Post 512150 (https://www.thevespiary.org/talk/index.php?topic=8556.msg51215000#msg51215000)
(wareami: "Squidippy's Plug...", Stimulants)[Materials and Methods
Drug Release Data
Dissolution data (125 profiles) obtained from SR minitablets, both coated and uncoated, were gathered for analysis. The original outputs for the neural network were the dissolution profiles for these SR tablets. A USP apparatus type 2 containing 900 mL of 0.015 M citrate/phosphate buffer, pH 6.8, was used for tablet dissolution testing. The complete dissolution of both enteric coatings (Eudragit L and S) required that the medium pH was greater than pH 6.5. Sodium lauryl sulphate (SLS), 0.5%, was introduced to achieve sink conditions for the CEL50.
The profiles were normalized between 0.0% and 100%. The times taken for 10% of the dosage form and 90% of the dosage form to be released were calculated from the profiles. These were converted to 2 outputs for use in the neural network:
1. Time taken for 10% of the drug to be released (Tlag) was taken to be an indication of the lag time before release began (ie, the extent of the delaying effect of the enteric coating).
2. Time taken for 90% of the drug to be released less the time taken for 10% of the drug to be released (T90-10) was taken to be an indication of the time taken for the drug to be released from the SR matrix once the outer coat had dissolved.
Similar simple dissolution parameters, such as the time to 50% drug dissolution have been used in other studies.2,4
The 2 separate phases in the dissolution profile (lag time followed by SR) are illustrated in Figure 1.
Software
The Trajan Neural Network Simulator, version 4, (Washington, Tyne and Wear, UK) was the software package used in this study. This is a Windows-based package, which supports numerous types of neural networks along with the fastest state-of-the-art training algorithms. The package includes other algorithms that carry out a variety of tasks such as pre- and post-processing of data, input feature selection, network design, and selection. The software also gives extensive statistical feedback on each network.15
Input variables
The number of inputs used for the network was 19 (see Tables 2, 3, and 4). The inputs included 11 variables related to the composition of the formulation and 8 variables related to the processing conditions: percentage Cel50 in the formulation, percentages of methocel K15M, methocel K100M, methocel K100LV, klucel, aerosil, magnesium stearate, avicel PH101, percentage weight gain after Eudragit L or Eudragit S coating, percentage weight gain after Opadry coating, spray rate of the coating solution (g/min), blend size of the original tablet batch (kg), batch size at that stage of processing (kg), time (minutes) the formulation was blended with the lubricant (magnesium stearate), press speed (revolutions per hour), Filomatic speed setting (setting on dial that governs how fast the blend was fed down from the hopper into the tablet press), tablet hardness (newtons), and tablet weight (mg).
http://www.aapspharmscitech.org/view.asp?art=pt040226&pdf=yes (http://www.aapspharmscitech.org/view.asp?art=pt040226&pdf=yes)
http://www.indianpharma.org/pt/index.php/2004/2February/article6.htm (http://www.indianpharma.org/pt/index.php/2004/2February/article6.htm)
http://www.usp.org/ (http://www.usp.org/)
http://store.usp.org/OA_HTML/usp_ibeCSrdSrchResults.jsp?cg=-200&kw=pseudoephedrine&ds=0&dr=20&st=kw&cpg=0 (http://store.usp.org/OA_HTML/usp_ibeCSrdSrchResults.jsp?cg=-200&kw=pseudoephedrine&ds=0&dr=20&st=kw&cpg=0)