When the permeation enhancer sets the pH: insulin release from enhancer-rich minitablets in simulated gut fluids

Background and motivation

Therapeutic peptides combine high target selectivity, high potency and low inherent toxicity, and since the 1920s they have been developed for oncological, metabolic and genetic diseases (Wang et al. 2022, ref 1). Patients prefer oral dosing, but peptides are a poor fit for the gut: digestive enzymes degrade them, and their size and hydrophilicity limit permeation across the intestinal epithelium, so oral bioavailability is typically below 2% (Nicze et al. 2024, ref 2). Most peptides are therefore still injected. Nanoparticles, hydrogels and ionic liquids have been proposed (refs 3–5), but the authors note that their clinical and commercial viability is "yet to be proven" (Section 1).

The approach that has reached patients is the addition of a permeation enhancer (PE) to the formulation. PEs include bile salts, medium-chain fatty acids and chelating agents. They act either paracellularly, by opening the tight junctions between epithelial cells, or transcellularly, by increasing the hydrophobicity of the peptide or by disrupting the cell membrane. This route led to the approval of oral semaglutide and oral octreotide (Maher and Brayden 2021, ref 6). Two PEs dominate clinical work: sodium caprate (C10) and salcaprozate sodium (SNAC). SNAC is the enhancer in Rybelsus, the oral semaglutide tablet approved in 2019, and the FDA quality review lists Rybelsus as 75% w/w SNAC plus microcrystalline cellulose (MCC) and polyvinylpyrrolidone (PVP) in undisclosed amounts (ref 13). Most of the research effort on C10 and SNAC has gone into their efficacy, mechanism of action and safety (Kim et al. 2022, ref 7; refs 18–21). The authors state the gap plainly: it is not understood how adding these enhancers to a tablet changes the release of the peptide from the dosage form, nor how the composition of the gastrointestinal milieu affects that behaviour.

Two physicochemical properties make this question more than bookkeeping. First, C10 and SNAC are both sodium salts of weak acids with a pKa of about 5. At least two pH units above the pKa they are extensively ionised and highly soluble; below the pKa they exist mainly as poorly soluble free acids (Twarog et al. 2019, ref 8; Fattah et al. 2020, ref 9). Their own dissolution should therefore differ sharply between gastric and intestinal fluid. Second, at the high concentrations found at the surface of a dissolving tablet, SNAC has been reported to raise the pH of its surroundings, lifting simulated gastric fluid from acidic towards neutral (Buckley et al. 2018, ref 10). What was not known is how that shift depends on the medium, and how an elevated pH at the tablet surface feeds back onto the release of the peptide sitting in the same tablet.

The peptide side of the problem is equally pH-sensitive. Peptide solubility depends on pH, buffer type and ionic strength (Grossmann and McClements 2023, ref 11). It passes through a minimum at the isoelectric point (pI), the pH at which the molecule carries no net charge, and rises one to two pH units above and below it, where the molecule carries net negative or net positive charge (Tokmakov et al. 2021, ref 12). For human insulin the pI is about 5.3 (Thomas and Thevis 2019, ref 14). Fasted intestinal fluid sits at pH 6.5 and fed intestinal fluid at pH 5.0, both within about one pH unit of that minimum, while fasted gastric fluid sits at pH 1.6, far from it. On paper, then, insulin release from a tablet should be governed not only by the pH of the fluid it meets but also by any pH change the enhancer itself produces.

The study was built to test that proposition in vitro with human insulin as a model peptide. The authors first measured the equilibrium solubility of insulin, C10 and SNAC in simulated fasted and fed intestinal fluid and simulated fasted gastric fluid, together with the blank versions of those media (the same media without sodium taurocholate and lecithin), and recorded the pH that excess enhancer produced in each. They then prepared minitablets with a high enhancer load by direct compression, compared insulin release from them with release from enhancer-free minitablets in each medium, and finally put a drop of universal indicator solution on the dry tablets to look for a difference in surface pH (Section 1).

Background concepts

Permeation enhancer. An excipient whose job is not to carry the drug but to make the gut wall more permeable to it, either by loosening the tight junctions between epithelial cells (paracellular route) or by acting on the cell membrane itself (transcellular route). C10 is sodium caprate, listed in the paper's abbreviations as sodium decanoate, a medium-chain fatty acid salt. SNAC is salcaprozate sodium, the N-acylated amino acid derivative used in Rybelsus. Both are surface-active weak-acid salts, so their behaviour in a dissolution vessel includes ionisation, micelle or aggregate formation, and precipitation of the free acid at low pH.

