorphous drug it can be calculated from the enthalpy and entropy differences between amorphous and crystalline forms, or approximated with the Hoffman equation, which the paper names but does not print C17·E1. When a polymer is mixed in at the molecular level, ΔG_c of the drug is no longer easy to quantify; one way to approach it is through the additional free energy change on mixing, ΔG_mix, estimated from melting point depression (Marsac et al. 2006) C17·E2. When water is also present, the authors call a reasonable estimate "extremely difficult", because drug, polymer and water amounts and all three pairwise interactions change the result C17. Kinetically, a polymer lowers the molecular mobility of the drug, seen as a higher glass transition temperature (T_g) of the mixture, and lower mobility "can be expected" to slow crystallization; specific drug–polymer interactions and molecular motif recognition also play a part C18.

Nucleation and growth. Crystallization is described as two consecutive steps: nucleation, in which nuclei that can act as centres of crystallization form, and crystal growth, in which these nuclei grow. Both depend on the driving force and on mobility; nucleation also depends on the interfacial energy of the nucleus C19. A bulk measurement such as PXRD sees neither step directly. It sees crystals only once they make up a detectable fraction of the sample, so a delay before crystals appear, an induction period or lag time, is "commonly associated" with the time that nucleation and early growth need to reach that level C26.

Water as plasticizer and as a demixing agent. Sorbed water lowers T_g and so raises molecular mobility. In some drug–polymer pairs it can also do something more drastic: make the drug and polymer immiscible, so that a homogeneous dispersion separates into a drug-rich and a polymer-rich amorphous phase. The paper calls this moisture-induced amorphous–amorphous phase separation, or moisture-induced drug–polymer immiscibility, and cites it for felodipine–PVP at around 75% RH and above R24. If it happens, the drug in the drug-rich phase is no longer protected by the polymer, and the authors expect it to crystallize much as the pure amorphous drug would C21.

Quantifying crystallinity by PXRD with a PLS model. Crystalline felodipine gives sharp Bragg peaks; the amorphous drug and the polymers give a broad halo. To turn a diffractogram into a percentage, the authors made calibration standards with known ratios of crystalline to amorphous felodipine in each drug:polymer ratio, and fitted a partial least squares (PLS) regression, a multivariate model that relates the whole diffractogram to the known crystallinity P5. Two statistics describe such a model: R² and Q² measure how well it fits and predicts the calibration set, and the root-mean-squared error of prediction (RMSEP) gives the typical error of a single prediction, here in percentage points of crystallinity R4. The RMSEP matters for reading the results: values within a few RMSEP of zero cannot be told apart from zero, and some plotted values are slightly negative for that reason R4·X4.

Methods in detail

The study is a long-term storage experiment with one read-out. Its design varies three factors: polymer type (PVP or HPMCAS), polymer level (10, 25 or 50 wt% of the dispersion) and storage RH (six levels between 0 and 93%), and follows each combination over time A1·A2·A3·A4. The flow below summarizes the work.

Making the dispersions

Felodipine (a gift from AstraZeneca) was combined with PVP K29-32 (Sigma-Aldrich) or HPMCAS AQOAT AS-MF (Shin-Etsu) at drug:polymer ratios of 90:10, 75:25 and 50:50 by dry weight; the polymers had been dried over phosphorus pentoxide for at least a week M1·M2·M3·P1. Each mixture was dissolved in a 1:1 (by weight) mixture of dichloromethane and ethanol, and the solvent was removed in a rotary evaporator; three replicate samples of each ratio were made independently P1. After at least 72 h under vacuum, the material was cryo-milled in liquid nitrogen for a total of 4 min and sieved, keeping particles below 150 µm, and PXRD showed no detectable crystalline material P1·R1. In the text and figures the dispersions are named by their polymer content: "25% PVP" is a 75:25 felodipine:PVP dispersion.

