Abstract
The aim of the work is to develop a data fusion model using near-infrared (NIR) and process parameters for the predictions of drug dissolution from controlled release multiparticulate beads. Using a design of experiments, ciprofloxacin-coated beads were manufactured and critical process parameters such as air volume, product temperature, curing temperature, and curing time were measured; environmental humidity was monitored usinga Pyrobuttons®. The NIR spectra were decomposed using principal component analysis (PCA). The PCA scores were fused with process measurements and all variables were autoscaled. The autoscaled variables were regressed against measured dissolution data at 1h and 2h time points; the PLS regression used quadratic and cross terms. The NIR spectra only model using data collected at the end of bead curing generated a PLS model using 5 latent variables with R2 equal to 0.245 and 0.299 and RMSECV 13.23 and 13.12 for the 1h and 2h dissolution time points, respectively. The low R2 and high root mean square error of cross validation (RMSECV) values indicate that NIR spectra alone were insufficient to model the drug release. Similar results were obtained for NIR model using data collected at the end of spraying phase. Models with fused spectral and process data yielded better prediction with R2 above 0.88 and RMSECV less than 5% for the 1h and 2h dissolution time points. The data fusion model predicted dissolution profiles with an error less than 10%.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have