Industrial Crystallization of Pharmaceuticals: Capability Requirements to Support an Outsourcing Paradigm

• Eli Lilly and Company

Introduction

The demands for increasing control of drug substance physical properties in the pharmaceutical industry have evolved considerably in the past decade. Regulatory agencies must be assured that a continuous supply of drug substance can be delivered to the patient, and that the drug substance attributes (e.g., physical properties, chemical purity) used in clinical trials is equivalent to the attributes in the commercial dosage form. Increasing cost pressure, decreasing development time, and increasing speed to the market are driving larger companies to leverage CMOs to achieve delivery timelines for clinical supplies.

Control of crystallization processes requires control of the crystallization kinetics (both nucleation and growth), and the ability to manipulate the kinetics to achieve the desired process result. In most cases, traditional drug substance physical properties such as particle size and specific surface area are no longer achieved by execution of the common top-down mantra of “grow it big and then mill it small”. Today, physical property control strategies focus on delivering the final particle size or specific surface area specifications through a controlled, well- defined crystallization process. In addition, emphasis is placed upon maintaining the physical properties through the subsequent isolation operations such as filtration, drying, and pneumatic conveyance. Crystallization is no longer used to just isolate the drug substance or to improve the impurity profile. The demand for increasing control has shifted crystallization process design to favor crystal growth over nucleation, with the critical factors being the use of seed and control of supersaturation. The successful design of a crystallization process to produce a predetermined physical property (such as particle size) is grounded in population balance theory, and direct application of the appropriate design equations [1, 2]. The primary teaching is that physical properties can be controlled through maximizing crystal growth on welldefined seed. Paul et al provides a detailed summary of the critical issues that must be addressed in order to achieve the desired level of control [3].

Crystal Engineering

Physical property control of the drug substance is achieved through the design of the crystallization process. Two of the most common physical properties that are controlled include crystal form and particle size distribution. Crystal shape (or habit) can also be important since it impacts other physical properties such as specific surface area, bulk/tap density, and powder flowability. The first steps in the design and ultimate control of crystallization is the definition of the physical property control requirements.

Physical Properties

Crystal form refers to the arrangement of the constituent molecules in an orderly repeating pattern extending spatially in all directions. Multiple forms or polymorphs of a material can exist due to different ordering of the constituent molecule within the lattice, both intra-molecularly and intermolecularly. The different arrangements of molecules in the lattice can be detected by x-ray powder diffraction (XRPD) resulting in unique diffraction patterns. Melting points are also different between crystal forms due to the differences in free energy. The free energy difference between crystal forms also affects the solubility of one form versus another and as a result can directly impact bioavailability. Therefore, control of crystal form is one of the most critical requirements for a drug substance crystallization process, and having the ability to identify (through solid state analysis) which crystal form is present is also important. From a characterization perspective, CMOs should have access to a variety of solid state analysis tools to discriminate between crystal forms such as XRPD and DSC.

Particle size is perhaps the most commonly thought of property when physical properties are discussed. However, particle size is the most ambiguous measurement and the absolute value depends on how it is defined and measured. Crystals from industrial crystallization processes have different shapes and different sizes, and although the crystals have a three dimensional length, a one dimensional PSD function is often used in practice to capture and describe the distribution [3]. Typically, the characteristic length is defined as an equivalent diameter of a sphere which has the same behavior under the measurement conditions. Sieving, laser diffraction, image analysis, and sedimentation are measurement techniques that are commonly used to measure PSDs. Although there is no absolute standard for a preferred measurement technique, laser diffraction is gaining wide use throughout the pharmaceutical industry. As a result, it is important for CMOs to have access to this type of analysis equipment.

Particle shape is also important to control for drug substances and there are multiple technologies emerging to quantify particle shape with both in situ and offline sample analysis. However, a simple qualitative understanding of shape is often all that is necessary. To satisfy this requirement, access to an optical microscope with 500 to 1000x magnification and image capture ability is typically adequate.

