Does a low drug loading make an amorphous dispersion more supersaturating? What dissolving felodipine and indomethacin dispersions really produce

Background and motivation

Amorphous solid dispersions (ASDs), in which a poorly soluble drug is dispersed in a polymer, are widely used to raise the effective solubility and bioavailability of active pharmaceutical ingredients (APIs) C1. In 2011 the processes that govern how they dissolve were still poorly understood. Simonelli et al. had pointed out three complications early on: the drug–polymer ratio, the relative dissolution rates of drug and polymer, and crystallization of the drug while it dissolves (Simonelli 1969, ref 4) C1.

Most dissolution work on dispersions had compared dissolution rates between systems. The authors focus on a different quantity: the degree of supersaturation that a dispersion generates and maintains when it dissolves under nonsink conditions, that is, with more drug added than the medium can hold at equilibrium C2. Their central question uses the term "solubility advantage" coined by Hancock and Parks for the transient concentration gain an amorphous solid offers over its crystal (Hancock 2000, ref 1): how does the solubility advantage of a dispersion compare with that of the pure amorphous drug C2?

The question is not trivial. A glass is not an equilibrium state, so strictly it has no defined solubility, and amorphous drugs often crystallize quickly in water C3. The group's own earlier work (Alonzo 2010, ref 10) had shown that the observed concentration–time profile of an amorphous solid depends on two kinds of stability, of the amorphous solid in contact with the medium and of the supersaturated solution, and that when both are stabilized the peak concentration can approach the theoretically estimated amorphous solubility C3. The authors argue that every measured supersaturation should therefore be compared with that estimate, which serves as a "measuring stick" for performance and a warning of instability C4.

The estimate itself rests on one assumption. All methods take the free energy difference between the amorphous and crystalline forms as the driver of the advantage, whether calculated from the enthalpy of fusion and melting temperature (Hoffman 1958, ref 13), from configurational heat capacity (Marsac 2006, ref 14; Matteucci 2008, ref 15), or with corrections for water uptake and ionization (Murdande 2009, ref 11) C5. They all assume that the drug interacts with the solvent in the same way whether it leaves an amorphous or a crystalline solid. This is reasonable for a pure drug. For a dispersion, in which the drug interacts with a polymer, the authors write that "it would be reasonable to speculate" that the estimates may not apply C6. Dissolution of dispersions was also known to depend on polymer type, drug-to-polymer composition, miscibility and physical stability, and dispersions had often given higher concentrations than crystalline or pure amorphous drug (refs 3, 4, 16–20) C7.

The paper therefore sets out to do two things: measure the supersaturation generated on dissolution as a function of drug–polymer composition, and relate the physical behaviour of the dissolving dispersion to the polymer loading C2·A1. The model drugs are felodipine, dissolved at pH 6.8, and indomethacin, dissolved at pH 2.0 where it is not ionized; the polymers are hydroxypropylmethylcellulose (HPMC, Pharmacoat 606) and polyvinylpyrrolidone (PVP, K29/32) M1·M2·M3·M4. Every pair was made at 90:10 and 50:50 polymer:drug by weight, that is, at 10% and 50% drug loading P1.

Background concepts

Supersaturation and the amorphous solubility advantage. The supersaturation ratio is the observed solution concentration divided by the equilibrium solubility of the crystal. The highest supersaturation an amorphous solid can give is set by how much higher its free energy is than that of the crystal. The paper estimates this with the Hancock–Parks relation (Eq. 2) E2:

In plain terms: the amorphous form is more soluble than the crystal by the exponential of its excess free energy $\Delta G$ divided by $RT$, where $R$ is the gas constant and $T$ the absolute temperature; ${\sigma }^{\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{h}}$ and ${\sigma }^{\mathrm{c}\mathrm{r}\mathrm{y}\mathrm{s}\mathrm{t}\mathrm{a}\mathrm{l}}$ are the two solubilities E2. The excess free energy is approximated with the Hoffman equation (Eq. 3) E3:

