Malvern software allows recomputing the particle size distribution in volume to number frequency. So, to compare you only need to recalculate your volume distribution and then compare any statistical parameter D10, D15, D90 etc..., in a correct way.

Good luck

Jyotsna Kamala Gutam

This is not as easy a question as it first appears. Why do this comparison at all? The 2 techniques are providing different and complementary information. There are obvious differences in ensemble versus one-by-one techniques of which number and volume may be one the most important as you state. If all particles are the same (monodisperse) and spherical, and within the range of both techniques, then identical results should be obtained. You say nothing of the material, its shape, and approximate size distribution and this is vital for a specific answer. I’d put representative sampling at the top of the list. See this webinar (registration needed):

Keys for Successful Analysis – Representative Sampling & Estimation of Standard Error Calculation

https://www.malvernpanalytical.com/en/learn/events-and-training/webinars/W190613KeysLD

A few questions:

• AQL of company/specification? Purpose of comparing? Scarlett quote: Brian Scarlett: ‘A particle size measurement does not have a meaning unless the objective of the measurement is also specified’ B Scarlett ‘25 Years of Particle Size Conferences’ Particle Size Analysis, Eds. N G Stanley-Wood, R W Lines, Royal Society of Chemistry (1992)

• What parameters being specified? x90/x95/x99 etc. What is the size range of vision system lens?

• Sample preparation in both cases – dispersion? Are the systems being measured in the same form of dispersion?

• What is the shape/form of the material – are we comparing circular equivalent versus volume?

• Is the distribution narrow or broad? Small material (< 2 μm) will better on laser diffraction rather than manual microscopy

• Representative sampling – see link above. Are the 2 samples being compared in the same manner? What is the best standard error based on the amount of material taken for the respective measurements?

• Vision – number of particles examined? Lens/magnification (link with distribution width)? What parameters used and specified? On the Malvern Morphologi, results are collected by number (one-by-one) but also can be displayed in volume format (for sparse distributions: few particles at larger sizes and this has statistical implications)

It is not, as you state, a simple conversion from D[1,0] to D[4,3] conversion... Even ignoring representative sampling, any errors in a number distribution are magnified 3-fold in the converted-to-volume distribution and this gets worse and worse as the distribution widens. So, we need more information to give you a specific answer.

Alan F Rawle Thank you Dr. Alan for a very detailed explanation.

Here are the answers to your questions:

1) The objective of particle size measurement is to determine if the D90 of a given API (active pharmaceutical ingredient) sample is within the specifed range or not

2) Image analysis is done on Microscope images. The microscope slide is prepared either by using a dry powder or by dispersing the dry powder in oil. In Malvern, the testing could be either a wet method or a dry method.

3) we encounter typical particle shapes of needles, cylindrical rods, plates etc

4) The distribution is generally broad.

5)

Jyotsna Kamala Gutam

OK, thank you for your response. I attach the 4Q’s which are de rigeur before one enters the laboratory… Try and answer these. Are you familiar with the concept of AQL? Some comments on your responses:

1) It isn’t a scientific objective ‘to determine if the D90 of a given API (active pharmaceutical ingredient) sample is within the specified range or not’. How is the specification set in terms of ‘fit for purpose’ and ‘just good enough’? What happens to the physical properties of the powder or material when the numbers are OOS? For example, is the dissolution impaired so that the drug loses its required efficacy? Is the compacting/compressive process in tableting impacted by having too much large material present? Is your specification a number volume/mass based one? Are your products sold or processed on a number or mass basis? Do you guarantee or specify (for example) 4.6 billion particles of drug to the customer or 250 mg (say)? See, this webinar (registration needed):

Basic Principles of Particle Size Analysis

https://www.malvernpanalytical.com/en/learn/events-and-training/webinars/W170216BasicsParticleSize

In order to have a 1% standard error on the x90 (the preferred ISO term) then on a number basis you’ll need to count 10000 particles in the x90+ part of the distribution. Have you done this? This statistical fact has minimum mass and representative sampling issues and you state that your distribution is broad. You may need kg’s of material to satisfy this requirement. See:

Keys for Successful Analysis – Representative Sampling & Estimation of Standard Error Calculation

https://www.malvernpanalytical.com/en/learn/events-and-training/webinars/W190613KeysLD

2) OK, you’re testing via oil or aqueous based methods. How do you know that the sample is the same (representative) and in the same state of dispersion in both cases? Laser diffraction will take much more sample than imaging typically but there is still a minimum mass needed in line with 1) above especially in line with your ‘broad’ distribution. If every particle was the same, then the answers should be identical for both diffraction and imaging (assuming several things such as the size ranges being applicable and correct optical properties). How are the samples being dispersed (if at all)?

3) So how are you getting a single number from acicular and plate-like material? This applies to both systems where imaging will see a projected area (2D representation of a 3D particle) and, if flow alignment occurs in the wet (or dry) diffraction measurement, then also a projected area (but it’s the z-axis RI that’s important here). There is also a shape distribution (indeed many shape distributions, depending on the parameter selected) as well as a particle size distribution

4) Your distribution is broad. This has implications for the best standard error you can achieve based on the point in the distribution you wish to specify (x90 in your case), the mass of sample actually analyzed or examined, and the density of the sample (for minimum mass considerations)

5) You have no question for this part, so I have no answer...