When an amorphous product crystallizes on the shelf: are the crystals it grows as harmful as reference crystals?

Background and motivation

An amorphous solid dispersion (ASD) holds a poorly soluble drug in an amorphous state, usually dispersed in a hydrophilic polymer. On dissolution it can produce a supersaturated solution, that is, a concentration above the solubility of the crystalline drug. The polymer has two jobs: it improves wetting and dissolution, which at low drug load can push concentrations above the crystalline and even above the amorphous solubility, and it inhibits crystallization both in the solid matrix and in the supersaturated solution (Introduction, refs 1–10) C1. Absorption depends on how high the supersaturation is and how long it lasts, so crystallization has to be prevented or delayed in the solid product and during dissolution C2.

Crystallization runs in two steps, nucleation and growth. Primary nucleation, in the absence of crystals, must overcome a free-energy barrier because a new surface has to form. Crystals that are already present act as templates for growth and can trigger secondary nucleation, for which the barrier is minimal. If nothing inhibits it, seed crystals in an ASD therefore speed up crystallization and the loss of supersaturation, which "may" reduce bioavailability (Mullin 2001, ref 11) C3. From a quality and regulatory point of view, a company then has to know how much a given level of crystallinity in an amorphous product costs in exposure.

The usual way to find out is to add known amounts of reference crystals to a fully amorphous dispersion and test it C4. Reference crystals are made by crystallization from a solvent, under well-controlled conditions. Crystals that grow inside an ASD during storage form under very different conditions, in a solid matrix full of polymer that is there precisely to inhibit crystallization. Que et al. had shown that paclitaxel seed crystals with different size distributions, available growth area and interface structure grew at different rates when added to a dissolving ASD, which changed the whole supersaturation–time profile (Que 2018, ref 12) C5. Crystals grown among inhibiting polymers "might be imperfect" or poisoned at their growth sites (refs 13–15), so it is "plausible", in the authors' word, that such endogenously derived crystals behave differently from the reference crystals used to stand in for them C6.

Tacrolimus, an immunosuppressant, offers a real case. The innovator capsule, Prograf, contains amorphous tacrolimus and gave much better bioavailability than crystalline drug (Yamashita 2003, ref 17); it is dose proportional from 3 to 10 mg (Bekersky 1999, ref 18), which suggests the amorphous formulation is not solubility limited in that range C7. A generic amorphous product, made by Intas for Accord Healthcare, is known to crystallize at moderate temperature and high humidity (Trasi 2017, ref 16) C7. The authors' group had already shown in vitro that even small amounts of crystalline tacrolimus in this generic product caused significant desupersaturation in non-sink dissolution (Purohit 2018, ref 19) C8. What was missing was the in vivo side: the impact of partial crystallinity on the in vivo performance of amorphous formulations "has not been widely explored", and the one recent study they cite had blended reference crystals with an ASD (Knopp 2018, ref 20) C9·C44.

The paper therefore sets out to measure, in dogs, the exposure from amorphous, partially crystallized and maximally crystallized generic tacrolimus capsules, with Prograf (fresh and stored) as a control, and to compare the crystals formed inside the product with reference crystalline tacrolimus monohydrate A1·A4.

Background concepts

Supersaturation and its upper limit. The paper defines supersaturation as a ratio (Discussion, p.11) E1:

In plain terms: $S$ says how many times the actual solution concentration $C$ exceeds the concentration at saturation $C^{*}$, the crystalline solubility. An amorphous solid can push $C$ up to its own, higher solubility, so the largest supersaturation it can give is the ratio of the two solubilities E2:

In plain terms: the amorphous solubility caps the concentration, and dividing it by the crystalline solubility gives the maximum supersaturation. For tacrolimus in the dissolution medium used here, the amorphous solubility is 50 µg/mL and the crystalline solubility is given as 1.4 µg/mL in the text and 1.5 µg/mL in the Figure 2 caption, both cited from the group's earlier work (refs 19, 23) M3. The authors state the ratio as "approximately 33" C30. That number matches 50/1.5 = 33.3; with the text value of 1.4 µg/mL the ratio would be 35.7, a small inconsistency inside the paper that the card records E2·X1.