In plain terms: the further the temperature lies below the melting point $T_m$ ($\Delta T=T_m-T$), and the larger the enthalpy of fusion $\Delta H_f$, the larger the free energy stored in the amorphous form E3. The enthalpies of fusion were measured by DSC, but the paper does not print them or the melting temperatures P8·M1. The results it does print are these: for felodipine at 37 °C an amorphous solubility of about 9.1 µg/mL, nine times the crystalline solubility of about 1 µg/mL (taken from Konno et al., ref 2), and 5–6 µg/mL at 25 °C; for indomethacin at pH 2.0 an amorphous solubility of 21 µg/mL against a measured crystalline solubility of about 1.2 µg/mL M1·M2. (Note by the AI writer, not in the paper: a ratio of 9 at 37 °C corresponds to $\Delta G$ of about 5.7 kJ/mol.) E2

Apparent concentration from an in-situ UV probe. A fiber-optic probe dipped in the dissolution vessel records a UV spectrum every minute. To reduce the effect of light scattered by particles, the analysis uses second-derivative spectra, calibrated with drug dissolved in methanol P2. What the probe reports is an "apparent" concentration: if small particles of drug absorb light as dissolved drug does, the probe counts them too. A companion study from the same group had shown that crystalline felodipine nanosuspensions absorb UV light in proportion to the amount of solid, that this makes the solution concentration look higher than it is, and that the second derivative does not remove this absorption; as particles get smaller, absorption grows relative to scattering (van Eerdenbrugh 2011, ref 35; Bohren and Huffman, ref 34) R25.

Measuring only dissolved drug: flux across a membrane. The paper's main check is a dialysis cell. Only dissolved molecules cross the membrane, and their flux follows Eq. 1 E1·C9:

In plain terms: the flux $J_i$ of component $i$ through the membrane equals a constant (the diffusion coefficient $D_i$ times the sorption coefficient $K_i$, divided by the membrane thickness $l$) times the concentration difference between outside ($c_{io}$) and inside ($c_{il}$) the membrane E1. With the receptor kept at low concentration, the rate at which drug builds up inside is proportional to the dissolved concentration in the donor, so any true difference in supersaturation between two dispersions shows up as a difference in this rate C9. When they report the result, the authors accordingly speak of "solution concentrations (or more correctly solution activities)" C22.

Measuring only dissolved drug: solution NMR. Proton NMR of a dissolution sample, with the water signal suppressed, gives sharp signals only from highly mobile, dissolved molecules; drug inside particles is not detected P7·C27. Comparing the integrated aromatic felodipine signal of two samples compares their dissolved concentrations R15.

Dynamic light scattering (DLS). DLS measures how fast scattered light fluctuates as particles diffuse, and from this an intensity-weighted average (z-average) diameter P5. It detects submicron particles; very large or settling particles make the size estimate unreliable, as the paper notes C14. With fixed instrument settings, the count rate of scattered photons compares the amount of scattering material between samples R9.

Polymer-controlled and drug-controlled dissolution. Older work on PVP–sulfathiazole coprecipitates distinguished two regimes (Simonelli 1969, ref 4; Corrigan 1985, ref 37) C37. At high drug loading, the water-soluble polymer leaves the surface faster than the drug, a drug-rich layer, "possibly in the amorphous state", remains at the surface and governs dissolution; the initial dissolution rate was faster than that of the crystal but did not depend on the polymer content up to 50 wt% polymer C37·R26. At high polymer loading there is not enough drug to form such a layer, the drug dissolves congruently with the polymer, and its rate is a fraction of the polymer's (Eq. 4) C37·R26·E4:

In plain terms: the dissolution rate of the drug $G_D$ equals the dissolution rate of the carrier $G_C$ scaled by the ratio of drug to carrier ($A_D/A_C$, the component concentrations) in the solid E4. The paper uses this relation qualitatively; it computes no dissolution rates with it E4.

Methods in detail

The study combines one apparent measurement with several that are blind to particles. The flow below shows how they fit together.

