How lecithin and the bile salt mix of simulated intestinal fluid change crystallization from supersaturated drug solutions

Background and motivation

Many new drugs are poorly soluble in water, and supersaturating dosage forms are one way around this. Amorphous solid dispersions, lipid-based formulations, salts and cocrystals, prodrugs and the gastric-to-intestinal pH shift of weak bases can all produce a drug concentration above the crystalline solubility in the gut (Introduction, refs 1–9). The authors set this against solubilization by micellar surfactants or cyclodextrins: a solubilizer raises the crystalline solubility but lowers the drug's thermodynamic activity, whereas a supersaturated solution with the same total drug concentration carries more free drug. Because transport across a membrane follows activity, not total concentration, a supersaturating system can outperform a solubilizing one, provided the drug does not crystallize first (refs 10–14).

Crystallization is the price of supersaturation. A supersaturated solution has a driving force to crystallize, and when it does the concentration falls and the advantage is lost (refs 15, 16). Additives change the outcome: polymers tend to inhibit crystallization, the most effective ones balancing hydrophilic and hydrophobic groups and carrying bulky side groups (Ilevbare et al., ref 19), while common surfactants such as sodium dodecyl sulfate and polysorbate 80 can induce it, through heterogeneous nucleation or by lowering the interfacial energy between solute and the emerging crystal (refs 20, 21). The composition of the medium matters as well (refs 22–24).

The gut fluid itself is one such medium. Human intestinal fluid (HIF) contains bile salts and phospholipids, mainly lecithin, that form micelles, mixed micelles and vesicles (refs 26–28). Biorelevant media try to mimic this. The commercial fasted-state simulated intestinal fluids, FaSSIF-V1 and FaSSIF-V2, both contain 3 mM sodium taurocholate (STC) as the only bile salt; they differ in lecithin (0.75 and 0.2 mM) and in buffer (phosphate and maleate). STC was chosen because its low pKa (1.8) keeps it dissolved across gastrointestinal pH (refs 29, 30). Riethorst et al. (ref 31) showed that real fasted HIF contains a mix: STC together with sodium glycocholate (SGC), glycodeoxycholate (SGDC), glycochenodeoxycholate (SGCDC), glycoursodeoxycholate (SGUDC), taurodeoxycholate (STDC), taurochenodeoxycholate (STCDC) and tauroursodeoxycholate (STUDC). SGUDC and STUDC make up only 1.3 and 0.6% of the total, and the median total bile salt concentration, 3.3 mM, is close to the STC level in FaSSIF. STC itself is only 12% of the bile salt in HIF.

Earlier work from the same group showed that bile salts matter for crystallization. Chen et al. found that STC delays or inhibits crystallization of several compounds (ref 32), and Li et al., testing thirteen bile salts, found that they differ in inhibitory power in a way "somewhat related" to their hydrophobicity (ref 33). From this the authors infer that a simulated fluid containing many bile salts "may" act differently from commercial SIF. How lecithin affects crystallization kinetics had not been studied. On the solubility side, Augustijns et al. reported that crystalline solubility in SIF and in HIF correlate well (R² = 0.85, ref 34), so crystalline solubility seems insensitive to which bile salts are present. For a supersaturating formulation, however, the amorphous solubility, the concentration in metastable equilibrium with the amorphous drug, is the more relevant ceiling, and its dependence on bile salt composition was not known.

The paper therefore has two aims: to find how (1) the lecithin content and (2) the bile salt composition of a simulated intestinal fluid affect solubility and crystallization kinetics in solution. For the second aim the authors built a new medium, Composite-SIF, from the six most common HIF bile salts in the mean proportions reported by Riethorst et al., with the FaSSIF-V1 lecithin level. Four structurally different, poorly soluble drugs were used: atazanavir, ritonavir, tacrolimus and cilnidipine (Figure 2). The comparisons of crystallization were made at the same supersaturation ratio in every medium, measured thermodynamically rather than inferred from concentration.

Background concepts

Crystalline and amorphous solubility. The crystalline solubility is the concentration in equilibrium with the stable crystal. The amorphous solubility is the highest concentration a supersaturated solution can reach: there the dissolved drug is in metastable equilibrium with the amorphous drug, and adding more drug produces liquid–liquid phase separation (LLPS) into a continuous solution at the amorphous solubility plus nanometre-sized drug-rich droplets (ref 35). The ratio of the two solubilities sets the largest supersaturation a drug can reach in a simple solution. In a micellar medium, however, both solubilities include drug held in micelles, so their ratio is no longer a clean measure of the driving force.