Crystallization Process Control

Crystallization is a thermodynamically driven process. Crystalline solids in contact with a liquid establish an equilibrium between the concentration of solute in solution and the solid state, and the concentration of solute in solution is the solubility. Solubility is a function of the solution composition (both solvent and impurity levels), temperature, and pressure. If the solute concentration is higher than the solubility, the system is supersaturated. If no solids are present in a supersaturated solution, the system is metastable, and a “small” perturbation to the system can cause crystallization to occur. If the solute concentration is “high”, homogeneous nucleation can occur which, in some cases, can lead to the formation of undesired crystal forms, inclusion of impurities and solvent, and promote conditions that favor particle agglomeration. These issues often translate into a lack of process robustness and variability in the control of physical properties and chemical purity. For most industrial crystallizations, homogenous nucleation is avoided by strict control over supersaturation and the use of seed.

Once crystals are present in a supersaturated solution, supersaturation can be relieved either by crystal growth (desired) or by secondary nucleation. If the crystallization process is maintained in a growth dominated state, final particle size is directly related to the size and amount of seed added to the crystallization, provided that agglomeration, attrition, and secondary nucleation mechanisms are minimized. If the supersaturation is too high (or uncontrolled), secondary nucleation mechanisms can become active and dominate the process outcome. As a result, it is critical to control the supersaturation level in the system. This is done by precise control of the temperature profile during a thermal crystallization, or the rate of addition for an antisolvent or reactive crystallization. In the case of an antisolvent or reactive crystallization, control of mixing can be critical in determining the success of the process. If the state of mixing of the feed stream(s) is imperfect, zones with different levels of supersaturation can develop, resulting in regions of different crystallization kinetics and ultimately polydisperse PSDs.

Mixing in crystallization processes impacts heat, mass, and momentum transfer, and many excellent reviews of the subject have been published. A short list of mixing considerations is outlined below and has been adapted from Tung et al [3] to highlight the importance of mixing and the requirements for equipment design, selection, and operation. It is important to note that mixing requirements for crystallizers incorporate a range of issues grounded in both liquid/liquid and solid/liquid mixing:

• Blending of solution and antisolvent to the molecular level to achieve supersaturation
• Blending of reagents to the molecular level to achieve reactive crystallization/precipitation
• Distribution of local and global supersaturation throughout the crystallizer
• Impact of mixing on crystallization kinetics (nucleation and growth)
• Shear rate management to avoid damage (attrition and breakage) to crystals
• Shear rate management to understand agglomerate formation and break-up
• Impact of mixing on rates of heat transfer
• Maintenance of a crystal slurry of the required solid/liquid ratio
• Avoidance of encrustation/fouling on equipment surfaces
• Avoidance of entrainment of gas/vapor from the head space
• Maintenance of crystal slurry during operations and discharge of the slurry without significant retention of product crystals
• Balancing scale-up criteria for both liquid/liquid and solid/ liquid mixing

Control of crystallization processes and the selection of equipment for both pilot-plant and commercial manufacturing is critical in order to ensure consistency in both chemical and physical properties of the drug substance. An objective of this article is to list common equipment sets and design features that should be in place in order to achieve the desired level of control, and enable teams to address the key considerations outlined above.

The three types of crystallizers that are often encountered for the crystallization of pharmaceuticals include the stirred vessel, fluidized bed, or an impinging jet. The selection of the most appropriate design depends on the specific needs of the process in terms of supersaturation control, mixing quality requirements, and the targeted physical properties of the drug substance [3, 5]. For the greatest fl exibility, access to a range of crystallizer designs is preferred, but conventional pilot-plants and commercial facilities typically standardize on stirred vessels. Figure 1 is an example of a conventional equipment set that can be used for batch and semi-continuous crystallization processes [3, 5, 6-12]. Key components of the layout include: stirred vessel crystallizer, feed vessels (both feed concentrate and antisolvent) with fl ow rate and feedback control capability, wet-milling equipment (e.g., rotor stator mill) for seed conditioning, an optional recycle loop for in-line mixer, PAT, and wet-mill installation, isolation equipment for fi ltration and drying (e.g., agitated fi lter dryer), and a comill to de-lump the drug substance prior to bulk packing. For the stirred vessel crystallizer, having the following functionality is preferred [4,5]:

