How a trace of crystals left by hot melt extrusion undermines posaconazole supersaturation

Background and motivation

An amorphous solid dispersion (ASD) holds a poorly soluble drug in amorphous form inside a polymer. Under nonsink conditions it dissolves to concentrations above the crystalline solubility, which raises the driving force for flux across the gut membrane; but such supersaturated solutions are metastable, and once crystals nucleate and grow the advantage is lost (Section 1, refs 3–10). C1 Purohit et al. had argued that crystalline drug left in an ASD "might" speed up this loss, because existing crystals bypass the barrier to primary nucleation (ref 11). C2

Hot melt extrusion (HME) is a common way to make ASDs. Below the drug's melting point, the crystalline drug must dissolve in the softened polymer, and whether it does so completely depends on the melt viscosity, the drug particle size, the processing temperature, the drug's solubility in the polymer and the residence time (refs 12, 15). Incomplete dissolution leaves "residual" crystals, which can later seed crystallization in the matrix during storage or from solution during dissolution (refs 13, 16). C3 Extruding above the drug's melting point avoids the problem but risks thermal degradation (refs 18, 19). Specific drug–polymer interactions depress the melting point, so complete amorphization is possible at lower temperatures; the minimum temperature for a given drug loading can be estimated with Flory–Huggins theory (refs 13, 20–23, including the temperature–composition phase diagrams of Moseson and Taylor, ref 13). C4 For tacrolimus, capsules with partly crystallized drug gave notably lower absorption in beagle dogs (ref 17), and a physiologically based pharmacokinetic (PBPK) model of the same system suggested that 20% crystallinity could already give lower than acceptable bioavailability (ref 11). R24

Two gaps motivated the study. First, the level of residual crystallinity that does no harm is poorly understood: it "likely" depends on how fast the drug crystallizes and how well the polymer inhibits growth, it can be extremely low, and in practice specification limits follow what the available methods can detect. C5 Powder X-ray diffraction (pXRD), the most common method, "may not" detect low crystallinity, especially at low drug loading, and ASDs that looked amorphous by pXRD have been shown to contain crystals by more sensitive methods (ref 11). C6·R26 Second, earlier work on residual crystallinity used a slowly crystallizing drug: Moseson et al. saw only a minor effect for indomethacin–PVPVA ASDs (ref 28). Fast crystallizers behave differently; Elkhabaz et al. found that posaconazole–HPMCAS ASDs crystallize more readily as drug loading rises in a pH-shift dissolution test (ref 29), and residual crystallinity "is expected" to accelerate crystallization further. C7·R26 Most dissolution studies also follow only the solution concentration, which hides the phase transformations underneath; techniques such as second harmonic generation (SHG) microscopy and two-photon excitation ultraviolet fluorescence (TPE-UVF) can follow the solid state during release (refs 29, 30). C8

The authors therefore set out to measure how low levels of residual crystallinity in ASDs of a fast-crystallizing drug affect the generation and maintenance of supersaturation and the crystallization kinetics. They chose posaconazole (PCZ), a weakly basic antifungal (pKa 3.6 and 4.6) that crystallizes fast, has a glass transition temperature, melting point and thermal stability suited to HME, and has a noncentrosymmetric crystal structure that makes SHG possible; the polymer was hydroxypropyl methylcellulose acetate succinate of MF grade (HPMCAS MF), chosen for its suitability for HME and its ability to inhibit PCZ crystallization (Figure 1, Table 1). C10·M1·M2 Their hypothesis was that residual crystals, even at low levels, "will accelerate crystallization" in the hydrating ASD matrix and lead to rapid desupersaturation once the polymer dissolves and releases the drug. C11

Background concepts

Melting point depression and the critical solution temperature. When crystalline drug is heated with a polymer in which it can dissolve, the last crystals disappear at a temperature below the melting point of the pure drug. The paper calls the temperature at which all drug of a given loading can dissolve, given enough time, the critical solution temperature, $T_c$, and reads it as the end (offset) of the drug's melting endotherm in a drug–polymer physical mixture (Section 3.1). C18·P2 Because the dissolution of crystals in the polymer during heating is a rate process, the measured offset depends on the heating rate; the authors therefore extrapolate the offset to zero heating rate (ref 39). C12 The Flory–Huggins equation relates $T_c$ to composition (Eq. 1; SI Eq. S1):

