Two Case Studies of the Transfer of Near-Infrared Methods for the Analysis of Pharmaceutical Solid Dosage Forms

Significance of the topic

Near-infrared (NIR) spectroscopy is increasingly used in pharmaceutical quality control and process analytical technology for non-destructive, rapid analysis of solid dosage forms. Reliable transfer of chemometric calibration models between instruments is critical for routine deployment across production sites and for scaling laboratory methods to manufacturing. Demonstrating robust method transfer reduces the need for complex instrument-matching algorithms, lowers validation workload, and supports consistent product quality across laboratories and plants.

Objectives and study overview

This application note evaluates the feasibility of transferring calibration models developed on a primary Thermo Scientific Antaris FT-NIR analyzer to a second (target) Antaris instrument. Two distinct tablet types with different physical and compositional characteristics were used to test transferability: a small, thin tablet with low active pharmaceutical ingredient (API) fraction (<10% w/w) measured in transmission mode, and a larger, thicker oval tablet with high API fraction (>40% w/w) measured in diffuse reflectance mode. The core goal was to determine whether models built on one instrument could be applied directly on a matched target instrument with minimal prediction error and no systematic bias.

Methodology

Measurements were performed non-destructively on the Antaris FT-NIR Method Development Sampling System using both the tablet detector and integrating-sphere modules so samples could be analyzed in transmission and reflectance without relocation. Key acquisition and chemometric settings were:

• Reflectance: 50 scans per tablet, resolution 8 cm-1, spectral range 4000–10000 cm-1.
• Transmission: 100 scans per tablet, resolution 8 cm-1, spectral range 6000–12000 cm-1 (analysis used 8924–11209 cm-1 for transmission models).
• Preprocessing: second derivative spectral pretreatment with Norris smoothing (Norris 11-point for reflectance model; Norris 25-point for transmission model).
• Modeling: Partial Least Squares (PLS) regression. Reflectance model used 3 latent factors; transmission model used 4 latent factors.

All reflectance measurements were taken on the unscored tablet side to reduce variability arising from tablet scoring.

Used instrumentation

• Thermo Scientific Antaris FT-NIR Method Development Sampling System.
• Tablet detector module for transmission sampling and integrating sphere module for reflectance sampling; internal laser-based frequency calibration providing high x-axis stability across instruments.

Main results and discussion

• Small, low-API tablets (transmission): The calibration exhibited very high correlation (reported correlation coefficient ~0.9996) with RMSEC = 0.247 mg per tablet. When the model was applied on the target instrument, RMSEP was 0.486 mg/tab and no significant bias was observed, indicating successful transfer for transmissive tablet forms.
• Large, high-API tablets (reflectance): The reflectance PLS model achieved a calibration correlation coefficient of 0.9864 and RMSEC = 1.65% (w/w). After transfer to the target instrument, RMSEP was 1.64%, demonstrating nearly seamless transfer. A noted caveat was a 3.5% difference between reflectance measurements on scored versus unscored tablet sides, highlighting surface-related variability for reflectance sampling.
• Direct comparison of primary versus target instrument performance on the same tablet samples showed close agreement, with percent differences generally within about ±1.8% of the primary instrument values.

Discussion points:

• The Antaris FT-NIR’s internal laser frequency calibration and matched instrument design contributed to strong x-axis stability and instrument sameness, reducing the need for corrective transfer algorithms.
• Choice of sampling mode is critical: transmission is preferable for small, more translucent tablets with low API load, while reflectance is better for larger, less transmissive, high-API tablets.
• Surface features (scoring) can introduce measurable bias in reflectance measurements and should be controlled in sampling protocols or compensated for in modeling.

Benefits and practical applications

• Enables rapid deployment of validated NIR methods across instruments and sites, supporting decentralized QC and process monitoring.
• Reduces dependency on transfer-correction algorithms and extensive instrument matching procedures when using stable, well-matched FT-NIR platforms.
• Supports non-destructive, high-throughput testing of tablets for content uniformity, release testing, and in-process control.
• Minimizes rework associated with method re-development, although regulatory re-validation requirements must still be considered in pharmaceutical contexts.

Future trends and possibilities

• Standardization of instrument designs and internal frequency references will further simplify calibration portability across platforms and manufacturers.
• Development of robust universal calibrations and transfer strategies leveraging advanced chemometrics, domain adaptation, or machine-learning approaches to accommodate remaining instrument and sample variability.
• Integration with manufacturing execution systems and cloud-based model management to enable centralized calibration updates and remote validation across sites.
• Increased emphasis on sampling protocols (e.g., handling of scored tablets, orientation control) and automated sample presentation to reduce surface-related variability in reflectance measurements.
• Expansion of PAT deployments where validated transferable NIR models enable real-time release testing and continuous monitoring in line with regulatory frameworks.

Conclusion

This study demonstrates that, on a purpose-designed FT-NIR platform with high frequency stability and matched instrument design, chemometric models for tablet assay can be transferred between instruments with minimal loss of performance. Appropriate selection of sampling mode (transmission vs. reflectance), careful modeling (PLS with derivative pretreatment and smoothing), and attention to sample-surface effects are key to robust transfer. While regulatory re-validation considerations remain, the results indicate that well-engineered FT-NIR systems can greatly simplify multi-instrument deployment of NIR tablet assays.

Reference

Thermo Scientific Application Note 50646: Two Case Studies of the Transfer of Near-Infrared Methods for the Analysis of Pharmaceutical Solid Dosage Forms. Thermo Fisher Scientific, 2008.