Isoelectric point and the pH–solubility curve of a peptide. A peptide carries acidic and basic groups. At the pI the positive and negative charges balance, electrostatic repulsion between molecules is at its weakest, and solubility reaches a minimum; moving the pH away from the pI in either direction charges the molecule and raises solubility. For insulin (pI about 5.3) this predicts low solubility across the intestinal pH range and high solubility in acid or in mild alkali. The practical consequence for formulation is that the pH the peptide actually experiences, not the pH written on the dissolution medium, sets the ceiling on how much can dissolve.

Biorelevant media and their blank versions. Fasted state simulated intestinal fluid (FaSSIF), fed state simulated intestinal fluid (FeSSIF) and fasted state simulated gastric fluid (FaSSGF) reproduce the pH, buffer system, ionic strength and colloidal content of gut fluid. FaSSIF contains 3 mM sodium taurocholate (NaTC) and 0.75 mM lecithin in a phosphate buffer at pH 6.5; FeSSIF contains 15 mM NaTC and 3.75 mM lecithin in an acetate buffer at pH 5.0 with 203 mM NaCl; FaSSGF contains only 0.08 mM NaTC and 0.02 mM lecithin with HCl to pH 1.6 (Table 1). The bile salt and lecithin form mixed micelles. The blank versions leave both out while keeping pH and buffer, which separates a colloidal effect from a pH effect.

Microenvironmental (surface) pH. As a tablet dissolves, a thin, poorly stirred layer forms at the solid–liquid interface in which the dissolving solids approach saturation. If one of them is a strong enough acid or base, that layer sits at its own pH rather than the pH of the bulk, so a 40 mL vessel can read unchanged while the layer the drug dissolves into is several pH units away. This is the mechanism the authors invoke, and the reason they treat the pH of a saturated enhancer suspension as a ceiling for the surface pH.

Specific ion effect. At the same nominal ionic strength, different salts suppress protein solubility to different degrees, a salting-out series that reflects ion–water and ion–protein interactions. The authors use this to account for a result that proximity to the pI alone cannot explain, citing evidence that dihydrogen phosphate salts proteins out more strongly than acetate (Mazzini and Craig 2017, ref 15).

Coning. In a paddle apparatus at low stirring speed, insoluble excipient powder from a disintegrating tablet can pile up under the paddle instead of dispersing. The mound is hydrodynamically dead, so drug inside it sees little fresh medium; it is a recognised artefact for formulations rich in insoluble fillers such as MCC (Terashima and Ozeki 2024, ref 16).

Temporal synchronicity of enhancer and peptide. An enhancer only works while its local concentration at the epithelium exceeds a threshold, and the peptide has to be present in that same window before the enhancer is absorbed, diluted or carried onward by transit (refs 22–24). Media composition can interfere: sodium dodecyl sulfate was less effective in fed than in fasted luminal conditions, attributed to sequestration by micelles (Roos et al. 2019, ref 26), whereas C10 enhancement was reported to be unaffected by bile salts and phospholipids (Berg et al. 2022, ref 27). Peptides can bind micelles too: octreotide and, less so, desmopressin interacted with micelles in biorelevant media, which lowered their flux (Dening et al. 2021, ref 31). This literature frames the paper's closing argument about staggered release.

The paper prints no numbered equations; every quantity in it is a directly measured solubility, pH or percent released.

Methods in detail

The work divides into three contributions. The first builds a physicochemical map: the equilibrium solubility of insulin and of each enhancer in six simulated media, plus the pH that an excess of each solid produces in those media, plus insulin solubility in buffer at the two pH values the saturated enhancer suspensions reach. The second measures insulin release from three minitablet formulations in all six media, with two control experiments designed to separate a hydrodynamic limitation from a solubility limitation. The third follows the enhancer itself rather than the peptide: its release profile, its recovery once the vessel is made basic, and a colour test for the pH at the tablet surface. The flow below shows how the three feed into the final argument.

Contribution 1: A solubility and pH map of peptide and enhancer in biorelevant media

① Problem. Any argument about release has to start from how much insulin and how much enhancer can dissolve in each region of the gut, and from what pH a concentrated enhancer imposes on the fluid around it. Neither had been measured for C10 and SNAC across simulated gastric and intestinal media, and SNAC's reported pH effect had been shown only for gastric fluid.

