Hello,
I am unable to get in touch with [email protected] in order to
join the network, therefore I am attempting to directly reach out to you,
as I have a question that I would like to post to the nmusers list. I have
written it down below. Thank you,
Kind regards,
Mélanie Karlsen
Hello nmusers,
I am currently working on automating the checks performed at each step of
the SCM forward inclusion pass. This requires setting up precise
quantitative thresholds. Once a candidate relationship has been identified
to be included in the model based on its LRT p-value, what is your strategy
in order to confirm covariate inclusion before moving to the next forward
inclusion step? Namely:
1. What threshold of omega decrease do you use for the omega on which
the covariate is included? If omegas of other parameters increase or
decrease, does this impact your decision? And if yes, how? (i.e what
threshold do you use?)
2. All other quantitative checks that you might perform on:
- the epsilon or residuals
- the extent of the covariate effect on the PK parameter
- other (please specify)
Thank you very much for your help.
Kind regards,
Mélanie Karlsen
PhD student @ LIRMM / Sanofi (Montpellier, France)
SCM forward inclusion checks
Dear Mélanie,
You are looking into applying other criteria (in addition p-value) for each
step of the forward search.
For convenience I would only look at the final SCM model in comparison with the
base model before SCM, which may then include several covariate-parameter
relations. In model finalization one can then consider deletion of covariate
relation that (everyone agrees) are clinically irrelevant. If there was an
automatic procedure for including this in the forward search, I think that
could sometimes be an appealing alternative.
Previously, there was functionality in PsN for omitting clinically irrelevant
covariate relations during the forward search, but I believe that functionality
has not been maintained. See publication by Tunblad et al., below.
I would not require that the magnitude of residual error is reduced by the
included covariates - not even if covariate(s) were time varying, but it is of
course encouraging seeing a lower residual-error magnitude.
Same goes for reduction in IIV. A small reduction is expected and provides more
confidence in the selection of covariates, but sometimes there could be a good
reason for no change in IIV. In particular, it may be OK to test covariates
also on parameters without IIV, if there is a sound mechanistic case. For
example, if data is sparse and (FOCE-based estimation) supports IIV on V but
not on the rate or absorption, a covariate for food effect should still be
tested on absorption parameters (and not on V).
Best regards
Jakob
Reference
Tunblad et al.: The use of clinical irrelevance criteria in covariate model
building with application to dofetilide pharmacokinetic data
Jakob Ribbing, Ph.D.
Principal Consultant & Client Operations Expert
[email protected]
+46(0)705-14 33 77
www.pharmetheus.com

Quoted reply history
> On 28 Oct 2024, at 14:15, Mélanie Karlsen <[email protected]> wrote:
>
> Hello,
>
> I am unable to get in touch with [email protected]
> <mailto:[email protected]> in order to join the network, therefore
> I am attempting to directly reach out to you, as I have a question that I
> would like to post to the nmusers list. I have written it down below. Thank
> you,
>
> Kind regards,
> Mélanie Karlsen
>
> Hello nmusers,
>
>
> I am currently working on automating the checks performed at each step of the
> SCM forward inclusion pass. This requires setting up precise quantitative
> thresholds. Once a candidate relationship has been identified to be included
> in the model based on its LRT p-value, what is your strategy in order to
> confirm covariate inclusion before moving to the next forward inclusion step?
> Namely:
>
> 1. What threshold of omega decrease do you use for the omega on which
> the covariate is included? If omegas of other parameters increase or
> decrease, does this impact your decision? And if yes, how? (i.e what
> threshold do you use?)
>
> 2. All other quantitative checks that you might perform on:
> - the epsilon or residuals
>
> - the extent of the covariate effect on the PK parameter
>
> - other (please specify)
>
> Thank you very much for your help.
>
> Kind regards,
>
> Mélanie Karlsen
>
> PhD student @ LIRMM / Sanofi (Montpellier, France)
>
--
*This communication is confidential and is only intended for the use of the
individual or entity to which it is directed. It may contain information
that is privileged and exempt from disclosure under applicable law. If you
are not the intended recipient please notify us immediately. Please do not
copy it or disclose its contents to any other person.*
*Any personal data
will be processed in accordance with Pharmetheus' privacy notice, available
here https://pharmetheus.com/privacy-policy/.**
*