Two controls test whether the preparation affects the read-out. Dry dispersions and dispersions equilibrated at 93% RH gave identical diffractograms, "indicating" that sorbed water does not change the diffraction pattern R2·C5. In a milling test, 25% PVP and 25% HPMCAS samples with half the drug added as crystals were milled for 2, 4 and 6 min: their infrared spectra overlapped and their diffractograms were almost identical, so milling did not turn the physical mixture into a dispersion; the crystalline peaks became slightly lower, which the authors read as defects that cryo-milling "may have" introduced, and the calculated crystallinity changed by less than 3% between 2 and 6 min, well within the error of the method R3·C6.

The storage design

Samples were stored at room temperature (22–25 °C) in glass desiccators over phosphorus pentoxide (0% RH) or saturated solutions of magnesium chloride (33%), magnesium nitrate (52%), sodium nitrite (66%), sodium chloride (75%), potassium chloride (86%) or potassium nitrate (93%) P2. The grid is not complete: the 10% PVP dispersions were stored at 66% RH instead of 52% RH, and only they were stored at 66% RH P2·X10. The abstract gives the magnesium chloride condition as 32% RH; the methods, the figures and Table I give 33% X1. The 10% PVP samples were followed for 724 days, the 25% and 50% PVP samples for 493–494 days and the HPMCAS samples for 487–491 days P2·R9.

Turning diffractograms into crystallinity

The diffractometer (Shimadzu XRD-6000, Cu-Kα radiation, Bragg–Brentano geometry) scanned 5–35° 2θ at 8°/min with a 0.04° step P3. Each measuring day began with a silicon standard: its [111] peak had to lie between 28.423 and 28.463° 2θ, and its intensity was used to normalize the data collected that day; each stored sample was read three times in a row P3. For calibration, drug and polymer were co-dissolved and evaporated, checked to be X-ray amorphous, and then mixed with crystalline felodipine, cryo-milled and sieved, giving five crystalline-to-amorphous ratios for each drug:polymer ratio P5. One PLS model per drug:polymer ratio was built in SIMCA-P+ v.11; all models reached R² of 0.975 or more and Q² of 0.973 or more P5·R4. Their prediction errors differ between systems and are listed in Table A R4.

Table A. Root-mean-squared error of prediction of the five PLS models (from the methods section, p.2600) R4.

No model was built for 10% HPMCAS: these samples crystallized to a mixture of two felodipine crystal forms, one of them identified as Form II, and were therefore not quantified R13.

Statistics and the rate analysis

Differences were tested with an F-test at 95% confidence, usually at the final common time point of the samples compared P6·E3. Where the RMSEP of the model was larger than the spread between the three replicates, the RMSEP was used as the denominator of the F-test, so that the largest source of error set the test; the same rule sets the error bars in the time-course figures (SD of the triplicates, or the RMSEP where it is larger) P6. The paper reports the outcomes of these tests in words but prints no F values or p-values R10·R18·R19·X3.

For the PVP dispersions at 93% RH, the authors also calculated the incremental mass of felodipine that crystallized per unit time, dM/dt in g/day, and plotted it against log time (Figure 8) P7·R21. The paper does not say how dM/dt was calculated, on what sample mass it is based or over which time intervals X6.

Experimental analysis

The central data of the paper are the time courses in Figures 2–4 (PVP) and 6–7 (HPMCAS) and the final values collected in Table I, reproduced here as Table B R6·R7·R8·R9·R15·R16. Figures 1 and 5 show representative diffractograms behind the numbers R5·R13·R14.

Table B. Percentage of felodipine crystallized at the last time point, mean (standard deviation), n = 3, with the storage time in days (Table I, p.2602; the column heads there read "HPMAS") R9·X12.

PVP dispersions: humidity switches them from stable to fast-crystallizing

Figures 2–4 show the same pattern at all three PVP levels: the higher the RH, the steeper the rise of % crystallized with time R6·R7·R8·C7. At 0, 33 and 52% RH the 25% and 50% PVP dispersions stayed close to zero for about 493 days (1.1–5.0%), values of the same size as the 4.5–5.3 RMSEP of their models, and the two levels did not differ significantly R9·R10·R4. At 75% RH and above the picture changes completely: after about 494 days the 25% PVP dispersion had crystallized 68.6, 77.2 and 82.1% at 75, 86 and 93% RH, and the 50% PVP dispersion 50.8, 78.4 and 75.1% R9. Every dispersion stored at 66% RH or above crystallized significantly more than the same dispersion at lower RH R10. Figure 1 shows what this means in the raw data: after about a month at 75% RH the 10% PVP sample (20 days) shows sharp felodipine peaks, tallest near 10.3 and 23.3° 2θ, the 25% PVP sample (29 days) shows the same peaks weaker, and the 50% PVP sample (29 days) is still mostly an amorphous halo R5.

