Merging Real-Time NIR and Process Parameter Measurements in a Fluidized Bed Granulation Process to Predict Particle Size

<b>Background/Objectives</b>: Controlling the critical quality attributes (CQAs), such as granule moisture level and particle size distribution, that impact product performance is essential for ensuring product quality in medicine manufacture. Oral solid dosage forms, such as tablets, often require appropriate powder flow for compaction and filling. Spray-dried fluidized bed granulation (FBG) is a key unit operation in the preparation of granulated powders. The determination of particle sizes in FBG using near-infrared spectroscopy (NIR) has been considered in the literature. Herein, for the first time, NIR is combined with process parameters to achieve improved prediction of the particle sizes in FBG. <b>Methods</b>: An inline model for particle size determination using both NIR and FBG process parameters was developed using the partial least square (PLS) method, or 'merged-PLS model'. The particle size was predicted at the end point of the process, i.e., the last 10% of the particle-size data for each batch run. An additional two analyses included a merged-PLS model with 12 batches: (1) where nine batches were training and three batches were a test set; and (2) where 11 batches were training and one was a test batch. <b>Results</b>: For all considered particle size fractions, Dv10, Dv25, Dv50, Dv75, and Dv90, an improved root-mean-squared error of prediction (RMSEP) is obtained for the merged-PLS model compared to the NIR-only PLS model and compared to the process parameters alone model. Improved RMSEP is also achieved for the additional two analyses. <b>Conclusions</b>: The improved prediction performance of endpoint particle sizes by the merged-PLS model can help to enhance both the process understanding and the overall control of the FBG process.