Benchtop NIR method development for continuous manufacturing scale to enable efficient PAT application for solid oral dosage form

An Optimization Approach Coupling Preprocessing with Model Regression for Enhanced Chemometrics

“…Overall, models built on robustness are anticipated in the literature to give simpler models, i.e., models with a fewer number of latent variables, that might not initially give the most accurate predictions but are in general less amenable to degradation with time. This is crucial as it can decrease the frequency and effort for model updates and maintenance and is in alliance with what was reported in Alam et al 38…”

“…In this direction, also multimodel approaches that can split the known variability realizations into clusters are conceivable. , However, approaches including all variations in the data can be generally impractical, as they may require very large data sets, which can increase experimental efforts . To avoid this problem, design of experiments for the calibration set to obtain maximum information on the variation has been also investigated in the literature, e.g., Alam et al and Li et al Furthermore, sometimes incorporating all variations might not be possible, as it may require data from the future . Such an example may for instance occur when there is data available for a process on lab and pilot scale, but we are interested in building a model that performs well in a fully industrial scale, although we still do not have any data for this scale beforehand.…”

“…Model accuracy is typically put forward in terms of goodness of fit of the prediction. As previously mentioned, the use of RMSEP is commonly utilized in this regard. , Alternative/complementary metrics for accuracy, such as root-mean-square error of calibration, or cross validation and/or bias are conceivable. These measures are sufficient for understanding the average errors in a data set; however they should not be interpreted as an estimate for the uncertainty of a specific prediction at a new input .…”

Raman and near-infrared spectroscopy for in-line sensors

“…In the beginning of the twenty-first century, these spectroscopic in-line sensors were developed to investigate manufacturing process of solid dosages [5] and for quantification of active pharmaceutical ingredients (APIs) in a dosage form [6,7]. Today, these techniques are utilized as PATs not only for batch manufacturing but also for continuous manufacturing [8]. On another front, continuous synthesis and manufacturing of APIs and biopharmaceuticals are innovative processes and expected to be cost-effective and time-efficient production processes.…”