The 10% PVP dispersion behaves somewhat differently. Even dry it crystallized more than the others: 14.2% at 0% RH and 23.1% at 33% RH after 724 days R9. At 33% RH the authors found its long-time extent significantly higher than for 25% and 50% PVP; at 0% RH they report no significant difference between the PVP levels in the results section, whereas the discussion uses 0% RH as its example that more PVP gives less crystallization (less than 15% for 10% PVP after about 730 days, none within error for 25% and 50% PVP after 490 days) R10·C8·X2. At 66% RH the 10% PVP dispersion climbed slowly, to about 14% at roughly one to two months and about 30% at about 180 days, and then reached about 62% by about 260 days (read from Figure 2, approximate) R6·X14.

How fast the high-RH crystallization happens depends on the RH. Table C collects readings from Figures 2–4.

Table C. Percentage of felodipine crystallized over time at 93% and 75% RH, read from Figures 2–4 (approximate; sampling days are not printed and are read from the time axis); last points from Table I R6·R7·R8·R9·X13.

At 93% RH most of the crystallization happened within days and was over within about two weeks; at 75% RH the same dispersions needed months R7·R8·X5. Figure 2 shows two 93% RH points of 10% PVP at the same early time position, about 19% and about 44%, which the paper does not explain R6·X13.

The plateau at high humidity

At 86 and 93% RH the PVP dispersions did not crystallize completely. The text says that crystallization at 93% RH was rapid during the first 20 days and that no additional crystallization was seen afterwards, even though some drug remained amorphous after 724 days (10% PVP) and 494 days (25% and 50% PVP) R11. The authors put the plateaus at approximately 70% (10% PVP), 80% (25% PVP) and 75% (50% PVP) and found no significant difference between them R11·R10. In Table B the corresponding values at 86 and 93% RH are 73.5 and 73.8%, 77.2 and 82.1%, and 78.4 and 75.1%; the 50% PVP value at 93% RH carries the largest standard deviation of the table, 16.2 R9. In the abstract and conclusions this becomes crystallization that "seemed to approach a kinetic plateau" C15.

HPMCAS dispersions: little crystallization at any humidity

The HPMCAS dispersions crystallized far less. After 487–491 days the 25% HPMCAS dispersion had crystallized 8.6–9.7% at 0–52% RH and 13.9, 16.4 and 27.7% at 75, 86 and 93% RH; the 50% HPMCAS dispersion stayed between 9.2 and 16.6% at every RH R9. The authors describe the overall extent as "quite low", with a maximum of about 30% for 25% HPMCAS at 93% RH after about 500 days and less than 10% after 500 days at 0 and 33% RH C25. Table B gives 10.7% for 50% HPMCAS at 33% RH, slightly above that "less than 10%"; the results section says "around 10%" X11. Unlike the PVP dispersions, the HPMCAS dispersions at 86 and 93% RH were still crystallizing after about 490 days: 25% HPMCAS at 93% RH went from about 11% at about 15 days to about 20% at about 134 days, about 23% at about 198 days and 27.7% at 491 days (read from Figure 6, approximate) R20·R15·C14.

The statistical statements about HPMCAS need care. The paper states that for a given polymer level RH had no significant effect on the extent of crystallization, and that 25% and 50% HPMCAS did not differ at a given RH after long times R18·C10. For 50% HPMCAS this matches the numbers, 9.2–16.6% across all RHs, and Figure 7 shows it clearly R9·R16. For 25% HPMCAS the values rise from 8.6% at 0% RH to 27.7% at 93% RH, and at 93% RH the two levels read 27.7% and 16.6%; with RMSEPs of 3.7 and 5.3 and no printed test statistics, these "no significant difference" statements cannot be checked from the paper R9·R4·X3.

