Log in
Register|
Alex ☆ Austria, 2014-09-05 11:42 (3514 d 09:10 ago) (edited by Alex on 2014-09-05 12:42) Posting: # 13460 Views: 5,756 |
|
|
Dear all! Hopefully you can help me regarding the analysis of serial sampling PK data in WinNonlin 6.3. Usually I am analyzing such data using R package 'PK' (Jaki and Wolfsegger, 2011) which conveniently provides asymptotic SEs for the most important PK parameters in serial sampling design. For PK modeling, I am using a bootstrap approach to quantify the uncertainty in PK estimates. Using NCA in WinNonlin doesn't provide SEs for most of the PK estimates (I think only for AUC to infinity and Cmax) which is undesirable as the uncertainty plays an important role in serial sampling designs and adds substantial value to interpretation of estimates. To avoid this drawback, one can switch to compartment modeling for which WinNonlin provides SEs. But I am not sure how these SEs were calculated (no hint in the manuals) and if they are appropriate in serial sampling designs?! Can anybody help? Thanks a lot and I am looking forward to read your opinions! All the best, Alex Jaki T and Wolfsegger MJ (2011). Estimation of pharmacokinetic parameters with the R package PK. Pharmaceutical Statistics, 10(3):284-288. R package version 1.2-4. URL: http://CRAN.R-project.org/package=PK |
|
SDavis ★★  UK, 2014-12-18 12:05 (3410 d 07:48 ago) @ Alex Posting: # 14105 Views: 4,317 |
|
|
Sorry for the late response here – the SE error calculation in WNL Classic models is based on least squares fit and it is calculated by taking the square root of the variance where D is the design matrix. ![[image]](img/uploaded/image283.png) Does that help? Simon — Simon Senior Scientific Trainer, Certara™ [link=https://www.youtube.com/watch?v=xX-yCO5Rzag[/link] https://www.certarauniversity.com/dashboard https://support.certara.com/forums/ |
|
AngusMcLean ★★ USA, 2014-12-18 20:20 (3409 d 23:33 ago) @ Alex Posting: # 14115 Views: 4,310 |
|
|
Using NCA Phoenix WinNonlin you can get estimates of the standard deviation and or the standard errorof the PK paramaters. Why not avoid modelling and use this NCA derived parameter data. At that point you could use a bootstrapping program. Why not avoid modelling? Why is it that serial sampling is presenting a problem? Angus |
Ing. Helmut Schütz