How much supersaturation can drug-loaded mesoporous silica deliver? Ritonavir in SBA-15, measured by dissolution and by membrane flux

Background and motivation

Many drug candidates are poorly soluble in water, and formulation scientists have two broad ways to help them. Solubilizing formulations raise the crystalline solubility with surfactants, cosolvents or complexing agents. Supersaturating formulations instead produce dissolved concentrations above the crystalline solubility; the dissolved drug then has a higher chemical potential, or thermodynamic activity, than in a saturated solution, and this raises its flux across the gut membrane C6. Lipid systems, salts and amorphous solid dispersions (ASDs) can all supersaturate. In an ASD a polymer is used to slow crystallization of the amorphous drug in the solid and after dissolution.

Ordered mesoporous silica (OMS) is an alternative carrier for amorphous drug. It has a regular arrangement of uniform pores, 2–30 nm wide, and a surface area often of 500–1000 m²/g. The introduction explains that adsorbing drug on this large surface lowers the free energy of the system, so the adsorbed drug is amorphous and described as thermodynamically stable, and that pores small enough to stop drug clusters from reaching the critical nucleus size prevent crystallization inside them (Qian and Bogner 2012; Rengarajan et al. 2008) C7.

Many studies had shown that OMS speeds up dissolution compared with the crystalline drug, but few had looked at supersaturating (nonsink) conditions. The study the authors build on most directly is Mellaerts et al. (2008): itraconazole loaded at 20% w/w in SBA-15 stayed well above its crystalline solubility in simulated gastric fluid for 60 min, yet release plateaued at about 55% R28. Mellaerts et al. suggested a dynamic adsorption equilibrium but did not test it, and incomplete release from OMS has been reported under both sink and supersaturating conditions C8.

The paper therefore asks how well SBA-15 generates and maintains supersaturation, and why its release is incomplete. The stated hypothesis was that SBA-15 can generate and maintain a supersaturated ritonavir solution under nonsink conditions, with the maximum supersaturation set by the amorphous solubility of ritonavir C9. Ritonavir was chosen because it crystallizes slowly from solution, so its supersaturation can be followed without interference from crystallization (Ilevbare et al. 2013) C5. The hypothesis was tested in two ways: by measuring ritonavir in the dissolution medium, and by measuring its flux across an artificial membrane in comparison with neat amorphous ritonavir C9. The title says "Part 1"; this article covers only what is in this paper.

Background concepts

Crystalline and amorphous solubility. The crystalline solubility is the concentration in equilibrium with the crystal. The amorphous solubility is the highest concentration of molecularly dissolved drug that can be reached before the solution separates into a drug-lean aqueous phase and a drug-rich liquid phase, a process called liquid–liquid phase separation (LLPS). Above it, extra drug goes into drug-rich nanodroplets rather than into solution. For ritonavir in 50 mM phosphate buffer pH 6.8 at 37 °C, the paper measured a crystalline solubility of 3.6 ± 0.6 µg/mL and an amorphous solubility of 27.9 ± 0.3 µg/mL, 7.8 times higher; in fasted state simulated intestinal fluid (FaSSIF) the values were 8.8 ± 0.8 and 42.7 ± 3.2 µg/mL M1. Solutions became visibly turbid above about 30 µg/mL, and the drug-rich nanodroplets at 56 µg/mL had a mean diameter of 740 ± 45 nm by dynamic light scattering R2·R5.

Supersaturation ratio. The paper defines the supersaturation ratio $S$ through the chemical potential of the dissolved drug:

Here $\mu$ is the chemical potential of the solute, $R$ the gas constant, $T$ the temperature, $a$ the activity, $\gamma$ the activity coefficient and $C$ the concentration; the asterisk marks the values in equilibrium with the crystal E5. In plain terms: $S$ measures how far the dissolved drug is above its crystal in chemical potential. For dilute solutions the activity coefficients cancel and

In plain terms: $S$ is simply the concentration divided by the crystalline solubility. With the measured values, the highest possible $S$ in buffer is 27.9 / 3.6 = 7.8 E6·M1. The relation stops holding above the amorphous solubility, where two phases coexist and the activity of dissolved drug no longer rises C54.

Membrane flux as a measure of free drug. In a side-by-side diffusion cell, drug crosses a membrane from a donor to a receiver compartment. The paper uses

where $J$ is the flux, $dM/dt$ the rate of mass transfer across the membrane, $A$ the membrane area (7.07 cm²), $D$ the diffusion coefficient, $h$ the membrane thickness, $y_m$ the activity coefficient of the drug in the membrane and $y$ that of ritonavir in the donor (the paper prints these coefficients as $y$, while equation 4 uses $\gamma$) E1·E2. In plain terms: since $D$, $A$, $h$ and $y_m$ are the same in every experiment, the rate at which drug appears in the receiver is proportional to the activity of drug in the donor, that is, to the free, molecularly dissolved drug C11. Drug held on silica or inside droplets does not count. Flux therefore rises with concentration up to the amorphous solubility and stays at a maximum above it C54·C20. This is why the authors compare every formulation with amorphous ritonavir made by adding a methanolic stock solution to buffer ("solvent shift" or "antisolvent addition"), which at the amorphous solubility gives the maximum flux possible, 9.20 ± 0.15 ng/min in their set-up R8.a·C20.

Theoretical monolayer coverage. To decide how much drug to load, the authors estimate the amount that would cover the silica surface exactly once, with an equation adapted from Azaïs et al. (2006):

where $X_m$ is the mass of adsorbate per gram of adsorbent for monolayer coverage, SSA the specific surface area of the silica (m²/g), MW the molecular weight of the drug, $S_c$ the contact area of one drug molecule (Ų), $N_A$ Avogadro's number, and ${10}^{20}$ converts m² to Ų E3. In plain terms: divide the silica surface by the footprint of one molecule to get the number of molecules in one layer, then convert to mass. The calculation assumes that the whole surface is available and that ritonavir lies flat, so that its footprint is the product of its two largest dimensions C17.

Dose expressed as a percentage of the amorphous solubility. Throughout the paper a dose is given as the concentration that would result if all drug were released, relative to the amorphous solubility. A 100% dose means enough formulation to give 28 µg/mL ($S$ of 7.8) on full release; a 1000% dose would give 280 µg/mL P6·P10. Percent release is the measured concentration divided by this theoretical concentration.

Methods in detail

The paper is experimental throughout; its one calculation (the monolayer) sets the drug load. Its design is to keep the drug, carrier and buffer fixed and to vary the dose, the medium, the order in which drug, silica and solution meet, and the drug load, reading out free drug by two independent methods A1·A2·A3·A4·A5·A6.

Materials and the loaded silica

The carrier was SBA-15 from Glantreo, with a pore diameter of 7.1 nm, a pore volume of 0.80 cm³/g and a specific surface area of 586 m²/g M3. The model drugs were ritonavir and its structural analogue lopinavir, both from ChemShuttle; hydroxypropyl methylcellulose (HPMC, Pharmacoat 606) at 5 µg/mL served as a crystallization inhibitor where needed, and all aqueous work used 50 mM sodium phosphate buffer, pH 6.8 M1·M2·M4·M8. Ritonavir's dimensions, measured on its crystal structure, are 18.2 × 15.2 × 5.0 Å (Figure 1) M1·P3. With the two largest dimensions as the footprint, equation 3 gives monolayer coverage at 0.253 g/g, which the paper restates as "25.3% w/w", and the formulation for all dose experiments was loaded at 25% w/w R4. With the standard molecular weight of ritonavir, 720.94 g/mol (not printed in the paper), equation 3 reproduces the stated 0.253 g/g E3.

Loading used incipient wetness impregnation: 25 mg ritonavir in 400 µL methanol was added dropwise to 75 mg SBA-15 and mixed with a spatula, then dried at 40 °C for 24 h and under reduced pressure for 48 h; drug content was checked by extraction in methanol and HPLC P4. For the drug-load study, six batches were made at 25 to 150% of the calculated monolayer (Table 2). Their measured drug loads were 4.9 ± 0.3, 8.8 ± 0.1, 16 ± 1, 20.1 ± 0.7, 30 ± 2 and 38 ± 2% w/w against targets of 6.2, 12.4, 18.6, 24.8, 31 and 37.2% w/w P4.

