Abstract
Dual-chamber systems can offer self-administration and home care use for lyophilized biologics. Only a few products have been launched in dual-chamber systems so far—presumably due to dual-chamber systems' complex and costly drug product manufacturing process. Within this paper, two improved processes (both based on tray filling technology) for freeze-drying pharmaceuticals in dual-chamber systems are described. Challenges with regards to heat transfer were tackled by (1) performing the freeze-drying step in a needle-down orientation in combination with an aluminum block, or (2) freeze-drying the drug product “externally” in a metal cartridge with subsequent filling of the lyophilized cake into the dual-chamber system. Metal-mediated heat transfer was shown to be efficient in both cases and batch (unit-to-unit) homogeneity with regards to sublimation rate was increased. It was difficult to influence ice crystal size using different methods when in use with an aluminum block due to its heat capacity. Using such a metal carrier implies a large heat capacity leading to relatively small ice crystals. Compared to the established process, drying times were reduced by half using the new processes. The drying time was, however, longer for syringes compared to vials due to the syringe design (long and slim). The differences in drying times were less pronounced for aggressive drying cycles. The proposed processes may help to considerably decrease investment costs into dual-chamber system fill-finish equipment.
LAY ABSTRACT: Dual-chamber syringes offer self-administration and home care use for freeze-dried pharmaceuticals. Only a few products have been launched in dual-chamber syringes so far—presumably due to their complex and costly drug product manufacturing process. In this paper two improved processes for freeze-drying pharmaceuticals in dual-chamber syringes are described. The major challenge of freeze-drying is to transfer heat through a vacuum. The proposed processes cope with this challenge by (1) freeze-drying the drug product in the syringe in an orientation in which the product is closest to the heat source, or (2) freeze-drying the drug product outside the syringe in a metal tube. The latter requires filling the freeze-dried product subsequently into the dual-chamber syringe. Both processes were very efficient and promised to achieve similar freeze-drying conditions for all dual-chamber syringes within one production run. The proposed processes may help to considerably decrease investment costs into dual-chamber syringe fill-finish equipment.
- Freeze-drying
- Dual-chamber syringe
- Dual-chamber system
- Aseptic processing
- Biologics
- Biopharmaceutical
- Small-volume parenterals
Introduction
Pre-filled syringes are increasingly used for biologics when self-administration and home care use are desired (1). However, standard pre-filled syringes can only be used for products that are stable as a liquid formulation. Unfortunately, some biological drugs are not sufficiently stable in the liquid state and must therefore be freeze-dried for stable storage and reconstituted shortly prior to administration.
The reconstitution from a vial is complex and includes handling of a single- or multi-dose vial with the drug, a second container with the diluent, a needle for liquid transfer, and a different needle for injection. This setup would make self-administration by a patient very complicated, if not impossible.
Dual-chamber pre-filled syringe systems (DCSs) which allow both simple reconstitution and self-administration, have been previously described (2,3). The DCS contains the freeze-dried drug in a syringe or cartridge head chamber and a diluent in a separate second chamber. Both chambers are separated by a middle plunger that, when pressed, moves into a bypass position and allows the diluent to enter the head chamber where it reconstitutes the freeze-dried cake. After reconstitution, the product solution can be administered directly.
However, only a few products have been launched in a DCS so far (4). One of the main reasons for this is likely the costly and complex drug product manufacturing process, that is, filling and lyophilisation. It was thus the scope our study to evaluate and identify novel and improved processes for freeze-drying pharmaceuticals in DCSs.
Established Drug Product Manufacturing Process
The established needle-up drug product manufacturing process (Figure 1A) uses DCSs as a bulk cartridge, which require that the inner walls of the glass barrels are washed/siliconized (step I) and sterilized/dehydrogenized (II). Prior to filling, a middle plunger is inserted into the glass barrel, creating two chambers (III). In a next step, the product solution is filled through the narrow opening of the glass barrel into the head chamber (IV), and a closure is placed onto that head chamber opening (V) in such a way that water sublimation during subsequent freeze-drying is still possible. Alternatively the DCS can be crimped outside the freeze-dryer after the lyophilization step. This, however, necessitates controlled atmosphere in terms of inert gas and/or relative humidity. The product is then freeze-dried inside the glass barrel, and the head chamber is closed prior to removal from the dryer (VI). After capping of the head chamber opening (VII), the DCS is inverted (VIII) to allow filling of the second chamber with diluent (IX). Finally, the end plunger is inserted from above (X) and the DCSs are placed in trays (XI) to move on to visual control and final assembly (XII).
Established (A) and proposed (B, C) process schemes (simplified). (A) Process schemes of the established bulk needle-up DCS fill finish process, (B) a nested needle-down process with freeze-drying in an aluminum block, and (C) freeze-drying in a metal cartridge (LPC-process).
Insertion of the middle plunger before product filling guarantees that the diluent chamber remains free of product. However, the consequence is that product filling as well as freeze-drying have to be performed in a needle-up orientation (5). This leads to a number of limitations. Apart from the complex drug product manufacturing process, limitations concerning head opening design, poor heat and mass transfer during freeze-drying, and potential batch-to-batch and unit-to-unit variability, for example, in terms of residual moisture levels.
Process complexity is caused by the large number of processing steps. These are due to the individual processing of the DCS, involving turning steps, and requiring different transportation systems such as clip systems, magazines, and trays.
Needle-up processing of a DCS is limited to systems where the head opening is sufficiently large for the filling needle to pass. This is fulfilled by, for example, a cartridge-like head opening design with a crimped neck. Apart from this, needle-up processing requires DCS head openings with special and complex closure caps, as it is often desired to close the system in a controlled atmosphere, for example, inside the freeze-dryer.
