Reading drug–polymer miscibility from a drug's own fluorescence, and what it says about crystallization in storage

Background and motivation

An amorphous drug dissolves faster and reaches higher concentrations than its crystal, but it tends to crystallize over time and lose that advantage. The usual remedy is an amorphous solid dispersion (ASD): the drug is mixed at the molecular level into an amorphous polymer, which slows crystallization. Several marketed products rely on this approach, and choosing the polymer is a central formulation decision (Introduction, refs 1–6). C1

The paper frames polymer choice through two thermodynamic limits, taking the polymer as a solvent for the drug (Qian 2010, ref 7). The first is the crystal solubility of the drug in the polymer: below it an ASD is thermodynamically stable and cannot crystallize, but this solubility is "virtually always too low" for practical loadings (refs 8–11). The second is the miscibility limit, the amorphous solubility of the drug in the polymer: the highest loading at which drug and polymer still form one homogeneous amorphous phase. Above the miscibility limit, amorphous–amorphous phase separation is favoured and a drug-rich phase forms. Crystallization from that phase "is anticipated", because it holds less polymer, and this would harm dissolution and bioavailability (refs 4, 12, 13). Between the two limits there is a driving force to crystallize but not to phase separate, and the drug "may remain" amorphous because nucleation and growth are slow in a glass (refs 14, 15). From this the authors draw the rule that the polymer "should be miscible with the drug". C2·C3·C4·C5·C6

Testing that rule needs a reliable way to detect phase separation, and the authors list the weaknesses of the usual methods. DSC is the most common: one glass transition (Tg) is read as miscible, two as phase separated. But domains smaller than about 30 nm, phases with similar Tgs, or phases that remix on heating all hide a second Tg, and both a single Tg for a separated system and two Tgs for a miscible one have been reported (refs 16–21). Electron and atomic force microscopy resolve nanoscale domains but cannot say which phase is which; Raman mapping identifies composition but is slow and resolution-limited; XRD, IR and solid-state NMR (ssNMR) are also used (refs 22–30). C7·C8

The second gap is the link itself. Miscible ASDs "should" be more stable (Ivanisevic 2010, ref 31), but the correlation between miscibility and crystallization tendency had not been studied much. Earlier Taylor-group work connected fast crystallization of felodipine–PVP dispersions at high humidity to moisture-induced phase separation (Rumondor 2009, ref 40), and solvent-dependent miscibility of spray-dried naproxen–PVP to differences in stability at high humidity (Paudel 2012, ref 41); in both, water was part of the story. C9·C51

This paper therefore does two things. It uses the autofluorescence of indomethacin (IDM) as a built-in probe of the drug's surroundings, alongside fluorescence imaging, IR and DSC, to find the miscibility limit of IDM in hydroxypropyl methylcellulose (HPMC), hydroxypropyl methylcellulose acetate succinate (HPMCAS) and polyvinylpyrrolidone (PVP). Then it stores dispersions over desiccant at 40 °C, so that absorbed water cannot change miscibility, and asks whether the phase-separated ones crystallize sooner. C10·C52·P8

Background concepts

Miscibility and the drug-rich phase. A miscible ASD is "a homogeneous single amorphous phase" in which drug and polymer are mixed at the molecular level (Baird 2012, ref 19). It is still metastable with respect to the crystal, but the polymer slows crystallization kinetically. If the loading exceeds the miscibility limit, or if processing leaves the material inhomogeneous, a drug-rich phase with less polymer forms, and crystallization "would be anticipated to occur more readily" there. C39·C4 In this paper "miscibility limit" is reported as a range of loadings: the last loading that looked homogeneous and the first that did not, with the step between loadings mostly 10%. R21

Environment-sensitive fluorescence. Some molecules emit at a wavelength that depends on the polarity of their surroundings. In earlier work, an added probe registered a hydrophilic environment in pure polymer, a hydrophobic one in pure amorphous drug and an intermediate one in miscible mixtures, changing fairly smoothly with loading (Purohit 2015, ref 12). Autofluorescent felodipine became brighter and shifted to shorter wavelength (a blue shift) when drug-rich nanodroplets formed in water (Raina 2015, ref 37). R25·R24 IDM is fluorescent through its indole ring (the authors write "tryptophan ring"; Figure 1 draws an indole), so it can act as its own probe. C11·X11 The authors' expectation follows from the probe work: in a mixture that stays miscible across all loadings, the emission should drift gradually from a polar to a less polar environment as drug replaces polymer around each IDM molecule; phase separation should instead give "more abrupt changes", with part of the IDM emitting like the pure drug. C12