Note by the AI writer (not discussed in the paper). Two bookkeeping points matter for the numbers that follow. First, equation 3 gives grams of drug per gram of silica; 0.253 g/g corresponds to about 20.2% w/w of a loaded formulation, not 25.3%, and Table 2 places 100% coverage at 24.8% w/w, so the "monolayer" is stated on more than one basis X4. Second, the drug content of the batch used in the dose experiments is not printed; the dose masses (9.4 mg in 75 mL for 28 µg/mL) imply about 22% w/w, and the 100%-coverage batch of Table 2 measured 20.1% w/w X5. Percent release values in the paper are therefore relative to a nominal 28 µg/mL.

Solubility measurements

Crystalline solubility was measured by shaking excess ritonavir (form II) in 15 mL buffer at 37 °C for 48 h, centrifuging at 21 100 × g for 30 min and analysing the supernatant by HPLC; it was repeated with 1 mg/mL SBA-15 and in FaSSIF P1. Amorphous solubility was measured by adding 75 µL of a 10 mg/mL methanolic ritonavir solution to 15 mL buffer at 37 °C (300 rpm) and immediately centrifuging the turbid solution at 21 100 × g for 40 min to pellet the drug-rich phase; it was repeated with 5 µg/mL HPMC, with 100 µg/mL SBA-15 (the concentration present during dissolution) and in FaSSIF P2.

Dissolution experiments

The standard test dosed the 25% w/w formulation into 75 mL buffer in a jacketed vessel at 37 °C, stirred at 300 rpm: 9.4 mg for a 100% dose ($S$ of 7.8) or 4.7 mg for a 50% dose ($S$ of 3.9). Samples (0.7 mL) were taken over 120 min, centrifuged at 21 100 × g for 10 min to remove silica, diluted with acetonitrile and analysed by HPLC; each run was done in triplicate P6. The same protocol was used in FaSSIF, dosed at the FaSSIF amorphous solubility (43 µg/mL), and with lopinavir–SBA-15 dosed at the lopinavir amorphous solubility (16.4 ± 0.1 µg/mL) and half of it P6·M2. For doses of 200, 400 and 1000% (theoretical 56, 112 and 280 µg/mL), samples were passed through a 1 µm PTFE filter, which lets drug-rich nanodroplets through, so that both free drug and any nanodroplets would be counted P10·C10.

Four further experiments were designed to probe why release stops:

• Reverse dissolution. Ritonavir was first dissolved at 28 µg/mL, then blank SBA-15 was added at about 100 µg/mL, the silica concentration present in a normal run (duplicate) P8.
• Loaded silica in a saturated medium. The medium was first brought to the amorphous solubility with dissolved ritonavir, then the loaded formulation was added to a total of 56 µg/mL (duplicate) P8.
• Residue. After an $S$ 7.8 run the medium was filtered, the residue dried under vacuum, examined by X-ray powder diffraction (XRPD) and redissolved at $S$ 7.8 (single run) P9.
• Sequential dilution. The formulation was dosed at $S$ 7.8 into 50 mL buffer and, every 30 min, fresh buffer was added so that the theoretical maximum $S$ fell stepwise toward 1 over 180 min, with sampling every 15 min (duplicate) P7.

Diffusion-cell experiments

Donor and receiver compartments of 34 mL buffer at 37 °C were separated by a regenerated cellulose membrane with a 6–8 kDa molecular weight cutoff and an area of 7.07 cm²; 5 µg/mL HPMC was added to both compartments in runs above the amorphous solubility P12. Receiver samples were analysed by HPLC, and the flux was the slope of receiver concentration against time from 30 to 120 min (Figure 9), all with $R^2$ above 0.99 P12·R11. Receiver concentrations stayed below the crystalline solubility, so the receiver remained a sink C12. Each formulation was compared with amorphous ritonavir (or amorphous lopinavir) added from methanolic stock at the same nominal concentration P12.

Note by the AI writer (not discussed in the paper). The values reported as "mass flow (ng/min)" equal the receiver-concentration slopes in ng/mL per min read from Figure 9 (for amorphous ritonavir, about 235 rising to 1055 ng/mL between 30 and 120 min, about 9.1 per min, against 9.20 ng/min); the 34 mL volume and 7.07 cm² area are not applied X3. All comparisons in the paper are ratios, which this does not affect. The paper also calls the membrane "cellulose acetate" in one place (Section on drug load) while the methods say regenerated cellulose X2.

