Jump to content

- - - - -

Q&A from Lesson 16: Warfarin IV bolus Turnover models

  • Please log in to reply
No replies to this topic

#1 Christopher Mehl

Christopher Mehl


  • Administrators
  • 27 posts

Posted 12 June 2017 - 02:56 PM

Q: How did the results of the built-in model compare to both Text models? Or, why did you need to run the built in?

A: The built-in model had some difficulty finding precise estimates for kout and gamma, and was used to find an initial estimate of gamma for the other models. The model with tlag improved the parameter precision and had better diagnostic results (AIC, BIC, -2LL).

Q: Can the model compare option be used instead of appending the worksheets?

A: the model comparer is only available with an NLME license for models run in Population mode. Since this was a single profile (Individual modeling), the Append Worksheet tool was used to compare model results.

Q: Why does kloss eq kout give better fit?

A: For this type of transit model, you cannot estimate separate values for kloss and kout as it is an identifiability problem. If you do try to estimate them separately, you will see you will get similar values and very high CVs. In model #3, Kloss and Kout had very similar final estimate (0.92 and 0.99) and were also highly correlated. Because of this, the model was reduced to a single k parameter "kout", and indeed the model fit better with one less parameter.

Q: Why does IC50 occur at the recovery phase?

A: There is a good discussion of this in the Gabrielsson and Weiner reference. A sensitivity analysis of IC50 shows that it has the most influence on the recovery phase. Please see the figure below, which shows the response time course of PCA at standard (0.35), 100% increase (0.7), and 50% reduction (0.175):

The time of 50% recovery (80 hrs) was then used to get the initial estimate for IC50 from the warfarin concentration function, as discussed in the presentation slides.

Attached Thumbnails

  • fig4.5.jpg

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users