NONMEM Users Network Archive

Hosted by Cognigen

FW: VPC appropriateness in complex PK

From: Martin Bergstrand <martin.bergstrand>
Date: Mon, 21 Sep 2009 18:15:35 +0200

Dear NMusers,

For some reason my last message to NMusers got lost in www-space. Since =
Leonid and Nick have responded to my initial message I repost this =
so that you can follow the discussion (see email below).

In addition to this message I would also like to comment on the messages =
Diane, Leonid and Nick.

Nick and Leonid: I agree that it would be useful if one could also =
that adaptive design (e.g. dose adaptations) and show the observations =
the non transformed scale. However this will in many cases be very hard
since dos adaptations are often done not according to a strict algorithm
and/or all information supporting the dose alterations is not available. =
is to my experience quite commonly written I study protocols that dose
adjustments can be done “by the discretion of the investigator”.

Diane and Leonid: If I understood the SVPC procedure correctly from =
presentation it utilizes a principle similar to that behind Numerical
Predictive Check (NPC). Most of all SVPC seem to have a striking =
to the first version of the prediction discrepancies as described by =
et al (1). The prediction discrepancies have been further developed into =
normalised prediction distribution errors (NPDE) (2). From my experience
both NPC and NPDE are useful diagnostic tools but not applicable to data
from studies with adaptive dos adjustments (correlation between ETAs and
design). What is the unique feature with SVPC that sets it apart from =
prediction discrepancies and makes it applicable to studies with =
dos adjustments?

Nick: Regarding this sentence “The empirical PRED-corrected VPC does =
give this kind of support for future use of the PK model under an =
design scenario”. Why is this? If the PC-VPC can verify that you have =
acceptable structure model and unbiased parameter estimates you can then
simulate any type of adaptive design scenario.

Best regards,

1. Prediction discrepancies for the evaluation of … Mentré F, =
S. JPKPD. 2006
2. Computing normalised prediction distribution errors ... Comets E,
Brendel K, Mentré F. CMPB. 2008
From: Martin Bergstrand [mailto:martin.bergstrand
Sent: den 20 september 2009 19:32
To: 'Leonid Gibiansky'; 'Nick Holford'
Cc: 'Dider Heine'; 'nmusers
Subject: RE: [NMusers] VPC appropriateness in complex PK

Dear Leonid and Nick,

You have both written that there is no simulation based diagnostic that =
be applied in the case of adaptive designs (unless you can simulate the
adaptations). Below I will try to describe why I think that PC-VPCs can =
used under these circumstances.

The example that Leonid describe is very similar to one of the example =
the abstract about PC-VPCs that I referred to previously (see example =
With this example we demonstrate that PC-VPCs can be used in the =
presence of
adaptive designs such as TDM. The prediction corrected dependent =
variable in
a PC-VPC is unaffected by changes in independent variables included in =
model such as dose and covariate effects. It can be seen as if the =
median in
a PC-VPC represent a typical individual with a typical dose and a =
set of covariates. If we look at a prediction interval for a PC-VPC that
represent only the variability that is explained by random effects in =
model and nothing that comes from fixed effects (dose, covariates and =
For this reason PC-VPCs can be used also in the cases when we do not =
the exact algorithm for the adaptations made (e.g. dose adjustments). In =
very simple case where we have linear kinetics, no covariates in the =
and no binning across the independent variable on the x-axis (e.g. time)
PRED correction will be the same a dose normalization of both the =
and simulated data. However the PRED correction can be more universally
applied than a dose normalization. PRED correction does not handle all =
of adaptive designs that you could think of. For instance

The above described feature of PC-VPCs are one of reasons I find it =
In the cases with adaptive designs PC-VPCs will in my mind replace
traditional VPCs whereas in many other cases it will only be a =
complement to
stratified VPCs to better diagnose the random effect components of a =

More about this can be read in the ACoP abstract:
Bergstrand M, Hooker AC, Wallin JE, Karlsson MO. Prediction corrected
visual predictive checks ACOP. =

Ps. PRED correction does not handle all types of adaptive designs that =
could think of. For instance adaptive censoring of data (i.e. study
discontinuation) will not be this easily handled.

Kind regards,


Received on Mon Sep 21 2009 - 12:15:35 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: