Which bile salts are in the test medium? How biorelevant media composition, lecithin and a polymer change the crystallization of supersaturated telaprevir

Background and motivation

Many oral drug candidates are poorly soluble, and a common way to raise their absorption is to create supersaturation in the gut, either by dissolving the drug first (cosolvent or lipid formulations) or by giving a high-energy solid (an amorphous solid dispersion, a cocrystal or a salt) (Brouwers 2009, ref 4). Supersaturation raises the chemical potential of the dissolved drug, which drives membrane transport, but the same excess chemical potential makes crystallization favourable; crystallization in vivo is expected to remove the gain (Taylor and Zhang 2016, ref 7). C1 Formulators therefore add crystallization inhibitors, most often polymers. C2

The intestine supplies inhibitors of its own. Earlier work from the same group and collaborators showed that bile salts are, in many cases, effective crystallization inhibitors: as monomers they slow nucleation and/or crystal growth of poorly soluble drugs about as well as polymers (Chen 2015, ref 11; Li 2016, ref 12; Lu 2017, refs 13 and 14). C2·R11·R12 For telaprevir in particular, all tested biorelevant bile salts were "extremely good inhibitors" as monomers, whereas their micelles acted differently depending on the bile salt; sodium taurochenodeoxycholate (STCDC) was an extremely effective growth inhibitor and sodium taurodeoxycholate (STDC) one of the better nucleation inhibitors. R11·R12

In the upper intestine, however, bile salts are not alone: they form mixed micelles with phospholipids and cholesterol (Hofmann 1987, ref 15). C3 Biorelevant media used for in vitro testing try to mimic these fluids, yet many commercial recipes, including fasted state simulated intestinal fluid (FaSSIF), contain sodium taurocholate (STC) as their only bile salt, although the biologically relevant bile salts are known to differ both thermodynamically (Lu 2017, ref 16) and as crystallization inhibitors (refs 12, 13). C4 The authors therefore ask whether current biorelevant media "might be oversimplified" for testing supersaturating dosage forms. C4

The paper sets out to measure how the composition of the test medium changes crystallization kinetics of supersaturated telaprevir, chosen because the effect of single bile salts on this drug had already been characterized for thermodynamics and for nucleation and growth inhibition. C4·R11·R12 It does so in three steps: binary and ternary bile salt mixtures; full media with lecithin, comparing FaSSIF with media containing six human bile salts; and the same media with the polymer hydroxypropyl methyl cellulose acetate succinate (HPMCAS-MF), to ask whether "the presence of an effective polymeric crystallization inhibitor override[s] the impact of testing media differences". C33

Background concepts

Supersaturation measured by flux. In a plain buffer, supersaturation is simply concentration divided by crystalline solubility. In a medium with micelles this fails, because part of the drug sits in micelles and does not contribute to the driving force for crystallization. The authors therefore measure supersaturation through the drug's thermodynamic activity, using a side-by-side diffusion cell with a regenerated cellulose membrane (molecular weight cutoff 6–8 kDa): only free drug crosses the membrane, and its flow is proportional to its activity. P1·M11·C5 The paper's Equation (1) defines the supersaturation ratio $\delta$:

Here $F$ is the solute mass flow rate across the membrane in the solution of interest, $F^{\circ }$ the mass flow rate of the standard-state system (a solution in equilibrium with the crystalline drug), and $a$ and $a^{\circ }$ the corresponding solute activities. E1 In plain terms: the drug flow from a test medium, divided by the flow from a saturated solution, tells how far the free drug is above saturation, whatever the micelles do to the total concentration. The calibration for telaprevir in buffer came from the group's earlier study (ref 16). E1·R1

Induction time. The authors follow crystallization by the time from creating supersaturation to the detectable onset of crystals, the induction time $t_{ind}$. They write it, as is common practice, as the sum of two parts (printed as Equation (3); the manuscript has no Equation (2)):

Here $t_n$ is the nucleation time for forming a critical nucleus and $t_g$ the growth period until crystals are large enough to detect. E2·C7·X1 In plain terms: a long induction time can mean slow nucleation, slow growth to detectable size, or both; the measurement alone does not separate them. The paper's conclusion nevertheless speaks of "nucleation-induction times". C32

Growth-inhibition ratio. To look at growth alone, the authors seed a supersaturated solution and follow the initial drop in concentration. Equation (4) compares the rates without and with bile salt:

