Same ingredients, different stability: why one generic amorphous tacrolimus capsule crystallizes

Background and motivation

An amorphous drug dissolves to a higher concentration than its crystal because it has a higher free energy, and that solubility advantage can be estimated from the enthalpy and entropy differences between the two forms (Murdande et al., ref 1). The same higher free energy drives the amorphous solid back toward the crystal. If a commercial amorphous product crystallizes during processing or storage, its solubility and dissolution rate fall; the batch may fail, and if quality control does not detect it, patients may absorb less drug. The authors stress that how much partial crystallinity changes bioavailability is not well understood and is "likely to vary" between drugs and products. C1·C2

The usual defence is a polymer. Storage far below the glass transition temperature (Tg) slows molecular motion (Hancock and Zografi, ref 2), but glasses can still crystallize. Mixing the drug with a polymer at the molecular level, as an amorphous solid dispersion (ASD), slows crystallization in three ways: drug–polymer hydrogen bonds must be broken before the drug can crystallize (Kestur and Taylor, ref 4), the high-Tg polymer couples its slow motion to the drug (antiplasticization, ref 5), and mixing lowers the chemical potential of the drug. Nucleation and growth must both be slowed, and the best polymer differs between drugs and between the two steps: for acetaminophen, HPMCAS inhibited nucleation best, polyacrylic acid and PVP inhibited growth best (Trasi and Taylor, ref 7). Marketed examples are PVPVA in Kaletra and HPMC in Sporanox. C3

Many other factors act on top of the polymer. Polymers can raise or lower nucleation rates (ref 7), surfactants can speed crystal growth (Mosquera-Giraldo et al., ref 8), and mere contact with mannitol or magnesium stearate raised the nucleation rate of amorphous nifedipine about 50 times (Zhang et al., ref 9). The route used to make the amorphous solid matters too: amorphous indomethacin made by spray drying, melt quenching, ball milling or cryomilling differed in stability (Karmwar et al., ref 10), and spin-coated felodipine–Eudragit EPO films resisted surface crystallization better than extrudates (Yang et al., ref 11). Storage conditions "can have the greatest impact": absorbed water lowers Tg and raises mobility (Ahlneck and Zografi, ref 12), heat without water causes physical aging (ref 13), and water can split a dispersion into drug-rich and polymer-rich amorphous phases, in which the drug-rich phase loses the protection of the polymer (Purohit and Taylor, ref 14). Underneath all of this sits the drug's own crystallization tendency, which depends on conformational flexibility, configurational entropy, driving force, viscosity and hydrogen-bond patterns (refs 15–20). C4·C5·C6

Small amounts of crystal matter more than their mass suggests. When an amorphous product dissolves under non-sink conditions it creates a supersaturated solution, and residual crystals "may" grow quickly in it and seed further nucleation. C7

Tacrolimus is an immunosuppressant first approved in 1994 as Prograf (Astellas). The first generic, from Sandoz, followed in 2009, and at least five generic capsules are sold in the United States. The drug has a narrow therapeutic window and highly variable pharmacokinetics between patients (FDA draft guidance, ref 21; Venkataramanan et al., ref 22), and a review of brand-to-generic switches of immunosuppressants recommended enhanced therapeutic drug monitoring and more outcomes research (Ensor et al., ref 23). Prograf and its generics all contain the drug in amorphous form, so the authors ask how susceptible each product is to crystallization, since physical instability "could potentially" add to variability in vivo. C8·C9

The work builds directly on an earlier study by the same group of these commercial products (Trasi et al., ref 25), which gave the amorphous solubility of tacrolimus (50 µg/ml at 37 °C and pH 4.5), its crystalline solubility (1.4 µg/ml) and a nucleation induction time of more than 400 minutes for tacrolimus at 50 µg/ml with 50 µg/ml HPMC in solution. The innovator process had been described by Yamashita et al. (ref 27) as solvent impregnation of tacrolimus into HPMC. M1·R22·R23

The stated goal was to test the robustness of Prograf 5 mg capsules and five generic capsules to crystallization after storage at stress conditions, and to evaluate their dissolution under non-sink conditions before and after storage. Once one product failed, the goal widened to finding out why. Only dissolution of fresh capsules is reported. A1·R3

