How the surfactant in an acidic suspending liquid decides whether spray-dried celecoxib–HPMCAS particles stay amorphous

Background and motivation

Reaching adequate plasma exposure with a poorly water-soluble drug is hard, and the authors note that it is hardest in early development, when toxicity studies need very high exposure to judge the safety of a new molecular entity (Introduction). Among the solubilization routes they list (pH adjustment, micelles and lipids, cosolvents, solid dispersions and nanocrystals), the amorphous solid dispersion (ASD) relies on the amorphous form of the drug, whose theoretical solubility is usually several times that of the crystal (Murdande 2010). The amorphous form is thermodynamically unstable and tends to crystallize, which removes the solubility gain; heat, moisture and solvents speed this up (Rumondor 2011; Rumondor and Taylor 2009). Stabilizing additives, usually polymers, are therefore mixed with the drug, and spray drying is one of the common ways to make the dispersion (Dhirendra 2009).

In toxicology studies, ASD particles are usually suspended in a liquid and given to animals by oral gavage. If the polymer in the dispersion has a pH-dependent solubility, as the enteric polymer hydroxypropyl methylcellulose acetate succinate (HPMCAS) does, acidifying the liquid keeps the polymer largely undissolved, so the particles stay suspended rather than dissolving. This approach works only if the drug does not crystallize during the life of the suspension. The authors add that the same requirement holds for marketed ASDs made with enteric polymers: these sit in the acidic stomach before they reach the higher pH of the small intestine, where they are meant to dissolve.

A suspension vehicle is rarely just acidified water. Polymers are added to raise the viscosity and keep particles suspended, surfactants to improve the wetting of the particles and the flow of the suspension, and polymers can also serve as anti-nucleants to slow crystallization, a risk that rises sharply once the formulation meets water (Ilevbare 2012; Ozaki 2013). The authors describe the formulation of ASD suspensions as largely empirical, with high-throughput screening used to find a workable vehicle for a given dispersion. Because the liquid is a multi-component system, how each additive and their interactions affect stability is poorly understood.

Earlier studies of additives in amorphous suspensions concentrated on polymers as inhibitors of nucleation and crystal growth; Yokoi 2005, for example, examined HPMC and a sugar ester in an aqueous suspension of amorphous cefditoren pivoxil. The role of surfactants in this setting had been largely neglected. Work in crystallization and crystal engineering had nonetheless shown that surfactants can enhance nucleation, accelerate solution-mediated polymorph transformation and change crystal growth (Canselier 1994, for γ-aminobutyric acid; Rodríguez-Hornedo and Murphy 2004, for carbamazepine dihydrate; Garti and Zour 1997, for glutamic acid). A wetting agent added for practical reasons could therefore be the component that makes an amorphous suspension fail, and the authors conclude that this deserves attention.

The paper sets three objectives: (1) to find out whether celecoxib crystallizes in or at the surface of the suspended ASD particles, or from the solution after the drug leaches out; (2) to evaluate how surfactants affect crystallization through their effects on celecoxib nucleation and on drug leaching; and (3) to relate surfactant structure to suspension stability. One spray-dried dispersion of celecoxib in HPMCAS LF at 33.3 w/w% drug loading (DL) was suspended in pH 3 HCl with no surfactant or with one of five commonly used surfactants: sodium dodecyl sulfate (SDS), sodium taurocholate (STC), Polysorbate 80, Poloxamer 188 and Triton X100 (Figure 1). Each surfactant was used just above its critical micelle concentration (CMC), and the work was done at 25 °C.

Background concepts

An ASD in an acidic suspension. Here the ASD is celecoxib (COX in the paper's sample names) dispersed in HPMCAS, which barely dissolves at pH 3, so in acidified water the particles persist as a suspension. Some of the amorphous drug can still pass into the liquid; the paper calls the passage of drug or polymer from the particles into the liquid "leaching". Because amorphous drug is more soluble than the crystal, leached drug can raise the liquid above the crystal solubility, and this excess is the driving force for crystals to form in the liquid. The polymer, by contrast, simply dissolves up to its solubility in the medium. The paper considers two routes to crystals: from the bulk solution after the drug has dissolved, or from the matrix or surface of the particles themselves.

