NONMEM Users Network Archive

Hosted by Cognigen

Re: ETAPvalues

From: Jeroen Elassaiss-Schaap <jeroen>
Date: Mon, 31 Aug 2015 11:10:38 +0200

Dear Ahmad,

The ETA P-value is one of the diagnostics available for modelers. What it me=
ans and how important it is depends on your modeling objective and data. Hav=
ing said that, in this case it seems that you have fixed thetas and omegas b=
ut left parent concentrations in the dataset. The lower ETA P-value indicate=
s that the parent posthoc ETA distribution has shifted away from zero. This m=
ost likely means that your metabolite submodel has 'pulled' the parent posth=
oc away. You might want to look at the parent posthoc graphs to further inve=
stigate how the model shifted.

Generally speaking this could mean that the metabolite model has some mismat=
ch that is compensated by the parent etas. You might be able to update the m=
etabolite (or parent) model to avoid this. If needed, you could also use a s=
lightly different approach where you completely fix the parent model by appe=
nding your original posthocs to your dataset and apply them as a fixed input=
 to the metabolite model and adjust. Afterwards you can see whether the upd=
ated model has improved ETA P-values.

Hope this helps!

Best regards,
+31 6 23118438
-- More value out of your data!

> Op 31 aug. 2015 om 03:18 heeft Abu Helwa, Ahmad Yousef Mohammad - abuay010=
> Dear Nonmem users,
> My question is about ETAPval. I am modelling a parent-metabolite data us=
ing the 2-stage approach. After modelling the parent drug alone, parameter e=
stimates (for the parent) were fixed in the metabolite model and metabolite p=
arameters were estimated. All ETAPvalules for metabolite parameters were acc=
eptable > 0.2 and all metabolite model parameters were estimated precisely; h=
owever, some of the ETAPvalues of parent drug parameters became < 0.05 while=
 they were acceptable when the parent drug was modelled alone.
> Should ETAPvalues for parent parameters stay above the acceptable limit ev=
en if the parameters are fixed or we care only about the ETAPvals for estima=
ted parameters only?
> N.B. Simultaneous modelling of the parent-metabolite wasn’t possib=
> Thank you,
> Ahmad Abuhelwa
> Email: ahmad.abuhelwa
> Mobile: +61 413118743

Received on Mon Aug 31 2015 - 05:10:38 EDT

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to:

Once subscribed, you may contribute to the discussion by emailing: