| Objective The near-infrared spectroscopy and chemometrics methods was combined with five kinds of oral solid preparations,azithromycin tablets,nimodipine tablets,nifedipine tablets,folic acid tablets and indapamide tablets as the research object,to explore the method of rapid quantitative analysis of the content of principal and auxiliary materials in drugs,which will provide analysis methods for the consistency evaluation of generic drug on sprescription content,and will provide a simpler and more accurate reference method to check the quality of marketable preparations.Methods In the experiment,the quantitative models of five drugs were established separately.In the study,the concentration of the main drug and the excipient were determined by analyzing the balance range,thus avoiding the use of alternative methods to obtain their reference concentration.The five drugs were modeled and designed according to their prescriptions,and then the simulated preparations were prepared.The spectrogram was collected on the Fourier transform near infrared spectrometer.When modeling,different chemometrics methods were selected,such as spectral pretreatment,elimination of abnormal variables,screening of modeling variables and selection of quantitative calibration models according to instrument background noise and spectral interferences,specific characteristic peaks of each component,and ranges of main and auxiliary materials,etc.According to the parameters,an optimal quantitative model of near-infrared principal and auxiliary materials was determined.Samples not involved in modeling were put into the model to predict the content.The errors between the known principal and auxiliary material content and the predicted value of the model were compared,and the accuracy of the model was evaluated.Finally,the stability,repeatability and linearity of the model were verified by the sample of the test set.Results In this paper,five quantitative analysis models of drugs were established by means of near infrared spectroscopy and chemometrics,and the prediction results were accurate and rapid.(1)In the process of establishing azithromycin model,it used the standard normal transformation variable(SNV)+ De-trend algorithm(Detrending)to preprocess the spectrum,which effectively eliminated the noise and improved the spectral resolution.After sifting out the modeling variables by using Monte Carlo(MC-UVE)eliminating information variables,the complexity of the model was reduced and the computation was reduced.Finally,the parameters of the partial least squares(PLS)model are significantly improved compared with the full wavelength modeling.There was no significant difference between the main drug content predicted by NIR method and the measured results of HPLC.The error between the predicted value and the reference value of each excipient material was small.All these indicated that the prediction performance of the model was good and could be used as an effective alternative to the conventional quantitative method.(2)In the process of establishing a quantitative model of nimodipine,some abnormal samples of drugs were eliminated by Chauvenet test,combined with lever value and student residual.The partial least square mathematical model was established after pretreatment of the remaining sample spectra and interval selection of modeling.It was found that the correlation parameters of the four kinds of excipients were better and the linear relationship of the main drug nimodipine was poor.The regression model of the support vector machine(SVM)of nimodipine was also established,and the model parameters and prediction performance were better than the PLS model,which shows that the SVM method has more advantages than the PLS method in dealing with nonlinear problems.(3)In the process of nifedipine modeling,it is found that the calibration model obtained by combined interval partial least squares method(siPLS)was superior to that obtained by application interval partial least square method(iPLS)and inverse interval partial least squares method(siPLS).which showed that the siPLS method was more advantageous in selecting valid variables during the modeling of this experimental sample.SiPLS method was applied to the other five components of the sample,and the parameters in the quantitative model all met the requirements.(4)In the establishment of folic acid quantitative model,Monte Carlo cross validation method(MCCV)was used to remove the abnormal samples,and after that,the model quality was improved.After the standard normal transformation variable and multivariate scattering correction were preprocessed,the quantitative model was established according to the weighted regression coefficient to screen the variables with strong correlation in each component.The performance of the model was good and stable.(5)In the establishment of the calibration model of indapamide,the abnormal samples were eliminated by combining the lever value and the chemical error.The sample set was divided by DUPLEX method,and the best principal factor number of the model was selected by using the retention method,and a better model was established.The T-test method was used to test the mean of the NIR prediction value and the reference value in the verification set.There was no significant difference between the two methods.Finally,the quantitative method was investigated and the results were satisfactory,which could be used as a reference for the future application.Conclusions In this paper,five quantitative calibration models of drugs were established by means of near infrared spectroscopy and chemometrics.The parameters of each model were better,which can realize the simultaneous content of the main drug and auxiliary materials,and predict the content of unknown samples.The result was accurate and the error was small,which can meet the requirement.The method was environmentally friendly,simple,fast and nondestructive.It can provide reference for scientific research institutions and grass-roots testing departments,and also can assist in the consistency evaluation of generic drug. |
PDF Full Text Request