② What they did. Excess insulin, C10 or SNAC was added separately to 1 mL of FaSSIF, FeSSIF, FaSSGF or the blank version of each in 2 mL Eppendorf tubes, shaken at 350 rpm for 24 h at 37 °C, then centrifuged at 18626 RCF for 2 min before assay; all in triplicate (Section 2.2). Insulin was measured in the six media and, additionally, in 0.1 M phosphate buffer isotonic with FaSSIF at pH 7.6 and pH 9.2, chosen to match the pH of saturated SNAC and C10 solutions respectively; phosphate has no buffering capacity near pH 9.2, so the authors verified that it held that pH for at least 48 h. The pH of every saturated suspension was read with a pH probe after the same equilibration (Section 2.6). Quantitation throughout the paper used size-exclusion HPLC (AdvanceBio SEC 130 Å, 2.7 µm, 7.8 x 300 mm; isocratic L-arginine 1 g/L, acetonitrile and 0.15 M phosphate buffer, 65/20/15% v/v, at pH 7.5; detection at 210 and 280 nm) (Section 2.5).

③ Why it works. Putting excess solid into the medium is exactly the condition that exists in the diffusion layer of a tablet that is 72% enhancer, so the pH of a saturated suspension is a defensible upper bound for the pH at a dissolving surface. Measuring insulin solubility in a plain buffer held at those same pH values isolates the pH variable: it answers how much more insulin could dissolve if the local pH were 7.6 or 9.2, without the enhancer being present to confound the measurement. Running every medium in a blank version alongside the biorelevant one separates the contribution of bile salt and lecithin from that of pH and buffer.

④ How it differs from earlier work. The pH-raising effect of SNAC was known for simulated gastric fluid only (ref 10). Here both enhancers are tested in three media with different buffer systems and starting pH values, and insulin solubility is then measured at the resulting pH. That pairing is what the earlier literature did not supply.

Contribution 2: Insulin release from enhancer-rich and enhancer-free minitablets

① Problem. A solubility map predicts that insulin release should depend on the medium, and that a high enhancer load should override that dependence. The prediction has to be tested on a dosage form that resembles the marketed one rather than on a dispersion of powder.

② What they did. Three formulations were blended in a pestle and mortar and compressed by hand on a Fette E1 press with 4 mm tooling to 22 ± 3 mg, hardness above 1 MPa (Section 2.3). The enhancer minitablets held 72% C10 or 72% SNAC, 20% MCC, 5% PVP K30 and 3% insulin by weight, chosen to approximate Rybelsus (75% w/w SNAC with MCC and PVP); the enhancer-free minitablet held 92% MCC, 5% PVP K30 and 3% insulin. Dissolution used a Distek 2100B with 100 mL vessels and paddles, 40 mL of medium and 75 rpm; 350 µL samples were drawn at 1, 5, 10, 15, 30, 45, 60 and 120 min with equal-volume replacement and were centrifuged at 18626 RCF for 2 min before assay; triplicate (Section 2.4). All three tablets were run in all six media. Two further sets used the enhancer-free tablet only: in blank FaSSIF at 150 and 250 rpm, and in 0.1 M phosphate buffer at pH 7.6 at 75 rpm. No dissolution temperature is printed anywhere in the paper, although the solubility work states 37 °C.

③ Why it works. Insulin is fixed at 3% of a 22 mg tablet in 40 mL, so the dose and the maximum attainable bulk concentration are the same in every run; any difference between curves belongs to the tablet or the medium. The two controls are the informative part of the design. Raising the paddle speed changes only the hydrodynamics, so a recovery implicates dispersion; switching to a pH 7.6 buffer changes only insulin's solubility, so a recovery implicates solubility.

④ How it differs from earlier work. Earlier studies of C10 and SNAC characterised permeation, efficacy and safety. This is a release study: the same tablet composition is carried through six media that span pH 1.6 to 6.5 and two bile-salt levels, with the enhancer loaded at a level typical of a marketed product rather than at a trace level.

Contribution 3: The fate of the enhancer, and a colour test for the surface pH

① Problem. The explanation offered for the release data is a pH rise at the tablet surface. The bulk pH of the vessel does not change, so the explanation needs independent support. In parallel, the enhancer's own release profile matters, because an enhancer that has left the tablet before the peptide cannot be at the epithelium with it.

