# Re: Covariate modelling, covaraite effects

From: Gastonguay, Marc <marcg>
Date: Tue, 13 Apr 2010 22:08:56 -0400

Dear Joann,
If your goal is to learn about the the effect of CLCR on CLM2, you'll =
need to keep this effect in the model.

If your goal is to reduce IIV, then it appears you could exclude this =
effect, but you won't learn anything about the effect of CLCR on CLM2.

Keep in mind that your estimate of IIV may be biased or imprecise, and =
using the estimated value of this parameter as the only criterion for =
exclusion of a covariate effect is risky and leads to a model that is =
not very useful... especially when goodness of fit and clinical interest =
would lead you to include the covariate effect in the model.

Marc

On Apr 13, 2010, at 5:07 PM, joan hern wrote:

> Dear NMusers,
>
> I would like to have your help on the following issue:
>
> I am trying to model the PK of a parent compound and two metabolites =
(M1 is the major and M2 is the minor metabolite), both metabolites are =
mainly eliminated by renal excetion although a minor biliar excretion =
pathway exists. Creatinine clearance (CLCR) values in the studied =
population range form 10 to 100 mL/min. The plots of eta vs CLCR show a =
relationship in both cases The inclusion of CLCR on CLM1 as covariate =
decreases significantly the OFV value in 1779 units with respect to the =
base model, and a decrease of about 30% is observed in the IIV of this =
parameter (CLM1).
>
> In the case of the second metabolite (M2) the inclusion of CLCR on =
CLM2 decreases the OFV around 700 units but no reduction of IIV of CLM2 =
is observed when compared with the base model. For the second =
metabolite, reduction in the residual error with respect to the base =
model is very low (around 10%).
>
> The way I entered the covariates was in both cases
> TVCL= theta(x)*(CLCR/mean population CLCR value)**(theta(y)).
>
> In both cases, all the model parameters were estimated correctly. Then =
my question is if I should leave CLCR in CLM2 or I should remove it. I =