Non-sink dissolution and the dose-to-volume ratio. A dissolution test is "non-sink" when the medium cannot dissolve the whole dose at the crystalline solubility, so supersaturation and its loss can be watched. The authors chose the dose and volume so that the dose, if fully dissolved, would just reach the amorphous solubility (Results, p.9) E3·C15:

In plain terms: 5 mg of tacrolimus in 100 mL gives at most 50 µg/mL, exactly the amorphous solubility, so the amorphous drug can dissolve completely to the highest supersaturation it can sustain without exceeding it. The authors call this "at-sink" with respect to the amorphous solubility. They chose it because maximizing the possible supersaturation also maximizes the chance of crystallization, so differences between formulations in their robustness to crystallization show up clearly C35.

Two ways in which crystals in a product can act. Drawing on earlier work (Sun 2016, ref 33), the authors distinguish two scenarios C32. In the least harmful one, the crystals in the product neither nucleate new crystals nor grow, for example because a polymer inhibits them well. The amorphous fraction dissolves to a concentration above the crystalline solubility, so the crystals have no driving force to dissolve; they just reduce the effective dose until absorption has lowered the local concentration below the crystal solubility. In the more damaging scenario, the product's crystals seed further crystallization by secondary nucleation and growth, at rates set by the supersaturation and by any inhibitors present. Which scenario applies is the key to how much crystallinity costs.

Endogenous and reference crystals. In this paper, "endogenously derived" crystals are crystals that formed inside the dosage form, here during storage at 40 °C/75% RH. "Reference" crystals are the purchased tacrolimus monohydrate, made by solvent crystallization C6. The question is whether the second can stand in for the first.

The two products. Both capsules contain about 5 mg tacrolimus and about 5 mg hydroxypropyl methylcellulose (HPMC), with lactose monohydrate, croscarmellose sodium (CCS) and minor excipients such as magnesium stearate M1·M2. Based on earlier studies, Prograf is an intimately mixed ASD of tacrolimus with HPMC, whereas Accord is "(mainly)" a physical mixture of amorphous tacrolimus with HPMC and the other excipients (refs 16, 17) C21. In Accord, tacrolimus is only about 3.7 wt% of about 135 mg of capsule contents, of which about 100 mg is lactose M2.

Pharmacokinetic read-outs. $C_{max}$ is the highest blood concentration, $T_{max}$ the time at which it occurs, and AUC_0-24 the area under the blood concentration–time curve from 0 to 24 h, a measure of total exposure. Table 2 gives means with the standard error of the mean (SEM, not to be confused with scanning electron microscopy) for six dogs R9.

Methods in detail

The study has one simple design idea: instead of adding reference crystals to an amorphous product, let the real product crystallize, set the crystallinity by blending crystallized and fresh product, and compare the result in vitro and in dogs with the fresh product, a non-crystallizing innovator and a suspension of reference crystals at the same dose P1·P3·P7. The figure below shows the flow.

Crystallizing the product and measuring its crystallinity

Capsules of both products were taken out of their primary packaging, placed one per 4 mL glass scintillation vial and stored at 40 °C and 75% RH; Accord capsules were opened periodically and their contents scanned by X-ray powder diffraction (XRPD; Rigaku TTRAX III, Cu Kα, 50 kV, 100 mA) P1·P2. Crystallinity was read from the diffraction peak at 9.3° 2θ. The calibration was built by adding purchased tacrolimus monohydrate to fresh Accord contents at 10 to 100% crystallinity in 10% steps, with 5 mg (the drug content of a capsule) taken as 100% P2. The calibration was linear from 15 to 100% crystallinity (R² = 0.9905), with an uncertainty of about ±8% P2. The authors note that quantification was hard because tacrolimus makes up only about 3.7 wt% of the capsule contents C12. Potency of fresh and stored capsules was measured with the USP monograph assay: the pooled contents of 10 capsules were dissolved to 50 µg/mL tacrolimus in acetonitrile and aqueous Brij-35, and the summed peaks of tacrolimus and its 19-epimer were compared with the USP reference standard by HPLC at 205 nm P4.

