While updating Users Guides VI (PREDPP), IV (NM-TRAN), and V
(Introduction) for NONMEM 7.4, I have noticed that the documentation
does not match the actual code of PREDPP.
PREDPP has always allowed output-type compartments. These are
compartments defined as (e.g.)
COMP=(METABURI,INITIALOFF,NODOSE)
Just as with the default output compartment, CMT may be negative
on an observation record, allowing the observation to be obtained,
and then the compartment turned off, with a single record.
Here are some details.
Chapter 6 of Guide V gives this example of data:
ID TIME EVID UVOL DV CMT AMT
1 6.00 1 0 0 1 100
1 8.00 2 0 0 3 0
1 9.50 0 75 .058 -3 0
1 9.50 2 0 0 3 0
1 11.75 0 100 .067 -3 0
1 11.75 0 100 5.80 2 0
With ADVAN2, compartment 3 is the default compartment for output,
and the observation at TIME 9.50 may have CMT=-3. But suppose a
general linear or non-linear model is used (ADVAN 5,6,7,8,9,13)
and there are more than 2 user-defined compartments.
If the $MODEL statement describes the 3d. compartment as simply
$MODEL COMPARTMENT=NAME, then the default compartment attributes
apply (initial on, off/on allowed, dose allowed) and the compartment
is not an output-type compartment. PREDPP produces this error
message for data record 3:
SPECIFIED COMPARTMENT MAY NOT BE TURNED OFF WITH AN OBSERVATION RECORD
There are two ways to avoid this error message. First, it is always
possible (for any compartment that may be turned off, even the
output compartment) to use two records instead of one, e.g., first
the observation, then a record with EVID=2 that turns off the
compartment:
1 9.50 0 75 .058 3 0
1 9.50 2 0 0 -3 0
This is what people have been doing.
Second, leave the data as-is, and change the compartment attributes
in $MODEL to INITIALOFF NODOSE. This was not documented, but was
always possible.
With this change to $MODEL, it is important that there be a record
in the data that turns the compartment on explicitly, rather than
relying on the fact that the compartment was initally on by default.
No existing problem specification that avoided the error message
by using extra event records is incorrect. It is simply that the
data files could have been left as-is and the $MODEL statements
changed instead.
--
Alison Boeckmann
[email protected]