Development of Screening Methodology for the ... - ACS Publications

For a more comprehensive list of citations to this article, users are encouraged to ... Development and Scale-Up of a Crystallization Process To Impro...
1 downloads 0 Views 1MB Size
Subscriber access provided by La Trobe University Library

Full Paper

Development of screening methodology for the assessment of the agglomeration potential of APIs Charles Dimitrios Papageorgiou, Marianne Langston, Frederick Hicks, David J am Ende, Eric Martin, Sarah Rothstein, Jerry Salan, and Robert Hunter Muir Org. Process Res. Dev., Just Accepted Manuscript • DOI: 10.1021/acs.oprd.6b00201 • Publication Date (Web): 18 Jul 2016 Downloaded from http://pubs.acs.org on July 24, 2016

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Organic Process Research & Development is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

Development of screening methodology for the assessment of the agglomeration potential of APIs Charles D. Papageorgiou1*, Marianne Langston1, Frederick Hicks1, David am Ende2*, Eric Martin2, Sarah Rothstein2, Jerry Salan2, Robert Muir2 1 Takeda Pharmaceuticals International Co., Chemical Development Laboratories-Boston, 40 Landsdowne St., Cambridge, MA, 02139 2 Nalas Engineering Services, Inc, 85 Westbrook Rd, Centerbrook, CT 06409

ACS Paragon Plus 1 Environment

Organic Process Research & Development

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Table of Contents Graphic

ACS Paragon Plus 2 Environment

Page 2 of 25

Page 3 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

Abstract A simple screening protocol has been developed for assessing the agglomeration potential of active pharmaceutical ingredients (API’s) using resonant acoustic mixing that minimizes the quantity of API used. This methodology improves upon existing ones as it allows for multiple conditions to be screened in parallel, saving time and allowing for the study of agglomeration and optimization of the drying unit operation to take place early in development. In addition to a qualitative (visual) assessment, quantitative data can be obtained after the material has been dried therefore accounting for a measure of cake hardening. This methodology was also extended to assess the friability of the generated agglomerates and was validated using a scaled-down agitated filter dryer (AFD). The impact of particle size, particle size distribution, solvent selection, and solvent loading on the agglomeration potential for a Takeda API is also discussed which allowed for the development of an improved drying process that was successfully scaled-up in the pilot plant.

Keywords: resonant acoustic mixing, agglomeration, friability, drying, agitated filter dryer

ACS Paragon Plus 3 Environment

Organic Process Research & Development

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Introduction In the production of APIs, drying is one of the last steps following crystallization. With an increasing number of poorly soluble drugs being developed and a desire to produce solid dosage formulations, much focus has recently been devoted to the concept of particle engineering in order to design desirable bulk powder and physical properties into the drug substance (DS).1, 2 In many instances these properties cannot be easily achieved or controlled through subsequent milling or other dry processing steps, increasing the burden on the crystallization and drying unit operations.3 There are a number of undesirable outcomes that can be encountered as a result of a poorly understood and controlled drying operation, such as agglomeration, attrition and chemical and physical instability including polymorphism, loss of crystallinity as well as dehydration or desolvation.4-7 All of these can have a detrimental impact not only on product quality and downstream drug product (DP) manufacturing, but also on the dryer itself. Agglomeration, for example, can result in variable, or out of specification, particle size distributions (PSD) and/or entrainment of impurities impacting downstream DP manufacture and even bioavailability. In extreme cases known as snowballing it can place excessive stress on the impeller causing damage to the processing equipment and/or lead to difficulty in removing the batch. A number of Takeda Pharmaceutical’s APIs have recently posed some challenges during the scale-up of the drying unit operation in agitated filter dryers (AFD) with respect to excessive agglomeration that has impacted DP manufacture, in vitro dissolution and resulted in significant API material losses. As is often the case, scale-up in the plant relied upon visual observation by the operators, especially during the initial stages of the process, with agitator speed, height and direction being manually adjusted to control and break down any agglomerates generated. While solvent composition, solvent content, particle size, particle size distribution and particle morphology have been found to affect agglomerate formation in AFDs, the fundamental mechanisms of agglomeration are still poorly understood. This is primarily as a result of the lack of representative scaledown models that would allow for its study in the laboratory, and it is only recently that suitable methodologies have emerged in the literature.8-11 Hapgood et al. recently demonstrated, using a 2L scaledown AFD, the impact of solvent composition on agglomerate friability as well as that of cake height and PSD on the extent of agglomeration.12 am Ende et. al., developed an agglomeration risk assessment strategy based upon the evaluation of a number of Pfizer compounds for which scale-up experience existed. Using mixer torque rheometry (MTR), the torque profiles of a number of solid/solvent systems were compared employing Avicel (microcrystalline cellulose PH102) as the reference standard to determine the relative risk of agglomeration for each system.9 It was also shown that the MTR was not able to predict agglomeration due to cake hardening during drying of threonine. To further evaluate the MTR method, Zhang and Lamberto systematically looked at the effect of mixing speed, mixing time,