Here $\Delta H$ is the enthalpy of fusion of PCZ (50323.87 J/mol), $R$ the gas constant (8.314 J/(mol K)), $T_m$ the melting temperature of neat PCZ (441.78 K), $\varphi$ the volume fraction of PCZ, $m$ the ratio of the molar volume of HPMCAS to that of PCZ (25.95) and $\chi$ the Flory–Huggins interaction parameter (SI Section S3; the printed equation uses square brackets around the second term). E1·M1·M2 In plain terms: the left side collects how far the dissolution temperature is depressed and the ideal entropy of mixing a small drug with a long polymer; whatever remains is attributed to the drug–polymer interaction, $\chi$, weighted by the square of the polymer volume fraction. A positive $\chi$ means mixing is mildly unfavourable, so the depression is small. The molar volumes come from molecular weight divided by true density: 701 g/mol and 1.28 g/mL give 546 cm³/mol for PCZ, and 18,000 g/mol and 1.27 g/mL give 14,164 cm³/mol for HPMCAS (Table 1). E4·M1·M2 To obtain $\chi$, the authors rearrange Eq. 1 into a straight line (SI Section S3):

In plain terms: each measured mixture gives one point $(B,A)$, and $\chi$ is the slope of the line through them. E3·R4

Glass transition of the dispersion. The glass transition temperature, $T_g$, of a single-phase ASD lies between those of its components. The paper predicts it with the Gordon–Taylor equation, with the constant $K$ estimated from the true densities $\rho$ and the glass transitions of drug ($d$) and polymer ($p$) (Eq. 2):

Here $w$ is the weight fraction. In plain terms: the $T_g$ of an ideally mixed dispersion is a weighted average of the two pure $T_g$ values (61 °C for amorphous PCZ, 122 °C for HPMCAS MF), with $K$ setting how much weight the polymer carries. E2·M1·M2 A single $T_g$ that follows such a curve is the usual sign of a single amorphous phase. C17

Supersaturation, LLPS and amorphous solubility. For a weak base like PCZ, the crystalline solubility falls steeply as the pH rises from gastric to intestinal values. C22 If the concentration of a supersaturated solution exceeds the amorphous solubility, the solution separates into a drug-rich phase dispersed as nanodroplets and a drug-lean aqueous phase; this liquid–liquid phase separation (LLPS) sets in at a concentration that approximates the solubility of the water-saturated amorphous drug (refs 47, 48). C14 In this paper the "solution" during release is taken to hold molecularly dissolved drug, drug associated with micelles of the biorelevant medium and, above the LLPS concentration, drug in amorphous nanodroplets. C50

Primary and secondary nucleation, growth and matrix crystallization. New crystals can form in a supersaturated solution without any crystal present (primary nucleation) or with the help of crystals already present (secondary nucleation); existing crystals can also simply grow and consume supersaturation. In an ASD particle, crystallization can also happen inside the solid matrix, before the drug is released. The paper's interpretation uses all four routes (Figure 11). C2·C45·C49

Three orthogonal solid-state probes. SHG is generated only by crystals that lack a centre of symmetry; PCZ, with four chiral centres, crystallizes in a chiral space group and is SHG-active, while amorphous PCZ, HPMCAS and dissolved drug give no signal (refs 42–45). C13·M1 TPE-UVF lights up any region with a high concentration of PCZ, whether amorphous ASD or crystal, but not the dilute dissolved drug, so a fading TPE-UVF image means particles are dissolving. C13 Focused beam reflectance measurement (FBRM) is a probe placed in the vessel that measures chord lengths of particles passing its window; here the counts were sorted into 1–10, 10–100 and 100–250 µm classes every minute, so falling counts mark particle "death" by dissolution and rising fine counts mark particle "birth" by nucleation. P14·C57

Methods in detail

The study has four parts that build on each other: a phase diagram that predicts which extrusion temperature leaves crystals; extrusion at two temperatures with crystallinity measured by two methods; a two-stage dissolution test watched by solution assay, SHG, TPE-UVF and FBRM; and a seeded desupersaturation test that isolates the effect of crystal seeds and polymer. The flow below summarizes them.

