Using On-Line Mass Spectrometry to Predict the End Point During Drying of Pharmaceutical Products

Downstream processing of Active Pharmaceutical Ingredients (APIs) typically involves extraction, crystallization, filtration and drying. In addition to being the most energy intensive unit operation, drying is often a major manufacturing bottleneck due to its relatively long processing times. Furthermore, drying conditions must be carefully selected and maintained to avoid problems such as API degradation. As a result, the optimization of dryer operation is needed to ensure product quality and minimize operating time. A key requirement for any optimization strategy is the ability to determine the drying end point, the time at which all solvent has been evaporated from the solid cake. While a variety of mathematical models have been developed to describe drying processes, available models are not sufficiently accurate to predict the end point and/or are too complex to be used for real-time applications in a pharmaceutical manufacturing environment. An alternative approach is to develop Process Analytical Technology (PAT) tools that take advantage of emerging on-line measurement capabilities. Near Infrared (NIR) spectroscopy and Mid Infrared (MIR) spectroscopy have been proposed as suitable methods for monitoring dryer performance and detecting the end point. However, these approaches require the development of chemometric models to map the measured spectra to solvent concentrations. Moreover, spectroscopic methods are only capable of measuring gas phase solvent concentrations, which can be quite different than solvent concentrations in the cake due to the gas residence time of the drying oven.

In this presentation, we describe the development and testing of a novel method for determining the end point for pharmaceutical dryers based on on-line mass spectroscopy. The proposed method offers several advantages over existing spectroscopic methods, including the ability to detect when the cake is dry from vapor phase measurements and very simple implementation that does not require chemometric models. The drying end point for each solvent is determined as the time at which the gas phase solvent concentration measurement from the mass spectrometer converges to a predicted value computed from a solvent mass balance on the oven assuming zero flow rate from the cake. The method was tested on a laboratory scale vacuum dryer over a range of temperatures and pressures using glass beads with three different particle sizes. Drying end points were detected for acetone, methanol and methanol-MTBE solvents well before the unprocessed gas phase solvent concentration measurements suggested that drying was complete. We found that the drying rate increased and the end point was reached more quickly as the mean bead size increased. The method was validated by performing Loss On Drying (LOD) experiments for one combination of pressure, temperature and bead size. Application of the method to an API with methanol-MTBE solvents produced a substantially reduced drying rate compared to the glass bead, most likely due to interactions between the API and solvents. We concluded that the proposed method represents a powerful Quality by Design (QbD) approach for pharmaceutical drying processes.