Figure 5 explains why 10% HPMCAS has no time course. After 491 days at 93% RH its diffractogram shows peaks, marked by asterisks near 11.6 and 13.2° 2θ, that belong to Form II of felodipine as described by Rollinger and Burger, next to peaks of a second crystalline phase; the 25% HPMCAS sample shows sharp peaks without the Form II markers, and the 50% HPMCAS sample is mostly amorphous R13·R14.

The two polymers compared

Put side by side, the two polymers are indistinguishable at low RH and far apart at high RH C12·C13. At 0, 33 and 52% RH the authors found no significant difference in stabilizing ability between PVP and HPMCAS at the same level, in line with earlier nucleation-rate measurements that showed little difference between these polymers at 0 or 33% RH R19·R22·C12. The HPMCAS values at low RH (8.6–10.7%) are in fact a few points higher than those of 25% and 50% PVP (1.1–5.0%), but these differences are within about two RMSEP R9·R4·X4. At 75, 86 and 93% RH HPMCAS was "clearly a better stabilizer": 13.9, 16.4 and 27.7% against 68.6, 77.2 and 82.1% at 25% polymer, and 10.6, 9.2 and 16.6% against 50.8, 78.4 and 75.1% at 50% polymer C13·R9. The authors also stress the difference in time course: the PVP dispersions at high RH reached their plateau quickly, the HPMCAS dispersions were still rising after about 490 days C14·R11·R20.

At 93% RH the amount of PVP no longer matters

Figure 8 plots the incremental crystallized mass of felodipine per day for the three PVP dispersions at 93% RH on a log time axis R21. The first points are about 6.4 g/day for 10% PVP at about 2.5 days (error bar about 3.2–9.6) and about 5.8 and 5.5 g/day for 25% and 50% PVP at about 4 days; by about 14 days the values have fallen to about 0.9, 0.4 and 0.2 g/day, and from about 34 days on they are about zero (read from Figure 8, approximate) R21. From the overlap of the first values the authors conclude that at 93% RH the absolute crystallization rates of the three dispersions are "independent of the amount of polymer added", in contrast to 33% RH, where the 50% PVP dispersion crystallized significantly less than the 10% PVP dispersion C20·R10. Because the basis of dM/dt is not given, the values cannot be recomputed from Figures 2–4 X6.

With 50% HPMCAS, crystals appear only after a delay

The 50% HPMCAS dispersions showed negligible crystalline drug after 100 days at 52% RH, 75 days at 75% RH, 120 days at 86% RH and 35 days at 93% RH, whereas the 25% HPMCAS dispersions began to crystallize after a short storage time (about 11% at about 15 days at 93% RH, read from Figure 6) R17·R15. The authors describe this as an induction period ("appeared to exhibit"; in the discussion, "seemed to display a lag time") C11. Once the 50% HPMCAS curve left zero, its slope was comparable to that of 25% HPMCAS, "especially obvious" at 93% RH: from about 76 to about 198 days the 50% sample rose from about 3 to about 11%, the 25% sample from about 16 to about 23% (read from Figures 6 and 7, approximate) C27·R15·R16.

Discussion

Why PVP fails at high humidity: the authors' explanation

The authors explain the behaviour of the PVP dispersions with a chain of arguments that rests partly on this paper's data and partly on earlier work C21·C22. Sorbed water lowers T_g and raises molecular mobility, and PVP dispersions take up more water than HPMCAS dispersions: the cited thesis gives 0.36 g water per 0.5 g PVP for a 50% PVP dispersion at 93% RH C22·R25. More importantly, felodipine–PVP dispersions exposed to around 75% RH and above have been shown to form drug-rich and polymer-rich amorphous phases R24. If that happens, the stabilization of the amorphous drug by the polymer "would be expected" to fall sharply, or vanish if the separation is complete, and the overall crystallization rate to become independent of how much polymer the dispersion contains, because the composition of the separated phases is set by the extent of immiscibility, not by the starting ratio C21.