Contribution 1: the same dispersions at two drug loadings, dissolved under nonsink conditions

Drug and polymer were dissolved together in 50:50 methanol–dichloromethane, the solvent was removed by rotary evaporation, and the solid was cryomilled (30 s at four impacts per second, four cycles) for a reasonably uniform particle size and stored dry in a refrigerator P1. The particle size after milling is not reported P1. All drug–polymer pairs were taken as miscible on the basis of earlier work (refs 21–23) C8.

For dissolution, about 5–20 mg of dispersion was added to 18 mL of 50 mM phosphate buffer (pH 6.8 for felodipine, pH 2.0 for indomethacin), stirred at 300 rpm with a cross-shaped stir bar at 37 or 25 °C, and followed by UV spectra from 270 to 450 nm every minute P2. All polymer in the medium came from the dispersion itself, so a 90:10 dispersion brings nine times more polymer than drug P2. Runs were in triplicate except a series at 37 °C in which increasing amounts of the 90:10 HPMC–felodipine dispersion were added (Figure 3) P2·R2. The nominal concentration, the concentration if everything added had dissolved, was 30 µg/mL for the 90:10 dispersion and 150 µg/mL for the 50:50 dispersion in the main 37 °C comparison (Figure 2), the same mass of solid being added R1. The nominal concentrations of the 25 °C runs (Figure 5) and of the indomethacin runs (Figure 12) are not printed X17·X12.

Contribution 2: watching the dispersion particles and the precipitate

Single dispersion particles on a microscope slide were wetted with buffer and photographed under crossed polars with a 530 nm tint plate at room temperature (Figure 4) P6. DLS (Malvern Zetasizer ZS, 173° backscatter) followed the average particle size over time in dissolving dispersions and in solutions made supersaturated by the methanol shift: a concentrated methanol solution of felodipine pipetted into buffer, with or without HPMC already dissolved in it P5·P3. For the main comparison at 25 °C (Figure 7), the 90:10 HPMC–felodipine dispersion was added to 20 mL buffer at nominal 30 or 50 µg/mL felodipine, which brings 270 or 450 µg/mL HPMC; the methanol-shift solutions at 30 and 50 µg/mL contained 250 and 500 µg/mL predissolved HPMC P5·R8. In one pair of runs the attenuation (6) and measurement position (1.25 mm) were fixed so that count rates could be compared P5·R9.

Contribution 3: two measurements that see only dissolved drug

This is the paper's decisive experiment. In a 600 mL jacketed beaker, 300 mL of pH 6.8 buffer at 37 °C served as the donor, monitored by one UV probe (Figure 1) P4. The receptor was a dialysis bag holding 10 mL of the same buffer with 500 µg/mL HPMC, to stop the drug crystallizing once inside, and a second UV probe P4·C10. The text gives the membrane a molecular weight cutoff of 6–8 kDa, while Figure 1 labels it 10 kDa M11·X1. Donors were 90:10 or 50:50 HPMC–felodipine dispersion, pure amorphous felodipine or crystalline felodipine (the last two with 250 µg/mL HPMC predissolved), and methanol-shift solutions at nominal 5 and 10 µg/mL with 250 µg/mL HPMC P4. For solids, the receptor was introduced only after the donor UV reading had reached a plateau, after about 40 min for the amorphous samples and about 60 min for the crystal; for the methanol-shift solutions it was present from the start P4. The slope of the receptor concentration over 10–40 min was fitted by linear regression and divided by the slope of the crystalline donor (Table 1) R14. How the pure amorphous felodipine was prepared, and how much solid went into each donor, is not stated X17·X3.

For NMR, 530 µL of suspension was withdrawn 50 min into dissolution of each HPMC–felodipine dispersion at 37 °C, mixed with 20 µL D₂O and measured on a Bruker AVANCE DRX500 with a cryoprobe at 37 °C; successive spectra over about 40–60 min (256 scans each, WATERGATE water suppression) were integrated over the aromatic felodipine peaks P7. The number of samples per dispersion is not stated P7.