Part 1: a phase diagram that predicts the safe extrusion temperature

Problem. To make ASDs with and without residual crystals from the same drug, polymer and loading, the authors needed to know which extrusion temperature leaves crystals undissolved. C18

What they did. Crystalline PCZ (Form I) and HPMCAS were cryomilled together at weight ratios from 10:90 to 90:10 in steps of 10% (1 g precooled 5 min in liquid nitrogen, milled 2 min at 10 Hz). P1 Samples of 3–5 mg were heated in a TA Q2000 DSC from 25 to 180 °C, cooled at 10 °C/min and reheated to 180 °C at 10 °C/min; the first heating was run at 10, 5, 3, 2, 1 and 0.5 °C/min, in triplicate, and the melting offset temperature was extrapolated linearly to 0 °C/min to give $T_c$ (SI Figures S1–S3). P2·R2·R3 The offsets fell with slower heating, for neat PCZ from about 170.9 °C at 10 °C/min to about 168.7 °C at 0.5 °C/min, and for the 40:60 mixture from about 165.6 to about 162.0 °C (read from Figure S3, approximate). R3 Measured $T_c$ values are reported only for PCZ weight fractions from 0.4 to 1.0; the paper does not say why the lower fractions are missing. R1·X7 The linearised plot of $A$ against $B$ for six mixtures gave $\chi$ = 0.43 with $R^2$ = 0.991 (Figure S4), and Eq. 1 with this $\chi$ then gave $T_c$ = 135, 155 and 165 °C at 10, 25 and 50 wt% PCZ (Figure 3). R4·R5 The second heating gave the $T_g$ of each mixture, compared in Figure 3 with Eq. 2 using the pure-component $T_g$ and densities measured by helium pycnometry (1.28 g/mL for PCZ, 1.27 g/mL for HPMCAS MF). P4·R6·M1·M2

Checks by the AI writer. Recalculating Eq. 1 with the printed inputs reproduces the three $T_c$ values within about 1 °C (135.9, 155.3 and 165.2 °C, converting weight to volume fraction with the two densities). E1 Two details deserve a note. First, the fitted line in Figure S4 crosses $B=0$ at $A$ of about −0.05, so 0.43 is the slope of a line with a free intercept rather than of $A=\chi B$; a line forced through the origin would have a slope of about 0.24 (AI calculation from plot readings), which would still put $T_c$ above 125 °C at all three loadings (about 132, 152 and 164 °C). X8 Second, the dashed Gordon–Taylor line in Figure 3 is reproduced only when $K$ is computed with temperatures in °C ($K$ = 0.504); with absolute temperatures, as the density form of $K$ requires, $K$ = 0.852 and the predicted $T_g$ at 0.5 weight fraction would be 89.1 °C instead of the plotted value of about 81.5 °C. The methods also say the measured $T_g$ "were fitted", whereas the results say pure-component values "were used". E2·X2

Part 2: two extrusion temperatures and two ways to measure crystallinity

What they did. Physical mixtures at 10, 25 and 50 wt% drug loading (DL) were extruded in a 5 mL corotating twin-screw extruder with recirculation at 100 RPM for 2 min, at 175 °C (above the calculated $T_c$ for all three loadings) or 125 °C (below it) (Table 2). The extrudates were cooled at ambient temperature, cryomilled and sieved, and the 105–250 µm fraction was used throughout. P5·A1·A2

