Method development guidance for nasal spray analysis using Morphologically-Directed Raman Spectroscopy (MDRS)

Introduction

The Morphologi 4-ID uses Morphologically-Directed Raman Spectroscopy to measure the particle size distribution (PSD) of the active pharmaceutical ingredient (API) in nasal spray formulations. The particle size and shape information is provided by automated image analysis and chemical identification using Raman spectroscopy.

This technical note provides guidance on method development for component-specific particle size and shape analysis of nasal spray samples using the Morphologi 4-ID. 

Optimization of the method requires consideration of the following:

• Sample preparation
• Microscopic assessment (morphological analysis)
• Classification for chemical analysis (selection of particles for targeting)
• Chemical analysis and API classification
• Verification by orthogonal techniques

Each step is described in more detail below.

Sample preparation

Appropriate sample preparation is key to all analytical methods. Current guidance¹ recommends measuring nasal sprays as wet suspensions to make minimal changes to the state of the sample. Wet suspension samples containing fine particles are presented between a slide and a coverslip. To prepare the sample it is necessary to:

• Shake and prime the nasal spray
  • The effect of shaking and priming on the actuation of a nasal spray may well have been investigated as part of droplet size measurements (there is more information about droplet size testing in our application note²)
  • Shaking and priming should be consistent and documented
• Actuate the device several times into a beaker 
• Using a cleaned quartz slide, use a micropipette to place an appropriate volume of the dispensed nasal spray onto the slide (Figure 1). The volume of liquid required will depend on the viscosity of the suspension (the lower the viscosity, the lower the volume)
  • A typical volume is 4μl
  • The volume used on the quartz slide should be validated by measuring a range of volumes (e.g., +/- 1μl)

Figure 1 (left) & 2 (right): Left: Dispensing nasal spray sample onto a quartz slide. Right: Using vacuum tweezers to cover the sample with a quartz coverslip

• Use vacuum tweezers to place a cleaned quartz coverslip over the sample (Figure 2)
  • Avoid forming bubbles
• Apply pressure using the vacuum tweezers to encourage the formulation to spread out and fill the area under the coverslip (and minimize bubbles)
  • Avoid excess sample around the edges of the coverslip
• Seal the edges of the coverslip with silicon grease, mineral oil, nail varnish or similar, to prevent evaporation (Figure 3)

Figure 3: Sealing the edges of the coverslip to prevent evaporation

• Allow the sample to settle until no particle movement is observed
  • Typically, this takes between 15 minutes and 1 hour
  • Evidence that the settling time has been optimized will be required for validation purposes

Figure 4: Field of view of nasal spray dispersion

From the field-of-view images of the dispersed nasal spray the concentration level of particles can be assessed (Figure 4). Ideally, the sample concentration should not be too high, such that there are many touching particles which cannot be distinguished from aggregates, or too low, such that it is impractical to measure enough particles. If the concentration is too high, try reducing the sample volume further, or consider if dilution is possible, which is dependent on the solubility of the particles in the dispersing medium. If the concentration is too low, then the approach to the morphological analysis will have to be adjusted.

Analysis of the formulation pre- and post-spray may be requested. Analysis of the sample pre-spraying follows the same sample preparation process, but the sample is taken directly from the bottle.

Morphological analysis

The aim of nasal spray analysis is usually to measure the particle size distribution of the active particles; hence, the morphological analysis settings are tailored to analyzing the active particles. 

The particle size of the active particles is expected to be within the range of the 50x magnification objective lens (0.5μm to 50μm). The limit of optical microscopy prevents particles smaller than 0.5μm from being measured on the Morphologi. It is generally considered impractical to carry out routine MDRS on particles in a nasal spray smaller than 1μm due to increased acquisition times and the potential for particle movement under Brownian motion. 

Table 1: Recommended SOP settings

The settings in Table 1 can be used to run an initial morphological measurement of a nasal spray. Morphological filters, such as solidity and convexity, are then applied to the data to remove touching particles. Typical values for the solidity and convexity filters are between 0.8 and 0.9. For validation purposes, images of the excluded particles should be collected along with the number of particles excluded by the filters (see Figure 5 to Figure 8). The particle size distribution with and without the filters applied should also be assessed. 

Figure 5: Number-based CE diameter distribution with different solidity filters

Figure 6: Volume-based CE diameter distribution with different solidity filters

Figure 7: Particles excluded by the 0.85 solidity filter

Figure 8: Particles excluded by the 0.9 solidity filter

The filters applied to remove touching particles should be investigated on both the test and reference product, and if appropriate the same filters can be applied to both materials. 

Filters such as intensity standard deviation may need to be applied to remove bubbles from the analysis. A filter should also be applied to remove particles below the minimum size to be targeted for chemical analysis (generally 1μm). Examples of filtered morphological results are shown in Figure 9 and Figure 10.

Figure 9: Number-based CE diameter distribution with all filters applied before chemical analysis

Figure 10: Volume-based CE diameter distribution with all filters applied before chemical analysis

The number of particles in the morphological result should be at least 10,000 for statistical significance³. In addition, the repeatability, reproducibility, and robustness of the morphological measurement could be assessed at this point. It may also be useful to assess the batch-to-batch variability using the morphological analysis only.

Classifying particles for chemical analysis

MDRS enables the results of the morphological analysis to be used to classify particles according to their size and shape to then target the chemical analysis at particles which are more likely to be active. The morphological filters used to classify particles as active should be selected to ensure that a minimal number of active particles are excluded from further analysis. If active particles are excluded, then this should not have a significant effect on the particle size distribution of the active particles.

There is more than one approach to classifying the particles for chemical analysis:

10,000 particles

In this approach to method development, 10,000 particles in a nasal spray are measured by MDRS, with filters applied only to remove touching particles and particles outside of the size range of analysis. Analyzing this number of particles by MDRS is best achieved by combining multiple smaller measurements. A typical overnight measurement can analyze ~2000 particles, multiple overnight runs could then be combined to create a single MDRS result containing 10,000 particles. 

The particles in this result can then be classified based on their chemical correlation score to the reference active material (setting up a chemical library is discussed later). Once classified, the morphological properties of the group of particles identified as active can then be used to target future MDRS analysis at particles more likely to be active⁴ and reducing the overall number of particles targeted in routine testing. This will enable a routine MDRS measurement of between one and two thousand particles to measure a greater proportion of active particles and provide good statistical significance for the particle size distribution of the active ingredient. Any filters will need to be verified for both test and reference and a 10,000 particle measurement may be needed for both.

Placebo

Morphological targeting may also be achieved by measuring a placebo formulation (a formulation containing only excipient and carrier) in addition to the formulation containing the active. This should identify which morphological parameters can be used to differentiate between the active and placebo formulation, and hence the morphological parameters that can be used to target the Raman analysis mainly at active particles. Figure 11 shows an illustration of the overlay of the elongation and High Sensitivity (HS) circularity distributions for a typical active ingredient and placebo.

Figure 11: Illustration of the comparison of elongation and HS circularity distributions for samples containing API and Placebo

Shape parameters such as elongation and HS circularity can be used to target the chemical analysis towards particles more likely to be active. This can be achieved either by creating a class of particles that appear to be active or filtering out particles that are unlikely to be active. The effect of these filters on the resulting particle size distribution should be investigated. Figure 12 shows the effect of varying the elongation filter between 0.3 and 0.5 on the particle size distribution. Looking at images of the particles excluded by the filter is also recommended. 

Figure 12: Impact of elongation filters from 0.3 to 0.5 on the volume-based CE diameter distribution

The Elongation and/or HS circularity descriptors can be used to morphologically classify particles which are likely to be active and to target the chemical measurement more effectively towards active particles.

Chemical Analysis

Chemical identification of particles is achieved by acquiring a Raman spectrum from each targeted particle. These spectra are compared to a chemical library of reference spectra. A correlation score to each component in the library is calculated, where a score close to 1 indicates a good match and a score close to zero indicates no match.

The setup of a chemical measurement is split into four sections:

• Spectrometer settings
• Particle selection
• Chemical library
• Chemical analysis

Spectrometer settings: Exposure time and laser power

The optimum exposure time and laser power are best determined using the Manually Target for Chemical Analysis tool. This enables manual spectral acquisition from a small number of particles during which multiple spectra per particle (e.g. with different exposure times and laser powers) can be over plotted to compare spectral features, illustrated in Figure 13. Laser power is generally only reduced from 100% if there is a danger of damaging the particles. This is unlikely to be the case for nasal sprays so a laser power of 100% is generally used. 

At this stage, it is worth targeting smaller particles as these will often require a longer acquisition time. In most samples, the active ingredient is a much stronger Raman scatterer than the excipients, in which case a decision must be made as to whether the excipients need to be positively identified or if a lack of correlation to the active ingredient is sufficient.

Figure 13: Over plotted API spectra with varying acquisition times

The acquisition time is often in the range of 5s to 60s depending on the minimum particle size to be targeted. 

Particle selection

The section titled classifying particles for chemical analysis describes how morphological parameters are used to focus the chemical analysis on particles which are more likely to be active. At this stage you will have either decided to target only particles that look active, in which case you should select particles from that classification. Otherwise, to choose particles at random from the whole population, you should have no classifications pre-set and select particles from the unclassified group. This would be appropriate to carry out the 10,000 particle training method or if you used filters to exclude particles unlikely to be active. With an exposure time of 30s, a measurement of 1000 particles will take approximately 8 hours. 

Chemical Library

A chemical library is created by acquiring reference spectra from individual pure ingredients. These reference spectra should be acquired at least from the active ingredient, but it is recommended to also acquire spectra from any strongly scattering excipients, as well as the substrate. Spectra from the pure ingredients can be acquired using the manual microscope allowing for the acquisition time to be optimised without the need to carry out morphological analysis. The spectra used in the chemical library should be of high quality and representative of the bulk material; reference spectra can be a mean spectrum created by averaging multiple highlighted spectra. 

Chemical analysis

The chemical analysis section deals with how the measured spectra are processed to enhance spectral features and reduce noise. Correlations scores are calculated using the processed spectra to achieve the best possible correlations.

Figure 14: Schematic of the spectral processing steps

The processing steps (illustrated in Figure 14) can include background subtraction, spectral range selection, and pre-processing (smoothing and differentiation). These processes are described in more detail in a separate technical note⁵. This document provides some suggestions for the analysis of nasal sprays.

Background correction

Background correction removes the signal from the substrate and the continuous phase of suspensions. If the full spectral range is to be used, then a background spectrum should be measured from a region of the sample containing no active particles and used with the substrate background subtraction option to remove features not associated with the API. However, if the spectral features of the active particles are well separated from the background features, then spectral range selection can be used instead.

Spectral range selection

The spectral range over which correlation scores are calculated can be specified. This tool can be used as an alternative to background subtraction, when the main spectral features of the active component are well separated from any background features, as illustrated in Figure 15.

Figure 15: Example of spectral range selection highlighting the spectral features of the active ingredient

In this example, if correlation is only carried out over 1325 cm^-1 to 1800 cm^-1 then it may not be necessary to subtract a background signal. 

Pre-processing

The default pre-processing options use a 2nd differential and wide smoothing. These options work well when the spectra contain sharp features, such as the active ingredient in a nasal spray, and when the spectral range selection is not significantly narrowed.

Smoothing

The spectral data is smoothed to reduce noise whilst preserving the underlying signal. A sliding window approach is used to produce a weighted average over a number of points. There are three different levels of smoothing using different sized sliding windows: narrow (7points), intermediate (31 points), and wide (57 points). Where the full spectral range is used, wide smoothing is recommended. If spectral range selection is used, then depending on the width of the selected range, intermediate and narrow smoothing options should be investigated. 