Contribution 4: the UV spectrum as a fingerprint of particles

The spectrum recorded by the probe after 60 min of dissolving the 90:10 HPMC–felodipine dispersion was compared with a 10 µg/mL solution of felodipine in methanol and with an aqueous crystalline felodipine nanosuspension of 25 µg/mL solid content, whose spectrum comes from Alonzo's thesis (ref 36) (Figure 13) P10·R19. Because absorbances add, the two reference spectra were summed and scaled down by 10% so that the peak heights matched, and the sum was compared with the dissolution spectrum (Figure 14) P10·R21·C35.

Experimental analysis

By UV, 90:10 dispersions read far above 50:50 dispersions

At 37 °C the 90:10 HPMC–felodipine dispersion rose within 10 min to about 23 µg/mL and reached about 27 µg/mL at 60 min; the authors give the peak as 25–30 µg/mL, approximately 100% of the 30 µg/mL added (Figure 2) R1. The 50:50 dispersion, with the same mass and five times more felodipine (150 µg/mL), rose slowly and levelled at about 8.9 µg/mL, close to the estimated amorphous solubility of about 9.1 µg/mL R1·M1. The 90:10 dispersion also dissolved much faster R1·C12. Adding more of the 90:10 dispersion raised the apparent peak to about 28, 53 and 75 µg/mL for 30, 60 and 90 µg/mL added, each 80–100% of the dose, reached within about 10 min; at 90 µg/mL the reading then fell to about 23 µg/mL by 50 min (Figure 3) R2. These were single runs, and at 30 µg/mL the reading rose faster than in the triplicate run of Figure 2 at the same dose X15.

The same pattern held at 25 °C, for PVP, and for indomethacin (Table A) R5·R16·R18. At 25 °C the 90:10 PVP–felodipine dispersion gave the highest early reading, about 36.5 µg/mL at 3 min, but it fell to about 9.2 µg/mL within an hour, while the 90:10 HPMC dispersion went only from about 26.8 to about 20.7 µg/mL; both 50:50 dispersions stayed at about 6–8 µg/mL, "just above" the predicted 5–6 µg/mL (Figure 5) R5·C12. From these data the authors concluded that the polymer-to-drug ratio was "an extremely important factor" in dissolution behaviour C12.

Table A. Highest apparent concentration measured by the in-situ UV probe, µg/mL (read from Figures 2, 5 and 12, approximate; HPMC–indomethacin values from the text) R1·R5·R16·R18·M1·M2.

Taken at face value, these readings say that a 90:10 dispersion produces 2.5 to 6 times the supersaturation of a 50:50 dispersion and far more than pure amorphous drug can, whichever polymer or drug is used. The authors put it more cautiously: it "appears at first glance" C27. For indomethacin they write that elevated supersaturations "appear to be generated" at 90% polymer, with the 50:50 PVP–indomethacin peak (about 22 µg/mL) near the predicted 21 µg/mL and the 90:10 peak "more than twice" that C25·R16. The fluctuating 80–120 µg/mL readings of 90:10 HPMC–indomethacin were near the upper limit of the probe's linear range, which "most likely" explains their scatter R18·C26.

Only the 90:10 dispersions fall apart within minutes and fill the medium with submicron particles

Under crossed polars the difference is visible at once (Figure 4). A 90:10 HPMC–felodipine particle had disappeared within 5 min of wetting, whereas a 50:50 particle kept its shape and integrity for at least 30 min R4. PVP–felodipine behaved alike (data not shown): 90% PVP particles disappeared almost instantly, 50% PVP particles stayed intact throughout R4. A few minutes after addition, the media of 90% polymer dispersions turned uniformly cloudy; with 50:50 dispersions individual undissolved particles were visible but there was no uniform cloudiness R3·C11. The authors read the cloudiness as very small scattering particles C13.

