Monitoring and Controlling Powder Blending Online at AstraZeneca

Significance of the Topic

Near-infrared (NIR) spectroscopic monitoring of powder blending addresses a persistent quality risk in solid-dosage pharmaceutical manufacture: non-uniform blending. Traditional thief sampling is labor-intensive, potentially hazardous, and can introduce sampling bias and additional inhomogeneity. Implementing robust on-line spectral monitoring can reduce reliance on intrusive sampling, improve reproducibility of blend uniformity assessment, accelerate decision making, and support PAT (Process Analytical Technology) initiatives required by regulatory guidance.

Objectives and Study Overview

This study evaluated a MEMS-based NIR spectrometer as an on-line blend-monitoring tool to replace or augment thief sampling. The goals were to demonstrate the instrument’s capability to capture representative spectra of active and excipient components during blending, to define spectral metrics that indicate blend endpoint, and to explore process factors (fill level, scale, component variability) that affect blending time and spectral homogeneity.

Methodology

• Model formulation: acetaminophen (API) blended with microcrystalline cellulose, spray-dried lactose monohydrate, crospovidone and magnesium stearate.
• Blender: lab-scale 20 L Bohle bin blender equipped with a modified lid containing a sapphire window for spectral sampling at the bottom of rotation.
• Data acquisition: the MEMS spectrometer recorded five scans per rotation; these were averaged to yield one spectrum per rotation. Data collection was triggered by a MEMS accelerometer at a predefined blender angle and required ~500 ms per rotation.
• Spectral preprocessing: second-derivative transformation was applied to minimize baseline effects and enhance subtle features associated with components.
• Endpoint algorithm: a moving block standard deviation approach was used. For a selected spectral region and a block size (seven rotations in the study), the standard deviation across wavelengths was computed for each block and summed to produce a single metric per block. Plotting that summed spectral standard deviation versus rotation count provided blend curves for endpoint determination.

Instrumentation Used

• Device: Thermo Electron Antaris Target Series Blend Monitor (MEMS-based NIR spectrometer).
• Spectral range: 1350–1800 nm.
• Optical components: semiconductor NIR tunable laser source; Fabry–Pérot tunable filter providing high resolution (specified options 4 or 8 cm–1); single-element InGaAs photodiode detector.
• Mechanical/reliability features: hermetically sealed spectrometer bench under dry nitrogen; internal reference for wavelength and absorbance reproducibility; no moving parts; battery-powered; MEMS accelerometer for synchronization with blender position; sapphire sampling window integrated into blender lid or vessel; spot size ~40 mm corresponding to ~600 mg dose equivalence; instrument insensitive to vibration and position.

Key Results and Discussion

• Raw and second-derivative spectra of the individual excipients and the API showed distinct features and combined to form the blended spectrum, enabling chemical sensitivity to both active and excipient contributions.
• Spectral variation was large at the start of blending and decreased substantially after approximately 15 rotations for the tested formulation and conditions, indicating convergence toward homogeneity as measured spectrally.
• Fill level markedly affected blending time: a 90% filled bin required roughly three times more rotations to reach the same spectral-homogeneity metric as a 60% filled bin, demonstrating that time- or rotation-based control without real-time measurement can be misleading.
• Other process variables identified as influential include blender size, blender speed, active concentration, loading technique, and excipient lot-to-lot variability (bulk density, moisture, surface area). These factors can change the number of rotations required to reach the spectral endpoint and underscore the value of in-line measurement for scale-up and production changes.
• The moving block summed standard deviation metric provided an intuitive, quantitative blend curve that converts spectral data into an operational endpoint signal suitable for process control.

Benefits and Practical Applications

• Non-contact, nondestructive, rapid and reproducible measurements performed on-line: enables real-time feedback without manual sampling and associated safety/exposure concerns.
• Reduction or elimination of thief sampling: minimizes sampling biases, reduces operator labor and exposure risk, and improves traceability and consistency of blend assessments.
• Process optimization and control: spectral endpoints allow adaptive control of blending duration across varying fill levels, lot changes and scale-up scenarios, potentially reducing over-processing or under-blending.
• Downstream quality assurance: once blend uniformity is confirmed spectrally, the approach can be extended to monitor the material through subsequent steps (e.g., feed to compression and content uniformity of uncoated tablets), which may reduce the need for some off-line QC tests.

