Real-time release testing of in vitro dissolution and blend uniformity in a continuous powder blending process by NIR spectroscopy and machine vision

Continuous manufacturing is gaining increasing interest in the pharmaceutical industry, also requiring real-time and non-destructive quality monitoring. Multiple studies have already addressed the possibility of surrogate in vitro dissolution testing, but the utilization has rarely been demonstrated in real-time. Therefore, in this work, the in-line applicability of an artificial intelligence-based dissolution surrogate model is developed the first time. NIR spectroscopy-based partial least squares regression and artificial neural networks were developed and tested in-line and at-line to assess the blend uniformity and dissolution of encapsulated acetylsalicylic acid (ASA) – microcrystalline cellulose (MCC) powder blends in a continuous blending process. The studied blend is related to a previously published end-to-end manufacturing line, where the varying size of the ASA crystals obtained from a continuous crystallization significantly affected the dissolution of the final product. The in-line monitoring was suitable for detecting the variations in the ASA content and dissolution caused by the feeding of ASA with different particle sizes, and the at-line predictions agreed well with the measured validation dissolution curves (f₂ = 80.5). The results were further validated using machine vision-based particle size analysis. Consequently, this work could contribute to the advancement of RTRT in continuous end-to-end processes.

Introduction

In recent years, an extensive transformation of pharmaceutical manufacturing has been initiated, aiming for more flexible and efficient research and development, production, and quality control. From the technological perspective, it manifests in introducing continuous manufacturing steps [1] or even integrated, end-to-end continuous production [2]. From the quality point of view, the risk- and knowledge-based manufacturing initiated by the Quality by Design (QbD) concept, the real-time monitoring by Process Analytical Technology (PAT) Food and Drug Administration [3], and the potential of real-time release testing (RTRT) Agency [4] gain increasing interest. Although these concepts also apply to batch processes, they became indispensable for continuous production to truly benefit from operational flexibility [5], [6]. The growing interest in continuous manufacturing and advanced quality control by QbD, PAT, and RTRT is well demonstrated by the vast number of publications from both the academia and industry, which are extensively reviewed in [6], [7].

Powder blending is a crucial unit operation of downstream processing of final solid dosage forms, e.g., tablets and capsules, to obtain homogeneous mixtures of components in the proper constitution. While in the case of batch blending, it is generally sufficient to analyze the endpoint of the mixing, the continuous operation entails the need to monitor the quality of the powder stream in-line [7]. Several studies have already dealt with the analysis of the critical quality attributes (CQAs) of continuous blending, i.e., concentration, blend uniformity (BU), and the consequent content uniformity (CU) of the final product (usually tablets) [6], [8]. For instance, the effect of different measurement setups of in-line NIR spectroscopic measurement on the BU results has been demonstrated [9], as well as the relationship of the BU and the tablet CU has been studied [9], [10], [11]. NIR spectroscopy has also been applied in the development of residence time distribution (RTD) model of the blending process [12], and feedback control of blend concentration has been achieved by applying in-line NIR [13], Raman spectroscopy [14], as well machine vision [15].

The particle size of the blends can also account for a critical material attribute (CMA), such as in further feedability [16] or in the occurrence of segregation [17], which can affect the CU or even the in vitro dissolution of the final products. In [18], the authors have demonstrated that continuous blending decreases the risk of segregation and increases the blend homogeneity compared to the batch operation. Several studies have already demonstrated the applicability of NIR spectroscopy to quantify particle size differences in powder samples, which is reviewed in detail e.g., in [19], [20], [21]. Several theories have attempted to describe the relationship between the NIR spectra and the particle size, but it still remained a complex field of study. It has been also shown that common preprocessing methods cannot entirely eliminate this effect [20], [21]. Barajas et al. concluded that NIR spectroscopy could be used to detect post-blending segregation of flowing powder, not only due to the concentration change during the segregation but thanks to the particle size information carried in the NIR spectra [22].

A direct approach for detecting in-process particle size variations is the application of particle size measurement tools. Several measurement techniques exist for quantifying particle size distribution (PSD) in-line. Apart from the commonly used laser diffraction, there are a few alternatives within the field of machine vision or imaging technologies. These techniques utilize advanced image processing or artificial intelligence [23], [24] and are gaining increasing interest in the pharmaceutical industry [23] due to their cost-efficiency, speed, non-invasiveness, and non-destructive nature [25]. Imaging techniques have the capability to analyze challenging samples that have a very diverse range or limited representation of particle sizes. They can serve as at-line, non-destructive techniques, similar to a microscopic particle size measurement. Nevertheless, when compared to a microscopic measurement, the developed system offers high-resolution images, easy accessibility in a cost-effective manner. The potency of novel technologies such as photometric stereo imaging, Eyecon®, and spatial filtering velocimetry were also compared, and the dissimilarities were explained based on their working principles [24]. Recently, Ficzere et al. utilized an AI-based machine vision system to evaluate the PSD of acetylsalicylic acid and calcium hydrogen phosphate, which involved the detection of particles exceeding 100 µm in size [26]. Madarász et al. also employed AI-based imaging techniques with an endoscope to perform in-line determination of particle size for sodium-chloride crystals ranging from 200-1000 µm [27].