Supersaturation ratio as an activity ratio. The paper defines the supersaturation ratio (SR) through activities (Eq. 1):

In plain terms: $a$ is the activity of the drug in the solution and $a^{*}$ its activity in the reference state, which here is the crystal, so $a^{*}$ is the activity of a solution saturated with crystals. SR is how far the drug's chemical potential sits above that of the crystal, independent of how much drug is hidden in micelles.

Measuring activity by flux. Activity cannot be read from a concentration when micelles are present, but it can be read from transport across a membrane. Under sink conditions on the receiver side, the diffusive flux $J$ is proportional to the donor activity (Eq. 2, after Higuchi, ref 12):

In plain terms: $D$ is the diffusion coefficient of the drug, $h$ the membrane thickness and ${\gamma }_m$ the activity coefficient of the drug in the membrane; for a given drug and system these are constant, so the flux tracks the activity of the drug in the donor; in the authors' words, it reflects the free drug available for mass transport across the membrane. Combining Eqs. 1 and 2 gives Eq. 3:

In plain terms: $J^{*}$ is the flux measured with the donor at the crystalline solubility, so the supersaturation ratio is simply the ratio of two fluxes. Measured with the donor at the amorphous solubility, it is called $S{R}_{\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{h}}=J_{\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{h}}/J^{*}$, the largest supersaturation the medium allows. The flux itself is obtained from the rate of mass transfer per unit membrane area (Eq. 4):

In plain terms: $dm/dt$ is the rate at which drug appears in the receiver and $A$ the membrane area (7.07 cm² here); in practice the slope of receiver concentration against time, multiplied by the receiver volume and divided by $A$, gives $J$.

Nucleation induction time. The nucleation induction time (NIT) is the time from creating a supersaturated solution to the first detectable sign of crystallization. Here detection is a rise in light scattering, so the measured time contains two parts (Eq. 5):

In plain terms: $t_{\mathrm{n}\mathrm{u}\mathrm{c}\mathrm{l}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}}$ is the time for the first nuclei to form and $t_{\mathrm{g}\mathrm{r}\mathrm{o}\mathrm{w}\mathrm{t}\mathrm{h}}$ the time for them to grow large enough to scatter light measurably. A longer NIT means the supersaturation lasts longer.

Why the driving force has to be held equal. In the Discussion the authors recall the classical expression for the rate $J_N$ at which crystal nuclei form (Eq. 6, ref 58):

In plain terms: $v^{*}$ is the rate at which a molecule attaches to the nucleus, $z$ the Zeldovich factor, $n$ the number density of molecules in solution, $\gamma$ the interfacial tension between nucleus and solution, $v$ the molecular volume, $k_B$ the Boltzmann constant, $T$ the temperature and $S$ the supersaturation ratio. Because $S$ sits squared inside a logarithm in the denominator of the exponent, small changes in supersaturation change the nucleation rate enormously; any comparison of media must therefore hold $S$ fixed. The interfacial tension $\gamma$ enters as a cube, which is why an additive that lowers it, for example by adsorbing on the nucleus, can speed nucleation sharply. The paper uses the equation only qualitatively.Card check: the text does not define $T$, and describes $n$ as a number density "per unit time" (story note X13).

Bile salts and lecithin. Bile salts are steroid acids conjugated to glycine or taurine (Figure 1). Cholates (SGC, STC) carry three hydroxyl groups on the steroid; deoxycholates (SGDC, STDC) and chenodeoxycholates (SGCDC, STCDC) carry two, at different positions. The authors call these classes "trihydroxy", "dehydroxy" and "chenodehydroxy". Lecithin is a phospholipid that forms mixed micelles with bile salts. Table A gives the three media compared in the paper.

Table A. Composition of the biorelevant media (from Table I, bile salts as a share of total bile salt; rows reordered by class). Table I prints the 27% salt as "SGCGC"; the text and Figure 1 name it SGCDC.Card check: misprint in Table I (story note X1). The text also says that Figure 1 shows the structures of the eight HIF bile salts; it draws six.

The two-hydroxyl salts, chenodeoxycholates and deoxycholates together, make up 60% of Composite-SIF and none of FaSSIF. This single fact carries much of the paper's later explanation.