② What they did. C10 and SNAC were assayed in the same dissolution samples as insulin, giving enhancer release profiles over 120 min in each medium. After the 120 min sample was drawn, 1 M NaOH was added to the FaSSIF, FeSSIF and FaSSGF vessels of the enhancer tablets to raise the pH above 10, and the enhancer was assayed again to close the mass balance. The appearance of the vessels was noted during the runs and after basification. For the surface pH, a single drop of universal indicator solution, which is acidic and red, was applied to one dry blank, one C10 and one SNAC minitablet, outside any medium, and the colour was photographed with a Canon EOS 1500D against a printed pH 4.0 to 10.0 colour scale (Sections 2.6 and 3.2.3).

③ Why it works. The basification step converts a negative result into a positive one. If the enhancer is missing from an acidic sample because it precipitated as the free acid and was spun down, raising the pH above 10 should redissolve it and recover the full dose; if it were still locked in an undissolved tablet, base would not help on that timescale. The change from turbid to clear supports the same reading. The indicator puts a reagent whose colour depends only on pH onto a surface that is almost entirely enhancer.

④ How it differs from earlier work. Quantifying both the peptide and the enhancer from the same vessel, and then closing the enhancer mass balance by changing the pH, is a more complete accounting than a peptide-only release profile. Note by the AI writer (not discussed in the paper): the indicator test is also the weakest link in the chain, because it is run on a dry tablet outside any medium, with no dissolution medium to buffer against the pH change, and only one representative image is shown.

Experimental analysis

Setup in brief. Three tablet formulations, six media, 40 mL, 75 rpm, 120 min, triplicate, with insulin and enhancer quantified by size-exclusion HPLC from the same centrifuged samples; two control series on the enhancer-free tablet, one end-of-run basification series, and one indicator photograph of three dry tablets. The paper prints no numeric data tables: only four insulin solubilities (0.1, 0.8, 1.2 and 2.1 mg/mL) and the two phosphate-buffer values (14.0 and 23.7 mg/mL) appear as numbers. Every other value below was read from a plot, which is why "about" is attached to it and why differences smaller than about 5 percentage points should be treated as within scatter.

Table A. Saturation solubility after 24 h at 350 rpm and 37 °C, n = 3 (from Figure 1; the four intestinal insulin values are printed in the text and the two buffer values in the abstract, all other values read from Figure 1, approximate).

Result 1: insulin solubility tracks distance from the pI, with a colloidal contribution at intestinal pH. Insulin was least soluble in blank FaSSIF at 0.1 mg/mL, pH 6.5 being about one unit above the pI of 5.3, and most soluble in the two gastric media at about 23 mg/mL, pH 1.6 being nearly four units below it. Plotted against pH (Figure 3), the values form a minimum near the dashed pI line and rise on both sides; the authors' wording is that "as expected, it appeared" that solubility rose with distance from the pI, and that proximity to the pI was "the dominant factor". Two secondary effects sit on top of that trend. Bile salt and lecithin raised solubility 8-fold in FaSSIF (0.1 to 0.8 mg/mL) and about 2-fold in FeSSIF (1.2 to 2.1 mg/mL) but made no difference in the gastric media, which the authors read as insulin interacting with micellar species, the nature of the interaction being "unclear". And insulin was more soluble in FeSSIF (pH 5.0) than in FaSSIF (pH 6.5), the reverse of what the pI alone predicts; the authors say "it is likely that the unexpected behavior observed was due to a specific ion effect", phosphate salting proteins out more strongly than acetate, and add that the twice higher NaCl in FeSSIF "may have" contributed. Neither explanation was tested: no binding measurement and no single-salt series were run.

The buffer experiments give the number that drives the rest of the paper. At pH 7.6 insulin dissolved to 14.0 mg/mL and at pH 9.2 to 23.7 mg/mL, that is 140 and 240 times the 0.1 mg/mL of blank FaSSIF, with the pH 9.2 value about twice the pH 7.6 value and comparable with the gastric media. The enhancer solubilities move in the opposite direction to insulin: both C10 and SNAC are highest in FaSSIF and lowest and most variable in the gastric media, as their pKa of about 5 predicts, and bile salt and lecithin "did not appear" to change them, which the authors take to mean "pH dominates PE solubility, rather than the mixed micelles". The absolute levels matter for the surface-pH argument: C10 dissolves to 117–260 mg/mL, more than an order of magnitude above SNAC at about 10–20 mg/mL.Card note: the enhancer bars in Figure 1 are labelled with the starting pH of the medium (6.5, 5.0, 1.6), but those same saturated suspensions had equilibrated to pH 7.6–9.2, so the labels identify the medium rather than the pH at which the solubility was measured (extractor note X2).

Table B. pH of the saturated suspensions after the same 24 h equilibration, n = 3 (SNAC and C10 values printed in the text; insulin values read from Figure 2, approximate). The blank media gave similar values, so only the biorelevant ones are plotted.

Result 2: excess enhancer overrides the medium. Table B is the sharpest result in the paper. Three media whose starting pH values span 4.9 units, buffered by phosphate, by acetate and by hydrochloric acid respectively, all end up within 0.4 units of each other once excess C10 is present (pH 8.8–9.2) and within 0.3 units once excess SNAC is present (pH 7.6–7.9). FaSSGF moves the furthest, from 1.6 to 9.2 with C10. Excess insulin, by contrast, leaves each medium at its starting pH. The authors call this "a significant increase", say the data "confirm" the earlier report that SNAC lifts simulated gastric fluid from acidic to neutral and extend it to C10, whose rise was larger, and conclude that the result "suggested that the ability of C10 and SNAC to alter pH was independent" of the composition and starting pH of the medium. The weak point is translation rather than measurement: these are 24 h suspensions of excess solid in 1 mL, not the diffusion layer of a tablet in 40 mL.Card note: the Table 1 row labelled "buffer capacity (Osm/L)" carries 270, 670 and 120, which are the osmolality figures usually quoted for these media rather than buffer capacities in mmol per litre per pH unit, so the buffer capacity should be read as unreported (extractor note X1).

Table C. Percent insulin released at 120 min, n = 3 (read from Figures 4, 5, 6 and 7, approximate).

Result 3: without enhancer the medium decides; with 72% enhancer it does not. Reading down the first column of Table C, the enhancer-free tablet separates into two groups: about 103–104% in the two gastric media and about 67–83% in the four intestinal media, a gap of roughly 20–35 percentage points.Card note: the abstract quotes 67 to 82% for the simulated intestinal media, which matches the two biorelevant panels of Figure 4 but not their blank equivalents, where the plot gives about 72 and 83% (extractor note X7). The shapes differ as well as the heights: in gastric media about 51% (blank FaSSGF) and 38% (FaSSGF) was out at 1 min and 98–99% by 30 min, whereas the intestinal curves start near 0% at 1 min and reach their plateaus by about 30 min. The authors say insulin "appeared to exhibit solubility-dependent release", which matches the solubility ranking, and then raise the objection themselves: the highest bulk concentration attainable in the vessel is far below insulin's solubility in every medium, so the reason for the shortfall was "unclear". They observed undissolved material settling at the bottom of the vessels and "suspected that incomplete disintegration of the minitablets may have restricted insulin release", pointing to coning as a known artefact of formulations rich in insoluble excipients such as MCC.

Reading across the other two columns, the medium effect disappears. Both enhancer tablets released about 88–92% of their insulin at 120 min in all six media, with the three curves within each panel of Figure 7 overlapping. The intermediate points are also close: C10 tablets reached about 70–77% at 15 min and 80–86% at 30 min, SNAC tablets about 72–82% at 15 min and 80–87% at 30 min, with maximum release at about 30 min in every medium. In FaSSGF the SNAC tablet had already released about 45% at 1 min. The authors' wording is that the medium "did not appear to affect the release of insulin" from the C10 tablets, that SNAC behaved likewise, and that "it is clear that the presence of C10 and SNAC had a significant impact", in direct contrast with the enhancer-free tablet.Card note: the abstract states "greater than 90%" release from the enhancer tablets, while the plotted 120 min values run from about 88 to 92% (extractor note X7, and see R12 and R13).

What the controls rule out, and what they do not. The last three rows of Table C are the most informative in the paper, because they attack the enhancer-free shortfall from two independent directions. Raising the paddle speed in blank FaSSIF from 75 to 150 to 250 rpm lifted release at 120 min from about 72 to 85 to 94%, while the time to the plateau stayed near 30 min at every speed; the authors note that a "very high" speed of 250 rpm was needed to exceed 90%, and read the result as insoluble MCC settling instead of dispersing and possibly having "trapped insulin above its saturation solubility, thereby preventing its dissolution" into the bulk. Changing the medium instead of the hydrodynamics gave a similar recovery: in phosphate buffer at pH 7.6 the same tablet released about 79% at 15 min and 91% at 120 min against about 56 and 72% in blank FaSSIF. With insulin about 140 times more soluble at pH 7.6, the authors conclude "it is clear that solubility played an important role" and that the complete gastric release "offers a plausible explanation for the limited release" in intestinal media. Note by the AI writer (not discussed in the paper): the two controls do not discriminate between the explanations, since both succeed. Stirring recovers about 22 of the roughly 28 missing percentage points with no change in solubility, and a pH change recovers about 19 of them with no change in stirring. That pattern fits a local, coupled limitation inside the settled bed, not a bulk solubility ceiling: 3% of a 22 mg tablet in 40 mL cannot exceed about 0.017 mg/mL, below even the lowest measured solubility of 0.1 mg/mL.

Table D. Percent enhancer released from its own minitablet, and percent recovered after the vessel was made basic at the end of the run, n = 3 (read from Figures 8 and 9, approximate). Gastric SNAC points scatter between about 2 and 16% at individual time points.

Result 4: the enhancer leaves first, and in acid it leaves the solution. In the four intestinal media C10 was essentially fully released within 10–15 min (about 82–93% at 10 min, about 95–102% at 15 min), and SNAC reached a plateau of about 66–84% over the same period, both ahead of insulin's maximum at about 30 min.Card note: the results section puts maximum enhancer release at about 15 min, while the discussion gives 10 min for C10 and 20 min for SNAC; treat the time to maximum enhancer release as 10 to 20 min (extractor note X8). In the two gastric media the picture inverts: no C10 was detectable at any time point, and SNAC never exceeded about 16%, with large scatter. The vessels were turbid in gastric media and clear in intestinal media during the runs. The authors "suspected" that at low pH C10 converts to its unionised free acid, which is known to form an oily precipitate (ref 17) that centrifugation removed before assay, and think "it is possible" that SNAC also formed a second phase; poor disintegration is "unlikely" because insulin release from the same tablets was near-complete. Adding 1 M NaOH at the end of the run recovered about 99–103% of the C10 and about 93.5–96.5% of the SNAC, and the gastric vessels turned clear, read as the second phase being "disrupted at the elevated pH"; in the intestinal media the recovered SNAC is attributed to colloidal self-assemblies or precipitate ("is suspected"). On that basis "it is proposed" that both enhancers were released rapidly and completely, but were not freely available as monomers or micelles, "which may restrict their efficacies".Card note: complete recovery of both enhancers is stated in the text, while Figure 9 gives about 93 to 96% for SNAC with error bars that stay below 100% (extractor note X9).

Result 5: the surface pH evidence. The authors state that there was "no direct evidence" for a pH change during dissolution, since the bulk medium showed no detectable change in pH before and after a run, and argue that the pH insulin meets at the surface could be "potentially reaching as high as pH 9.2" with C10 and 7.9 with SNAC. The indicator test is their substitute for a measurement. The text reports no colour change on the blank tablet, dark green on the C10 tablet and pale green on the SNAC tablet, which it reads as a surface pH close to 8 and 7 respectively, far above the blank, and concludes that this "likely accounts for" the difference in release.Card note: the printed photograph shows a red-orange blank, a pale green C10 tablet and a yellow-green SNAC tablet, and yellow-green sits near pH 6 on the printed scale, so the assignments of pH 8 and pH 7 should be checked against the original colour image before being used (extractor note X6). Two caveats belong here: only one representative image is shown, so there is no measure of variability, and the drop sits on a nearly pure enhancer surface with no medium present, the most favourable case possible for a large pH shift.

Comparison across systems, and what the design does not separate. Three comparisons in the data are clean. Blank against biorelevant versions of the same medium: enhancer-free release at 120 min differs by at most about 5 percentage points (72 against 67%, 83 against 82%, 103 against 104%), so the bile salt and lecithin that raise insulin solubility 8-fold in FaSSIF do not measurably raise release. C10 against SNAC: C10 raises the pH about 1.3 units higher and dissolves more than ten times better, yet both tablets give the same insulin release within scatter, about 88–92% in every medium. Insulin against enhancer in the same vessel: the enhancer finishes 10–20 min before the peptide. The comparison that is not clean is the one carrying the main conclusion. The enhancer-free tablet contains 92% MCC and the enhancer tablets 20%, so the two differ in insoluble filler as much as in enhancer, and the authors themselves blame settled MCC for part of the enhancer-free shortfall. A filler-matched blank, or a series of lower enhancer loadings, would have separated the two effects; neither was run.