Background concepts

Amorphous solid dispersion versus physical mixture. In an ASD the drug and the polymer form one glass, mixed molecule by molecule, so the polymer can interact with every drug molecule. In a physical mixture, amorphous drug and polymer sit in separate particles; the polymer can only act where the particles touch. The whole paper turns on the claim that two products with the same ingredients can fall on different sides of this line. The authors cite the view that molecular-level mixing and drug–polymer interactions "are considered to be essential" for stabilizing the amorphous form (Taylor and Zografi, ref 36). C50

Glass transition and water. Below Tg the amorphous solid is a glass with low molecular mobility; above it, mobility rises steeply. Sorbed water lowers Tg (plasticization). The paper names the modified Gordon–Taylor equation as a way to estimate this drop (ref 2) but does not print or use it, so no equation is given here. E7·C5

Hydrate crystallization and critical humidity. Tacrolimus crystallizes as a monohydrate, one water molecule per drug molecule, which corresponds to 2.2 wt.% water. A hydrate can only form when enough water is available, so the authors expect crystallization from the amorphous state to need a relative humidity (RH) above some critical value (Burnett et al., ref 33) and to speed up at higher water content through plasticization. M1·C43

Non-sink dissolution and desupersaturation. The dissolution test was designed so that the whole dose can dissolve only if it stays amorphous. The relation is simple and stated in words in the paper:

In plain terms: when one 5 mg capsule dissolves completely in V = 100 ml, the concentration C_max equals the amorphous solubility of tacrolimus (a "1x sink" for the amorphous form) and is about 36 times the crystalline solubility of 1.4 µg/ml. A product that keeps its drug amorphous should therefore climb toward 50 µg/ml; a fall after a peak (desupersaturation) means drug is leaving solution, which the authors read as crystal growth. E6·C12·M1

Measuring a little crystal in a lot of excipient. Tacrolimus makes up only 2–5 wt.% of the capsule contents, so a crystalline fraction of a few percent of the dose is a few hundredths of a percent of the powder. X-ray powder diffraction (XRPD) quantifies crystallinity by comparing the strength of a tacrolimus reflection with calibration mixtures. The laboratory diffractometer used a peak-area ratio around the 10.32° 2θ reflection:

In plain terms: A is the area under the diffraction pattern over the stated angular range; the narrow window holds the tacrolimus peak, the wide window holds the peak plus its surroundings, so the ratio rises as more crystal is present. The synchrotron instrument used a peak-to-baseline ratio:

In plain terms: I is the intensity at the 8.57° reflection of the monohydrate and Ī the mean intensity of a nearby peak-free baseline; both angles are on the Cu Kα scale after conversion from the synchrotron wavelength. Both ratios are stated in words in the paper and converted to % crystallinity with the calibration lines given in the next section. E1·E2·P8·P9

Limit of detection and quantification. The limit of detection (LOD) is the smallest crystallinity that can be seen; the limit of quantification (LOQ) is the smallest that can be measured with confidence. For the synchrotron method the paper gives an LOQ of 4% and an LOD of about 1% of the drug. R11

Seeds. Crystal seeds are tiny crystals that let growth start without a nucleation step. They can come from incomplete amorphization (milling is thought to leave residual crystalline seeds; refs 26, 29) or form during processing. A product with seeds can crystallize far faster than its amorphous ingredients would on their own. C32·C33

Methods in detail

The study has three parts: a stress screen of six marketed products, a quantitative follow-up of the one product that failed, and an attempt to reproduce that failure by building model formulations along different manufacturing routes. The flow below shows how they connect.

Contribution 1: a stress screen of six marketed products

The problem. All six products declare the same kind of ingredients, so a label comparison cannot say whether they protect the amorphous drug equally well. The authors first measured what is actually in each capsule and then exposed all six to the same harsh condition.