Crystallinity was measured in two ways. pXRD (Rigaku SmartLab, Cu Kα, 5–40° 2θ at 0.2°/min) used the area of the strongest PCZ peak at 18.0° 2θ, calibrated against ASDs spiked with 0–25 wt% crystalline PCZ ($R^2$ = 0.998); the limit of detection (LOD) was 0.3 wt% and the limit of quantification (LOQ) 1.2 wt% crystalline PCZ in the ASD (Figure S6). P6 SHG microscopy used a SONICC instrument with a 1064 nm femtosecond laser; particles were held in a thin spacer between cover glasses, imaged in planes 50 µm apart over a 1925 × 1925 µm field, and the summed intensity of the brightest plane and two planes on either side was calibrated against the same kind of spiked ASDs (0–25%, $R^2$ = 0.999; Figure S7). P7

Part 3: a two-stage dissolution test watched in solution and in the solid

Problem. A concentration profile shows only the net result of particle dissolution and crystallization. To tell them apart, the authors followed the solid as well as the solution. C57

What they did. First the media were characterized: the equilibrium solubility of Form I after 48 h at 37 °C and the LLPS concentration by solvent shift, adding PCZ stock at 4 µg/min and watching UV extinction at 400 nm (n = 3). P8·P10 For release, ASD particles were added to 40 mL fasted-state simulated gastric fluid (FaSSGF, pH 1.6) at 37 °C in a 100 mL jacketed beaker; after 60 min, 40 mL of double-strength fasted-state simulated intestinal fluid version 2 (FaSSIF V2) was added and the pH set to 6.5 with 1 M NaOH (Figure 2, adapted from Elkhabaz et al.). P11 The dose was 10 times the amorphous solubility of PCZ in FaSSIF V2, about 200 µg/mL at complete release, given as 160 mg of the 10%, 64 mg of the 25% or 32 mg of the 50% DL ASD, so that the polymer concentration in the medium differs between loadings while the drug dose stays fixed. P11·A2 Samples of 0.5 mL were taken for up to 5 h, centrifuged at 3000 rpm for 3 min to remove undissolved particles and assayed by HPLC with fluorescence detection (n = 3). P12·P9 Aliquots taken up to 3 h were imaged by SHG and TPE-UVF, and an FBRM probe in the beaker recorded chord-length counts every minute. P13·P14 A separate set of samples was imaged by SHG before and after 60 min in FaSSGF to measure crystallization inside the ASD particles during the acid stage. P15

Part 4: seeded desupersaturation

What they did. To separate the roles of seeds, drug concentration and polymer, HPMCAS was predissolved in FaSSIF V2, Form I seeds (5 or 10 µg/mL, or none) were equilibrated in it for 48 h at 37 °C, and supersaturation was created by adding a 10 mg/mL PCZ stock in methanol to 20 or 100 µg/mL. P16 HPMCAS was added at 3 or 9 times the PCZ concentration (60 or 180 µg/mL at 20 µg/mL PCZ; 300 or 900 µg/mL at 100 µg/mL), stated to match the drug:polymer ratio of fully dissolved 25 and 10 wt% DL ASDs, with controls without seeds and without polymer; samples at 5–30 min were centrifuged at 10,000 rpm and assayed by HPLC (n = 3). P16·C15 Note by the AI writer (not discussed in the paper): 20 µg/mL is about the LLPS concentration in FaSSIF V2 (19.1 µg/mL), while 100 µg/mL is about five times it, so the two starting points differ in phase state as well as in supersaturation. R11·X16

Experimental analysis

The phase diagram and the extrudates

The extrapolated $T_c$ of neat PCZ was about 168.6 °C and fell only to about 161.7 °C at 0.4 PCZ weight fraction, a small depression (Figure 3, read, approximate). R1 The authors read $\chi$ = 0.43 as "weak interactions" in the results and as "moderately unfavorable interactions" in the discussion, with the small depression driven by the favourable entropy of mixing; for comparison, $\chi$ was −1.97 for indomethacin/PVPVA and −1.57 for bicalutamide/PVPVA (refs 13, 49). C16·R23·X17 The measured $T_g$ fell smoothly from about 121.7 °C for HPMCAS to about 61.2 °C for PCZ, which the authors read as "consistent with" a single-phase ASD; the points lie a few degrees below the predicted line. R6·C17 DSC of the three ASDs extruded at 175 °C showed one glass transition each and no melting peak (Figure S9). R7 Note by the AI writer: in Figure S9 the trace labelled 50% DL carries a $T_g$ of 110.72 °C and the trace labelled 10% DL 79.82 °C, the reverse of Figure 3 and of the mixing rule, so these two labels appear to be swapped. X1

Extrusion at 175 °C gave clear amber strands, extrusion at 125 °C opaque ones, white at 50% DL (Figures 4 and S5). R8 Table A collects the crystallinity of the six extrudates.