DLS confirmed the particles. From a 90:10 HPMC–felodipine dispersion at 25 °C (nominal 25 µg/mL), the average diameter was about 218 nm at 5 min, above 300 nm by 90 min and about 720 nm near 10 h, when the particles began to settle (Figure 6) R6·C14. A 50:50 dispersion at nominal 250 µg/mL gave no correlation function for submicron particles, and the authors took them to be absent R7·C15. For PVP–indomethacin the 90:10 dispersion gave 210 nm at 30 min and 217 nm at 45 min, while the 50:50 dispersion gave no usable correlation function, "presumably due to the absence of submicron particles" R17·C15.

Particles from dissolving dispersions resemble particles precipitated from a methanol shift

Figure 7 is the authors' key piece of evidence on how the particles form. At 25 °C, particles from the 90:10 HPMC–felodipine dispersion at 30 and 50 µg/mL and from methanol-shift solutions at the same nominal concentrations with HPMC started at about 183–203 nm at 5 min and reached about 280–307 nm at 90 min (Table B) R8. The authors describe both the initial size and the growth as "very similar" C16. With the count rate fixed, the 50 µg/mL dispersion scattered 148,000 ± 8100 kcps against 71,000 ± 7500 kcps at 30 µg/mL, which they read "as expected" as roughly twice the amount of scattering particles R9·C17. Note by the AI writer: the nominal felodipine ratio of the two samples is 1.67, not 2, and scattered intensity also depends on size, so the factor of two is not derived X13.

Table B. Average particle diameter by DLS, nm (read from Figures 7, 8 and 9, approximate; the no-polymer values at 25 °C are from the text) R8·R10·R11.

Polymer type mattered for growth. Without any polymer, particles from a 30 µg/mL methanol shift were 450 nm at once and 936 nm at 7 min, already settling; with a 90:10 PVP dispersion they grew to about 459 nm by 90 min, against about 289–307 nm with HPMC (Figure 8) R10. The authors read this as rapid growth in the absence of an inhibitor and "considerably faster" growth with PVP than with HPMC, as their earlier work on felodipine had led them to expect (refs 2, 10) C18·R24. The 5-min point of the PVP run, a bimodal distribution with a peak at about 5 µm, was removed as a dust particle or an undissolved dispersion particle R10·C19.

At 37 °C (Figure 9), the temperature of the flux and NMR tests, the picture changes in one respect. Particles from the HPMC dispersions started larger, about 259–274 nm in the first 10 min, than those from methanol-shift solutions with HPMC, about 194–219 nm, and stayed larger to 90 min (about 320–352 against about 294–296 nm) R11·C20. Growth was slower and levelled off sooner than at 25 °C, and PVP-dispersion particles again grew fastest, to about 441 nm by 20 min R11·C20. Note by the AI writer: the "very similar" behaviour of the two particle sources therefore holds at 25 °C but not at 37 °C, and the statement that growth without polymer was "even more rapid" at 37 °C rests on a single point (about 514 nm at about 2 min) X19·X8.

Flux and NMR show the same dissolved level for 90:10, 50:50 and pure amorphous felodipine

The controls first show that the cell can tell concentrations apart: methanol-shift donors at 10 and 5 µg/mL gave receptor slopes 10.0 and 4.8 times the crystalline slope, in the ratio 2.07, as Eq. 1 requires for a twofold concentration difference R14·C21·E1. From the UV donor readings (Figure 10), the 90:10 dispersion should then have diffused about 2.5 times faster than the 50:50 dispersion, as the authors put it R12·C24. Note by the AI writer: the plotted UV values, about 27.0 and 8.9 µg/mL, give a ratio closer to 3 X2. It did not: the slopes were 0.0858 ± 0.0094 and 0.0782 ± 0.015, ratios to crystal of 10.5 and 9.5, and pure amorphous felodipine gave 8.5 (Table C) R14. The pure amorphous donor's UV profile was almost identical to that of the 50:50 dispersion R12·C24.

Table C. Receptor slopes and their ratio with the crystalline slope, 37 °C, mean ± SD, n = 3 except crystalline (n = 1); slope units are not printed (from Table 1) R14.