Heat and mass transfer for freeze-drying of a DCS in needle-up position is poor, as the product resides on the isolating middle plunger at a distance from the heat-controlled shelf. This is the reason why heat conduction and convection (normally the main mechanisms for heat transfer in freeze-drying of conventional vials) become marginal and heat transfer is mainly driven by radiation. In addition, with radiation being the main heat transfer mechanism for this setup, also batch-to-batch and unit-to-unit homogeneity can be adversely affected, as the DCS in the periphery of the shelf receives additional radiation from the chamber walls and doors compared to the DCS located in the center.
Last but not least, being designed to hold both product and diluent, a DCS needs more volume per unit compared to conventional vials, and therefore a general drawback of a DCS is that only a smaller batch size, compared to vials, is feasible with a given size of a freeze drier.
It was the aim of this study to explore novel processes and approaches towards drug product manufacturing of a DCS in order to overcome current limitations. This includes assessing the use of a DCS in a tray format (Figure 1B, 1C), by testing and comparing process efficiency and the product impact to the current process.
Materials and Methods
Primary Packaging Containers and Model Formulations
Vials:
Two milliliter, type 1 glass vials from Schott AG (Mainz, Germany) and suitable lyophilization (lyo) stoppers from Daikyo Seiko Ltd. (Tokyo, Japan) were used as standard lyophilization container closure system (Figure 2A).
Schematic illustration of drying configurations used in this work. (A) vials, (B) DCS in needle-up process, (C) DCS in needle-down process, and (D) LPC. Physical properties such as empty container mass (m), heat capacity (C), area relevant for heat transfer from the shelf (A1), area relevant for heat transfer from the aluminum block/LPC to the product (A2), sublimation front area (As), and fill height (h) for V = 1.2 mL are given below each illustration.
Dual-Chamber Systems (DCSs):
Three millimeter DCSs from Nipro Glass Germany Ltd. (Muennerstadt, Germany), manufactured from FIOLAX® clear (alkaline-earth containing borosilicate glass), with a barrel length of 73.1 mm, an outer diameter of 10.85 mm, an inner diameter 8.65 mm, a bypass length of 13 mm, and different head designs (cartridge opening and luer cone) were used in this study (see also Figure 2B and 2C). The closures used were either 7.65 mm aluminum crimped caps from Dätwyler Holding (Altdorf, Switzerland), including a polytetrafluoroethylene (PTFE)-coated rubber septum or PTFE-coated Luer Tip caps from Daikyo Seiko Ltd. Middle and end plungers were supplied by Daikyo Seiko Ltd.
Lyophilization Process Cartridges (LPCs):
Five hundred precision tubes of AISI type 304 stainless steel made by Interalloy (Schinzach-Bad, Switzerland) were used as LPCs. The tubes had a total length of 38 mm, an inner diameter of 8 mm, and a wall thickness of 1 mm (Figure 2D). To easily insert a plunger, the tubes have one flared end with a diameter of 12 mm. To enable better shelf contact and the ability to stand freely, the flared ends were mechanically reworked.
Aluminum Block:
As part of this work, an aluminum block carrier system (for efficient primary packaging transportation during filling and optimized heat transfer during freeze-drying) was designed and tested.
The custom-made metal block, with a total height of 66 mm, was manufactured from anodized aluminum with an emissivity value of 0.77 (see also Figure 2C). The honeycombed outer shape of the block allows easy scale-up. Each block comprises 24 cavities that are arranged in the same hexagonal pattern and distance as in a standard 2.25 mL syringe tray. For optimum contact between the metal block and the DCS, the cavities are stepped bores with a diameter of 13 mm at the surface and 11 mm diameter at the bottom. In the needle-down orientation the block-to-glass distance is approximately 0.1 mm for the product chamber of the DCS, whereas the DCS bypass position resides on the conical tapering, freely rotatable.
Model Formulations:
Bovine serum albumin (BSA) Fraction V from Roche Diagnostics (Penzberg, Germany) was used as a model protein. It was formulated as 75 mg/mL BSA in a 10 mM Histidine buffer of pH 6.0 (Ajinomoto; Tokyo, Japan) with 125 mM Sucrose (Ferro Pfanstiehl Co.; Mayfield Heights, OH, USA) and 0.02% Polysorbate 20 (Croda International; Snaith, UK). The formulation was sterile-filtered using 0.22 μm hydrophilic polyvinylidene fluoride (PVDF) filter units (Millipore; Bedford, MA, USA).
In addition, two monoclonal antibodies provided by F. Hoffmann-La Roche Ltd. (Basel, Switzerland) were investigated in freeze-drying studies. Testing these formulations was performed to demonstrate the pharmaceutical relevance and applicability for highly concentrated antibody formulations in DCSs, Formulation A contained 150 mg/mL of monoclonal antibody 1 (mAb1) in a 20 mM Histidine/HCl buffer of pH 5.5, 200 mM Trehalose, and 0.04% Poloxamer 188 and was diluted to 75 mg/mL using water for injection (WFI). Formulation B contained 25 mg/mL of mAb2 in a 5 mM Histidine/HCl buffer of pH 6.0, 60 mM Trehalose, and 0.01% Polysorbate 20. Both formulations were sterile-filtered using 0.22 μm hydrophilic PVDF filter units before filling.
For the freeze-drying tests, 1.2 mL of each test solution was filled into washed and sterilized containers.
Lyophilization Cycles
Sublimation Tests:
In sublimation tests with pure water, the mass flux during primary drying at different chamber pressures Pc was determined gravimetrically.
For the experiments a volume of 1.5 mL WFI was filled into each container and the weight loss was measured individually using a XP105 balance from Mettler Toledo (Greifensee, Switzerland). Because water has no product resistance, the fill volume for sublimation experiments has no impact on sublimation rate.
