In USP 27 ( the latest I have immediate access to ), the related substances method used is normal phase (iC8, IPA, HOAc ), with silica column, but the EP uses a reverse phase with C18 column and methanol-based acidic mobile phase. The EP strongly cautions about heat, light, and oxidation, the USP about light. The compounds are known to be unstable in solution.
You could try the ( older? ) USP normal phase related substance method, but if they have changed to reverse phase, the method may be fragile. I'd be inclined just to follow the standard related substances methods and perform the analysis quickly trying to eliminate the known degradants.
If you want, you could contact the authors of the following study and ask about their methods, more details available at Pubmed.
Pharmaceutical quality of generic isotretinoin products, compared with Roaccutane.
Taylor P.W., Keenan M.H.
Curr Med Res Opin 2006 Mar;22(3):603-15.
Good luck,
Bruce Hamilton