Future Trends and Potential Applications

• Advanced algorithms and chemometrics: development and validation of more sophisticated endpoint-detection algorithms (multivariate control charts, PCA, PLS, change-point detection) will improve sensitivity, robustness and specificity for different formulations and process disturbances.
• Integration into PAT and manufacturing execution systems: wireless, battery-powered MEMS spectrometers with embedded accelerometers can be more easily deployed across lab, pilot and production equipment, enabling continuous data capture and automated decision-making linked to process control systems.
• Extended process monitoring: spectral tracking from raw material blending through compression and coating can allow end-to-end verification of content uniformity and batch release strategies with fewer destructive laboratory assays.
• Scalability and modular sensors: MEMS-based units may be positioned at multiple locations (inlets, outlets, transfer lines), permitting spatially resolved monitoring and improved root-cause analysis for blend heterogeneity.
• Regulatory alignment: validated in-line NIR approaches that demonstrate equivalence or superiority to traditional sampling will support PAT adoption and potentially streamline regulatory submissions focused on process understanding and control.

Conclusion

The study shows that a MEMS-based NIR spectrometer can reliably monitor powder blends on-line, translating spectral variability into actionable blend endpoints. Key advantages include rapid, noninvasive acquisition, immunity to vibration, and operational flexibility for lab-to-production deployment. Spectral endpoint metrics reveal sensitivity to fill level and other process variables that time-based controls miss. With continued development of algorithms and integration strategies, MEMS NIR monitoring has strong potential to reduce or replace thief sampling and to support more efficient, data-driven blend control and product quality assurance.

References

1. FDA. Current Good Manufacturing Practice: Amendment of Certain Requirements for Finished Pharmaceuticals; Proposed Rule (61 FR 20103), May 1996.
2. Boehm G., Clark J., Dietrick J., Foust L., Garcia T., Gavini M., Gelber L., Goeffroy J.M., Jimenez P., Mergen G., Muzzio F., Planchard J., Prescott J., Timmermans J., Takiar N. The Use of Stratified Sampling of Blend and Dosage Units to Demonstrate Adequacy of Mix for Powder Blends. PDA J Pharm Sci Tech 57:59–74, 2003.
3. FDA. Guidance for Industry: Powder Blends and Finished Dosage Units – Stratified In-Process Dosage Unit Sampling and Assessment, October 2003.
4. Hwang R.-C., Wu S.-J. Challenges of Blend Uniformity Testing for Tablet Formulation. American Pharmaceutical Review 7:101–103, Jan/Feb 2004.
5. FDA. Guidance for Industry: PAT – A Framework for Innovative Pharmaceutical Development, Manufacturing and Quality Assurance, September 2004.
6. Sekulic S., Ward H., Brannegan D., Stanley E., Evans C., Sciavolino S., Hailey P., Aldridge P. On-Line Monitoring of Powder Blend Homogeneity by Near-Infrared Spectroscopy. Analytical Chemistry 68:509–513, Feb 1996.
7. Berntsson O., Danielsson L.-G., Lagerholm B., Folestad S. Quantitative In-line Monitoring of Powder Blending by Near Infrared Reflection Spectroscopy. Powder Technology 123:185–193, 2002.
8. Cogdill R., Anderson C., Delgado-Lopez M., Molseed D., Chisholm R., Bolton R., Herkert T., Afnan A., Drennen III J. Process Analytical Technology Case Study Part I: Feasibility Studies for Quantitative Near-Infrared Method Development. AAPS PharmSciTech 6(2):E262–E272, 2005.
9. Crocombe R. MEMS Technology Moves Process Spectroscopy into a New Dimension. Spectroscopy Europe, July 2004.
10. Parris J., Airiau C., Escott R., Rydzak J., Crocombe R. Monitoring API Drying Operations with NIR. Spectroscopy 20(2):34–42, Feb 2005.
11. Sullivan M. The Use of NIR as a PAT Tool for Measuring Blend Uniformity. Spectroscopy Supplement: Process Analytical Technologies, Feb 2006.
12. Sekulic S.S., Wakeman J., Doherty P., Hailey P.A. Automated System for the On-line Monitoring of Powder Blend Processes Using Near-Infrared Spectroscopy Part II: Qualitative Approaches to Blend Evaluation. J Pharm Biomed Anal 17:1285–1309, 1998.