Products. Prograf 5 mg capsules (Astellas) and five generics, from Mylan, Panacea Biotech, Dr. Reddy's, Accord (manufactured by Intas Pharma) and Sandoz, were bought from a pharmacy; one lot of each was tested, and three more Accord batches were bought later (R13629, S03163, S12165, in addition to R09717). The paper prints the same lot number, 047354, for Prograf and Sandoz, which looks like a misprint. M5·M6·M7·M8·M9·M10·X4

Reverse engineering. The ingredient list came from the prescribing information, and only components expected to affect stability were quantified; the drug was taken as its 5 mg label value. Croscarmellose sodium (CCS), an insoluble disintegrant, was weighed as the dried filter residue left after stirring 2 capsules in 250 ml water for 2 hours. HPMC was measured by the phenol–sulfuric acid colorimetric method (absorbance at 490 nm) after dissolving the drug and polymer in dichloromethane:ethanol (1:1), with n = 3 capsules. Lactose, together with minor excipients such as magnesium stearate, was taken as the remainder: P11·P12·P13

In plain terms: whatever mass of the capsule fill is not drug, HPMC or CCS is counted as lactose and minor excipients. For Accord, 132 − 5 − 5.5 − 21.5 = 100 mg, exactly the value in Table I; the other rows agree within 1 mg. E5·R1

Dissolution. The whole contents of one capsule were stirred at 300 rpm in 100 ml of 5 mM phosphate buffer at pH 4.5 and 37 °C, and the drug concentration was followed for 2 hours by a UV dip probe (absorbance at 204 nm, minus scattering at 350 nm). The volume gives the 50 µg/ml ceiling explained above. P6·M11·E6

Water sorption. 15–20 mg of each product's capsule contents, melt-quenched amorphous tacrolimus and a 1:1 tacrolimus:HPMC dispersion were run from 5 to 95% RH at 25 °C in a vapour sorption analyzer. P4

Stress storage. Capsules were taken out of their packs and kept in uncapped vials in desiccators over saturated sodium chloride (75% RH) or sodium bromide (54% RH), in ovens at 40, 35 or 30 °C. The main screen was 4 weeks at 40 °C/75% RH; the capsule contents were then scanned by laboratory XRPD from 5 to 40° 2θ. P7·P8

Contribution 2: quantifying crystallization in a 2–5 wt.% drug product

The problem. With so little drug in the powder, the tacrolimus reflections are weak, and a product could hold a few percent of crystalline drug without showing it in a routine scan.

Calibration. The authors spiked Prograf capsule contents with 0.25–5 mg of crystalline tacrolimus monohydrate per capsule content, giving calibration samples of 5–100% crystalline content relative to the 5 mg label; the drug:excipient ratio was not otherwise adjusted, so each sample also kept Prograf's own amorphous drug. Prograf was chosen because its weight and composition are similar to Accord's and because, being stable in storage, it was assumed to contain no initial crystallinity. The two calibration lines printed in Figure 6 are: P10·C20

In plain terms: x is the % crystallinity, y_syn the synchrotron peak-to-baseline ratio and y_lab the laboratory area ratio. Both lines pass close to their calibration points (for example, at 50% the synchrotron line gives 1.816 against a plotted point of about 1.79, and the laboratory line 0.4023 against about 0.4024). The laboratory ratio changes by only about 0.02 over the whole 0–100% range. E3·E4·R9

Two instruments, two trade-offs. The laboratory diffractometer (Cu Kα, scans of 9.5–11.5° 2θ at 0.5°/min for kinetics) samples a large area, so its calibration points scatter little. The synchrotron (beamline 12-ID-B, about 10 scans per calibration mixture) has a beam of only 300 × 20 µm: it gives far better signal to noise, but each scan sees few particles, so a sample with very little crystal can show none in one scan and a lot in another. Some scans of calibration samples at 50% crystallinity and above gave very high crystalline signals and were discarded. The authors therefore used the synchrotron to look for traces of crystal in fresh Accord and to follow its early crystallization, and called that analysis "semiquantitative". P8·P9·C21·C23

Follow-up studies. One Accord batch (the lot is not named) was stored at 40 °C/75% RH and scanned by synchrotron for up to 42 days. Four Accord batches were stored at 40 °C/75% RH and scanned by laboratory XRPD weekly for 4 weeks, n = 3 capsules per point. A fast and a slow batch (S12165 and S03163) were then stored at 30/75, 35/75, 40/54 and 40/75 (°C / % RH). R12·R13·R14·X11

Contribution 3: taking the failure apart

The problem. Once Accord alone crystallized, the question became which of its differences was responsible: more CCS, crystal seeds, mechanical stress, or the way the amorphous drug was made.