Supersaturation. The central quantity in the paper is the supersaturation ratio, stated in words in the Methods and in the caption of Figure 7:

Here $C$ is the measured concentration of dissolved celecoxib (µg/mL) and $C_{\mathrm{e}\mathrm{q}}$ is the equilibrium solubility of crystalline celecoxib Form III in the same medium at 25 °C. In plain terms: $S$ says how many times more drug is dissolved than the crystal can hold at equilibrium; $S=1$ is a saturated solution and $S=5$ is five times saturated. Form III is the crystalline form used as the solubility reference, and it grows as thin needles. The normalization is done separately in each medium, because each surfactant raises $C_{\mathrm{e}\mathrm{q}}$ by a different amount: from 1.1 µg/mL without surfactant to 14.5 µg/mL in 0.25% SDS. The authors argue that running all nucleation tests at the same initial $S$ "ensured" the same driving force in every medium. Note by the AI writer (not discussed in the paper): this convention treats drug held in micelles and free drug alike; it measures the driving force correctly if micelles take up drug in proportion to the free-drug concentration, which the paper does not test.

Nucleation induction time. The induction time $\tau$ is the time between creating a supersaturated solution and the first detectable crystals. The paper creates the solution by solvent shift: a small volume of celecoxib stock in methanol is added to the stirred aqueous medium, which is clear at first. Light extinction at 350 nm stays near zero until particles appear and then rises. Straight lines are drawn through the flat and the rising parts of the trace, and their intersection is taken as $\tau$ (Figure 3). At the same $S$, a longer $\tau$ means the medium slows nucleation and a shorter $\tau$ means it promotes nucleation. In the single trace shown for 0.25% SDS, extinction stays about zero to about 30 min, reaches about 0.19 at 60 min, and the construction gives $\tau$ = 38 min.

Critical micelle concentration from a solubility break-point. The CMC is the surfactant concentration at which micelles form; above it, the surfactant solubilizes a hydrophobic drug much more strongly. Because a dissolved drug can shift the CMC, the authors measured the CMC with celecoxib present at saturation, from the solubility of Form III over a range of surfactant concentrations. The paper states the rule in words: below and above the CMC the solubility rises linearly with surfactant concentration, but at different rates, and the CMC is where the two lines meet. Written in symbols (the symbols are the AI writer's; the relation is the paper's):

Here $c_s$ is the surfactant concentration (% w/w), $a_1$ and $b_1$ are the intercept and slope of the line through the low-concentration points, and $a_2$ and $b_2$ those of the line through the high-concentration points. In plain terms: the solubility curve has a knee, and the CMC is the surfactant concentration at the knee. Setting the two lines equal gives the crossing point:

In plain terms: the CMC is the surfactant level at which both straight lines predict the same solubility. Table A shows the SDS data. The solubility hardly moves from 0% to 0.175% SDS and then climbs steeply. The card's recomputation reproduces the printed SDS value: a line through the points from 0% to 0.175% and a line through the points from 0.25% to 0.40% meet at 0.227%, against the printed 0.23%. If the 0.210% point is added to the upper line, the crossing moves to 0.215%, so the CMC depends slightly on how the points are split.

Table A. Solubility of celecoxib Form III in pH 3 HCl containing SDS at 25 °C (from SI Table S1; mean ± SD, number of replicates not stated). The break in slope gives the CMC of 0.23% (Figure S1).

Five surfactants of two shapes. The paper describes the five surfactants by charge and by the shape of their hydrophobic part (Figure 1). SDS is anionic, with a charged head and a linear C12 tail. STC is also anionic; it is a bile salt with a charged head and a rigid steroid ring system that has a hydrophilic and a hydrophobic face. Polysorbate 80 is neutral, with polyoxyethylene groups as its hydrophilic part (w+x+y+z = 20 oxyethylene units in Figure 1) and, in the text, a "chain-like C20 tail".Card check: Figure 1 draws an 18-carbon unsaturated acyl chain, and the only "20" in the drawing is the number of oxyethylene units, so the "C20 tail" is unverified (story note X15). Poloxamer 188 is a block copolymer of ethylene oxide and propylene oxide (EO–PO–EO), in which the PO block is the hydrophobic part. Triton X100 is nonionic, with a polyoxyethylene head like Polysorbate 80 and Poloxamer 188, but a bulky aromatic hydrocarbon group as its hydrophobic part. Read as a design, the set contains two useful contrasts: SDS and STC share an anionic head but differ in the shape of the hydrophobic part, and Polysorbate 80 and Triton X100 share a polyoxyethylene head but differ in the same way.

Detecting crystals. Under a polarized light microscope (PLM) with crossed polarizers, crystals appear bright (birefringent) while amorphous particles stay dark. X-ray powder diffraction (XRPD) gives sharp peaks for a crystal, which identify its form, and a broad halo for an amorphous solid. Differential scanning calorimetry (DSC) shows the glass transition temperature ($T_g$) of the dispersion and any melting; Form III melts at 165 °C.