• Material of Construction: Glass-Lined or alloys of Hastelloy
• Multiple impeller designs: pitched blade turbine, Ekato Intermig, retreat curve
• Multi-stage agitators
• Baffl es
• Variable speed drive
• Subsurface addition lines: access to wide range of inner diameters to avoid backmixing in the feed pipe
• Tickler blade: facilitate discharge of the slurry
• Split jacket services: avoid potential for encrustation/fouling on the crystallizer surfaces at elevated temperatures
• Programmable jacket control: ability for controlled cooling profi les (linear or cubic temperature profi les)
• Automated fl ow rate controlled additions through heat traced lines
• Automated temperature control for jacket services to the vessel
• pH control within the vessel
• Heat traced recycle loop with a low shear rate, low pulsation pump

Figure 1. Typical equipment layout for pilot-plant and commercial scale crystallizations, the equipment set can be used for batch and semi-continuous crystallization processes. Key components of the layout include: stirred vessel crystallizer, feed vessels (both feed concentrate and antisolvent) with flow rate and feedback control capability, an optional recycle loop for in-line mixer and PAT installation, wet-milling equipment (e.g., rotor stator mill) for seed conditioning, isolation equipment for fi ltration and drying (e.g., agitated fi lter dryer), and a comill to de-lump the drug substance prior to bulk packing.

When required, modifi cations to the mechanical setup of the stirred vessel are made to tailor the needs for a particular operation. For example, when intense micro and mesomixing of a feed stream is required, stirred tanks are often not the ideal type of equipment to carry out a robust, reproducible process. Depending on the crystallization kinetics relative to the local mixing time, the additive stream can be charged to a zone of high energy dissipation rate (intense mixing) within an in-line mixing device. Having the fl exibility to install an in-line mixer (e.g., impinging jet [13], mixing tee, mixing elbow [14], or static mixer [9]) for zones of brief, high intensity mixing are generally preferred. Depending on the desired process result, once-through and recycle operations can be carried out with and without the use of seed. In addition, external recycle loops around the crystallizer facilitates the installation PAT equipment, sampling ports, wet milling equipment, and aids in the transfer of slurry to isolation equipment. The selection of a low shear rate pump for the recycle loop is important to ensure crystals that have been grown to achieve the desired physical properties are not attritted or undergo breakage while being recycled. Low shear rate pump designs such as rotary lobe and disc fl ow are often preferred, although diaphragm pumps (equipped with a pulse dampener) are suitable in some cases.

For controlled, growth dominant crystallizations, seed material is milled (conditioned) to reduce the particle size prior to beginning the crystallization. It is preferable to mill the drug substance as a slurry (vs. dry powder) for numerous reasons [15] including increased containment, reduced operator exposure, temperature control for thermal sensitive materials, ability to active the seed surface, and ease of transfer of the seed slurry into the crystallizer. Rotor-stator mills are often used for the initial PSD control and the operation is often termed wet milling (or slurry milling) and is most suitable for high aspect ratio crystals. The milling effi ciency and ability to maintain monodisperse seed PSDs decreases for some drug substances that have a rod-like crystal habit or have a high intrinsic crystal density. In these cases, alternative wet milling strategies can be employed such as stirred bead mills [10, 15].

The ability to control the fl ow rate of feed streams at a precise level is imperative in many antisolvent and reactive crystallization processes. Typical operations involve a pressure transfer of the reagent or feed stream from the feed vessel to the crystallizer, and a fl ow meter in conjunction with a fl ow control valve are used to regulate the fl uid fl ow rate. In addition, having the ability to conduct a “hot” addition by keeping the transfer line heated is generally preferred.

Isolation Equipment (Filtration, Drying) and Comilling

Emphasis has been placed on achieving the desired physical properties in the crystallizer, but it is equally important to maintain the physical properties through the product isolation (fi ltration and drying). Two common equipment sets used for fi ltration and drying include agitated fi lter dryers or a centrifuge coupled to a “pan” dryer. Drug substance that is discharged from either equipment set is typically de-lumped to disperse any large, loosely-bound “clumps” of particles that may have formed during fi ltration and drying. A comill with a broad selection of impeller types and screen sizes is often used for de-lumping prior to bulk packaging.

