Do simulated intestinal fluids tell us how fast a supersaturated drug crystallizes in the human gut?

Background and motivation

Many new drug candidates dissolve poorly in water. Supersaturating formulations, such as amorphous solid dispersions (ASDs), lipid-based formulations, nanocrystals, cosolvent systems and prodrugs, raise the concentration in the intestinal lumen above the crystalline solubility; a weakly basic drug moving from the acidic stomach into the small intestine can supersaturate on its own. Because a supersaturated solution has a higher chemical potential, it can drive more drug across the gut wall, provided the drug permeates well. Prolonged supersaturation has been linked to better absorption of abiraterone, celecoxib, danazol, halofantrine, paclitaxel and posaconazole, and for fenofibrate even "transient" supersaturation supported absorptive flux for up to an hour (Introduction; refs 1–18). C1

The catch is that a supersaturated solution is unstable or metastable. Sooner or later it crystallizes and falls back to the crystalline solubility, so the benefit lasts only as long as crystallization can be held off for a physiologically relevant time (refs 19, 20). Crystallization from solution has two stages, nucleation and growth: a critical nucleus must form against an energy barrier, and the new crystals then grow and consume the supersaturation. C2

Formulators add polymers and surfactants to control this. Polymers delay the onset of nucleation or slow crystal growth by adsorbing to crystal surfaces, while surfactants delay or promote crystallization depending on the surfactant and the drug (refs 22–29). The dissolution medium matters as much as the excipients. Biorelevant fasted-state media contain bile salts and phospholipids, the endogenous surfactants of the gut, and earlier work from the Taylor group showed that bile salts inhibit nucleation to different extents depending on the structures of both bile salt and drug, that monomeric bile salts can inhibit crystallization while micellar ones are less effective, and that increasing amounts of lecithin induced crystallization (Chen 2015, ref 33; Li 2016, ref 34; Lu 2017, ref 35; Indulkar 2018, ref 36). C3·C4

Commercial simulated intestinal fluids are simple: they contain sodium taurocholate (STC) and lecithin at set concentrations. Human intestinal fluid (HIF) is not. In fluid aspirated from 20 healthy volunteers, taurocholate made up on average only 12% of the total bile salt, next to taurodeoxycholate, taurochenodeoxycholate, glycocholate, glycodeoxycholate and glycochenodeoxycholate (Riethorst 2016, ref 39), and bile salts are present with lecithin and its hydrolysis products as micelles and mixed micelles. C5·R29 The authors had shown earlier, for these same two drugs and media, that composition changes crystalline and amorphous solubility, micellar partitioning and membrane transport (Elkhabaz 2019, ref 43). C6 What had not been studied much was whether simulated media reproduce the kinetics of crystallization in human fluid.

The paper therefore asks a practical question: if a formulator measures how long a supersaturated solution survives in a buffer or a simulated fluid, does that number say anything about human intestinal fluid? The authors studied two poorly soluble weak bases (BCS class II), atazanavir (p$K_a$ 4.52) and posaconazole (p$K_a$ 3.6 and 4.6), in six fasted-state media including pooled aspirated human fluid. They measured nucleation induction times, nucleation and growth rates by second harmonic generation (SHG) microscopy, and the morphology of the crystals by scanning electron microscopy (SEM). M1·M2·A1

Background concepts

Amorphous solubility and supersaturation

When drug is added to an aqueous medium faster than it can crystallize, the dissolved concentration can rise far above the crystalline solubility until it reaches the amorphous solubility, the concentration at which the drug separates as a disordered, drug-rich phase. This is the highest supersaturation a solution can reach, and every experiment in this paper started there. In precentrifuged FaHIF at 37 °C, the crystalline and amorphous solubilities were 6.6 ± 0.3 and 100 ± 5 μg/mL for atazanavir and 2.4 ± 0.4 and 13.0 ± 0.3 μg/mL for posaconazole (Table 1, n = 3). R1

As its measure of driving force, the paper uses the activity-based supersaturation ratio taken from the earlier study (ref 43); its definition is not printed here. (Note by the AI writer: in micellar media a plain concentration ratio also counts drug held in micelles, which is why an activity-based ratio is the better measure of driving force.) At the amorphous solubility this ratio was 54 ± 2 for atazanavir in all six media within experimental variation, and 36 ± 1 for posaconazole in all media except composite-FaSSIF and FaHIF, where it averaged 30 ± 1. R4·E8 The concentration ratios in FaHIF from Table 1 are much smaller, about 15 for atazanavir (100/6.6) and about 5.4 for posaconazole (13.0/2.4), which shows how much the two definitions differ in a solubilizing medium. X14

Induction time

The nucleation induction time is the time from creating a supersaturated solution until crystallization is detected. The paper defines it as the sum of two times (Eq. 1):

In plain terms: the measured induction time $T_{\mathrm{i}\mathrm{n}\mathrm{d}}$ is the time $T_{\mathrm{c}\mathrm{r}\mathrm{i}\mathrm{t}}$ needed for a critical nucleus to form plus the time $T_{\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}}$ for the new crystals to grow large enough to be detected. The two parts were not measured separately; the paper reads the total as the onset of nucleation. E1

Classical nucleation theory

The authors use classical nucleation theory to discuss why media differ. Eq. 2 is printed as:

Here $J$ is the solution nucleation rate, $A$ a pre-exponential kinetic parameter, ${\nu }_o$ the solute molecular volume, $\gamma$ the interfacial tension between the crystal nucleus and the solution, $k_B$ the Boltzmann constant, $T$ the temperature and $S$ the supersaturation. In plain terms: the exponential term is the nucleation barrier, which grows steeply with $\gamma$ and shrinks as $S$ rises, so a lower interfacial tension or a higher supersaturation gives faster nucleation. E2 The printed exponent has no minus sign; with the sign as printed, $J$ would rise with $\gamma$, which contradicts the text's own statement that $\gamma$ is a barrier and that lowering it increases the nucleation rate. The equation should be read with a negative exponent, as the text uses it. X1·C20

The pre-exponential factor is expanded in Eq. 3:

In plain terms: $A$ is the product of $f$, the rate at which monomers attach to the critical nucleus (by diffusion through the solution or transfer across the interface), $Z$, the Zeldovich factor (in the paper's words, the probability of formation or dissolution of the crystal nucleus), and $n$, the number density of molecules in solution. Anything that hinders attachment lowers $A$ and so the nucleation rate, independently of the barrier. E3·C22

Bile salts, lecithin and simulated intestinal fluids

Bile salts are the surfactants of bile; the prefixes tauro- and glyco- name their taurine and glycine conjugates. In intestinal fluid they occur as micelles and as mixed micelles with phospholipids such as lecithin (phosphatidylcholine) and its hydrolysis products (refs 41, 42). C5 The commercial fasted-state simulated intestinal fluids used here are FaSSIF-V1 (3 mM STC and 0.75 mM lecithin in sodium phosphate buffer) and FaSSIF-V2 (3 mM STC and 0.2 mM lecithin in maleic acid buffer), both at pH 6.5. The authors also made composite-FaSSIF, which keeps the 3 mM total bile salt and 0.2 mM lecithin but splits the bile salt among six species in human-fluid proportions: STC 12% (0.36 mM), sodium taurochenodeoxycholate (STCDC) 12% (0.36 mM), sodium taurodeoxycholate (STDC) 6% (0.18 mM), sodium glycocholate (SGC) 28% (0.84 mM), sodium glycochenodeoxycholate (SGCDC) 27% (0.81 mM) and sodium glycodeoxycholate (SGDC) 15% (0.45 mM) (Figure 2). P1 The blank buffers of the two commercial fluids, buffer-V1 (phosphate) and buffer-V2 (maleic acid), serve as surfactant-free controls. P1

SHG microscopy

Second harmonic generation is a nonlinear optical effect in which intense laser light at one wavelength (here 1064 nm) produces light at half the wavelength (532 nm). It arises only in materials without a centre of symmetry, such as crystals of chiral molecules. Both drugs have four chiral centres, so their crystals light up while dissolved or amorphous drug stays dark. M1·M2·R11 The instrument used, SONICC (second-order nonlinear imaging of chiral crystals), gives images with low background and high contrast, from which crystals can be counted and sized over time; the smallest detectable crystalline domain was estimated earlier to be about 90 nm (ref 47). P9

Crystal growth and step pinning

After nucleation, crystals grow by adding molecules at steps on their faces. Additives that adsorb on a face can slow or block growth. In the step-pinning model of Cabrera and Vermilyea (ref 67), adsorbed impurities act as obstacles that pin advancing steps, which then bow between the pinning points and slow down; how well an additive works depends on the supersaturation, how much of it adsorbs, and the spacing of adsorbed molecules compared with the critical nucleus radius. Additives can stop growth, slow it, or change the crystal habit (refs 64–68). C37 Because different faces bind additives differently, a medium can change not only how fast a crystal grows but also its shape, for example the aspect ratio (length divided by width) of a needle. E4·C45

Methods in detail

The study has one design idea: start both drugs at the same, maximal supersaturation (their amorphous solubility in each medium) and change only the medium, in steps of increasing complexity, from blank buffer to commercial simulated fluid to a six-bile-salt fluid to pooled human fluid, each without and with the polymer HPMC-AS. Everything was done at 37 °C with stirring at 300 rpm. A1·A3·P7·C8

Contribution 1: six media and a pooled human fluid at one starting supersaturation

