High throughput in-line content uniformity measurement of tablets based on real-time UV imaging

Abstract

This paper presents a precursor of a novel, high-throughput, in-line system, which utilizes ultraviolet (UV) imaging in order to predict the active pharmaceutical ingredient (API) content of tablets in real-time, non-destructive manner. Pimobendan, cardiovascular drug used in veterinary medicine was chosen as a fluorescent model API. Two experiments were carried out using different measurement setups, where the tablets were moving at different speeds. The blue colour components of the images were used to predict the pimobendan content of the tablets, and as a reference method, traditional UV spectroscopy was used to measure the API content of the dissolved tablets. In the case of the first, slower experiment (with a conveyor belt speed of 83 mm/s), a second order polynomial was fitted to the calibration tablets containing a nominal dose of 1.25, 5 and 10 mg of pimobendan and it was used to predict the API content. The RMSEP obtained was 0.428 mg for the validation tablets with a relative error as low as 7 % for the target level. For the second, faster experiment (1000 mm/s) the same polynomial was used to predict the pimobendan concentration of a different set of tablets, achieving a relative error of 2.03 %. Finally, the throughput of both systems was calculated to assess their applicability to meet the requirements of a pharmaceutical manufacturing line. The first system could inspect up to 93 375 tablets per hour, while the second was able to process up to 360 000 tablets in an hour, making it suitable for industrial application. By using these developed systems, the API content of all produced tablets could be determined non-destructively, which can greatly improve patient safety.

Introduction

Quality-by-testing (QbT) is the currently widespread approach in the pharmaceutical industry as a quality control strategy during manufacturing. During this, a small set of samples undergo slow and destructive off-line testing (Nagy et al., 2022) and if the evaluated parameters (for example their content uniformity) do not meet the desired specifications, the entire batch could be discarded, causing severe financial loss (Rossi, 2022). The process analytical technology (PAT) concept was introduced in order to better understand the pharmaceutical production process and to combat this phenomenon. The initiative is supported by the authorities (Food and Drug Administration, 2004) and over the years has attracted the significant interest of the pharmaceutical manufacturers. These tools can comprehensibly understand the products and processes, and they are also able to rapidly analyze and control any deviations that may occur during production (Casian et al., 2022).

Content uniformity is considered as a high-risk critical quality attribute, and recalls have been initiated regarding it (Food and Drug Administration, 2024). Therefore, several faster methods have been developed to replace the traditional off-line, destructive measurements. Spectroscopic methods, such as near infrared (NIR) and Raman spectroscopy have emerged as PAT tools for the measurement of API content in tablets. NIRS has already been utilized to assess the API content and hardness of tablets (Kandpal et al., 2017) and to predict the API concentration of tablets produced at different levels of compression force (De Man et al., 2023). Transmission Raman spectroscopy has also been used for content uniformity measurement coupled with NIRS (Wang et al., 2021) or as a standalone method (Belay et al., 2021). However, despite developments in the speed of Raman spectroscopy, new models could still only inspect up to 100 tablets per hour in transmission mode (Agilent, 2024). Systems using NIRS have been reported with higher speeds up 120 000 tablets/hour (Brouckaert et al., 2022), but this value is still below the productivity of pharmaceutical tablet presses.

Machine vision is an innovative analytical tool in pharmaceutical manufacturing, thanks to its numerous advantages, such as its low investment cost and fast data acquisition. By using colour cameras, the content of various coloured components can be monitored throughout the pharmaceutical manufacturing process (Galata et al., 2021b, Péterfi et al., 2024). Machine vision has already been used to accurately monitor the content of the yellow-coloured API riboflavin during continuous twin-screw wet granulation (Ficzere et al., 2021) and continuous blending (Galata et al., 2021a). Durão et al. found that powder mixtures containing coloured multivitamins could be monitored accurately inside the feed frame of a tablet press as well (Durão et al., 2017). Machine vision could also be used for the investigation of tablets containing coloured components (Wagner et al., 1999). These systems all applied visible illumination, which however cannot be used in the case of colourless APIs for direct concentration measurement.

UV imaging utilizes ultraviolet illumination, creating a non-destructive and fast tablet inspection method. One type of it is multispectral UV imaging, which was successfully used by Wu et al. to visualize glibenclamide in tablets (Wu et al., 2014). This same approach was used for the visualization of pellets in tablets (Novikova et al., 2016) and for the determination of tablet physical properties (Klukkert et al., 2016). However, due to the consecutive process of data acquisition at different wavelengths, in all of these cases the acquisition time was 18 s per multispectral image. This greatly limits the number of samples, making multispectral UV imaging not suitable for an in-line, real-time system.

By exposing a pharmaceutical dosage form to only a single wavelength of UV light, its components can change into different colours thanks to differences in their fluorescence and absorption. This phenomenon if combined with an RGB camera, could be used for API content prediction with high speed and resolution (Mészáros et al., 2022). Mészáros et al. (Mészáros et al., 2020) used single wavelength UV/VIS imaging to determine the drug content, the crushing strength and the dissolution profile of tablets containing meloxicam, a yellow model drug. They used an acquisition time of 1.2 ms, and were able to predict the API content of tablets with RGB and CIELAB colour space-based algorithms with a relative error of 4.9 %. In their previous work, the authors (Ficzere et al., 2024) utilized single wavelength UV imaging to determine the content of two colourless APIs in their respective tablets. They used artificial neural networks to predict the amlodipine and valsartan content of the tablets, and have achieved relative errors of 4.41 % and 3.98 %, respectively. However, there have been no cases reported in the literature, where UV imaging was used an in-line, high-speed tool for the API content determination of tablets in real-time.

The authors have demonstrated in their prior publications that the API content can be determined in tablets through an offline approach. The goal of this study was the development of a precursor of a first of its kind, high-throughput, real-time system which utilizes UV imaging in order to determine the API content of colourless tablets accurately. The authors aimed to create a system that could be utilized for efficient image acquisition in industrial settings, incorporating a concept of a rapid evaluation algorithm. We intend to use pimobendan, as a fluorescent API. The applied excipient matrix also models commercially available formulations. Finally, we also intend to check if the developed system is capable of matching the productivity of the already existing pharmaceutical manufacturing lines.

Materials

Pimobendan, a cardiovascular drug used in veterinary medicine was chosen as a model API and was obtained from Sigma-Aldrich (St. Louis, Missouri, USA). Lactose monohydrate (EMPROVE® ESSENTIAL) and magnesium stearate were acquired from Merck (Budapest, Hungary). Microcrystalline cellulose (MCC, Vivapur® 200) was purchased from JRS Pharma (Rosenberg, Germany), while cross-linked sodium carboxymethyl cellulose (Ac-Di-Sol® SD-711) was obtained from IFF Inc. (New York, USA).

Máté Ficzere, Anna Diószegi, Attila Farkas, Béla Weiss, Lilla Alexandra Mészáros, Zsombor Kristóf Nagy, High throughput in-line content uniformity measurement of tablets based on real-time UV imaging, International Journal of Pharmaceutics, 2024, 125066, ISSN 0378-5173, https://doi.org/10.1016/j.ijpharm.2024.125066.