Methods in detail

The work can be split into four contributions. The first measures how each surfactant changes nucleation of celecoxib from solution at an equal driving force, which needs the solubility and CMC in each medium first. The second tests the real outcome: whether and when suspended ASD particles crystallize, and where the crystals form. The third measures how much drug and polymer leach from the particles into each liquid. The fourth puts the pieces back together by adding HPMCAS to the nucleation test and by relating surfactant structure to the results. The figure below shows the workflow.

Contribution 1: Nucleation from solution at an equal driving force

① Problem. If celecoxib crystallizes in the liquid after leaching from the particles, the surfactant can act on that step directly. But each surfactant also raises the solubility of the drug, by a different amount, so comparing nucleation at one fixed concentration would mix two effects: how much the surfactant solubilizes and how it changes the kinetics of nucleation. The authors needed a test that isolates the kinetic effect and simulates drug that has already leached into the liquid.

② What they did. First, they measured the solubility of Form III over a range of each surfactant: excess crystals were stirred in 15 mL of medium for 48 h at 25 °C, the mixture was ultracentrifuged at 40,000 rpm and the supernatant was assayed by HPLC with UV detection at 250 nm. The ranges were SDS 0–0.40%, STC 0.10–1.50%, Polysorbate 80 0–1%, Triton X100 0–0.15% and Poloxamer 188 0.002–3% (SI Tables S1–S5). The CMCs read from these curves were 0.23% (SDS), 0.69% (STC), 0.0058% (Polysorbate 80), 0.16% (Poloxamer 188) and 0.022% (Triton X100), and the test levels, described as "11% above" each CMC, were 0.25%, 0.75%, 0.0064%, 0.17% and 0.025%.Card check: the printed test levels lie 6–14% above the printed CMCs, not exactly 11% (story note X2); the STC CMC is 0.69% in the text but 0.668% on the label of Figure S2 (X1). Two levels below the CMC were added for the nucleation study only: SDS 0.175% and STC 0.4%. The induction time was then measured by the method of Ilevbare 2012. A methanol stock of celecoxib (usually about 100 µL) was added to 50 mL of medium stirred at 300 rpm with a cross-shaped stirrer at 25 °C, in an amount that gave an initial $S$ of 5 relative to the Form III solubility in that medium. Extinction at 350 nm was followed with a fibre-optic dip probe of 10 mm path length, and $\tau$ was read as in Figure 3. Each medium was tested at least three times; the STC runs were followed for 16 h. Crystals were collected at the end and examined by microscopy.

③ Why it works. Holding the initial $S$ at 5 in every medium removes the solubilization effect from the comparison, on the authors' assumption that equal $S$ means equal driving force. Any remaining difference in $\tau$ then reflects how the surfactant changes the rate of nucleation itself, and against the surfactant-free reference each surfactant can be classed as a promoter (shorter $\tau$) or an inhibitor (longer $\tau$). The levels below the CMC test whether micelles are needed for the effect.

④ How it differs from earlier work. Ilevbare 2012 used the constant-supersaturation induction-time method to compare polymers. This paper applies it to surfactants, which requires the solubility and the CMC in each medium with the drug present, and adds sub-CMC controls.

Contribution 2: Suspension stability and where the crystals form

① Problem. The nucleation test uses drug in solution without any particles or polymer. The practical question is different: do suspended ASD particles crystallize in a given medium, how quickly, and do the crystals grow in the liquid or in the particles? Answering it needs a well-defined amorphous starting material and a suspension test close to the real use.

② What they did. Celecoxib (20 mg/mL) and HPMCAS LF (40 mg/mL) were dissolved in acetone and spray-dried on a ProCepT Microspray dryer (0.8 mm two-fluid nozzle, 6 mL/min, inlet 85 ± 5 °C, outlet 55 ± 5 °C, atomizing air 4 L/min), collected by a small cyclone and dried overnight under vacuum at room temperature. This gave a DL of 33.3 w/w%. The powder showed no birefringence (Figure 2a), one glass transition at 61 °C (midpoint) with no thermal event near 165 °C (Figure 2b), and only a broad XRPD halo (Figure 2c). For the stability test, 100 mg of ASD was added to 10 mL of medium (10 mg/mL solids) in a 20 mL vial, held at 25 °C and stirred at 150 rpm. Six media were used: pH 3 HCl alone and with each surfactant at its test level. A drop was examined by PLM at 0.5, 1, 2, 4, 6, 24 and 48 h, and the presence of birefringent particles was scored as crystallization (Table 1). Solids filtered from the SDS suspension after 24 h were dried and examined by XRPD (Figure 6). Some suspensions were also examined after about 7 days; these observations appear only in the Discussion (Figure 10).