The problem. To compare media fairly, the driving force for crystallization should be the same in each. Solubilizing media raise both the crystalline and the amorphous solubility, so a fixed drug concentration would mean a different supersaturation in each medium. C6

What they did. Every run started at the amorphous solubility of the drug in that medium, which the earlier study had shown corresponds to the same activity-based supersaturation in all media for atazanavir (54 ± 2) and nearly the same for posaconazole (36 ± 1, or 30 ± 1 in composite-FaSSIF and FaHIF). R4·C8 Starting concentrations for the five buffers and simulated fluids come from that study (ref 43) and are not printed here. X15 FaSSIF-V1 was prepared from commercial FaSSIF/FeSSIF/FaSSGF powder in sodium phosphate buffer and FaSSIF-V2 from its own powder in maleic acid buffer; composite-FaSSIF used the phosphate buffer, the six bile salts weighed in, and lecithin dissolved in dichloromethane, which was evaporated by stirring at 500 rpm and 50 °C for 30 min (dichloromethane always below 1% of the volume). All of these media were at pH 6.5. P1

The human fluid (FaHIF) came from an earlier study at University Hospitals Leuven: duodenal fluid was aspirated near the ligament of Treitz every 10 min for 90 min from 20 healthy volunteers (10 men, 10 women, 18–31 years) after an overnight fast of at least 12 h and 250 mL of water, pooled per volunteer, and 16 of these volunteer pools were mixed into one pool and frozen at −80 °C. The pool had pH 7.3. P2·M13 Before use it was centrifuged at low speed (3000 rpm, 2 min). This removed visible particles, made the fluid clear enough for UV measurements, and removed a trace SHG signal seen in native FaHIF that would otherwise have given false positives. P3·R3 Crystalline solubility (about 2 mg drug in 1 mL, 48 h at 37 °C) and amorphous solubility (solvent shift in the presence of 10 μg/mL HPMC-AS for atazanavir or 100 μg/mL for posaconazole) were remeasured in the precentrifuged fluid, after ultracentrifugation at 50 000 rpm (135700g) for 30 min at 37 °C, by HPLC with UV detection at 247 nm (atazanavir) or fluorescence detection (posaconazole, excitation 240 nm, emission 385 nm). P4·P5·P6 The values did not differ significantly from earlier values in uncentrifuged FaHIF (two-tailed t test, p > 0.05), which the authors read as precentrifugation leaving the fluid's solubilizing capacity unchanged ("This suggests"). R2·C7

How it differs from earlier work. Earlier studies compared single bile salts or simulated fluids with each other (refs 33–36); this one adds pooled human fluid and a composite fluid designed to match its bile salt profile, and measures nucleation and growth kinetics side by side in all of them. C5·C6

Contribution 2: induction time and desupersaturation from an in situ probe

What they did. Atazanavir in DMSO (50 mg/mL) or posaconazole in methanol (10 mg/mL) was spiked into 4 mL of medium in an 8 mL glass vial at 37 °C and 300 rpm to reach the amorphous solubility. A fiber-optic UV dip probe followed absorbance at 280 nm (atazanavir) or 260 nm (posaconazole). In buffer-V2 and FaSSIF-V2, maleic acid interferes with UV absorbance, so light scattering at 400 nm, a wavelength where the drugs do not absorb, was followed instead. The probe path length was 0.5 cm for atazanavir and 1 cm for posaconazole, except 0.2 cm in FaHIF. HPMC-AS was added at 0 or 10 μg/mL for atazanavir and at 0, 10 or 50 μg/mL for posaconazole; each condition was run in triplicate for up to 480 min. P7

Figure S1 shows how the profiles were read, for atazanavir in buffer-V1. The apparent concentration starts at about 70 μg/mL, rises to about 82 μg/mL within about 20–25 min, then declines slowly along a line labelled L1. At about 100 min the slope changes and the concentration falls steeply along a second line, L2, to about 15 μg/mL by about 155 min, levelling off at about 6–8 μg/mL from about 175 min. R5 The induction time is the point where the absorbance changes slope at the same time as the scattering at a nonabsorbing wavelength changes; the desupersaturation rate is the negative slope of L2, taken where L2 was linear over 5–10 time points. Such a rate could not be measured for every sample and is reported for buffer-V1, FaSSIF-V1 and composite-FaSSIF only. P7·P8·E7 Read from Figure S1, L2 falls from about 65 to about 15 μg/mL between about 105 and 155 min, about 1.0 μg/mL·min, consistent with the 1.0 ± 0.4 μg/mL·min in Table 2. E7·R8