Methods in detail

The study has three building blocks: a set of media that vary one thing at a time, a thermodynamic measurement of supersaturation in each medium, and crystallization tests that start from that measured supersaturation. All work was at 37 °C in a pH 6.5 phosphate buffer of about 270 mOsmol.

Contribution 1: media that separate lecithin from bile salt identity

Problem. Commercial FaSSIF-V1 and V2 differ in lecithin and in buffer at the same time, and both contain only one bile salt. To see what lecithin does, and what the bile salt mix does, each has to be varied alone.

What they did. FaSSIF-V1 was made from the commercial FaSSIF/FeSSIF/FaSSGF powder in the pH 6.5 phosphate buffer. Diluting it with a 3 mM STC solution gave a series with 3 mM STC and 0.01, 0.025, 0.05, 0.075, 0.1, 0.2, 0.3, 0.4, 0.5 or 0.75 mM lecithin, plus 3 mM STC with no lecithin. The 0.2 mM member has the STC and lecithin amounts of FaSSIF-V2, but in phosphate rather than maleate buffer, so that buffer and ionic strength stay the same throughout. Composite-SIF was made by dissolving the six bile salts of Table A to 3 mM total; lecithin (0.75 mM) was dissolved in dichloromethane and added to the bile salt solution, and the turbid emulsion was stirred at 500 rpm and 50 °C for 30 min to evaporate the solvent, leaving a clear micellar solution with no smell of dichloromethane. A 0.2 mM lecithin version was made by diluting Composite-SIF with the 3 mM bile salt mix. To separate the bile salts, each of the six was also prepared alone at 3 mM with 0.75 mM lecithin by the same route. Micelle size in FaSSIF-V1 and Composite-SIF was measured by dynamic light scattering (Malvern Zetasizer Nano ZS, backscatter at 173°).

Why it works. Every medium has the same total bile salt (3 mM), the same buffer and, where compared, the same lecithin, so a difference between two media can be put down to the one factor that differs. Card check: the "FaSSIF-V2" of Figures 3, 5, 6 and 9 is this phosphate-buffer analogue, not the commercial medium (story note X11). The Composite-SIF shares sum to 100%, so the two minor HIF salts were left out; how the remainder was rescaled is not stated (story note X12).

Contribution 2: supersaturation measured by flux, not by concentration

Problem. In a micellar medium the total drug concentration includes drug inside micelles, so the ratio of amorphous to crystalline solubility does not give the true supersaturation. Earlier comparisons of FaSSIF and HIF (Bevernage et al., ref 49) inferred supersaturation from concentration, which the authors' own earlier work showed can be wrong when solubilization and supersaturation occur together (Indulkar et al., ref 11).

What they did. Crystalline solubility was measured by adding excess crystal to each medium, holding it at 37 °C for 24 h, filtering through 1 µm syringe filters (glass fibre for atazanavir, PTFE for the others) and assaying by HPLC (UV 210 nm, or 240 nm for cilnidipine; calibration 0.05 to 10 µg/mL, R² = 0.999). Amorphous solubility was measured by the solvent-shift method (ref 35): a stock of about 25 mg/mL drug in organic solvent was pumped into the stirred medium (300 rpm, 37 °C) while extinction at a non-absorbing wavelength was read every 10 s; the concentration at which scattering rose was taken as the amorphous solubility, and each run lasted under 10 min. Both were measured in triplicate in buffer, FaSSIF-V1 and Composite-SIF for all four drugs, and for atazanavir also across the lecithin series and in the single bile salts.

The supersaturation ratio at the amorphous solubility was then measured in a side-by-side diffusion cell (PermeGear). A regenerated cellulose membrane (Spectra/Por 1, molecular weight cut-off 6–8 kD, area 7.07 cm²) separated two 34 mL compartments of the same medium at 37 °C. The donor was dosed from a methanol stock to either the amorphous or the crystalline solubility, 200 µL samples were taken from the receiver over about 40 to 60 min, and the slope gave the flux; $S{R}_{\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{h}}=J_{\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{h}}/J^{*}$ (Eq. 3). No crystallization occurred during these runs.

Why it works, and its limit. The membrane sees only the activity of the dissolved drug (Eq. 2), so the flux ratio gives a supersaturation that micelles cannot inflate. For tacrolimus and cilnidipine, however, $J^{*}$ could not be measured: the authors cite the limits of their analytical method combined with slow diffusion. For these two drugs only $J_{\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{h}}$ was compared across media.

