NONMEM Users Network Archive

Hosted by Cognigen

Units of SIGMA estimates for urine data scaled by urine volume

From: Sarah F Cook <cook.sarah>
Date: Tue, 18 Aug 2015 12:04:07 -0400

Hi all,

I am working on a PK model that includes plasma and urine data for
parent drug and some metabolites.

The DV for the urine data are provided in concentrations (mg/L) and
scaled according to each urine sample volume (mL) as outlined in the
NONMEM Users Guide - Part V - Introductory Guide (Chapter 6, Section
9, The Output Compartment: Urine Collections and Observations; and
Chapter 7, Section 4.3.3, Scaling by a Data Item). To define the urine
sample collection time, the urine compartments are turned on with
EVID=2 and CMT=n at the start time, and turned off with EVID=2 and
CMT=-n at the end time. So the data records for each urine sample look
something like the following:

0.01 2 5 . .
3 0 5 9.03 22
3.005 2 -5 . .

And for this example, the $PK statement includes:


My understanding is that NONMEM should then be scaling any
record-specific output (e.g., PRED, IPRE, RES) so that the units are
in mg/L.

My question relates to the SIGMA estimates for the urine compartments.
I have an additive component for the residual unexplained variability
of the urine compartments. Are the SIGMA estimates also scaled so
that, when expressed as a standard deviation, they should also be
interpreted as concentrations (mg/L)? Or are they not scaled so that,
when expressed as a standard deviation, they should be interpreted as
amounts (mg)? The former makes more sense to me, but I’m rather

A review of the literature has turned up mixed reporting practices in
this regard, so any advice would be greatly appreciated.

Kind regards,
Sarah Cook

Sarah F. Cook
Ph.D. Candidate
University of Utah
Department of Pharmacology & Toxicology
Center for Human Toxicology
30 South 2000 East, Room 105
Salt Lake City, UT 84112
E-mail: cook.sarah
Phone: (570) 362-2981
Received on Tue Aug 18 2015 - 12:04:07 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: