Using scattering and absorption spectra as MCR-hard model constraints for diffuse reflectance measurements of tablets

Analytical comparison between batch and continuous direct compression processes for pharmaceutical manufacturing using an innovative UV-Vis reflectance method and chemometrics

Advancements in industrial technologies and the application of quality by design (QbD) guidelines are shifting the attention of manufacturers towards innovative production techniques. In the pharmaceutical field, there is a significant focus on the implementation of continuous processes, in which the production stages are carried out continuously, without the need to interrupt the process and store the production intermediates, as in traditional batch production. Such innovative production techniques also require the development of proper analytical methods able to analyze the products in-line, while still being processed. The present study aims to compare a traditional batch manufacturing process with an alternative continuous one. To this end, a real pharmaceutical formulation was used, substituting the active pharmaceutical ingredient (API) with riboflavin, at the concentration of 2 %w/w. Moreover, a direct and non-destructive analytical method based on UV-Vis reflectance spectroscopy was applied for the quantification of riboflavin in the final tablets, and compared with a traditional absorbance analysis. Good results were obtained in the comparison of both the two manufacturing processes and the two analytical methods, with R2 higher than 0.9 for all the calculated calibration models and predicted riboflavin concentrations that never significantly overcame the 15 % limits recommended by the pharmacopeia. The continuous production method demonstrated to be as reliable as the batch one, allowing to save time and money in the production step. Moreover, UV-Vis reflectance was proved to be an interesting alternative to absorption spectroscopy, which, with the proper technology, could be implemented for in-line process control.

Machine Learning-Enabled NIR Spectroscopy. Part 3: Hyperparameter by Design (HyD) Based ANN-MLP Optimization, Model Generalizability, and Model Transferability

Data variations, library changes, and poorly tuned hyperparameters can cause failures in data-driven modelling. In such scenarios, model drift, a gradual shift in model performance, can lead to inaccurate predictions. Monitoring and mitigating drift are vital to maintain model effectiveness. USFDA and ICH regulate pharmaceutical variation with scientific risk-based approaches. In this study, the hyperparameter optimization for the Artificial Neural Network Multilayer Perceptron (ANN-MLP) was investigated using open-source data. The design of experiments (DoE) approach in combination with target drift prediction and statistical process control (SPC) was employed to achieve this objective. First, pre-screening and optimization DoEs were conducted on lab-scale data, serving as internal validation data, to identify the design space and control space. The regression performance metrics were carefully monitored to ensure the right set of hyperparameters was selected, optimizing the modelling time and storage requirements. Before extending the analysis to external validation data, a drift analysis on the target variable was performed. This aimed to determine if the external data fell within the studied range or required retraining of the model. Although a drift was observed, the external data remained well within the range of the internal validation data. Subsequently, trend analysis and process monitoring for the mean absolute error of the active content were conducted. The combined use of DoE, drift analysis, and SPC enabled trend analysis, ensuring that both current and external validation data met acceptance criteria. Out-of-specification and process control limits were determined, providing valuable insights into the model's performance and overall reliability. This comprehensive approach allowed for robust hyperparameter optimization and effective management of model lifecycle, crucial in achieving accurate and dependable predictions in various real-world applications.

Diffuse reflection mid-infrared spectroscopy of neat powders by using a wire mesh "Button" sample holder

A new approach for neat powder mid-infrared spectroscopy measurements is described. Powder thin layers are created and supported by wire meshes attached to stainless steel backings, or buttons. Stainless steel buttons are durable, resistant to corrosion, and are easily cleaned and reused. To obtain spectra, buttons containing neat samples are placed at the infrared beam focus of a praying mantis diffuse reflection optical system. Wire diameter and mesh opening size dictate the powder particle size range that can be accommodated and maximum sample layer thicknesses, which determine radiation penetration path lengths. For powder layer thicknesses of several microns, infrared spectra that are free of artifacts can be obtained without dilution in a non-absorbing matrix. Nanogram detection limits are demonstrated for a benzoic acid coating prepared by evaporating an aqueous solution from a button surface.

Spatially Resolved Spectral Powder Analysis: Experiments and Modeling

Understanding the behavior of light in granular media is necessary for determining the sample size, shape, and weight when probing using fiber optic setups. This is required for a correct estimate of the active pharmaceutical ingredient content in a pharmaceutical blend via near-infrared spectroscopy. Several strategies to describe the behavior of light in granular and turbid media exist. A common approach is the Monte-Carlo simulation of individual photons and their description using mean free path lengths for scattering and absorption. In this work, we chose a complementary method by approximating these parameters via real physical counterparts, i.e., the particle size, shape, and density and the resulting chord lengths. Additionally, the wavelength dependence of refractive indices is incorporated. The obtained results were compared with those obtained in an experimental setup that included the SAM-Spec Felin probe head by Indatech for detecting spatially resolved spectra of samples. Our method facilitates the interpretation of the acquired experimental results by contrasting the optical response, the physical particle attributes, and the simulation results.