Table A. Residual crystallinity of the six extrudates (text of Sections 4.2–4.4, Figures 4 and 5). R5·R9·R10

The ranking follows the phase diagram: more drug above its solubility in the polymer at 125 °C left more crystals, which the authors take as confirmation that the 125 °C ASDs kept crystals "because they were prepared below the extrapolated solubility temperature". C20 They see in this a "potential utility" of the phase diagram for choosing "safe" processing temperatures, from a thermodynamic view only, since mixing intensity and residence time distribution also matter. C21 At 10% DL, pXRD could not detect the crystals that SHG measured at 0.3%; the authors write that pXRD "was not sensitive enough". C19 Note by the AI writer: the stated pXRD LOD is itself 0.3 wt%, so the 0.3% lies at the detection limit and clearly below the LOQ of 1.2 wt%; the abstract calls it below the "quantification limit" and the conclusions below the "detection limit". X4 The two methods also disagree at the higher levels (23.1% by pXRD against 19.3% by SHG at 50% DL, 2.4% against 3.2% at 25%); the paper uses the SHG values. X5

The gastric stage

Table 3 sets the scale of the test. In FaSSGF the equilibrium solubility of Form I was 86.5 µg/mL and the LLPS concentration 2326 µg/mL; in FaSSIF V2 they were 1.2 and 19.1 µg/mL (n = 3). R11 In FaSSGF, where HPMCAS is effectively insoluble, all six ASDs released PCZ slowly, and at 60 min the 125 °C ASDs had released more (about 38, 43 and 73.5 µg/mL at 10, 25 and 50% DL) than the 175 °C ASDs (about 30, 28.5 and 39.5 µg/mL; read from Figures 6a, 7a, 8a, approximate). R12·R15·R18·C23 All these values lie below the gastric equilibrium solubility, so the gastric solution was subsaturated. C40 FBRM counts of 100–250 µm particles fell by about 10–20% during this hour (for example from about 810 to about 720 counts/s for 10%-175 °C), which the authors read as some shrinkage of the particles ("indicating"). R14·R17·R20·C24 At 60 min TPE-UVF still showed undissolved particles for every ASD, while SHG showed crystals only for the 125 °C ASDs, "suggesting" that the 175 °C ASDs contained only amorphous material. R13·R16·R19·C25

Why should crystal-containing ASDs release more in acid, when amorphous drug is normally the more soluble form? The authors' "likely reason" is that in the fully amorphous ASDs the drug is molecularly mixed with HPMCAS (one $T_g$ in Figure S9), which does not dissolve at gastric pH and so holds the drug back, in line with a thermodynamic explanation given by Li and Taylor for enteric-polymer ASDs (ref 50); crystalline PCZ, phase-separated from the polymer and far below its solubility, dissolves freely. C41 This explanation was not tested directly; release of neat amorphous or neat crystalline PCZ in FaSSGF was not measured. C41

SHG crystallinity also rose during the acid hour, but only where crystals were already present: from 0.3 to 0.4% (10% DL), from about 3 to 4% (25% DL) and from about 19 to 26% (50% DL), while the 175 °C ASDs stayed at 0 (Figure 9, n = 3). R21·C36 The authors read this as matrix crystallization promoted by the residual crystals, "likely through growth of existing crystals present in the matrix following absorption of water and plasticization"; nucleation "does not appear to be favored" within 1 h, since the amorphous ASDs stayed amorphous. C49 Note by the AI writer: the rises at 10 and 25% DL are small and no statistical test is given, and the schematic in Figure 11 lists matrix growth only for the 25% and 50% DL ASDs. X11

The intestinal stage, loading by loading

After the pH shift, supersaturation forms in two ways, as the authors describe it: drug already released in acid loses its charge and solubility, and the now-soluble polymer releases further drug, a release that "can be considered as being controlled by the polymer" (ref 51). C42 Table B condenses the concentration profiles.