Contribution 3: nucleation and growth rates from SHG images

What they did. Supersaturated solutions were prepared as above, and small aliquots were taken at time points chosen from the onset of nucleation in each medium and pipetted onto glass slides. A SONICC instrument (1064 nm laser, 51 MHz, 166 fs pulses; SHG collected at 532 nm in transmission through a 10× objective, NA 0.25; 350 mW; 894 ms exposure) imaged fields of 1925 μm × 1925 μm in several focal planes, at least three fields per sample. All media gave no SHG background before drug was added. P9·R3 ImageJ counted and sized the SHG particles (circularity 0.00–1.00, single pixels excluded), averaged over fields of view; the change in count per minute was taken as the nucleation rate and the change in size per minute as the growth rate (n = 3). P10·E5·E6

Limits of the measurement. The methods define the nucleation rate as particles formed per unit time per unit volume, but the reported quantity is SHG signal count per minute per field, and the aliquot volume and imaged depth are not printed; the growth rate is a change in detected signal size, limited by the roughly 90 nm detection limit and the 10× objective. X13 Atazanavir with polymer was not imaged because its induction times were too long, and posaconazole with 50 μg/mL HPMC-AS gave signals too low for reliable analysis. R14·P10

Contribution 4: crystal habit and form

What they did. Solutions were kept past their induction time at 37 °C and 300 rpm; aliquots were centrifuged (14 800 rpm, 10 min), and the precipitate was coated with platinum and imaged by SEM at 5 kV. P11 The length and width of atazanavir needles gave the aspect ratio (n = 15 per medium, n = 10 in FaSSIF-V1), compared by two-tailed t tests; the sampling time for these needles is not stated. P12·X15 PXRD (Cu Kα, 5–50° 2θ) compared atazanavir precipitates from buffer-V1, FaSSIF-V1 and FaHIF with the starting material, with starting material slurried in acetonitrile, and with the crystal structure LISTEP in the Cambridge Structural Database (Figure S4). P13 No PXRD of posaconazole precipitates is reported. X16

Experimental analysis

Induction times without polymer: human fluid longest, FaSSIF-V1 shortest

Table A. Nucleation induction times at 37 °C, starting at the amorphous solubility, no polymer; means of n = 3 read from Figures 3 and 4 (approximate) unless the text gives the value. R6·R7·P1

Atazanavir is a slow nucleator, in agreement with earlier work (ref 36). In the two blank buffers it began to crystallize after about 90–100 min. It was fastest in FaSSIF-V1 and somewhat slower in FaSSIF-V2 (72 min), and it held supersaturation longest in FaHIF (∼140 min), with composite-FaSSIF close behind (∼130 min). R6·C9·C10 One number in the text does not match the figure: the text puts the FaSSIF-V1 induction time "around 60 min", while the bar in Figure 3 reads about 49 min with an error bar of roughly 44–55 min. X2

Posaconazole crystallized much faster, within minutes in every medium. Again the blank buffers held supersaturation longer than their FaSSIF counterparts (the authors summarise this as about 2-fold), FaHIF was longest, and the authors give FaHIF and composite-FaSSIF as 6-fold and 3-fold longer than FaSSIF-V1. R7·C11 So for both drugs the extremes are the same: the commercial FaSSIF-V1 shortest, human fluid longest. R6·R7

Two details matter when reading Table A. First, the detection mode was not the same in all media: UV absorbance in four media but 400 nm scattering in buffer-V2 and FaSSIF-V2, and a shorter probe path (0.2 cm) in FaHIF; FaHIF was also at pH 7.3 while all other media were at pH 6.5. X12 Second, no statistical test was reported for induction times, and the atazanavir error bars for FaHIF and composite-FaSSIF overlap in Figure 3. X11

Lecithin and bile-salt diversity

FaSSIF-V1 and FaSSIF-V2 contain the same 3 mM taurocholate and differ in lecithin, 0.75 versus 0.2 mM. The higher-lecithin FaSSIF-V1 gave the shorter induction time for both drugs (atazanavir about 49 versus 72 min; posaconazole about 3 versus about 4 min), and the authors conclude that a higher lecithin concentration promoted more rapid crystallization, "as reported previously". R6.b·R6.d·R7.b·R7.d·C24 The two commercial fluids also differ in buffer (phosphate versus maleic acid) and in how induction was detected, so this comparison is not lecithin alone; the blank buffers themselves differ, for example in posaconazole growth rate (about 112 versus 156 nm/min, Figure 9). X19·R15.a·R15.c

Composite-FaSSIF tests bile-salt diversity at the same total bile salt and lecithin as FaSSIF-V2. For atazanavir it lengthened induction from 72 to ∼130 min, close to FaHIF; for posaconazole it did not reach FaHIF (about 9 versus about 18 min). R6.d·R6.e·R6.f·R7.e·R7.f·C27 Earlier work had found 2–3-fold longer induction times in composite-FaSSIF than in FaSSIF-V1 for atazanavir, ritonavir and tacrolimus but none for cilnidipine (ref 36), and the authors conclude that "the impact of media composition may vary from compound to compound". R30·C28