Here $R_0$ and $R$ are the desupersaturation rates (initial slopes over the first 10 minutes) in the absence and presence of bile salt. E3·P3 In plain terms: $E_G$ = 1 means the additive does nothing to bulk growth, and $E_G$ = 10 means growth is ten times slower. This reading assumes that bulk growth on the seeds dominates the loss of supersaturation; the authors note that secondary nucleation "can occur" at the supersaturation used. C8

Monomers, micelles and mixed micelles. Bile salts are surfactants: below a critical concentration they are dissolved as single molecules (monomers); above it they assemble into micelles, which can take up a hydrophobic drug. Mixtures of bile salts form mixed micelles, and the phospholipid lecithin enters bile salt micelles as well. C42 The paper chose total concentrations for its bile salt mixtures so that the low level (1.86 mM) would give monomers and the high level (3.72 or 5.58 mM) micelles. P5·C9 Critical micelle concentrations are not printed in the manuscript, and micelle formation in the mixtures was not measured directly. C17

Amorphous solubility and glass–liquid phase separation. Above a certain concentration a supersaturated solution of telaprevir separates into a dispersed, glassy drug-rich phase and a drug-lean solution; the solution then stays at the amorphous solubility, the highest supersaturation reachable without crystals. For telaprevir this corresponds to $\delta$ = 28.8 (ref 13), and the authors call this glass–liquid phase separation (GLPS) (Mosquera-Giraldo and Taylor 2015, ref 30). M1·C6·C35

The media. Table 1 of the paper lists every medium. FaSSIF contains 3 mM STC with 0.75 mM lecithin. The authors' mixed micelle with lecithin solutions (MWL-1 and MWL-2) contain 3 mM of six bile salts with the same 0.75 mM lecithin: the taurine and glycine conjugates of cholate (STC, SGC), deoxycholate (STDC, SGDC) and chenodeoxycholate (STCDC, SGCDC). P5 STC is the only trihydroxy bile salt among those tested singly; STDC and STCDC are dihydroxy. M3·M5·M7

Methods in detail

The study is a sequence of induction-time experiments at constant driving force, in media of increasing complexity, with one growth experiment and one morphology survey alongside. The flow below shows the stages.

Step 1: the same driving force in every medium

Problem. Comparing crystallization in buffer and in micellar media at the same total concentration would be unfair: micelles take up drug, lower its activity, and so lower the driving force. Independent experiments are needed in each medium to set comparable driving forces (Raina 2015, ref 17). C5

What they did. For each medium the authors measured the telaprevir mass flow rate across the membrane at several concentrations and converted it to $\delta$ with Equation (1), using the buffer data of ref 16 as the calibration (Figure 2). P1·R1·E1 Two levels were then chosen: $\delta$ = 11.5 for a homogeneous supersaturated solution, used for all bile salt mixtures and media without polymer, and $\delta$ = 28.8, the amorphous solubility, where a glassy telaprevir-rich second phase is present, used for the experiments with polymer. C6 The manuscript does not state the temperature, stirring, volumes or number of runs of the diffusion-cell measurements. P1

What Figure 2 shows. In buffer, $\delta$ rises from about 1.6 at 5 µg/mL to about 28.0 at 150 µg/mL and then stays near 27 up to 170 µg/mL; the authors mark $\delta$ = 28.8 at about 150 µg/mL. R1.a·R1.h For $\delta$ = 11.5 the arrows mark about 60 µg/mL, where buffer, FaSSIF and the MWL media lie, and about 72 and 77 µg/mL, where the dotted lines of binary-H and ternary-H cross. R1.h In words, the authors report no significant change of mass flow rate with the low (1.86 mM) mixtures and a decrease with the high mixtures, "consistent with the presence of solubilizing micelles". R2·C10 Note by the AI writer (not discussed in the paper): the per-medium concentrations are given only as unlabelled arrows, and the asterisk markers of the low mixtures at about 90 µg/mL plot at $\delta$ of about 14.3–14.8, below the buffer series (about 17.5), which does not fit the "no change" reading in the text; the overlapping markers cannot be told apart at the figure's resolution. X2·X3