The estimated amorphous advantage of felodipine at 37 °C is about 9 M1. The authors conclude that dissolution of pure amorphous felodipine and of both dispersions generated solution concentrations, "or more correctly solution activities", "remarkably close" to that estimate, and that "at least over the timeframe that these experiments were conducted" there is no significant difference in the maximum thermodynamic activity of dissolved felodipine among the three amorphous systems C22. They add that the result shows felodipine was not complexed with HPMC in solution, as had been reported for other systems C23.

NMR supports the flux result in the sense that matters: the integrated aromatic felodipine signal was 1370 for the 90:10 and 1014 for the 50:50 sample, a relative difference of approximately 26%, far from the threefold difference of the UV readings R15·C24. Note by the AI writer: 26% is the difference relative to the 90:10 value (35% relative to the 50:50 value), there is one pair of integrals without replicates, no statistical test supports "no significant difference", the three flux ratios fall in the same order as the UV readings, and every ratio is normalised by a single crystalline run although the methods state triplicates X6·X5·X4.

The spectrum of the 90:10 solution carries the signature of crystalline nanoparticles

Felodipine dissolved in methanol shows a peak at 360 nm and a valley at 294 nm, and amorphous felodipine films 1.5 µm to 200 nm thick give almost the same positions; dissolved in buffer, the peak moves slightly, to about 364 nm (Figure 13) R19·R20. The crystalline nanosuspension differs clearly, mainly through a red shift of the valley to 325 nm, with the peak at 373 nm R19. The spectrum of the 90:10 HPMC–felodipine solution at 60 min lies between the two: peak at 367 nm, valley at 312 nm (Table D) R19·C34. The shift appeared within 5 min of the start of dissolution, the timeframe in which DLS found particles, and it never appeared for the 50:50 dispersion, pure amorphous felodipine or the 5 and 10 µg/mL controls R20·C34.

Table D. Peak and valley positions of the felodipine UV spectra (printed values; the sum read from Figure 14, approximate; on the plot the nanosuspension valley sits nearer 330 nm) R19·R20·R21.

The summed reference spectrum puts its peak and valley within a few nanometres of the 90:10 spectrum, although it lies higher above about 385 nm (Figure 14) R21·C35. Together with three other observations, continuous particle growth, single birefringent particles at about 1 µm under polarized light (data not shown) and the known rapid crystallization of amorphous felodipine in water even with polymeric inhibitors, the authors conclude that the nanoparticles from the 90:10 HPMC–felodipine dispersions "were indeed crystalline" C33·R22·R24. They add that particles formed by crystallization of nanometre-scale amorphous drug regions in the polymer matrix "cannot be entirely ruled out but seems unlikely", given the high polymer content and the known molecular miscibility C36.

Discussion

Why the UV probe and the flux test disagree

The authors' "most likely explanation" for the contradiction is the submicron particles generated during dissolution of the 90:10 dispersions C27. Particles of 200–400 nm absorb light much as the solution does; at the initially observed size of about 200 nm a substantial absorption is expected, and taking the second derivative does not remove it C27·R25. They find it "reasonable to hypothesize" that the probe's concentration for the 90:10 dispersions includes drug in particles, whereas flux and NMR see only highly mobile, dissolved molecules C27. In the conclusions this becomes firmer: the particles "prevented the accurate measurement of solution concentration via in situ UV spectroscopy as demonstrated by the diffusion and NMR experiments" C27.

The authors' explanation: two dissolution regimes that end at the same level

The authors connect the observations through the polymer-controlled and drug-controlled regimes described above; the felodipine and indomethacin data "appear to agree" with that model, since dispersions with 90% polymer dissolved much faster than those with 50% (Figures 4, 5 and 12) C38.