Setting the crystallinity by blending

Once the stored Accord capsules had crystallized "to the maximum extent", which the authors describe as close to 100% given the XRPD error, the contents of 10 such capsules were pre-mixed to even out capsule-to-capsule differences and then blended geometrically with fresh Accord contents to give 0, 50 and 100% crystalline tacrolimus P3·C10. For dog dosing the blends were refilled into size 3 clear gelatin capsules; for dissolution the contents were used directly P3. Prograf capsules, fresh and stored for 5 weeks at 40 °C/75% RH, were used as they were P3. Note by the AI writer (not discussed in the paper): the 50% level is set by the mixing ratio and is not reported as checked by XRPD afterwards, and "100%" means maximally crystallized on a scale calibrated with reference crystals X8·X11.

Non-sink dissolution

Capsule contents holding 5 mg tacrolimus were added to 100 mL of USP test 1 medium, water at pH 4.5 (adjusted with phosphoric acid) containing 50 µg/mL hydroxypropyl cellulose (HPC), in jacketed beakers at 37 °C, stirred at 300 rpm with cross-shaped magnetic bars P5. Samples were taken by syringe at 10 to 120 min, passed through 1 µm glass filters and analysed by HPLC at 210 nm; every test was run in triplicate P5. As shown in the concepts section, this dose and volume let the amorphous drug reach, but not exceed, its solubility of 50 µg/mL E3·C15.

Looking at the crystals: SEM and XRPD

To see the drug particles in a capsule that is mostly lactose, the authors vortexed the contents of fresh and stored Accord in 10 mL water for 2 min, which dissolves the lactose (about 100 mg of the 135 mg), filtered the suspension through a 0.45 µm cellulose membrane, dried the solids (mainly tacrolimus and CCS) for 4 h under vacuum and imaged them by scanning electron microscopy (SEM) after platinum coating P6·C11. Reference tacrolimus monohydrate was sieved through a 110 µm screen and imaged the same way P6. Figure 3 also shows the residue of fresh Prograf, although the method section lists only the Accord samples and the monohydrate R6·X13. The crystal form of the material that crystallized in the capsules was checked by XRPD against the reference R8.

The dog study

One group of six male and female beagle dogs (6–11 kg) received each of six formulations as a single 1 mg oral dose in successive dosing periods: fresh, 50% and 100% crystalline Accord, and fresh and stored Prograf, each weighed out as formulation equivalent to 1 mg and filled into clear gelatin capsules, plus a suspension of 1 mg tacrolimus monohydrate (particles below 110 µm) in 50 µg/mL HPMC, dosed in a size 12 capsule P7. The dogs had free access to water P7. Blood was taken from the jugular vein at 0.25, 0.5, 1, 1.5, 2, 3, 4, 6, 9, 12 and 24 h and stored at −20 °C P7. Whole blood was mixed with zinc sulfate containing internal standard, precipitated with methanol and analysed by LC-MS/MS (AB Sciex API 6500, transition m/z 821.3/768.4) against standards prepared in dog blood P8. Cmax and AUC were compared between formulations with a paired one-tailed t-test, with significance at p ≤ 0.05 P7·R9. The paper does not state the washout interval, the dosing order or whether the dogs were fasted X12.

Note by the AI writer (not discussed in the paper): because every dog received every formulation, each comparison is within-animal, which suits tacrolimus, whose exposure varies widely between individuals. The reference arm, however, differs from the crystallized product in more than crystal origin: it is a neat suspension in dilute HPMC, without the lactose, CCS and roughly 1:1 HPMC of the capsule matrix X9.