Table B. PCZ concentration (µg/mL) in the pH-shift test, read from Figures 6a, 7a and 8a (approximate; n = 3); the last column paraphrases the text of Section 4.6. R12·R15·R18·C27·C28·C29·C34

10% DL. Both ASDs dissolved quickly after the shift: concentrations rose, the TPE-UVF images went dark because the dissolved drug is too dilute to detect, and the FBRM total fell from about 890 to about 50 counts/s within about 15 min for the amorphous ASD. R12·R13·R14·C26 Both reached about nine times the LLPS concentration, which the authors take as release up to 10 times the amorphous solubility, initially with no crystals. R12·R11·C50 The amorphous ASD then showed no SHG signal up to the last image at 180 min and no rise in FBRM counts, and the authors call its supersaturation "maintained for the duration of study" (Figure 6a–c). R13·R14·C27 Note by the AI writer: its concentration nevertheless fell from about 172.5 to about 125 µg/mL by 300 min, a loss of about a quarter that the paper does not discuss. X10 The ASD with 0.3% crystals peaked at about 178 µg/mL and then collapsed to about 18.5 µg/mL by 150 min; its 1–10 µm FBRM counts started to rise at about 78 min and reached about 480 counts/s near 165 min, and SHG spots multiplied from 90 to 180 min (Figure 6d, e). R12·R13·R14 The authors write that "secondary nucleation was detected after 15 min following pH shift". C28

25% DL. Both ASDs peaked at about 172.5–177 µg/mL at 75 min. R15 The amorphous ASD held its supersaturation "for around 1.5 h after pH shift" and then crystallized: about 143.5 µg/mL at 150 min, about 13 µg/mL by 180 min, with a jump of fine FBRM counts at about 150 min and faint SHG spots at 180 min. R15·R16·R17·C29 With about 3% crystals the induction time "was reduced to 45 min after pH shift": the concentration was about 154.5 µg/mL at 105 min and about 14 µg/mL at 120 min, and the 1–10 µm counts jumped to about 390 counts/s at about 118 min, then fell to about 100 while the 10–100 µm counts rose to about 235 by 300 min. R15·R17·C29 The authors read the jump as rapid secondary nucleation and the later shift to larger particles as crystal growth, and conclude that "both (secondary) nucleation and crystal growth kinetics appeared faster" with residual crystallinity. C30·C31 Note by the AI writer: the FBRM signature of the amorphous ASD (flat, then a sudden jump of fine counts) looks the same, although Figure 11 assigns that run to primary nucleation; the assignment to secondary nucleation rests on knowing that crystals were present, not on a measurement that separates the two. X14

50% DL. Neither ASD dissolved well at pH 6.5: the amorphous one stayed near 27 µg/mL until 75 min and the partially crystalline one dropped from about 48.5 µg/mL at 65 min to about 23.5 at 75 min; both were below about 20 µg/mL from 90 min on, large particles persisted in FBRM and TPE-UVF, and the authors call the dissolution "incomplete". R18·R19·R20·C32 The amorphous ASD kept more large particles (100–250 µm counts of about 105 against about 40 counts/s for the 125 °C ASD after the shift), and the authors write that the partially crystalline ASD particles "appear to disintegrate into smaller particles". R20·C33 Note by the AI writer: the paper cites Figure 8b and 8c, the panels of the amorphous ASD, for this disintegration; the FBRM data show it more strongly in Figure 8e. X12 The amorphous 50% ASD recrystallized from solution about 15 min after the shift, and its undissolved particles showed SHG signal 120 min after the shift, which the authors read as crystalline PCZ "generated because of matrix crystallization". R18·R19·C34·C35 Their explanation for the poor dissolution is that at high loading the molecularly mixed hydrophobic drug "substantially" hinders hydration and dissolution of the polymer: at low pH the polymer holds back the drug, at high pH a large amount of drug holds back the polymer. C43

