Determination of moisture content in freeze-dried materials by FT-NIR spectroscopy

Determination of moisture content in freeze-dried materials by FT-NIR spectroscopy — Summary

Significance of the topic

Residual moisture in lyophilized pharmaceutical products is a critical quality attribute that affects chemical stability, physical integrity and shelf life. Traditional reference methods such as Karl Fischer titration are accurate but destructive, reagent-dependent and slow, which is problematic for expensive or limited lyophilized batches. Rapid, non-destructive at-line monitoring of moisture enables real-time process control of lyophilization, reduces waste and supports Process Analytical Technology (PAT) initiatives. FT-NIR spectroscopy is particularly attractive because water produces strong overtone and combination bands in the near-infrared, enabling fast moisture quantification without sample preparation.

Objectives and overview of the study

The study evaluated the feasibility of Fourier-transform near-infrared (FT-NIR) spectroscopy to quantify moisture in lyophilized folinic acid calcium salt (Leucovorin Ca). The goal was to develop and validate a chemometric calibration that predicts residual water content across a practical range for lyophilized batches and to assess FT-NIR as a non-destructive at-line method to determine the end point of freeze-drying.

Methodology

- Sample set: Eight batches of lyophilized Leucovorin Ca with gravimetrically determined moisture (loss on drying) spanning approximately 2.65 to 8.04 % water.
- Spectral acquisition: Diffuse reflectance FT-NIR spectra collected through sealed clear glass vials to preserve hygroscopic samples; multiple spectra per vial to capture measurement variability.
- Instrument settings: Spectra recorded from 12,000 to 4,000 cm-1, collection time ~1 minute, spectral resolution 4 cm-1; internal gold flag in the integrating sphere used as the background reference.
- Preprocessing and modeling: Second derivatives applied to spectra to remove baseline offset and slope; no pathlength correction required due to similar particle size and composition across samples; stepwise multiple linear regression (SMLR) used to select predictive spectral regions (two component regions included) and build the quantitative model; leave-one-out cross-validation employed for internal validation.

Used instrumentation

- Thermo Scientific Antaris FT-NIR Analyzer with Integrating Sphere sampling accessory.
- Thermo Scientific TQ Analyst Method Development Software (SMLR implementation).
- Note: the application note references the Antaris model and mentions Antaris II as a more recent instrument with improved speed and performance.

Main results and discussion

- Calibration performance: Correlation coefficient (R2) for calibration = 0.9996; root mean square error of calibration (RMSEC) = 0.0610 % over the 2.65–8.04 % moisture range.
- Cross-validation: Leave-one-out cross-validation produced R2 = 0.9989 and RMSECV = 0.0987 %, indicating strong predictive ability and robustness within the tested set.
- Spectral observations: Clear spectral differences with varying moisture content, most pronounced in OH overtone and combination regions; baseline shifts were observed, attributable to vial/morphology differences, handled via derivative preprocessing and region selection.
- Practical implications: The method provides rapid (sub-minute), reagent-free, non-destructive moisture estimates through vial glass, enabling at-line or near-line decision-making (e.g., stopping lyophilization at optimal end-point) and reducing reliance on destructive gravimetric or titrimetric assays.

Benefits and practical applications

- Non-destructive, reagentless measurement suitable for valuable lyophilized pharmaceuticals.
- Fast turnaround (<1 minute) supports process monitoring and reduced batch losses during lyophilization.
- Ability to analyze samples in their sealed containers minimizes sample handling and contamination risk.
- Applicable to at-line or near-line PAT workflows for lyophilization cycle optimization and quality control.

Limitations and practical considerations

- Calibration dependency: Accurate FT-NIR quantification requires a robust calibration built against a reliable reference (here, gravimetric loss on drying). Calibration transfer across instruments, formulations or vial types may require additional work (standardization, transfer models or spiking strategies).
- Sample variability: Baseline shifts from vial geometry and sample morphology were observed; careful sampling protocols and preprocessing are necessary to control these influences.
- Sample set size: The study used eight batches; broader validation across more batches, different production lots and formulation variants is advisable before routine deployment.
- Hygroscopicity: Because samples were hygroscopic, they were measured sealed; real-time inline application may need controlled environments to avoid moisture exchange artifacts.

Future trends and potential applications

- Instrument advances: Higher-performance FT-NIR instruments (e.g., Antaris II and successors) offer faster acquisition, improved signal-to-noise and better stability, which can enhance model robustness and throughput.
- Advanced chemometrics: Multivariate approaches such as partial least squares (PLS), variable selection algorithms, and ensemble/regularized models can further improve prediction accuracy and generalizability. Preprocessing pipelines, wavelength selection and model validation strategies will remain important.
- Calibration transfer and model updating: Techniques for spectral standardization and transfer learning will ease deployment across manufacturing sites and instrument platforms.
- Inline/online integration: Development of fiber-coupled probes or integrated sphere assemblies for true inline moisture monitoring could enable continuous process control during lyophilization.
- Regulatory and PAT integration: Wider adoption will be supported by demonstration of method equivalence to reference techniques, robust validation protocols and alignment with regulatory expectations for PAT tools.

Conclusion

The feasibility study demonstrates that FT-NIR spectroscopy, combined with appropriate preprocessing and SMLR chemometrics, can accurately and rapidly quantify residual moisture in lyophilized Leucovorin Ca within the tested range (2.65–8.04 %). High calibration and cross-validation statistics indicate the technique is suitable for at-line or near-line monitoring to support lyophilization end-point decisions. Broader validation, attention to calibration transfer and consideration of sample variability are recommended steps before routine manufacturing implementation.

References

• Thermo Fisher Scientific. Application note AN50780: Determination of moisture content in freeze-dried materials by FT-NIR spectroscopy. 2022.