③ Why it works. PLM detects crystals against dark amorphous particles, and repeated sampling gives the time of first appearance. The location of the crystals indicates the route: free needles in the liquid point to growth from dissolved drug, birefringence on or in the particles to crystallization from the particle. XRPD identifies the form.

④ How it differs from earlier work. The test uses spray-dried dispersion particles in an acidified vehicle rather than the drug alone, and it asks where crystallization starts, which matters for choosing a remedy.

Contribution 3: Drug and polymer leaching from the particles

① Problem. The nucleation test fixes the supersaturation, but in a suspension the supersaturation is set by how much drug leaches out, and a surfactant can change that too. The authors suspected that leaching of the drug follows leaching of the polymer, since the drug is dispersed in the polymer. Both quantities had to be followed in the same suspensions.

② What they did. Suspensions were prepared as in Contribution 2 and sampled up to 24 or 48 h. At each time point a 10 mL volume of suspension was filtered through a 0.45 µm nylon syringe filter. Celecoxib in the filtrate was assayed by HPLC and divided by the Form III solubility in that medium to give $S$ (Figure 7). The authors state that "it is assumed" the small amount of leached HPMCAS does not change the solubility of celecoxib to a large extent. HPMCAS in the filtrate was assayed by phenol–sulfuric acid colorimetry: 2 mL of filtrate with 0.05 mL of 89% phenol and 5 mL of sulfuric acid (above 51%), left for 1 h, absorbance at 490 nm. The calibration used known HPMCAS concentrations printed as "2–20 mg/mL" in pH 3 HCl. Polymer was measured in the SDS, STC, Triton X100 and surfactant-free media only (Figure 8).

③ Why it works. For an amorphous drug, a rise in dissolved concentration means drug is leaching from the particles, and a fall means some of it has crystallized, because a crystal holds less in solution. Following $S$ over time therefore shows both processes, and the times of the falls can be checked against the PLM record. Measuring the polymer in the same filtrate tests whether a surfactant that dissolves more HPMCAS also releases more drug.

④ How it differs from earlier work. The polymer is not treated as a fixed matrix: its dissolution in the vehicle is measured as a possible cause of drug release, the link the authors later use to explain why SDS fails first.

Contribution 4: Recombining the factors

① Problem. With the first three data sets in hand, one surfactant did not fit. Poloxamer 188 promoted nucleation in solution and, in the authors' reading, caused drug leaching similar to Polysorbate 80, yet its suspension stayed crystal-free for 48 h. The solution test contains no polymer, whereas the suspension liquid always contains some leached HPMCAS. The authors state that they "suspected" the leached polymer might change the effect of the surfactant on the induction time.

② What they did. Induction times at $S$ = 5 were measured again in the SDS, Polysorbate 80 and Poloxamer 188 media with 3 µg/mL HPMCAS added (Figure 9). The level was chosen as the minimum polymer concentration expected in the suspension liquid, just below the amount seen to leach at pH 3 without surfactant (Figure 8). The surfactant levels are not restated for Figure 9; the no-polymer values match the 0.25% SDS, 0.0064% Polysorbate 80 and 0.17% Poloxamer 188 bars of Figure 4. Separately, the authors compared the structures of promoters and inhibitors (Figure 1) and drew on the literature to propose mechanisms.

③ Why it works. Adding the polymer at a level that the suspension itself produces brings the solution test one step closer to the suspension. If the polymer changes $\tau$ in one surfactant medium but not in others, the solution ranking alone cannot predict suspension stability.

④ How it differs from earlier work. Earlier studies looked at one additive at a time, typically a polymer; here the surfactant and the polymer that leaches from the dispersion are tested together.

The account the authors build from these four contributions can be drawn as one chain that splits by surfactant.

