(35c) A Development of Combined PBM-RTD Technique to Simulate Continuous Wet Granulation Systems for Pharmaceutical Manufacturing Processes

The healthcare requirements of the developing countries in the world are expected to grow in keeping pace with their burgeoning populations. In efforts to meet the current and future projected supply-demand figures, there is a paradigm shift in pharmaceutical manufacturing towards producing the drug products (most commonly tablets) in a continuous operation as opposed to traditional batch methods of manufacturing in the industry. Granulation is a key unit operation that is responsible for the reduction in the segregation of the components, increase in the blend composition and, is responsible for producing particles with homogeneous particle size distributions. In an effort to predict the outlet compositions of the particles of a continuous granulator, the study of residence time distributions (RTDs) is crucial in order to predict the effects of variabilities in the inlet parameters such as feeder fluctuations and blend inhomogeneity on the outlet granule particle sizes and compositions. Population Balance Models (PBMs) are developed and extensively employed in modelling Wet Granulation systems in order to track the particle size distributions (PSDs) and porosities of the particles. Therefore, this work has initially studied two (2) different RTD models, compared them, and has applied them to a continuous granulation system. The RTD models are being combined with a PBM model in order to develop a model that quantifies the processes and attempts to describe the phenomena in as mechanistically as possible.

RTD models in granulation literature have been modelled in similar vein as the models for continuous liquid reactor systems [1]. The models account for plug flow volume fractions (conveying of material), number of continuous stirred tanks (back-mixing within mixing and kneading sections), and dead volume fraction (stagnant material within the unit operation). RTD models for hot-melt extrusion process (HMEs), which occur in similar equipment as granulation in polymeric media, have been given similar treatment in literature. The HME models too are described as reactor trains, and have even been modelled as cascading series of stirred tanks with recycle flows added from the subsequent reactor to the preceding tanks [4].

The present study has initially done a case study of two kinds of cascading reactor models for granulation operation. The first one, that has been published and studies for HME systems assumes that the ratio between the amount of material recycle to the previous tank to the amount of fresh feed coming into the current tank is constant for all the tanks in the cascade. This model holds true when the flow of material in either direction is uniform in any particular local mixing section, but the degree of mixing decreases gradually as the material is conveyed forward. The other approach that was developed in this work models the system such that the amount recycle to a previous stage is a fraction of the total inlet coming into the section. This accounts for the case when the degree of mixing is uniform in all the sections of the operation but the amount of material exchanged is different.

The two different models are being developed and studied in detail. They are applied to an experimental case of continuous granulation whereby, the model predictions and their closeness to experimental results are being compared.

An initial experimental study on a twin screw granulator has been performed using a blend of caffeine, micro-crystalline cellulose (MCC), lactose and polyvinylpyrrolidone (PVP) across a DOE consisting of variation in powder feed rate, liquid content and screw rotation speed. The RTD experiments were carried out on the DOE using the pulse method of tracer study using caffeine (one of the components of the formulation). The exit concentration was measured in real time using an inline near infrared probe (NIR). The NIR was calibrated with blends containing the aforementioned components in different combinations of proportions in order to measure real time changes in caffeine concentration due to addition of pulse tracer. The initial test results showed that the mean residence time (MRT) decreased with an increase of either the screw rotation speed (RPM) or increase in the feeder powder flow rate.

Further refined studies were carried out using the same components and experimental methods on the full DOE and the data is being fitted to the various RTD models. It is hoped that the RTD model would rely less on the similarity and approximation to the various TIS models developed for traditional liquid reactor systems and would be more inherently mechanistic and derived from basic principles applicable to granulation and pharmaceutical material systems.

In order to achieve the above mentioned goals, it is envisioned that the traditionally developed PBM equations [3] would be employed for estimating the particle size growth and porosity change as the primary powder particles are transformed into granules. The PBM equations take into account the binary aggregation of the smaller particles, breakage of larger particles and granules into smaller ones due to shear from the screws, liquid addition due to binder spraying, liquid evaporation due to hot air drying, and gas consolidation due to compression of the particles. The aggregation rate is generally considered to be proportional to the number of particles and the fractional area of the wetted region in the colliding particles. The breakage rate is primarily proportional to the screw speed and a power law factor to the particle size. These four elementary particulate rates are responsible for the changes in particle sizes and porosity.

Combining PBM with RTD models hasn’t been reported yet in pharmaceutical granulation modelling literature. However, a combined approach has been employed in modelling the rate processes of colloidal aggregation of beta-lactose globulin in calcium chloride solutions [2]. Therefore, it is envisioned that a similar approach would lead to a combined PBM/ RTD model for simulating continuous pharmaceutical granulation. It is hoped that the model developed would be as mechanistic as possible that would lead to reduction in dependence on experimental data and parameter fitting/ assumptions often involved in the current semi- empirical models.

Such models would then be employed in the simulation of a continuous pharmaceutical process line with integrated process control models. The integrated models would then be useful for designing quality manufacturing setups which would provide the advantages of lower equipment footprint, reduced manufacturing cost and reduced human intervention in the entire pharmaceutical manufacturing process thereby, leading to high- quality pharmaceutical products at affordable prices to the uninsured and under-insured consumers in the developed and developing countries.

References:

[1] A. Kumar et al., ‘Conceptual framework for model-based analysis of residence time distribution in twin-screw granulation’, European Journal of Pharmaceutical Sciences, vol. 71, pp. 25–34, Apr. 2015.

[2] N. Erabit, F. T. Ndoye, G. Alvarez, and D. Flick, ‘Coupling population balance model and residence time distribution for pilot-scale modelling of β-lactoglobulin aggregation process’, Journal of Food Engineering, vol. 177, pp. 31–41, May 2016.

[3] A. Chaudhury, ‘Mechanistic modeling, simulation and optimization of wet granulation processes’, Rutgers University - Graduate School - New Brunswick, 2015.

[4] J. Puaux, G. Bozga, and A. Ainser, ‘Residence time distribution in a corotating twin-screw extruder’, Chemical Engineering Science, vol. 55, no. 9, pp. 1641–1651, May 2000.