A universal quantitative model was developed to determine the azithromycin content in granules using near infrared （NIR） diffuse reflectance spectroscopy. The diffuse reflection spectra were recorded with the integr...A universal quantitative model was developed to determine the azithromycin content in granules using near infrared （NIR） diffuse reflectance spectroscopy. The diffuse reflection spectra were recorded with the integrating sphere at 8 cm-1 resolution in 4000–12 000 cm-1 spectral range. During each measurement, 32 co-added scans were performed. This quantitative model was constructed with 103 batches of azithromycin granules from 21 different manufacturers. The azithromycin content ranges from 3.0% to 24.5%. The root mean square error of prediction （RMSEP） of model was 0.613. In addition, the quantitative model was evaluated in terms of specificity, linearity, accuracy, and precision according to ICH guidelines. In conclusion, it is feasible to construct a universal quantitative model for azithromycin granules by choosing suitable training set samples and selecting an appropriate wavelength range. The quantitative model could be applied in the quick assay of azithromycin granules produced by domestic manufacturers （content： 3.0%–24.5%）.展开更多
The training set of a universal near infrared (NIR) model for quantitative analysis of a drug should cover as many samples of this drug in the market as possible. Inevitably the model may fail for new products that ha...The training set of a universal near infrared (NIR) model for quantitative analysis of a drug should cover as many samples of this drug in the market as possible. Inevitably the model may fail for new products that have different excipients and production processes. In such circumstances the model should be updated. We here propose a new strategy to iteratively update a universal NIR quantitative model for azithromycin. We prove that universal quantitative models generated from this new strategy are comparably effective for azithromycin injection powders and azithromycin tablets, compared to the strategy using hierarchical clustering method which we reported previously. Furthermore, we establish the correlation coefficient r between a new sample and the training set samples can be used to decide whether or not the model should be updated.展开更多
基金National Science and Technology Major Project of the Ministry of Science and Technology of China (Grant No.2010ZX09401-403)
文摘A universal quantitative model was developed to determine the azithromycin content in granules using near infrared （NIR） diffuse reflectance spectroscopy. The diffuse reflection spectra were recorded with the integrating sphere at 8 cm-1 resolution in 4000–12 000 cm-1 spectral range. During each measurement, 32 co-added scans were performed. This quantitative model was constructed with 103 batches of azithromycin granules from 21 different manufacturers. The azithromycin content ranges from 3.0% to 24.5%. The root mean square error of prediction （RMSEP） of model was 0.613. In addition, the quantitative model was evaluated in terms of specificity, linearity, accuracy, and precision according to ICH guidelines. In conclusion, it is feasible to construct a universal quantitative model for azithromycin granules by choosing suitable training set samples and selecting an appropriate wavelength range. The quantitative model could be applied in the quick assay of azithromycin granules produced by domestic manufacturers （content： 3.0%–24.5%）.
文摘The training set of a universal near infrared (NIR) model for quantitative analysis of a drug should cover as many samples of this drug in the market as possible. Inevitably the model may fail for new products that have different excipients and production processes. In such circumstances the model should be updated. We here propose a new strategy to iteratively update a universal NIR quantitative model for azithromycin. We prove that universal quantitative models generated from this new strategy are comparably effective for azithromycin injection powders and azithromycin tablets, compared to the strategy using hierarchical clustering method which we reported previously. Furthermore, we establish the correlation coefficient r between a new sample and the training set samples can be used to decide whether or not the model should be updated.
