Log in
Register|
jag009 ★★★ NJ, 2025-11-19 21:29 (196 d 05:17 ago) Posting: # 24509 Views: 2,714 |
|
|
Hi all, I have a simple question about analyzing two treatments from a >2 arms study, as per ICH M13a recommendation below: "...The analysis for each comparison should be conducted excluding the data from the treatments that are not relevant to the comparison in question” I have a 4-treatment 4-sequence BA study (ABCD) and I want to carry out a comparison C vs D. How should I prepare the PK parameter dataset? 1) Should I just leave the pk parameters as blank (missing) for treatment A and B and then run proc mix? 2) or I should remove treatment A and B data from the dataset so that it only contains C and D data before running proc mix? I also got an advice from a stat fellow who told me to convert the study into a 2-way 2-sequence study before running proc mixed(????). What he meant was remove treatment A and B data and also convert the 4 sequences - ABDC, BCAD, CDBA, DACB, by removing the A + B and leave only C + D so to end up with 2 sequences CD and DC. Thx J |
|
Relaxation ★ Germany, 2025-11-20 09:47 (195 d 16:59 ago) @ jag009 Posting: # 24510 Views: 2,426 |
|
|
Dear J. Just very briefly, your second option is the one to use, so "just" remove the treatments that are not of interest from the data set. Thinking about it, maybe the first option may give the same information, but I would have to try that in a dummy data set. I think, even if that works, the other treatments may still be mentioned in the SAS output as levels then, which would certainly trigger questions. In any case, the last option including changes in the sequence or even period allocation for me seems risky and was also recommended against (Not sure, whether this was in a guideline, but it was discussed (edit: and advised against) when the EMA-IR was introduced at the 3rd EGA conference, if you want to search for that). Best regards! |
|
Shuanghe ★★ Spain, 2025-11-20 14:59 (195 d 11:47 ago) @ jag009 Posting: # 24512 Views: 2,464 |
|
|
Hi J, ❝ 2) or I should remove treatment A and B data from the dataset so that it only contains C and D data before running proc mix? ❝ I also got an advice from a stat fellow who told me to convert the study into a 2-way 2-sequence study before running proc mixed(????). What he meant was remove treatment A and B data and also convert the 4 sequences - ABDC, BCAD, CDBA, DACB, by removing the A + B and leave only C + D so to end up with 2 sequences CD and DC. — All the best, Shuanghe |
|
jag009 ★★★ NJ, 2025-11-21 00:47 (195 d 01:59 ago) @ jag009 Posting: # 24513 Views: 2,427 |
|
|
Thank you guys. I tried the first option in Phoenix and I received an error related to Satterthwaite... I will try the 2nd option and see. J |
|
Helmut ★★★   Vienna, Austria, 2025-11-23 12:26 (192 d 14:20 ago) @ jag009 Posting: # 24514 Views: 2,272 |
|
|
Hi John, ❝ I tried the first option in Phoenix and I received an error related to Satterthwaite... I will try the 2nd option and see. Say, subject 1 is randomized to sequence \(\small{\text{ACBD}}\). In Phoenix (as well as in in SAS) you need a flat file structure. Therefore, for this subject you should drop the data of periods 1 and 3 (i.e., treatments A and B) and code:$$\small{\begin{array}{c|cccc} \textsf{subject} & \textsf{period} & \textsf{sequence} & \textsf{treatment} & \textsf{PK}\\\hline 1 & 2 & \text{ACBD} & \text{C} & foo\\ 1 & 4 & \text{ACBD} & \text{D} & bar\\ \cdots & \cdots & \cdots & \cdots & \cdots\\ \end{array}}$$ — Dif-tor heh smusma 🖖🏼 Довге життя Україна! ![[image]](https://static.bebac.at/pics/Blue_and_yellow_ribbon_UA.png) Helmut Schütz ![[image]](https://static.bebac.at/img/CC by.png) The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
jag009 ★★★ NJ, 2025-11-23 23:37 (192 d 03:09 ago) @ Helmut Posting: # 24515 Views: 2,280 |
|
|
Thanks Helmut! J |
Ing. Helmut Schütz