Contribution 3: crystallization tests started at equal driving force

Problem. Since nucleation depends steeply on supersaturation (Eq. 6), comparing induction times in different media at the same concentration would mix up two effects: the medium's effect on nucleation and its effect on the driving force.

What they did. A concentrated methanol solution of the drug was added to 20 mL of medium so that the drug concentration equalled the amorphous solubility of that medium. The solution was stirred at 300 rpm at 37 °C, extinction at a non-absorbing wavelength was read every minute with the fibre optic probe, and the NIT was taken as the time at which scattering rose above the noise. All four drugs were tested in buffer, FaSSIF-V1 and Composite-SIF; atazanavir alone was tested across the lecithin series, in the six single bile salts with lecithin, and in FaSSIF-V2 and Composite-SIF at 0.2 mM lecithin.

Why it works. Starting every test at the amorphous solubility places each medium at its own maximum supersaturation, and Contribution 2 shows that this maximum is the same in the main media (next section). The induction times can then be compared as properties of the medium. Card check: replicate numbers for the induction times are not stated, although the figures carry error bars (story note X8). No solid-state or microscopic check identifies the scattering particles as crystals (story note X14).

How it differs from earlier work. Li et al. (ref 33) compared bile salts as crystallization inhibitors; the authors point out that HIF also contains lecithin, so here lecithin is present throughout the bile salt comparison, and the media are compared at a supersaturation measured by flux rather than assumed from concentration.

Experimental analysis

The two media look different at the micelle level

Composite-SIF was a clear solution, FaSSIF-V1 slightly translucent. FaSSIF-V1 gave one population of micelles with a z-average of 49 ± 4 nm, in line with earlier reports (refs 37, 38). Composite-SIF gave two: 56% of the micelles at 63 ± 4 nm and 44% at 3.9 ± 0.4 nm. The authors expect "structures of varying sizes" from such a complex mixture; the two populations were not characterized further, and the basis of the percentages (intensity, volume or number) is not stated.

Result 1: lecithin moves the amorphous solubility of atazanavir, not its crystalline solubility

Figure 3 follows atazanavir in 3 mM STC as lecithin rises from 0 to 0.75 mM. The crystalline solubility stays at about 1 to 1.8 µg/mL throughout; STC, with or without lecithin, "does not seem to solubilize" the crystal. The amorphous solubility behaves differently: 65 µg/mL in plain buffer, 38 µg/mL in 3 mM STC alone, and then a steady climb with lecithin, to about 44 µg/mL at 0.01 mM, 57 at 0.025, 65 at 0.075, 72 at 0.2 and 82 to 83 µg/mL from 0.4 mM upwards. In FaSSIF-V1 the amorphous solubility is therefore only 1.3 times the buffer value. With 0.75 mM lecithin, the six single bile salts gave a similar picture (Figure 4): crystalline solubility about 0.9 to 1.2 µg/mL, amorphous solubility 82 to 93 µg/mL (about 81 µg/mL for SGDC up to 93 for SGC), which the authors express as 1.3 to 1.5 times buffer.Card check: the buffer bar in Figure 3 reads about 62 µg/mL, against 65 µg/mL in the text and Table II; the 1.3 to 1.5-fold range matches the figure's buffer value better than 65 (story note X2).

Result 2: FaSSIF-V1 solubilizes more than Composite-SIF, and the crystal more than the amorphous form

Table II compares the four drugs in the three main media; Table B reproduces it.

Table B. Crystalline and amorphous solubility, mean (SD), n = 3 (from Table II). Table II prints no unit; the text gives the atazanavir values in µg/mL.