HPMC-AS: effective everywhere, most in buffers

Table B. Induction times (min) with HPMC-AS, read from Figures 3 and 4 (approximate); ">480" means no nucleation within the 480 min run. For atazanavir in FaHIF one of three runs did not nucleate within 480 min, so the mean of the other two is a lower bound. R6·R7·X17

With 10 μg/mL HPMC-AS, atazanavir stayed supersaturated beyond 8 h in both buffers and in composite-FaSSIF, and FaHIF again exceeded the commercial simulated fluids. R6·C10 For posaconazole, 10 μg/mL gave only minor increases, whereas 50 μg/mL extended the induction time substantially, most in the buffers; FaHIF was about 2-fold longer than the simulated fluids, which the text places between 35 and 42 min (Figure 4 reads about 43 for FaSSIF-V2). R7·C11 The authors call HPMC-AS an effective nucleation and growth inhibitor "even when present a low concentration", with a clear concentration dependence for posaconazole. C31

The polymer's relative effect was largest in the blank buffers. The authors read this as "This suggests that polymer−solute interactions are maximized in the absence of solubilizing species": in the presence of bile salts and lecithin, the polymer's affinity for the drug "may be reduced" by competing interactions. C32 They single out the extension of supersaturation in FaHIF as "perhaps the most significant finding" with HPMC-AS, in line with earlier observations in human fluid (ref 62). C33

Desupersaturation after nucleation

Table C. Desupersaturation rates (μg/mL·min, mean ± SD, n = 3), from Table 2 of the paper. R8

Once crystals had formed, atazanavir lost supersaturation fastest in FaSSIF-V1 (4.2 μg/mL·min) and slowest in composite-FaSSIF (0.3 μg/mL·min). Posaconazole behaved differently: its desupersaturation was notably slower in buffer-V1 (0.6) than in the media with solubilizing species (1.9 and 1.7). HPMC-AS at 50 μg/mL slowed posaconazole desupersaturation most in buffer (0.6 to 0.10 μg/mL·min), less in FaSSIF-V1 and little in composite-FaSSIF. R8·C12

What SHG imaging adds: nucleation and growth rates

The SHG images match the induction times. At 90 min, atazanavir showed similar amounts of crystals in buffer-V1 and FaSSIF-V1 (Figure 5A, B) and, according to the text, in buffer-V2 and FaSSIF-V2 (Figure S2); at 120 min composite-FaSSIF and FaHIF showed only trace crystallinity (Figure 5C, D). R9·R10·C13 In Figure S2 the FaSSIF-V2 image shows visibly more small spots than the buffer-V2 image, although the nucleation rates of the two media are similar. X20 Posaconazole at 16 min gave many SHG spots in the FaSSIFs and larger domains in the blank buffers (Figure 6); in composite-FaSSIF the count rose notably by 24 min, while in FaHIF only trace spots appeared up to 40 min (Figure 7; the text cites this as "Figure 6C,D"). R11·R12·C14·X4

Table D. Nucleation rate (SHG counts/min) and growth rate (nm/min) from SHG images, n = 3; read from Figures 8 and 9 (approximate; Figure 8 has a broken axis). HPMC-AS at 10 μg/mL. R14·R15

For atazanavir, the nucleation rate was similar in the buffers and FaSSIFs and fell about 4-fold in FaHIF and 16-fold in composite-FaSSIF compared with FaSSIF-V1. Posaconazole varied far more, from about 48 counts/min in FaSSIF-V1 to about 0.06 in FaHIF, which the authors read as "suggesting slow crystallization kinetics in this medium". With 10 μg/mL HPMC-AS the posaconazole rates dropped notably. R14·C15 Growth was fastest in the buffers for both drugs; atazanavir grew slowest in FaSSIF-V1 and posaconazole in FaHIF, and HPMC-AS suppressed posaconazole growth in all media, in three of them below what could be measured. R15·C16

Two statements in the discussion stretch these numbers. The posaconazole nucleation rate is called "somewhat similar for all of the simulated media", but the values span about 6.5 to 48 counts/min; and composite-FaSSIF is called similar to FaHIF for atazanavir with reference to Figures 3 and 8, which holds for induction time (about 131 versus 141 min) but not for the SHG nucleation rate (about 0.1 versus 0.5). X7·C27 The statement that the nucleation rate was higher in FaSSIF than in buffer holds for induction times and for posaconazole in FaSSIF-V1, but the posaconazole SHG rate in FaSSIF-V2 (about 6.5) is below that in buffer-V2 (about 9). X6·C21

