Simultaneous prediction of the API concentration and mass gain of film coated tablets using Near-Infrared and Raman spectroscopy and data fusion

Abstract

This study investigates the simultaneous prediction of active pharmaceutical ingredient (API) concentration and mass gain in film-coated tablets using Partial Least Squares (PLS) regression combined with three data fusion (DF) techniques: Low-Level (LLDF), Mid-Level (MLDF), and High-Level (HLDF). Near-Infrared (NIR) and Raman spectroscopy were utilized in both reflection and transmission modes, providing four types of spectral data per tablet. Transmission models proved more effective for API prediction by capturing data from the entire tablet, while reflection models excelled in assessing mass gain by focusing on the surface layer. Among the DF strategies, MLDF with Principal Component Analysis (PCA) offered the most significant improvements in predictive accuracy by filtering out irrelevant information. Variable selection methods further enhanced model performance by reducing the number of latent variables required. Overall, the integration of multiple spectral datasets and DF techniques resulted in models that gave predictions for evaluation samples with lower errors, demonstrating their potential to optimize quality control in pharmaceutical manufacturing.

Highlights

  • The study aims to evaluate the effectiveness of Partial Least Squares (PLS) models combined with different spectroscopic techniques (NIR and Raman spectroscopy) for predicting Active Pharmaceutical Ingredient (API) concentration and mass gain during film coating process. It explores low-, mid-, and high-level data fusion (DF) techniques to enhance prediction accuracy.
  • Transmission spectroscopy models showed superior performance in predicting API content, capturing comprehensive information about the tablet’s volume. Data fusion strategies, especially mid-level data fusion (MLDF) using Principal Component Analysis (PCA), significantly enhanced prediction accuracy. The best results were achieved by combining all four spectra types.
  • Reflection spectroscopy models were most effective for predicting mass gain, focusing on the tablet’s top layer where the film coating resides. Data fusion approaches consistently improved model performance, with MLDF models integrating two reflection spectra showing exceptional results. High-level data fusion models (HLDF), using Extreme Gradient Boosting (XGBoost), generally performed well but did not significantly outperform low-level data fusion (LLDF) models.

Introduction

Orally administered solid dosage forms are the simplest and most commonly used formulations in the pharmaceutical industry, providing numerous advantages, including relatively easy manufacturing, cost-effectiveness, and high patient compliance. Tablets are the most important and widely used members of this group (Augsburger and Hoag, 2016, Zaid, 2020). In recent decades, it also became common to utilize film coatings to enhance certain properties of the tablets (Sastry et al., 2000).

Film coating typically occurs in the final stages of pharmaceutical manufacturing. During this process, the surface of the products is covered with a continuous and uniform film layer (Kapoor, 2020). The purpose of film coating can include modifying the kinetics of drug release (e.g., achieving extended drug release), protecting the drug and product from physical and environmental factors (mechanical damage, light, air, moisture, gastric acid, etc.), improving patient compliance through the enhancement of tablet appearance (e.g., colored coating), masking unpleasant tastes and odors, and making swallowing easier (Felton, 2007).

Film coating methods offer great flexibility, allowing for a wide range of products to be coated, such as tablets, granules, powders, capsules, and nonpareils. Film coating is typically applied gradually to a moving mass of product, usually using a spray atomization technique. The success of the process can be attributed to various factors (Wang, 2012). Firstly, by applying a coating of only 2–3 % of the tablet core weight, the attributes of the product can be significantly altered. Moreover, film coating opens up opportunities for branding and identification, further establishing its significance in the pharmaceutical industry (Porter, 2021, Felton and Porter, 2013). Tablet film coating is a commonly used but critical process that provides various functions to tablets, thereby meeting different clinical needs. The evolution of dosage forms with film coatings is based on the development of coating technology, equipment, analytical techniques, and coating materials (Seo, 2020).

Two very important quality attributes of film-coated tablets are their active pharmaceutical ingredient (API) concentration and the mass gain that occurs during the film coating process. The cores must be adequately layered to ensure the product’s safety and efficacy (Barimani and Kleinebudde, 2018). Acquiring real-time information during the process helps to enhance the production process and to detect possible problems (Peng, 2015, Bakeev, 2010). This can be accomplished using process analytical technology (PAT) tools, as recommended by the Food and Drug Administration in a guidance in 2004 (Food and Drug Administration, 2004).

In the past decade, near-infrared (NIR) (Togashi, 2015, Tabasi, 2008, Römer, 2008) and Raman spectroscopy (Barimani and Kleinebudde, 2018, Nagy, 2019, Wabuyele, 2017, Müller, 2012) have become increasingly utilized for assessing critical attributes in pharmaceutical processing (Kandpal, 2017). These techniques offer rapid, non-destructive analysis without requiring sample preparation, and their growing use is driven by their ability to provide multivariate qualitative and quantitative data (De Beer, 2011, Ciza, 2019). Full-spectrum calibration methods like partial least-squares (PLS) regression have been widely validated for developing fast spectral screening techniques (Esposito Vinzi and Russolillo, 2013, Pirouz, 2006).