Experimental analysis

Result 1: the generic capsule crystallizes within four weeks; the innovator does not

Stored open at 40 °C/75% RH, the Accord capsules showed little crystallinity at day 9, then a steady rise, and a jump to the plateau between day 26 and day 28 (Figure 1, Table A) R1. The authors describe the capsules as "maximally crystalline after 4 weeks" and stored all test capsules for 5 weeks to make sure the maximum was reached C13·P1. Prograf showed no tacrolimus crystallization after 35 days under the same conditions (data not shown) R2.

Table A. Crystallinity of tacrolimus in Accord capsules during storage (read from Figure 1, approximate; the type of error bar and n are not stated) R1.

Two details matter for reading Table A. The first two points, about 2% and 11%, lie below the 15% lower end of the linear calibration, and the plateau reads slightly above 100%, within the stated ±8% uncertainty P2·X8. Storage did not degrade either product: potency was 100.1 ± 1.4% fresh and 99.4 ± 3.4% stored for Accord, and 96.4 ± 1.8% and 95.7 ± 1.2% for Prograf (mean ± SD, p > 0.05; Table 1) R3. The authors read this as no significant degradation, and as amorphous tacrolimus being "clearly stable" under these conditions, since it stayed amorphous in Prograf with no loss of potency C14. So any exposure loss in the crystallized product cannot be blamed on chemical degradation.

Result 2: product crystallinity lowers the concentrations reached in dissolution

Table B. Tacrolimus concentration (µg/mL) during non-sink dissolution of 5 mg in 100 mL pH 4.5 medium with 50 µg/mL HPC at 37 °C (read from Figure 2, approximate; triplicate). Amorphous solubility 50 µg/mL; crystalline solubility 1.4–1.5 µg/mL R4·M3.

Both fresh products came close to the amorphous solubility, but by different routes (Figure 2, Table B) R4. Fresh Accord dissolved fast, reaching about 44 µg/mL at 10 min and about 49 µg/mL at 20–30 min, then lost concentration from 1 h on, ending at about 20 µg/mL at 2 h R4. The authors ascribe this decline to crystallization during the test on "undetectable residual crystalline seeds" of tacrolimus monohydrate present in the fresh product, as described in their earlier work (ref 16) C17. Prograf dissolved more slowly, reaching about 41 µg/mL at 40 min and ~47 µg/mL later, which the authors expect from its lower potency, and showed no sign of crystallization over 2 h R4·C16·C19.

The crystallized products behaved differently again. The 50% crystalline product peaked at about 25 µg/mL, roughly half the amorphous solubility, and fell to 15 µg/mL at 2 h R4. The authors read this as the crystalline half being unable to dissolve once the amorphous half has pushed the concentration above the crystalline solubility, with further crystal growth "may occur" during the test C36. The 100% crystalline product rose to about 5 µg/mL at 10 min and ended at 4 µg/mL (about 3.7 in the figure), still above the crystalline solubility R4. In the authors' words, crystallinity in the product is "extremely detrimental" to the concentrations reached in this test C34.

The paper describes the 100% curve in two ways. The Results section says that the 50% and 100% crystalline products "also lead to crystallization during the dissolution test" (p.9) C18. The Discussion says that the maximally crystallized sample keeps a "constant and low" concentration, "indicating" that whatever can dissolve does so quickly and that no further crystallization can occur (p.12) C37. The figure shows a slow fall from about 5.1 to 3.7 µg/mL over 110 minutes, and no analysis of the solid after dissolution is reported, so the question stays open R4·X2.