Why two read-outs

Dissolution sampling with centrifugation measures what stays in the supernatant; it can miss or include fine particles depending on the separation step. The diffusion cell measures only drug that is free to cross a membrane. The authors argue that where the two agree, the diffusion cell is "an excellent alternative for measuring free drug release" from complex systems such as OMS (Indulkar et al. 2017) C22. The ratio of formulation flux to amorphous-drug flux at the same dose estimates the fraction of the dose present as free drug; multiplied by 28 µg/mL it gives an estimated free concentration R10.

Experimental analysis

Silica neither solubilizes nor changes the amorphous solubility

With 1 mg/mL SBA-15 present, the crystalline solubility of ritonavir was 3.2 ± 0.3 µg/mL, not significantly different from 3.6 ± 0.6 µg/mL without it R1. With 100 µg/mL SBA-15, the amorphous solubility was 27.5 ± 0.1 µg/mL against 27.9 ± 0.3 µg/mL (p > 0.05), and 5 µg/mL HPMC had no effect R3. The authors read the first result as confirmation that OMS is "a supersaturating delivery system and not a solubilizing one" C14. They explain the absence of any effect on either solubility by the excess of solid or drug-rich phase in these tests, which replaces whatever the silica adsorbs; a solution held at the amorphous solubility without such a reservoir behaves differently, as shown below C44.

Release is fast, then stops at about half

At the 100% dose ritonavir reached about 13.3 µg/mL at 5 min and about 14 µg/mL from 30 min on, staying there for 2 h (Figure 2A): about 50% of the dose R6. At the 50% dose it reached about 9.8 µg/mL, or 70% (Figure 2B) R6. Both solutions were supersaturated (crystalline solubility 3.6 µg/mL), and the plateau was reached within 5 min, which the authors call "arguably faster" than typical polymer ASDs at high drug load and attribute, with the hedge "may be attributed", to the ordered, untortuous pores in which water displaces adsorbed drug C40·C41. A run extended to 240 min, in which withdrawn samples were replaced with buffer, drifted from about 14.3 to 12.9 µg/mL and never approached the crystalline solubility R7.

Membrane flux told the same story (Table 1). At the 100% dose the formulation gave 5.16 ± 0.06 ng/min, 56.1 ± 0.6% of the 9.20 ± 0.15 ng/min of amorphous ritonavir; at the 50% dose it gave 3.80 ± 0.05 ng/min against 4.87 ± 0.05 ng/min for amorphous ritonavir at the same concentration, or 78% R8.a–b·R9. The authors note the "reasonable correlation" between 50% and 70% released and 56% and 78% flux, and take the lower flux to mean lower ritonavir activity in solution, which "can be accounted for by the incomplete release" C21·C22.

Silica takes ritonavir out of solution

The two reverse experiments are the core of the argument (Figures 3 and 4). When blank SBA-15 (about 100 µg/mL) was added to ritonavir dissolved at 28 µg/mL, the concentration fell to about 22.7 µg/mL within 5 min and stayed near 22 µg/mL for 2 h; the discussion gives 22.7 ± 0.5 µg/mL after 15 min, so about 20% was taken up and 80% remained free R13. The authors read this as "suggesting strong adsorptive interactions between ritonavir and SBA-15 in solution" C23. When the loaded formulation was added to a medium already at the amorphous solubility (56 µg/mL in total), the concentration did not rise above 28 µg/mL but fell, to about 17.4 µg/mL at 5 min, recovering only to about 20 µg/mL by 2 h R14. "It is clear", the authors write, that SBA-15 adsorbed additional ritonavir even though it was already loaded at the calculated monolayer C24. From the drop from 28 to about 20 µg/mL they compute that about 600 µg more ritonavir was adsorbed, giving 27% w/w drug on the formulation and, in their words, "multilayer surface coverage" R15·C57.

Crystallization was checked as an alternative cause. The residue left after an $S$ 7.8 run, estimated at about 12% w/w drug, showed only a broad halo in XRPD (Figure 5A), so crystallization during dissolution "appears unlikely" R16·C25. Redissolved at $S$ 7.8, this residue gave about 12 µg/mL after 2 h, 43% of its drug (Figure 5B, single run), which the authors take as evidence "that an equilibrium indeed exists between free drug and adsorbed drug" R17·C4.

Table A. Ritonavir in solution after 2 h in the release and adsorption experiments, pH 6.8 buffer, 37 °C (read from Figures 2–5, approximate; percentages from the text) R6·R13·R14·R17.

