Application Note

Real-time blending optimization with NIR technology

VIAVI Solutions

Blend uniformity testing has traditionally been a laborious  and time- consuming process. NIR spectroscopy can monitor the blending process in real time to determine the optimum end-point, thereby reducing effort, time, and cost while improving process understanding.

Blend uniformity testing in the pharmaceutical industry is traditionally carried out by stopping the blender after a given time, commonly defined by existing procedures. The operator then withdraws 10 samples at predefined locations specified by the current International Conference on Harmonization (ICH) guidelines. Samples are then analyzed offline in the QC/QA laboratory

by means of HPLC, UV-Vis and/or other laboratory techniques. The approach has several disadvantages:

•      The blender must be stopped to withdraw samples. Continuous monitoring is not thereby possible

•      Thief sampling causes perturbation of the blend process

•      Laboratory characterization is labor intensive and time-consuming

•      Operator exposure to API and solvents may require extra safety measures

•      The blend is exposed to the environment

Thief sampling is the most significant issue on this list. In fact, introducing a thief and withdrawing samples causes perturbations in the blend, creating segregation in the powder beds. This is well explained in a recent

article1 by Kim Esbensen, Rodolfo Romañach and coworkers.

Installing a near infrared (NIR) sensor online can be an effective solution to overcome the drawbacks listed above. The MicroNIR™ PAT-W from Viavi Solutions monitors tumble blending continuously in real time,  with no need to stop the blender for sample withdrawal and further laboratory analysis. In the pharmaceutical industry, the direct monitoring approach based on

PAT (Process Analytical Technology) enables the implementation of QbD (Quality by Design), which is now encouraged by the FDA2. In other industries, the ”Real-time blend monitoring is a fundamental step in the transition from process validation to continuous process verification, in which every step of every batch is controlled during the production phase.”

1     Esbensen et al., International Journal of Pharmaceutics (2016) 499, 156 DOI: 10.1016/j.ijpharm.2015.12.038

2     PAT - A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance https://www.fda.gov/regulatory-information/ search-fda-guidance-documents/pat-framework-innovative-pharmaceutical-development-manufacturing-and-quality-assurance

direct monitoring of the blending process represents a significant advance toward LEAN manufacturing.

This note describes how the MicroNIR PAT-W can be integrated into a commercial manufacturing environment to optimize the endpoint of a blend.

System Configuration

Setting up the MicroNIR PAT-W to monitor powder blending is quick and simple:

•      Mount the NIR sensor on the blender lid equipped a standard triclover clamp

•      Connect the instrument to the computer through a dedicated Wi-Fi connection

•      Adjust the acquisition parameters so that data are collected exclusively when the blend covers the instrument’s sapphire process window. The PAT-W includes an onboard accelerometer used to trigger data acquisition.

With some minor additional software settings, the system is ready to display the progress of blend uniformity, allowing a direct and non-invasive interpretation of the dynamics occurring in the blender. Typically, 50-100 acquisitions during a single rotation, requiring a total time of ~450-900 ms, are

averaged into a single spectrum. A single spectrum thus represents a single blender rotation.

The MicroNIR PAT-W is IP 65/67 rated and is completely isolated from the blend by a sapphire window, allowing non-contact operation. When the blending process is complete, CIP (Cleaning in Place) and SIP (Sterilization  in Place, if required) can be accomplished without disassembly. Materials used in the PAT-W are FDA approved.

From real-time data to executable information

Unlike most NIR applications, blend uniformity can be monitored with no need for chemometric data modeling. The time-consuming acquisition of reference standard spectra and PLS regression model building are not required. Starting from the initial loading of pure ingredients, blend uniformity is monitored throughout the blending cycle by plotting the standard deviation of the acquired spectra over time. This provides an easily interpreted numerical value that reflects the degree

of uniformity of the mixture: as the blend becomes increasingly uniform, successive spectra become increasingly similar.

The data processing, based on the calculation of standard deviation as an index of similarity, is called Moving Block Standard Deviation (MBSD) (Figure 1). At the beginning of the blending cycle the materials are well separated. Spectral variability from one rotation to the next and the resulting block standard deviation are high. As the material becomes well blended the MBSD decreases to a low, nearly constant level. Blending is complete when the calculated value falls below a user-defined threshold and remains consistent over a meaningful number of blender revolutions.

The threshold may be manually set or mathematically defined by Fisher statistics at a 95% confidence interval. Data acquisition and analysis are accomplished by Viavi’s MicroNIR Pro software, which is compliant with Title 21 CFR Part 11, USP chapter 1119 and E.P. chapter 2.2.40.

Figure 1: How the MBSD algorithm works

Case study

The product used in this application consisted of an active ingredient and four excipients. The standard blending time for the chosen process was 20 minutes, or approximately 450 rotations. After the endpoint was reached, the last steps of the process included the addition of a lubricant, followed by a final blending

time of 1 minute. A MicroNIR PAT-W was installed on theblender lid (Figure 2) equipped with a welded sanitary flange and sapphire process window.

The built-in rechargeable Li ion battery of the MicroNIR PAT-W allows >8 hours of continuous operation and the dedicated wireless channel addresses network security concerns.

Figure 2: Interface of the MicroNIR PAT-W into the lid of a bin blender

The MicroNIR PAT-W uses Viavi’s Linear Variable Filter (LVF) technology and has no moving parts, free space optics, or optical fiber couplings, permitting the instrument to deliver highly reproducible spectral data at a rapid rate. Dual onboard tungsten lamps provide stable operation with an extremely long lamp life.

Acquisition parameters used to collect data are reported in Table 1.

Table 1: Acquisition parameters

Delay time refers to the point in rotation when scanning starts, corresponding to the time when the blend

first covers the instrument window. The NIR spectral profiles of the entire process are shown in Figure 3a. Visibly different spectra reflect the inhomogeneity of the powder mix at the beginning of the cycle, clearly showing that blend uniformity is embodied within the collected data.

Figure 3. a) Spectra acquired during blending. b) Moving Block Standard Deviation of spectra plotted over time.

In the present case study, the MBSD profile of Figure 3b indicated that the process endpoint occurred after 150 rotations, while the traditional approach standard procedure was set to 450 rotations with subsequent laboratory analysis. Once validated, the PAT approach greatly reduces the need for lab analysis and increases uptime and product throughput. In many cases, NIR monitoring can also reveal overblending, resegregation and the production of fines.

Advanced approach

The ability to monitor the homogeneity of a blend with no need for data modeling and prior analytical method development applies to a wide range of formulations, regardless of the chemical composition, number, or nature of the ingredients. When complex blending dynamics need to be understood, more advanced data analysis tools can be applied. Mixtures can be monitored by narrowing the spectral region to the unique signatures of specific ingredients while simultaneously displaying the profile of multiple constituents. Alternatively, data reduction with PCA (Principal Component Analysis) prior to performing the MBSD computation allows the monitoring of ingredients characterized by chemical similarities or by particle size inhomogeneity among the mixture constituents, again with no need for chemometric modeling.

Conclusions

In summary, blend uniformity continuous monitoring by means of the MicroNIR PAT-W is a critical step toward the implementation of QbD in pharmaceutical manufacturing and/or LEAN manufacturing in any industry and can be successfully used to:

•      Enable process monitoring from laboratory to production through a scalable path

•      Acquire actionable information within the process experimental domain

•      Avoid overprocessing, such as production of fines and material segregation

•      Reduce blending time and increase manufacturing productivity

The MicroNIR PAT-W is designed to suit multiple applications, including small-batch pilot laboratory projects to identify the optimal blending time of new formulations, scale-up on the production floor, and OPC automated process controller integration. The compact, lightweight, and robust MicroNIR PAT-W is IP 65/67 compliant and designed for long service life and low cost of ownership.

Figure 4. MicroNIR PAT-W Instrument