That the fully crystallized product stays above the crystalline solubility is used later in the argument. The concentration is "always higher than the crystalline solubility" over 2 h, "by a factor of 2.5" relative to the solubility of the reference crystals R5. The factor matches 3.7/1.5; with the text values (4 and 1.4 µg/mL) it would be about 2.9 E1·X1. The authors explain the absence of further crystallization in this sample by its low supersaturation together with the HPMC that comes from the dosage form into the medium: crystallization inhibitors have been shown to work better at low than at high supersaturation (refs 12, 34–37), which "provides an explanation" for HPMC failing in fresh Accord but working in the nearly fully crystallized sample C39.

Result 3: crystals grown in the product are much smaller than reference crystals

Table C. What SEM of the washed capsule residues and of the reference crystals shows (Figure 3 and text p.10) R6·C22.

Tacrolimus crystals appeared only in the stored Accord, and they were "much smaller" than the reference crystals: 1–5 µm against 5–25 µm (Table C) R6·C20. The text gives an average size of 10.5 µm and a D50 of 6.5 µm for the reference powder without saying how they were measured R6·X13. XRPD showed that the crystals formed in the capsules are the same form as the reference, tacrolimus monohydrate (patterns not shown), so differences between the two cannot come from a different polymorph R8·C49. The SEM also shows how different the two fresh products are: the residue of fresh Accord "indicates" porous 5–10 µm fragments of amorphous tacrolimus, in the authors' words, while that of Prograf shows ~50 µm particles that "appear to be" undissolved ASD presumed to consist of tacrolimus and HPMC (Figure S1 shows HPMC particles for comparison) C22·R7. Note by the AI writer (not discussed in the paper): the panels are at very different magnifications, with scale bars of 4–5 µm for the Accord samples and 50–100 µm for Prograf and the reference crystals, so the images have to be read with the scale bars R6.

Result 4: in dogs, crystallinity in the product lowers exposure, but only about two-fold

Table D. Pharmacokinetic parameters after a 1 mg oral dose, mean (SEM), n = 6 (Table 2). Letters mark significant differences at p ≤ 0.05: a against fresh Accord, b against 50% crystalline Accord, c against 100% crystalline Accord. The crystalline suspension differed from all formulations in Cmax, Tmax and AUC (p < 0.05). The Cmax column is printed with the unit "ng"; the values are concentrations in ng/mL R9·X7.

Crystallinity in the product cut Cmax sharply: from 25.3 ng/mL for fresh Accord to 9.31 and 3.84 ng/mL for the 50% and 100% crystalline products, with all three pairs marked as significantly different (Table D) R9. AUC fell less: from 42.1 to 22.9 and 21.6 ng·h/mL R9. The authors note that although the "100% crystalline" product is maximally crystallized, its AUC drops "only by a factor of two" C28·R11. In Table 2 the AUC difference is marked significant for the 50% product, but the 100% product's AUC carries no letter against fresh Accord, so the abstract's statement that both crystallized products gave lower AUC rests, for the 100% product, on the means rather than on a marked test C54·X3.

The two crystallized products reach about the same AUC by very different routes (Figure 4a, 4b) R10. Fresh Accord peaks sharply at about 23 ng/mL at 1 h. The 50% product also peaks early, at about 7.6 ng/mL at 1 h (Tmax 0.96 h), and is already measurable at 0.25 h. The 100% product rises to a flat plateau of about 2 ng/mL between 1 and 3 h (Tmax 2.8 h) and stays near 1 ng/mL to 12 h R10·R9. The authors first consider that the equal AUCs "might suggest" further crystallization of the 50% product in vivo, but read its fast absorption as showing that at least part of its remaining amorphous fraction dissolves rapidly and is absorbed C50. For the 100% product they read the blunted Cmax as "suggesting" a prolonged period of absorption along the gastrointestinal tract, which "may contribute" to its AUC matching the 50% product C51.

Note by the AI writer (not discussed in the paper): the mean curves read from Figure 4, integrated by the trapezoid rule from 0.25 to 24 h, give about 42, 22, 21.5, 33, 44 and 5.3 ng·h/mL for the six formulations, close to the Table 2 AUCs, so the figure readings and the table agree R10·R9. The same check suggests that the Figure 4 error bars, which the caption calls standard deviations, are more likely SEMs: the upper bar at the fresh Accord peak is about 11 ng/mL above the mean, close to the Table 2 SEM of Cmax (10.5) and well below the standard deviation it implies (about 25.7) X6. The caption also speaks of plasma, whereas whole blood was analysed X6.