Across the three loadings the authors conclude that ASDs with residual crystallinity desupersaturated and crystallized faster than those with no initial detectable crystallinity. C44 Note by the AI writer: in all six runs the concentrations after crystallization levelled off at about 11–16 µg/mL within 5 h, still about 9–13 times the crystalline solubility of 1.2 µg/mL, so "desupersaturation" here means a fall below the LLPS concentration, not a return to equilibrium. X15

Seeded desupersaturation

Table C gives the fraction of PCZ left after 30 min in the seeded tests.

Table C. PCZ remaining at 30 min (% of initial) in FaSSIF V2 at 37 °C, read from Figure 10 (approximate; n = 3). R22

Desupersaturation was faster at the higher drug concentration and with more seeds, and slower with more predissolved HPMCAS. R22·C37 The polymer's effect was weaker at 100 µg/mL, and HPMCAS was "unable to prevent desupersaturation when a high concentration of seeds was present", although it was "quite effective" at delaying crystallization without seeds or with few. C38·C39 With 10 µg/mL seeds at 100 µg/mL and 300 µg/mL HPMCAS, about 58% of the drug remained after only 5 min (Figure 10e). R22 The authors use this fast seed growth, even with polymer present, to argue that growth of existing crystals is "likely important" in the ASD test too. C46

Discussion

How the authors connect the findings. The authors' account runs from processing to release, and splits by whether seeds are present and by loading. Residual crystals consume the supersaturation generated in intestinal conditions by crystal growth and by inducing secondary nucleation, and they can "potentially" promote matrix crystallization in gastric medium; "each of these mechanisms is likely in play, albeit to different extents". C45 The balance between growth and secondary nucleation "probably changes" with the amount of residual crystallinity, growth "likely" dominating when there is a lot of it. C48 Figure 11 summarizes their pathways, and the chain below follows it.

The causal chain as the authors give it (Sections 5.1.1–5.1.2, Figure 11); steps 2 and 3 and the nucleation type are interpretations, not direct measurements. C41·C42·C43·C45·C49

Why orthogonal methods. The authors stress that under nonsink conditions PCZ can sit in several phases at once: dissolved, associated with micelles, in amorphous nanodroplets, in undissolved ASD particles (amorphous or crystalline) or as new crystals from solution; the dissolved, micellar and nanodroplet forms give no SHG, TPE-UVF or FBRM signal. C50 SHG then maps crystallization, TPE-UVF maps particle dissolution and FBRM tracks particle death, birth and size, so that together they give what the authors call "an unprecedented level of information" on particle disappearance, appearance and phase. C51 In their conclusions they write that a combination of orthogonal techniques "was necessary" to separate the loss of particles by dissolution from the formation of new particles by crystallization. C57 Note by the AI writer: the nanodroplets themselves were not measured (for example by light scattering); their presence is inferred from concentrations above the LLPS concentration. X13

How much crystallinity is acceptable? The authors propose that the acceptable level of residual crystallinity "could be considered as the maximum level" at which no significant further crystallization occurs over a physiologically relevant period, the time over which most of the drug is absorbed, and which does not notably reduce bioavailability; above it a system "is expected" to crystallize within that period. C53 They cite in vivo evidence that crystallinity matters: tacrolimus ASDs with 50% crystalline drug gave significantly lower absorption in dogs, a PBPK model suggested that 20% could already be too much, and 25 wt% DL celecoxib–PVP ASDs lost rate and extent of absorption in rats as crystallinity rose (refs 11, 17, 58). R24·R25 Finding the acceptable level, keeping crystallinity below it during manufacture and measuring it with sensitive techniques are, in their words, "essential for a successful ASD performance". C54

