口服固体制剂辅料近红外光谱定量方法的初步研究

发布时间：2018-04-27 07:31

本文选题：近红外光谱 + 药用辅料　；参考：《佳木斯大学》2017年硕士论文

【摘要】：目的探索固体药物制剂复杂分析系统中辅料的近红外光谱同时定量,尝试打破由于国内辅料现有检测标准缺少及固体制剂辅料多难以制成供分析检测的合适溶液而形成的仿制药一致性评价实施中多种辅料难以同时定量、无恰当的分析方法的初始瓶颈,探讨仿制药一致性评价中几种药品的近红外定量分析的光谱预处理、波长选择、模型优化与确定。为固体制剂仿制药一致性评价提供药用辅料的快速检测提供实用可行的分析方法。方法(1)将近红外光谱和化学计量学相结合快速检测苯磺酸氨氯地平片辅料含量。通过随机青蛙法、变量投影重要性和竞争自适应重加权采样筛选特征波长变量点,采用9种光谱预处理方法对原始光谱进行处理后,分别建立偏最小二乘法模型和支持向量回归分析模型,并进行两种模型的比较及应用优选模型测试样品。(2)将蒙特卡洛无信息变量消除结合遗传算法(MCUVE-GA)用于优选格列吡嗪药用辅料特征光谱变量,建立主成分分析(PCA)与人工神经网络算法(ANN)模型,比较分析所建立模型与偏最小二乘(PLS)模型的性能,优选模型建立方法。(3)利用近红外光谱法分别结合间隔偏最小二乘(iPLS)、反向区间偏最小二乘(BiPLS)、联合区间偏最小二乘算法(SiPLS)进行缬沙坦胶囊中药用辅料含量建模,比较分析各算法在不同划分区间数及区间选择时对建立模型的影响。(4)利用近红外漫反射技术结合偏最小二乘法建立多潘立酮药用辅料定量分析模型。通过考察光谱区域选择、光谱预处理及最佳主因子数选择等方面对模型进行不断优化,最终确定了最佳建模参数。结果(1)近红外漫反射光谱法快速检测苯磺酸氨氯地平片辅料结果表明:对于所涉及的样本,在最优特征波长变量选择上,随机青蛙法效果较好;在模型预测结果上,与支持向量回归分析模型相比,5个指标的偏最小二乘定量模型的决定系数,预测均方根误差评价参数效果较好,相对分析误差值均大于3.0。样品测试值与实测值标准误差均小于1.30,配对t检验表明,在α=0.05显著性水平上,两者无显著性差异。(2)近红外光谱法测定格列吡嗪片辅料结果表明:辅料淀粉、糊精、硬脂酸镁的偏最小二乘(PLS)模型的评价性能参数优于主成分分析(PCA)与人工神经网络算法(ANN)模型。而蔗糖经过共轭梯度学习算法训练得到性能参数优于PLS模型。配对t检验表明,在α=0.05显著性水平上,两者无显著性差异。(3)近红外光谱法测定缬沙坦胶囊辅料结果表明:辅料微晶纤维素和羧甲基淀粉钠BiPLS模型预测精度较好于iPLS和SiPLS模型精度,辅料聚维酮和十二烷基硫酸钠iPLS模型预测精度较好于BPLS和SiPLS模型精度。配对t检验表明,在α=0.05显著性水平上,两者无显著性差异。(4)所建多潘立酮片辅料PLS定量分析模型性能良好,验证集相关系数和预测均方根误差分别为0.9657,1.29;0.9870,0.877;0.9734,0.688;0.9474,0.734;0.9303,0.880;0.9777,0.0495。结论近红外漫反射光谱法结合化学计量学可快速检测苯磺酸氨氯地平片、格列吡嗪片、缬沙坦胶囊和多潘立酮片辅料含量,通过选取未参与建模的6组样品对模型有效性进行确认并通过配对t检验分析得出近红外测定的结果与实测值之间无显著性差异。分析方法操作简便快速、绿色环保、结果准确可靠,重复性、中间精密性、线性、精确性良好可为其他药用辅料含量快速检测提供了借鉴。在实际应用中,通过校正集和预测集样品容量增加,对模型进一步再优化和验证完善,可以不断提高模型的适用性和可靠性,更好的满足实际生产需求,对在线检测有重要的指导意义,同时,也进一步扩大近红外光谱分析的应用、有望助力于仿制药一致性评价。
[Abstract]:Objective to explore the near infrared spectroscopy of the auxiliary materials in the complex analysis system of solid drug preparation, and to try to break through the lack of the existing inspection standards for the domestic excipients and the difficulty in making the suitable solution of the solid preparation to make the suitable solution for analysis and testing. The initial bottleneck of the method is analyzed. The spectral preprocessing, wavelength selection, model optimization and determination of near infrared quantitative analysis of several drugs in generic drug consistency evaluation provide a practical and feasible analysis method for the rapid detection of pharmaceutical excipients for the consistency evaluation of solid preparations. Method (1) near infrared spectroscopy and chemometrics To detect the content of Amlodipine Besylate Tablets excipient quickly, the feature wavelength variable points are selected by random frog method, variable projection importance and competitive adaptive heavy weight sampling. 9 spectral preprocessing methods are used to process the original spectrum, and the partial least two multiplication model and support vector regression analysis model are set up respectively. Two models were compared and applied to optimize the model test samples. (2) Monte Carlo non information elimination combined with genetic algorithm (MCUVE-GA) was used to optimize the characteristic spectral variables of glipizide medicinal excipients, and the principal component analysis (PCA) and artificial neural network algorithm (ANN) model was established, and the model established and partial least squares (PLS) were compared and analyzed. The performance of the model and the optimal model establishment method. (3) using the near infrared spectroscopy (NIR) combined with spaced partial least squares (iPLS), reverse interval partial least squares (BiPLS) and joint interval partial least squares (SiPLS) to model the content of the auxiliary materials used in Valsartan Capsules. The influence of the model was established. (4) the