This study focuses on defining an early-stage development strategy to be followed in a R&D environment to optimize the feeding and blending processes using the Continuous Dosing and Blending (CDB-0) unit from GEA. The CDB-0 is a small footprint and flexible equipment ideal for testing the suitability of a formulation for CDC, as well as for pilot-scale or scale up studies prior to transfer to production. This approach reduces materials consumption needs and minimizes the manufacturing line occupation time.
The process development phase is crucial and involves (i) optimizing the feeder setup and refill strategy, (ii) defining the feeding process parameters, and (iii) determining the blender speed and configuration to ensure a homogeneous blend. Moreover, with the aid of Process Analytical Technology (PAT) tools, preliminary process controls such as Residence Time distribution (RTD) and funnel plots, can be defined and later used in the CDC line. These controls are required to understand the process dynamics by predicting the effect of the propagation of a material disturbance on the final product as well as ensuring product traceability. All these critical development steps can be tested and defined in laboratory using a CDB-0 unit and will be discussed in this study.
To test the proposed development strategy, a model API formulation was chosen as a proof of concept. Feeding performance for the API was estimated using predictive models developed at Hovione, which correlate feeding processes with material flow properties. The use of these models significantly reduces material waste and time spent on trial-and-error experiments. The API material properties were characterized by conducting the following analytical tests: bulk density, tapped density, true density, charged density and shear cell (i.e., yield locus and wall yield locus). The predictive models were then employed to select the optimal feeding setup and initial feed factor (i.e., the mass of material dispensed per screw revolution). The selected set-up and estimated feeding parameters were successfully confirmed during the CDB-0 runs.
The CDB-0 unit, which includes LiW feeders and a linear blender, operates by transferring raw materials from polybottles to the feeders by gravity. A rotating wiper at the inlet of blender facilitates the product transfer. The raw materials are then mixed inside the 70 mm linear blender, whose shaft configuration can be tuned by choosing between transporting and mixing elements. A Near-Infrared (NIR) probe, namely SentroProbe DRLS from Sentronic, can be coupled to the outlet of blender to collect real-time NIR spectra. A waste divert unit is also used during initialization and shutdown for purging.
Blending process development was conducted using the CDB-0 unit and consisted in fluidization trials to determine optimal blender speed and validate the appropriate impeller configuration. At the end of each trial, the hold-up-mass (HUM) was determined by weighing the amount of blend remaining inside of the blender. The optimal blender speed was obtained as the speed that maximized the number of blade passes (i.e. the product of the impeller speed and mean residence time), which is related with mixing strain. Additionally, a Principal Component Analysis (PCA) using the NIR spectra collected during the fluidization trials was performed to confirm the uniformity of the blend.
Afterwards, a RTD study was carried out to calibrate the RTD parameters. In a CDC line, the RTD model can be implemented considering an ideal plug flow reactor (PFR) with two ideal continuous stirred tank reactors (CSTR) and can be used as a part of the control strategy to divert non-conform material. Therefore, the RTD parameters must be calibrated for both blenders, feed tube and feed fame of CDC line, resulting in a total of nine parameters. Using the CDB-0, it is possible to calibrate the RTD parameters for both blenders by applying disturbances in the composition and monitoring the outcome of the perturbation using a NIR probe. Three types of perturbations were tested - namely step changes, spikes and relay test - and compared in this work to assess the system’s response. Step changes consisted in changing the feeders setpoints, in this case, by increasing and decreasing the API feeder setpoint and adjusting the fillers accordingly to keep the same line feed rate. Spike or impulse perturbation was performed by dumping a controlled amount of API at the inlet of CDB-0 equipment. Relay test was done by performing a sequence of small perturbations with short durations. A Digital Twin webapp developed at Hovione was used to fit the experimental results obtained using the NIR probe with the RTD model prediction. The results showed that NIR probe successfully captured the performed perturbations and that the RTD parameters calibrated with these different methodologies were comparable.
The RTD model was further used to create the funnel plots, which represent the maximum predicted API concentration disturbance at the exit of each blender as a function of the API feeder deviation and duration.
In summary, the development of a robust and efficient CDC process for pharmaceutical applications relies on the optimization of feeding, blending and process controls. The use of CDB-0 unit in a laboratory setting has proven to be effective in refining key process parameters, demonstrating the power of predictive models, design of experimental runs, process controls, and PAT tools in early-stage process development. The results highlight the potential of the CDB-0 unit to support the development and scale up of CDC processes, providing valuable insights for future large-scale pharmaceutical manufacturing.