We have implemented a balanced cross-over design (6 sequences Williams ) for a three way cross-over BE study (fed, fasted and sprinkled on food). One subject dropped out at Period 3. I am wondering whether it is the best thing to do prior to BE data analysis in Phoenix to drop the PK data from the other two treatments of this subject (Period 1 and Period 2). Or should I perform the data analysis with all of the data I have? It is not a marginal situation. Please can you comment on this issue. Thank you, Angus
Angus for such a statistical enquiry you might be better cross-posting at a dedicated BABE forum like Bioequivalence and Bioavailability Forum Simon. [size=4]EDIT, [color=#0080BF]In case anyone else has the misfortune of a similar occurrence here is the direct link to the suggested cross posting with a prompt response by Helmut as always ;0)[/color] Bioequivalence and Bioavailability Forum • BE and drop out subjects Phoenix
Simon: Thank you for your reply. I have just read Helmut’s reply post. It is a statistical issue, but it is also a regulatory one. The Phoenix program originates here in the US and the work is of course subject to FDA review. I would have thought that Pharsight would have some good information on this topic in view of the importance of the BE aspect of the program. Drop out subjects, as you may well imagine, are very common so it is an issue people constantly have to deal with. Angus
As Helmut answered on his forum, WNL Bioequivalence is very similar to PROC MIXED, so it will handle unbalanced data appropriately if you include the subjects with missing data. We in Phoenix Development of course stay familiar with FDA guidances (as well as EMA and other countries). However, you can read the one paragraph that is provided about dropouts in the FDA 2001 BE guidance and see that it is not a one-size-fits-all answer. Furthermore, although we provide software that our clients use to make FDA and other Regulatory submissions, ultimately it is the users who are responsible for their interpretation and report, so it seems practical to discuss this with others who have much more experience dealing daily with regulators; hence the suggestion to post on the BEBAC forum where you’re more likely to learn of others’ real-world experiences independent of software used. Regards, Linda Hughes
Linda: Thank you for the response. I find it most reassuring. I followed Simon’s advice by posting on the forum and I also found Helmut’s response very helpful indeed. I completely agree with all of your remarks. I understand perfectly the disposition of Pharsight. I think many members of this forum will find this discussion most useful. Angus