• In silico formulation: Application of artificial intelligence in support of hard gelatin capsule formulation of biopharmaceutics classification system class II drugs

      Wilson, Wendy I.; Augsburger, Larry L. (2004)
      Guo et al have demonstrated that a prototype hybrid expert network (PEN) for capsule formulation is capable of yielding formulations that meet specific manufacturing and performance criteria for the model Biopharmaceutics Classification System (BCS) Class II drug piroxicam (Pharm Tech NA (26:9) 2002, p. 44). Purpose. The overall objective of this project is to assess the application of artificial intelligence in capsule formulation support of BCS Class If drugs. The phases of this project include: (1) Characterize model BCS Class II drugs based on physiochemical properties associated with solubility; (2) Evaluate the dissolution performance of the capsule formulations; (3) Generalize the PEN for formulation of model drugs; (4) Apply a Bayesian Network (BN) to formulation development of the model drugs. Methods. The model drugs used in this project were carbamazepine (CAR), chlorpropamide (CHL), diazepam (DIA), ibuprofen (IBU), ketoprofen (KET), naproxen (NAP) and piroxicam (PIR). The micromeritic properties, aqueous solubilities, contact angles, specific surface areas (SSA) and intrinsic dissolution rates (IDR) of these actives were characterized. Capsule formulations of each active were manufactured based on a Box-Behnken experimental design, varying the filler type/ratio and the amounts of lubricant, wetting agent and disintegrant. Dissolution performance of these capsules was characterized using USP standard dissolution media. This data was used to generalize the PEN and to develop a BN. Results. The model drugs were subdivided into weak acids and weak bases. A dataset containing SSA, contact angles, IDRs and percent drug dissolved at 10, 30 and 45 minutes for the model drugs was used as the basis for training the PEN. The system was successfully able to predict the dissolution performance of the model drugs. A BN was successfully developed to model the relationships between formulation variables and dissolution performance. Conclusions. Testing of the PEN indicates that the system can predict the dissolution performance of BCS Class II drugs and can be successfully generalized. The BN has proven to be a useful tool for formulation development. Within the scope of this research, this project proves that artificial intelligence can be successfully applied to pharmaceutical research and development.
    • Quality by design: Application of a segregation tester to identify the critical factors affecting the weight and content uniformity of capsules filled on the two main types of automatic filling machines

      Xie, Lin; Hoag, Stephen W. (2006)
      The overall objective of this study is to use the experimental design approaches to investigate the relationship between quality uniformity and formulation parameters such as segregation tendency. The phases of this project include: (1) Optimize the experiment methodology for segregation testing following ASTM D 6940-04 standard and evaluate the segregation property of multiple formulations with varying particle ratios. (2) Investigate the effect of segregation tendency on the product quality (content, weight and drug concentration) indices in capsule filling process and the correlation of the parameters. A segregation tester was built following the ASTM D 6940-04 standard. Formulations containing 4% aspirin with coarse and fine particle sizes, microcrystalline cellulose (AvicelRTM PH200 and PH301) and magnesium stearate were used for the correlation study. A full factorial protocol was used to select the amount of materials needed (lot size, 200 g or 400 g) and the number of segregation cycle (Ns = 1, 5). The accuracy and robustness of the testing method was evaluated by comparing the L/F ratio of the mean values when one experimental run was repeated for different times (n). The content and weight uniformity quality indices of each formulation were assessed using the coefficient of variation. The coefficient of variation was estimated for within sampling points (CVi), between sampling points (CVj) and overall sampling points (CVij). Multiple regression and principal component analysis were used to analyze the data set and build the statistical models. Statistical analysis shows that the sample size is the critical parameter for segregation tendency testing. Segregation cycle is not statistical significant for the 400 g sample size testing. Segregation tendency measured is highly correlated to particle size ratio of drug to excipient. The results obtained demonstrate the reliability of the test method and standardize the critical details in the testing procedures of using the segregation tester to examine segregation. Machine type was the major parameter affecting the weight variation in this study. The segregation tendency correlated with the content variation. All these information will help in designing quality into formulation and a good control or monitoring on the critical parameters will enhance the product quality.