Case Study

Utilizing the teaching, guidance, and practices provided in this article, the authors recently partnered with a CMO and manufactured over one metric ton of a drug substance with full replication of physical properties from the laboratory scale (~20 g) to the pilot-plant scale (~100 kg/batch). The crystallization was a mixing sensitive reactive crystallization that utilized conditioned seed, in-line mixing, PAT, and automated pH control. Figure 2 shows representative PSDs (measured by laser diff raction) for ten batches of the drug substance and an SEM micrograph of the product crystals.

Figure 2. A representative SEM micrograph and volume weighted PSDs for ten batches of drug substance. The PSD data illustrate the batch-to-batch consistency and reproducibility of the crystallization and subsequent isolation process.

This example highlights that utilization of appropriate design principles, equipment, and operating conditions to achieve control of nucleation and growth are key to the successful development and ultimately determine success or failure on scale-up of crystallizations. It is our hope that the teaching, references, equipment sets, and overall approach to crystallization design presented here can be adopted by the broader CMO community and leveraged with other large pharmaceutical companies to accelerate the development of new medicines for waiting patients.

References

• Randolph, A.D., Larson, M.A., Theory of Particulate Processes, 2nd edition, Academic Press, 1988.
• Larson, M.A., Garside, J., Crystallizer Design Techniques Using the Population Balance, The Chemical Engineer, 1973, 318-328.
• Tung, H.H., Paul, E., Midler, M., McCauley, J., Crystallization of Organic Compounds, An Industrial Perspective, John Wiley and Sons, 2009.
• Paul, E.L., Tung, H.H., Midler, M., Organic Crystallization Processes, Powder Technology, 2005, 150, 133-143.
• Myerson, A.S., Handbook of Industrial Crystallization, 2nd Edition, Butterworh-Heinemann, Newton, MA, 2002
• Mullin, J.W., Crystallization, 4th edition, Butterworth-Heinemann, 2001.
• Perry, R.H., Maloney, J.O., Green, D.W., Perry’s Chemical Engineers’ Handbook, 7th edition, Mc Graw Hill, 1997
• Baldyga, J., Bourne, J.R., Turbulent Mixing and Chemical Reactions, Jon Wiley and Sons, 1999.
• Paul, E.L., Atiemo-Obeng, V., Kresta, S.M., Handbook of Industrial Mixing: Science and Practice, John Wiley and Sons, New York, 2003.
• Cote, A.S., Variankaval, N., Doherty, M.F., From Form to Function: Crystallization of Active Pharmaceutical Ingredients, AIChE, 2008, 54(7) 1682-1688.
• Mersmann, A., Crystallization Technology Handbook, 2nd Ed., CRC Press, 2010.
• Kamahara, T., Takasuga, M., Tung, H. H., Hanaki, K., Fukunaka, T., Izzo, B., Nakada, J., Yabuki, Y., Kato, Y., Generation of Fine Pharmaceutical Powders Via Controlled Secondary Nucleation Under High Shear Environment during Crystallization – Process Development and Scaleup, Organic Process Research and Development, 11, 2007, 699-703.
• Johnson, B.J., Prud’homme, R. J., Chemical Processing and Micromixing in Confi ned and Impinging Jets, 2003, AIChE, 49, 2264-2282.
• Singh, U.K., Spencer, G., Osifchin, R., Tabora, J., Davidson, O.A., Orella, C.J., Incorporation and Characterization of a Mixing Elbow on the Pilot Plant Scale for a Mixing Sensitive Crystallization of an API, Ind. Eng. Chem. Research, 2005, 4068-4074.
• Fisher, E.S., Milling of Active Pharmaceutical Ingredients, Encyclopedia of Pharmaceutical Technology, 3rd Ed., 2006, 2339-2351.

Author Biographies

Christopher Burcham, Ph.D., Engineering Research Advisor at Eli Lilly and Company, leads the Particle Design Laboratory in the Small Molecule Design and Development department within Product Development. Chris received his PhD from Princeton University in 1998, and a BS from the University of Illinois, both in Chemical Engineering.

Daniel J. Jarmer, Ph.D., is an Associate Engineering Advisor and co-directs the Particle Design Laboratory in the Small Molecule Design and Development Division at Eli Lilly and Company. Daniel received a B.S. in chemical engineering from the University of New Mexico (1998) and a doctorate in chemical engineering from the University of Colorado at Boulder (2004).