Using the Thermo Scientific MarqMetrix All-In-One Process Raman Analyzer for real-time monitoring of a hot-melt extrusion process

Significance of the topic

Hot-melt extrusion (HME) is an increasingly important formulation route in pharmaceutical development for improving the bioavailability of poorly soluble Class IV active pharmaceutical ingredients (APIs). Real-time monitoring of HME supports consistent product quality, rapid detection of deviations, and regulatory documentation. Process Raman spectroscopy, implemented as a process analytical technology (PAT), furnishes non-destructive, inline chemical and solid-state information (API concentration, crystallinity, homogeneity) that can enable closed-loop control and reduce risk of out-of-specification batches.

Objectives and overview of the study

This application note evaluated the feasibility of using the Thermo Scientific MarqMetrix All-In-One Process Raman Analyzer to monitor API concentration during HME in real time. A model API was co-processed with a polymer in a twin-screw extruder across a range of loadings (15–60% API). Raman spectra collected at the die were correlated with off-line HPLC assays to develop and assess chemometric calibration models capable of predicting API content online.

Used instrumentation

• Thermo Scientific Pharma 11 Twin-Screw Extruder with two gravimetric feeders (separate feeding of polymer and API).
• Thermo Scientific MarqMetrix All-In-One Process Raman Analyzer equipped with a ball-probe sampling optic extruder probe mounted at the extruder die outlet.
• HPLC used as a reference assay for API concentration (two sample masses tested per point: ~2 mg and ~30 mg).

Methodology

Extrusion runs were performed at constant total throughput while varying feed rates to achieve target API mass fractions from 15% to 60%. After each change, the process was allowed to equilibrate (~4 minutes) before resuming spectral acquisition. Raman data were collected continuously in a monitoring mode with an 800 ms integration time, averaging 10 measurements per reported spectrum, using a 300 mW laser; each reported scan represented 16 s of acquisition. Spectra in the 800–1800 cm-1 region were selected because they capture dominant features from both polymer and API. Parallel extrudate samples were taken and analyzed by HPLC to provide reference concentrations for model calibration and validation.

Pre-processing of spectral data included first derivative (order 2, 15-point window, polynomial interpolation of tails), standard normal variate (SNV), and mean centering. Partial least squares (PLS) regression was used to build quantitative models correlating Raman spectra to HPLC-determined API content. The calibration set comprised ten concentration levels; two independent test points (40% and 50% API) were used to evaluate predictive performance.

Main results and discussion

Raman-derived PLS models demonstrated good quantitative performance across the tested concentration range. Two calibration models were developed based on two rounds of HPLC reference data; the model built from the first HPLC round had slightly lower RMSE values for calibration, cross-validation and prediction compared with the second-round model. HPLC duplicate analyses at each point (2 mg and 30 mg sample sizes) produced similar results, indicating sample homogeneity. Measured deviations between intended feed and HPLC-determined API content were observed at the lowest (24 g/h dosing) and highest dosing conditions, attributed to the very low feed rate and limited absolute mass of API, respectively. These dosing issues were peripheral to the Raman monitoring focus and were not pursued further.

The study demonstrates that process Raman, when combined with appropriate chemometrics, can track API concentration in-line and in near real time, enabling automated documentation and potential intervention to prevent faulty production lots.

Benefits and practical applications of the method

• Non-destructive, non-contact (or probe-separated) measurement that requires no sample preparation, minimizing waste.
• Real-time chemical and solid-state information (API loading, polymorphism/crystallinity, homogeneity) supports PAT and continuous manufacturing initiatives.
• Rapid measurements (sub-minute reporting) enable earlier detection of process deviations and corrective actions.
• Fiber-coupled probe flexibility allows safe distancing of analyzer from process environment and ease of integration.
• Automated data logging supports GMP documentation and regulatory traceability.

Limitations and considerations

Calibration models must be validated with independent, unknown data sets prior to deployment for routine control decisions. Low absolute feed rates and extremes of formulation may introduce sampling and dosing bias that should be addressed in process design and feeder specification. Signal quality can be affected by probe placement, melt opacity, and matrix spectral overlap; robust preprocessing and variable selection are necessary to ensure reliable predictions.

Future trends and possibilities for application

Advances that will further expand the utility of process Raman in HME include:

• Integration with multivariate statistical process control (MSPC) and model-predictive control (MPC) for closed-loop process adjustments.
• Use of larger, more diverse calibration sets and data-augmentation approaches (transfer learning) to improve model robustness across formulations and equipment.
• Application of machine learning methods for non-linear calibration and enhanced outlier detection.
• Improved probe designs and beam delivery to handle highly scattering or multi-phase melts, and to enable spatially resolved sampling.
• Regulatory acceptance trends favoring inline PAT for continuous manufacturing, reducing release testing burden through demonstrated process understanding.

Conclusion

The application note demonstrates that the MarqMetrix All-In-One Process Raman Analyzer can successfully monitor API concentration during HME in real time when combined with appropriate chemometric modeling. The technique delivers fast, non-invasive measurements that support documentation, process understanding, and timely intervention, offering clear value for pharmaceutical continuous manufacturing and PAT implementation. Further validation with independent data sets and attention to low-rate dosing challenges are required before routine process control deployment.

References

1. Yüce C., Hauch D., Chen L. Using the Thermo Scientific MarqMetrix All-In-One Process Raman Analyzer for real-time monitoring of a hot-melt extrusion process. Thermo Fisher Scientific Application Note 1498; 2025.