Discussion

The chain of reasoning. The authors assemble their explanation in four steps. As an enhancer-rich tablet begins to dissolve, enhancer at the solid–solution interface reaches concentrations that overwhelm the buffer of the surrounding medium. The local pH then approaches that of a saturated enhancer solution, up to about 9.2 for C10 or 7.9 for SNAC. Insulin at that local pH is 140 to 240 times more soluble than in blank FaSSIF, so "this elevated pH likely caused a considerable increase in the saturation solubility of insulin", and release becomes near-complete. Since the enhancer, not the fluid, sets that pH, "if correct, this hypothesis proposes" that the pH insulin experiences early in dissolution is independent of the medium, which is what the release data show. The strength of the language rises through the paper: "hypothesized" in the results, "compelling evidence to suggest" in the discussion, and "was determined" in the abstract and conclusions. The card flags that escalation; the middle wording is the defensible one.

Where the approach applies. The boundaries are narrow: one peptide (human insulin, pI about 5.3), two enhancers, one enhancer level (72% w/w), one tablet geometry (22 mg, 4 mm, direct compression, hardness above 1 MPa), one apparatus and volume (paddle, 40 mL), and 120 min of release. The media span pH 1.6 to 6.5 with phosphate, acetate and hydrochloric acid buffers; a medium with a higher buffer capacity, or a larger volume per tablet, would resist the local pH shift more strongly, and nothing in the data says where that resistance begins. The dissolution temperature is not printed, so comparison with other release studies carries that uncertainty. Within those bounds the finding that matters for practice is simple: at a loading typical of a marketed enhancer product, the enhancer is also the pH modifier of the formulation, and the medium stops mattering.

What is unresolved. First, the surface pH was never measured during a run. The whole explanation rests on a saturated-suspension pH measured in a tube, an insulin solubility measured in an enhancer-free buffer at that pH, and a colour photograph of dry tablets. A microelectrode or a pH-sensitive fluorophore in the diffusion layer would convert the hypothesis into a measurement. Second, the comparator is confounded: the enhancer-free tablet differs in MCC content, and the authors' own coning explanation for its shortfall means that part of the release difference may be a disintegration and dispersion difference rather than a pH difference. Third, insulin solubility in the presence of dissolved C10 or SNAC was not measured. The authors raise this themselves, noting that enhancers form peptide-specific noncovalent and hydrophobic ion pairs that lower peptide solubility (refs 35, 10), so "it is not clear without further studies" whether the behaviour transfers to other peptides. For insulin itself, the same ion pairing could work against the solubility gain the pH shift provides. Fourth, the second phase formed in acid was never identified; it is inferred from turbidity, from the known behaviour of decanoic acid, and from the recovery after basification. Fifth, no permeability, cell or animal experiment tests whether the staggered release lowers insulin uptake.

What it suggests for further work. The authors make corelease the priority: because insulin peaks at about 30 min and the enhancers at about 10–20 min, the staggered release "would likely result in incongruous presentation at the epithelial walls", and research "should" focus on bringing the two together, for example by slowing enhancer dissolution or speeding peptide dissolution. They also flag that an enhancer sitting in precipitate or colloidal aggregates rather than as monomer or micelle may not be available to act on the epithelium, so its association state in each medium becomes a design variable. A third line follows from their data rather than their text: C10 and SNAC gave the same insulin release despite differing by about 1.3 pH units and by an order of magnitude in solubility, which suggests the release benefit saturates once the local pH clears the pI by a unit or two. If so, a lower enhancer load might buy the same benefit, which matters because enhancer load drives tablet size.

For formulation practice. Three things follow directly. A weak-acid-salt enhancer at high loading should be treated as an alkalising pH modifier, so the peptide's pH–solubility curve belongs alongside the enhancer's saturated-solution pH (about 8.8–9.2 for C10 and 7.6–7.9 for SNAC in these media) when the formulation is set; where the pI sits near intestinal pH, the pairing can make release largely independent of the prandial state. Dissolution methods need care: a centrifugation step will under-read enhancer release in acid, a final basification step closes the mass balance, and a filler-rich comparator at 75 rpm can generate a coning artefact that looks like a solubility limit. And an enhancer-free comparator should match the filler, otherwise the pH effect and the filler's dispersion behaviour cannot be separated. Note by the AI writer (not discussed in the paper): the authors frame the pH rise as beneficial, and for release it is; a sustained local pH near 9 at an epithelial surface is a separate question that this in vitro design cannot address.