Process design and control of a twin screw hot melt extrusion for continuous pharmaceutical tamper-resistant tablet production

Hot melt extrusion (HME) is a well-known process within the plastic and food industries that has been utilized for the past several decades and is increasingly accepted by the pharmaceutical industry for continuous manufacturing. For tamper-resistant formulations of e.g. opioids, HME is the most efficient production technique. The focus of this study is thus to evaluate the manufacturability of the HME process for tamper-resistant formulations. Parameters such as the specific mechanical energy (SME), as well as the melt pressure and its standard deviation, are important and will be discussed in this study. In the first step, the existing process data are analyzed by means of multivariate data analysis. Key critical process parameters such as feed rate, screw speed, and the concentration of the API in the polymers are identified, and critical quality parameters of the tablet are defined. In the second step, a relationship between the critical material, product and process quality attributes are established by means of Design of Experiments (DoEs). The resulting SME and the temperature at the die are essential data points needed to indirectly qualify the degradation of the API, which should be minimal. NIR-spectroscopy is used to monitor the material during the extrusion process. In contrast to most applications in which the probe is directly integrated into the die, the optical sensor is integrated into the cooling line of the strands. This saves costs in the probe design and maintenance and increases the robustness of the chemometric models. Finally, a process measurement system is installed to monitor and control all of the critical attributes in real-time by means of first principles, DoE models, soft sensor models, and spectroscopic information. Overall, the process is very robust as long as the screw speed is kept low.

Rapid characterization of tanshinone extract powder by near infrared spectroscopy

Chemical and physical quality attributes of herbal extract powders play an important role in the research and development of Chinese medicine preparations. The active pharmaceutical ingredients have a direct impact on the herbal extract's efficacy, while the physical properties of raw material affect the pharmaceutical manufacturing process and the final products' quality. In this study, tanshinone extract powders from Salvia miltiorrhiza which are widely used for the treatment of cardiovascular diseases in the clinic are taken as the research object. Both the chemical information and physical information of tanshinone extract powders are analyzed by near infrared (NIR) spectroscopy. The partial least squares (PLS) and least square support vector machine (LS-SVM) models are investigated to build the relationship between NIR spectra and reference values. PLS models performed well for the content of crytotanshinone, tanshinone IIA, the moisture, and average median particle size, while, for specific surface area and tapped density, the LS-SVM models performed better than the PLS models. Results demonstrated NIR to be a valid and fast process analytical technology tool to simultaneously determine multiple quality attributes of herbal extract powders and indicated that there existed some nonlinear relationship between NIR spectra and physical quality attributes.

Effect of experimental design on the prediction performance of calibration models based on near-infrared spectroscopy for pharmaceutical applications

Near-infrared spectroscopy (NIRS) is a valuable tool in the pharmaceutical industry, presenting opportunities for online analyses to achieve real-time assessment of intermediates and finished dosage forms. The purpose of this work was to investigate the effect of experimental designs on prediction performance of quantitative models based on NIRS using a five-component formulation as a model system. The following experimental designs were evaluated: five-level, full factorial (5-L FF); three-level, full factorial (3-L FF); central composite; I-optimal; and D-optimal. The factors for all designs were acetaminophen content and the ratio of microcrystalline cellulose to lactose monohydrate. Other constituents included croscarmellose sodium and magnesium stearate (content remained constant). Partial least squares-based models were generated using data from individual experimental designs that related acetaminophen content to spectral data. The effect of each experimental design was evaluated by determining the statistical significance of the difference in bias and standard error of the prediction for that model's prediction performance. The calibration model derived from the I-optimal design had similar prediction performance as did the model derived from the 5-L FF design, despite containing 16 fewer design points. It also outperformed all other models estimated from designs with similar or fewer numbers of samples. This suggested that experimental-design selection for calibration-model development is critical, and optimum performance can be achieved with efficient experimental designs (i.e., optimal designs).

Penetration of light into multiple scattering media: model calculations and reflectance experiments. Part I: the axial transfer

The article presents two general equations of radiation penetration into layers of diffuse reflectors. One of the equations describes the depth origins of reflection, the other the depth profiles of absorption. The equations are evaluated within the theory of radiative transfer applying various degrees of analytical approximations and Monte Carlo simulations. The data are presented for different scattering and absorption coefficients, arbitrary layer thicknesses, collimated and diffused irradiation, and anisotropic forward scattering. The calculated mean depths of reflection are always lower than the mean depths of absorption. For nearly non-absorbing layers, the mean depths of absorption are about one third of the physical layer thickness. In contrast, penetration saturates for strong absorbers at very low depth levels. From the simulated data, methods are derived for the determination of the penetration depth from reflectance and transmittance data of thin layers or from radially diffused reflectance profiles upon spot irradiation. The methods are experimentally verified for a series of metal oxide powders with particle sizes ranging from much smaller to much larger than the wavelength of irradiation and for microcrystalline cellulose stained with different concentrations of an organic dye.