Step 2: induction times and seeded growth

Induction times. Supersaturation was created by solvent shift: a small aliquot of a 12 mg/mL methanolic telaprevir stock was titrated into 50 mL of medium at 37 °C, stirred at 300 rpm. P2 A fibre-optic UV dip probe (1 cm or 0.5 cm path) followed the signal at 270 nm, where telaprevir absorbs, and at 370 nm, where it does not; the induction time was the onset of a sudden change in the signal. P2 Runs were followed for 16 hours, so bars marked ">960" mean no crystallization within the observation. P2·X5 Pairwise differences were tested by t-test (p < 0.05) and the media with polymer by one-way analysis of variance; the number of runs and the type of error bar are not stated. P6·X5

Seeded growth. Seeds were produced in the vessel by the in situ "common history" seeding method of ref 14, so that every run starts from comparable crystals. P3 The bile salt mixture was then added and allowed to equilibrate, and supersaturation was regenerated to $\delta$ = 7.9 with the methanolic stock. P3 Full UV spectra (200–450 nm) were taken every minute; to suppress scattering by seed crystals and micelles, second-derivative spectra were used for calibration and samples (linear standard curve, R² > 0.99), and the initial slope over 10 minutes at 270 nm gave the desupersaturation rate $R$. P3 $E_G$ followed from Equation (4). E3

Step 3: media that mimic human bile

The binary mixtures combined STDC and STCDC at a 1:1 molar ratio, chosen because STCDC is an extremely effective growth inhibitor and STDC one of the better nucleation inhibitors; the ternary mixtures added STC, chosen because it inhibits telaprevir crystallization even in the presence of micelles. C9·P5 For the full media, the authors combined the six most abundant bile salts of human intestinal fluid with lecithin, keeping 3 mM bile salt and 0.75 mM lecithin "for direct comparison with the FaSSIF recipe". C23 MWL-1 follows a reported human duodenal composition (ref 22), with cholate and chenodeoxycholate salts predominant; MWL-2 shifts the balance to cholate and deoxycholate salts, which "appear to be" more effective nucleation inhibitors for telaprevir, to test whether more of them prolongs supersaturation. C23 M-2 has the MWL-2 bile salts without lecithin. P5 All media were made in 50 mM pH 6.5 sodium phosphate buffer, and the lecithin solutions were stirred for 2 days before use. P5·M12

Table A. Compositions of the test media (from Table 1 of the paper). The basis of the MWL percentages, molar or by mass, is not stated. P5·X13

Note by the AI writer (not discussed in the paper): the 0.75 mM lecithin of FaSSIF and the MWL media lies above the jejunal range listed in Table 1 and inside the duodenal range, which matters because lecithin turned out to change the bile salt effect. X12

Morphology. After the induction-time runs, about 10 µL of each crystallized solution was dried on a glass coverslip in a vacuum oven for at least a week, sputter-coated with platinum and imaged by field-emission SEM at 5 kV and 5000–40000× magnification. P4

Experimental analysis

Bile salt mixtures: strong as monomers, weak as micelles, no synergy

In buffer at $\delta$ = 11.5, telaprevir crystallized in about 30 minutes (Figures 3 and 4). R3.a·R4.a Every bile salt system at 1.86 mM total, single STC, STDC or STCDC, binary-L and ternary-L, kept the solution supersaturated for the whole 16-hour observation. R3.b-d·R4.b-e At micellar levels the picture changed. The binary mixture at 3.72 mM (binary-H) gave about 257 minutes and the ternary mixture at 5.58 mM (ternary-H) about 254 minutes, against about 137–140 minutes for 12 mM STDC, about 179–180 minutes for 12 mM STCDC, and more than 960 minutes for 12 mM STC. R3·R4

Table B. Induction times of telaprevir with single bile salts and their mixtures at δ = 11.5, 37 °C (read from Figures 3 and 4, approximate). R3·R4

The authors read these results through the monomer–micelle distinction. Binary-L "appears to be a mixture of monomers, which are extremely effective crystallization inhibitors", while binary-H contains mixed micelles and is less effective; the flux and induction-time results "suggest" that STC, STDC and STCDC are monomers at 1.86 mM and form mixed micelles at 5.58 mM. C13·C17 They find "no evidence of synergistic interactions" between two bile salts: "in general" the mixtures act like their single components. C14 Single STC still inhibited beyond 16 hours as micelles, whereas the micelles of the dihydroxy bile salts "strongly" negate the inhibition of their monomers. C15 In the ternary micelles the STC effect is lost: ternary-H was only slightly better than the dihydroxy micellar solutions and "considerably lower" than STC alone, which the authors describe as an "intermediate effect" in which "the inhibitory impact of STC is reduced by the presence of the other bile salts". C16·C18