The IDM carbonyl bands. Amorphous IDM shows three carbonyl bands in the IR: 1733 cm^-1 (free carboxylic acid), 1709 cm^-1 (acid carbonyl in a cyclic dimer, two IDM molecules hydrogen bonded acid to acid) and 1683 cm^-1 (amide carbonyl), as assigned by Taylor and Zografi (1997, ref 32). C25 A polymer that hydrogen bonds to the acid group breaks the dimers, so the 1709 cm^-1 band fades; if a drug-rich phase forms, the dimers, and the band, come back. C26·C27

Tg as a miscibility test. A miscible mixture shows one Tg that moves with composition; a phase-separated mixture would be expected to show two Tgs that stay roughly fixed (refs 33, 34). C30 The paper prints no equation for this; the composition dependence is read directly from the measured Tg values in Figure 10. R13

Crystallization onset. Here, the first storage time at which sharp Bragg peaks appear on the amorphous halo of the X-ray diffractogram. Because samples were scanned at intervals, an onset of "15 days" means crystals appeared between the previous scan and day 15. R17·X9

Methods in detail

The study has two halves that use different sample types. Miscibility was read mainly from thin films cast from solution; physical stability mainly from bulk powders made by rotary evaporation, with films also followed in storage. The flow below shows how the pieces fit together. P1·P2·P8

Materials and the two sample types

Indomethacin came from Letco Medical; the polymers were HPMC (Methocel E5, Dow), HPMCAS (MF grade, Shin-Etsu) and PVP K30 (BASF). Molecular weights and substitution levels are not given. M1·M2·M3·M4

Films. IDM and polymer at eleven ratios, from 5:95 to 95:5, were dissolved at 10 mg/mL total in methanol:dichloromethane (1:2, v/v). A 30 µL drop was placed on a quartz slide (for fluorescence) or a KRS-5 window (for IR), dried at about 20 °C under a nitrogen purge that kept the relative humidity at 20% or less, and then dried overnight under vacuum at ambient temperature. Drug-only and polymer-only films were made the same way as controls. P1

Bulk powders. Nine ratios, 10:90 to 90:10, in the same solvent mixture were rotary evaporated at 60 °C under reduced pressure, dried overnight under vacuum and ground in a cryogenic mill cooled with liquid nitrogen. P2

Contribution 1: the drug's own fluorescence as a miscibility probe

Problem. Added probes can change the system they report on, and DSC misses small domains. If the drug fluoresces and its emission depends on its surroundings, the drug itself can report whether it sits among polymer or among other drug molecules, "without adding probes". C40·C7

What they did. Emission spectra of fresh films were recorded from 410 to 670 nm with excitation at 355 nm on a Shimadzu RF-5301PC spectrofluorophotometer. For each film the spectrum was normalized to its maximum, and the wavelength of maximum emission was plotted against loading as the mean of three films (Figures 3–5). The quartz slide, drug-only and polymer-only films were measured as controls (Figure 2). Some films, not named, were also annealed for 5 h at Tg − 10 °C; their spectra matched the fresh ones. P3·R1·R5 The same films were viewed under a fluorescence microscope (excitation 330–380 nm, emission above 420 nm); none showed birefringence, so the films were taken to contain amorphous drug. P4·R6·C20

Why it should work. The authors' premise is that IDM in a miscible film is surrounded by polymer, a more polar environment than in pure amorphous IDM; above the miscibility limit a drug-rich, less polar phase forms and part of the IDM should emit like the pure drug. C10 The imaging step is limited by diffraction: domains smaller than about 300–500 nm cannot be resolved. The spectra, in contrast, report the local environment of each molecule, so the authors consider them "potentially sensitive" to domains too small to image. C44·C45

How it differs from earlier work. Earlier solid-state studies added an environment-sensitive probe (ref 12), and the felodipine work used autofluorescence in solution (ref 37). Here the drug's autofluorescence is used in the solid dispersion itself. R24·R25

Contribution 2: cross-checking with IR and DSC

IR spectra of the films were collected in transmission (500–3000 cm^-1, 4 cm^-1 resolution, 64 scans, purged with conditioned air). The carbonyl region of IDM–HPMC is shown in Figure 9. IDM–HPMCAS could not be evaluated because HPMCAS itself absorbs in that region, and IDM–PVP spectra are described but not shown because they had been published before (ref 32). P5·R10·R11·R12 The Tg of the freshly made powders was measured by modulated DSC on a TA Q2000: about 4 mg in a pinholed Tzero pan, equilibrated at −10 °C, heated to 190 °C at 5 °C/min with ±1 °C modulation every 60 s; each point in Figure 10 is the mean of two runs. P6·R13

