Near-Infrared Analysis of Critical Parameters in Lyophilized Materials

Significance of the Topic

Lyophilized products are widely used in pharmaceutical and biotechnological manufacturing to extend shelf life and enable ambient storage of labile biomolecules. Reliable, non-destructive quality control of lyophilized cakes inside sealed vials is a critical operational need: conventional techniques (Karl Fischer titration for moisture, activity assays for potency) are destructive, slow, consumable-intensive, and provide only batch-level information. Near-infrared (NIR) spectroscopy combined with multivariate analysis offers rapid, through-container, non-destructive measurement of multiple critical quality attributes from a single spectrum, enabling more frequent or 100% inspection and improving process and product control.

Objectives and Overview of the Study

The study evaluated the performance of the Thermo Fisher Antaris Fourier Transform Near-Infrared (FT-NIR) spectrometer for non-destructive determination of two critical quality attributes of lyophilized thrombin: residual moisture and biological potency. Additional objectives were to demonstrate rapid measurement through sealed serum vials, to develop chemometric models (using TQ Analyst software) correlating spectra with reference destructive assays, and to assess whether morphological changes in lyophilized cakes (settling) can be spectrally discriminated and accommodated in calibrations.

Methodology

Experimental design and data acquisition:

• Samples: Two sets of ten finished-product thrombin vials were used—one set for moisture calibration and one for potency calibration. Moisture reference values were obtained by Karl Fischer titration; potency references by colloidal light-scattering plasma titration.
• Instrument and acquisition: Antaris FT-NIR analyzer with Autosampler RS. Spectra acquired through sealed vials across 4000–10000 cm⁻¹. Each sample: 32 co‑averaged scans at 4 cm⁻¹ resolution; analysis time ≈ 20 s per vial.
• Chemometrics: TQ Analyst software. Moisture calibration used Stepwise Multiple Linear Regression (SMLR) on the second-derivative spectra (Norris smoothing 9,2) focused near 7000 cm⁻¹ (first water overtone). Potency calibration used Partial Least Squares (PLS) on second-derivative spectra (Norris smoothing 9,5) with Multiplicative Scatter Correction (MSC) and spectral region 6000–6800 cm⁻¹.

Instrumentation Used

• Thermo Fisher Antaris FT-NIR spectrometer.
• Autosampler RS accessory for unattended sample throughput.
• TQ Analyst chemometric software for model development (SMLR, PLS, discriminant/PCA).

Main Results and Discussion

Moisture:

• Calibration model built from ten standards with moisture values ≈0.5–0.8% (Karl Fischer reference).
• Model statistics: correlation coefficient (R) = 0.998, RMSEC = 0.005% (excellent fit).
• Cross-validation: R = 0.984, RMSECV = 0.018% — indicates somewhat increased error on leave‑one‑out validation but still acceptable for low-level moisture screening.

Potency:

• Potency reference range ≈29,000–33,000 (units as per light-scattering assay).
• PLS calibration statistics: R = 0.999, RMSEC = 21.9 (potency units).
• Prediction residuals typically ±0.09% for most samples though one outlier had −0.18%; cross-validation residuals increased to ≈±2.0%, with larger errors at underrepresented extremes of the calibration range.
• PRESS behavior consistent with a reasonable model (initially high, trends to a minimum, then flattens/increases).

Sample morphology / cake settling:

• Principal component analysis and discriminant analysis readily separated intact lyophilized cakes from vials deliberately shaken to produce settled/powdery cakes, indicating measurable spectral differences (largely scattering and baseline offsets).
• Including representative settled-cake spectra in calibrations or applying pathlength/scatter compensation (e.g., MSC, derivative pretreatments) can mitigate bias introduced by physical variability.

Method considerations:

• NIR operates as a secondary (indirect) technique: its accuracy is limited by the quality and representativeness of the primary reference assays used for calibration.
• Minimal spectral pretreatment was sufficient for moisture because water overtone bands are strong; potency required more elaborate pretreatment (derivative + smoothing + MSC).

Benefits and Practical Applications

• Non-destructive, through‑vial measurement enables 100% inspection or more representative lot screening without consuming product.
• Rapid analysis (≈15–30 s per vial) reduces turnaround compared with destructive methods (Karl Fischer titration: 30–60 min per sample) and lowers labor and consumable costs.
• Single-spectrum multivariate models can simultaneously predict multiple attributes (moisture and potency) and also flag physical changes such as cake settling.
• Automated sampling (Autosampler RS) permits high-throughput, operator-independent workflows suitable for QA/QC environments.

Future Trends and Potential Uses

• Broader adoption of through-container NIR for release testing and stability monitoring of lyophilized biologics could enable more frequent in-process and end-of-line checks and reduce batch destruction.
• Refinement of calibration strategies: expanding calibration sets to cover manufacturing and shipping variability (e.g., different cake morphologies, vial types, fill volumes) to improve robustness and reduce prediction bias.
• Hybrid models and transfer learning approaches to accelerate model deployment across product lines and instruments, and to compensate for inter-batch and inter-instrument variability.
• Integration with process analytical technology (PAT) frameworks and real-time stability monitoring for predictive shelf-life estimation.

Conclusion

The Antaris FT-NIR system, combined with appropriate chemometric modeling, provided accurate, rapid, and non-destructive predictions of residual moisture and potency in lyophilized thrombin vials. High correlation and low calibration errors demonstrate the method’s suitability for routine QC screening, while PCA/discriminant analysis permits detection and compensation of physical sample variability such as cake settling. Limitations inherent to secondary calibrations (dependence on primary reference accuracy and calibration representativeness) should be addressed by rigorous reference assays and inclusion of variability in the calibration set.

Reference

• Hirsch J. Near-Infrared Analysis of Critical Parameters in Lyophilized Materials. Application Note 50911. Thermo Fisher Scientific; 2007.