5 replies to this topic

[LLLi]

Hi everyone,

 

I have a question about choosing the lambda Z calculation method.

 

Should I choose best fit or time range? If to use time range, how to decide the first time point and the last time point?

 

Thank you!

 

LLLi

[Simon Davis]

Personally I use best fit first and then review those selections to adjust and confirm them bearing in mind the following;

a) aim to estimate Lz over at least 2 half-lives, i.e. if I can go back further for a v small reduction in rsq-adj then I will do so.

b.) consistency with other, comparable, profiles, i.e. starting at similar times for similar individuals, beware in an SAD design with 2compartment model drugs since sometines in the lower doses you actually see only distribution phase.

c) if your SOP permits potentially excluding outliers if you feel there is justification e.g suspected switched sample.

Hope that helps - there are many views and disucssions on this if you search PharmPK archives or BEBAC forum e.g.

http://forum.bebac.a...6619&order=time

But really think of the fact you trying to get the best, most reasonable, estimate of apparent rate of elimination in order to extrapolate AUC with confidence. Then if you've written your SOP/protocol with that in mind and followed it you can be confident in answering any reviewer.

I think it's possible to spend a lot of time debating the exact points chosen for little real gain in your quality of analysis.

Simon

There isn't a defienr

Edited by Simon Davis, 23 September 2016 - 02:41 PM.

[LLLi]

Hi Simon,

 

Thank you for your reply. I have some more questions.

 

a) aim to estimate Lz over at least 2 half-lives, i.e. if I can go back further for a v small reduction in rsq-adj then I will do so.

This means that the time range should be larger than at least 2 half-lives and the influence of the change on the rsq-adj should be small? 

b.) consistency with other, comparable, profiles, i.e. starting at similar times for similar individuals, beware in an SAD design with 2compartment model drugs since sometines in the lower doses you actually see only distribution phase.

I thought the consistency means the similar lambda Z for similar individuals under the same situation. 

If consistency means the similar starting time for similar individuals, how do we know we need adjust the time range? Plot the starting and ending time points of all subject and adjust the "outliers"?  How about plotting the lambda Z and adjusting the outliers?

 

Thank you!

LLLi

[Simon Davis]

a) aim to estimate Lz over at least 2 half-lives, i.e. if I can go back further for a v small reduction in rsq-adj then I will do so.

This means that the time range should be larger than at least 2 half-lives and the influence of the change on the rsq-adj should be small? 

Yes - that is what I would aim for (2 half-lives), i wouldn't put a hard limit on it but above0.85 for Rsq-adj is a guide I often see implemented in SOPs.

 

b.) consistency with other, comparable, profiles, i.e. starting at similar times for similar individuals, beware in an SAD design with 2compartment model drugs since sometines in the lower doses you actually see only distribution phase.

I thought the consistency means the similar lambda Z for similar individuals under the same situation. 

If consistency means the similar starting time for similar individuals, how do we know we need adjust the time range? Plot the starting and ending time points of all subject and adjust the "outliers"?  How about plotting the lambda Z and adjusting the outliers?

Agreed - it is consistent Lz that I would expect, and sometimes over-riding the 'automatic' selection to go back a few points earlier will help.
By outliers I meant points within an individual's time-conc profile that seem not to fit with expected/other's behaviour.  I would investigate further with lab and clinic to see why this might happen e.g. is a switched sample suspected/possible.

 

Does that help - they're my personal opinions/experiences, maybe some others can contribute theirs too.

 

 Simon

[LLLi]

Hi Simon,

 

I am reading the NCA analysis strategies from the Introduction to Phoenix WinNonlin 6.3 Course Materials. One suggestion is that visual inspection should support a mono-exponential decrease. Dose this means that when there is a two-compartment model, we should choose the time points during the beta phase?

 

Thanks,

LLLi

[Simon Davis]

That is certainly my opinion, you are trying to define the apparent elimination half-life, typically the alpha is realtively short-lived and would more likely correspond to distribution.

 

 Simon