The sublimation tests were performed using a FTS Lyostar II freeze-dryer from SP Scientific (Stone Ridge, NY, USA). The Plexiglas door was covered with aluminum foil on the inside to reduce radiation effects. During each test run, the freeze-dryer contained:
96 vials directly placed on the freeze-dryer shelf in a closely packed hexagonal profile;
a total of 4 times 96 dual-chamber containers (total 384): (a) 192 were inserted into aluminum blocks (needle-down), and (b) 192 were directly placed on the shelf in typical nest orientation and container distance (needle-up);
192 LPCs placed on the shelf in the same nest orientation and with the same container distance as the DCSs.
The freezing conditions were kept the same for all test runs (15 min precooling at T = 5 °C, ramping at 1 °C/min to T = –5 °C, holding for 15 min/ramping at 1 °C/min to T = –40 °C, holding at T = –40 °C for at least 3 h), but chamber pressure set points were varied between 6.7 and 40.0 Pa (6.7, 13.3, 20.0, 26.7, 33.3, or 40.0 Pa). After the freezing step, the chamber pressure (Pc) was lowered to the designated set point and the shelf temperature (Ts) was raised to Ts= –5 °C to allow a considerable mass flow. To measure the product temperature, thin wire thermocouples type T from Omega Engineering Inc. (Stamford, CT, USA), made from a Cu-CuNi alloy with a diameter of 0.08 mm, were placed centrally at the bottom of at least two center and two edge containers for each of the above described configurations. Additional thermocouples were used to probe temperatures on the carrier systems, the shelf surface, and LPCs. Each sublimation experiment was repeated at least once. Mass loss during the ramping phase was determined and subtracted as a blank.
Freezing Studies:
The influence of the freezing rate on the cake structure and the primary drying time was assessed for the BSA model formulation. During testing, the freezing rate was varied from slow (0.3 °C/min, Cycle A) to medium (0.5 °C/min, Cycle B) to fast (1.5 °C/min, Cycle C), and an additional test was performed that combined fast freezing with an annealing step (1.5 °C/min + annealing, Cycle D). Further details are shown in Table I (Cycles A to D). Each container was filled with a target volume of 1.2 mL BSA formulation. For each of the above described Cycles A through to D, the following systems were used:
(I) 96 vials;
(IIa) 192 DCS inserted needle-down into aluminum blocks and (IIb) 192 DCSs were processed needle-up during Cycle A;
(III) 192 LPCs were processed.
Overview of Performed Drying Cycles (Shelf Temperature, Ts; Chamber Pressure, Pc)
The detailed setup of the containers in the freeze-dryer was as described above for the sublimation tests. For each Cycle A–D and each configuration I–III, the product temperatures of three containers placed centrally and two containers placed at the edges were monitored using thin wire thermocouples of type T to obtain the time at freezing equilibrium and the primary drying time. Monitoring temperatures of edge and center containers allowed confirmation of batch homogeneity within sublimation tests (data not shown).
From the resulting temperature curves, the time at freezing equilibrium was defined as the time between start of ice nucleation and the first inflection point of the product temperature curve after reaching freezing equilibrium temperature. The drying time was defined as the time between the start of primary drying and the time when the product temperature reached a plateau value close to the shelf temperature.
Drying Studies:
Drying cycles with product temperatures in the primary drying phase which were lower or equal to the formulations Tg' were defined as gentle, others as aggressive. The influence of gentle and aggressive drying conditions on the primary drying time of the BSA model formulation and the two mAb formulations, mAb1 and mAb2, was assessed. Details of the performed drying cycles are shown in Table I (gentle, aggressive). Here, gentle drying conditions refer to a shelf temperature of Ts = –20 °C and aggressive conditions to a shelf temperature of Ts = +10 °C.
Each container was filled with a volume of 1.2 mL of the individual formulation. For each of the three above described formulations (BSA, mAb1, mAb2) and two mentioned cycles (gentle, aggressive) (I) 96 vials, (II) 96 DCSs inserted needle-down into aluminum blocks, and (III) 96 LPCs were processed.
For each cycle and each configuration I–III the product temperatures of four centrally placed containers were monitored using thin wire thermocouples of type T to obtain the primary drying time.
The drying time was defined as the time between the start of primary drying and the time when the product temperature reached a plateau value in the range of the shelf temperature.
Fill Height Studies:
Fill height studies were performed for the different containers and configurations to assess the ratio of influence of fill height to surface area on primary drying efficiency. BSA formulation was used as model system, and the following fill heights were tested: (I) 4.0, 5.0, 7.9, and 10.0 mm for the vials, where the fill height of 7.9 mm corresponds to a fill volume of 1.2 mL; (II) 7.9, 10.0, 11.2, and 22.4 mm for the DCSs inserted needle-down into aluminum blocks, where the fill height of 22.4 mm corresponds to a fill volume of 1.2 mL; and (III) 4.4, 7.9, 8.7, 10.0, 13.1, and 26.2 mm for the LPCs, where the fill height of 26.2 mm corresponds to a fill volume of 1.2 mL.
For a better comparison of the different containers or configurations, the drying conditions were adapted to achieve comparable product temperatures. For vials, primary drying was performed at Ts = 10°C and Pc = 20 Pa (M); needle-down processing conditions for the DCS in the aluminum block were Ts = –10 °C and Pc =10 Pa (E); and primary drying conditions for the LPCs were set to Ts = –10 °C and Pc =5 Pa (K). Details of the drying cycles are shown in Table I (Cycles M, E, K). For each ratio and each configuration I –III, the product temperatures of four centrally placed containers were monitored using thin wire thermocouples of type T to obtain the primary drying time.
Product Characterization
Scanning Electron Microscopy (SEM):
SEM images of the lyo cakes were acquired on a Sigma VP system from Zeiss (Oberkochen, Germany). The secondary electron detector under high vacuum was used for the image acquisition with an acceleration tension of 3 kV. Line averaging of 17 scans per frame was applied to reduce noise, resulting in a full acquisition time of 44.6 s per image. In order to allow good conductivity, the samples were gold-sputterred using the Auto Sputter 108 from Cressington Scientific Instruments Ltd. (Watford, UK). The parameters for sputtering were set to 120 s with a current intensity of 30 mA under a flow of Argon at 0.1 bar.