ACS Paragon Plus 4 Environment

Page 4 of 25

Page 5 of 25

bowl load and solvent on the registered torque in a four-factor half factorial design and determined that solvent is the only statistically significant factor affecting the value of the maximum torque.8 1 0.9

Relative Torque (Nm)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

0.8 Avicel/Water

0.7 0.6 0.5 0.4 0.3 Alisertib sodium/Ethanol 0.2 0.1

TAK-117/Water

0 0

10

20

30

40

50

60

70

% LoD

Figure 1: Rheological torque profile for TAK-117/water, alisertib sodium/ethanol and Avicel/water systems. The MTR data suggests a low agglomeration risk for both the TAK-117 and alisertib sodium systems, particularly at higher moisture contents. Agglomeration has been a significant problem in the manufacturing process of two of Takeda’s APIs, TAK-117 and alisertib sodium. In the case of alisertib sodium, small agglomerates were formed that were only detected by a polymodal particle size distribution determined by a dry dispersion laser diffraction particle size method and found to impact both in vitro dissolution and DP manufacturing.11 In the case of TAK-117, snowballing was observed resulting in significant losses (>30%) during the delumping step in the DP manufacturing process. Therefore the agglomeration potential of both compounds was studied using the MTR in an attempt to better understand and define the drying process. Figure 1 shows the torque data collected for various solvent contents for the two APIs studied in the solvent system present at the beginning of drying (water for TAK-117 and ethanol for alisertib sodium) against Avicel in water. Both compounds exhibited a relatively low and flat torque profile across the range of moisture contents tested. Above 30% loss on drying (LoD), the torque for both TAK-117 and alisertib sodium began to drop, which typically occurs when a slurry is formed. In this case, it was found that the wet powder had begun to stick to the blades of the MTR and therefore register little resistance (Figure 2 -TAK-117 behaved similarly). This issue is sometimes overcome by increasing the dry solids loading but this was not successful in this case. Because of this tendency to stick to the metal blades and the sides of the MTR bowl, the MTR method was deemed ineffective in predicting the agglomeration potential for these compounds and potentially other similar ones. It is believed that this illustrates how different agglomeration mechanisms occur depending on the physical characteristics of the solids and solvent.

ACS Paragon Plus 5 Environment

Organic Process Research & Development

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

dry powder

39% wet

Page 6 of 25

49% wet

Figure 2: Pictures of the MTR bowl containing 10 g of alisertib sodium wetted with various amounts of ethanol. Left: dry powder, middle: 39% wet, right: 49% wet. The material is coating the blades with very little “free “material to measure torque resistance. Solid loadings greater than 10 g behaved similarly. Polarized light microscopy (PLM) images showed that the particles of both compounds shared a common morphology: they consisted of very small needles (Figure 3), which are known to exacerbate agglomeration as a result of the high surface area available for effective contact between each individual crystal. To ensure robust scale-up and manufacture, the agglomeration in both processes needs to be better understood and ideally minimized and it is clear that alternative methodologies are needed.

Figure 3: PLM image of TAK-117 (Left) and alisertib soidum (Right) at 500x magnification. The calibration bar (upper left corner in each photo) is 50 µm. This paper describes the subsequent studies conducted to better understand the agglomeration potential for TAK-117 that led to the development of a new methodology that was successfully used to determine the impact of the key process parameters on the agglomeration potential. This method builds upon the development work of Zhang and Lamberto who adapted the LabRAM with a vacuum drier attachment