Note by the AI writer (not discussed in the paper): the micellar mixtures are compared with single bile salts at 12 mM, not at the mixtures' 3.72 or 5.58 mM, and the manuscript does not say whether the single-salt bars were re-measured or taken from refs 13 and 14. X4 The "comparable" monomer effects are all above the 960-minute limit, so differences beyond 16 hours could not be seen. X6

Seeded growth: an intermediate effect for the monomer mixture

In seeded solutions at $\delta$ = 7.9 (Figure 5), monomeric STDC (1.86 mM) slowed growth modestly, with $E_G$ of about 3.8, while the bar for monomeric STCDC runs off the top of the axis, so its $E_G$ is above 60 but not printed. R5.a-b Binary-L gave about 35.5, with a wide error bar of about 16.7–54.5, and binary-H about 7.4. R5.c-d The authors call binary-L intermediate between its components and note that doubling the concentration decreased the inhibition. C19 They consider three ways in which two additives may act at a crystal surface: independent adsorption, competition for sites, or adsorption as a complex (refs 18–21). C20 Based on the data, they "speculate" that both monomeric bile salts adsorb, with different effectiveness, which gives the intermediate value, and that at higher concentration mixed micelles "serve to counteract the inhibitory effect of the monomers". C21·C22 Adsorption was not measured. C21

Table C. Growth-inhibition ratio of bile salts and binary mixtures (read from Figure 5, approximate). R5·X7

FaSSIF against media with six bile salts

This is the central result (Figure 6). At $\delta$ = 11.5 and without polymer, telaprevir crystallized in about 29 minutes in buffer and about 21 minutes in FaSSIF; the text rounds both to "around 30 minutes". R6.a-b·X14 In MWL-1 and MWL-2, which share FaSSIF's 3 mM bile salt and 0.75 mM lecithin but contain six bile salts, no crystallization occurred within 16 hours. R6.c-d·C24 The authors connect this to their earlier finding that several other biorelevant bile salts "appear" to be better nucleation inhibitors than STC, and to STC being a minor component of human bile and of their mixed media (12% of the bile salt). C25·P5 They conclude that the difference "further demonstrates that STC may not be a good surrogate for human bile composition". C26 The result agrees "in general" with Li et al., who found a two-fold increase in celecoxib induction time in mixed bile salts compared with STC alone, without lecithin (ref 12). C27·R13

MWL-1 and MWL-2, rich in chenodeoxycholate and deoxycholate salts respectively, both stayed beyond 960 minutes; the authors read this as "minimal impact" of the hydroxylation pattern at the supersaturation studied. C28 Since both values are capped by the observation window, the comparison cannot show differences beyond 16 hours. X6

Adding HPMCAS-MF: the medium effect disappears among the lecithin media

The question here was whether an effective polymer overrides the medium. HPMCAS-MF is one of the most potent polymeric inhibitors of telaprevir crystallization (Mosquera-Giraldo 2016, ref 9). C33·R16 The authors dissolved 150 µg/mL of it in each medium and worked at the amorphous solubility, assuming that the polymer at this concentration does not change solution thermodynamics, so that $\delta$ = 28.8. C34 All these solutions were turbid, read as glass–liquid phase separation. R8·C35

In buffer without polymer, telaprevir at its amorphous solubility crystallized immediately (Figure 7). R7.a With HPMCAS-MF, the induction times were about 433 minutes in buffer, about 474 in FaSSIF, about 357 in MWL-1 and about 394 in MWL-2, and one-way analysis of variance found no difference between these media (F < F_critical). R7.b-f·C37 The authors describe this as "contradictory" to the strong medium effect without polymer at lower supersaturation. C38 Relating the times to a small-intestinal transit of about 3–4 hours (ref 31), they write that the solubility enhancement of telaprevir formulations "can likely be guaranteed over a biorelevant time frame" with proper selection of the inhibitory excipient, regardless of fluctuations in medium composition. C39·R14 Note by the AI writer (not discussed in the paper): supersaturation (11.5 to 28.8) and polymer (absent to present) changed together between Figures 6 and 7, so the loss of the medium effect cannot be assigned to the polymer alone. X10

