In pharmaceutical manufacturing processes multimodal particle size distributions (PSDs) are often considered undesirable because to date there is no accepted way to characterize them as for example normal distributions. However, there is no firm evidence to indicate that a multimodal distribution will not ensure quality of the final product with respect to criteria such as dissolution and/or bioavailability profiles, processibility performance, etc. In view of this, there is a definite need to address the characterization of pharmaceutical products in terms of easily measured physical properties that can be used to judge "sameness” of the product. Although the situation is arguably more critical for particle size, a sound methodology for comparing and establishing batch-to-batch sameness of physical attributes is needed which forms the objective of the presentation.

We propose that the diagnostics of the product quality should be based on the desired functionality, which can be either of the dissolution profile or any other processibility performance index. The idea behind the proposed methodology is to identify the region of acceptable space around the measured properties from the view point of the acceptability region around the desired functionality. This can be done by either empirical approach or the model driven approach. In empirical approach various particles size distributions can be generated by physically mixing the samples followed by the measurements of the performance criteria (generally not amenable to model) for each of the samples. One can then establish the sameness criteria by mapping the acceptable performance to the acceptable particle size distributions. Oftentimes, the empirical methodology may be time consuming and expensive. On the other hand one can adopt the model based approach where the functionality can be modeled mathematically.

We illustrate here the model based approach to arrive at the simple diagnostic test on the parameters of PSD for the product acceptability based on the chosen functionality of dissolution.

• We use a mathematical model to simulate the reference dissolution profile and propose a computational protocol to generate the reference PSD for the desired dissolution profile. The reference can be very well generated through experiments or experience.

• The reference PSD was characterized using the parameters, d50 and d90 (defined as the particle sizes corresponding to the cumulative volume percentages of 50% and 90%, respectively).

• Nearly fourteen thousand sample PSDs were generated using Monte Carlo simulations satisfying specified 1) d50 in a range spanned by +/-10% of the d50 of the reference PSD and 2) a set of d50 and d90 spanning the range around +/-20% of the d50 and d90 of the reference PSD. These artificial distributions can be looked upon as the representatives of the different batch samples with different looking PSDs but with the gross parameters (d50 and d90) lying in a certain range around the reference values.

• Dissolution profiles were simulated for each of the distributions and compared with the reference dissolution profile using the standard similarity and difference (f1-f2) test (Moore and Flanner, 1996) to judge its acceptability.

• Based on the comparison, the domain of desired dissolution performance can be transformed into acceptable play off region around the parameters (d50 and d90) of the reference PSD. The play off region can be then used to judge the sameness of the product.

It is of interest to note that the methodology presented here can very well be extended to propose the diagnostic test on the measurable product properties for any other performance objective (e.g. granulation end point, etc).

References

Moore J, and Flanner, H., Mathematical Comparison of Dissolution Profiles. Pharm Tech., 20, 64-74 (1996).