Derivative

The derivative can be set to either 1st or 2nd. In general, if the spectra show sharp features, then the 2nd derivative is more effective. Using the 1st derivative may be more appropriate if the spectra contain broad features. 

Chemical classification

Particles can be classified by their chemical identity using correlation scores to the reference library. Particles can be added to a class using their best correlation to the reference material, or if the correlation score is above a minimum value. To positively identify particles, it is recommended that the correlation score should be above a minimum value. This minimum correlation score should be determined by inspecting the processed spectra, where a signal to noise ratio of 3 to 1 for the characteristic peaks has been used as a limit of detection⁶ (the S/N ratio is not calculated directly in the Morphologi software so can be estimated). At this stage you may also assess more than one set of processing conditions to see which results in the higher number of positively identified particles. Different processing conditions can be assessed by using the reanalyze feature to re-process the same set of data. Figure 16 shows the distribution of correlation scores to the API in a nasal spray sample where two populations are visible alongside example spectra, and correlation scores around the lower limit for positive correlation to the active ingredient.

Figure 16: Example of the distribution of correlation scores, and example particles and spectra with correlation scores

Once the particles have been classified a record can be created which contains only the particles identified as active (as shown in Figure 17), enabling the particle size distribution of the active component to be compared between the generic and innovator products.

Figure 17: Particle size distribution in volume and number for all particles and the active ingredient

Orthogonal techniques

The minimum particle size which can be routinely targeted for chemical analysis using the Morphologi 4-ID is around 1µm. This is due to the weak signal from particles smaller than the spot size of the laser (2µm) and the difficulty in returning to fine particles moving under Brownian motion. To compare generic and innovator products below 1um in size requires an orthogonal technique. Although not a chemically specific technique, laser diffraction has been suggested as a potential orthogonal technique¹. The particle size distribution of a nasal spray measured by laser diffraction is shown in Figure 18. This would allow the particle size distribution of the whole sample, active and excipient, to be compared. Although the chemical identity of the particles smaller than 1um cannot be identified by laser diffraction, the quantity of material in that size range can be compared between generic and innovator.

Figure 18: Example of a nasal spray suspension sample measured by laser diffraction

Conclusion

By combining image analysis and Raman spectroscopy, MDRS offers unique insights into the bioavailability and bioequivalence of complex generics like nasal sprays. In this document, we have covered aspects of sample preparation and measurement settings that will impact MDRS results with the goal of developing robust methods for MDRS to enable enhanced analytical precision and workflow efficiencies. For further guidance on implementing these methods, or to explore how Malvern Panalytical’s solutions can support your analytical needs, please contact us.

References

1. Scientific Considerations for the Review and Approval of First Generic Mometasone Furoate Nasal Suspension Spray in the United States from the Bioequivalence Perspective https://www.ncbi.nlm.nih.gov/pubmed/30617594
2. Understanding the performance of nasal spray drug delivery systems using the Malvern Spraytec https://www.malvernpanalytical.com/en/learn/knowledge-center/application-notes/an101104nasalspraydrugdeliveryspraytec
3. A L Dragoo, C R Robbins, S M Hsu et al “A critical assessment of requirements for ceramic powder characterization” Advances in Ceramics Vol 21 Ceramic Powder Science (1987).The American Ceramic Society Inc. (p711 - 720)
4. https://www.fda.gov/drugs/news-events-human-drugs/new-insights-product-development-and-bioequivalence-assessments-generic-orally-inhaled-and-nasal  See Session 2: Novel Analytical Tools for Characterization of Nasal Suspensions
5. Chemical Acquisition and Raman Spectral Processing methods used with the Morphologi G3-ID https://www.malvernpanalytical.com/en/learn/knowledge-center/technical-notes/tn121214ramanspectralprocessing
6. B J Thomas et al “Analytical Method Development for Characterizing Ingredient-Specific Particle Size Distributions of Nasal Spray Suspension Products” Journal of Pharmaceutical Sciences, Volume 110, Issue 7, July 2021, Pages 2778-2788