Experimental analysis

Setup in brief. All experiments used one spray-dried celecoxib–HPMCAS LF dispersion at 33.3 w/w% DL, pH 3 HCl as the base medium and 25 °C. The six suspension media were the surfactant-free medium and SDS 0.25%, STC 0.75%, Polysorbate 80 0.0064%, Poloxamer 188 0.17% and Triton X100 0.025% (% w/w). The nucleation test used these six media plus SDS 0.175% and STC 0.4%. Suspensions held 10 mg/mL solids in 10 mL at 150 rpm, while the nucleation test used 50 mL at 300 rpm with no particles or polymer. Replicate numbers are given only for the induction times (at least three runs); they are not stated for the solubility, suspension, drug-leaching or polymer-leaching data, and Figure 7 has no error bars. Table B collects the per-medium data.

Table B. CMCs and solubilities from the Results text and SI Tables S1–S5; induction times are the printed bar labels of Figure 4 (means of at least three runs) with spreads read from the error bars (approximate); suspension results from Table 1 (10 mg/mL ASD, 25 °C, 150 rpm).

Main result 1: two surfactants make the suspension crystallize within an hour. By PLM, the SDS suspension contained crystals from the first check at 0.5 h and the Polysorbate 80 suspension from 1 h; both stayed crystalline to 48 h (Table 1). In SDS, needle-like crystals floated in the liquid next to dark aggregates of ASD particles at 0.5 h; in Polysorbate 80 there were many thin needles at 1 h (Figure 5a, b). The surfactant-free, Poloxamer 188, Triton X100 and STC suspensions showed only particles and aggregates up to 48 h (Figure 5c–f). Solids recovered from the SDS suspension after 24 h gave XRPD peaks at the Form III positions on top of a raised halo, which the authors read as a mixture of Form III and remaining ASD (they say the patterns "suggest" this; Figure 6). The peaks are weaker than in the Form III reference, for example about 252 against about 607 counts near 19.7° 2θ (read from Figure 6, approximate). The authors also state that the needles grow along the crystal c-axis, citing the labelled reflections, but the paper gives no basis for the indexing. An XRPD pattern for the Polysorbate 80 solids is not shown.

Main result 2: in solution, three surfactants speed nucleation and two slow it. Against 88 min without surfactant, the induction time at $S$ = 5 was 7 min with Polysorbate 80, 12 min with Poloxamer 188 and 41 min with 0.25% SDS, so the authors rank nucleation promotion as Polysorbate 80 > Poloxamer 188 > SDS (Figure 4). Triton X100 lengthened $\tau$ to 273 min, and in 0.75% STC no rise in scattered light was seen in 16 h, so its induction time, "although not accurately determined, appears to be" longer than 960 min. The effects held below the CMC, which the authors flag with "Interestingly": 0.175% SDS still shortened $\tau$ to 29 min, and 0.4% STC still lengthened it.Card check: the STC 0.4% bar in Figure 4 carries the label 273 min, the same as Triton X100, but is drawn to the 400-min axis limit with a lower whisker near 333 min, and the text gives no value; its induction time is unresolved (story note X3). Crystals collected after every run were thin needles, which the authors call "consistent with" Form III; no XRPD was done on them. Note that the promoter–inhibitor order in solution does not match the suspension outcome one to one: SDS is the weakest of the three promoters in solution, yet its suspension failed first, and Poloxamer 188 is a stronger promoter than SDS, yet its suspension did not crystallize. The leaching data explain part of this.

Table C. Celecoxib supersaturation S in the suspension liquid (dissolved drug divided by Form III solubility in the same medium), read from Figure 7, approximate; 10 mg/mL ASD, 25 °C; no error bars plotted and replicates not stated.

Main result 3: the supersaturation curves match the crystal record. Table C shows the drug side of the story. In SDS the supersaturation spiked within the first minutes and then fell to about 3.2 by 1 h, the period in which crystals appeared; the text gives about 107 µg/mL dissolved within 10 min, an $S$ of 7.4, followed by a fall to 3.2 and a slight rise over the next 23 h.Card check: the text value (S 7.4 at 10 min) and the figure (about 8.6 at about 5 min, about 6.2 at 10 min) disagree; the card treats the SDS peak as 7.4–8.6 (story note X4). The printed 7.4 does follow from 107 µg/mL divided by 14.5 µg/mL. In Polysorbate 80, $S$ fell from 3.4 to 1.1 between 0.5 and 3 h (text), again matching the crystals seen from 1 h; by 24 h it read about 0.85, slightly below saturation, which lies within the uncertainty of the solubility value used. The authors read the rises as leaching and the falls as crystallization (caption of Figure 7). In the four crystal-free media the curves behave differently. Triton X100 held a low $S$ of about 1.5–1.8 throughout, so leaching was minor. Poloxamer 188 rose for 4 h to about 4.2 and then fell slowly to about 2.8 by 48 h with no crystals seen; the authors propose this fall "could be" crystallization so slow that the few crystals escaped the microscope. The authors call $S$ in the surfactant-free and STC media "relatively constant" over 48 h and place the maxima of four media within the first six hours.Card check: Figure 7 shows both media still rising and highest at 48 h (surfactant-free about 4.3 to 6.7, STC about 3.5 to 4.9); the card uses the figure values (story note X5). Read from the figure, the surfactant-free suspension had the highest supersaturation of all six media from 1 h on, yet it stayed crystal-free.

