中药过程分析中NIR技术的基本理论和方法研究

【作者】 吴志生；

【导师】 乔延江； 史新元；

【作者基本信息】 北京中医药大学 ， 中药学， 2012， 博士

【摘要】 基于“顶层设计”理念的中药过程分析学从中药制药过程全局出发,对生产过程的单元、要素进行统筹考虑和系统设计,最终实现质量稳定、均一、安全、有效。近红外光谱(NIR)技术作为过程分析领域发展最快、应用最广、最受瞩目的技术之一,可满足过程分析技术快速、无损、可靠、简便的要求。三文针对中药过程分析技术面临的理论、方法、技术及应用等重点问题开展了研究工作,取得了如下成果：理论方面：①基于两类误差多变量分析理论,建立了中药复杂体系中多变量检测的理论公式；②借鉴总误差分析理论,研究了NIR技术用于中药复杂体系定量分析的评价参数,包括准确性、真实性、精密度、定量限、风险性、线性和回收率等；③研究了NIR频率谱区的吸收理论,并对不同近红外频率谱区的定量性能进行了评价。通过上述研究,为NIR用于中药过程分析奠定了理论基础。方法学方面：①以清开灵注射液中间体银黄液、金银花浓缩液乙醇沉淀中间体和银黄颗粒与淀粉混合物三者为载体,结合间隔偏最小二乘(iPLS)、向后间隔偏最小二乘(BiPLS)、移动窗口偏最小二乘(mwPLS)等变量筛选方法,研究了不同体系的近红外多变量检测限估计值,结果表明水溶液体系的近红外多变量检测限比醇溶液体系的检测限高(醇溶液体系中检测限估计值可低至几十个ppm),固体体系的近红外多变量检测限比液体体系的检测限高；②采用符合FDA和ICH验证标准的准确性轮廓(Accuracy Profile)分析方法,引入β-期望容差区间参数,以上述三种体系为载体,建立了中药液体制剂醇溶液体系、水溶液体系和固体体系中近红外定量分析方法；③采用准确性轮廓方法的定量分析参数作为评价标准,根据近红外定量分析流程,以金银花浓缩液乙醇沉淀中间体为载体,解决了低含量中药样本集的NIR模型可靠性间题,提出了准确性轮廓验证方法的NIR模型筛选新策略；④采用准确性轮廓方法考察了柠檬酸在不同溶剂体系中的定量限,阐明了溶剂和温度对NIR定量分析的影响,指出FCOT和CR频率谱区存在饱和吸收,SCOT频率谱区是变量筛选最佳光谱区域之一等结论。上述方法的建立,为中药过程分析中NIR定量分析研究提供方法学依据。技术方面：①以清开灵注射液为载体,结合间隔偏最小二乘、向后间隔偏最小二乘、移动窗口偏最小二乘变量筛选方法,系统考察了全息光栅型、傅里叶变换型、声光可调型和微电子阵列型等不同类型近红外光谱技术的多变量检测限,研究表明不同类型近红外光谱技术的多变量检测限估计值在几十到几百ppm。其中,新类型的近红外光谱技术,在中药分析领域多变量检测限估计值达到了ppm级,肯定了近红外技术在本领域广阔的运用前景；②采用准确性轮廓分析方法对不同类型和不同采样方式的近红外技术的定量分析结果进行了评价,建立了相应的定量分析方法体系,为本领域近红外技术的选择提供参考；③运用光谱分析新技术------近红外光谱成像技术,以乳块消素片和复方甘草片为载体,构建了复杂体系中活性成分的空间分布,研究了基础相关性分析方法(BACRA)、直方图法和像元块标准差法(SDMT)在片剂成分空间分布均匀性评价中的应用。以银黄片中金银花提取物、黄芩提取物和淀粉三元体系的混合为载体,构建三元体系中活性成分空间分布图,建立了二值化图像中移动式像元块标准差方法(MBMSTDEV),以此进行中药制药混合过程均匀度的评价方法。以不同产地的甘草为载体,选取甘草酸为指标成分,采用线性模型(PLS)和非线性模型(LS-SVR),阐明了利用高维数据进行低含量成分定量分析的准确性,为近红外成像技术在中药定量分析领域的运用提供铺垫。上述研究结果为NIR研究过程仪器及采样方式的选择提供了依据,也为不同研究需求提供了更多的技术方法。应用方面,①建立了中试规模的清开灵注射液在线近红外过程分析技术平台；②集成近红外过程分析技术平台的优势,以清开灵注射液生产过程金银花提取单元为例,结合间隔偏最小二乘算法,运用前期建立的多变量检测限评价方法,确定了金银花提取过程中近红外光谱在线分析方法的多变量检测限估计值(达到了十几ppm)；③建立了金银花提取过程近红外光谱在线分析方法的验证体系。这些应用不仅为近红外技术在清开灵注射液生产过程的实际运用奠定了坚实的基础,也为其在中药液体制剂生产过程的应用提供了依据。通过上述研究,本文较全面地解决了NIR技术用于中药过程分析领域的几个关键问题,所形成的理论、技术、方法及装备体系为中药生产过程质量控制体系水平的升级提供了支撑。更多还原