Reference amorphous systems. Pure amorphous tacrolimus was made by melt quenching followed by cryomilling. A 1:1 tacrolimus:HPMC dispersion was cast from 1:1 ethanol:dichloromethane onto a Teflon film, dried under vacuum at room temperature and cryomilled (the sorption section calls it "prepared by rotary evaporation"). Both were measured by DSC as prepared and after 3 days at 75% RH in DSC pans that were then sealed (cooled to −10 °C, heated to 100 °C at 10 °C/min, n = 3), and both were stored at 40 °C/75% RH and scanned after 1, 2, 3 and 4 weeks. P1·P2·P3·X15

Model formulations. Each of the following was stored at 40 °C/75% RH and checked by XRPD after 4 weeks: P20

• Extra disintegrant: about 15 mg CCS added to Prograf contents, bringing them to Accord's level. P14
• Cryomilling: 100 mg crystalline drug, 100 mg HPMC and 400 mg CCS cryomilled until X-ray amorphous, then mixed with lactose; cryomilling can leave residual seeds. The milling time is given as 2 hours in the text and 60 minutes in Table III. P15·R20·X2
• Rotary-evaporated drug: amorphous tacrolimus made by rotary evaporation, 50 mg of it mixed with 50 mg HPMC, 200 mg CCS and 1 g lactose and briefly cryomilled (30 seconds in the methods, 10 seconds in the results); in a variant, CCS was suspended in the ethanolic drug solution before evaporation. P18·X3
• Ball milling: Prograf contents milled in a Wig-L-Bug (5 minutes in the methods, 2 minutes in the results and Table III) to create seeds or high-energy sites. P16·X1
• Seeding: 1 or 2 mg crystalline monohydrate mixed into the contents of 5 Prograf capsules, i.e. 4% or 8% of the drug. P17
• Ethanol wet granulation: 20 mg tacrolimus in ethanol, alone or with 20 mg HPMC "dispersed in the solvent", added to a stirred bed of 400 mg lactose with 80 mg CCS (drug:CCS 1:4, close to the 5 mg : 21.5 mg of Accord), and dried at room temperature or at 60–70 °C with a heat gun; in one variant HPMC was mixed in only after drying. P19

The granulation experiments follow from the public assessment report of the Medicines Evaluation Board (Netherlands) for Tacrolimus Intas 5 mg capsules, which describes the Accord contents as made by wet granulation, followed by sizing, blending and encapsulation (ref 30). The authors infer that ethanol is "most likely" the solvent and, since the product is amorphous, assume the drug is dissolved in it. R25·C34

Scanning electron microscopy. To see what the drug looks like inside each product, capsule contents were stirred in 10 ml water for 4–5 minutes to dissolve lactose and HPMC, filtered and dried. The residue should hold only tacrolimus and CCS; CCS was recognised by its fibrous shape, so the remaining particles were taken as tacrolimus. The same was done for two model materials, tacrolimus deposited from ethanol on lactose and tacrolimus impregnated into HPMC with ethanol. Images were taken at 10 000×. P5·C36·R21

Experimental analysis

What is in the capsules

Table A summarizes the reverse engineering. Every product is mostly lactose, with tacrolimus at 2–5 wt.% of the fill. HPMC is present at close to 1:1 with the drug in all six (4.8–5.5 mg per 5 mg drug). The fill weight varies more than twofold, from 100 mg (Mylan) to 242 mg (Sandoz). The one composition outlier is Accord, with 21.5 mg CCS against 4.1–8.2 mg in the others, "nearly 4 times" the amount in the innovator. R1·R2·C11

Table A. Ingredients per capsule, from Table I (standard deviation in parentheses, n = 3 capsules; tacrolimus taken as the label claim; no standard deviation is printed for CCS). R1

Note by the AI writer (not discussed in the paper): the CCS value comes from the weight of an insoluble residue, and the method does not say whether undissolved drug or other insoluble matter was subtracted, so the CCS figures are approximate; the difference between Accord and the rest is nonetheless large. X17

