Development and validation of global prediction models for monitoring the manufacturing process of herbal medicine by ultraviolet spectroscopy

Abstract Objective: Process monitoring for traditional Chinese medicine (TCM) preparations is necessary to ensure quality of the product. A typical pharmaceutical process of TCM preparations consists of multiple manufacturing units, such as ethanol precipitation, concentration, and water precipitation, among others. Compared with the traditional practice of one prediction model for one unit, the global model covers the variation from samples with different backgrounds or processes and can be used to monitor intermediates from substeps. Methods: We used ultraviolet (UV) spectroscopy to establish global models for a typical TCM preparation—Danhong injection. The concentrations of danshensu, protocatechualdehyde, rosmarinic acid, salvianolic acid A, salvianolic acid B, and hydroxyl safflor yellow A and the total phenolic and total sugar contents were quantified for every intermediate from operation units of Danhong injection. New samples prepared by mixing different intermediates were introduced for the calibration set to cover more variations. An accuracy profile was employed to validate the developed method from the aspects of specificity, trueness, precision, accuracy, linearity, and robustness. Results: The developed models showed a high determination coefficient ( R 2 ) value up to 0.97 and a low root-mean-square error of the prediction set. Five components of the models passed all validation tests, whereas the total sugar was not suitable for modeling with UV and was not applicable to the whole process. Conclusions: This study indicates that the global models of UV spectroscopy for the quantitative determination of phenolic acids are feasible and reliable with a simple, rapid, and non-destructive method.