Light Microscopy:
Light microscopy samples were prepared as described by Lam et al. (6) using the Sylgard 184 polydimethylsiloxane (PDMS) polymer kit from Dow Corning Corp. (Midland, MI, USA). Having been removed from the lyophilizer, the containers were carefully shattered and the embedded cake was sectioned using a custom-made razor blade assembly. A fluorescent dye was applied to the surface of the embedded lyo cake slices using a yellow highlighter (Schneider Schreibgeräte GmhH, Schramberg, Germany), and the excess dye was wiped off. A UV-LED flashlight, at a wavelength of 395 nm, was used to illuminate the samples, and images of the embedded lyo cake slices were acquired using a Stemi 2000-C light microscope in combination with the Axio Vision LE software, both supplied by Karl Zeiss AG (Oberkochen, Germany).
Heat and Mass Transfer
During the freeze-drying cycle, energy in terms of heat is needed to overcome the specific sublimation enthalpy ΔHice of water to enable the removal of water from the system. Energy contributions are from conduction (ΔQ/Δt)cond, convection (ΔQ/Δt)conv, and from radiation (ΔQ/Δt)rad, for example, from the surrounding walls and shelves of the dryer and the holder system itself.
The amount of energy per unit time ΔQ/Δt or heat flux that is necessary to remove water during steady state primary drying can be described by the following equation:
Here, the sublimation rate Δm/Δt describes the amount of water Δm that is removed during the time period Δt, and the specific sublimation enthalpy of water is ΔHice =2805 J/g (7). It defines the amount of energy that is necessary to transfer 1 g of water from the solid state (ice) to the gas phase (water vapor).
During primary drying the energy for the sublimation of ice is provided by the heated shelves and the surrounding walls. The sublimation rate, or mass flux Δm/Δt, is primarily driven by the difference in water vapor partial pressure between the sublimation interface and the condenser, the available sublimation area. It is limited by any mass transfer resistance between the sublimation front and the condenser surface (e.g., dry layer resistance, container resistance, chamber resistance, etc.).
In order to compare the heat transfer efficiency of the different investigated processes, a heat transfer coefficient Kx was used that took all heat transfer contributions into account.
Tsource and Tsink are the temperatures of the heat source (e.g., the shelf, carrier system, or LPC) and the heat sink (e.g., product, carrier system, LPC), Ax represents the area of heat input, and Kx the heat transfer coefficient for the investigated container and processing conditions (chamber pressure and shelf temperature). As described in the literature Ax for a vial can be attributed to its cross-sectional area or, in other words, the bottom of the vial that is in direct contact with the shelf (8). For the aluminum block we calculated the block footprint relative to the number of cavities. In the case of an LPC, the contact area to the shelf is only the ring-shaped metal-to-metal contact, as the plunger can be regarded as isolating material. For syringes a different definition was published by Korpus et al. (9). In the case of heat transition from shelf to a DCS without carrier (or from carrier system to frozen liquid, or LPC to frozen liquid), the average of the surface areas of the container that are in contact with the product between the beginning and end of primary drying was defined as the heat transfer area ADCS.
h0 and hend PD represent the height of the frozen solution at the beginning and the end of primary drying, and d is the diameter of the DCS/LPC.
Combining eqs 2 and 1, Kx can be defined as the ratio of the area normalized heat flow to the temperature difference between Tsource (e.g., shelf) and Tsink (e.g., product) (8), multiplied by the specific sublimation enthalpy of water. In other words, it describes how effective energy transfer between heat source and heat sink takes place, but there is no differentiation between the contributions of conduction and radiation.
Here, Kx can be obtained from measuring the weight loss during primary drying and monitoring the product temperature (Tsink) and the shelf temperature (Tsource). The overall heat transfer coefficient (Ktotal) of a series of heat transfers (e.g., from the shelf to the carrier system/LPC and subsequently from the carrier/ LPC to the frozen product) was defined according to eq 5, where Atotal represents the mean of the two areas of heat input (9).
The overall heat transfer can be used to compare drying systems (9).
Assessment of Batch (Unit-to-Unit) Homogeneity
Batch homogeneity, in a freeze-drier, can be affected when the sublimation rates of containers located in the center of the shelf and peripheral containers differ during primary drying due to additional heat fluxes, caused, for example, by radiation from the surrounding walls. In such cases, a so-called edge effect can be observed.
Two numerical indices, Cohen's effect size d and the coefficient of variation cv, are useful to quantify batch homogeneity.
Cohen's effect size d compares the two mean sublimation rates (M) of a collection of edge and center containers relative to their pooled standard deviations (SDs). A larger absolute value for d indicates a stronger batch inhomogeneity.
For the calculation of means and SDs, sublimation rates at 13 Pa of 216 centrally located containers and 216 edge containers were used for each configuration (vials, DCSs, LPCs).
The coefficient of variation cv, center is defined as the ratio of SD and mean of the centrally located container sublimation rates. It describes the variability of the centrally located containers, which is important especially for larger chamber sizes, where the number of centrally located containers relative to the peripheral containers increases. A large value for cv, center indicates a poor homogeneity of centrally located containers.
For the calculation of cv, center the sublimation rates at 13 Pa of 192 centrally located containers were used for each configuration (vials, DCSs, LPCs).