Contribution 3: storage without water

Problem. In earlier studies, humidity both changed miscibility and drove crystallization, so the two could not be separated. C51

What they did. Films and powders were stored in desiccators over desiccant at 40 °C. Powders were scanned by XRD (Rigaku SmartLab, Cu Kα, 2θ 5–40°, 4°/min, 0.02° steps) at intervals that began at 2–10 days for the high loadings and became monthly; the study lasted 14 months for HPMC, 13 months for HPMCAS and 3 months for PVP. Films were imaged after 5, 12 and 26 days (30 days for PVP). The tube current is printed as "20 mV", a unit misprint. P7·P8·X2

Why it works. Removing absorbed moisture, which "can impact the miscibility", lets the correlation between miscibility and crystallization be probed directly, in the authors' words. C52 The water content of the stored samples was not measured. P8

Experimental analysis

Main result 1: the probe works. In Figure 2 the pure amorphous IDM film emits strongly with a maximum near 476 nm (intensity about 400 on the instrument scale). Films with 5% IDM emit at longer wavelength: about 510–515 nm in HPMC (about 255), about 508 nm in HPMCAS (about 250) and a weak, broad band near 515–520 nm in PVP (about 87). The neat polymer films emit only weakly near 435–440 nm (about 48–65) and the quartz slide about 15, so the IDM emission dominates. These values are read from Figure 2 and are approximate. The authors read the blue shift of the pure drug as showing that IDM emission "is sensitive to its local environment" and can probe miscibility. R1·C13 Note by the AI writer (not discussed in the paper): the Figure 2 caption lists only HPMC and HPMCAS, but the legend and the text also include the PVP film. X3

Main result 2: the emission drops abruptly at a polymer-specific loading. Table A collects the emission maxima read from Figures 3B, 4B and 5B.

Table A. Wavelength of maximum emission of fresh films, mean of three, all values about (read from Figures 3B, 4B and 5B, approximate); bold marks the first loading with the blue-shifted maximum. R2·R3·R4

For HPMC, the maximum stays at about 511–515 nm from 5 to 50% (the text says "~515 nm", with a slight fall at higher loading), then falls to about 488 nm at 60%. At 60% the normalized spectrum shows a new peak near 480 nm, similar to pure IDM, with a shoulder remaining near 515 nm; the three 60% films scatter widely (about 478–499 nm), against a few nm at the other loadings. From 70% the maximum stays at about 476–483 nm. The authors read the new peak as IDM in an IDM-rich, less polar phase next to IDM still in the polymer, so that IDM–HPMC is miscible up to 50% and the limit "appears to be" between 50 and 60%. R2·X15·C15·C16·C17 For HPMCAS the drop comes between 20% (about 513 nm) and 30% (about 480 nm), giving a limit between 20 and 30% ("appears"). R3·C18 For PVP the maximum drifts from about 518 nm at 5% to about 507–511 nm at 60–70%, then drops to about 476 nm at 80%; the ~480 nm band "is not observed until" 80%. R4·C19

The authors' reading of the whole set is that at higher loadings the spectra "appear to have contributions" from IDM in two environments, the polymer matrix and an IDM-rich phase. C41 Note by the AI writer (not discussed in the paper): the gradual drift of the PVP maximum from about 518 to 507–511 nm before the drop matches what the authors expect from a miscible system, and the drop itself is about 23–35 nm for all three polymers. R4·C12

Main result 3: imaging, IR and DSC put the limit in the same place, within about one step. In the fresh-film images (Figures 6–8), IDM–HPMC films are uniform from 5 to 50% and show green and dark domains from 60%; the authors take the green domains as IDM-rich (IDM fluoresces) and the dark ones as HPMC-rich ("are expected to be", "are thought to be"). IDM–HPMCAS films are uniform at 5 and 10%, "reasonably homogeneous" at 20%, heterogeneous at 30% and clearly separated above. IDM–PVP films are clearly heterogeneous from 70%. R7·R8·R9·C21·C22·C23 The images also show features below these limits: sparse dark spots in the 30% and 40% HPMC films, dark branched marks in part of the 20% HPMCAS film, and a network of thin dark lines in the 60% PVP film; no domain areas were measured. X12