The 90:10 branch rests on classical crystallization theory: nucleation rate rises with supersaturation, so a dispersion that releases nearly all its drug at once should create ideal conditions for rapid nucleation C29. "We speculate," the authors write, that this is what happens, and they take the match between dispersion-born and methanol-born particles in Figure 7 as support; their similar behaviour "suggests" a similar mechanism of formation C29·C31. That precipitation also happened in methanol-shift solutions at a supersaturation ratio of 60 with HPMC present shows, in their reading, that nucleation cannot be fully suppressed at that level, while faster growth without HPMC is "consistent with" HPMC slowing felodipine crystallization C30·C32. Note by the AI writer: the crystalline solubility of felodipine at 25 °C, needed to check the ratio of 60, is not printed X9.

The weak link is time. Flux and NMR both need at least 40 min of data, so neither can show what happens in the first minutes; "the generation of a short-lived, elevated solution concentration (most likely on a local scale) ... cannot be discounted" C28. The local supersaturation at the heart of the 90:10 branch was not measured, and neither was the composition or physical state of the drug-rich layer in the 50:50 branch C28·C46.

A thermodynamic side argument

The authors also ask whether a dispersion should give the same solution activity as pure amorphous drug at all. If drug and polymer are miscible, the free energy of the dispersion is lower than that of pure amorphous drug, so by Eq. 2 its solubility advantage "might actually be expected to be lower" (refs 40–43) C41. But Eq. 2 assumes the same solute–solvent interaction for both solid forms, which is "unlikely to be true when considering an ASD", because the different intermolecular bonding in the matrix changes both the entropy and the enthalpy of solution C41. They write the activity of a solute in solution, ignoring heat capacity, as printed in Eqs. 5 and 6 E5·E6:

In plain terms: the activity $a$ of the dissolved drug, written as activity coefficient $\gamma$ times equilibrium solubility $c$, is set by the free energy needed to take the drug out of its crystal, estimated from melting data; the authors call this "the energy of the crystal lattice" and take it to be the origin of the amorphous advantage E5·C42. Note by the AI writer: as printed, Eq. 5 carries an extra factor $T_m$ in the numerator, which gives the right side units of kelvin instead of a dimensionless number; the usual ideal-solubility expression has $\Delta H_f/R$ there. The paper uses Eq. 5 only in words, so no number depends on it X14. The estimate assumes $\gamma$ is the same for amorphous and crystalline drug (regular solution theory), which the authors say would not hold for a miscible dispersion C42.

The enthalpy of solution makes the point concrete (Eqs. 7 and 8) E7·E8:

In plain terms: for pure amorphous drug mixing with water, only the water–drug interaction parameter ${\chi }_{12}$ matters; for a dispersion, the water–polymer (${\chi }_{13}$) and drug–polymer (${\chi }_{23}$) terms enter as well. Here $n$ is the mole fraction (as the paper defines it), $V$ the volume fraction, and the subscripts 1, 2 and 3 are water, drug and polymer E7·E8. With 90% polymer there are far fewer drug–drug contacts, which "in large part" dictate ${\chi }_{12}$, and the authors argue this "would lead to a reduction in the barrier" to the drug entering solution C43. They also caution that the binary ${\chi }_{12}$ is not the ${\chi }_{12}$ of the ternary, which makes Eq. 8 hard to use in practice; Rumondor et al. had found that binary parameters could not produce a reasonable felodipine–PVP ${\chi }_{23}$ (ref 50) C44·R27. No $\chi$ values are given E7·E8.

Their conclusion from this argument is that "there is no reason to expect" the solution activity from a dispersion to equal that from pure amorphous drug, yet the flux test says it does C45·C22. For the 50:50 dispersion the drug-rich surface layer makes this easy to rationalize; for the 90:10 dispersion the activity falls to about the pure amorphous value through precipitation, and "the fact that these values are similar may be purely coincidental (at this point we see no theoretical reason why they should be)" C46·C47·C48.

What it suggests for formulation

Three practical points follow directly from the paper. First, apparent concentrations from in-situ UV fiber-optic probes are not reliable once submicron particles form, and second-derivative processing does not correct them C27. Second, observed supersaturation should be compared with the estimated amorphous solubility as a reference C4. Third, for felodipine HPMC held the supersaturation better than PVP, which "appeared to be a less effective crystallization inhibitor"; the abstract states that HPMC maintained it "for biologically relevant timeframes" C51. The authors leave the in vivo consequences open: they are "not clear at this point" C50.