Results and Discussion
Needle-Down Process
Figure 1B and Figure 3A describe an alternative process for freeze-drying in a DCS. Using a so-called needle-down process, the DCS is processed in a tray format, where the DCS head is pointing downwards and does not change orientation during the complete manufacturing process. This is the most notable difference to the established needle-up process (10). For this alternative process, ready-to-fill (RTF) tubs containing a DCS and middle plungers in a special tray system, shown in Figure 3A, were used. [7]After being unpacked, the RTF tray was placed on the tray filler and product solution was filled through the open end of the DCS. For the freeze-drying process, the trays were placed onto a metal block, which comprises cavities to hold the DCS barrels to assure effectual heat transfer to the DCSs. At the end of the freeze-drying cycle, the lyophilization shelves moved towards each other, pushing the plunger tray out of its filling position and centering the middle plungers onto the DCS end openings (Figure 3A). The DCSs are then closed. Releasing the partial vacuum from the drying chamber leads to underpressure inside the DCS, drawing the middle plunger into the barrel (Figure 3A).
Proposed process mechanisms for (A) needle-down processing and (B) LPC processing.
After unloading the trays from the freeze-dryer, the same filling equipment was used to fill the diluent through the open end of the DCS. The diluent chambers were subsequently closed by inserting end plungers using a standard stoppering station. In a final step, the DCS trays were placed back into the tubs, closed, and transferred to visual inspection.
The needle-down processing, in a tray format, offers many possible advantages for the process, enabling the manageability of many different DCS head designs such as the Luer cone, staked-in needle, or cartridge and the use of standard tray filler technology for manufacturing (characterized by a relative compact and flexible processing line). Heat transfer from the shelves to the product is considered to be improved for freeze-drying in a needle-down orientation, as the product sits relatively close to the shelf and is surrounded by the aluminum block. As already shown in the literature for vials, the use of a metal block offers increased heat transfer due to its excellent thermal conductivity. Heat transfer due to conduction and convection are anticipated to be the determining heat transfer mechanisms. The block assures that each individual DCS experiences approximately the same amount of energy, which leads to enhanced homogeneity across the batch (11).
Needle-down processing further allows better use of the space within the DCS product chamber. The product chamber can be completely filled with drug product, leading to smaller syringe sizes compared to needle-up processing.
One potential risk still remains in the proposed needle-down process in the case of a DCS with staked-in needle or Luer cone being used. When filling the active pharmaceutical ingredient (API) into the DCS, the API may unintentionally also enter the needle or Luer cone and dry out there. This could result in the risk of prolonged reconstitution times. As a needle is not transparent and a Luer cone is covered by a tip cap, dried product in the needle/Luer cone cannot be detected by visual/optoelectronic inspection.
Lyophilization Process Cartridge (LPC) Process:
A completely different approach to overcome the limitations for current and proposed DCS manufacturing processes are shown in in Figure 1C and Figure 3B. Here, freeze-drying is not performed in the DCS but in a separate cylindrical metal tube. This is termed the lyophilization process cartridge (LPC) (12). After washing and sterilizing the LPC, a plunger is inserted into the tapered end of the metal tube to seal it. Then, the liquid drug product is filled through the upwards pointing end, and the filled LPC is placed into the lyophilization chamber for the freeze-drying process. As described above, for the needle-down process, RTF DCSs and standard tray filling lines were used. For the transfer of the freeze-dried product from the LPC into the DCS, a vent tube was aligned with the DCS and the LPC. The LPC, containing the freeze-dried product, was placed on top of the vent tube. The vent tube was lowered so that it is partially inserted into the DCS and its lower end resides in the later position of the middle plunger (Figure 3B). A piston moved downwards, pushing the plunger and the lyophilizate through the vent tube, until the lower end of the plunger was arranged into its final position within the DCS (but still inside the vent tube). In the next step, the lyophilizate was transferred into the DCS and the vent tube was moved upwards while the piston remained in place, so that the vent tube releases the plunger that—upon being released—seals the inner wall of the DCS above the bypass. Subsequently, the plunger that sealed one end of the LPC is transferred together with the lyophilizate into the DCS, where it finally acted as middle plunger. The liquid solvent was then filled into the diluent chamber above the plunger. Finally, an end plunger was inserted into the DCS followed by further process steps, for example, visual inspection.
The already cited advantages of a relatively simple process, using RTF DCSs and standard tray processing equipment, also apply here. Further, the very good heat transfer properties (already described for the aluminum block) also holds true for the LPCs made out of stainless steel. Hence, the heat conduction is likely to be the determining heat transfer mechanism over radiation and assures good homogeneity across the batch.
Opposite to the above described needle-down process, the space in the freeze-dryer can be very efficiently used, as the LPC that needs to only house a plunger and the product is much shorter than a DCS. The above described risk of residual product remaining in the diluent chamber is solved for the LPC process due to the vent tube. This arrangement assures that the inner part of the glass barrel, which later forms the diluent chamber, is never in contact with the product. The other potential risk of liquid product entering geometries such as a staked-in needle (SIN) or Luer cone prior to drying is addressed here by performing the freeze-drying step outside the DCS, and filling the already freeze-dried product. Furthermore, the LPC offers the possibility to fill multiple lyo cakes into a single DCS. This allows a combination of two or more freeze-dried products in a single DCS, for example, in fixed dose combinations.
In the subsequent sections, the efficiency of the freeze-drying process and possible impact on the product are investigated for the above described established and newly proposed processes for freeze-drying in a DCS.
Sublimation Tests
Heat and Mass Transfer:
According to the steady-state theory of heat and mass transfer, the heat input during primary drying is proportional to the amount of water that is removed via sublimation (8). Thus, the sublimation rate can be used to describe the energy transfer efficiency during primary drying (9).
Figure 4A illustrates the specific sublimation rates for (I) vials, (II) DCS during the needle-up process, (III) DCS during the needle-down process (including a metal carrier), and (IV) for the LPC process. All containers that were standing centrally located in a hexagonal packed array were gravimetrically measured before and after processing under different chamber pressures.
