Tscale parameter calculation in winnonlin

IVIVC

2 replies to this topic

#1 jessicalee

jessicalee

Newbie

• Members
• 5 posts

Posted 16 June 2022 - 02:54 AM

Repost the topic of Vijay. I have the same question as Vijay, and still wait for the response. Here is the full text.

Dear all,
I am using winnonlin v. 6.2 with ivivc toolkit for ivivc,
I want to know how Time scalling parameter calculated in this software also i didn't understand the correlation parameter estimation specified in the manual for eg. Abscale and Tscale etc..initial value or bound should be provided for them..

Why there is need to give the upper bound and lower bound to be given
for Tscale parameter or Tlag?when it can be calculated only from Tvitro vs Tvivo Levy plot BY LINEAR REGRESSION or is it nonlinear regression ,

also can anyone elaborate the correlation model which are provided in
manual some of them as per i understood

(1) Fabs = Diss(Tvivo)

means :- regression between Fabs (amount absorbed at time invivo for given formulations) vs Diss (Tvivo) (amount dissolved at the corresponding invivo time for given formulations)
y=x;
y=amount absorbed at time invivo
x=amount dissolved at the corresponding invivo time(which can be
calculated easily from any dissolution fitted function which already
i have done in invitro tab)

(2) Fabs = AbsScale * Diss(Tscale * Tvivo)

means :- regression between Fabs (amount absorbed at time vivo for given formulations) vs Diss (Tvivo*Tscale) (amount dissolved at the (corresponding invivo time*Tscale) for given formulations)
here what is the need of intial value or bound to be given for Tscale which can be measured from Tvitro vs Tvivo Levy plot.

y=mx

y= Fabs(amount absorbed at time invivo)
m= Absscale parameter

x=amount dissolved at the (at time invivo*Tscale)
this amount(X) can be calulated from any dissolution data fitted model eg. hill, weibull which already i have done in invitro tab if i have Tscale
parameter obtained from Tvitro vs Tvivo Levy plot

Also when performing regression (most probably linear or is it NONLINEAR  if it is in nonlinear regression i think Absscale parameter's initial value or bound should be provided) between Fabs(amount absorbed at time invivo) vs amount dissolved at the (at time invivo*Tscale) Absscale parameter which can be calculated by simple regression-analysis equation THAN WHY BOUNDS OR INITIAL VALUE
SHOULD BE GIVEN
y=mx

y= Fabs(amount absorbed at time invivo)
x=amount dissolved at the (at time invivo*Tscale)
m= Absscale parameter

Please correct me if i am wrong..i will highly appreciate your response

Vijay

#2 Simon Davis

Simon Davis

• 1,230 posts

Posted 16 June 2022 - 06:03 AM

Hi Jessica, I'm sorry I didn't see Vijay's original post,where/when was it made?

You have quite a few questions, so apologies if I miss any, have you taken the IVIVC course on Certara University or followed the free simple tutorial that is included with Phoenix 8.3.5 https://www.certarau...c-8-3-tutorial.(Note upgrade to the latest version of phoenix is free and your version, 6.2 is no longer supported).

Tscale is simply making the adjustment between the duration of your in vitro experiment vs the in vivo data, it is optimised using the WinNonlin classic modelling engine which is used for all the IVIVC modelling; least squares regression.

One challenge of learning IVIVC is that if you areused to NCA, yu now have to learn some basic approaches in modelling, in the case of adding bounds, these guide the algorithm away from physiolically improbable/impossibe solutions that might appear to be mathematically possible. e.g. a negative Tlag would not be useful.

Similarly these algorithms work best when a reasonable initial estimate is given as a starting point and so we use features like the Levy plot to see what the slope and intercept are when comparing dissolution data to the deconvolved absorption in vivo.

We do all this model building with the first 4 tabs of the wizard so ultimately we can simulate new in vivo profiles for formulations for which we only have dissolution data.

If your observed data is rich enough you might get away ith simple regressions and then making some assumptions but modelling and simulating you data has the possibility of qunifying and extrapolating with more confidence.

Simon.
take a look at the second attachment on this post;
https://support.cert...election/?p=969

Edited by Simon Davis, 16 June 2022 - 08:05 AM.

#3 jessicalee

jessicalee

Newbie

• Members
• 5 posts

Posted 16 June 2022 - 07:57 AM

Dear Simon,

Thank you for your response.  I have other questions about the calculation of the correlation model.

According to my understanding, levy plot can also calculate Tscale and Tshift, which are the slope and intercept of the plot respectively. I found out that the Tscale parameter from the correlation equation constructed by winnonlin is not the same as the intercept of levy plot.

1. How does winnonlin calculate the Tscale parameter and what is the difference in principle between this method and the levy plot?

2. In order to have a high correlation between each Fab and Fdiss after the treatment of the correlation equation, the Tscale used for each time point wound be different. How does Winnonlin deal with this problem and come up with a Tscale?

I will Highly appreciate if you can clear it.

Cheers,

Jessica

Edited by jessicalee, 16 June 2022 - 07:58 AM.

Also tagged with one or more of these keywords: IVIVC

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users