Three patterns stand out. First, atazanavir hardly responds to either simulated fluid: its crystalline solubility stays at 1.1 to 1.3 µg/mL and its amorphous solubility rises only from 65 to 82 µg/mL in both. Second, for the other three drugs both fluids raise both solubilities, and FaSSIF-V1 raises them more than Composite-SIF. The authors write that FaSSIF values are "nearly 1.5 to 2 fold higher"; from the table the ratio is 1.37 and 1.47 for ritonavir (crystalline, amorphous), 1.35 and 1.31 for tacrolimus, and 1.89 and 1.78 for cilnidipine. They suggest the difference "could be attributed" to different extents of solubilization by the two kinds of bile salt–lecithin micelles.Card check: the stated 1.5 to 2-fold overstates the lower end, and there is no FaSSIF advantage for atazanavir (story note X3). Third, for ritonavir and cilnidipine the crystalline solubility gains more than the amorphous one: cilnidipine's crystalline solubility rises 40-fold (Composite-SIF) and 80-fold (FaSSIF-V1) but its amorphous solubility only 16 to 30-fold, and ritonavir's crystalline solubility 2 and 3-fold against 1.3 and 1.9-fold. Tacrolimus rises about 3 to 4-fold in both forms. The authors call this "consistent with" their earlier finding that micellar solubilization works differently at crystalline and amorphous concentrations (ref 11). As a result the amorphous-to-crystalline ratio, read naively as the maximum supersaturation, falls in the simulated fluids for ritonavir (12 to 8.4 and 7.8) and cilnidipine (36.5 to about 13).

Result 3: the true supersaturation at the amorphous solubility hardly changes

The flux measurements tell a different story from the concentration ratios (Table C).

Table C. Supersaturation ratio at the amorphous solubility, or flux at the amorphous solubility where J* could not be measured, mean (SD), n = 3 (from Table III).

For atazanavir, SRamorph is 63 to 65 in all three media, and for ritonavir 9.2 to 10. Yet ritonavir's concentration ratio ranged from 7.8 to 12. The authors draw the general lesson that "solubility or concentration values are not accurate measures of supersaturation" and cannot replace a measured SR. For tacrolimus and cilnidipine, $J_{\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{r}\mathrm{p}\mathrm{h}}$ was similar in buffer and in both simulated fluids; from this "it can be supposed" that these drugs do not mix with the bile salts and lecithin, that their solubility gain is "purely" micellar entrapment, and "it can be assumed" that $J^{*}$, and so SRamorph, is also equal across media. That last step is an assumption, not a measurement.

Across the lecithin series (Figure 5), atazanavir's SRamorph was about 63 in buffer but about 40 in 3 mM STC alone. Lecithin restored it: about 44 at 0.01 mM, 55 at 0.025, 60 at 0.05, and 62 to 66 from 0.075 to 0.75 mM, which the authors describe as "fairly constant". The six single bile salts with 0.75 mM lecithin all gave about 65, and the FaSSIF-V2 analogue and 0.2 mM Composite-SIF matched the 0.75 mM media; neither of these sets of values is shown.

Result 4: taurocholate alone delays crystallization, and lecithin undoes this

Figure 6 gives the induction time of atazanavir across the lecithin series (Table D).

Table D. Atazanavir supersaturation and induction time at the amorphous solubility of each medium (read from Figures 5 and 6, approximate; Figure 6 has a broken axis).

In plain buffer atazanavir crystallized after 170 min. In 3 mM STC alone crystallization was held off for up to about 600 min, so STC alone is, in the authors' words, "an effective crystallization inhibitor" of atazanavir. Lecithin "as low as" 0.01 mM diminished that effect, and from 0.025 mM onwards the induction time fell below the buffer value, reaching about 43 min in FaSSIF-V1: lecithin made the medium "worse than in buffer alone". The abstract summarizes that crystallization "was accelerated" as lecithin increased.

The supersaturation was not equal across this series: it was about 40 in STC alone and about 44 at 0.01 mM, against 60 to 66 from 0.05 mM upwards. The authors recognized this and repeated the measurement "at select lecithin concentrations where an equivalent SR was maintained", reporting that the NIT did not change with SR; these data are not shown and the lecithin levels used are not named.Card check: the long induction time in STC alone was measured at a lower supersaturation than buffer, and the control that addresses this is not shown (story note X5). Note by the AI writer: from 0.075 mM lecithin upwards, where SRamorph is nearly constant (about 62 to 66), the induction time still fell from about 98 to about 43 min, so in that range the shortening is not explained by a change in driving force.

Result 5: bile salts with two hydroxyl groups delay crystallization more than cholates

With 0.75 mM lecithin and SRamorph about 65 in every case, the six single bile salts gave very different induction times for atazanavir (Figure 7, Table E).

Table E. Atazanavir in single bile salts with lecithin (read from Figures 7 and 4, approximate). Buffer alone: about 170 min.