Illustration of area-related mass flux (A) and calculated steady-state heat transfer coefficients (B) obtained from gravimetric sublimation experiments with water for injection (WFI) at different chamber pressures for the investigated configurations: (♦) vials, (
) DCS in needle-up position, (■) DCS in needle-down position + metal carrier, (
) LPC process). Each data point represents the mean value of n= 32 centrally located containers. (A) The mass flux data are normalized to the sublimation front area. (B) The heat transfer coefficients are normalized to the area of relevant heat transfer which, in the case of vials, is the container cross-sectional area, and for the DCS and LPC the average container surface area that is in contact with the product.
Figure 4A shows that the average area-related sublimation rate was similar for the vial process and the established DCS needle-up process. Both processes show an average sublimation rate of 22 μg s–1 cm–2 at a chamber pressure of 16 Pa. It has to be noted, however, that the relevant sublimation area of the vial was a factor of 3 larger (Figure 4 and Figure 2) than that of the DCS, resulting in considerably shorter drying times for the vial if fill volumes are similar.
For both, the DCS needle-up process and the vial process, the specific sublimation rate increased with chamber pressure by approximately 29%. This is in good agreement with Korpus et al., who measured an increase of 20% during their investigation of the energy transfer in the needle-up process (9). The area-related sublimation rates for the DCS needle-down and LPC process, however, were more than 3 times higher compared to the needle-up and vial process. This confirms good heat transfer capabilities of the metal block and the LPC. Here, heat transfer is driven by the excellent heat conduction in the metal with, at the same time, almost negligible radiation influence.
The sublimation rate for the DCS needle-down process increases with chamber pressure by 26%. This is in agreement with data from Patel and Pikal, who completely immersed syringes into an aluminum block and found an increase of 31% (13). At the same time, the obtained values for LPCs do not show distinct pressure dependence, which can be explained as a result of the decreased driving force for sublimation. This driving force is the pressure difference between the vapor pressure of ice at the sublimation front and the chamber pressure, which in this case did not largely change with an increase in chamber pressure. With a decrease in sublimation rate, there was less self-cooling of the LPCs and hence product temperature increased (by 12 K/30 Pa).
Overall heat transfer coefficients Kx versus chamber pressures are shown in Figure 4B; these were calculated according to eq 4 and corrected for non-steady-state conditions during the ramping phase. The y-intercept of Figure 4B corresponded to the sum of pressure-independent terms.
Heat transfer coefficients in DCSs in the needle-up orientation were poor compared to heat transfer coefficients in vials (e.g., at chamber pressure of 16 Pa Kx this was 17 W m–2 K–1 for vials and 6 W m–2 K–1 for DCS needle-up). However, heat transfer using the suggested processes that utilize a metal block or container are much more effective, for example, at 16 Pa Kx this was 28 W m–2 K–1 for the needle-down process and 21 W m–2 K–1 for the LPC process (Figure 4B).
For typical chamber pressures between 6 and 26 Pa (as commonly employed for freeze-drying of pharmaceuticals), both conduction and radiation are the main heat transfer mechanisms for vials and DCSs without carrier as reported by Korpus et al. (9). A needle-down orientation improved heat and mass transfer due to the proximity of the product to the shelf. Heat transfer efficiency of the metal block and heat transfer of the LPC is excellent due to the direct contact of metal and product. Heat for the LPC process is mainly driven by conduction; contributions of convection (moderate slope of heat transfer coefficient versus chamber pressure, Figure 4B) and radiation (small emissivity constant, Figure 2) are minor. Heat transfer in the needle-down orientation (immersed in an aluminum block) is driven by conduction and convection. Convection is dominant for the DCS due to the cavities in the block being slightly larger than the DCS barrel and thus the glass barrel is very close to the metal but with minimal direct contact. It is known that carrier systems improve heat transfer and alter the heat transfer mechanism from radiation to a larger contribution of conductive heat transfer (11,13). From a heat and mass transfer perspective, the needle-down process and the LPC process are thus preferred although these are currently not an industry standard.
Batch (Unit-to-Unit) Homogeneity:
It has been reported that a DCS placed at the edge of commercially used magazines during the needle-up processes experiences additional radiation from freeze-dryer walls and doors, which can compromise batch homogeneity (14⇓–16). In order to qualify batch homogeneity, two index numbers were introduced. These are the edge effect size (eq 6) and the variance of centrally located containers (eq 7). The edge effect describes the difference in sublimation of containers around the periphery and center. The variability of centrally located containers describes the spread of their sublimation rate.
For the used 2 mL vials, an edge effect was found in the range of 2.4 (Details are shown in Table II). Centrally located vials showed a variability of 14%. The needle-up process was found to show numbers in the same order of magnitude but slightly smaller compared to vials. An edge effect can be minimized by shielding the containers. Here, the use of metal blocks is shown for both DCS and vials (11,13).
Batch Homogeneity Based on Mass Flux of Pure Water Measured at Chamber Pressures of 16 Pa for the Proposed Processes and Containers
The DCS in the needle-down process resides within a cavity that fully surrounds the product and completely shields it from the other containers (17). It is thus not surprising that the proposed needle-down process shows very little edge effect (0.1) and minimal centrally located container variability (6%). The LPC process shows an edge effect of 1.2, which is approximately half the size obtained for vials and a variation of centrally located containers (12%) compared to vials (Details are shown in Table II). One explanation for this observation might be the difference in emissivity. Emissivity for glass is close to 1, but the LPCs (made out of stainless steel) showed an emissivity coefficient of 0.66. Therefore LPCs (even unpolished) show a lower tendency to absorb heat via radiation.
Freezing Considerations
Rapid freezing commonly leads to smaller ice crystals, which corresponds to smaller pores and thus larger product resistance and a higher specific surface area. In the case of slow freezing, however, concentration gradients might form with potentially negative impact on product quality. Because primary drying takes the most time and is an energy-consuming process step, it would be desirable to have large ice crystals with a low product resistance for fast primary drying. All of the above mentioned aspects need to be considered for the freeze-drying cycle development, and special care must be applied to the design of the freezing step because it considerably affects product quality and also has a large impact on the efficiency of primary drying.