Result 5: the crystallized product still gives about four times the exposure of reference crystals

The suspension of reference crystals gave the lowest exposure of all: Cmax 0.69 ng/mL, Tmax 3.7 h and AUC 5.29 ng·h/mL, significantly different from every other formulation (Table D) R9. In Figure 4e no drug appears in blood until 1.5 h; the curve peaks at about 0.67 ng/mL at 4 h and falls to about 0.16 ng/mL at 12 h R10. The fully crystallized Accord, against this, gave Cmax 3.84 ng/mL and AUC 21.6 ng·h/mL R9. In the authors' summary, the maximally crystallized sample keeps about half the AUC of the fresh product, while the crystalline suspension gives about one eighth; the endogenously crystallized tacrolimus yields an AUC about four-fold higher than the reference crystals, and "does not have as much impact on the biological exposure as might be anticipated" R11·C46. The Table 2 means bear these ratios out (AI arithmetic): 21.6/42.1 = 0.51, 5.29/42.1 = 0.13 and 21.6/5.29 = 4.1 R9·R11.

The authors propose three explanations, without choosing ("This suggests either"): amorphous tacrolimus cannot fully crystallize in the powder and the remainder is below the XRPD detection limit; the crystals formed are highly defective; or their small size allows fairly rapid redissolution, which partly offsets the conversion to the crystal C47. They find the first two "somewhat supported" by the dissolution result, in which the fully crystallized product stays 2.5 times above the reference crystalline solubility for 2 h C48·R5. Because the crystal form is the same as the reference, a different polymorph is ruled out as a cause C49.

Result 6: fresh generic and innovator give similar exposure; the stored innovator is unchanged

Fresh Accord and fresh Prograf gave similar AUCs, 42.1 and 33.7 ng·h/mL, and fresh Accord and stored Prograf did not differ significantly in AUC or Cmax (Table D) R9·C25·C26. Prograf peaked later and lower: Cmax 12.3 ng/mL at 1.6 h against 25.3 ng/mL at 0.83 h for fresh Accord (Figure 4c) R9·R10. The authors call the lower Prograf Cmax "most likely" a result of its slower dissolution (Figure 2) and say its longer Tmax "appears to reflect" that slower dissolution, which shows that "not all amorphous formulations dissolve at the same rate" C27·C43. Table 2 does not mark the Cmax difference between the two fresh products as significant X5. Stored Prograf gave Cmax 16.8 ng/mL, Tmax 1.3 h and AUC 44.0 ng·h/mL R9. The authors call this "comparable" to fresh Prograf, "confirming the robustness of the innovator product to drug crystallization", and they read the lower AUC of stored Accord than stored Prograf as "consistent with" Accord crystallizing and Prograf not C55·C25. Table 2 reports significance only against the three Accord arms, so neither the fresh–stored Prograf comparison nor the stored Accord–stored Prograf comparison is tested in the printed table X4.

Discussion

How the authors connect the results

The authors' reading links the dissolution and dog results through the level of supersaturation. Their chain is shown below; the good and bad cases are the two scenarios of the concepts section, applied to these products C32·C39.

The bad branch rests on the decline seen for fresh Accord and the 50% blend in Figure 2 and on residual seeds reported earlier, the good branch on the flat, above-solubility curve of the 100% product and on literature showing that inhibitors work better at low supersaturation C17·C36·C37·C39. Neither crystal growth nor the HPMC concentration in the medium was measured in this paper C36·C39.

