428862 Determination of Reaction Profiles Using Online FTIR Spectroscopy Coupled with Kinetics Model-Based Multivariate Curve Resolution

Monday, November 9, 2015: 1:38 PM
Ballroom D (Salt Palace Convention Center)
Zhihao Lin, PAT, Merck & Co. Inc., Rahway, NJ

Fourier Transform Infrared (FTIR) spectroscopy is an important tool for the monitoring and controlling  Active Pharmaceutical Ingredients (API) manufacturing processes. High spectral resolution and information content are particularly suitable for monitoring chemical reactions where differentiation of structurally similar molecules is required to maintain the quality of products. With a brief introduction to the setups of online FTIR in pharmaceutical reaction monitoring, this presentation will focus on a novel approach that uses the kinetics model-based multivariate curve resolution algorithm as a “calibration-free” tool for real-time analysis of batch reaction processes.

 In this application, online FTIR was used to collect data from a hydrogenation process that makes an API (drug substance). The FTIR spectra are fed into the kinetics model-based multivariate curve resolution algorithm to generate the concentrations of product and starting material in real-time without using traditional calibration models. Experiments in the lab have shown that the results from kinetics model-based multivariate curve resolution were similar to those generated by Partial Least Square (PLS) prediction models in terms of analytical accuracy. This offers a “calibration-free” alternative to traditional multivariate (chemometrics) calibration for overcoming spectral overlapping in quantitative analysis. Because the process parameters are explicitly expressed in the kinetics model, their values can be changed easily to reflect the changes in the process condition. Therefore, unlike statistical calibration models that often require adjustment or complete rebuild once process conditions are changed, the kinetics model-based multivariate curve resolution algorithm is able to adjust to the new conditions and perform normally as before. This has been demonstrated in the scale up of the hydrogenation process to pilot plant, in which the batch surface area, hydrogen pressure and agitation were different from those in the lab experiments. The kinetics model-based multivariate curve resolution algorithm worked well without any adjustment. This study has demonstrated an alternative approach for meeting the challenges in supporting fast changing process development and scale up without using multivariate (chemometrics) calibration models that require lengthy adjustment or rebuild once process conditions are changed.

Extended Abstract: File Not Uploaded