Study On The Analysis Of Lyophilized Powder For Injection With Diffuse Reflectance FT-Nirs And Chemometrics

Lyophilized powder for injection is a sterile preparation prepared by freeze drying technology. It is widely used for the preparation of injections which were sensitive for moisture and heat, because of its characteristics that high stability, low moisture content and good rehydration. On the quality control of lyophilized powder for injection, residual water may drastically affect the drug stability and the content of active component can indicate the drug effectivity, Karl-Fischer titration (KFT) and high performance liquid chromatography (HPLC) are the regular approaches to determine moisture and active component of lyophilized powder for injection. However, these methods are destructive, polluting and time-consuming. Especially, the measured values of moisture are high for the cylindroid vials are opened during KFT and atmospheric moisture is absorbed. On the other hand, the discrimination of cake structures is also a significant test for lyophilized powder for injection, the general method for the discrimination of cake structures is visual inspection (VI) although the method is subjective. Therefore, it is imperative to establish an accurate, non-destructive and rapid method for the quality control of lyophilized powder for injection.Combined with chemometrics, diffuse reflectance Fourier transform near infrared spectroscopy (FT-NIR) can simultaneously extract multiple chemical and physical information of an analyte. It is widely used for the analysis of agricultural products, foods, petrochemicals and pharmaceuticals. Compared with the traditional methods, near infrared spectroscopy (NIRS) possesses many advantages that sample-preparation-free, non-destructive, pollution-free, rapid and easy-to-use.In this work, the feasibility of the simultaneous analysis of moisture, active component and cake structure of lyophilized powder for injection, such as lyophilized potassium sodium dehydroandroandrographolide succinate for injection, with diffuse reflectance FT-NIR chemometrics was investigated while KFT, HPLC and VI were used as the reference methods, respectively. Root mean square error of cross validation (RMSECV), root mean square error of prediction (RMSEP) and correlation coefficient (R) were used for the evaluation of moisture and active component, prediction accuracy and performance index were used for cake structure.OBJECTIVES1. Establish a method to determine moisture contents of lyophilized potassium sodium dehydroandroandrographolide succinate for injection with diffuse reflectance FT-NIR.2. Establish a method to determine active component contents of lyophilized potassium sodium dehydroandroandrographolide succinate for injection with diffuse reflectance FT-NIR.3. Establish a method to discriminate cake structures of lyophilized potassium sodium dehydroandroandrographolide succinate for injection with diffuse reflectance FT-NIR.METHODS1. Measurement of Fourier transform near infrared diffuse reflectance spectra (NIRDRS) of lyophilized potassium sodium dehydroandroandrographolide succinate for injection Scan region: 10 000–4000 cm^-1; number of scans: 64; resolution: 8 cm^-1; temperature: 22±1°C; humidity: 55±5% RH.2. Determination of the reference values of lyophilized potassium sodium dehydroandroandrographolide succinate for injection(1) Determination of moisture contents: the reference values of moisture contents were obtained by KFT.(2) Determination of the contents of active component: the reference values of the contents of active component were measured by HPLC. Column: C₁₈ column (250mm×4.6mm, 5μm), mobile phase: methanol-potassium dihydrogen phosphate (pH 3.0; 0.05 mol/L) (64:36, v/v), flow rate: 1.0 ml/min, column temperature: 40°C, injection volume: 20μl, detection wavelength: 251 nm.(3) Discrimination of cake structures: the cake structures were discriminated by VI.3. Calibration and validation of the models for the analysis of lyophilized potassium sodium dehydroandroandrographolide succinate for injection(1) Calibration and validation of the moisture models: the moisture models of NIRS were established by back-propagation artificial neural network (BP-ANN) and partial least squares (PLS).(2) Calibration and validation of the active component models: the active component models of NIRS were established by BP-ANN and PLS.(3) Calibration and validation of the cake structure models: the cake structure models of NIRS were established by self-organizing map (SOM) and discriminant analysis (DA).RESULTS1. Determination of moisture contents of lyophilized potassium sodium dehydroandroandrographolide succinate for injection with diffuse reflectance FT-NIR(1) BP-ANN model: the network structure parameters were 6-5-1, RMSECV was 0.2376, RMSEP was 0.1471, R of the prediction set was 0.9553.(2) PLS model: the numbers of factors was 5, RMSECV was 0.2267, RMSEP was 0.1576, R of the prediction set was 0.9293.2. Determination of active component contents of lyophilized potassium sodium dehydroandroandrographolide succinate for injection with diffuse reflectance FT-NIR(1) BP-ANN model: the network structure parameters were 7-4-1, RMSECV was 0.0470, RMSEP was 0.0082, R of the prediction set was 0.9891.(2) PLS model: the numbers of factors was 3, RMSECV was 0.0479, RMSEP was 0.0242, R of the prediction set was 0.7964.3. Discrimination of cake structure of lyophilized potassium sodium dehydroandroandrographolide succinate for injection with diffuse reflectance FT-NIR(1) SOM model: NIRDRS data were reduced to 10 principle components (PCs), the cumulative contribution rate was 98.4%, the prediction accuracy was 100%.(2) DA model: NIRDRS data were reduced to 10 PCs, the cumulative contribution rate was 97.8%, the performance index was 86.8%.CONCLUSIONThe simultaneous analysis of moisture, active component and cake structure of lyophilized potassium sodium dehydroandroandrographolide succinate for injection using diffuse reflectance FT-NIR and chemometrics was established. The BP-ANN models showed better performances for content predictions of moisture and active component; the SOM model appeared much more suited for the discrimination of cake structures.