Iterative model‐based experimental design for spherical agglomeration processes

Abstract

Spherical agglomeration (SA) is a process intensification strategy, which can reduce the number of unit operations in pharmaceutical manufacturing. SA merges drug substance crystallization with drug product wet granulation, reducing capital, and operating costs. However, SA is a highly nonlinear process, thus for its efficient operation model‐based design and control strategies are beneficial. These require the development of a high‐fidelity process model with appropriately estimated parameters. There are two major problems associated with the development of a high‐fidelity process models—(i) selection of the appropriate model corresponding to the underlying process mechanisms, and (ii) accurate estimation of the parameters. This work focuses on the identification of the best fitting model that correlates with experimental observations using cross‐validation experiments. Further, an iterative model‐based experimental design strategy is developed, which uses D‐optimal experimental design criterion to minimize the number of experiments necessary to obtain accurate parameter estimates.

Parameter Estimation and Estimability Analysis in Pharmaceutical Models with Uncertain Inputs

“…[7][8][9] In all the studies listed in Table 1, model inputs (independent variables) were assumed to be perfectly known during parameter estimation and all of the experimental uncertainty was assigned to the model outputs (dependent variables). This assumption enabled modelers to use either Least Squares (LS) 15,18,30 or Weighted Least Squares (WLS) estimation, 10,12,16,17,19,23,27,28,32,34 which is applied when there are multiple dependent variables with different levels of variability.…”

Parameter estimation and estimability analysis in pharmaceutical models with uncertain inputs

“…In all the studies listed in Table 1, model inputs (independent variables) were assumed to be perfectly known during parameter estimation and all of the experimental uncertainty was assigned to the model outputs (dependent variables). This assumption enabled modelers to use either least squares (LS) 12,15,27 or weighted least squares (WLS) estimation, 7,9,13,14,16,20,24,25,29,31 which is applied when there are multiple dependent variables with different levels of variability. Sometimes, however, uncertainties in independent variables can be large due to measurement errors in process inputs or other difficulties in achieving the desired experimental settings.…”

“…[32][33][34] In all the studies listed in Table 1, model inputs (independent variables) were assumed to be perfectly known during parameter estimation and all of the experimental uncertainty was assigned to the model outputs (dependent variables). This assumption enabled modelers to use either least squares (LS) 12,15,27 or weighted least…”

Spherulitic Growth Strategy for Agitation-Induced Formation of Spherical Amoxicillin Sodium Products

“…Three main crystallization approaches have been used to produce spherical crystals . (1) spherical agglomeration (SA): spherical crystals are obtained in the complex ternary solvent systems composed of good solvent, poor solvent, and bridging liquid in spherical agglomeration. , The bridging liquid acts as an interparticle binder to promote the agglomeration of crystals suspended in the system. , (2) Quasi-emulsion solvent diffusion (QESD): unstable emulsion droplets are formed by dispersing a good solvent solution into a poor solvent. Spherical crystals crystallize in the droplets driven by the diffusion of the poor solvent into the droplets and good solvent out of the droplets.…”

“…11 (1) spherical agglomeration (SA): 12 spherical crystals are obtained in the complex ternary solvent systems composed of good solvent, poor solvent, and bridging liquid in spherical agglomeration. 13,14 The bridging liquid acts as an interparticle binder to promote the agglomeration of crystals suspended in the system. 15,16 (2) Quasi-emulsion solvent diffusion (QESD): 17 unstable emulsion droplets are formed by dispersing a good solvent solution into a poor solvent.…”