Log in
Register|
jag009 ★★★ NJ, 2013-01-22 22:11 (4548 d 13:41 ago) Posting: # 9900 Views: 9,580 |
|
|
Hi all, Can someone give me some advice on how to incorporate intrasubject variation into BE simulations? For example, if one wants to simulate (monte carlo sim) a 2x2 crossover trial for a drug, one would have on hand the PK parameters obtained from fitting the PK profile with a model, the intersubject CV of each PK parameters, one can also get the intrasubject CV too. When it comes to doing the simulation, how does one incorporate the within-subject CV though? Thanks John |
|
ElMaestro ★★★ Denmark, 2013-01-22 23:59 (4548 d 11:53 ago) @ jag009 Posting: # 9901 Views: 8,583 |
|
|
Hi John, ❝ Can someone give me some advice on how to incorporate intrasubject variation into BE simulations? For example, if one wants to simulate (monte carlo sim) a 2x2 crossover trial for a drug, one would have on hand the PK parameters obtained from fitting the PK profile with a model, the intersubject CV of each PK parameters, one can also get the intrasubject CV too. When it comes to doing the simulation, how does one incorporate the within-subject CV though? It is quite well described in the paper by Diane Potvin. In reality, depending on your needs you might only need to simulate the log(T/R)=log(T)-log(R), with no regard to between-subject variability. Otherwise, you can simply roll out the standard model Y=Trt+Per+Seq+Subj+e, where you assign whatever fixed values you find appropriate to the four fixed factors, and define the e(rror) on basis of your desired CV. For most practical purposes you can ignore the subject, sequence and period effects, and we are effectively back at the Potvin approach. — Pass or fail! ElMaestro |
|
jag009 ★★★ NJ, 2013-01-23 00:13 (4548 d 11:38 ago) @ ElMaestro Posting: # 9902 Views: 8,524 |
|
|
Thanks ElMaestro, ❝ It is quite well described in the paper by Diane Potvin. In reality, depending on your needs you might only need to simulate the log(T/R)=log(T)-log(R), with no regard to between-subject variability. You meant to wrote "... with no regard to within-subject variability."? Can you provide reference to the paper by Potvin? Thanks John |
|
ElMaestro ★★★ Denmark, 2013-01-23 05:52 (4548 d 06:00 ago) @ jag009 Posting: # 9903 Views: 8,507 |
|
|
Hi John, ❝ You meant to wrote "... with no regard to within-subject variability."? Actually, I meant between. In Potvins paper (see PubMed ref. here) the approach was to just simulate the log(T)-log(R) plusminus intrasubject variability and forget about the between-subject variability; for the purpose of her simulations this variability -just like period and sequence effects- isn't necessary. — Pass or fail! ElMaestro |
|
jag009 ★★★ NJ, 2013-01-23 16:03 (4547 d 19:49 ago) @ ElMaestro Posting: # 9905 Views: 8,342 |
|
|
Many thanks ElMaestro. |
|
Helmut ★★★   Vienna, Austria, 2013-01-24 07:44 (4547 d 04:08 ago) @ jag009 Posting: # 9907 Views: 8,261 |
|
|
Hi John, ❝ […] one would have on hand the PK parameters obtained from fitting the PK profile with a model, the intersubject CV of each PK parameters, one can also get the intrasubject CV too. Your other questions were already answered by ElMaestro, so I’ll concentrate on this interesting one. What do you mean by “CV of each PK parameter”? They don’t help because in my experience model-based AUC and Cmax tend to have a smaller CV than ones obtained from an actual sampling schedule (remember: you will use NCA in the study). I assume you are using classical two-step PK, right? Another option would be PopPK. If I recall it correctly Emily Colby posted a cross-over PopPK model on Pharsight’s Extranet. — 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, 2013-01-24 20:54 (4546 d 14:58 ago) @ Helmut Posting: # 9908 Views: 8,067 |
|
|
Hi Helmut and ElMaestro, Correct me on the following steps if I am wrong (it's just an experiment). I want to see if this approach can be used to determine "What if" when we increase the subject sample size. 1) Obtain Test and Reference plasma data from 24 subjects 2) For each product, fit each subject data with a pk model and obtain the model parameters. Determine the mean(CV). 3) Per product, carry out Monte carlos to simulate plasma concentration data for 'n' number of subjects with the PK model parameter (V,D,Cl,K01,K10) means(CV). 4) Randomly assign subject to period and sequence. 5) NCA the data and evaluate BE. JOhn |
|
ElMaestro ★★★ Denmark, 2013-01-24 21:35 (4546 d 14:17 ago) @ jag009 Posting: # 9909 Views: 8,199 |
|
|
Hi John, ❝ Correct me on the following steps if I am wrong (it's just an experiment). I want to see if this approach can be used to determine "What if" when we increase the subject sample size. ❝ 1) Obtain Test and Reference plasma data from 24 subjects ❝ 2) For each product, fit each subject data with a pk model and obtain the model parameters. Determine the mean(CV). ❝ 3) Per product, carry out Monte carlos to simulate plasma concentration data for 'n' number of subjects with the PK model parameter (V,D,Cl,K01,K10) means(CV). ❝ 4) Randomly assign subject to period and sequence. ❝ 5) NCA the data and evaluate BE. Here I am a little in doubt about what you are trying to achieve. However, I have a vague feeling you might wish to use bootstrapping for this; perhaps you can simpy forget all the model stuff and just bootstrap the Cmax and AUCt's directly and do parametric anova for each bootstrapped dataset. It depends on the specifics. Let's hear a little more, please. — Pass or fail! ElMaestro |
|
jag009 ★★★ NJ, 2013-01-24 23:19 (4546 d 12:32 ago) @ ElMaestro Posting: # 9910 Views: 8,022 |
|
|
Hi ElMaestro, ❝ Here I am a little in doubt about what you are trying to achieve. However, I have a vague feeling you might wish to use bootstrapping for this; perhaps you can simpy forget all the model stuff and just bootstrap the Cmax and AUCt's directly and do parametric anova for each bootstrapped dataset. It depends on the specifics. Let's hear a little more, please. It's one of this off the top of my head experiment  I actually have done what you suggested (bootstrapping). Thanks John |
|
jag009 ★★★ NJ, 2013-01-29 17:48 (4541 d 18:04 ago) @ jag009 Posting: # 9926 Views: 7,624 |
|
|
Hi ElMaetro, I think I sent the message to the wrong post and person (see 5x5 crossover thread) as I was meaning to ask you the following question. Helmut, don't kill me  In a nutshell here is what I am trying to do. I have a set of data from a parital 3-way replicate study (1 Test, reference given twice). I want to find out what happens if I simulate this to a full replicate study. Therefore I need to 1) Generating a 2nd set of data for the test product 2) I also want to introduce some intrasubject CVs on the test data Question 1 can be done with bootstrap on the existing data(Correct?). But how should I approach question 2 so that I can see BE/SABE scenarios with different intrasubject CV on the test? thanks John |
Ing. Helmut Schütz