The collected data frequently encompasses extraneous variables that necessitate segregation from the primary variables. Algorithms for variable selection (VS) eradicate noisy spectral segments and redundant data to enhance predictive precision. VS improves the understanding and interpretability of these multivariate classification models (Alsberg et al., 1998). The interval partial least squares (iPLS) is a local regression method, which can give prediction models with improved precision by selecting the optimum interval for the spectral data. Genetic algorithm (GA) is also a suitable method for selecting wavelengths in PLS, enabling the calibration of mixtures with nearly identical spectra without compromising prediction capacity, utilizing spectrophotometric data (Xiaobo, 2010, Nørgaard, 2000, Ji, 2015). Extreme Gradient Boosting (XGBoost) is a scalable end-to-end tree boosting system, which is widely utilized by data scientists to achieve state-of-the-art results on numerous challenges in machine learning (Chen and Guestrin, 2016, Hayashi, 2023).

These advanced analytical platforms that are readily accessible offer extensive and diverse datasets linked to manufacturing processes that can be used for monitoring and predictive purposes (Casian, 2022, Casian, 2023, Casian, 2019, Azcarate, 2021). Data fusion (DF) refers to the process of combining multiple data sources, typically to increase the precision and accuracy of downstream predictive models. It has become a popular method in recent years due to the increased use of various spectroscopic analysis techniques (Hayes, 2023). Casian et al. gave an overview of opportunities of implementing DF in PAT (Casian, 2022). For example Zomer et al. employed chemometric methods for multivariate statistical process modeling to oversee the ongoing wet granulation tableting process of a pharmaceutical product currently under development (Zomer, 2018). A study by Casian et al. represents an investigation of NIR spectroscopy, Raman spectroscopy, colorimetry and image analysis methods which were tested and compared considering the ability to quantify the API concentration and to detect production errors (Casian, 2019).

This study explores the use of NIR and Raman spectroscopy to simultaneously predict API concentration and mass gain in film-coated tablets. To accommodate different dosages with a single model, it is crucial that the model remains robust to variations in both coating thickness and API concentration. Both spectroscopic techniques were employed in reflection and transmission modes, resulting in four measurements per tablet. Data fusion techniques, which have not been thoroughly tested before for film-coated tablets, were applied and compared to identify the most reliable modeling approach. Exploring various combinations of data fusion may lead to the development of a more robust model that is applicable across different dosages and unaffected by variations in API concentration. This is particularly important for products manufactured in multiple dosages, where the model must predict a quality attribute other than API content, despite changes in the API signal.

Download the full article as PDF here Simultaneous prediction of the API concentration and mass gain of film coated tablets using Near-Infrared and Raman spectroscopy and data fusion

or read it here

Materials

In this study, anhydrous caffeine (BASF, Ludwigshafen, Germany) was used as model API. The formulated tablets contained three more excipients: microcrystalline cellulose (MCC) (Vivapur grade 200, JRS Pharma GmbH, Rosenberg, Germany) as filler, croscarmellose sodium (Ac-Di-Sol®, FMC BioPolymer, DuPont, Leiden, Netherlands) as disintegrant promoting dissolution and magnesium stearate (MgSt) (grade S, Faci S.P.A., Carasco, Italy) as lubricant to protect the tablet press from mechanical degradation due to repeated usage.

The material Opadry® QX (Colorcon, Budapest, Hungary) was used for the film coating, which is a commercially available preformulated yellow colored polyvinyl alcohol (PVA) based coating material. It is used to enhance the stability, appearance, and patient acceptability of tablets and capsules.

Film coating

Film coating was carried out in a Glatt GB2 L50-10026 pan coating machine (Glatt GmbH, Binzen, Germany) equipped with a perforated drum and winglike baffles were installed on it to facilitate the thorough mixing of the tablet bed and ensure uniform coating.

Bence Szabó-Szőcs, Máté Ficzere, Orsolya Péterfi, Dorián László Galata, Simultaneous prediction of the API concentration and mass gain of film coated tablets using Near-Infrared and Raman spectroscopy and data fusion,
International Journal of Pharmaceutics, 2024, 124957, ISSN 0378-5173, https://doi.org/10.1016/j.ijpharm.2024.124957.


Are you looking for excipients in commercial quantities?

Order your commercial quantities of excipients with Pharma Excipients
Order your commercial quantities of excipients with Pharma Excipients
You might also like