In the IR (Figure 9), the 5% HPMC film loses the 1709 cm^-1 dimer band and gains intensity at 1733 cm^-1, which the authors read as HPMC disrupting IDM self-association ("indicate"). The spectra change little up to 50%; from 60% the 1709 cm^-1 band returns and grows with loading, "consistent with" an IDM-rich phase, so the IR limit is placed "around 50%". In the figure the 40% and 50% traces already lie between the low-loading traces and the 60% trace near 1705–1709 cm^-1, so the change may begin gradually. R10·C26·C27·X17 For PVP, films from 5 to 80% lack the 1733 and 1709 cm^-1 bands and show new bands at 1724 and 1640 cm^-1, a strong IDM–PVP interaction; above 80% the spectra resemble pure IDM. The authors estimate the IR limit as "at least 80%" and say it "can be attributed to" the strong interaction. R12·C28·C29

DSC gave one Tg for every powder (Table B). For HPMC, Tg falls steeply from about 395 K at 10% to about 328 K at 40% and then stays at about 320–324 K, close to pure IDM (about 320 K). The authors suggest the second Tg expected for a separated system "might result from" remixing of the domains during heating, and call the DSC limit of about 50% "not definitive". For HPMCAS, Tg falls to about 324 K at 30% and then varies little, supporting a limit of about 30%. For PVP, Tg falls almost linearly over the whole range. R13·M1·C31·C32·C33

Table B. Single Tg of the freshly made powders, mean of two, all values about (read from Figure 10, approximate); 0% is the neat polymer and 100% pure amorphous IDM. No HPMC point is visible at 90%. R13·M1·M2·M3·M4·X14

The authors summarize the four methods in their Table 2, reproduced as Table C. They see "good alignment" between techniques and conclude that fluorescence spectroscopy and imaging "appear to be useful techniques" for the phase behaviour of autofluorescent drugs. R21·C43

Table C. Miscibility of the three systems by four methods, as printed in Table 2 (the printed title names only HPMC and HPMCAS). R21·X3

Note by the AI writer (not discussed in the paper): the entries mix two conventions. For imaging, "~50%" (HPMC) and "~20%" (HPMCAS) are the last homogeneous films, but "~70%" (PVP) is the first heterogeneous one; the text gives the DSC limits as "around 50%" and "approximately 30%" where the table gives ranges, and the IR limit for PVP as "at least 80%". Read per loading, all methods for HPMC and HPMCAS place the change within one 10% step, while for PVP the spectra (80%), images (70%), IR (above 80%) and DSC (none) disagree more. X4·R12·R13

Main result 4: HPMC and HPMCAS powders above the limit crystallize within days. Table D gives the crystallization onsets from Table 1. HPMC powders at 10–50% stayed X-ray amorphous for the whole 14 months; at 60% Bragg peaks appeared by day 15, and at 70–90% by day 6. HPMCAS powders at 10–30% stayed amorphous for 13 months; at 40–90% peaks appeared within 2–10 days. The authors conclude that IDM–HPMC powders are more stable than IDM–HPMCAS powders ("demonstrate"). R14·R15·R17·C34

Table D. Crystallization onset of powders stored at 40 °C over desiccant (Table 1); a = amorphous for the whole study (14 months for HPMC, 13 months for HPMCAS). R17

Each onset is the first scan with Bragg peaks, so it is bounded by the scan before: 10 and 15 days at 60% HPMC, 4 and 6 days at 70–90% HPMC, and between day 0 and the first scan at 2 days for 80% HPMCAS. X9·P8 The crystals in the 40% HPMCAS powder give a few peaks (about 11.6, 16.7, 21.8 and 26.6° 2θ), while those at 80–90% give a fuller pattern from about 8.5° upward (read from Figure 12); the authors do not identify the crystal forms. R15 Note by the AI writer (not discussed in the paper): the 10% and 30% HPMC traces in Figure 11 carry one sharp reflection near 31.7° 2θ from about 1–2 months onward; it is not mentioned, and these samples are reported as amorphous. R14·X13

Main result 5: PVP dispersions resisted crystallization. No PVP powder showed Bragg peaks over the 3 months it was followed (Figure 13), so the authors suggest PVP dispersions "may have the best physical stability", in particular at high loading. R16·C35 In films, PVP showed no crystals from 5 to 70% within 30 days, the best of the three polymers; the text says crystals appeared at higher loadings "after longer time periods (30 days)", although Figure 16 shows crystals in the 90% film at 12 days and faint clusters in the 95% film already at 5 days. R20·C37·X10 The authors suggest the difference from XRD arises because fluorescence microscopy "may be more sensitive" to small amounts of crystal, especially crystals that form at the film surface (Wu and Yu 2006, ref 36). C38