Figure 5 shows the time at the freezing equilibrium (A) and the primary drying time (B) for the investigated processes when varying freezing conditions (model formulation BSA; Cycles A through D as described in Table I). The data and pore structure images for the needle-up DCS configuration, at a slow freezing rate (Cycle A), are included here for comparison reasons. It is clear that the heat transfer for the needle-up process is poor and leads to very slow freezing and also long drying times. This is due to the limited heat transfer during primary drying compared to all other configurations. The product pore sizes for needle-up processing at a slow freezing rate were nevertheless found to be comparable to the vial and LPC process (Figure 6, Figure 7). For all configurations, the time during freezing equilibrium decreases with an increasing freezing rate (Figure 5A, Cycles A through C). The effect, however, varies in intensity depending on the overall heat transfer efficiency of the configuration. For the LPC, the time required for freezing equilibrium is significantly reduced from 28 to 15 min with increasing freezing rate; however, this time is considerably reduced from 7 to 5 min for the needle-down process in the aluminum block.
Freezing equilibrium time (A) and primary drying times and product temperatures at steady-state (B) for different freezing conditions (Cycles A through D) obtained from freezing experiments with the model formulation BSA.
Scanning electron microscopy (SEM) images of lyo cake structures having applied different freezing rates (model formulation BSA; Cycles A through D).
No impact of the freeze rate on primary drying times was observed for vials (Figure 5B). However, it was noted that a linear increase in product temperature occurred. SEM images of the pore structures (Figure 7) show a decrease in pore size, and light microscopy images of the lyo cake (Figure 6) showed evidence of crack formation in the cake with increased freezing rate. The constant drying time, for vials with varying freezing rates, maybe due to the balanced effects of pore size and crack formation in the lyo cake. Increased product temperature is a result of smaller pore sizes (increased product resistance), increased crack formation thus reducing product resistance/drying time.
Light microscopy images of lyo cake structures, having applied different freezing rates (model formulation BSA; Cycles A through D).
Due to the freezing equilibrium generally being shorter for the needle-down process and not changing with freezing rate, it was expected that neither cake structure, product temperature, or the primary drying time would change (Figure 5). This expectation was confirmed by the SEM images (Figure 7), which show consistently small pore sizes for all freezing rates. The above described observations can be explained by the large heat capacity of the aluminum block in combination with its excellent heat transfer. This interaction buffers the heat during freezing. The physical properties of the aluminum block lead to effective cooling and result in short freezing equilibrium times. However, the current block design made it impossible or excessively difficult to control ice crystal size by varying freezing conditions. This lead to relatively high product temperatures during primary drying. The slightly longer drying times (for the needle-down process) compared to the vial process are caused by the smaller sublimation area in combination with increased product fill height.
For the LPC process, the freezing equilibrium time considerably decreased with higher freezing rate, and a steep increase in product temperature was observed to have only a small impact on the primary drying time (Figure 5B). SEM images also showed the expected smaller pore structure for higher freezing rates, leading to higher product resistance (Figure 7).
Contrary to the vials, where the strongest temperature gradient is in the vertical direction, freezing in a DCS and LPC starts at the container side walls and propagates towards the center. An anisotropic, lamellar structure was observed for the LPC processing using fast freezing rates (Figure 6C), which resulted in ice crystals developing on the side walls of the container and growing towards the center as the liquid was cooled. For fast cooling rates, the orientation of the lamellar structure is parallel to the temperature gradient from the metal wall towards the center, as heat transfer in the metal tube is much faster than heat transfer within the freezing liquid. Despite the observed inhomogeneous structure formation, SEM images do not indicate any micro-collapse or melt-back structures. The overall orientation of the ice structure does not completely consist of horizontal lamellae. There are slight inclines into the vertical direction, towards the center of the cake (Figure 6C).
With these considerations concerning pore structure in mind, the slight reduction in primary drying time for faster freezing rates might be explained by a lower product resistance in the vertical ice crystal channels that results in an increased sublimation rate within the center. In order to achieve large ice crystals, an annealing step maybe useful (Cycle D). Light microscopy and SEM images (Figures 6 and 7) confirmed the pore size was consistently larger for cycle D (fast freezing + annealing) compared to cycle C (fast freezing only). The necessary annealing time, however, contributes to the overall cycle time. Hence, the overall cycle time might be negatively affected. Comparing Cycles C and D, the annealing step, in this study, increased the overall cycle time by 3 h and reduced the primary drying time by only 1 h. This was dependent on the configuration (Table I and Figure 5). For the configurations, formulations, and process conditions that were used in this study, an annealing step did not lead to a decrease in overall cycle time.
Drying Considerations
Figure 8 compares the different processes for an aggressive and a gentle drying cycle. For each protein formulation the drying times and the corresponding product temperatures are shown.
Drying times and product temperatures for the BSA model formulation (A, B), mAb1 (C, D), and mAb2 (E, F). Black bars represent vial data, hatched bars represent the needle-up process, white bars represent the needle-down process, and gray bars represent LPC processing.
Drying in the needle-up configuration results in approximately double the drying time compared to a vial process for both aggressive and the gentle cycle conditions. This can be explained by the lack in heat transfer. The drying times for the BSA model formulation are comparable for the vial process, needle-down process, and LPC process (Figure 8A). The product temperatures are, however, approximately 2 °C higher in the needle-down and LPC process compared to vials using gentle drying conditions and up to 9 °C higher (LPC versus vial) using aggressive drying conditions. Such an increase in product temperature is caused by the unfavorable slim aspect ratio in combination with the excellent heat transfer of the LPC and the process. However, caution must be applied. Despite the good heat transfer, the product temperature always stays below its glass transition or collapse temperature. Gentle drying conditions result in fewer product temperature differences; however, the differences increase when looking at the drying times. This may be explained by the obvious difference in fill heights and product resistances. The drying cycles were repeated using different antibody formulations (Figure 8C to 8F). The trends in temperature and drying time were similar; however, the absolute numbers were different.
