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

Importance of the Topic

Hot-melt extrusion (HME) is an established manufacturing route to enhance solubility and bioavailability of poorly soluble active pharmaceutical ingredients (APIs), particularly Biopharmaceutics Classification System class IV compounds. Real-time analytical control during HME is critical to ensure correct API loading, homogeneity and solid-state properties (e.g., crystallinity and polymorphism). Implementing robust process analytical technology (PAT) methods such as process Raman spectroscopy enables manufacturers to monitor, document and react to process deviations inline, supporting product quality, process understanding and regulatory compliance.

Objectives and Study Overview

This application note evaluated the feasibility of using the Thermo Scientific MarqMetrix All-In-One Process Raman Analyzer for online monitoring of API concentration during HME. A model API was blended into a polymer matrix across a targeted concentration range (15–60% w/w) using a twin‑screw extruder. Raman spectra acquired at the extruder die were correlated with off-line HPLC assays to build multivariate calibration models for real-time API quantification and to demonstrate the utility of Raman as a PAT tool in HME operations.

Methodology

Experimental design and data collection steps included:

• Materials and feeding: Separate gravimetric feeders delivered API and polymer to a Thermo Scientific Pharma 11 twin‑screw extruder to achieve nominal API loadings from 15% to 60% (constant total throughput).
• Inline spectral acquisition: A ball‑probe sampling optic (Dynisco probe style) was mounted at the extruder die and coupled by fiber to the MarqMetrix Raman analyzer. Spectra were recorded at 1‑minute intervals during process changes; each acquisition used 800 ms integration, averaged over 10 means, with 300 mW laser power (≈16 s per analysis scan).
• Off‑line reference: Extrudate pellets (≈1 mm × 1 mm) were sampled in parallel and analyzed by HPLC. Two sample masses (≈2 mg and ≈30 mg) were assayed at each setpoint to assess sample homogeneity and validate Raman predictions.
• Data preprocessing and modelling: Spectral region 800–1800 cm−1 was selected. Preprocessing combined a first derivative (order 2, window 15 points, polynomial interpolation for tails), standard normal variate (SNV) and mean centering. Partial least squares (PLS) regression models were developed using the ten calibration levels and validated via cross‑validation and test samples (including two independent tests at 50% and 40% API).

Used Instrumentation

• Thermo Scientific Pharma 11 Twin‑Screw Extruder with gravimetric feeders.
• Extruder die mounted ball probe sampling optic (Dynisco-style) for inline Raman coupling.
• Thermo Scientific MarqMetrix All‑In‑One Process Raman Analyzer (fiber-coupled).
• HPLC system for off‑line reference assays (model not specified in the note).

Main Results and Discussion

Key findings from the study were:

• HPLC results confirmed generally good sample homogeneity: assays from ~2 mg and ~30 mg sample masses produced similar API percentages across most setpoints.
• Discrepancies between nominal feed setpoints and HPLC results were observed at the lowest (15%) and highest (60%) target loadings. These deviations were attributed to extremely low dosing rates at the low end (e.g., 24 g/h) and limitations in high‑end dosing accuracy rather than spectroscopic error.
• PLS models built from two rounds of HPLC reference data produced accurate calibration across the 15–60% API range. The model based on the first HPLC round showed slightly lower root mean square errors for calibration, cross‑validation and prediction compared with the second round. Both models demonstrated that Raman spectra in the 800–1800 cm−1 window contain sufficient information to quantify API content during extrusion.
• The authors emphasize that the models require external validation with independent datasets before routine deployment; nonetheless, inline Raman provided continuous, non‑destructive concentration monitoring and automatic documentation suitable for GMP traceability.

Benefits and Practical Applications

Practical advantages demonstrated or discussed include:

• Real‑time API concentration monitoring enabling rapid detection of process deviations and potential intervention to avoid out‑of‑spec batches.
• Simultaneous assessment of multiple quality attributes from a single scan (e.g., API content, crystallinity/polymorphism indicators, and potentially moisture-related bands).
• Non‑invasive, no sample preparation workflow that minimizes waste and preserves sample integrity.
• Flexible deployment due to fiber‑optic coupling between probe and analyzer, allowing safe, remote placement of the analyzer.
• Automated data logging that supports regulatory documentation and PAT strategies for continuous manufacturing.

Future Trends and Potential Uses

Opportunities to expand and improve the approach include:

• Robust model validation and transfer: building larger calibration sets, including independent validation batches, to improve model robustness and enable model transfer between lines or facilities.
• Closed‑loop process control: integrating Raman predictions with feed control to automatically correct API dosing or process settings in real time.
• Advanced chemometrics and machine learning: exploring non‑linear models or hybrid approaches to capture complex matrix effects, polymorphism, and temperature‑dependent spectral changes.
• Multimodal PAT: combining Raman with NIR, terahertz or imaging sensors for more comprehensive control of solid‑state properties and content uniformity.
• Regulatory integration: aligning spectral model development, validation and data handling with regulatory expectations for PAT and continuous manufacturing documentation.

Conclusion

This application note shows that inline process Raman spectroscopy, implemented with the MarqMetrix All‑In‑One Process Raman Analyzer and a die‑mounted probe, can reliably monitor API concentration during hot‑melt extrusion across a broad loading range. The method is non‑destructive, supports continuous documentation and has the potential to reduce batch failures by enabling timely process corrections. Prior to production use, multivariate models must be validated with independent datasets and process conditions to ensure robust quantitative performance.

Reference

Thermo Fisher Scientific. Using the Thermo Scientific MarqMetrix All‑In‑One Process Raman Analyzer for real‑time monitoring of a hot‑melt extrusion process. Application Note 1498. 2025.