Bioequivalence and Bioavailability Forum

Dear Helmut,

❝ don’t want to interfer with your enthusiasm, but it seems that currently (anything?) except f₂ is not acceptable (in the EU).

Thanks for your viewpoint.
I'm a bloody newbie on the field of dissolution. Usually I have nothing to do with the evaluation of dissolution profile. Therefore I'm not familiar with used and/or accepted methods. I simply was asked to implement Sarandanasa/Krishnamoorthy method (of course in SAS ) for what end ever.

Your reasoning about f2 (there can be only one?) is nevertheless astonishing for me. Let me cite Appendix I of the EMA guideline:
"When the ƒ2 statistic is not suitable (however this has to be justified, remark by D.L.), then the similarity may be compared using model-dependent or model-independent methods e.g. by statistical multivariate comparison of the parameters of the Weibull function or the percentage dissolved at different time points.
Alternative methods to the ƒ2 statistic to demonstrate dissolution similarity are considered acceptable, if statistically valid and satisfactorily justified."

Ok, ok, I see: Acceptable, but if not satisfactorily justified (howsoever it should be done) ...

As far as I noticed the 1997 FDA guidance "Dissolution testing for Immediate Release Solid Dose Forms" has also a recommendation of a multivariate measure of similarity (Tsong's Mahalanobis distance?)

Tsong, Y., Hammerstrom, T, Sathe, P., Shah, V. P.
Statistical assessment of mean differences between two dissolution data sets.
Drug Information Journal 30:1105–1112 (1996).

There is a free EXCEL Add-in out there (DDSolver and supplemental material) which implements Tsongs method. But seems the source code is not free. Its safeguarded by a password . Anybody out there to guide me to the upper 90% CI of the Mahalanobis distance?