Dissolution of the fresh products

Under the 1x-sink conditions (Figure 1), Accord dissolved fastest: about 33 µg/ml at 5 minutes and about 39 µg/ml at 10 minutes, roughly twice Prograf (about 13 and 20 µg/ml). It peaked at about 45 µg/ml around 40 minutes ("45 µg/ml within 40 min" in the text) and then fell steadily, to about 27 µg/ml at 2 hours (the text says "~25"), never reaching the 50 µg/ml ceiling. Mylan, Dr. Reddy's and Panacea dissolved faster than Prograf and Sandoz; at 2 hours Prograf, Mylan, Dr. Reddy's and Panacea had reached about 47–49 µg/ml, close to the amorphous solubility, while Sandoz, the slowest, had reached about 43 µg/ml. R3·C13

The authors read the fall as crystallization of tacrolimus during dissolution of Accord ("indicates"), and the other products "appear to be stable" over the 2 hours. C14 They later argue that this is unexpected for an HPMC product: in their earlier work, tacrolimus at 50 µg/ml with 50 µg/ml HPMC (about what one dissolved capsule gives in 100 ml) had a nucleation induction time of more than 400 minutes, and Yamashita et al. found that innovator-type dispersions held about 50 µg/ml for several hours. Hence "There is some evidence" that Accord carries crystal seeds before storage, or that seeds "quickly develop in the dissolving matrix". R22·R23·C44

Water sorption

Accord contents sorbed more water than any other product at every RH (Figure 2): about 5.1% at 80% RH, 9.2% at 90% RH and 12.8% at 95% RH. The other products took up about 4.0–5.7% at 95% RH and Sandoz the least, about 1.6%. R4·C15 Pure melt-quenched tacrolimus sorbed about 4 wt.% at 95% RH and the 1:1 dispersion about 6.5% (Figure 3). The amorphous drug thus passed the 2.2 wt.% of the monohydrate (between about 80 and 85% RH) without crystallizing during the sorption run. R5·C17 Scaled to the small drug-plus-HPMC fraction of the capsules, these values cannot explain Accord's uptake, so the authors attribute its hygroscopicity "most likely" to the extra CCS. C16

Note by the AI writer (not discussed in the paper): sorption was measured at 25 °C, whereas storage was at 30–40 °C, so these curves rank the products rather than give the water content they held during storage. X18

Which products crystallize under stress

After 4 weeks open at 40 °C/75% RH, only Accord showed changes in its XRPD pattern consistent with tacrolimus crystallization (Figure 4): new reflections at 8.5, 10.3 and 13.7° 2θ. Subtracting the fresh Accord pattern from the stored one left a pattern that matches crystalline tacrolimus monohydrate, which in turn matches the pattern calculated from its single-crystal structure (Cambridge Structural Database code FINWEE10; Figure 5). R6·R7·C18 Panacea's pattern also changed, but without tacrolimus reflections; the authors assign the change to anhydrous lactose turning into lactose monohydrate, and the extra peaks of Mylan likewise to anhydrous lactose. R8·C19 Prograf capsules stored for up to 2 months stayed free of crystals (data not shown), in line with Honbo et al., who found a tacrolimus–HPMC dispersion amorphous after 3 months at 40 °C/75% RH. R17·R24·C40

Synchrotron scans of fresh Accord (Figure 7) showed no reflection at 8.57°, whereas a Prograf sample spiked with 5% crystalline drug (0.2 wt.% of the powder) showed it clearly. Any crystal in fresh Accord is therefore "considerably less than 5%" of the drug. R10·C22 Note by the AI writer: in the same figure, fresh Prograf shows small reflections near 8.47° and 10.52° that lie away from the tacrolimus positions; the paper does not comment on them. R10

How fast, and in which batches

Followed by synchrotron at 40 °C/75% RH (Figure 8), one Accord batch showed monohydrate reflections within 12 hours, below the LOQ of 4%. It reached more than 20% after 2 days and was nearly completely crystalline after 3 weeks. R11·R12 The discussion says crystallinity was detectable "after only a few hours"; the results text and the first plotted point (about half a day) support "within 12 hours". X14