Lecithin cancels the bile salt contribution

Figure 8 removes the lecithin. With the same 150 µg/mL HPMCAS-MF at $\delta$ = 28.8, 3 mM STC alone gave about 678 minutes, against about 472 minutes for FaSSIF, which is the same STC with 0.75 mM lecithin; the six-bile-salt mix without lecithin (M-2) gave more than 960 minutes, against about 396 minutes for MWL-2 with lecithin; buffer with polymer gave about 432 minutes. R9 The buffer, FaSSIF and MWL-2 bars repeat the values of Figure 7. R9

Table D. Induction times at the amorphous solubility with HPMCAS-MF (read from Figures 7 and 8, approximate; ranges cover the two readings of the repeated bars). R7·R9

The authors conclude that "it is evident" lecithin decreases the inhibitory effect of STC, that the same holds for the mixed bile salts, that bile salts and HPMCAS-MF show "a synergistic effect" with mixed bile salts stronger than STC, and that this synergy "is largely negated" by lecithin, since buffer and FaSSIF with polymer gave very similar times. C43·C44·C45 With the lower lecithin concentration of the updated FaSSIF-V2 (0.2 mM instead of 0.75 mM, ref 37), they expect a "more pronounced" medium effect in the presence of HPMCAS-MF. C46·R15 Note by the AI writer (not discussed in the paper): no bile salt-only run at $\delta$ = 28.8 and no statistical test of Figure 8 are reported, so "synergistic" here means longer with bile salt and polymer than with polymer alone; whether the combination exceeds the sum of the single effects was not tested. X9

Crystal morphology

Figure 9 shows SEM images of crystals collected after the induction-time runs. In buffer, telaprevir formed bouquet-like aggregates of prismatic particles with smooth surfaces radiating from a central origin (9a); HPMCAS-MF reduced the number of particles per aggregate and gave rough surfaces (9b). R10.a-b·C47 Crystals from MWL-1 and MWL-2 were similar in shape and smoothness to those from buffer but less aggregated (9c, 9e), and with HPMCAS-MF they were rough aggregates resembling those from buffer with polymer (9d, 9f). R10.c-f·C47 With polymer, FaSSIF gave smooth, more plate-like crystals (9h), although its induction time matched the MWL media; the authors conclude that the additives "clearly interact differently with the crystal surface". R10.h·C49 They relate the roughening to earlier reports: PVP was suggested to form a net-like structure on sulfathiazole crystals, and HPMCAS roughening of felodipine was suggested to come from pinning of growth steps (refs 42–44). C48·R17 Because morphology affects dissolution rate, as shown for dipyridamole and phenytoin (refs 45, 46), they suggest that the redissolution of precipitates "may vary" with the medium in which they form. C50·R17 Note by the AI writer (not discussed in the paper): Figure 9c and 9e show crystals from the MWL media without polymer, although Figure 6 reports no crystallization in these media within 960 minutes at $\delta$ = 11.5; the conditions under which those crystals formed are not stated, and panel 9g (FaSSIF without polymer) is not discussed. X11

Discussion

The authors' chain of reasoning. The paper's interpretation can be drawn as one chain with two switches: whether the bile salts are free monomers or bound in micelles, and whether lecithin is present when a polymer is used. C13·C22·C43

Each link in the chain rests on different evidence. The induction times themselves are direct measurements. R3·R4·R6·R7·R9 The monomer–micelle assignment is inferred from flux and induction times, and the adsorption picture is the authors' explicit speculation. C17·C21·C22 The explanation of the FaSSIF result relies on the earlier single bile salt data ("it appears"). C25

What the authors take from it for testing. In their words, the observations "suggest" that the in vivo crystallization tendency of poorly soluble compounds "may not be adequately predicted" by experiments in commonly used simulated media. C29 This "may be of particular relevance" for two-stage dissolution tests of weakly basic drugs, in which the acidic sample is diluted into intestinal media containing STC and lecithin (refs 26–28). C30 Because excipient selection is still largely empirical, and because the medium effect "may not always be outweighed" by stabilizing excipients, they recommend that "more attention should be paid to testing media selection" when evaluating supersaturating formulations. C41 They also connect their result to a rat study with halofantrine in polyethylene glycol 400 and polysorbate 80, where exogenous surfactant above a threshold was suggested to make bioavailability more independent of bile composition (ref 32). C40