Note by the AI writer (not discussed in the paper): for HPMC–felodipine, lowering the drug loading from 50% to 10% made release faster and more complete but did not raise the dissolved drug above the amorphous level; the extra drug ended up as particles, likely crystalline, that could seed further crystallization. In practice, a medium that turns cloudy early in dissolution is a sign that UV probe values include particles and that a particle-blind method is needed before supersaturation is stated.

Limitations

Limits stated by the authors

• Flux and NMR need at least 40 min of data and cannot resolve dissolution kinetics, so a short-lived local supersaturation from the 90:10 dispersions "cannot be discounted" C28.
• The equality of activities holds "at least over the timeframe that these experiments were conducted" C22.
• Formation of the particles from nanometre-scale amorphous regions of the matrix "cannot be entirely ruled out" C36.
• The match between the 90:10 and pure amorphous activities "may be purely coincidental" C48.
• The impact on in vivo performance "is not clear at this point and will require further investigation" C50.

The AI writer's own judgement

• Only HPMC–felodipine at 37 °C was tested with a particle-blind method. The abstract's statement that bulk supersaturation did not depend on the drug–polymer ratio, and the extension to PVP and indomethacin, rest for the other three systems on UV and DLS alone X16.
• "No significant difference" is not supported by a statistical test; the three flux ratios overlap within their SDs (n = 3) but follow the order of the UV readings, and all share one crystalline normaliser (n = 1) although the methods state triplicates X5·X4.
• NMR is one pair of integrals differing by 26% (relative to 90:10) or 35% (relative to 50:50); it excludes a threefold difference but does not show equality X6.
• The crystallinity of the early 200–300 nm particles was inferred from UV spectra, from birefringence at about 1 µm that is not shown, and from the known behaviour of amorphous felodipine; no diffraction was done, and the nanosuspension reference spectrum comes from a thesis X7.
• The similarity of dispersion-born and methanol-born particles holds at 25 °C; at 37 °C dispersion-born particles were consistently larger X19.
• Several printed details disagree: the apparent 90:10/50:50 ratio is given as about 2.5 while the plotted values give about 3; the membrane cutoff is 6–8 kDa in the text and 10 kDa in Figure 1; the donor traces of Figure 10 repeat those of Figure 2; Eq. 5 carries an extra factor $T_m$; a reference to Figure 3 should point to Figures 2 and 10 X2·X1·X3·X14·X11.
• "Biologically relevant timeframes" rests on UV profiles of about an hour (two hours for PVP–indomethacin), and the 90:10 UV values include particle absorbance X10.
• Only two drug loadings, 10% and 50%, were tested, so where between them the behaviour switches is unknown A1.

Conclusion

The paper is a cautionary tale about measuring supersaturation. By in-situ UV, a dispersion with 90% polymer looked as if it dissolved to about three times the concentration of one with 50% polymer, with any polymer and either drug R1·R5·R16. For HPMC–felodipine at 37 °C, two methods that see only dissolved drug, membrane flux and solution NMR, showed instead that 90:10, 50:50 and pure amorphous felodipine all reach about the same activity, close to the estimated amorphous solubility, nine times the crystal R14·R15·M1·C22. What differs is the route. The 90:10 dispersion falls apart within 5 min and, in the authors' reading, releases nearly all its drug at once into a highly supersaturated local solution from which felodipine immediately precipitates as submicron particles, which they conclude are crystalline and which the UV probe counts as dissolved drug R4·R8·C47·C33·C27. The 50:50 dispersion dissolves slowly behind a drug-rich surface layer that "presumably" caps the concentration at the amorphous level R4·C46. The local supersaturation and the crystallinity of the early particles were inferred rather than measured, and whether the equal end point is more than a coincidence is left open C28·X7·C48.