Why the dissolution test overstates the in vivo effect

The authors stress that it is "hard to directly correlate" the dissolution results with the dog results, for two reasons C40. First, the dose-to-fluid volume ratio in vivo is unknown and keeps changing. The dogs received 1 mg with free access to water; the gastric fluid volume of dogs has been reported as 45 mL and the small intestinal volume as ~50 mL (GastroPlus v 9.0), so "it is likely that some supersaturation is theoretically possible", but it "may not" reach the maximum C40·R12. Second, after the drug leaves the stomach, absorption across the gut wall keeps removing it, which "may" lower supersaturation further, weaken the drive for secondary nucleation and growth, and let the crystalline portion dissolve once luminal concentrations fall low enough C41.

Note by the AI writer (not discussed in the paper): with the volumes the authors cite, 1 mg in 45 mL is about 22 µg/mL and 1 mg in 95 mL about 10.5 µg/mL, below the 50 µg/mL amorphous solubility measured in the pH 4.5 test medium (the solubility in dog intestinal fluid is not given) R12·M3. The in vitro test, which puts 5 mg in 100 mL, therefore drives the system to a supersaturation the dogs probably never reach, which fits the authors' explanation of why crystallization looks more damaging in the beaker than in the animal E3·C40.

The clearest example is fresh Accord. It loses supersaturation in vitro, yet gives the same AUC as fresh Prograf, which does not R4·R9. The authors conclude that crystallization of fresh Accord "either does not occur, or has no significant impact in vivo": given its short Tmax, it is "likely" that most of the absorbable drug is absorbed before any crystallization, or crystallization is slower in vivo, "perhaps" because supersaturation is lower C42. A second example is the pair of crystallized products: in vitro the 100% product ends far below the 50% product (4 against 15 µg/mL), but in dogs their AUCs are the same (21.6 against 22.9 ng·h/mL) R4·R9.

Why endogenous crystals differ from reference crystals

The authors argue that crystals grown inside a solid matrix, among excipients that "may substantially impact" crystal size and quality, can have very different properties from reference material grown from a solvent, and that this is why endogenously derived crystallinity has to be studied as such C45. The SEM result, 1–5 µm crystals in the product against 5–25 µm reference crystals, supports the size part of this argument directly R6·C20. The other parts, residual amorphous drug and crystal defects, are offered as possibilities with no direct measurement C47·C48.

The authors also caution that the smaller impact of endogenous crystals "may be highly dose and compound dependent" C52. In another study by the group (Wilson 2018, ref 30), an ASD that crystallized rapidly on contact with the dissolution medium gave a lower AUC than the reference crystalline material, and there the crystals formed from the ASD were larger than the reference crystals; in other cases the crystals formed may even be a different crystal form C52. They therefore conclude that "more research is warranted" on partial crystallinity in ASDs, and that this study is "particularly relevant given the ease of inducing crystallization in the Accord product" C53.

What it suggests for formulation and quality assessment

The authors' own conclusion is that the nature of the crystalline material formed decides its in vivo impact: it is "insufficient to simply assess the impact of crystallinity per se", and crystalline material tested in vivo "should closely resemble" the crystals expected to form in the product C57. Conventional mixing of neat crystalline drug with an amorphous formulation "may not adequately mimic" crystals formed in the matrix C58. They expect these observations to improve risk assessment of residual crystallinity in amorphous products C59. They also draw a second lesson from Prograf: not all amorphous formulations dissolve at the same rate, and making a drug amorphous is not the only consideration for performance C43.

Note by the AI writer (not discussed in the paper): two practical points follow. First, the direction of the error from using reference crystals is not fixed: here a reference-crystal surrogate would have overstated the exposure loss about four-fold, while in the group's ref 30 case crystals formed from the ASD were worse than reference crystals. Characterizing the size, form and residual amorphous content of the crystals that form in a stressed product is a cheap step before choosing a surrogate. Second, the formulation microstructure mattered more than anything else for this drug: the intimately mixed Prograf dispersion did not crystallize at 40 °C/75% RH in 35 days, while the mainly physical mixture in Accord was fully crystalline in about four weeks R1·R2·C21.

