Investigation of Different Sampling Techniques for the Analysis of Capsule Contents by Fourier Transform Near-Infrared Spectroscopy

Investigation of Different Sampling Techniques for Capsule Analysis by FT-NIR — Summary

Significance of the topic

Capsules are a major solid-dosage form in pharmaceuticals and present analytical challenges because of the gelatin shell and demanding sample-preparation steps required for conventional assays. Fourier transform near-infrared (FT-NIR) spectroscopy offers rapid, precise, and low-preparation alternatives that can be deployed outside central laboratories. Evaluating sampling strategies (non‑destructive whole‑capsule reflectance versus destructive powder reflectance) is therefore important to establish practical FT‑NIR workflows for qualitative identity testing, content uniformity screening and quantitative assay of low‑level actives in capsules.

Objectives and overview of the study

• Assess FT‑NIR (Thermo Scientific Antaris MDS) viability for qualitative and quantitative analysis of conventional hard gelatin capsules using minimal or no sample preparation.
• Compare two sampling approaches: intact whole‑capsule reflectance and reflectance of emptied powder fills (measured in micro sample cups or small vials through glass).
• Test three sample sets representing progressively demanding use cases: qualitative discrimination of four formulations with the same active (sample set 1), quantitative distinction of clinical supplies (sample set 2), and assay calibration for a marketed formulation at low active loading (~5 mg/capsule, sample set 3).

Methodology

• Samples: Standard hard gelatin capsules (powder fills). Three datasets: set 1 (four formulations: 300 mg, 120 mg, 120 mg SR, 60 mg SR), set 2 (four clinical formulations suitable for quantitative models), set 3 (seven formulation levels spanning ~90–110% of a 5 mg nominal label claim to develop an assay calibration).
• Sampling approaches: non‑destructive whole‑capsule reflectance (capsules placed on integrating sphere window) and destructive powder reflectance (powder emptied into micro sample cups with glass window or small vials and measured through glass).
• Spectral region and processing: full near‑IR region (~4000–10,000 cm−1); common pre‑treatments included Norris second derivative (segment 11, gap 0) where beneficial. Data processing used discriminant analysis (DA) for qualitative class distinction and partial least squares (PLS) regression for quantitative models. Model quality assessed by correlation coefficient, RMSEC and RMSECV (errors in mg/capsule) and Mahalanobis distance for DA.

Instrumentation used

• Thermo Scientific Antaris Method Development Sampling (MDS) FT‑NIR analyzer equipped with an integrating sphere for reflectance measurements (also includes transmission compartment, fiber‑optic probe and tablet transmission accessory).
• RESULT software for data acquisition and TQ Analyst for chemometric model building; ValPro system qualification for instrument verification.
• Internal gold flag used as background reference to ensure stable, featureless background.

Results and discussion

• Sample set 1 (qualitative discrimination): Both whole‑capsule reflectance and powder reflectance provided spectra sufficient to distinguish the four dosage forms. A DA model built on the full spectral range reliably classified all calibration capsules with no mismatches. Mahalanobis distance values for the best class were well below the 3.0 threshold, and next‑best class distances were substantially larger, indicating clear separability without derivative preprocessing in the primary analysis (derivative spectra helped visualize differences).
• Sample set 2 (quantitative distinction of clinical supplies): Both sampling modes yielded viable PLS calibrations. Powder‑based models produced slightly better cross‑validation accuracy. Representative metrics: whole‑capsule PLS (3 factors) produced corr. coeff. ≈ 0.9991, RMSEC ≈ 0.143 mg/capsule, RMSECV ≈ 0.742 mg/capsule; powder PLS (2 factors) produced corr. coeff. ≈ 0.9993, RMSEC ≈ 0.126 mg/capsule, RMSECV ≈ 0.140 mg/capsule. The lower RMSECV for powder measurements indicates improved robustness and predictive performance.
• Sample set 3 (assay for low‑level active ~5 mg/capsule): Powder analysis delivered substantially better quantitative performance than whole‑capsule reflectance. Powder PLS (2 factors) achieved corr. coeff. ≈ 0.9960, RMSEC ≈ 0.0274 mg/capsule, RMSECV ≈ 0.151 mg/capsule. Whole‑capsule PLS (2 factors) was markedly weaker (corr. coeff. ≈ 0.8686, RMSEC ≈ 0.152 mg/capsule, RMSECV ≈ 0.302 mg/capsule). This case is challenging because the active comprises only ~4% of the formulation; spectral contributions from the gelatin shell and excipients reduce sensitivity in non‑destructive measurements.
• General findings: Non‑destructive whole‑capsule reflectance is attractive for rapid identity checks and cases where preserving capsule integrity is required. However, the capsule shell introduces spectral interference that commonly reduces quantitative accuracy. Destructive powder reflectance (no solvent extraction) improves signal quality and model performance, particularly for low‑level actives. All analyses are rapid (analysis time < 1 minute) and avoid solvent handling and disposal.

Benefits and practical applications

• FT‑NIR enables fast, precise screening and assay of capsules with minimal sample preparation and no solvents, offering major throughput and safety advantages over HPLC for many routine tasks.
• Whole‑capsule reflectance is suitable for non‑destructive identity verification, in‑process checks and situations where sample integrity matters (e.g., clinical supplies), reducing handling time and risk to operators.
• Powder reflectance provides superior quantitative accuracy for content uniformity and low‑level assay tasks and remains rapid and solvent‑free compared with traditional wet chemistry approaches.

Future trends and potential applications

• Improved chemometric methods and larger calibration sets will further close the accuracy gap between non‑destructive and destructive approaches, especially for low‑dose actives.
• Advances in spectral pretreatment, variable selection and machine‑learning algorithms could enhance sensitivity to minor active components in whole‑capsule spectra despite shell interference.
• Integration of FT‑NIR into PAT (process analytical technology) frameworks and at‑line/portable configurations will expand real‑time quality control for capsule manufacturing and clinical sample handling.
• Miniaturized sampling accessories and optimized sample cups/vials may increase powder‑based measurement reproducibility and success for small sample masses.

Conclusion

FT‑NIR (Antaris MDS) is a practical tool for both qualitative and quantitative capsule analysis. Non‑destructive whole‑capsule reflectance works well for many classification tasks and offers logistical benefits, while destructive powder reflectance generally yields better quantitative accuracy, especially when actives are present at low weight fractions. Both approaches provide rapid, solvent‑free analysis and can significantly streamline routine pharmaceutical testing when appropriate calibrations and sampling strategies are applied.

References

• Thermo Fisher Scientific. Application Note 51595: Investigation of Different Sampling Techniques for the Analysis of Capsule Contents by Fourier Transform Near‑Infrared Spectroscopy. AN51595_E, 04/2008.