Table B. Percentage of the drug crystallized during open storage at 40 °C/75% RH, read from Figures 9 (n = 3) and 8 (approximate). The synchrotron batch also gave about 7.5% at 1 day, 27% at 2 days and 88.5% at 6 weeks. R12·R13

The four batches differed "considerably" (Figure 9; Table B). R09717 and S12165 were already largely crystalline after 1 week; R13629 and S03163 showed crystals only from the second week. After 4 weeks all four had crystallized substantially, to about 57–95% of the drug (the abstract gives 50–100%). Some batches also varied strongly from capsule to capsule. The authors note that the laboratory kinetics of the fast batches are very similar to the synchrotron profile. R13·C24

Table C. Percentage of the drug crystallized at four storage conditions, laboratory XRPD, read from Figure 10 (approximate). R14

Storage condition mattered strongly (Figure 10; Table C). The fast batch crystallized at 75% RH even at 30 °C and 35 °C, but hardly at 54% RH and 40 °C; the slow batch crystallized clearly only at 40 °C/75% RH, with about 22% at 35 °C/75% RH. The authors conclude that "Clearly" both temperature and RH act, with humidity "likely" the more important because tacrolimus crystallizes as a hydrate. They point out that 30 °C/75% RH corresponds to ICH climatic zone IVb, a hot and humid climate. R14·R26·C25·C26 Two details need care. The text gives "around 40%" for the slow batch at 40 °C/75% RH after 4 weeks, but the figure shows about 57% (about 39% at 3 weeks); the figure values are used here. And the text says the fast batch showed crystals "after one week" at 30 °C/75% RH, while the figure shows about 1% at 1 week and about 17% at 2 weeks. X5·X13

Accord against the amorphous drug itself

The reference systems behave as expected for a good dispersion (Table II, Figure 11). Melt-quenched, cryomilled tacrolimus had a Tg of 63.6 °C, falling to 51.4 °C after 3 days at 75% RH; the 1:1 tacrolimus:HPMC dispersion fell from 77.2 °C to 49.0 °C. Both therefore stay glassy, about 9–11 °C above the 40 °C storage temperature. R15·C27 In storage at 40 °C/75% RH, the dispersion showed only an amorphous halo for 4 weeks, and the pure amorphous drug started to crystallize only after 2 weeks (sharp peaks visible at 4 weeks) and was still not fully crystalline at 4 weeks. R16

Accord, by contrast, was about 59% crystalline after 1 week (Table B). The authors call it "somewhat surprising" that tacrolimus in a formulated product crystallizes faster than the pure amorphous drug, and say this "strongly suggests" that a formulation component or the processing method accelerates crystallization. R12·C28

Note by the AI writer (not discussed in the paper): the Tg values come from the pure drug and the 1:1 dispersion after humidity exposure, not from the capsule contents, where CCS brings extra water; the statement that "all of the systems remain in the glassy state" therefore applies to these model systems. X16

Which change reproduces the failure

Table D. The model formulations of Table III, with the HPMC-contact and processing columns added by the AI writer from the descriptions in the text. Row c of Table III prints "Minor crystallization" and "Alone stable with CCS" on two lines; it is read here as the drug alone showing minor crystallization and the CCS variant staying stable. Replicate numbers are not printed. R18·R19·X8

Not the CCS. Adding CCS to Prograf, up to Accord's level, did not make it crystallize, and neither did drug rotary evaporated together with CCS. The authors conclude that their results "do not support" CCS as the cause, and that susceptibility "does not appear" to come from the mere presence of an excipient. R18·C30·C45

Not added seeds, not milling. Prograf seeded with 4% or 8% crystalline drug stayed stable, as did ball-milled Prograf, so mechanical stress "does not appear" to be the cause. Crystalline drug cryomilled with HPMC and CCS until X-ray amorphous, which might leave residual seeds, was also stable. R18·R20·C32·C33 To square the seeding result with their suspicion of seeds in Accord, the authors suggest that added seeds failed "most likely" because the drug is so dilute in the powder that seeds rarely touch drug particles; if Accord holds seeds, they must sit in intimate contact with the drug. C46