ACS Paragon Plus 6 Environment

Page 7 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

using resonant acoustic mixing to impart mixing energy as a means of providing conditions, similar to those present in AFDs, that are conducive to granule formation.8 The attachment consisted of a cylindrical glass vessel equipped with a temperature probe and vacuum port, mounted on top of the acoustic mixer. Low frequency, high-intensity acoustic energy was used to create a uniform shear field throughout the entire mixing vessel resulting in rapid fluidization of the powders (similar to a fluidized bed) and dispersion of the material. For each experiment, approximately 30 g of the wet-cake was vacuum-dried on the LabRam for 2 hrs, after which time the sample was removed from the vessel, visually observed and analyzed for particle size and morphology using laser diffraction and scanning electron microscopy (SEM). This methodology was extended and adapted to allow for the rapid screening of multiple conditions using readily available small glass vials using only 1g of material per sample (as opposed to 10-20 grams needed for the MTR method). The resonant acoustic mixer utilized for this work was the benchtop LabRAM, manufactured by Resodyn (Figure 4).13 Experimental Section Materials: Microcrystalline cellulose (PH102) was obtained from FMC Biopolymer. Deionized water was generated using a Barnstead GenPure xCAD Plus system by Thermo Scientific. TAK-117 was crystallized from a DMSO:Water solvent system according to a the standard manufacturing process, isolated using an AFD and washed with water. The wet-cake was dried at 50 °C under vacuum. All organic solvents were reagent grade and used as received. Wet-Cake generation: The wet-cake for the drying studies reported herein can be obtained either directly from the filter post crystallization or prepared by manually mixing dry powder and the desired solvent using the LabRAM. In the later case, 1.0 g of the solid dry sample was weighed into a 20 mL scintillation vial. A micropipette was used to dispense the desired amount of solvent (e.g. 250 µL of water to 1 gram of powder for a 20 wt% solvent loading) into the vial. After the materials were dispensed into the vials, the vials were capped and secured in the LabRAM (Figure 4A).14 The LabRAM was then set to 30% intensity (corresponding to approximately 30 g’s of force) for 3 minutes in order to evenly disperse the solvent throughout the solid sample.15 After the 3-minute cycle, the vials were removed from the LabRAM and uncapped. A spatula was used to simply loosen up the wet-cake sample from the glass and ensure that it was not adhering to the walls or the bottom of the vial. The samples were then recapped. In this way the wet-cake was mobile within the glass vial when placed back onto the LabRAM for the granulation step.

ACS Paragon Plus 7 Environment

Organic Process Research & Development

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Granulation of the wet-cake in the LabRAM: The capped vials were placed into the LabRAM. The LabRAM was set to 30% intensity for 3 minutes. This cycle caused the samples to agglomerate and “ball” together if the conditions of the samples were conducive to agglomeration. An example of balling can be found in Figure 4B. The vials were then removed from the LabRAM and the wet-cake samples were dispensed onto tared aluminum pans using a spatula as shown in Figure 4D. It was important to attempt to remove the sample while leaving as much of the already-formed agglomerates intact. This ensured that all of the agglomerates were present during the analysis. A qualitative analysis was made via visual inspection to assess the degree of agglomeration in each sample. Snowballing of the wet-cake could be discerned which is further discussed in the Results and Discussion section. Quantitative analysis, if desired can be performed by assessing the friability of the agglomerates.

ACS Paragon Plus 8 Environment

Page 8 of 25

Page 9 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

A A

C A

B A

D A

Ethanol – Large sticky agglomerate

THF – Small granular. Mostly powder

Methanol – Large sticky agglomerate

Heptane – Mostly powder

Isopropanol – Large sticky agglomerate

Acetone – Mix of agglomerates and powder Figure 4: A) The LabRAM holding six screening vials. B) An example of balling of a TAK-117/water system (40% moisture content) from the LabRAM method. C) The sieve stack secured in the LabRAM. D) Six samples from a solvent screen at 30% LoD after the two LabRAM cycles. The samples have been transferred to pans for drying. Friability experimental procedure: Each granulated sample of the wet-cake was weighed and the mass was recorded. The samples were then dried overnight in a fume-hood or vacuum oven until a constant weight was obtained, followed by hand sieving through a sieve assembly consisting of a stacked 2000 and 1000 µm brass hand sieve (supplied by

