Dear NM-users,
I have a question regarding NONMEM modeling of PK data from individuals
who vomited after drug ingestion. Basically, I have a set of data from
about 150 subjects (about 5 measurements per subject) who orally
ingested alcohol at time = 0 h. However, about 15% of this group
vomited more than 1 h post-dose. Based on NONMEM modeling of PK data
from subjects who did not vomit during the study, the mean elimination
half-life is about 30 min.
May I ask if I should model vomiting incidence (a "yes" or a "no") as a
categorical covariate for analysis of all the data? I have demographic
covariate information at hand. Or, should I introduce a mixture model
during NONMEM modeling? I would expect the vomiting subgroup to display
different values for Vd, Ka etc as compared to the non-vomiting
subgroup.
Thank you very much for your time and kind advice.
Best wishes,
Kok-Yong Seng, PhD
DSO National Laboratories
Republic of Singapore
PK data from individuals who vomited after drug ingestion
Dear NM-users,
I have a question regarding NONMEM modeling of PK data from individuals
who vomited after drug ingestion. Basically, I have a set of data from
about 150 subjects (about 5 measurements per subject) who orally
ingested alcohol at time = 0 h. However, about 15% of this group
vomited more than 1 h post-dose. Based on NONMEM modeling of PK data
from subjects who did not vomit during the study, the mean elimination
half-life is about 30 min.
May I ask if I should model vomiting incidence (a "yes" or a "no") as a
categorical covariate for analysis of all the data? I have demographic
covariate information at hand. Or, should I introduce a mixture model
during NONMEM modeling? I would expect the vomiting subgroup to display
different values for Vd, Ka etc as compared to the non-vomiting
subgroup.
Thank you very much for your time and kind advice.
Best wishes,
Kok-Yong Seng, PhD
DSO National Laboratories
Republic of Singapore
Kok-Seng,
I would not use a mixture model because you already have the information about which group vomited.
I am surprised you describe the elimination of ethanol with a half-life. If you gave an ethanol dose big enough to induce vomiting I would expect you to describe elimination with a mixed order (saturable) elimination process.
Best wishes,
Nick
Seng Kok Yong wrote:
> Dear NM-users,
>
> I have a question regarding NONMEM modeling of PK data from individuals who vomited after drug ingestion. Basically, I have a set of data from about 150 subjects (about 5 measurements per subject) who orally ingested alcohol at time = 0 h. However, about 15% of this group vomited more than 1 h post-dose. Based on NONMEM modeling of PK data from subjects who did not vomit during the study, the mean elimination half-life is about 30 min.
>
> May I ask if I should model vomiting incidence (a “yes” or a “no”) as a categorical covariate for analysis of all the data? I have demographic covariate information at hand. Or, should I introduce a mixture model during NONMEM modeling? I would expect the vomiting subgroup to display different values for Vd, Ka etc as compared to the non-vomiting subgroup.
>
> Thank you very much for your time and kind advice.
>
> Best wishes,
>
> Kok-Yong Seng, PhD
>
> DSO National Laboratories
>
> Republic of Singapore
--
Nick Holford, Professor Clinical Pharmacology
Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand
[email protected] tel:+64(9)923-6730 fax:+64(9)373-7090
mobile: +64 21 46 23 53
http://www.fmhs.auckland.ac.nz/sms/pharmacology/holford
Dear Kok-Yong Seng,
With my naive thinking what happens if you vomit is that a part or all of
what is remaining to be absorbed will disappear. If you know the time of
vomiting I can see a nice and mechanistic way of handling this.
>From your depot compartment you should in this case have an elimination
pathway instead of only an absorption pathway. As long as the subject
doesn't vomit the elimination pathway is turned off (elimination rate
constant=0). If and when a vomiting event is recorded the elimination
pathway is turned on for a short and fix amount of time (with the fix amount
of time a first order rate constant can conveniently be used to describe
this). I have provided a short coding example for this in the end of the
email.
If you for some reason doesn't have the exact time of the vomiting events I
would suggest a covariate for the number of vomiting events within the
relevant timeframe (while there is significant absorption). This covariate
should then be included for the bioavailability so that the relative
bioavailability is 1 for zero vomiting events and relatively different for
1, 2 and 3 vomiting events. This should however in my mind be a far cruder
way of handling this than my firs suggestion.
;; Mechanistic vomiting effect on absorption -------------------------
ID AMT TIME DV EVID CMT VOM(vomiting) Comments
1 100 0 . 1 1 0
Dosing
1 . 0.25 0.5 0 2 0
Observation before vomiting
1 . 0.4 . 4 0 0
Vomiting event (obs.VOM=0)
1 . 0.5 . 4 0 1 End
of vomiting event (obs.VOM=1)
1 . 0.5 1.25 0 1 0
Observation after vomiting
; In the above described case the vomiting covariate was turned on between
TIME 0.4-0.5.
