Data-driven full-chain quality control for Xiaoer Chiqiao Qingre granules: from process risk management to intelligent product testing

BACKGROUND: Xiaoer Chiqiao Qingre granules (XECQ) is a widely used traditional Chinese medicine for the treatment of pediatric wind-heat colds. However, its multi-step manufacturing process results in substantial inter-batch variability, posing a challenge for consistent quality control. Guided by the Quality by Design (QbD) concept, this study developed an integrated framework that combines process analysis, parameter optimization, and intelligent monitoring to improve batch-to-batch consistency and ensure medication safety. METHODS: Within the QbD framework, HPLC-based multi-component quantification, chromatographic fingerprinting, and physical fingerprinting were employed to monitor quality across the entire manufacturing process, including extraction, concentration, alcohol precipitation, and decoction collection. Eleven active constituents were quantitatively characterized, and their relationships with key physicochemical properties were analyzed. The retention rates of terpenoids, flavonoids, and phenylethanoid glycosides were defined as critical quality attributes (CQAs). Critical process parameters were screened using a Plackett-Burman design, and the design space for the alcohol precipitation process was established through Box-Behnken response surface methodology. In addition, a machine vision system was developed for rapid, non-destructive evaluation of particle appearance. RESULTS: > 0.85) were established to define a design space that maximized CQA retention while minimizing batch-to-batch variability. Furthermore, the machine vision system achieved 100% accuracy in both identification and abnormal batch rejection. CONCLUSION: This study proposes a holistic "Process Analysis-Parameter Optimization-Intelligent Monitoring" strategy, providing a robust framework for enhancing the quality control of XECQ and facilitating the transition of TCM manufacturing from empirically driven practices to data-driven processes.