Films in storage. In HPMC and HPMCAS films (Figures 14 and 15), no crystals appeared from 10 to 40% within 26 days, and crystals appeared in all films of higher loading. HPMCAS films crystallized earlier and more heavily: many crystals at 5 days in the 60% and 80% films, against none visible at 5 days in HPMC films below 95%. The authors call this "consistent with the XRD results". R18·R19·C36 Note by the AI writer (not discussed in the paper): film and powder thresholds differ by one step. The 50% HPMC film had crystals by 12 days while the 50% powder stayed amorphous for 14 months; the 40% HPMCAS film had none at 26 days while the 40% powder crystallized by 10 days. X8

The correlation. Figure 17 lays the onsets under bars that mark each polymer's amorphous, miscibility and crystallization ranges: for HPMC the miscibility band spans about 50–60%, for HPMCAS about 20–30%. The authors write that ASDs "determined to be miscible have excellent stability to crystallization", while those with a detected drug-rich phase "undergo crystallization at short time periods"; in the abstract this is "a good correlation". R22·C53 Table E sets the thresholds side by side.

Table E. First loading at which each signal appears (from Figures 3–8 and 14–16 and Table 1). R2·R3·R4·R7·R8·R9·R16·R17·R18·R19·R20

For HPMC the match is exact: 50% is both the last loading without a drug-rich band and the last amorphous powder, and 60% the first with both. For HPMCAS the first loading with a drug-rich band by spectra, images and DSC (30%) stayed amorphous for 13 months, and crystallization began at 40%. For PVP, powders at loadings with a drug-rich band (80% and 90%) did not crystallize within 3 months. R2·R17·R3·R8·R13·R4·R16·X5·X6

Discussion

The causal chain the authors propose. The paper's argument can be drawn as one chain that splits at the miscibility limit. Steps 1 and 2 are measured in this paper; the branch outcomes are measured as a correlation across loadings; the mechanism inside each branch (how much polymer the drug-rich phase holds, where crystals start) is taken from the literature and not measured here. C4·C39·C57

The authors' own words in the conclusion are that the formation of drug-rich domains "led to an enhanced susceptibility to crystallization", while the abstract says the correlation is "suggesting that miscibility is indeed an important ASDs characteristic". C57·C53 They draw two practical messages: the correlation "highlights the need to evaluate the homogeneity of different formulations during formulation screening", and the study "reinforces the importance of formulating homogeneous drug-polymer dispersions". They single out HPMCAS and, "to some extent", HPMC, where the IDM-rich phase forms at loadings relevant to commercial ASDs (20–30% and 50–60%). C55·C58·C54

What sets the limit. The paper offers one explicit explanation: the high IDM–PVP limit "can be attributed to" strong drug–polymer interaction, seen in the IR as new bands at 1724 and 1640 cm^-1 and the loss of the dimer band up to 80%. HPMC also disrupts the dimers at low loading. No reason is given for the low HPMCAS limit, and interaction strength was not compared quantitatively across the three polymers. C29·R12·C26·C56

What the 480 nm band really reports. The authors are explicit that the molecular origin of the spectral change "has not been elucidated": it "could be due to" more IDM dimers or other IDM–IDM contacts forming with the drug-rich phase. C46 The question matters most for PVP. Earlier work established that PVP breaks IDM dimers through hydrogen bonds (refs 32, 38) and found IDM–PVP miscible at about 5 nm by ¹H T_1ρ relaxation for 10–50 wt% polymer (Yuan 2015, ref 38). The present spectra and images instead suggest phase separation at 70–80% IDM. R23·C47·C48 The authors give two hedged explanations. "One possible explanation" is preparation: miscibility can depend on solvent and method (refs 30, 39). "Another explanation might be" that the emission is very sensitive to IDM dimers, so the 480 nm band reports the balance of drug–polymer and drug–drug interactions rather than a separate phase; but this "does not explain" the heterogeneous images, and "Clearly more investigations are warranted". C49·C50