The authors then point to three observations as consistent with this picture. At 93% RH the first-interval crystallization rate is the same for 10, 25 and 50% PVP, "exactly what was observed in this study" C20·R21. The crystallization of the phase-separated dispersions up to the plateau was comparable to that of pure amorphous felodipine at 93% RH, which crystallized completely in less than 12 days C23·R12. And in the cited thesis, the linear growth rates of felodipine crystals in films with 2.5, 5 and 10% PVP were comparable to each other at 93% RH C24·R23. Their conclusion keeps a hedge throughout: the observations "support the supposition" that rapid crystallization at high RH "can be explained in part" by moisture-induced phase separation, "in addition to" higher mobility from sorbed water; the abstract says "It is hypothesized", the conclusions say the trend "can be attributed" to the greater moisture sorption of PVP dispersions C22. The same reasoning explains why the two polymers part ways only from 75% RH: below that, both keep the drug mixed and immobile enough; above it, the more hygroscopic PVP dispersions separate C12·C13·C22.

Note by the AI writer (not discussed in the paper). None of the links in this chain was measured on the stored samples: there are no water-sorption, T_g, infrared or imaging data for these dispersions X7. The independence of the 93% RH rate from PVP level is the only test of the mechanism in this paper, and it rests on Figure 8, whose calculation is not given X6. The comparison with pure amorphous felodipine rests on a statement without data X8.

The plateau: a solubility limit or a kinetic trap?

Why does crystallization stop at about 70–80% in the PVP dispersions at high RH R11? The "abrupt" way the plateaus were reached "seemed to suggest" a thermodynamic origin: felodipine might have reached its solubility limit in the remaining amorphous felodipine–PVP–water phase C30. The authors test this with a mass balance stated in words (p.2605) E4. Take a 50% PVP dispersion at 93% RH and suppose that 20% of the original felodipine stays uncrystallized. Per gram of dispersion, the remaining amorphous phase then holds about 0.1 g felodipine in 0.86 g of PVP–water mixture (0.5 g PVP plus 0.36 g water, the water content taken from the thesis) E4·R25. In dry PVP, felodipine dissolves to about 0.25 g per g of PVP (Marsac et al. 2006), and large amounts of water lowered felodipine solubility by at least an order of magnitude in the methylpyrrolidone system (Marsac et al. 2008) R26. "Assuming similar effects in PVP", the solubility in the PVP–water mixture would be no more than 0.025 g per g, well below the felodipine that remains at the plateau C31·E4. The authors therefore conclude that the plateau "possibly represents" drug in a polymer-rich phase that has not crystallized for kinetic reasons C32.

The arithmetic holds: 0.1 g in 0.86 g is about 0.116 g per g, roughly five times the 0.025 g per g bound, and with the measured 75.1% crystallized for 50% PVP at 93% RH (24.9% left) instead of the round 20%, the ratio is about 0.145 g per g, still above the bound E4·R9. The weak points are the assumptions, not the arithmetic: the tenfold drop in solubility is transferred from another system, and the composition of the residual amorphous phase was not measured C31·X9.

Delayed nucleation with 50% HPMCAS

For the HPMCAS dispersions the authors use the induction period to separate nucleation from growth, in an indirect way C28·C29. Lag times are "commonly associated" with the time nuclei need to grow to a detectable size C26. Because the 50% HPMCAS curves, once they leave zero, rise at about the same slope as the 25% HPMCAS curves, and "assuming" that growth below and above the detection limit is governed by the same factors, they conclude that nucleation was delayed in the 50% HPMCAS dispersions, and "It can thus be speculated" that the growth rates are similar at the two polymer levels C27·C28·C29. This is an inference from bulk curves; nucleation and growth rates were not measured here C29.

What it suggests for formulation

The authors draw three practical points. When moisture is limited, a felodipine–PVP dispersion has "excellent physical stability over pharmaceutically relevant time periods", and both polymers are effective inhibitors when the dispersion is protected from moisture C9·C12. The physical stability of solid dispersions as a function of RH is "highly dependent on the polymer used to form the solid dispersion" C16. And the results "highlight the importance of considering the effects of storage conditions on the miscibility" of a dispersion when interpreting crystallization data C33.