What it suggests for formulation (note by the AI writer, not stated in the paper). For this fast crystallizer at 10% DL, crystallinity at the pXRD detection limit was enough to lose the supersaturation within 1.5 h of the pH shift, so a release specification of "X-ray amorphous" would not have protected performance. R9·R10·R12 In the seeded tests even nine times more HPMCAS than drug did not stop desupersaturation with 10 µg/mL seeds, which suggests that adding polymer is a weak remedy for residual crystals and that the extrusion step has to prevent them. R22 At 50% DL the dispersions failed whether or not they contained crystals, so at such loadings controlling crystallinity alone would not secure release. R18 Only two temperatures were tested and the calculated $T_c$ at low loading is an extrapolation, so in practice a margin above the calculated $T_c$ would be prudent. X7·X8

Limitations

Limits the authors state. The phase diagram gives "safe" processing temperatures only "from a thermodynamic perspective whereby kinetic considerations (e.g., mixing intensity and residence time distribution) are also important". C21 The balance of growth and secondary nucleation is described only as "probably" changing with crystallinity, and matrix nucleation as something that "does not appear to be favored" only "over the time frame of the experiment (1 h)". C48·C49 The acceptable level of residual crystallinity remains to be defined for each drug–polymer system. C52·C53

The AI writer's own judgement (1): what is measured and what is inferred. The central result is direct: at 10 and 25% DL, three independent readouts (solution assay, SHG, FBRM) show the crystal-containing ASD crashing much earlier, in triplicate runs. R12·R13·R14·R15·R16·R17 The mechanisms are softer. Secondary nucleation is inferred from FBRM patterns that look the same in the amorphous 25% DL run, which the authors assign to primary nucleation. X14 The gastric release difference and the poor dissolution at 50% DL are explained by molecular mixing and by drug hindering polymer hydration, neither of which was measured. C41·C43 Matrix growth in acid is clear at 50% DL but small and untested at 10 and 25% DL. X11 The seeded tests (methanol solvent shift, pre-wetted seeds, 30 min) model the seed effect qualitatively rather than reproduce the conditions inside the ASD test. X16

The AI writer's own judgement (2): the phase-diagram rule and source slips. The processing rule rests on two temperatures, and $T_c$ at 10 and 25% DL is extrapolated below the measured range of 0.4–1.0 PCZ weight fraction; the fitted $\chi$ also depends on a free intercept in Figure S4, although a fit through the origin would not change the classification of 125 and 175 °C. X7·X8 Several small inconsistencies do not change the conclusions but matter for anyone reusing the numbers: the Gordon–Taylor line in Figure 3 corresponds to $K$ computed with Tg in °C; Table 1 gives inflection rather than onset $T_g$ values; the Figure S9 labels for 10% and 50% DL appear swapped; the pXRD limit is worded as both a detection and a quantification limit; the "complete release" line in Figures 6a–8a lies at about 187–188 µg/mL against a stated dose of about 200 µg/mL; and the Figure S3 caption names a single heating rate. X2·X3·X1·X4·X9·X6 The amorphous 10% DL ASD's "maintained" supersaturation still fell by about a quarter. X10 Finally, the work covers one drug, one polymer grade and one particle size, without absorption data or storage, so the size of the in vivo effect for posaconazole is not known from this paper. R24·R25

Conclusion

The paper links three things for posaconazole–HPMCAS MF dispersions. A Flory–Huggins temperature–composition diagram predicted that extrusion at 125 °C would leave crystals at 10, 25 and 50% DL and that 175 °C would not, and SHG confirmed it, measuring 0.3, 3.2 and 19.3% crystallinity, including a level that pXRD could not quantify. R5·R10·C20 In a two-stage dissolution test these residual crystals shortened or reduced the supersaturation at every loading, most clearly at 10% DL, where the amorphous dispersion stayed supersaturated through the 5-h test while the one with 0.3% crystals crashed within 1.5 h of the pH shift. R12·R15·R18·C56 Solid-state monitoring by SHG, TPE-UVF and FBRM, together with seeded solution tests, led the authors to attribute the loss to crystal growth and secondary nucleation, with crystal growth also inside the ASD matrix in acid ("likely in play"). C45·C47·C49·R22 Their practical message is to choose extrusion conditions from the phase diagram, measure crystallinity with a sensitive method such as SHG, and test release with a discriminating, two-stage method. C21·C60