(M0930-06-33) Wavelength Selection Method Based on Partial Least Square for Determination of Quetiapine in Tablets

Purpose: In the manufacture of drugs, control and monitoring of the concentration of the active pharmaceutical ingredient (API) must be carried out to ensure the quality and safety of the final product. Quetiapine is a drug used for the treatment of mental illnesses such as schizophrenia, bipolar disorder, and Parkinson's. Its manufacture must be subject to reliable and accurate analytical processes that allow quantifying its concentration at each stage of its fabrication. For this reason, there is a need to implement analytical technologies that provide fast and reliable answers on the concentration of this API. In the last decades, NIR spectroscopy together with chemometrics tools have been prime in the analytical development of pharmaceutical industries, due to the reduction of time and operating costs. But not all wavelength bands have the same predictive power for the primary parameter of interest. For this reason, the elimination of the least useful wavelength bands, often improve the quality and robustness of the model. In this study, the obtained spectroscopic data were used to compare the performance of several wavelength selection methods based on partial least squares (PLS) regression to quantify quetiapine in tablets.

Methods: The product used in this work was Quetiapine 400 mg tablets. A tablet is a pharmaceutical oral dosage form or solid unit dosage form. Tablets may be defined as the solid unit dosage form of medicament or medicaments with suitable excipients. It comprises a mixture of active substances and excipients, usually in powder form, pressed or compacted from a powder into a solid dose (see Figure 1). For this study, a set of calibration samples were prepared into the range of 0 to 150% of the target concentration of Quetiapine in the product. In addition, samples of five commercial batches of Quetiapine were obtained, three of these batches were included within the calibration set and two batches were included in an external validation set, used to evaluate the prediction capacity of the proposed models. The spectra of the samples were acquired using a NIR-FLEX-N-500 spectrophotometer (Brand: BUCCHI), by diffuse reflectance, 32 scans and 8 cm-1 resolution, with a spectral range of 4000-10,000 cm-1. Samples were analyzed in triplicate and 4 spectra were taken from each replicate. The spectral treatment software was NIRCAL 5.5V. In this study HPLC method was used as reference method to quantify the concentration of the samples used in the NIR calibration model. The spectra of the calibration samples were correlated with the independent variables that are the concentrations of the samples, through a regression by partial least squares (PLS). Several PLS regressions models were created, varying the wavelength selection methods, the type of pretreatment and the number of factors in order to find the equation that best models or predicts the spectral signal with respect to the concentration. The predictive capacity of the PLS model was evaluated using statistical parameters such as: Root mean square calibration (RMSE_CAL), prediction (RMSE_P) and validation (RMSE_val), R² of calibration (R²_Cal), prediction (R²_p) and validation (R²_Val), and percentage of recovery of the samples obtained with the PLS model Vs HPLC results.

Results: In Figure 2 shows the spectra samples and commercial batches of Quetiapine. The PLS methods tested include filter methods (loading weights, regression coefficient, and variable importance in the projection), wrapper methods (Genetic Algorithm-PLS, Backward Variable Elimination-PLS, The gray wolf optimizing algorithm-PLS). Additionally, three non-PLS-based methods Ridge regulation, Laso regulation, and elastic net regulation were used. Figure 2. Original Spectra of samples of quetiapine (Raw Data). Principal component analysis (PCA) was implemented to verify quetiapine concentrations correspond to spectroscopic data (see Figure 3). Figure 3. PCA for original Spectra of samples of quetiapine (Raw Data). All models create can be used in multivariate analysis for quantification Quetiapina in tablets, these models get the following results: RMSECAL: from 0.6000 to 0.6600; RMSEP: from 0.6500 to 0.6800; RMSEVAL: from 0.6500 to 0.7000; R2Cal: from 0.999 to 1.000; R2P: from 0.997 to 0.999 and R2Val: from 0.995 to .0999; the percentage of recovery was from 98.44 to 102.73. The wrapper methods and filter methods eliminated nearly 70% and 60% respectively, of the original spectral wavelengths.

Conclusion: The proposed methods of quantification of quetiapine, based on NIR spectroscopy of diffuse reflectance and PLS regression, gives accurate and precise results, which can be comparable with HPLC results, however, proposed method is less time consuming, since it reduces the sampling preparation step and it is environment-friendly, since it avoids the use of solvents. For this reason, the proposed method is an excellent alternative to be used in quality control analysis for this product.

References: [1] Zhong, L., Gao, L., Li, L., & Zang, H. (2020). Trends-process analytical technology in solid oral dosage manufacturing. European Journal of Pharmaceutics and Biopharmaceutics, 153, 187–199. https://doi.org/10.1016/j.ejpb.2020.06.008.
[2] Suttajit, S., Srisurapanont, M., Maneeton, N., & Maneeton, B. (2014). Quetiapine for acute bipolar depression: A systematic review and meta-analysis. Drug Design, Development and Therapy, 827. https://doi.org/10.2147/DDDT.S63779.
[3] Bakeev, K. A. (Ed.). (2010). Process analytical technology: Spectroscopic tools and implemented strategies for the chemical and pharmaceutical industries (2nd ed). Wiley.
[4] Mehmood, T., Liland, K. H., Snipen, L., & Sæbø, S. (2012). A review of variable selection methods in Partial Least Squares Regression. Chemometrics and Intelligent Laboratory Systems, 118, 62–69. https://doi.org/10.1016/j.chemolab.2012.07.010.
[5] Xu, H., Liu, Z., Cai, W., & Shao, X. (2009). A wavelength selection method based on randomization
test for near-infrared spectral analysis. Chemometrics and Intelligent Laboratory Systems, 97(2),
189–193. https://doi.org/10.1016/j.chemolab.2009.04.006.


Figure 1. Pharmaceutical Tablet.


Figure 2. Original Spectra of samples of quetiapine (Raw Data).


Figure 3. PCA for original Spectra of samples of quetiapine (Raw Data).