Table D. HPMCAS in the suspension liquid, read from the log axis of Figure 8 (approximate; error bars plotted for the surfactant-free medium only). The last column is the card's arithmetic: highest S in Figure 7 (or the text value for SDS) times the Form III solubility in that medium.

Main result 4: SDS dissolves far more of the polymer. Without surfactant, HPMCAS in the liquid stayed at a few µg/mL, as expected for an enteric polymer at pH 3 (Table D). With SDS it rose from about 73 µg/mL at 0.5 h to about 860 µg/mL at 24 h. STC raised it less, to about 43–48 µg/mL, and Triton X100 gave a profile close to the surfactant-free medium. The authors describe the SDS increase as "more than 3 orders of magnitude" above the surfactant-free medium.Card check: the plotted values give about 250-fold at 24 h and about 50-fold at 0.5 h, that is about 2–2.5 orders of magnitude (story note X7). The printed calibration range of 2–20 mg/mL also lies far above the low values measured, so the values of a few µg/mL carry uncertain accuracy (X8). They attribute the SDS effect "most likely" to polymer–surfactant complexes, as reported for HPMC with SDS: Nilsson 1995 described SDS clusters formed on single polymer chains at low polymer concentration and shared between chains at high concentration, and Qi 2013 showed HPMC–SDS complexes by fluorescence spectroscopy and faster ASD dissolution with an ionic surfactant. No complex was detected directly in this paper. The last column of Table D puts the polymer next to the drug. In absolute terms, the order of dissolved drug (SDS far above STC, STC above the surfactant-free and Triton X100 media) follows the order of polymer leaching, except that the surfactant-free and Triton X100 media are about equal. In supersaturation terms the link does not hold: the surfactant-free medium has the least polymer but a higher $S$ than STC.

Table E. Nucleation induction times at S = 5 with and without 3 µg/mL HPMCAS, read from Figure 9 (approximate); no statistical test or number of runs is given for these tests.

Main result 5: a trace of HPMCAS slows nucleation in Poloxamer 188 only. With 3 µg/mL HPMCAS added, the induction time in the Poloxamer 188 medium rose about fourfold, from about 12 to about 50 min, while SDS and Polysorbate 80 showed no change within experimental error (Table E, Figure 9). The authors conclude that this "most likely explains" the stability of the Poloxamer 188 suspension, especially since the real HPMCAS level there "is likely to be" higher than 3 µg/mL. HPMCAS was not measured in the Poloxamer 188 suspension, so this remains an inference.Card check: the no-polymer SDS whiskers in Figure 9 (about 21–61 min) differ from those of the 0.25% SDS bar in Figure 4 (about 29–52 min) for the same 41-min mean (story note X18). The result also shows that the nucleation test without polymer mis-ranked one of five surfactants with respect to the suspension.

Main result 6: where the crystals form. The authors take the long needles floating free of the particles in SDS and Polysorbate 80 as evidence that these surfactants promote crystallization of drug that has dissolved in the liquid. After about 7 days in 0.025% Triton X100, birefringent regions sat on or in the particles and no free needles were seen (Figure 10; the caption says the birefringence "suggested" crystallization in or on the particles). The Discussion adds that without surfactant crystallization was seen only after more than 7 days and "seems to" start from the particle matrix or surface, and that the STC and Triton X100 suspensions were stable for at least 7 days and crystallized from the particles rather than from solution. These observations are reported only in the Discussion, and only the Triton X100 image is shown. They answer the first objective qualitatively: the fast failures occur in the liquid, and the slow ones in the particles.

Comparison across systems. Combining Tables B to E, the authors sort the surfactants into the four cases of the flow diagram above. For SDS they add that the high supersaturation from substantial leaching "of course also favors nucleation", which makes it the least stable suspension. Their stability ranking is STC ≈ Triton X100 ≈ no surfactant > Poloxamer 188 > Polysorbate 80 > SDS. Poloxamer 188 is placed below the surfactant-free medium because its supersaturation fell slowly, not because crystals were seen; by PLM alone it ties with the surfactant-free medium over 48 h.