In summary, the drying time for the needle-up process is not comparable to vials and primarily limited by the lack in heat transfer. The other proposed processes are solely limited due to the high fill height (at constant volume), which results in higher product resistances. As expected, applying the same drying conditions for vials and other containers results in different product temperatures and drying times. Thus, each configuration and product requires its own optimized drying cycle.
Fill Height Implications
Figure 9 shows the drying times for the vial, DCS needle-down, and LPC configuration for different fill heights. For the experiment, the drying cycles were adjusted so that the same product temperature was reached during primary drying for all configurations. The unfavorable aspect ratio of syringes leads to large fill heights for the same fill volume. As an example, for the same 1.2 mL volume the different fill heights are 7.9 mm for vials, 22.4 mm for the DCS needle-down process, and 26.2 mm for the LPC process. For the equivalent fill volume and equivalent steady-state product temperature, the drying times in the LPC and needle-up process are similar. For the vials this is approximately 1.5 times longer.
Drying time versus fill height for the investigated configurations: (♦) vials, (■) DCS in needle-down position + metal carrier, and (
) LPC process obtained from experiments with the BSA model formulation.
For the LPC process and the needle-down process, no pronounced difference in drying times were observed, as the fill height in the LPC is marginally higher and the heat transfer does not differ significantly. It can be assumed that the product resistance in the LPC is potentially smaller compared to the DCS due to the formation of larger ice crystals and the ice crystal architecture mentioned in the freezing considerations section. For the equivalent 7.9 mm fill height, for all configurations, the primary drying time is 483 ± 39 min for vials, 411 ± 2 min for the DCS in the needle-down position, and 323 ± 5 min when processing the LPC. For the LPC process, there is the possibility to fill multiple lyo cakes into the same DCS. If two lyo cakes (0.6 mL each) were filled into one DCS using the LPC process, the drying time could be reduced from 724 min to approximately half, which is 423 min (Figure 9). This is even faster than drying 1.2 mL in the vial (483 min).
Conclusions
Two new processes for freeze-drying pharmaceuticals for a DCS were introduced. Heat transfer and processability were assessed and compared to the current standard needle-up process. These studies enabled the possibility of designing a freeze-drying cycle that may be implemented on an industrial production scale. As expected, drying cycles cannot simply be transferred from the vial into a DCS format, but specific knowledge of heat and mass transfer is needed to develop the optimum freeze-drying cycle.
Processing DCSs in a needle-down configuration offers several advantages. The process is similar to the established industry standard syringe processing in a tray format and does not need complex turning, DCS removal (de-traying), and DCS replacement (re-traying) steps. It is also independent from the DCS design. Heat transfer and homogeneity across batches is excellent. Furthermore, DSCs are completely filled with drug product, leading to smaller syringe sizes as compared to needle-up processing. Needle-down processing, however, cannot overcome the challenges attributed to the slim geometry of the DCS itself. One issue is that the volume of the freeze-dryer cannot be efficiently used, as the DCS consists of the product and the diluent chamber. The other challenge is that for the same product volume, the product fill height for the DCS is much higher compared to a vial and the available sublimation area is smaller. This results in considerably longer drying times. These challenges could, in part, be addressed by designing a LPC-based process. The LPC process can more efficiently use lyophilizer space and long drying times caused by the large product fill height. Only the product and the middle plunger must fit into the metal tube, which therefore can be shorter than any DCS.
Heat transfer with the LPCs is excellent due to the very good thermal conductivity of the steel cylinder, which is in direct contact with the product and the shelf. Further design optimization such as a larger contact area of the LPC with the shelf may even further increase heat transfer effectiveness, thus leading to vertical ice crystal structure formation that can be associated with very low product resistances. The LPC promises to simplify DCS processing and may help to considerably increase availability of the DCS on the market.
On the basis of heat transfer coefficients in combination with knowledge of formulation-specific mass transfer resistance, one can design an optimized lyophilization cycle within a short time period. Patel and Pikal showed that manometric temperature measurements lead to acceptable results for drying in a syringe configuration immersed in an aluminum block when using formulations with low solid content (13). In order to use these principles for highly concentrated protein formulations, new mathematical or process analytical methods and technologies have to be established. They will be instrumental in reducing and simplifying cycle development for highly concentrated formulations in non-vial containers and to allow for widespread acceptance and use of DCSs in the pharmaceutical industry.
Conflict of Interest Declaration
The authors declare that they have no competing interests (commercial, monetary, or intellectual property–related). There were no external funding sources for the study.
Acknowledgements
We greatly thank Sebastian Schneider (Technical Transfer EU) and Frank Bamberg (Pre-Filled Syringe Device Development) for their valuable input and discussion. We also thank our colleagues from Lab Technologies & Robotics, Tom Kissling and Jörg Völkle, for their help during design and development of the LPCs, the closing system, as well as the aluminum block. Many thanks to the mechanical group members Christian Müller, Bruno Häring, and Emre Cankurt for the realization of these processes at laboratory scale. Special thanks go to Pierre Goldbach (Late-Stage Pharmaceutical & Processing Development) for providing mAb formulations. Additional thanks go to Katja Schulze (Solid State Properties Analytics, SM Analytical Drug Development) for measuring and providing the SEM images. Special thanks to Helen Brown (Late-Stage Pharmaceutical & Processing Development) for her detailed manuscript review and helpful suggestions and recommendations.
- © PDA, Inc. 2016