Note by the AI writer (not discussed in the paper): what this suggests for formulation work. Three practical points follow from the data, within their limits. First, without a polymer, a fast crash of a telaprevir-like drug in FaSSIF may overstate its crystallization risk compared with media containing several bile salts, so a FaSSIF result alone is weak evidence of failure in vivo. R6 Second, with an effective polymer at the amorphous solubility, the choice among buffer, FaSSIF and the two lecithin media did not change the outcome for telaprevir; the choice of medium may matter more for formulations that rely on weaker inhibitors. R7 Third, both the lecithin level and the bile salt concentration act as hidden variables: 0.75 mM lecithin removed the bile salt contribution seen with polymer, and the same bile salt mixture was a strong inhibitor at 1.86 mM and a weak one at 3.72–5.58 mM, a range that overlaps the fasted intestinal levels listed in Table 1. R9·R3·R4·P5

Limitations

Limits the authors state. The authors note that their mixed micelle with lecithin recipe "is also a simplified biorelevant testing medium", and that dynamic fluid motion, concentration fluctuation and food effects further complicate drug behaviour in vivo, so that in vitro testing "might not fully capture" precipitation in the gastrointestinal tract. C31 In the growth experiments they assume that bulk growth dominates desupersaturation while noting that secondary nucleation "can occur" at $\delta$ = 7.9. C8 Their supersaturation at the amorphous solubility rests on the assumption that 150 µg/mL polymer has a negligible effect on solution thermodynamics. C34

The AI writer's own judgement (1): what the data can and cannot resolve. The large contrasts are robust: about 21–29 minutes against more than 960 minutes at equal $\delta$ cannot be an artefact of run-to-run scatter, and the medium comparison at equal supersaturation is a real strength of the design. R6·E1 But the manuscript does not state the number of runs or the error bar type, and many values are capped at 960 minutes, so "comparable" or "little difference" among capped conditions means only "both beyond 16 hours". X5·X6 The growth-inhibition ranking rests on a lower bound for STCDC and on a binary-L value with a wide error bar. X7 The mixtures are compared with single bile salts at a different total concentration, and the provenance of the single-salt bars is not stated. X4

The AI writer's own judgement (2): the mechanism and the scope of the claims. The monomer–micelle reading is plausible but indirect: no critical micelle concentrations or micelle measurements are given, and Figure 2 partly conflicts with the statement that the low mixtures do not change the flux. C17·X2 Two headline statements reach beyond their controls. The abstract's "differences between media were minimized" by the polymer holds for buffer, FaSSIF, MWL-1 and MWL-2, but with the same polymer the lecithin-free media gave about 678 and more than 960 minutes against about 432 in buffer. X8 The "synergistic" effect of bile salts with HPMCAS-MF lacks a bile salt-only run at $\delta$ = 28.8 and a statistical test. X9 Between the experiments without and with polymer, supersaturation also changed, so the polymer's role is confounded with the higher $\delta$. X10 Finally, the study covers one drug, one polymer at one concentration, one lecithin level and one temperature; the SEM conditions for the MWL crystals without polymer are not stated. X11·X12

Minor source issues. The equations are numbered (1), (3) and (4), with no Equation (2). X1 The text gives "around 30 minutes" for FaSSIF where Figure 6 reads about 21 minutes. X14 The manuscript prints no journal, year or DOI.

Conclusion

The paper shows, for telaprevir at equal activity-based supersaturation, that the bile salt composition of a test medium can change the induction time by more than an order of magnitude: FaSSIF, with sodium taurocholate as its only bile salt, allowed crystallization within about half an hour, whereas media built from six human bile salts with the same lecithin did not crystallize within 16 hours. R6·C24 Bile salt mixtures behaved like their components, inhibiting strongly as monomers and much less as mixed micelles, with no synergy between bile salts. R3·R4·C14 A polymeric inhibitor, HPMCAS-MF, made the lecithin media indistinguishable at the amorphous solubility, while bile salts without lecithin lengthened the induction time further, an effect the authors call synergistic and find "largely negated" by lecithin. R7·R9·C37·C44·C45 The mechanisms offered, monomer adsorption and micellar counteraction, remain untested, and the conclusions are bounded by one drug, one polymer and in vitro media. C21·C22·C31 The authors' own summary is that media composition plays a "critical role" in crystallization kinetics and that "attention should be paid to selecting biorelevant media". C32·C41