Application of NIR handheld transmission spectroscopy and chemometrics to assess the quality of locally produced antimalarial medicines in the Democratic Republic of Congo

• Main Authors: P.H. Ciza, P-Y. Sacre, M.R. Kanyonyo, C.T. Waffo, M.A. Borive, L. Coïc, J.K. Mbinze, Ph. Hubert, E. Ziemons, R.D. Marini
• Format: Article
• Language: English
• Published: Elsevier 2021-08-01
• Series: Talanta Open
• Online Access: http://www.sciencedirect.com/science/article/pii/S2666831920300254

In recent decades, more than 15% of the antimalarials marketed in low- and middle-income countries have been of poor quality, in which quinoline derivatives and quinine-based formulations account for 21%.Near infrared spectroscopy (NIR) was chosen for its fast and inexpensive test properties as well as the ability of using handheld devices to monitor drugs directly on the field. Data driven - soft independent modeling of class analogy (DD-SIMCA) and partial least squares (PLS) regression models were developed for qualitative and quantitative purpose, respectively. The specificity and selectivity tests were performed using the DD-SIMCA models on the placebo, the quinidine and cinchonine standard samples. Then, PLS regression methods have been developed and validated for the quality control of quinine dosage forms manufactured by a major local manufacturer in the Democratic Republic of Congo (DRC).Calibration and validation samples were prepared by dissolving quinine sulfate / quinine hydrochloride in the presence of excipients in HCl 1 M. The opportunity to work with dissolved quinine with a cheap and readily available medium in low- and middle-income countries allowed analysis of different pharmaceutical forms (oral drops, solutions for injection and tablets) with the same regression model. DD-SIMCA models have demonstrated, for both equipment, perfect authentication of quinine and good discrimination of the two alkaloids close to quinine namely cinchonine and quinidine.The NIR PLS regression models were successfully validated using the total error approach with acceptance limits set at ± 10% with a risk level of 5%. The predictive performance of the methods developed was tested in terms of robustness.