Only hot drying without HPMC contact. Of the four ethanol granulations, only the one in which the drug solution was added to lactose and CCS at 60–70 °C, with HPMC mixed in after drying, crystallized: it was amorphous at first and showed crystalline reflections within 14 days at 40 °C/75% RH. The same process with HPMC in the ethanolic liquid stayed stable, and so did both granulations dried at room temperature. From this the authors conclude that "it appears" the intimacy of contact between tacrolimus and HPMC, as well as temperature, "may determine" susceptibility to crystallization. R18·R19·C35

Note by the AI writer (not discussed in the paper): two features of this design limit what it can show. Both room-temperature granulations had HPMC present during granulation (in the bed or in the liquid), and no granulation was dried at room temperature with HPMC added afterwards, so drying temperature and HPMC contact are not separated; the only matched pair (rows h and i, both at 60–70 °C) varies HPMC contact. In addition, the HPMC in the ethanolic liquid is called "dispersed" in the methods, a "solution" in Table III and "in suspension" in the results, so whether it was dissolved is unclear. X7·X6 Table III also lists minor crystallization for rotary-evaporated drug simply mixed with HPMC, CCS and lactose, a plain physical mixture that the text does not discuss. X8

What the drug looks like

In the SEM images (Figure 12), tacrolimus particles left from Accord had a smooth surface with small holes, much like particles from tacrolimus deposited from ethanol onto lactose. Tacrolimus particles left from Prograf were larger, with a rough, porous surface, very similar to particles from tacrolimus impregnated into HPMC. The authors find the images "highly supportive" of an Accord process like their hot granulation and different from the Prograf process. R21·C37 The particles were identified as tacrolimus by excluding the fibrous CCS, not chemically. C36·X10

Discussion

The authors' chain of reasoning

The argument runs by elimination. HPMC is present in every product at about 1:1 with the drug, and from Yamashita et al. it "can be inferred" that HPMC forms an ASD with tacrolimus in Prograf; a 1:1 dispersion resisted crystallization for a month at 40 °C/75% RH, so all the products "would be anticipated" to be stable. C39·C40 Accord crystallized faster than even pure amorphous drug, so something in its formulation or processing must accelerate crystallization. C28 Its obvious difference, extra CCS, failed the test, and so did added seeds and mechanical stress. C30·C32·C45 What remained was the route by which the drug becomes amorphous. The chain below sets out the two outcomes the authors propose.

The authors' proposed chain ("we believe", "We thus speculate"); built from the model granulations and the SEM images, not measured in Accord itself. C47·C48·C49·C51

In the authors' words, they "believe" that Accord does not contain an amorphous solid dispersion with molecular-level mixing, but amorphous tacrolimus present as a coating on the lactose particles; their results "suggest" the product may be dried at an elevated temperature that induces a low level of crystalline tacrolimus; and they "speculate" that Accord is made by ethanol wet granulation with solvent removal at elevated temperature and HPMC added by physical mixing after the drug has been rendered amorphous, so that no solid dispersion forms and drug and HPMC do not interact. C47·C48·C49 Granulating the drug onto excipients "may" also raise its surface area and speed crystallization (Crowley and Zografi, ref 37). C51 The extra water sorbed through CCS "might be expected" to accelerate the transformation to the monohydrate, even though CCS alone did not destabilize Prograf. C43·C30 The abstract states the same conclusion more cautiously: "It seems probable" that the unstable generic contains amorphous drug physically mixed with polymeric excipients rather than an ASD. C47

What the paper supports, and how firmly (the AI writer's view)

The AI writer's assessment: the product comparison is direct; one of six products crystallized under the screen, the effect appeared in all four of its batches, was seen by two instruments and grew with humidity. The mechanism is supported indirectly, by elimination of three simpler causes, by one model process that reproduced the failure and by the resemblance of SEM residues. The authors' own wording follows that gradient, from asserted results to "we believe" and "We thus speculate" for the manufacturing route. R6·R13·R14·R18·R21·C47·C49

What it suggests for formulation

The authors draw two practical messages. The study "highlights" the impact of processing conditions on amorphous product stability and the need to control temperature and humidity over the product's lifespan; and the susceptibility of Accord "suggests that this product may undergo unwanted physicochemical changes" under similar storage, which reinforces the importance of packaging. C53·C41