The authors also point out a discrepancy themselves. Loaded silica retained about 50% of its ritonavir, but blank silica took up only about 20% of dissolved ritonavir at the same silica concentration C47. McCarthy et al. (2018) had found that drug retained on OMS after sink dissolution exceeded what adsorption isotherms predict R30. The authors conclude that retention "may therefore depend only in part on the adsorption equilibrium" and that further work is warranted C47.

Release stays incomplete under every condition

Dilution. Starting from an $S$ 7.8 dose, the medium was diluted every 30 min (Figure 6). The fraction released rose from 40% in the first 30 min to about 56% after the first dilution, and to about 60–66% from 75 min on, while the concentration fell stepwise from about 11 to 2.1 µg/mL R18. At the end, with the total drug about equal to the crystalline solubility, 2 µg/mL was in solution against a crystalline solubility of 3.6 µg/mL; the authors comment that "close to complete release should have been possible" R18·C27.

Note by the AI writer (not discussed in the paper). The methods list seven $S$ values (7.8, 3.9, 2.6, 1.9, 1.6, 1.3 and 1) for six 30-min periods; the percentages plotted for the last four time points match the plotted concentrations only if $S$ was about 1.3 and then 1, not 1.6 and 1.3 X6. The plateau of 60–65% is unaffected. "Sink" here means a total dose equal to the crystalline solubility, not a medium with excess capacity X7.

Lopinavir. Lopinavir–SBA-15 (25% w/w) dosed at the lopinavir amorphous solubility reached about 7–8 µg/mL, 40–50% release, and 51.4 ± 6.5% of the flux of amorphous lopinavir; at half the dose about 60% was released and the flux was 70.9 ± 4.8% of the dose-matched reference (Figure S1, Table SI) R19·R20. The authors conclude that incomplete release "is not specific to ritonavir" C28. Because lopinavir cannot ionize and ritonavir is un-ionized at pH 6.8 (its thiazole p$K_a$ values are 1.8 and 2.6), and because the pH inside the pores may differ from the bulk through release of silicic acid, they also take the similarity to show that ion exchange "is not responsible" C46·M1.

FaSSIF. Dosed at the FaSSIF amorphous solubility of 43 µg/mL, ritonavir rose more slowly, from 25 µg/mL at 5 min to about 30 µg/mL after 30–45 min, and then stayed constant: about 71% release, significantly more than the about 50% in buffer (p < 0.05) (Figure 7) R21·C30. The authors attribute the gain "likely" to micellar solubilization by taurocholate and lecithin and to better wetting, and they stress that release remained incomplete under every condition tested C48·C2.

Higher doses: more free drug, smaller fraction

To see whether SBA-15 can reach the amorphous solubility at all, the authors dosed 2, 4 and 10 times more formulation (Figure 8). The concentration peaked at 16 µg/mL at the 200% dose and reached about 24 µg/mL at 400% and 1000%, about 85% of the amorphous solubility R22. No dose produced more than the amorphous solubility in solution, which the authors read as release that "approaches but does not exceed" it, with no drug-rich phase formed C32. Because free drug rose so little, the fraction released fell steeply, from about 67% at the 50% dose to about 9% at the 1000% dose (Figure S2) R23.

Table B. Dose series for ritonavir–SBA-15 in pH 6.8 buffer, 37 °C. "% released" is relative to each dose; "% dissolution" and "% mass flow" in Table SII are both relative to amorphous ritonavir at 28 µg/mL (Figure S2 read approximately; Table 1, text and Table SII; mean ± SD, n = 3) R23·R24·R8·R9.

The flux results agree up to 400%: 6.29 ng/min at 200% (68.4%) corresponds to about 19 µg/mL free drug, and 7.62 ng/min at 400% (82.8%) to about 23 µg/mL, close to the dissolution values R8.c–d·R10. The diffusion data, the authors write, "confirm" that release at 200% does not reach the LLPS boundary C31. At 1000%, however, the flux was 9.12 ± 0.17 ng/min, 99.2% of amorphous ritonavir, which "suggests" that enough drug was released to reach the LLPS concentration of 28 µg/mL, while dissolution showed about 24 µg/mL R8.e·C33. The authors note the discrepancy and add that flux shows when the amorphous solubility is reached but not how far release goes beyond it C34. Plotted against dose, flux rises nonlinearly (Figure 10, natural-log trend line, $R^2$ = 0.979), which they attribute to incomplete release R12·C55.