$SUBROUTINE ADVAN5
$MODEL
COMP = (1_DEPOT) ; Depot compartment
COMP = (2_CENT) ; Central observation compartment
K1T2 = THETA(1) ; Absorption rate constant (KA)
IF(VOM.EQ.1) THEN
K1T0 = THETA(2) ; Elimination from depot compartment turned on at
vomiting event
ELSE
K1T0 = 0 ; No elimination from depot compartment
without vomiting
ENDIF
;; -------------------------------------------------------------------
>From the estimated K1T0 and the fixed time that VOM is set to 1 you can
calculate the average fraction of the remaining drug substance in the depot
compartment that is lost.
I hope that this is of some use to you.
Kind regards,
Martin Bergstrand, MSc, PhD student
-----------------------------------------------
Department of Pharmaceutical Biosciences,
Uppsala University
-----------------------------------------------
P.O. Box 591
SE-751 24 Uppsala
Sweden
-----------------------------------------------
[email protected]
-----------------------------------------------
Work: +46 18 471 4639
Mobile: +46 709 994 396
Fax: +46 18 471 4003
Quoted reply history
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Seng Kok Yong
Sent: den 27 augusti 2009 07:53
To: Nick Holford; nmusers
Subject: RE: [NMusers] PK data from individuals who vomited after drug
ingestion
Dear Nick,
Thank you for your reply. Yes, a single compartment model with
Michaelis-Menten elimination fits the data significantly better compared
to a single compartment model with linear elimination.
GAM analysis, using AIC, and covariate analysis within NONMEM do not
suggest that vomiting incidence is a significant covariate. Could this
be attributed to the fact that the systemic alcohol levels at the time
of vomiting (~1 h post-dose) is only about 25% of what they consumed
(assuming a elimination half-life of 30 min)? Or, could it be because
the number of vomiting cases makes up only 15% of the total number of
subjects?
Thank you and best wishes,
Kok-Yong Seng
-----Original Message-----
From: [email protected] [mailto:[email protected]]
On Behalf Of Nick Holford
Sent: Thursday, 27 August 2009 1:15 PM
To: nmusers
Subject: Re: [NMusers] PK data from individuals who vomited after drug
ingestion
Kok-Seng,
I would not use a mixture model because you already have the information
about which group vomited.
I am surprised you describe the elimination of ethanol with a half-life.
If you gave an ethanol dose big enough to induce vomiting I would expect
you to describe elimination with a mixed order (saturable) elimination
process.
Best wishes,
Nick
Seng Kok Yong wrote:
>
> Dear NM-users,
>
> I have a question regarding NONMEM modeling of PK data from
> individuals who vomited after drug ingestion. Basically, I have a set
> of data from about 150 subjects (about 5 measurements per subject) who
> orally ingested alcohol at time = 0 h. However, about 15% of this
> group vomited more than 1 h post-dose. Based on NONMEM modeling of PK
> data from subjects who did not vomit during the study, the mean
> elimination half-life is about 30 min.
>
> May I ask if I should model vomiting incidence (a "yes" or a "no") as
> a categorical covariate for analysis of all the data? I have
> demographic covariate information at hand. Or, should I introduce a
> mixture model during NONMEM modeling? I would expect the vomiting
> subgroup to display different values for Vd, Ka etc as compared to the
> non-vomiting subgroup.
>
> Thank you very much for your time and kind advice.
>
> Best wishes,
>
> Kok-Yong Seng, PhD
>
> DSO National Laboratories
>
> Republic of Singapore
>
--
Nick Holford, Professor Clinical Pharmacology
Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New
Zealand
[email protected] tel:+64(9)923-6730 fax:+64(9)373-7090
mobile: +64 21 46 23 53
http://www.fmhs.auckland.ac.nz/sms/pharmacology/holford
Dear NM-users,
I have found an important typo in the dataset example that I provided
previously. The vomiting event lines have EVID=4 but should really have
EVID=2. On top of this the last row should have CMT=2.
I am sorry for my sloppy writing.
Best regards,
Martin
; Corrected version -------------------------------------------------------
ID AMT TIME DV EVID CMT VOM MDV Comments
1 100 0 . 1 1 0 1 Dosing
1 . 0.25 0.5 0 2 0 0 Observation before vomiting
1 . 0.4 . 2 0 0 1 Vomiting event (obs.VOM=0)
1 . 0.5 . 2 0 1 1 End of vomiting event (obs.VOM=1)
1 . 0.5 1.25 0 2 0 0 Observation after vomiting
; -------------------------------------------------------------------------
Quoted reply history
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Martin Bergstrand
Sent: den 27 augusti 2009 11:13
To: 'Seng Kok Yong'; 'Nick Holford'; 'nmusers'
Subject: RE: [NMusers] PK data from individuals who vomited after drug
ingestion
Dear Kok-Yong Seng,
With my naive thinking what happens if you vomit is that a part or all of
what is remaining to be absorbed will disappear. If you know the time of
vomiting I can see a nice and mechanistic way of handling this.