Note by the AI writer (not discussed in the paper): the case shows that a qualitative and quantitative match of ingredients, including the drug:polymer ratio, does not guarantee a matching physical state. If the polymer must meet the drug at the moment the drug becomes amorphous, then the order of addition and the drying temperature behave as critical process parameters of such a granulation. A stress screen of the kind used here separated the products within days; at 2–5 wt.% drug, release testing by routine XRPD would not see seeds below a few percent of the dose, which is where the synchrotron was needed. R1·R11·R13·C35

Limitations

Limits the authors state

• The relationship between product crystallinity and in vivo performance has not been elucidated for this product, so it is unknown whether the observations have clinical significance. C42
• The products were tested outside their packaging under open-dish stress; the authors present the result as a reason for packaging and storage control rather than a prediction of shelf behaviour. C41·C53
• With tacrolimus at 2–5 wt.% of the contents, solid-state analysis of the products is "challenging", and the synchrotron kinetics are "semiquantitative". C10·C23
• Low-level crystallinity measured with the small synchrotron beam depends on the sampling strategy. C21

The AI writer's own judgement

• The mechanism is inferred, not measured. No measurement in Accord itself (for example Tg, spectroscopy or imaging of drug and HPMC) shows that the drug is unmixed with HPMC or sits on lactose. The support is one reproducing model granulation, with replicate numbers not printed, and SEM resemblance of particles identified visually. X10
• Drying temperature is not isolated. The room-temperature controls had HPMC present during granulation, so the claim that ambient drying prevents crystallization is not tested on its own; whether HPMC was dissolved or suspended in the ethanol is also unclear. X7·X6
• Seeds in fresh Accord are inferred. The synchrotron found no crystal above 5% of the drug in fresh Accord, and the dissolution data cannot separate pre-existing seeds from crystals formed during dissolution; the authors give both possibilities. X9
• Laboratory crystallinity values are semiquantitative. The laboratory area ratio changes by only about 0.02 over 0–100% crystallinity, its slope is printed to one significant figure, no LOD or LOQ is given for it, and the calibration matrix (Prograf contents) differs from Accord in CCS and lactose. X12
• Several method details conflict in the text. Ball-milling time (5 or 2 minutes), cryomilling time (2 hours or 60 minutes), the brief cryomilling of the rotary-evaporated mixtures (30 or 10 seconds), the slow-batch value at 4 weeks ("around 40%" against about 57% in the figure) and a lot number shared by Prograf and Sandoz. X1·X2·X3·X5·X4
• Unidentified and unreplicated details. The synchrotron kinetics batch is not named, and replicate numbers are missing for the dissolution, sorption, calibration and kinetics figures; the reference samples' preparation is described inconsistently. X11·X15
• Model systems stand in for products. Tg was measured on the pure drug and the 1:1 dispersion rather than capsule contents, and water sorption at 25 °C rather than at the storage temperatures. X16·X18
• A result left undiscussed. A plain physical mixture of amorphous drug with the excipients showed minor crystallization in Table III; it bears directly on the physical-mixture explanation but is not discussed. X8
• Scope. One lot of each product other than Accord was tested, and only fresh capsules were dissolved, although the stated aim included dissolution after storage. A1·R3

Conclusion

The paper asked whether six marketed amorphous tacrolimus capsules are equally robust to crystallization. Five, including the innovator, stayed amorphous for 4 weeks at 40 °C/75% RH; Accord crystallized to tacrolimus monohydrate within days, in every batch tested, at 75% RH even at 30 °C, and it lost supersaturation during dissolution. R6·R12·R13·R14·R3 Accord crystallized faster than pure amorphous drug, which pointed away from the drug and toward the product; extra CCS, added seeds and milling did not reproduce the failure, whereas drug granulated from ethanol onto lactose at 60–70 °C with HPMC added afterwards did. R16·R18·R19·C28 From this and from SEM of the residues the authors "believe" that Accord holds amorphous drug coated on lactose rather than an amorphous solid dispersion, a conclusion that rests on inference rather than on a measurement of the drug's state inside Accord. C47·C49·X10 The lesson the authors draw is that processing conditions, not only the ingredient list, decide the stability of an amorphous drug product, together with control of temperature and humidity over its life. C53