Note by the AI writer (not discussed in the paper). The 1000% samples were filtered through 1 µm, which should pass drug-rich nanodroplets of about 740 nm, so a dissolution value below 28 µg/mL is hard to reconcile with a flux at the amorphous maximum; whether the LLPS concentration was reached at 1000% remains unresolved X14. The paper's general statement that dissolution "appear[s] to underestimate" free drug relative to flux holds at four of five doses but not at 400% (86.5 vs 82.8%) C36·X8. The Table SII percentages at 120 min are also a little lower than the 120-min points of Figures 2 and 3 and the rounded text values, and Table SII gives n = 3 where Figures 3 and 4 give n = 2 X9.

Loading above the monolayer lowers flux

Each of the six loadings was dosed into the diffusion cell at the amorphous solubility (Figure 11). All gave significantly less flux than amorphous ritonavir (p < 0.05), 45–58% of it R27. The four loadings up to the calculated monolayer (4.9, 8.8, 16.0 and 20.1% w/w) did not differ from each other, at 5.1 ± 0.2 to 5.4 ± 0.3 ng/min (about 56–58%); the two above it (29.8 and 38.1% w/w) gave 4.37 ± 0.06 and 4.10 ± 0.08 ng/min (45–48%) R27. XRPD showed no crystalline ritonavir even at 38.1% w/w (Figure S4) R26. The authors see in the coincidence of the drop with the calculated monolayer a sign "that the calculation approach is reasonable" and that surface coverage governs release C38.

Table C. Flux against drug load, each formulation dosed at the amorphous solubility (Table 2; mass flow read from Figure 11, approximate, except the two printed values; percentages read from the right axis) P4·R27.

Note by the AI writer (not discussed in the paper). No loading lies between 20.1 and 29.8% w/w, so the data place the break somewhere in that interval; whether it sits at 20.2% w/w (0.253 g/g on a formulation basis) or at 25.3% w/w cannot be decided from them X16·X4.

Discussion

The authors' explanation: an adsorption–desorption equilibrium

The authors write the release step as an equilibrium:

In plain terms: water molecules displace drug from the silica surface, but dissolved drug can also go back onto the surface, so the amount in solution settles where the two rates balance E4. As the dose, and with it the theoretical supersaturation, rises, the equilibrium "shifts further to the left", toward adsorbed drug, which the authors offer as the reason the fraction released falls with dose C53. They add a thermodynamic argument: the chemical potential of dissolved drug equals that of drug in any phase in equilibrium with it, so if adsorption to silica lowers the drug's chemical potential, the solution cannot reach the activity of amorphous drug; since adsorption is favoured at higher solution concentration, "it is clear that as $S$ increases, the driving force for the drug to desorb from the silica surface and enter solution is reduced" C56·E5. The uptake of drug by already-loaded silica from a saturated medium is their main evidence for this lowered chemical potential R14·R15·C56.

The authors also rule out the obvious alternatives. Crystallization is excluded because ritonavir crystallizes slowly, no run showed desupersaturation toward the crystalline solubility (including the 4 h run), the residue was amorphous and they found no report of OMS inducing nucleation; they state that adsorption "is considered to be the mechanism" C45·R25·R16. Ion exchange is excluded through lopinavir C46.

Comparison with polymer ASDs

The authors set their results against earlier work on polymer ASDs of ritonavir, which released 100% under supersaturating conditions (Ilevbare and Taylor 2013) and exceeded the amorphous solubility to form drug-rich colloids R29·R35. From this contrast they conclude that "it would appear that OMS systems are not able to achieve the same extent of supersaturation" as polymer ASDs, because release is incomplete, and that the observations "point to the very different thermodynamic and structural properties of amorphous drug adsorbed to SBA-15" C43·C52. They add that the solid-state stability of OMS, its main advantage over most polymer ASDs, is lost if the drug is not released C42. This comparison rests on cited data; no ASD was tested in this paper.