The structural pattern. The three nucleation promoters (Polysorbate 80, Poloxamer 188, SDS) all have a chain-like hydrophobic tail, and the two inhibitors (Triton X100, STC) have a bulky hydrophobic group; the authors introduce this with "Interestingly". They note that a suspension has many interfaces on which surfactants can assemble into Langmuir–Blodgett-type films, that such assemblies have been reported to template nucleation and growth (Weissbuch 1988; Bujan 2001), and that a flexible polydiacetylene film was shown to align with nucleating calcite (Berman 1995). From this they find it "conceivable" that flexible linear chains align to promote celecoxib nucleation, which rigid bulky groups cannot, and suggest that drug–surfactant interactions "may be" responsible for the inhibition. Neither interfacial assembly nor drug–surfactant binding was measured, so the pattern rests on five surfactants and literature analogies.

Controls and what they rule out. The surfactant-free medium is the reference in every test, and it shows that the base system is stable for at least 48 h; any early failure is therefore caused by the surfactant. The constant-$S$ design rules out that the solution ranking simply reflects different solubilities. The sub-CMC levels show that the promoting effect of SDS and the inhibiting effect of STC do not require micelles. Note by the AI writer (not discussed in the paper): this argues against micellar solubilization as the cause of the nucleation effects and is compatible with the authors' molecular and interfacial explanation, though it does not prove it. The added-HPMCAS test rules out that the solution test alone predicts the suspension: it did not for Poloxamer 188.

Discussion

Where the findings apply. The results hold for one drug (celecoxib), one polymer (HPMCAS LF), one drug loading (33.3 w/w%), one level of each surfactant just above its CMC, pH 3 HCl, 10 mg/mL solids, 150 rpm and 25 °C, over 48 h with qualitative observations to about 7 days. The surfactant was always in the liquid; the authors note that surfactants are also added to the ASD particles themselves, but this was not tested. The Discussion places the work in two settings, acidified suspensions for animal dosing and ASD dosage forms in the low-pH stomach, but no gastric medium or in vivo study was used. The authors attribute the polymer control of drug release "most likely" to the "relatively low" drug loading, which suggests that the link between polymer leaching and drug leaching may differ at higher loadings.

What remains unresolved. The largest open point is the surfactant-free control. It held the highest supersaturation of all six media, about 4.3–6.7, for 48 h without crystals, although its induction time at $S$ = 5 in solution was 88 min. The two-factor account, nucleation plus leaching, does not explain this, and the paper does not discuss it. Note by the AI writer (not discussed in the paper): the paper's own Poloxamer 188 experiment points to one candidate, the few µg/mL of HPMCAS present in that liquid, but $\tau$ with HPMCAS was not measured in the surfactant-free medium, and differences between the solution test and the suspension (volume, stirring, methanol, the presence of particles) could also matter. Other open points are raised by the authors themselves: whether the slow fall in supersaturation with Poloxamer 188 was undetected crystallization, how much HPMCAS that suspension really held, and the true STC induction time. The tail-shape rule and the SDS–HPMCAS complex are both inferred, not measured. The solubility of celecoxib in the presence of the leached HPMCAS was not measured either, although the polymer reached about 860 µg/mL in SDS.

What it suggests for formulation. The authors' own recommendations are specific. A suspension for toxicity testing needs to be stable for a few hours, and celecoxib ASDs "would" meet this in a simple pH-adjusted medium. Surfactants are often needed for wetting and flow, but the "wrong" surfactant can reduce stability and cause loss of efficacy through crystallization, as SDS and Polysorbate 80 did here. STC and Triton X100 did not compromise stability and "should be assessed" as alternative wetting agents. Leaching of drug from the particles should be minimized, and the effect of surfactants on stability should be weighed in "a risk benefit analysis". Three further readings are the AI writer's and are not stated in the paper. First, a wetting agent for an acidified ASD suspension should be chosen by its effect on nucleation and on polymer dissolution, not by how well it solubilizes the drug: SDS solubilized the drug most and failed first. Second, a surfactant that dissolves the enteric polymer at low pH, as SDS did with HPMCAS, defeats the purpose of acidifying the vehicle, because the polymer no longer holds the drug back. Third, a quick solution induction-time test sorted the surfactants into promoters and inhibitors, but it mis-ranked Poloxamer 188 until the polymer was added; a screen should therefore include the ASD polymer, or a short suspension hold, before a vehicle is chosen.

