The Use of Transmission FT-NIR Analysis for the simultaneous analysis of Carbamazepine and PEG 2000 in Extrudates

Significance of the topic

Near-infrared (NIR) spectroscopy integrated into continuous manufacturing offers real-time, non-destructive monitoring of critical material attributes during pharmaceutical extrusion. Monitoring active pharmaceutical ingredient (API) content and plasticizer levels during twin-screw hot-melt extrusion (HME) is especially important for producing stable amorphous solid dispersions of poorly soluble drugs, ensuring product quality, reducing waste, and enabling process control under PAT (process analytical technology) frameworks.

Objectives and study overview

This application study evaluated the feasibility of using transmission Fourier-transform NIR (FT-NIR) spectroscopy to quantify carbamazepine (CBZ, API) and polyethylene glycol 2000 (PEG 2000, plasticizer) simultaneously in clear extrudates composed of CBZ, Kollidon VA64 (polymer), and PEG 2000. The aim was to demonstrate in-line, real-time measurement capability for transparent melts that are not amenable to diffuse reflectance probes.

Instrumentation used

• Twin-screw extruder: Thermo Scientific PRISM Pharmalab 16 HME (temperature 120 °C; feed 300 g·h−1; screw speed 50 rpm).
• FT-NIR spectrometer: Thermo Scientific Antaris II Method Development System (MDS).
• Probes and sampling: Transmission probes mounted in a custom die with a 2 mm optical path length; powdered raw materials measured with an integrating sphere in disposable glass vials (diffuse reflectance) for reference spectra.
• Acquisition settings: spectral range 4000–10,000 cm−1; spectral resolution 8 cm−1; 32-scan averaging (~16 s per spectrum); continuous-mode acquisition at ~30 s intervals on-line.
• Software: RESULT for acquisition/workflow and TQ Analyst for chemometric model building.

Methodology

• Formulations: mixtures prepared with CBZ between ~5–20% and PEG 2000 broadly 2.5–25% (calibration set: ten mixtures; validation: two mixtures).
• Spectral preprocessing: second-derivative transformation to sharpen peaks and reduce baseline drift from melt scattering/noise.
• Analytical approach: initial single-wavenumber multiple linear calibrations targeted characteristic peaks (CBZ at 5064 cm−1; PEG at 5912 cm−1). The study notes overlapping absorptions for PEG and recommends PLS regression with larger sample sets for improved robustness.
• Validation and demonstration: real-time predictions were performed on unknown extrudates during extrusion runs to monitor stabilization and composition trends at the die exit.

Main results and discussion

• Spectral features: Pure-component and extrudate transmission spectra showed distinct CBZ absorption near 5064 cm−1 and a weaker PEG band near 5912 cm−1; polymer (Kollidon VA64) had overlapping contributions requiring preprocessing and chemometrics.
• Calibration performance: CBZ single-wavenumber calibration achieved correlation coefficient ≈ 0.994 and RMSEC ≈ 0.48% (w/w). PEG calibration produced correlation ≈ 0.97 and RMSEC ≈ 0.82% (w/w), with higher scatter—particularly for samples impacted by cloudiness or overlap.
• Real-time monitoring: In-line predictions tracked concentration changes as formulations transitioned; examples include a 7.5% CBZ/10% PEG run where predicted CBZ values were within 0.5% of nominal despite slow melt stabilization. Another run (15% CBZ/5% PEG) showed trends converging toward target values though stabilization required many minutes.
• Limitations observed: At high CBZ load (20%) melts became cloudy due to incomplete API dissolution, increasing spectral noise and baseline shifts and reducing prediction precision. PEG quantification was affected by overlapping absorptions and calibration sample variability.

Benefits and practical applications

• Enables real-time in-line monitoring of API and plasticizer concentrations for transparent melts where reflectance probes fail.
• Supports PAT goals: faster process feedback, potential for reduced end-product testing, and improved batch-to-batch consistency during HME.
• Non-destructive and rapid measurement (spectra every ~30 s) compatible with continuous manufacturing time scales.
• Applicable for process troubleshooting (e.g., detecting incomplete dissolution/cloudiness) and process optimization (residence time, feed adjustments, plasticizer levels).

Future trends and opportunities

• Adopt multivariate chemometrics (PLS, robust cross-validation) with expanded calibration sets to reduce interference from overlapping bands and improve limits of detection and accuracy.
• Combine transmission NIR with other PAT sensors (Raman, inline melt temperature/torque monitoring) and multivariate process control to provide orthogonal information about solid-state form, dissolution, and process stability.
• Optimize probe/die geometry and path length for a broader range of formulations, and develop strategies to mitigate scattering from partially crystalline/cloudy melts.
• Integrate model maintenance workflows (adaptive or transfer learning) to accommodate raw material variability and scale-up from lab to production extruders.

Conclusion

Transmission FT-NIR spectroscopy mounted at the extruder die can successfully monitor API (carbamazepine) and plasticizer (PEG 2000) levels in transparent HME melts in real time. Single-wavenumber calibrations proved feasible and delivered good accuracy for CBZ and reasonable results for PEG under the tested conditions. For industrial deployment, expanding calibration libraries and applying multivariate regression will enhance robustness against spectral overlap and process variability. The approach supports PAT-driven continuous manufacturing and provides actionable in-line composition information for process control.

References

• Kelly A., Halsey S., Terrell M. The Use of Transmission FT-NIR Analysis for the simultaneous analysis of Carbamazepine and PEG 2000 in Extrudates. Thermo Fisher Scientific application note AN52565 (2014).