【Abstract】 Process analysis of Chinese material medica depends on the concept of "top-level design". Overall consideration and "top-level design" of process unit can ultimately assure that the product quality is stable, uniform, safe and effective. The objective of process analysis was to understand the whole procedure of Chinese material medica. Attributed to rapid, nondestructive, reliable and simple requirements, near infrared spectroscopy (NIR) has been widely used as one of the most important process analysis technologies for Chinese material medica. According to the key issues faced in this field, the thesis was to research the basic theory and methodology of process analysis of Chinese material medica by NIR.In the aspect of theory,①introduce both typeⅠand typeⅡerror and provide the theoretical formula of multivariate detection limits in the field of process analysis of Chinese material medica by NIR;②introduce total error theory and establish analytical methodology of process analysis of Chinese material medica by NIR, including accuracy, precision, limit of quantification, risk, linearity, and recovery;③research the NIR frequency absorption theory and assess the quantitative performance in each frequency spectral region.In the aspect of methodology,①take Qingkailing injection intermediate, Lonicera japonica solution in ethanol precipitation process and Yihuang particles as examples and research multivariate detection limits estimates of NIR based on the methods of interval partial least squares, backward interval partial least squares and moving window partial least squares. The results show that multivariate detection limits estimate is higher in solid system than in aqueous solution. On the other hand, multivariate detection limits estimate is higher in aqueous solution than alcohol solution system. Multivariate detection limits estimate attach dozens of ppm in alcohol system;②take Qingkailing injection intermediate, Lonicera japonica solution in ethanol precipitation process and Yihuang particles as examples and establish NIR quantitative analysis method by accuracy profile (AP) in alcohol system, aqueous solution system and solid system of Chinese material medica;③take quantitative analysis parameters of AP method as criteria, illuminate the reliability of NIR model by sample sets with low APIs content in CHM. A promising model selection strategy for the determination of low-content anlytes ishighly recommended by AP;④citric acid was used as solute, while water, ethanol and mixture of ethanol and water are used as solvent respectively. Taking quantitative analysis parameters of AP method as criteria, the result show that FCOT or CR region is preferred as a variables selection region, but if saturated absorption occurred in FCOT or CR region, SCOT region was a better choice. Technology,①take Qingkailing injection as an example, themultivariate detection limits estimates of different types of NIR technologies (Holographic grating, Fourier transform, AOTF, MEMS) were investigated based on the methods of interval partial least squares, backward interval partial least squares and moving window partial least squares. The result show that multivariate detection limits estimates of novel type of NIR technologies attach ppm level, which will broad the use of NIR in the process analysis of Chinese material medica;②NIR quantitative analysis method was established by AP in different types of NIR instruments and different sampling methods.③NIR-CI served as a process analysis technology in the field of Chinese material medica. Take Rukuaixiao tablets as an example, the spatial component distribution was visualized using BACRA method. Furthermore, histogram analysis method was applied to assess the homogeneity of the tablets. Taking compound liquorice tablets as an example, standard deviation method (SDMT) was introduced to characterize the compound distribution of compound liquorice tablets based on binary image and MBMRSTDEV was introduced as a suitable approach to evaluate the homogeneity in blending process. Taking different origins of licorice as a carrier, a rapid and non-destructive quantitative analysis of glycyrrhizic acid content was carried out depended on linear model (PLS) and nonlinear model (LS-SVR) by NIR hyperspectral imaging. The result proved that hyperspectral data could be effectively employed, which paved the way for NIR hyperspectral imaging in the field of quantitative analysis of Chinese material medica.Application aspects,①The pilot platform of process analysis for Qingkailing injection was established by on-line NIR;②integrate the advantages of pilot platform, theextraction process of Lonicera japonica was taken as an example and research the multivariate detection limits estimates of on-line NIR based on interval partial least squares were determined. Imply multivariate detection limits theory into the practical application and provide about 10 ppm level of multivariate detection limits estimate of on-line NIR;③establish on-line NIR quantitative analysis method.All in all, the thesis focusing on the basic theory, methodological research, technology promotion and application of process analysis of Chinese material medica, provides several key issues of NIR technology in strengthening the quality control in the process analysis of Chinese material medica.更多还原