Crystal habit and form

All atazanavir precipitates were needles. In buffer-V1 at 150 min, small needles grew on the surface of what appeared to be amorphous atazanavir; by 300 min well-defined needles of variable size had formed. FaHIF showed small needle bundles at 180 min and well-developed needles at 300 min (Figure 10). R16·C17 At higher magnification, needles from buffer-V1 were wide (about 1 μm) with secondary growth and surface imperfections, while needles from FaHIF were narrow (about 188, 303 and 192 nm in the annotated examples) with smooth ends and surfaces and no branching (Figure 11). Needles from FaSSIF-V1 and composite-FaSSIF branched like those from buffer but were narrower (Figure 12). R17·R18

Table E. Aspect ratio of atazanavir needles, from Table 3 of the paper. All pairs differ significantly (two-tailed t test, p < 0.05) except FaSSIF-V1 versus FaHIF. R21

With 10 μg/mL HPMC-AS, atazanavir still formed needles, and the authors report no major change in aspect ratio, although the crystals were less uniform, especially in composite-FaSSIF (Figure 13); no aspect-ratio values with polymer are printed. R19·C48·X10 Atazanavir precipitated in a different crystal form from the material supplied: the bricklike starting material had PXRD peaks at 4.4, 9.8 and a 14.8/15.1° 2θ doublet that were absent in the acicular precipitates, which showed peaks at 6.7, 10.6, 11.9, 12.8, 14.2 and 19° 2θ. Starting material slurried in acetonitrile converted to the same pattern, consistent with CSD LISTEP, and the authors "can assume" that the precipitate is the more stable polymorph. R20·C18

Posaconazole precipitated as very thin plates in every medium: smooth plates of different sizes in buffer-V1 (annotated thickness about 35–45 nm), plates with striations and irregular ends in FaSSIF-V1 (Figure 14), similar plates in composite-FaSSIF (annotated thickness about 25–30 nm), and in FaHIF thin plates with irregular, "finger-like" ends (Figure 15). In Figure 15C some particles are circled as "Possible FaHIF components". R22·R23·C19 All crystals were small (<5 μm). R24

In vitro onset versus human data

The authors compared two of their FaHIF results with earlier human studies. Noxafil delayed-release tablets, a posaconazole–HPMC-AS amorphous solid dispersion, gave a mean supersaturation of 93 min in the proximal jejunum of fasted volunteers (ref 17). In FaHIF with 50 μg/mL HPMC-AS (a drug-to-polymer ratio of about 1:3.8 at the amorphous solubility of 13.0 μg/mL), posaconazole nucleated after about 80 min, which the authors call "reasonably close". R26·R28·R7.f·C34 One Reyataz capsule gave duodenal atazanavir supersaturation for 1.25 h (ref 63), against about 2 h in vitro (Figure 3: about 141 min in FaHIF without polymer); the authors call this "not an exact match" and attribute the difference "likely" to the Reyataz excipients, a higher dose and the local crystallization environment in vivo. R27·R6.f·C34·C35 Together, they write, these comparisons "provide some assurance" that measuring supersaturation duration this way "may be a useful approach to approximate in vivo supersaturation timeframes". C36

Discussion

Why do surfactant-rich media differ? The nucleation puzzle

Because every run started at the amorphous solubility, the authors argue that differences in induction time "cannot be attributed" to differences in driving force; for one molecule, the only exponential factor in Eq. 2 that can change between media is the interfacial tension $\gamma$. C8·C20 Surface-active species that adsorb on a nucleus should lower $\gamma$ and therefore speed up nucleation. That fits FaSSIF versus buffer, but not FaHIF: human fluid also contains bile salts and phospholipids, yet it slowed nucleation. Bile salts had been seen to delay nucleation before (refs 33, 34), while other surfactants are enhancers (ref 52). C21

To reconcile this, the authors turn to the pre-exponential factor (Eq. 3). It has been suggested that molecules adsorbed at the nucleus–solution interface can slow monomer attachment, because an adsorbed molecule must leave before a solute molecule can join the nucleus, which adds a desolvation barrier (ref 51). A surfactant therefore pulls in two directions, lowering $\gamma$ but also lowering $f$, and "competing effects of adsorbed surface-active molecules on the interfacial tension and the monomer attachment rate may explain" why surfactants differ in their effect on nucleation. C22 Neither $\gamma$ nor $f$ was measured. C22

Note by the AI writer (from the card's extractor notes): the equal-driving-force premise holds for atazanavir but not exactly for posaconazole, whose cited activity-based supersaturation was 30 ± 1 in composite-FaSSIF and FaHIF against 36 ± 1 in the other media, that is, lower in exactly the two media where posaconazole held supersaturation longest. Part of the longer posaconazole induction times there may come from driving force rather than from the medium's components. X5·R4