Further work. The open points translate into direct experiments. The mechanism would be tested by $\tau$ with HPMCAS in the surfactant-free medium, HPMCAS leaching in the Polysorbate 80 and Poloxamer 188 suspensions, and celecoxib solubility with leached polymer present. The scope would be tested by more surfactants of each shape, other drug loadings and surfactant levels, and surfactant placed inside the particles.

Limitations

Limitations stated by the authors. The paper has no separate limitations section, but several caveats are stated in the text. The STC induction time was "not accurately determined" because the runs were followed for only 16 h. The slow fall in supersaturation with Poloxamer 188 "could be" crystallization that the microscope did not detect. The supersaturation values rest on the assumption that the small amount of leached HPMCAS does not change the solubility of celecoxib "to a large extent". The 3 µg/mL HPMCAS added to the nucleation test is described as the minimum level expected, and the real level in the Poloxamer 188 suspension "is likely to be higher". The XRPD of the SDS solids is said to "suggest" a mixture of Form III and residual ASD, and the 7-day crystallization route "seems to" start from the particles.

The AI writer's own judgement (1): the explanation is weaker than the observation. The main observation agrees across microscopy, the Form III XRPD of the SDS solids and the drops in supersaturation. The two-factor explanation is weaker, and the card rates the paper "moderate" overall for this reason. It leaves the surfactant-free control unexplained (high supersaturation, no crystals). The link between polymer leaching and drug leaching rests on polymer data from four media and holds for absolute drug concentration but not for supersaturation. The supersaturation in SDS assumes that about 860 µg/mL of HPMCAS leaves the solubility of celecoxib unchanged, which was not tested. The Poloxamer 188 polymer effect is a single comparison without a statistical test or a stated number of runs. The abstract says STC and Triton X100 "inhibit" crystallization of the suspensions, but the surfactant-free suspension was also crystal-free over 48 h, so inhibition relative to no surfactant is shown only for nucleation in solution.

The AI writer's own judgement (2): the data carry more uncertainty than the text suggests. Four text values disagree with the figures, as flagged above: the SDS peak supersaturation, the STC 0.4% induction time, the "relatively constant" supersaturation in the surfactant-free and STC media, and the "more than 3 orders of magnitude" of HPMCAS in SDS. The solubility without surfactant is 1.1 µg/mL in the text and Table S1 but 1.2 µg/mL in Tables S3 and S4, which shifts the surfactant-free supersaturation by about 9%. The solubilities used for Polysorbate 80 (2.0 µg/mL) and Poloxamer 188 (1.4 µg/mL) are at levels not listed in the SI tables, and their origin is not stated. The Poloxamer 188 CMC rests on solubility changes of 0.1–0.2 µg/mL; the card's own straight-line fit puts it nearer 0.10%, so whether 0.17% lies just above the CMC is weakly determined. Replicate numbers are not stated for the suspension, leaching and solubility data.

The AI writer's own judgement (3): comparability and the structural rule. The nucleation test (50 mL, 300 rpm, methanol stock, no particles or polymer) differs from the suspension (10 mL, 150 rpm, 10 mg/mL ASD with leached polymer), so the induction times should be compared as a ranking, not as absolute times to set against the suspension onset times. The tail-shape rule rests on five surfactants and no interfacial measurement, and the description of Polysorbate 80 as having a "C20 tail" does not match the drawn structure. The c-axis growth direction of the needles is stated without an indexing basis. None of these points overturns the main observation, but together they mean the paper establishes which surfactants failed in this system more firmly than why they failed.

Conclusion

The paper shows, for one celecoxib–HPMCAS spray-dried dispersion in pH 3 HCl, that the surfactant in the suspending liquid can decide the fate of the suspension. SDS and Polysorbate 80 caused celecoxib needles to grow in the liquid within 0.5 and 1 h, while the other four suspensions stayed crystal-free for 48 h. In solution at equal supersaturation, three surfactants shortened the induction time and two lengthened it; SDS also dissolved far more HPMCAS than the other media measured, and a trace of HPMCAS slowed nucleation in Poloxamer 188 only. The authors propose that the two surfactant effects, on nucleation and on polymer-driven drug leaching, "appear to play a critical role".

The observation is robust within its narrow scope; the mechanism is a plausible proposal that leaves the surfactant-free control unexplained and rests on limited polymer data. For formulation, the wetting agent in an acidified ASD suspension is part of the stability problem and should be screened for its effects on nucleation and on the dispersion polymer before it is chosen.