Limitations

Limits the authors state. The authors are explicit that there was "no direct evidence" of a pH change during dissolution, because measurement of the bulk media before and after dissolution "showed no detectable change in pH"; the surface pH is introduced as something that "would result in" a higher local value, "potentially reaching as high as pH 9.2". The central mechanism is offered as a hypothesis in the results section ("hypothesized to be due to high concentrations of C10 or SNAC") and as "compelling evidence to suggest" in the discussion, and its generalisation is prefaced "if correct, this hypothesis proposes". For the enhancer-free tablets they state that the reason for the incomplete release was "unclear", given that the concentrations attainable during dissolution were far below insulin's saturation solubility in every medium, and they offer coning as a suspicion ("it was suspected that incomplete disintegration of the minitablets may have restricted insulin release"). The missing enhancer in acid is likewise a suspicion: C10 "was suspected" to have converted to the free acid and precipitated, and "it is possible that SNAC also formed a second phase". Transfer to other peptides is left open, given peptide-specific ion pairing with enhancers.

The AI writer's own judgement. Two things weaken the conclusions beyond what the authors say. The first is the confounded comparator. The enhancer-free tablet holds 92% MCC against 20% in the enhancer tablets, so filler content and enhancer content change together, and the authors' own coning explanation makes that difference causally relevant rather than incidental. The stirring control makes this concrete: raising the paddle speed alone, with no change in pH or solubility, recovered release from about 72% to about 94%, which is almost the whole gap that the pH argument is invoked to explain. With a single enhancer level and no filler-matched blank, the data support the statement that a 72% enhancer tablet releases insulin completely and medium-independently, but they do not isolate how much of the difference from the blank is the pH. The second is that every link in the surface-pH chain is measured somewhere other than where it is claimed to act: the pH in a 24 h suspension of excess solid, the insulin solubility in an enhancer-free buffer, the colour on a dry tablet with no medium present. The bulk pH, the only pH measured during a run, did not change.

Several smaller issues bound the precision of the numbers. Nearly all values were read from plots; only six solubility numbers are printed, and no numeric data tables exist. Release values exceed 100% in places, up to about 105% for insulin in gastric media and about 104% for C10, and the gastric SNAC points scatter between about 2 and 16%, so differences smaller than about 5 percentage points should not be interpreted. The abstract's "greater than 90%" release and its 67–82% intestinal range are both slightly tighter than the plotted values, and the stated complete recovery of SNAC is about 93–96% in Figure 9. The time to maximum enhancer release is given as 15 min in the results and as 10 and 20 min in the discussion. The enhancer solubility bars carry the starting pH of the medium rather than the pH the suspension reached, the Table 1 "buffer capacity" row carries osmolality figures, and the dissolution temperature is not reported. None of these invalidates the central observation, but together they put the resolution of the study at the level of ranking and order of magnitude.

Conclusion

This is a release study rather than a permeation study, and its value lies in showing that a permeation enhancer at a loading typical of a marketed product is not a neutral passenger in the dissolution of the peptide beside it. Excess C10 drove simulated gastric and intestinal fluids to pH 8.8–9.2 and excess SNAC to pH 7.6–7.9, whatever the starting pH and buffer; insulin, whose pI is about 5.3, is 140 to 240 times more soluble at those pH values than in blank FaSSIF. Enhancer-free minitablets released insulin in a medium-dependent way, about 103–104% in gastric media against about 67–83% in intestinal media, while minitablets with 72% C10 or SNAC released about 88–92% in all six media. The authors attribute the difference to a pH rise at the tablet surface, an explanation that is coherent with every measurement they report and directly demonstrated by none of them, since the bulk pH never changed and the indicator test was run on dry tablets.

The second finding deserves equal weight for formulation: both enhancers left the tablet 10 to 20 min before the peptide, so the dosage form delivers its enhancer and its peptide to the epithelium at different times, which "would likely" blunt the enhancement the enhancer exists to provide. Added to the observation that in acid the enhancer sits in a precipitate or colloidal phase rather than as free monomer, this points the development problem away from enhancer potency and towards release engineering: matching the two release profiles, and keeping the enhancer in an available state where it is needed. For the pH mechanism itself, the obvious next experiment is to measure the pH in the diffusion layer during a run, with a filler-matched comparator and a series of enhancer loadings.