457484 CFD MODEL Based Near-Infrared Spectroscopy Implementation for in-LINE ASSAY Monitoring of a Pharmaceutical Suspension

Thursday, November 17, 2016: 3:15 PM
Continental 4 (Hilton San Francisco Union Square)
Maxim Verstraeten and Thomas De Beer, Laboratory of Pharmaceutical Process Analytical Technology, Ghent University, Ghent, Belgium

INTRODUCTION AND PURPOSE

During bottle filling of a pharmaceutical suspension formulation, a decreasing trend in assay is observed at the end of the batch. This leads to approximately 17% of the batch content which cannot be commercialized. The root cause can most likely be found in foam present in the top layer of the suspension in the holding tank which contains less API. The aim in this study is to implement a PAT solution for establishing a batch-dependent endpoint (I.e. assay ≤ 95%) determination while reducing in-process sampling drastically.

MATERIALS AND METHODS

Materials

Tests were performed with a high-dose (i.e. 100 mg/mL) pharmaceutical suspension formulation. Lab scale samples were prepared in a 10L mixing vessel. Full scale batches are manufactured in a 150L compounding vessel following a one-pan process. After the transfer to a holding tank, the full scale batches are gradually discharged to a specific bottle filling system with the help of a peristaltic pump. During this process step, the suspension is stirred continuously.

Methods

NIR spectra were recorded using a Fourier-transform Near-Infrared (NIR) spectrometer (Bruker Matrix-F Duplex) equipped with a TE-InGaAs detector, a quartz halogen lamp and a fiber-optic non-contact probe. Every 30 seconds, a spectrum was acquired in the 12000-4000 cm-1 region with a resolution of 8 cm-1, averaged over 64 scans. Ultra Performance Liquid Chromatography (UPLC) was selected as reference technique for API content determination. All UPLC measurements were performed at Janssen Pharmaceutica, Beerse, Belgium, using their developed and validated method (confidential).

RESULTS AND DISCUSSION

Samples in a range of 85-115% target concentration were analyzed with both NIR spectroscopy and UPLC. A lab-scale calibration model was developed which shows acceptable feasibility for NIR spectroscopy as potential PAT tool for the API assay determination in suspension.

For approaching the production measurement setup as good as possible, the calibration model was optimized by using a circular system with a similar flow as in production (i.e. 800 mL/min) with the help of a customized T-piece. Samples with concentrations varying between 90-103% were measured and included in the prediction model.

At production scale however, the developed model showed a weak performance with an unexpected variation in NIR assay results. It was assumed that the signal variation was created by disturbed flow patterns/turbulence in front of the lens due to possible back pulse formation at the system output. A Computational Fluid Dynamics (CFD) simulation of the flow pattern of the suspension inside the measuring interface confirmed this assumption.

CFD simulations demonstrated that an extension of the original T-piece could optimize the measurements by transferring the back pulse formation further away from the lens. This mechanical solution allowed a constant sample volume per time unit in front of the NIR laser. This was confirmed with additional lab-scale tests and the follow-up of several production scale batches.

CONCLUSIONS

Despite the dominantly present water absorption bands, NIR spectroscopy is feasible to establish good API assay predictions of this high solid-load suspension formulation.

Experiments indicated that a fine-tuning of the interfacing device is crucial to achieve adequate in-line NIR measurements. Therefore, sufficient process knowledge, including fluidum mechanics, is desired to rationalize the choice for an optimal measurement setup.

CHALLENGES AND FUTURE WORK

In the future it is aimed to investigate if NIR spectroscopy in combination with this specific interfacing device is applicable for other liquid pharmaceutical formulations.


Extended Abstract: File Not Uploaded