Where the approach applies. The method needs a drug whose emission depends on its surroundings; the authors frame it as useful for "drugs which exhibit autofluorescence". C43 Within this study it located the limit for two cellulose polymers in agreement with imaging and DSC, and it detected a change in PVP that DSC did not. R21 Note by the AI writer (not discussed in the paper): because the spectra respond to the local environment of each molecule, they may react to dimer formation even in a single phase; agreement with the images is therefore what supports the phase-separation reading, and for PVP that agreement is only partial (images change at 70%, spectra at 80%). X12·R9·R4

For formulation. Note by the AI writer (not discussed in the paper): for IDM the polymer moved the highest homogeneous loading threefold, from 20–30% in HPMCAS to 70% or more in PVP. In dry storage the detected limit acted as a conservative ceiling: below it nothing crystallized in 13–14 months, and in HPMCAS the first phase-separated loading still stayed amorphous. The fluorescence screen is fast, but it was run on drop-cast films while stability was tested on powders made differently, and the thresholds of the two sample types differ by one step; a screen should be run on material made as the product will be. Humidity, which lowers miscibility in many systems, was excluded on purpose, so these thresholds are an upper bound for humid storage. R21·R17·X5·X7·X8

Limitations

Limitations stated by the authors. The authors name four. Fluorescence imaging cannot resolve domains smaller than about 300–500 nm. C44 The molecular origin of the spectral change is unknown, and dimer formation without phase separation remains a possible explanation. C46·C50 Their DSC reading of the HPMC limit is "not definitive", since the expected second Tg was not seen. C32·C31 And the IDM–PVP results disagree with earlier ssNMR at high loading, for reasons not tested. C48·C49

The AI writer's own judgement (1): the correlation and its exceptions. The correlation rests on two polymers with nine loadings each. For HPMC it is exact, but for HPMCAS the 30% powder, which had a drug-rich band in the spectra, heterogeneous films and a Tg plateau, stayed X-ray amorphous for 13 months; crystallization started one step higher, at 40%. X5·R3·R8·R13·R17 For PVP the general statement is not tested in bulk: powders at 80% and 90%, where the spectra show a drug-rich band, did not crystallize, but they were followed for only 3 months against 13–14 months for the cellulose polymers. X6·R16 The abstract's wording ("samples with a drug loading of less than 50%" stayed amorphous; "higher than 50%" crystallized) also leaves the threshold loadings themselves out; Table 1 shows that 50% HPMC and 30% HPMCAS were among the amorphous samples. X16 The causal step, that crystals form in the drug-rich domains because they hold less polymer, was not measured: no phase compositions, nucleation sites or same-loading comparison with and without phase separation are reported. C57·C4

The AI writer's own judgement (2): the measurements. Miscibility was judged mostly on drop-cast films dried at about 20 °C, stability mostly on powders rotary evaporated at 60 °C and cryomilled; the film and powder thresholds differ by one step in both cellulose polymers. X7·X8 Most numbers in this article are read from plots, and the threshold image judgements are visual, with no domain areas or crystal counts. X12 Several textual slips appear: the methods state that fresh and annealed films gave similar spectra "indicating that these annealing conditions did change the miscibility" (a missing "not"); the XRD current is printed in mV; the Figure 2 caption and the Table 2 title omit PVP; and the fluorescence is attributed to a "tryptophan ring" although IDM contains an indole ring. X1·X2·X3·X11 Onsets are intervals set by the sampling schedule, not exact times. X9 None of these slips changes the ranking of the polymers or the main correlation for HPMC.

Conclusion

The paper makes three contributions. First, it shows that indomethacin's own emission shifts from about 510–518 nm in a polymer to about 476–488 nm when an IDM-rich environment forms, and uses this to place the miscibility limit at 20–30% (HPMCAS), 50–60% (HPMC) and 70–80% (PVP). R1·R2·R3·R4·C56 Second, it shows that imaging, IR and DSC place the limits in the same order, within about one 10% step for the cellulose polymers. R21·C43 Third, it shows that in dry storage at 40 °C, HPMC and HPMCAS powders crystallized within 2–15 days at or one step above the limit and stayed amorphous for 13–14 months below it, while no PVP powder crystallized within 3 months; the authors read this as "a good correlation" and conclude that drug-rich domains "led to" faster crystallization. R17·R16·C53·C57

Within its scope (one drug, three polymers, dry storage, films for miscibility and powders for stability) the study gives a simple screen and a clear practical rule: keep the loading below the detected miscibility limit. What it leaves open is the meaning of the 480 nm band (separate phase or dimers), the mechanism linking the drug-rich phase to crystallization, and whether PVP dispersions with a drug-rich band crystallize over longer storage or at humidity. C46·C50·C57·X6