Near-infrared Spectroscopic Quality Control on Coating Process of Vitamin C Yinqiao Tablets / 中国实验方剂学杂志

ABSTRACT

ObjectiveTo construct a quantitative prediction model of three indicators（moisture absorption rate， film thickness and coating weight gain） during the coating process of Vitamin C Yinqiao tablets（VCYT） by near-infrared spectroscopy（NIRS）， and to realize the endpoint judgment. MethodReal-time NIRS data of 4 batches of VCYT during the coating process were collected by diffuse reflection method. The coating method employed was the rolling coating method， and the samples were obtained at the spray stage from the coater's sampling port every 10 minutes， and 57 batches of samples（about 1 800 tablets） were collected at various coating times， the tablets were embedded in molten paraffin， cut longitudinally， and observed by stereomicroscope. The film thickness， with a target value of 38 μm， was then measured using Motic Images Advanced 3.2 software. Furthermore， the mositure absorption rate of samples， aiming for a target value of 3%， was determined in accordance with guiding principles for drug hygroscopicity testing in the 2020 edition of Chinese Pharmacopoeia， and 3 samples were randomly selected from each batch（10 tablets per batch）， and the coating weight gain was calculated（target value of 4%）. Partial least squares regression（PLSR） was used to construct a quantitative model of the 3 coating indicators， and the predicted values of the coating indicators were smoothed using the moving average method and used to determine the coating endpoints. ResultThe prediction determination coefficients（Rp2） for moisture absorption rate， film thickness and coating weight gain were 0.933 4， 0.932 6 and 0.965 9， the root mean square errors of prediction（RMSEP） were 0.163 5%， 1.870 9 μm and 0.240 3%， the relative percent deviations（RPD） were 3.711 0， 2.760 7 and 5.415 8， respectively. The results of the external validation set demonstrated that the real-time predicted values obtained by the models exhibited the same trend as the measured values， and the coating endpoint could be accurately predicted（with a prediction error of less than 7.32 min and a relative error of less than 5.63%）. ConclusionThe established NIRS model exhibits excellent predictive performance and can be used for quality control of VCYT during the coating process.