Applicability domains of a minimal-calibration model for effective online monitoring of pure components’ concentrations in the pharmaceutical continuous manufacturing processes

Abstract

Abstract Continuous manufacturing is an emerging topic in the pharmaceutical industry. To realize continuous manufacturing, quality assurance during processes is essential. To assure the product quality, calibration models are constructed, usually using data that are collected in advance. However, the cost for collection of calibration data is not so small, because model construction is conducted frequently to sustain the model’s accuracy. Iterative optimization technology (IOT) is a key technology involving infrared spectroscopy and predicting the concentrations of the pure components in the mixtures. IOT can be a powerful and useful tool, because the concentrations of the pure components in the mixtures, especially active pharmaceutical ingredient, are monitored through the pharmaceutical processes. However, IOT aims, in its original paper, to be applied for the process development step, not for the production step and its accuracy is not reliable enough for monitoring of the manufacturing process. In this work, we propose a method to evaluate quantitatively prediction accuracy for IOT prediction before acquiring the observed mole fractions. Assurance of the prediction accuracy is important as well as construction of accurate models in the actual operation. We estimate the prediction errors of the concentrations using the mixture infrared spectra simulated using Monte Carlo method.