>From your depot compartment you should in this case have an elimination
pathway instead of only an absorption pathway. As long as the subject
doesn't vomit the elimination pathway is turned off (elimination rate
constant=0). If and when a vomiting event is recorded the elimination
pathway is turned on for a short and fix amount of time (with the fix amount
of time a first order rate constant can conveniently be used to describe
this). I have provided a short coding example for this in the end of the
email.
If you for some reason doesn't have the exact time of the vomiting events I
would suggest a covariate for the number of vomiting events within the
relevant timeframe (while there is significant absorption). This covariate
should then be included for the bioavailability so that the relative
bioavailability is 1 for zero vomiting events and relatively different for
1, 2 and 3 vomiting events. This should however in my mind be a far cruder
way of handling this than my firs suggestion.
;; Mechanistic vomiting effect on absorption -------------------------
ID AMT TIME DV EVID CMT VOM(vomiting) Comments
1 100 0 . 1 1 0
Dosing
1 . 0.25 0.5 0 2 0
Observation before vomiting
1 . 0.4 . 4 0 0
Vomiting event (obs.VOM=0)
1 . 0.5 . 4 0 1 End
of vomiting event (obs.VOM=1)
1 . 0.5 1.25 0 1 0
Observation after vomiting
; In the above described case the vomiting covariate was turned on between
TIME 0.4-0.5.
$SUBROUTINE ADVAN5
$MODEL
COMP = (1_DEPOT) ; Depot compartment
COMP = (2_CENT) ; Central observation compartment
K1T2 = THETA(1) ; Absorption rate constant (KA)
IF(VOM.EQ.1) THEN
K1T0 = THETA(2) ; Elimination from depot compartment turned on at
vomiting event
ELSE
K1T0 = 0 ; No elimination from depot compartment
without vomiting
ENDIF
;; -------------------------------------------------------------------
>From the estimated K1T0 and the fixed time that VOM is set to 1 you can
calculate the average fraction of the remaining drug substance in the depot
compartment that is lost.
I hope that this is of some use to you.
Kind regards,
Martin Bergstrand, MSc, PhD student
-----------------------------------------------
Department of Pharmaceutical Biosciences,
Uppsala University
-----------------------------------------------
P.O. Box 591
SE-751 24 Uppsala
Sweden
-----------------------------------------------
[email protected]
-----------------------------------------------
Work: +46 18 471 4639
Mobile: +46 709 994 396
Fax: +46 18 471 4003
-----Original Message-----
From: [email protected] [mailto:[email protected]] On
Behalf Of Seng Kok Yong
Sent: den 27 augusti 2009 07:53
To: Nick Holford; nmusers
Subject: RE: [NMusers] PK data from individuals who vomited after drug
ingestion
Dear Nick,
Thank you for your reply. Yes, a single compartment model with
Michaelis-Menten elimination fits the data significantly better compared
to a single compartment model with linear elimination.
GAM analysis, using AIC, and covariate analysis within NONMEM do not
suggest that vomiting incidence is a significant covariate. Could this
be attributed to the fact that the systemic alcohol levels at the time
of vomiting (~1 h post-dose) is only about 25% of what they consumed
(assuming a elimination half-life of 30 min)? Or, could it be because
the number of vomiting cases makes up only 15% of the total number of
subjects?
Thank you and best wishes,
Kok-Yong Seng
-----Original Message-----
From: [email protected] [mailto:[email protected]]
On Behalf Of Nick Holford
Sent: Thursday, 27 August 2009 1:15 PM
To: nmusers
Subject: Re: [NMusers] PK data from individuals who vomited after drug
ingestion
Kok-Seng,
I would not use a mixture model because you already have the information
about which group vomited.
I am surprised you describe the elimination of ethanol with a half-life.
If you gave an ethanol dose big enough to induce vomiting I would expect
you to describe elimination with a mixed order (saturable) elimination
process.
Best wishes,
Nick
Seng Kok Yong wrote:
>
> Dear NM-users,
>
> I have a question regarding NONMEM modeling of PK data from
> individuals who vomited after drug ingestion. Basically, I have a set
> of data from about 150 subjects (about 5 measurements per subject) who
> orally ingested alcohol at time = 0 h. However, about 15% of this
> group vomited more than 1 h post-dose. Based on NONMEM modeling of PK
> data from subjects who did not vomit during the study, the mean
> elimination half-life is about 30 min.
>
> May I ask if I should model vomiting incidence (a "yes" or a "no") as
> a categorical covariate for analysis of all the data? I have
> demographic covariate information at hand. Or, should I introduce a
> mixture model during NONMEM modeling? I would expect the vomiting
> subgroup to display different values for Vd, Ka etc as compared to the
> non-vomiting subgroup.
>
> Thank you very much for your time and kind advice.
>
> Best wishes,
>
> Kok-Yong Seng, PhD
>
> DSO National Laboratories
>
> Republic of Singapore
>
--
Nick Holford, Professor Clinical Pharmacology
Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New
Zealand
[email protected] tel:+64(9)923-6730 fax:+64(9)373-7090
mobile: +64 21 46 23 53
http://www.fmhs.auckland.ac.nz/sms/pharmacology/holford