In vitro dissolution is one of the most critical CQAs of the solid dosage forms, which is used to show bioequivalence as well as consistent quality by characterizing the inter- and intra-batch variability. The in vitro dissolution curves are affected by several CMAs and critical process parameters (CPPs) from different processing units and, therefore, can serve as a fingerprint of the whole manufacturing [28]. Several studies have already addressed the challenge of developing RTRT alternatives of dissolution, both for immediate- and extended-release products, although the immediate-release is more prevalent in regulatory submissions [28], [29]. In these works, the CMAs and CPPs are first identified, then these factors are monitored by appropriate PAT measurement or process data collection. For example, NIR [30], [31], [32], [33] and Raman spectroscopy [30], [34] have been already successfully applied to dissolution surrogate modeling. Finally, mathematical modeling is carried out to establish the connection between the collected data and the in vitro dissolution curves, for which machine learning and artificial intelligence are gaining increasing interest [28], but mechanistic/white-box modeling is also a possibility [35], [36]. Despite the numerous successful proof-of-concept studies, the real-time utilization of surrogate dissolution modeling has rarely been demonstrated. In [37], real-time NIR measurements performed within a pan coater were used to predict the dissolution of the final product, i.e., a controlled-release tablet containing a functional coat. Su et al. [38] demonstrated the in-line application of a model-predictive dissolution prediction approach in an end-to-end extrusion-molding-coating manufacturing line. In their work, a mechanistic model was developed to account for the swelling and eroding of the immediate-release tablet, which was applied to monitor the variation of the dissolution caused by the changing mass fraction of the active pharmaceutical ingredient (API) measured by in-line NIR.

This work aims to elaborate on the in-line applicability of in vitro dissolution surrogate modeling by PAT tools. NIR spectroscopy-based partial least squares regression (PLS) and artificial neural network (ANN) models were developed and applied both in-line and at-line to simultaneously assess the blend uniformity and in vitro dissolution of encapsulated powder blends produced in a continuous blending process. To the best of the authors’ knowledge, this is the first published in-line, real-time dissolution monitoring based on artificial neural networks (ANNs). The studied model system, i.e., acetylsalicylic acid (ASA) − microcrystalline cellulose (MCC) blend, is the same as used for the development of an end-to-end continuous manufacturing line in our previous works [12], [36], [39], [40], where the ASA particle size was found to be affected by the continuous crystallization conditions of the flow reaction [40], and influencing the in vitro dissolution of the final product. As the particle size information in the NIR spectra can be carried in a relatively weak signal, it was also critically important to validate the particle size determination using an orthogonal particle size measurement technique. For this purpose, the application of an at-line machine vision technique was selected, which can provide a cost-effective, non-destructive, and high-resolution approach for particle size analysis. The comparison of the two techniques highlighted the validity of ANN model, as well as the potential of the machine vision approach for at-line analysis, which could be further developed into in-line PAT technique with suitable sampling interface and data analysis technique. Consequently, this work could contribute to the advancement of RTRT in continuous end-to-end processes.

Download the full article as PDF here: Real-time release testing of in vitro dissolution and blend uniformity in a continuous powder blending process by NIR spectroscopy and machine vision

or read it here

Materials

Acetylsalicylic acid (ASA) was used as a model drug, purchased from Sigma Aldrich (Germany), and microcrystalline cellulose (MCC, Vivapur® 200) was used as an excipient (JRS Pharma, USA). For dissolution testing purposes, concentrated hydrochloric acid solution was purchased from Merck Ltd. (Germany).

Lilla Alexandra Mészáros, Martin Gyürkés, Emese Varga, Kornélia Tacsi, Barbara Honti, Enikő Borbás, Attila Farkas, Zsombor Kristóf Nagy, Brigitta Nagy, Real-time release testing of in vitro dissolution and blend uniformity in a continuous powder blending process by NIR spectroscopy and machine vision, European Journal of Pharmaceutics and Biopharmaceutics, Volume 201, 2024, 114368, ISSN 0939-6411, https://doi.org/10.1016/j.ejpb.2024.114368.