Inline Monitoring of a Hot Melt Extrusion Process by Near Infrared Spectroscopy

Importance of the topic

The development of reliable in-line process analytical technologies (PAT) for hot melt extrusion (HME) addresses regulatory and manufacturing trends toward quality-by-design (QbD) and real-time release. Near-infrared (NIR) spectroscopy directly monitors critical product attributes such as drug load and homogeneity in continuous HME, enabling earlier detection of out-of-specification material than indirect process signals (e.g., torque) and supporting robust process control and reduced waste.

Objectives and overview of the study

This study demonstrates method development and implementation of Fourier-transform NIR (FT-NIR) as an in-line reflection measurement to quantify theophylline content in extrudates produced by a twin-screw HME. Key goals were to identify prerequisites for successful quantitative models, compare spectral discrimination across loadings, build and validate a PLS calibration, and illustrate real-time monitoring during process startup and steady-state operation.

Methods

The model system used an anhydrous theophylline API dispersed in polyethylene oxide (PEO, Sentry WSR N10) at nominal loadings of 0%, 5%, 10%, 15% and 20% w/w. Some batches included a fixed fraction of lactose to increase surface reflectivity and improve NIR signal quality. Materials were hand-blended in polyethylene bags for 3 minutes, fed by a single-screw feeder (FW 18, Brabender) at 500 g/h into a Pharma 16 HME twin-screw extruder (Thermo Fisher). Screw speed was 100 rpm with barrel temperatures up to 120 °C; the screw profile included two kneading sections to promote melting and mixing. Spectra were collected in reflection mode at a probe mounted near the die via a 1/2"-20 UNF Dynisco port; measurement cadence captured spectra during approximately 30 minutes per concentration for calibration and continuously during test runs.

Used instrumentation

• Antaris MX FT-NIR spectrometer (Thermo Fisher Scientific) with a fiber-optic reflection probe installed at the extruder nozzle port.
• Pharma 16 HME twin-screw extruder (Thermo Fisher Scientific).
• Single-screw feeder FW 18 (Brabender Technologies) operating at 500 g/h.

Data analysis and calibration strategy

Spectral pre-processing and multivariate analysis were used to extract quantitative information. Principal component analysis (PCA) highlighted spectral variance associated with drug loading and showed tighter cluster dispersion at higher loadings (improved spectrum consistency when lactose was present). Partial least squares (PLS) regression was used to build a quantitative calibration for theophylline content. Model quality was assessed by standard diagnostics including PRESS, RMSEC and cross-validation metrics. The PRESS analysis indicated that two PLS factors sufficiently captured the variance relevant to the theophylline signal, consistent with a robust, low-complexity model.

Main results and discussion

• The reflection-mode FT-NIR method could discriminate and quantify theophylline loadings across the investigated range; inclusion of lactose markedly improved reflectivity and signal quality and yielded a stronger calibration.
• Calibration statistics for the lactose-containing batches showed a correlation coefficient (R) of 0.9938 and RMSEC of 0.837, indicating high linearity and low calibration error.
• Only two PLS latent variables were needed to model theophylline content, suggesting a parsimonious and stable model appropriate for in-line trend monitoring.
• During a production run nominally containing 15% theophylline, the Antaris MX predictions tracked drug content in real time and correctly predicted the 15% level within the calibration error. Importantly, spectral predictions indicated that true compositional homogeneity in the extrudate was achieved later than the point where extruder torque reached a steady value, demonstrating that torque alone can be a misleading indicator of product uniformity.
• PCA cluster behavior showed that increased drug content (and the presence of lactose) reduced within-class spectral variability, improving discrimination and calibration performance.

Practical benefits and applications

The demonstrated FT-NIR in-line reflection approach provides direct, non-destructive monitoring of drug load and homogeneity in HME processes. Practical advantages include:

• Real-time feedback on product quality enabling earlier detection of out-of-spec material and more reliable determination of steady-state production.
• Support for QbD and PAT strategies by enabling operation within defined design spaces and potential for real-time release decisions.
• Low-complexity PLS models (few latent variables) reduce risk of overfitting and simplify transfer to automated workflows.
• Applicability to diverse formulations where surface reflectivity can be managed (e.g., using reflective excipients) or where transmission measurement is impractical due to opacity.

Limitations and practical considerations

• Choice of reflection versus transmission measurement depends on extrudate opacity; opaque or highly scattering melts favor reflection probes placed near the die.
• Formulation variables (e.g., excipients, particle size, surface texture) and process parameters (temperature, shear) can alter the NIR signal and must be considered during calibration development and validation.
• Manual pre-blending in this study may differ from production feeding strategies; robust calibrations should be developed with representative process variability and validated under multiple operating conditions.

Future trends and potential applications

Further work should systematically map the influence of process variables (melt temperature, shear, residence time distribution) and formulation parameters on NIR spectral features to build transfer-ready, robust models. Promising directions include:

• Integration of multi-sensor PAT (NIR plus Raman, imaging or torque/process analytics) to provide orthogonal confirmation of composition and physical state.
• Advanced calibration transfer strategies and adaptive modeling to accommodate scale-up, screw geometry changes, and formulation variability.
• Deployment of automated, validated software workflows (e.g., RESULT or similar) for continuous monitoring, trend analysis, and real-time release.
• Expansion to quantify additional quality attributes such as solid-state form (amorphous vs crystalline), API dissolution modifiers, and content uniformity at higher throughput.

Conclusion

The study illustrates that in-line FT-NIR reflection spectroscopy, combined with PCA and low-complexity PLS models, can accurately quantify theophylline load in PEO extrudates and detect process homogeneity later than conventional torque metrics indicate. Proper selection of measurement mode, attention to surface reflectivity (e.g., lactose addition), and rigorous model validation are critical prerequisites. Implemented within PAT frameworks, FT-NIR enables direct, real-time quality monitoring and supports QbD strategies in HME manufacture.

References

1. Tumuluri SV, Prodduturi S, Crowley MM, Stodghill SP, McGinity JW, Repka MA, Avery BA. Drug Dev Ind Pharm. 2004 May;30(5):505-11.
2. Rohe T, Becker W, Kölle S, Eisenreich N, Eyerer P. Talanta. 1999 Sep 13;50(2):283-90.
3. Gendrin C, Roggo Y, Spiegel C, Collet C. Eur J Pharm Biopharm. 2008 Mar;68(3):828-37. Epub 2007 Aug 10.