The ranking, "evident" to the authors, is chenodehydroxy above dehydroxy above trihydroxy, for glycine and taurine conjugates alike. Against plain buffer (about 170 min), the chenodeoxycholates leave the induction time unchanged, while SGC and STC with lecithin cut it to less than a third. The amorphous solubility does not follow the same order (SGC, with the shortest induction time, has the highest), and SRamorph was about 65 in all six media. Only atazanavir was tested in this way.

Result 6: Composite-SIF keeps three of four drugs supersaturated longer than FaSSIF-V1

The central comparison is Figure 8: each drug at its amorphous solubility in Composite-SIF and in FaSSIF-V1 (Table F). Figure 9 adds the lecithin level for atazanavir.

Table F. Induction times in minutes (SIF values read from Figures 8 and 9, approximate; buffer values from the text and the Figure 8 caption; the ratio column is computed by the AI writer from these readings).

For atazanavir, ritonavir and tacrolimus the induction time was longer in Composite-SIF, "readily apparent" in the authors' words, so supersaturation lasted longer there. For cilnidipine there was "no significant difference"; the authors offer that this "may" be because the different bile salts act alike on cilnidipine, or because its crystallization is governed by lecithin rather than by the bile salts. Figure 9 shows that the lecithin effect also holds within each medium type: less lecithin, longer induction time (Composite-SIF about 174 against 120 min; FaSSIF about 69 against 43 min), with Composite-SIF ahead at both levels. The abstract states the comparison was made "at equivalent supersaturation ratios", which was measured for atazanavir and ritonavir and assumed for tacrolimus.Card check: equal SR for tacrolimus and cilnidipine rests on similar Jamorph and an assumption (story note X4).

Note by the AI writer (not discussed in the paper). Set against plain buffer, both simulated fluids shortened the induction time of atazanavir (170 to about 120 and 43 min) and tacrolimus (340 ± 100 to about 101 and 53 min), whereas ritonavir in Composite-SIF and cilnidipine in both fluids crystallized later than in buffer. So the result is that Composite-SIF promotes crystallization less than FaSSIF-V1, not that it inhibits crystallization for every drug (story note X6).

Discussion

How the authors connect the results

The authors' explanation runs in two layers, one for solubility and one for crystallization. None of its steps was measured directly; the hedges below are theirs.

Step 1 rests on earlier work showing that a second component can mix into the amorphous drug-rich phase and lower its activity (Trasi and Taylor, ref 52). Step 2 is the authors' reading of why SRamorph stops rising at 0.075 mM: they write that the free drug concentration "increases from 38 µg/mL in the absence of lecithin to 65 µg/mL" at 0.075 mM, and that above this point the total concentration can exceed 65 µg/mL while the free concentration that sets the supersaturation stays at 65 µg/mL.Card check: 38 and about 65 µg/mL are the total amorphous solubilities of Figure 3; the free drug concentration itself was not measured (story note X9). Step 3 offers two routes joined by "and/or", and the authors do not choose between them. The branch at the bottom generalizes the atazanavir single-salt ranking to the other drugs.Card check: single bile salt induction times exist only for atazanavir (story note X10).

What the authors draw from it

On solubility, the authors make three points. The choice of biorelevant medium "can impact the solubility of the drug and these effects should be considered". Because the thermodynamic supersaturation at the amorphous solubility was equivalent across media while the concentrations were not, concentration is not a surrogate for the supersaturation ratio; for the same reason they ask that the earlier FaSSIF-V1 versus HIF crystallization correlation of Bevernage et al. be read cautiously ("inferences from this study need to be drawn cautiously"), since it inferred supersaturation from concentration. And because the medium shifts the amorphous solubility, "the same drug formulation may or may not undergo LLPS based on the choice of biorelevant medium used", which matters for formulation design, screening and selection, for in silico or pharmacokinetic modelling of the maximum absorptive flux, and for dissolution rates governed by the amorphous drug (refs 56, 57). They add that differences in lecithin content between people "can potentially result in inter-subject variability" ("it can be speculated").

On crystallization, they conclude that "the use of FaSSIF-V1 versus V2 to evaluate the crystallization propensity of a supersaturated solution can lead to different outcomes", that "a single bile salt may not be representative of HIF" for supersaturating systems, and that FaSSIF-V3, with SGC and STC in a 1:1 ratio plus lecithin, lysolecithin, sodium oleate and cholesterol (Fuchs et al., ref 48), "can be expected" to give kinetics similar to FaSSIF-V1. FaSSIF-V3 was not tested. Their next step, "Clearly", is to measure crystallization kinetics and amorphous solubility in real HIF and compare them with the simulated media.

