A modern strategy for digital real-time release testing in continuous tablet manufacturing

Abstract

The pharmaceutical industry is currently moving away from traditional approaches to quality control with off-line quality tests, limited in-line process monitoring and minimal control strategies towards more sophisticated methods. This transition addresses several critical aspects, including the reduction of ecological and economic footprints and ensuring the safety for patients and personnel. In that context, the initial step is the application of process monitoring tools, such as process analytical technology (PAT) and soft sensors, for real-time product quality assessment. This will enable real-time release testing (RTRT), which redefines conventional approaches by relying solely on the process data reported by equipment or collected from sensors to predict the product quality. However, the implementation of RTRT requires reliable material tracking algorithms, which align the process data with the product’s characteristics.

This study proposes a modern digital RTRT strategy that aligns process data collected from a state-of-the-art manufacturing line with a sophisticated process monitoring strategy for specific product quantities, i.e., single dosage units (tablets). To trace the material through the production line and align it to the collected process data, residence time distribution (RTD) models and material tracking algorithms were developed. The digital RTRT strategy was designed and demonstrated using the industrial manufacturing line ConsiGma^TM-25. The developed strategy makes full product quality information digitally available, including critical quality attributes (CQAs) and processing conditions experienced during the production. The obtained results were validated using traditionally established off-line methods.

Introduction

In traditional pharmaceutical manufacturing, the process steps are carried out in the batch mode and the quality is assured via in-process controls and end-product testing. Samples are drawn mainly at the end of each process step and analysed off-line in the quality control laboratory. After the approval, the batch is released to the next process step or, at the end of the line, for packaging and delivery to the patients. This is the quality by testing (QbT) approach, with no information acquired and used during processing. The implementation of the quality by design (QbD) principles offers more sophisticated options for quality and process control. Under QbD, pharmaceutical quality is assured by understanding and controlling the material attributes (MAs) and the process parameters (PPs) [32] to achieve critical quality attributes (CQAs) that are within specification. Especially, the process analytical technology (PAT) initiative (FDA, 2004) has led to a paradigm change. Data acquired by in-line monitoring devices allows to make decisions about the material quality or control actions in real-time. These are essential especially for continuous manufacturing, which requires enhanced control strategies rather than traditional off-line end product testing [23].

In-line data acquisition is an enabler of real-time release testing (RTRT), which is the ultimate option for quality control and product release. It evaluates and ensures the product quality based on the process data [15]. Real-time measurements and control of relevant in-process MAs and PPs are used to predict the corresponding finished product’s attributes. For common unit operations in solid dosage manufacturing, PAT tools have been successfully implemented [9], [29]. RTRT strategies have been developed for specific CQAs in commercial manufacturing, e.g., the tablet content and content uniformity obtained via near-infrared spectroscopy (NIRS) at-line measurements [13].
Since testing the dissolution performance is time-consuming and laborious, it would be a major step forward to implement RTRT for this CQA. Attempts have been made to predict dissolution via at-line NIR [26]. Dissolution profiles have been predicted via machine-learning approaches using NIR data, the compression force and the particle size distribution of the main excipient [12].

In a few cases, RTRT strategies have been developed for a process chain and several CQAs. For a wet-granulation tableting process in the batch mode, NIRS has been used for concentration-related CQAs and the design of experiments (DoE) for predicting the dissolution [19]. However, in continuous manufacturing lines, in addition to the PAT sensor data, a huge amount of in process data is available, which originates from univariate measurements and from equipment sensors, which can be used directly or as input for soft sensors. Matching the information to the material requires time-aligned data recording and knowing the residence of the material in the equipment. The probability of how long the material remains in a certain process step can be computed via residence time distribution (RTD) models. A substantial effort has been invested in the RTD modeling of pharmaceutical manufacturing equipment [7], [1]. One major application of RTD models in control concepts has been to coordinate the discharge actions [17], [2], [20]; Rehrl et al., 2018; [14]. The out-of-specification (OOS) material is detected either via content monitoring by NIRS or via soft sensors while the RTD model is used to trigger the discharge actions. Most studies, however, have involved simulations. The second application of RTD models has been to track the material on integrated manufacturing lines. Batch transitions in a feeding-blending system have been investigated [30]. Disturbances and raw material changes due to feeder refill have been tracked to the tablet in direct compaction (DC) line [7], [3]. Karttunen et al. [16] have determined the RTDs of three continuous tableting routes via wet granulation (WG), via dry granulation (DG) and via hot melt extrusion (HME) combined with pelletization to track the material and coordinate the discharge actions.

In this work, a novel method is introduced that connects process data generated with a state-of-the-art continuous manufacturing line with a specific portion of intermediate material or final product. A sophisticated process monitoring strategy was developed for the line, and the data acquired from the process equipment, PAT tools and soft sensors was collected in a data management platform. This data was then matched to individual material portions using the RTD models of the continuous unit operations and a material tracking algorithm for semi-continuous (mini-batch) unit operations. The proposed method made the complete history of process conditions and product quality attributes digitally available at the end of manufacturing cycle for a single dosage unit. Specifically, it was possible to assess which process conditions and quality attributes were experienced by each unit dose of the final product (i.e., tablets manufactured at specific timestamps) at various stages of production (e.g., during granulation, drying, tableting, etc.). The approach was implemented and demonstrated using the ConsigmaTM-25 tableting line.

Materials

The investigated pre-blend contained methyl 4-hydroxybenzoate (Sigma Aldrich, USA) as an active pharmaceutical ingredient (API) surrogate and VIVAPHARM® HPMC E5 (Demacsa, Mexico), Avicel® PH-101 (DuPont, Ireland) and GranuLac® 200 (Meggle, Germany) as excipients. Deionized water was used as the granulation liquid. External excipients, Primellose® (DFE Pharma, The Netherlands) and magnesium stearate (Merck kGaA, Germany), were used to create the final blend for tableting.

Selma Celikovic, Jakob Rehrl, Rúben Martins Fraga, Martin Steinberger, Johannes Khinast, Martin Horn, Stephan Sacher, A modern strategy for digital real-time release testing in continuous tablet manufacturing, European Journal of Pharmaceutics and Biopharmaceutics, 2025, 114700, ISSN 0939-6411, https://doi.org/10.1016/j.ejpb.2025.114700.