How lecithin and bile-salt mixtures push in opposite directions

For lecithin, the authors cite two mechanisms that "had been suggested" in earlier work (ref 55): lecithin forms mixed micelles with taurocholate and other bile salts at a lower critical micelle concentration, which leaves fewer monomeric bile salts free to inhibit crystallization, and lecithin adsorbs to the emerging nucleus and promotes nucleation by lowering the interfacial tension. Lecithin also raises crystalline and amorphous solubility through extra solubilization (as seen for posaconazole in ref 43), so it helps solubility while it shortens the life of the supersaturated state. C25 The authors add a caveat for physiological relevance: most phospholipid in human fluid is present as lysolecithin rather than lecithin, which "can potentially impact" crystallization differently. C26·R32

For bile-salt diversity, the authors note that bile salts are not equivalent as crystallization inhibitors, structure and hydrophobicity being determinants (Li et al., ref 34), and that composite-FaSSIF, although designed to match FaHIF's bile salt profile, lacks minor components: glyco- and tauroursodeoxycholate (about 1.6% of the total bile salt, insoluble at pH 6.5), cholesterol, fatty acids and monoacylglycerides. The aggregation state of the bile salts and the structure of the micelles (mono- versus mixed micelles) "could also" play a part. C28·C29·R29 They conclude that the interplay between medium composition, drug properties and crystallization kinetics "requires additional studies in order to optimize simulated intestinal fluids". C30

Polymer and media together

HPMC-AS inhibited crystallization in every medium, but its relative benefit was largest in the blank buffers. The authors read this as competition: polymer–drug interactions are greatest without solubilizing species, and bile salts and lecithin may reduce the polymer's affinity for the drug, noting that polymer–surfactant interactions are "common but extremely complex". C32 For growth, polymer adsorption at the crystal–water interface is generally taken to depend on the polymer's affinity for the solid relative to the medium (ref 70). The authors see the further growth reduction in FaHIF with polymer as a sign that "the polymer is still able to interact with the crystal surface" despite polymer–biological surfactant interactions, and compare it with earlier work in which HPMC-AS slowed felodipine growth by surface poisoning while sodium dodecyl sulfate weakened, but did not prevent, its adsorption (Schram 2016, ref 26). C42·C43·C44 Figure 9 marks the posaconazole growth rate with polymer in FaHIF as not quantifiable, so this "further reduction" is an absence of measurable growth rather than a measured rate. X9·R15.f

Growth and crystal shape

The authors read the morphology through the step-pinning model. Bile salts in the media cut posaconazole growth "4−8-fold" and changed crystal morphology, as seen earlier for telaprevir in bile-salt media (ref 69). C38 The finger-like protrusions at the ends of posaconazole crystals grown in bile-salt media, especially FaHIF, "suggests that step pinning may be occurring", which would give curved growth steps. C39 FaHIF gave the most altered morphology and the slowest growth, "suggesting that FaHIF contains additional components not found in the simulated media" that strongly affect growth, with the simulated media in between. C40 The fold change depends on which pair of media is compared: read from Figure 9, buffer-V1 versus FaSSIF-V1 is only about 2.3-fold (112 versus 48 nm/min), while buffer-V2 versus FaSSIF-V2 is about 5.8-fold and buffer-V2 versus FaHIF about 7.8-fold. X8·R15

For atazanavir, high-aspect-ratio needles mean growth mainly at the needle ends, so additives that adsorb on particular faces could change the aspect ratio. C45 Needles from buffer-V1 had the lowest aspect ratio and showed secondary nucleation, and needles from simulated media and FaHIF were longer relative to their width. The authors "hypothesize that bile salts suppress crystal growth on the faces forming the sides of the needle, rather than at the ends", and see the smooth, unbranched FaHIF needles as "supporting the conjecture" that bile salts or other components adsorb on the side faces. C46·C47 Adsorption on individual faces was not measured. C47 The small posaconazole plates (<5 μm, with a high surface-area-to-volume ratio) are, they suggest, likely to redissolve readily. C41

What it means in vivo

The authors frame the stakes as follows: crystallization is thermodynamically favoured, so retarding desupersaturation, which gives a larger area under the concentration–time curve, is what keeps a formulation's advantage. C49 In vitro crystallization kinetics are of interest as input to physiologically based absorption models, and earlier studies suggested that in vitro tests overestimate precipitation compared with in vivo (ref 71); "this discrepancy may arise due to the use of simulated media", in which the authors "clearly show" crystallization to be faster than in pooled FaHIF. C50 Other factors in vivo include an absorptive compartment, which can lower local supersaturation of highly permeable drugs (refs 72, 73), and hydrodynamics unlike a well-stirred vial. C51 Since nucleation and growth are highly sensitive to medium composition, variation in intestinal fluid between and within subjects "might also impact desupersaturation kinetics" and so add variability to absorption. C52