Drug load and pore filling

The literature is split on how drug load affects release from OMS: three itraconazole studies found more release at higher loads, one study the opposite R33·C58. Here flux fell above the calculated monolayer, and the authors propose that "the data suggest that drug release might be physically obstructed at higher drug loading levels": with more complete filling or blocking of the pores, water may not reach the drug–silica interface to displace drug, and drug may not get out; collisions of drug with water or pore walls may also become more likely C60. They cite Mellaerts et al. (2013), where etravirine loaded at 5, 10 and 20% w/w in SBA-15 released to about the same extent but 30% w/w (above an estimated 25% w/w monolayer) released less, as support R34·C60. They conclude that "caution must be exercised when aiming for high drug loading levels in OMS delivery systems" C62. Pore filling, water access and the state of the drug in the pores were not measured.

In vivo

Because release stayed incomplete even in FaSSIF, the authors "speculate that drug release from SBA-15 may be incomplete in vivo" C49. They then cite Vialpando et al. (2016), where an OMS formulation of flubendazole released the least drug in FeSSIF but gave the highest bioavailability in rats among three amorphous technologies, and suggest that in the gut absorption may keep removing dissolved drug and so drive further desorption, giving bioavailability "that cannot be predicted from in vitro studies" R32·C50. Both statements are hypotheses; no in vivo work or absorptive sink was part of this paper.

What it suggests for formulation

The authors' own recommendations are to measure free drug from silica formulations with a diffusion cell as well as by dissolution, to use caution with loads above monolayer coverage, and more generally to investigate the supersaturation potential of enabling formulations systematically C22·C62·C65.

Note by the AI writer (not discussed in the paper). For ritonavir in SBA-15 the dose–response is flat: a tenfold dose was needed to reach the flux that amorphous ritonavir gives at its amorphous solubility R8. Percent release from such a carrier is therefore not a property of the formulation alone; it depends on the dose and the medium volume, so formulations should be compared at matched dose and preferably with a free-drug read-out. Loading at or below the calculated monolayer gave the higher flux in this system R27.

Limitations

Limitations stated by the authors. The authors say that flux data cannot show the extent of release beyond the amorphous solubility, so it is "not possible to discern if a drug-rich phase was generated, or to what extent" at the highest dose C34. They acknowledge that forward release and reverse adsorption do not match and that retention "may therefore depend only in part on the adsorption equilibrium" C47. They describe the adsorptive interactions as "hypothesized" and call further studies "clearly warranted" C66, and, since buffer and FaSSIF are simpler than intestinal fluid, they leave open whether release is also incomplete in vivo C49·C50.

The AI writer's own judgement.

• The central finding, incomplete release that falls with dose, is well supported: two independent read-outs agree in trend, the effect is large, and lopinavir and FaSSIF give the same picture R6·R8·R23·R19·R21.
• The mechanism is inferred, not measured. No adsorption isotherm, no measurement of drug remaining on the silica and no surface analysis were reported, so the equilibrium and the lowered chemical potential rest on solution concentrations alone C4·C56.
• Several numbers need care. The mass-flow unit is effectively ng/mL per min X3; percent release rests on a nominal 28 µg/mL while the drug content of the main batch is not printed X5; the monolayer is stated in g/g and in % w/w as if equal X4; the discussion once gives the lower dose as $S$ 3.6 instead of 3.9 X1; and the dilution schedule does not match the plotted percentages at the last time points X6.
• Dissolution and flux disagree at the 400% and 1000% doses, and the question whether the amorphous solubility was reached at 1000% is open X8·X14.
• The residue redissolution was run once, and the adsorption experiments in duplicate R17·R13·R14.
• One silica (7.1 nm pores, 586 m²/g), one loading method and two structurally related drugs were studied.

Conclusion

This paper measured how much supersaturation a drug-loaded ordered mesoporous silica actually delivers. Ritonavir in SBA-15 supersaturated the medium within minutes, but at a dose equal to the amorphous solubility only about half of it came off, membrane flux was 56% of that of amorphous ritonavir, and the fraction released fell from about 67% to about 9% as the dose rose twentyfold; the amorphous solubility was approached but, by dissolution, never exceeded R6·R8·R23·R22. Blank silica took ritonavir out of solution and loaded silica took up more from a saturated medium, which the authors read as an equilibrium between adsorbed and dissolved drug in which adsorption lowers the drug's chemical potential; in their conclusions, strong ritonavir–silica interactions "result in incomplete drug release under all investigated conditions" R13·R14·C4·C56·C63. Above the calculated monolayer, flux fell further, possibly because filled pores hinder water access R27·C60. For formulators, the paper shows that silica carriers of this kind give a fast but capped supersaturation, and that membrane flux is a useful check on what dissolution tests report.