ACS Paragon Plus 9 Environment

Organic Process Research & Development

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 25

McMaster Carr -product numbers 34735K216, 34735K221, 34735K83, and 34735K85). A 2000 µm sieve was chosen because it was the size used with a conical mill in the DP manufacture of TAK-117. Agglomerates smaller than 2000 µm were not found to have a detrimental impact on the DP. The tare weight of each section of the sieve was recorded, and the sample from the drying pan was transferred to the top, 2000 µm sieve. The lid was placed onto the top sieve and the sieve stack was tapped on the bottom of the fume hood gently about ten times, until all particles smaller than the size of the sieves had passed through to the collection pan. The aim of the hand sieve tapping was to dislodge any fine particles from the larger agglomerates in the sample without causing the agglomerates to break apart. When successful, all of the particles under 1000 µm should have passed through to the collection pan while the larger agglomerates should have been retained on the two sieves. The 2000 and 1000 µm sieves as well as the pass-through collection pan were weighed and the masses recorded to determine the amount of material retained at each level, expressed as the percentage of retention on each sieve. The material retained on the sieves was then used to define the extent of agglomeration. An example of agglomerates retained at the 2000 µm and 1000 µm levels, as well as the pass-through material, is shown in Figure 5.

Figure 5: Agglomerates retained on a 2000 µm (Left) and 1000 µm (Center) sieve, as well as the material collected in the pass-through pan (Right).

The sieve stack was then reconstructed in the same configuration as before to test the friability or hardness of the agglomerates formed in the sample. The lid was placed on the sieve stack and the stack was secured in the LabRAM (see Figure 4C). The LabRAM was programmed to 25% intensity (corresponding to approximately 25 g’s of force) for 1 minute.16 .

ACS Paragon Plus 10 Environment

Page 11 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

After the friability cycle, the sieve stack was removed from the LabRAM and each section of the stack was weighed a second time. The weights recorded were then used to determine the percentage of the material retained at each level. The sieve stack was cleaned and the friability procedure repeated until all of the samples had been analyzed. Pilot plant studies: Approximately 15 kg of TAK-117 was crystallized according to the standard manufacturing process from dimethyl sulfoxide (DMSO) and de-ionized (DI) water. The wet-cake was isolated and dried using a 0.4 m2 Rosenmund Hastelloy C-22® AFD equipped with a 20 µm Hastelloy C-22® filter mesh, The cake was agitated with a heated 2-blade impeller with a diameter of 0.72 m fabricated with two orientations that was capable of plowing or smoothing the cake in the clock-wise and anti clock-wise directions, respectively. The agitation regime was fully programmable and pressure, temperature (both jacket and cake) were data-logged. Particle Size Distribution Particle size distributions (PSD) for TAK-117 were measured on a Malvern Mastersizer 2000. The dispersant used for the Malvern method was 1% (w/w) Triton ® X-100 in Water. To prepare the Malvern sample, ~200 mg of TAK-117 was weighed into a glass vial and 5 mL of 1% (w/w) Triton ® X-100 was added to the vial. The sample was mixed by inversion ~10 times and added dropwise into the instrument until an obscuration limit of 5-15% was obtained. The sample was allowed to mix for 30 seconds before the measurement was started. After the initial measurement, the sample was sonicated at 100% for 60 seconds, allowed to equilibrate for another 60 seconds, and the final measurement was taken. The average d10, d50 and d90 were reported. Results and Discussion: Friability Assessment: In order to quantitatively compare the agglomerate strength of different samples the relative friability of the agglomerates was calculated. This was defined as the fraction of material that passed through the sieve in question after both the hand-sieving and the LabRAM friability cycle. If, for example 100% of the material has passed through the sieve after the friability cycle, then the sample can be said to be 100% friable at 2000 µm. This percentage can be used to compare the “agglomeration risk” of each sample, or