The authors also bound the relevance of their result. For drugs that crystallize very fast, media differences are unlikely to matter, and for drugs that take several hours to crystallize no practical consequences are expected; but "for drugs with moderate nucleation induction times (longer than a few minutes), such variations could be significant". They also point to effects not captured by current models: residual supersaturation when growth stalls in a "dead zone" in the presence of additives (ref 69), and the size and morphology of the crystals, which affect redissolution. C53·C54

Note by the AI writer (not discussed in the paper): in these data FaSSIF-V1, the medium with the most lecithin, gave the earliest crystallization of both drugs, and blank buffer gave the largest apparent benefit of HPMC-AS. A precipitation-inhibitor screen run only in blank buffer, or a supersaturation test run only in FaSSIF-V1, would give a skewed picture in opposite directions: the first flatters the polymer, the second is pessimistic about the drug. Neither simulated fluid tracked FaHIF for both drugs.

Limitations

Limits the authors state

The authors acknowledge that composite-FaSSIF lacks several minor components of human fluid and may differ in micellar structure C29; that phospholipid in human fluid is mainly lysolecithin rather than the lecithin used in simulated media C26; that variation of human fluid composition between individuals needs further study C59·C52; that absorption and hydrodynamics in vivo may change crystallization kinetics C51; that media effects matter mainly for drugs with moderate induction times C53; and that residual supersaturation, crystal size and morphology are largely unexplored C54. They base the conclusion that none of the common media predicts crystallization times well explicitly on "these two compounds" ("it could be argued"). C23

The AI writer's own judgement

The ranking is consistent, the statistics are thin. Triplicate induction times rank FaHIF longest and FaSSIF-V1 shortest for both drugs, and the posaconazole SHG rates agree in direction. But no statistical test is reported for induction times, nucleation or growth rates, so the abstract's "significantly reduced" is descriptive; tests are reported only for solubility and aspect ratio. X11·R2·R21

The media were not measured identically. Detection switched from UV absorbance to 400 nm scattering in the two maleic acid media, the probe path length was shorter in FaHIF, and FaHIF was frozen, precentrifuged and at pH 7.3 while the simulated media were at pH 6.5. The lecithin comparison (FaSSIF-V1 versus FaSSIF-V2) is also a comparison of phosphate with maleic acid buffer and of two detection modes. X12·X19

SHG rates are relative. They are counts and detected sizes per field of view, without normalization to sample volume, and are best compared within this paper. X13

Some wording runs ahead of the data. The conclusions say commercial media "appeared to overestimate the onset times of nucleation" relative to FaHIF, but the measured onset times were shorter in the commercial media; the discussion's version, that crystallization kinetics "might be overestimated" in FaSSIFs, is the one the data support. X3·C56·C23 The "4−8-fold" growth reduction, "somewhat similar" posaconazole nucleation rates and the absence of aspect-ratio changes with polymer each hold only in part or without printed numbers. X8·X7·X10

The mechanisms are inferred. Interfacial tension, monomer attachment, face-specific adsorption, step pinning and polymer–bile salt competition are offered as explanations ("may explain", "suggests", "we hypothesize") without direct measurement. The posaconazole crystal form was not determined. C22·C39·C47·C32·X16

The in vivo comparison is approximate. It pairs one in vitro value per drug with one cited in vivo value, measured with different formulations, doses and sampling sites; it supports agreement in order of magnitude, not a validated correlation. X18

Scope. Two weakly basic drugs, one HPMC-AS grade at two concentrations, and one pooled sample of human fluid. M1·M2·M3·P2

Conclusion

This paper put a practical question to the standard tools of supersaturation testing. Started at the same high supersaturation, atazanavir and posaconazole crystallized latest in pooled aspirated human intestinal fluid and earliest in the commercial FaSSIF-V1, and human fluid also gave the slowest posaconazole nucleation and growth. R6·R7·R14·R15·C55 More lecithin sped crystallization, a six-bile-salt medium brought atazanavir, but not posaconazole, close to human fluid, and HPMC-AS delayed crystallization in every medium, most in blank buffer and still in human fluid, where posaconazole with polymer held supersaturation for about 80 min, near the 93 min measured in human jejunum after an HPMC-AS dispersion. C24·C27·C31·C32·C33·C34 Bile-salt media also changed crystal habit, which the authors attribute to adsorption on particular faces and step pinning. C38·C39·C47

The authors' conclusion is cautious in wording but clear in direction: buffers and current simulated fluids "appear to require further verification" as surrogates when crystallization kinetics matter, the most biorelevant composition and concentration of endogenous surfactants should be included in in vitro media, and structurally diverse drugs should be tested in simulated and aspirated fluids. C58·C59 The AI writer's reading for a formulation scientist: for these two drugs, the medium is part of the experiment: a supersaturation lifetime measured in FaSSIF-V1 or in buffer is a number for that medium, not yet a prediction for the gut.