(219c) Analysis of Full Scale Fixed Bed Adsorption of Active Pharmaceutical Ingredient By Applying a Validated Adsorption Process Model Based On Independently Determined Key Adsorption Parameters

Adsorption is a common separation method for isolation and purification of active pharmaceutical ingredients (APIs) produced by bio-syntheses, but despite its widespread use, published literature describing such adsorption processes, their mathematical modeling, and the use of models for design, scale-up, optimization and troubleshooting is rare. Therefore a mathematical model (a simple non-structural and more complex structural approach was used) for a continuous fixed bed column adsorption process of glycopeptide antibiotic vancomycin was developed as a tool for analysis of the operation of a full scale adsorber. The model allows a prediction of breakthrough curves by an application of adsorption kinetics and equilibrium data, determined independently in batch adsorption experiments. Parameters characterizing the flow pattern of the liquid in the adsorption column, determined independently as well, using a tracer method. Such an approach differs from a widespread practice of fitting experimental breakthrough data and lumping several parameters into a dispersion coefficient.

The model was validated with laboratory scale experiments. Two macro-reticular polymeric adsorbents (Rohm&Haas XAD16N and XAD1600N), differing in particle and pore size distribution, were evaluated and their vancomycin dynamic adsorption capacity was determined for different superficial velocities. The key model parameters, such as those related to adsorption equilibrium and kinetics, were determined separately in batch experiments. A description of the liquid flow pattern in the column was obtained on the basis of tracer experiments, while different correlations have been utilized for the estimation of liquid phase mass transfer coefficients. Comparison of modeled and experimental break through curves shown that virtually all the result fell within the ±25% margin, typically even within ±15%. A certain amount of a disagreement between the model-predicted and the measured data may primarily be the result of the non-ideal wetting of adsorbent particles and various channeling effects, which render the adsorption kinetics, determined by batch experiments, less realistic; nonetheless, within a certain margin of error, our approach with data based on batch experiments was shown to be adequate. Furthermore, the parametric sensitivity of the structural adsorption/desorption kinetics model was evaluated to determine the impact of changes in key model parameter on the model results.

Laboratory scale experiments confirmed that the use of Amberlite XAD1600N is advantageous, which, based on the batch adsorption data, was expected. The dynamic adsorbent capacity (determined at C/C₀ = 0.1) was approximately 2.5−3.0 times higher (50−90 g kg⁻¹ wet resin) compared to Amberlite XAD16N (20−35 g kg⁻¹ wet resin). The predicted values of the dynamic capacity were somewhat underestimated, but typically within the +10/−35% range. As expected, lower liquid velocity (i.e.longer residence time) was favorable for the dynamic adsorption capacity.

Finally, the validated model was applied for an analysis of the existing full scale adsorption process and predicted break-through curves revealed 20-30% of unused adsorbent capacity and therefore the column can be further loaded with no loss of API. To increase the capacity of the adsorption process even more, adsorbent has be replaced either with fresh XAD16N or with XAD1600N.

The laboratory scale model verification and the application of the model to analyze the operating boundaries of the existing adsorption process demonstrated that our methodology in modeling the column adsorption process based on independently determined key model parameters, assures a reliable prediction of break through curves and dynamic adsorption capacities. The approach therefore can be applied as a valuable tool for fast evaluation of the effectiveness of various adsorbents, experimental design, designing full-scale continuous column operation, scaling-up, and analysis and troubleshooting of the existing full-scale adsorption processes.