What it suggests for formulation work

Note by the AI writer (not discussed in the paper). Three practical readings follow from the data, within the limits below. First, of the two simulated fluids, FaSSIF-V1 was the harsher crystallization test for three of the four drugs; a formulation that keeps its supersaturation in FaSSIF-V1 has passed the stricter of these two tests, though neither has been checked against human fluid. Second, crystallization-inhibitor screens in bile salt solutions without lecithin can mislead: STC alone held atazanavir supersaturated about 3.5 times longer than buffer, yet 0.025 mM lecithin or more turned the same medium into a crystallization promoter. Third, when a formulation is assessed by concentration in a micellar medium, the apparent maximum supersaturation (the amorphous-to-crystalline ratio) can differ from the true one; for ritonavir in Composite-SIF the ratio was 7.8 while the measured SRamorph was 10. Activity-based measurements, such as membrane flux, transfer between media; concentrations do not.

Limitations

Limits the authors state. The authors did not test HIF and name a "careful evaluation" of crystallization kinetics and amorphous solubility in real intestinal fluid as "Clearly the next step". They note the analytical limit that prevented $J^{*}$ for tacrolimus and cilnidipine, and they present the equal SRamorph of those drugs as something that "can be assumed". Their mechanisms are hedged throughout ("can be attributed", "can be due to", "possibly", "may"), and their conclusion is that crystallization tendency in a given medium "may or may not correlate well with in vivo performance".

The AI writer's own judgement.

• Replicates and controls. No replicate number is given for the induction times, and Figure 5 has no error bars. Several supporting results are reported as "data not shown" or without values: the equal-SR induction times in the lecithin series, SRamorph in each single bile salt, and SRamorph in the 0.2 mM lecithin media (story notes X8, X5).
• Driving force in the lecithin series. The longest induction time (STC alone, about 595 min) was measured at an SRamorph of about 40, against about 63 in buffer, so part of that delay could reflect the lower driving force; the authors' control for this is not shown (story note X5).
• One drug carries the mechanism. The lecithin series and the single bile salt ranking used atazanavir only, yet the explanation of the Composite-SIF effect is applied to ritonavir and tacrolimus as well (story note X10).
• Relative, not absolute, inhibition. Composite-SIF beats FaSSIF-V1, but for atazanavir and tacrolimus both simulated fluids gave shorter induction times than plain buffer (story note X6).
• What the scattering is. The induction time is the first rise in light scattering; no X-ray diffraction or microscopy confirmed that the particles were crystals (story note X14).
• Reporting details. Table II has no unit, Table I misprints SGCDC as "SGCGC", the Figure 3 buffer value (about 62 µg/mL) differs from the 65 µg/mL in the text, and the FaSSIF advantage in solubility is 1.3 to 1.9-fold rather than "nearly 1.5 to 2 fold" (story notes X7, X1, X2, X3). None of these changes a conclusion.
• Scope. One pH (6.5), one temperature (37 °C), one total bile salt level (3 mM), fasted-state media only, drugs added from methanol stocks, and no polymer or other formulation excipient in the crystallization tests.

Overall the AI writer rates the evidence as moderate: the solubility and flux data are direct and internally consistent, and the key comparison is made at a measured equal supersaturation for two drugs, but the crystallization data lack stated replicates, several controls are not shown, and the mechanisms remain hypotheses.

Conclusion

The paper makes two experimental points and one methodological one. Lecithin, at fixed taurocholate, leaves the crystalline solubility of atazanavir unchanged, restores and raises its amorphous solubility, and shortens the time before it crystallizes from a supersaturated solution, cancelling the inhibition that taurocholate alone provides. Replacing taurocholate with the six-bile-salt mix of human intestinal fluid, at the same total bile salt and lecithin, lengthens the induction time of atazanavir, ritonavir and tacrolimus about two to three-fold, and the single bile salt tests with atazanavir point to the two-hydroxyl chenodeoxy and deoxy salts as the inhibitors. Throughout, supersaturation was measured as an activity ratio by membrane flux, which stayed nearly constant across media while the concentration ratios did not. The authors' summary is that "variations in crystallization tendency can be observed and these, may or may not correlate well with in vivo performance", and that media composition deserves close attention when supersaturating dosage forms are evaluated.