The Optimization Of Sustained-release And Controlled-release Based On Modified Quadratic Inference Functions And NSGA-Ⅱ

Objective:Compare the modeling of sustained and controlled release drug data by Modified quadratic inference functions(MQIF), Quadratic inference functions(QIF) and Generalized estimating equations(GEE);Use NSGA-II optimizing of sustained and controlled release drug at each time point cumulative release. Provide a more reasonable and feasible method of sustained and controlled release on modeling and optimizating process.Methods:Taking the sustained and controlled release preparation as the independent variable and cumulative release of each time as the dependent variable,choose the matrix by exchangeable correlation matrix,use GEE,QIF and MQIF to establish the model of the repeated measurement time(T). Evaluate fitting effect of the model by relative efficiency(SRE), red pool information(AIC) and Bias information(BIC).Point time variables into the fittest effect model,establish the model of sub objective function,take Pharmacopoeia standards as the optimization objective,use the improved non dominated sorting genetic algorithm(NSGA-II) for multi-objective optimization,and compare the results with the original optimization.Results:(1) Results on modeling: GEE haven’t likelihood function, so use SRE to compare the fitting effect. Compared with GEE,Dextromethorphan Hydrobromide sustained-release capsules MQIF and QIF’s SRE are 906.25,and 1058.01.Multi particle film controlled release tablets MQIF and QIF are all 1.839. It can be seen that modeling by MQIF and QIF is better than GEE; MQIF and QIF defined the likelihood function,so use AIC and BIC to evaluate the fitting effect; Example of Dextromethorphan Hydrobromide sustained-release capsules, the AIC and BIC of MQIF are 22 and 24, the AIC and BIC of QIF are 31 and33.17.Example of multi particle film controlled release tablets,the AIC and BIC of MQIF are 18.87 and 25.82, AIC and BIC of QIF are 26 and 36.04.The AIC and BIC of MQIF are less than QIF,visible,MQIF’s modeling effect is better than QIF. Therefore,the fitting effect of MQIF in three modeling methods is the best.(2) Analysis of optimization effect: Using NSGA-II to optimizate the model established by MQIF.The Pareto non inferior solution set of Dextromethorphan Hydrobromide sustained-release capsules,Scheme 7 display: When HPMC K4 M 21.90 mg,sodium BICarbonate 10.63 mg,hexadecanol 20.96 mg,2,4 and 8 hours’ s cumulated release were 39.2171%,58.8231%,76.7355%,close to Pharmacopoeia requirements:40%,60,80%;the original literature by direct method, it accumulate the three points release into a target, through than,the direct method it only can get a unique solution in the test level,contrary to the multi-objective Pareto,it cannot be given the predictive value of cumulative release degree at each time point, and can not evaluate the degree of the best release target.In Pareto non dominated solutions of Multi particle film controlled release tablets, plan 29shows: when the coating weight is 8.0%, the curing time is 43.8h, increasing plastic agent concentration is 65.0%, five point cumulative release are 16.2%, 32.2%, 49.4% and 72.3%,95.8%, at each time point close to Pharmacopoeia requirements:16.7%, 33.3%, 50%, 75%,100%, The deviation degree of the best release target are 3.0%, 2.2% and 1.3%, 3.6% and4.0%. The original document by means of artificial neural network analysis method(ANN)using quantitative weighted average method, transformed into a single objective through weighted values, then given the only solution, determine the optimal solution scheme with optimal release goal deviation respectively 9.6%,0.6% and 0.4%,0%,2.0%, the method using weighted method has certain rationality, but the method at the expense of the first point release cost to ensure the other point. At the same time,the original document also uses the response surface method(RSM),use the 3D surface map and contour map,the observation and analysis of the optimal solution,given the only solution and its subjectivity is strong,the determination of the optimal solution scheme with optimal release goal deviation respectively 6.6%,3.9%,5.0%,2.7%,3.1%, the overall deviation is greater than the method of this text.Conclusions:The modeling effect of Modified quadratic inference functions(MQIF) for the sustained and controlled release formulations is superior to that of the generalized estimating equation(GEE) and quadratic inference functions(QIF);As Modified quadratic inference function(MQIF) as the objective function,use NSGA-II optimize to sustained and controlled release preparation.The Pareto non dominated solutions are close degree and the degree of deviation is small with the standard of Pharmacopoeia optimal release;Therefore,the Modified quadratic inference functions(MQIF) and improved non dominated sorting genetic algorithm(NSGA-II) can be combined and used as a more reasonable method to controlled-released formulation optimization.