quantitative analysis model of domperidone medicinal excipients was established by using near infrared diffuse reflectance technique combined with partial least square method. The model was optimized through the selection of spectral region, spectral preprocessing and the selection of the best main factor number. Finally, the optimal modeling parameters were determined. Results (1) near infrared diffuse reflectance The results of rapid detection of Amlodipine Besylate Tablets excipients by spectral method show that the random frog method has better effect on the selection of the optimal characteristic wavelength variables for the selected samples. Compared with the support vector regression analysis model, the decision coefficient of the minimum two multiplying model of the 5 indexes is compared with the model prediction results, and the root mean square error is predicted. The effect of the price parameter is better, the relative error value is greater than the 3.0. sample test value and the measured value standard error is less than 1.30. The paired t test shows that there is no significant difference at the level of the alpha =0.05 significance. (2) the result of the determination of Glipizide Tablets adjunct by near infrared spectroscopy shows that the partial least squares (PLS) of the auxiliary starches, dextrin and magnesium stearate. The evaluation performance parameters of the model are superior to the principal component analysis (PCA) and the artificial neural network algorithm (ANN). While the sucrose is trained by the conjugate gradient learning algorithm, the performance parameters are better than the PLS model. The paired t test shows that there is no significant difference in the significant level of the alpha =0.05. (3) the results of the near infrared spectroscopy for the determination of the Valsartan Capsules excipient result The results showed that the prediction accuracy of the BiPLS model was better than that of iPLS and SiPLS model. The prediction accuracy of the iPLS model was better than that of BPLS and SiPLS. The paired t test showed that there was no significant difference in the significant level of alpha =0.05. (4) the Domperidone Tablets was built. The auxiliary material PLS quantitative analysis model has good performance, the correlation coefficient and the mean square root error of the validation set are 0.9657,1.29, 0.9870,0.877; 0.9734,0.688; 0.9474,0.734; 0.9303,0.880; 0.9777,0.0495. conclusion the near infrared diffuse reflectance spectroscopy combined with chemometrics can quickly detect Amlodipine Besylate Tablets, Glipizide Tablets, and Valsartan Capsules. And the content of Domperidone Tablets auxiliary materials, by selecting 6 groups of samples not participating in modeling, confirming the validity of the model and analyzing the results of near infrared by paired t test, there is no significant difference between the results and the measured values. The analysis method is simple, fast, green, reliable, repeatable, intermediate precision, linear and accurate. Good sex can be used for reference for the rapid detection of other medicinal excipients. In practical applications, the application and reliability of the model can be improved continuously through the increase of the sample volume of the correction set and the prediction set, and the model is further optimized and verified. It is of great guiding significance for on-line detection. At the same time, it will further expand the application of near infrared spectroscopy analysis, which is expected to contribute to the consistency evaluation of generic drugs.

【学位授予单位】：佳木斯大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R927;O657.33

【参考文献】