Limitations

Limits the authors state. They describe quantifying crystallinity as "challenging" because tacrolimus is only about 3.7 wt% of the capsule contents, with an uncertainty of about ±8% crystallinity C12·P2. They say the dissolution results are "hard to directly correlate" with the in vivo results, because the in vivo dose-to-fluid ratio is unknown and absorption keeps removing drug C40·C41. They note that the smaller impact of endogenous crystals "may be highly dose and compound dependent" and that "more research is warranted" C52·C53. Their explanations for the four-fold difference are offered as alternatives ("This suggests either"), not as measured results C47.

The AI writer's own judgement (1): what the reference arm compares. The headline comparison is between the crystallized product and a neat suspension of reference crystals in 50 µg/mL HPMC P7. The crystallized product also carries about 1 mg HPMC per mg of drug, lactose and croscarmellose sodium, and the authors themselves credit the HPMC from the dosage form with holding the supersaturation of the crystallized product in vitro M2·C39. A blend of reference crystals with fresh or excipient-only Accord, the practice the introduction describes, was not dosed. The four-fold AUC difference therefore measures crystal origin and formulation matrix together, and the paper's main recommendation rests on it X9·C57.

The AI writer's own judgement (2): statistics and nominal levels. Table 2 marks the AUC difference between fresh and 100% crystalline Accord as not significant, although the abstract and conclusions describe a reduced absorption for both crystallized products X3. The table tests nothing between fresh and stored Prograf, so the robustness of the innovator rests on the absence of crystallization and unchanged potency plus similar mean PK X4. The crystallinity levels are nominal: "50%" is a mixing ratio not checked afterwards, and "100%" is read on a scale calibrated with reference crystals, which may respond differently from small, possibly defective crystals X8·X11. The study used six dogs, one dose of 1 mg and one generic product, and the washout, dosing order and feeding state are not reported X12.

The AI writer's own judgement (3): smaller inconsistencies in the source. The crystalline solubility is 1.4 µg/mL in the text and 1.5 µg/mL in the Figure 2 caption; the stated maximum supersaturation of about 33 and the factor of 2.5 match the second value X1. The Results and the Discussion disagree on whether the fully crystallized product crystallizes further during dissolution X2. The Figure 4 caption names plasma and standard deviations, whereas whole blood was analysed and the bars behave like SEMs; the Cmax column of Table 2 carries the unit "ng" X6·X7. The SEM method lists no Prograf images though Figure 3 shows them, the table-of-contents graphic labels the Accord product an "ASD" though the text calls it mainly a physical mixture, and the method for the reference particle size statistics is not given X13. Finally, the dissolution samples were filtered through 1 µm glass filters while the endogenous crystals are 1–5 µm, so any sub-micrometre solid passing the filter would add to the above-solubility plateau that the authors use as support for the defective-crystal explanation X10. None of these changes the main ranking of the formulations.

Conclusion

This paper asks a practical question that the usual risk test answers by proxy: if an amorphous product crystallizes on the shelf, how much exposure is lost? By crystallizing the generic tacrolimus product itself and dosing blends to the same six dogs, the authors show that crystallinity in the product lowers concentrations in a non-sink test and lowers Cmax and AUC in vivo, but that the fully crystallized product still keeps about half of the fresh AUC, about four times more than a suspension of reference crystals at the same dose R4·R9·R11. The crystals formed in the product are smaller than the reference crystals and of the same form R6·R8. The authors conclude that the nature of the crystals formed matters and that crystals used to assess risk in vivo should resemble those expected in the product C57. The mechanism behind the difference, whether residual amorphous drug, defective crystals or fast redissolution, remains untested, and the comparison with the reference suspension mixes crystal origin with formulation matrix C47·X9. The innovator, an intimate tacrolimus–HPMC dispersion, did not crystallize under the same storage and kept its exposure R2·R9·C55.