Note by the AI writer (not discussed in the paper). Three further readings follow from the data, without being stated by the authors. First, for the hygroscopic polymer, more polymer bought no protection at high RH: at 86–93% RH the 50% PVP dispersion ended at the same level as the 10% PVP dispersion (Table B), so for such a pair humidity control and packaging carry the stability, not the polymer content R9·R10. Second, storage at 0–52% RH did not discriminate between the polymers; only storage at 75% RH or above did, so a polymer screen run under dry conditions would have missed the difference R19·R9. Third, a plateau below full crystallinity at high RH is not a safe residual amorphous fraction; the authors themselves argue it is kinetic, and HPMCAS dispersions were still crystallizing at the end of the study C32·R20.

Limitations

Limitations stated by the authors. The paper has no separate limitations section; its reservations are carried by its wording. The phase-separation mechanism is introduced as a hypothesis ("It is hypothesized") and supported only "in part" C22. The plateau explanation is labelled "Further speculation" and "possibly" C32. The growth-rate conclusion for HPMCAS depends on an explicit assumption and is put as "It can thus be speculated" C28·C29. On the method side, the authors use the RMSEP as the error term where it exceeds the scatter of replicates, acknowledging that the model error dominates at low crystallinity P6·R4.

The AI writer's own judgement (1): the mechanism is borrowed, not measured. The paper's data are kinetic only. The water content of the stored samples, their T_g and their phase state were not measured, so moisture-induced phase separation is a cited explanation consistent with the kinetics, not an observation of this study X7. The key observation offered in its support, a PVP-independent first-interval rate at 93% RH, comes from a dM/dt plot whose calculation, mass basis and time intervals are not given; the first points also sit at different times (about 2.5 and about 4 days) for dispersions that contain 90, 75 and 50% drug X6. The pure amorphous felodipine reference, used twice in the argument, has no data in the paper X8.

The AI writer's own judgement (2): the statistical statements. Several conclusions rest on F-tests whose statistics are not printed. Some of them sit uneasily with Table I: the statement that RH had no significant effect on HPMCAS dispersions does not fit 25% HPMCAS, which rose from 8.6% at 0% RH to 27.7% at 93% RH X3. The results and the discussion disagree on whether PVP level mattered at 0% RH X2. At low RH the two polymers differ by less than about two RMSEP, so "similar" there means "not distinguishable by this method" X4. The 10% PVP samples, stored at 66% instead of 52% RH and read last at 724 days, are not directly comparable with the other entries of Table I X10.

The AI writer's own judgement (3): time course and wording. The statement that PVP dispersions at high RH reached a steady plateau "within a few days" holds at 93% RH; at 75% RH, and at 86% RH for 10% PVP, the rise took months X5. The text groups 66% RH with the fast conditions, although 10% PVP at 66% RH crystallized only about 30% in the first half year X14. Smaller inconsistencies do not affect the conclusions: the abstract gives 32% instead of 33% RH for the magnesium chloride condition, the text says "less than 10%" where Table I shows 10.7%, and Table I labels HPMCAS as "HPMAS" X1·X11·X12. Finally, the study covers one drug, room temperature and two polymers, so the results describe this system rather than a general rule.

Conclusion

For felodipine dispersions stored at room temperature for up to about two years, the choice of polymer decided how much humidity mattered. Dry or at moderate RH, PVP and HPMCAS both kept felodipine largely amorphous for about 490 days; at 75% RH and above the PVP dispersions crystallized to 51–82% while the HPMCAS dispersions stayed at 9–28% R9·C12·C13. At 93% RH the PVP dispersions crystallized at the same initial rate whatever their PVP content and stopped at plateaus of about 70–80% R21·R11·C20. The authors hypothesize that water separates felodipine from PVP into drug-rich and polymer-rich phases and that the plateau is drug kinetically held in the polymer-rich phase; both ideas come from earlier work or from estimates, not from measurements on these samples C22·C32·X7·X9. For 50% HPMCAS, a delay of 35–120 days before crystals appeared is read as delayed nucleation R17·C29.