ACS Paragon Plus 11 Environment

Organic Process Research & Development

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 25

the chance of forming hard agglomerates larger than the sieve limit. Because this number is a percentage of the total material used in the sample, samples of different sizes can be compared. It is important to note that the “agglomeration risk” as defined herein encompasses both the extent of agglomeration and the strength of the agglomerates. Both need to be considered to fully capture the risk of agglomeration. This methodology was adapted from Birch and Marziano who developed the agglomerate brittleness index (ABI) to account for the hardness of the agglomerates.10 To illustrate the importance of this distinction, two extreme cases should be considered: a sample that forms a very large amount of agglomerates that are extremely friable, and a sample that forms a very small amount of low-friability agglomerates. In the first case, even though many agglomerates are formed, these will break up during the friability cycle resulting in most of the material passing through the sieve and a corresponding high % friability value. It is important to note that the % friability is simply the % of material passed through a given sieve size and 100 % pass equates to 100 % friable for that sieve size. In the second case, even though the agglomerates formed are hard agglomerates, most of the material will pass through during the first cycle, which will also result in a high % friability. Both examples would be low-risk for agglomeration in an AFD, but the powders exhibit very different bulk properties. A comparison of the % agglomeration vs. % friability can be found in Figure 7 and will be further discussed later in the paper. Qualitative assessment by visual inspection: The LabRAM method can effectively screen multiple solid and solvent combinations in parallel with all samples being granulated in the same way. By a simple visual inspection the combinations that tend to “ball” or remain as a free-flowing wet cake can be easily determined. From Figure 4, which shows a solvent screen for TAK-117 at 30% LoD, (which is a typical wet-cake solvent composition post filtration), it is clear that the alcohols (methanol, ethanol, isopropanol) all tended to result in balling of the wet-cake while THF, acetone and heptane remained more granular. Solvents that posed minimal risk for agglomeration did not significantly change the texture or consistency of the wet-solids when mixed in the LabRAM; these samples would stay loose and powdery when removed from the vials. In general, it was found that the more “balling” or stickiness the samples showed in the solvent screen, the lower the % friability and hence the higher the risk of agglomeration. For TAK-117 it was concluded that THF or heptane warranted further investigation toward reducing the agglomeration potential during agitated drying. Comparison of the resonant acoustic screening method to the MTR:

ACS Paragon Plus 12 Environment

Page 13 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

The LabRam methodology was found to be very reproducible and independent of how the wet-cake was prepared. A total of six repeats were performed for the TAK-117/water wet-cakes prepared in the LabRAM at 37% LoD. The average extent of agglomeration after drying, based on the % retained on the 2000 µm sieve, was 61% ±2%, which was in close agreement with the %friability determined when the wet-cake was prepared on a pressure filter with a similar residual LoD (% friability determined in this case to be 65% ±4%) . As expected from the qualitative assessment discussed previously, the TAK117/heptanes system (with an LoD of 25%) resulted in fewer agglomerates. Six repeats using TAK-117 with heptane after one LabRam cycle resulted in an average retain of 8% ±6% agglomerates on the 2000 µm sieve. The variability is higher as the quantity of retained agglomerates is much lower. The details of this reproducibility study can be found in the supplemental materials. The LabRAM methodology was then validated against the mixer torque rheometry methodology using the Avicel/water system as the aforementioned reference standard (Figure 6). Very good agreement was observed between the two methodologies. At around 40% LoD, the torque began to increase significantly and the friability dropped below 100%. This represents the critical moisture content at which agitation of the wet-cake during drying should be minimized as the agglomerates generated begin to harden. As the solvent content increased, the agglomerates continued to become harder, reflected by a lower %friability and higher torque, until finally reaching a maximum torque at around 62% LoD. The torque registered by the MTR then decreased as a slurry was generated, whereas the %friability which is a measure of the agglomeration risk of the dried powder, remained constant. While the Avicel/water system was tested up to 70% moisture content, this solvent loading is much higher than what is needed for the purpose of the agglomeration assessment. Typically wet-cakes post filtration will have a moisture content in the range of 30-50% depending on the solvent used, the API’s particle size distribution and how the material was deliquored (e.g. centrifugation typically reduces the LoD to a greater extent than pressure filtration). Therefore, the residual %LoD after filtration should be measured experimentally and those values used to inform the LoDs selected for the study in the LabRam.

ACS Paragon Plus 13 Environment

Organic Process Research & Development

1

100

Relative Torque (Nm)

0.9

90

%Friable

0.8

80

0.7

70

0.6

60

Avicel/Water 0.5

50

0.4

40

0.3

30

0.2

% Friable

20

Torque

0.1

10

0

0 0

10

20

30

40

50

60

70

80

%LoD (water) 1

100

TAK-117/Water

Relative Torque (Nm)

0.9

90

0.8

80

0.7

70

0.6

60

%Friable

0.5

50

0.4

40

0.3

30

0.2

% Friable

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 14 of 25

20

Torque

0.1

10

0

0 0

5

10

15

20

25

30

35

40

45

50

% LoD (water)

Figure 6: Comparison of the rheological torque profile measured by the MTR with the friability determined by the LabRAM screening method (Top: Avicel/water; Bottom: TAK-117/water) It is important to emphasize that the two methods are independent and measure different properties. The MTR merely measures the rheological behavior of the wet-cake which has been correlated to granulation or agglomerate formation. Because the MTR samples are never dried, cake hardening, which is a very common occurrence as API samples are dried especially in a tray dryer or during the static drying period

