Near-Infrared Analysis of Critical Parameters in Lyophilized Materials

Significance of the topic

Lyophilized products are widely used in pharmaceutical and biotech industries to extend shelf life and facilitate storage at ambient temperatures. Quality control of freeze-dried formulations is challenging because opening sealed vials destroys the sample and may invalidate the lot. Near-infrared (NIR) spectroscopy offers a nondestructive approach that can probe through common vial materials to assess critical stability parameters such as residual moisture and biological potency rapidly and without consumables.

Objectives and overview of the study

This application note evaluated the performance of an Antaris Fourier Transform NIR (FT‑NIR) spectrometer for noninvasive QC of lyophilized thrombin. Two primary quality attributes—residual moisture (target <1% w/w) and potency (enzymatic/coagulation activity)—were modeled from vial-through spectra using multivariate chemometrics. The study also examined spectral effects caused by cake settling during handling and shipping and assessed the ability to discriminate intact versus settled cakes.

Methodology

Two independent sets of ten finished-product thrombin vials were used to develop separate calibrations for moisture and potency. Each vial was scanned non-destructively through the sealed serum vial. After spectral acquisition, the same vials were destructively assayed by standard primary methods: Karl Fischer titration for moisture and a plasma‑titration light‑scattering assay for potency. Spectral and reference data were combined in TQ Analyst chemometric software to build predictive models.

Key analytical parameters and spectral processing steps:

• Instrumental acquisition: Antaris FT‑NIR, scan range 4000–10000 cm-1, 32 co‑averaged scans per sample, 4.0 cm-1 resolution; typical analysis time ~20 s per vial.
• Autosampler: Autosampler RS used for unattended, reproducible vial positioning.
• Moisture model: Stepwise Multiple Linear Regression (SMLR) on second‑derivative spectra with a Norris smoothing filter (9,2); spectral region centered on the first water overtone (~7000 cm-1).
• Potency model: Partial Least Squares (PLS) applied to second‑derivative spectra with Norris smoothing (9,5) and Multiplicative Scatter Correction (MSC) for pathlength compensation; spectral region 6000–6800 cm-1 (outside primary water bands).
• Validation approach: Leave‑one‑out cross‑validation to assess model robustness and identify needs for additional standards.

Used instrumentation

• Thermo Fisher Antaris FT‑NIR spectrometer
• Autosampler RS (for automated vial handling)
• TQ Analyst chemometric software (model building: SMLR, PLS, Discriminant Analysis)
• Reference laboratory methods: Karl Fischer titration (moisture) and plasma titration/light scattering (potency)

Main results and discussion

Moisture:

• Calibration dataset moisture range: ~0.5–0.8% (w/w).
• Calibration performance: correlation coefficient (R) = 0.998, RMSEC = 0.005%.
• Cross‑validation: R = 0.984, RMSECV = 0.018% — indicating a stable model though with slightly increased error on withheld samples.

Potency:

• Calibration potency range: approximately 29,000–33,000 activity units.
• Calibration performance: R = 0.999, RMSEC = 21.9 (activity units).
• Prediction residuals for potency were typically ±0.09% relative, with one outlier residual of −0.18%; cross‑validation residuals increased to approximately ±2% for some withheld standards, partly due to sparse representation at the extremes.

Cake settling and sample morphology:

• Principal Component Analysis / Discriminant Analysis clearly separated intact cakes from samples mechanically converted to a settled/powder morphology, demonstrating detectability of physical form changes by NIR.
• Spectral differences between intact and settled cakes were largely attributable to scattering (baseline offsets) and some absorbance changes; these effects can be mitigated by applying scatter‑correction pretreatments (e.g., MSC) or by including representative settled‑cake spectra in the calibration dataset.

Practical considerations:

• FT‑NIR is a secondary method: model accuracy depends on the quality and representativeness of the primary reference assays.
• Model robustness benefits from a sufficiently wide and well‑sampled calibration range and inclusion of physical variability (e.g., settled cakes) likely to be encountered in routine production and distribution.

Benefits and practical applications of the method

• Non‑destructive, vial‑through measurement enables 100% inspection or expanded sampling of lyophilized lots without consuming product.
• Rapid analysis (≈20 s per vial) greatly increases throughput relative to destructive assays such as Karl Fischer titration or potency titrations and reduces operator time and consumable costs.
• Simultaneous prediction of multiple quality attributes from a single spectrum simplifies QC workflows.
• Automated sampling (Autosampler RS) supports routine implementation and reduces operator variability.

Limitations

• As an indirect technique, FT‑NIR requires robust primary methods for calibration; errors in reference assays propagate to the NIR model.
• Calibration applicability is limited to the compositional and physical space represented by the standards; unusual formulations or severe cake damage may require model updates.
• Spectral effects from vial glass, headspace, or variable cake morphology must be anticipated and handled in preprocessing or calibration design.

Future trends and potential applications

• Expansion of NIR‑based QC to in‑line or at‑line PAT (process analytical technology) for real‑time lyophilization monitoring (moisture endpoint detection, uniformity checks).
• Use of larger, diverse calibration libraries and transfer protocols (transfer standards, instrument standardization) to enable multi‑site deployment and regulatory acceptance.
• Integration with advanced chemometrics and machine learning approaches to improve robustness against morphological variability and low‑level analytes.
• Combining NIR with imaging or hyperspectral methods to spatially resolve cake heterogeneity and identify localized defects.

Conclusion

The study demonstrates that Antaris FT‑NIR spectroscopy, coupled with appropriate multivariate models, provides a fast, accurate, and non‑destructive approach for monitoring critical quality attributes of lyophilized thrombin—specifically residual moisture and potency—directly through sealed vials. The method reduces reliance on time‑consuming, destructive assays and can detect physical changes such as cake settling; however, careful calibration design and quality reference assays are essential to ensure reliable routine performance.

References

• Hirsch J. Near‑Infrared Analysis of Critical Parameters in Lyophilized Materials. Thermo Fisher Scientific Application Note AN50911, 2007.