Log in
Register|
Nirav Jani ● 2011-02-03 20:32 (5199 d 02:02 ago) Posting: # 6550 Views: 5,743 |
|
|
Dear All, We have pilot data for partial reference replicate design (2 treatment, 3 period reference replicate design). Based on pilot results for N=9 subjects, intra-subject cv of Cmax and AUC for reference product is 28% and 25%, respectively. The ratio (T/R) for Cmax and AUC is 115% and 114%, respectively. The power of the study for Cmax and AUC is below 70%. It will be helpful if someone suggest me for the sample size calculation for pivotal partial replicated design (2 treatment, 3 period reference replicate design) using the above information? Thanks and regards, Nirav Jani |
|
ElMaestro ★★★ Denmark, 2011-02-03 21:41 (5199 d 00:53 ago) (edited on 2011-02-04 07:59) @ Nirav Jani Posting: # 6551 Views: 4,938 |
|
|
Hello Nirav, On basis of T/R 115% and CV 28% I get the following for a three-period study: ![[image]](img/uploaded/image22.jpg) (Edit: Sorry, misspelling of your name) — Pass or fail! ElMaestro |
|
d_labes ★★★ Berlin, Germany, 2011-02-04 13:00 (5198 d 09:34 ago) @ ElMaestro Posting: # 6557 Views: 4,952 |
|
|
Dear ElMaestro, were did your graph come from? The infamous PowerTOST package for R gives f.i. a sample size of 102 if target power is 80%, achieved power is 0.802180. Reading the value from your graph seems higher. Dear Nirav, The power curve given by ElMaestro is for the conventional ABE test. Since you have employed a partial replicate design I suppose that you aimed for going with scaled ABE or widened ABE acceptance limits. Concerning the sample size for this have a look at this thread and search the forum. Of course your CV results does not allow you to go for scaled ABE because the Reference variability is <30% for both PK metrics. But only if you see these values as "carved in stone". Given the fact that the CV is usually underestimated in small pilot's you have IMHO a fair chance to come out in the pivotal study with CV's >30%. BTW: An upper 80% CL for an estimated CV = 28% from a study with 9 subjects (df=6 for a 3-sequence study) is 39.9%. Thus I suggest you to go with 24-36 subjects in the pivotal study and cross your fingers  .The high point estimate obtained in your pilot study may of course decrease your chance to arrive at BE considerably. If you have reasons to see this as real, and not only by chance, the number of subjects must be much higher then above suggested. A tangible number I can't give you because there is no elaborate method for sample size or power estimation for scaled ABE up to now, at least as far as I know. If you don't believe in your chance to arrive at CV >30% and take your obtained point estimator of 1.15 as truth you have to go with above 100 subjects. A study with which much money can be earned, of course only if you find a sponsor who has that money in his big long pocket  .— Regards, Detlew |
|
ElMaestro ★★★ Denmark, 2011-02-04 14:23 (5198 d 08:11 ago) @ d_labes Posting: # 6558 Views: 4,870 |
|
|
Hi dlabes, ❝ were did your graph come from? ❝ The infamous PowerTOST package for R gives f.i. a ❝ sample size of 102 if target power is 80%, achieved power is 0.802180. ❝ Reading the value from your graph seems higher. I am sorry, I probably screwed up. I would at all times trust your calculations more than my own. — Pass or fail! ElMaestro |
|
d_labes ★★★ Berlin, Germany, 2011-02-04 15:00 (5198 d 07:34 ago) @ ElMaestro Posting: # 6559 Views: 4,792 |
|
|
My dear pirate, ❝ I would at all times trust your calculations more than my own. That is totally against Jaime's famous principle of faith: "Never trust in any calculations you haven't done yourself (and even then you should be cautious...)"  .(adapted correspondingly by me) — Regards, Detlew |
|
Nirav Jani ● 2011-02-04 20:15 (5198 d 02:19 ago) @ d_labes Posting: # 6560 Views: 4,783 |
|
|
Dear ElMaestro and d_labes, Thanks a lot for your reply. Here we are using pilot data of partial reference replicated (2-treatment, 3 period) crossover design for estimating sample size of pivotal study for same design and this is first time for me to face the situation and I don't know the method to calculate the sample size for 2 treatment, 3 period partial replicated design using the same design. If we use maximum intra-subject CV of Reference for Cmax (28%) and (T/R) ratio 115%, and calculate sample size as per ABE approach. Now, can we multiply the estimated number by 0.75 for 3 period partial replicated design? Can we go for this approach? Or we should not consider the (T/R) ratio for sample size for the pivotal study and should enroll 24 to 36 subjects for pivotal study for ISCV o 28%. Looking for your valuable suggestions. Thanks and regards, Nirav |
Ing. Helmut Schütz