ACS Paragon Plus 14 Environment

Page 15 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

in an AFD, is not accounted for. The LabRAM method on the other hand, has a granulation step, and the opportunity to observe the wet-mass after mixing, followed by drying of the wet solids and thus assesses the impact of the combination of granulation and subsequent cake hardening due to drying. As previously mentioned, the drying step, as part of this methodology is performed externally to the LabRAM in individual drying pans. This difference between the two methods is clearly demonstrated in the TAK-117/water example shown in Figure 6. While the friability of the LabRAM samples decreases between a moisture content of 1030%, the torque profile registered by the MTR sample remains flat and begins to gradually decrease. The friability test shows that the sample is forming more agglomerates while the MTR suggests that it is less likely to agglomerate. The LabRAM method is therefore able to capture the additional agglomeration due to cake hardening that the MTR method is unable to. Finally, the LabRAM method is a screening method reducing both the time and the quantity of materials required by the MTR method. A full solvent screen can be conducted with the LabRam process in a single day using 99% using a 2000 µm sieve. Ethanol wetcakes on the other hand which were originally used in the process were only 84% friable. Screening of more APIs with high agglomeration potential would be useful to determine whether this observation can be generalized so as to standardize the use of heptane with such compounds as the final wash solvent prior to drying. Effect of particle size and particle size distribution on agglomeration potential: The effect of the particle size and PSD of TAK-117 was then investigated. Interestingly solvent was found to have a much larger impact on friability than PSD (Figure 9). As expected the friability was found to be low for a TAK-117/water system of larger particle size (coarse: d10 = 4.6µm, d50 = 14.1µm, d90 = 40.5µm) across the full range of moisture contents explored, compared to that of a smaller particle size (fine: d10 = 2.5µm, d50 = 6.2µm, d90 = 13.8µm). Jet-milling the larger particle size lot of API (micronized: d10 = 0.4µm, d50 = 1.9µm, d90 = 4.3µm) resulted in material that had a high agglomeration potential. Interestingly a blend consisting of 75 wt% of the coarse lot and 25 wt% of the micronized lot, exhibited a significantly higher agglomeration potential compared to the coarse lot, demonstrating how a relatively small number of fines can have a dramatic impact on agglomeration. However, in heptane, there was little difference in friability between the coarse, micronized, and fine lots, further supporting the use of heptane in designing a more robust drying process.

ACS Paragon Plus 19 Environment

Organic Process Research & Development

100

Coarse

80 Blend

% Friability

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 20 of 25

60 Fine

40 Micronized

20

0 0

10

20

30

40

% LoD Water

Figure 9: Effect of PSD and solvent on the TAK-117 agglomerate friability. Friability was tested with a 2000 µm sieve on TAK-117 lots of different PSD in water and heptanes.

ACS Paragon Plus 20 Environment

Page 21 of 25

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Organic Process Research & Development

Pilot plant scale-up: Based on the LabRam agglomeration screening studies it was shown that wash solvent was critical in controlling and minimizing TAK-117’s agglomeration potential. Furthermore while PSD was not found to be critical, larger particles further reduced the agglomeration potential. The TAK-117 crystallization process was therefore revised to include a seeding step and temperature cycling (Ostwald ripening) to control the nucleation, reduce the number of fine particles and promote growth. This was successfully demonstrated on a 17 kg scale in the pilot plant improving the d50 from 6.9 µm to 19.9 µm. The TAK117/DMSO:Water slurry was filtered using a 0.4 m2 Rosemund AFD and the wash protocol was modified to include an ethanol displacement wash post the water reslurry, followed by two heptanes washes. This successfully removed residual inorganics and controlled residual DMSO levels. While it had been demonstrated using the LabRam methodology that the solvent content of the heptane wet-cake did not impact the API’s agglomeration potential, a period of static drying with a 1 m3/hr nitrogen flow through was incorporated in order to minimize the LoD of the wet-cake. Surprisingly, it was found that drying was complete within 1 hr meeting the residual solvent specifications. Nevertheless, the nitrogen flow through was switched off and the cake was submitted to an agitated drying regime according to which it was agitated for 1 min every 3 min at 11 rpm for a total of 8 hrs in order to assess the impact of agitation on the PSD. No significant change in the PSD was observed over the course of the stress test. A comparison between the original drying process conditions and the revised conditions developed based upon the process development outlined above is shown in Table 2. A sample of the dry powder was finally sieved using a 1000 µm sieve and % agglomeration was found to be