From: Nick Holford <*n.holford*>

Date: Wed, 16 Dec 2009 08:56:15 +1300

Mona,

Saik has replied to you about some of the technical issues related to

using NONMEM.

I would add that failure of the $COV step tells you nothing about the

suitability of your model. This has been discussed many times on

nmusers. It is in fact a failure of NONMEM which has been partially

fixed in NONMEM 7.

More generally the relationship between age and clearance( after

adjustment for differences in body size) in humans shows a very dramatic

INCREASE in utero and infancy (due to enzyme and organ maturation) to

reach adult values that PLATEAU in children and young adults with a slow

DECREASE in the elderly (perhaps due to declining renal function, poorer

nutrition, etc). Therefore you need to think about the reasons for the

age associated change when you try to use age to explain between subject

differences in clearance.

I would recommend that you use a sensible model to account for increases

in body size (weight) and then you can try empirically to see if AGE has

any influence (INCREASE, PLATEAU, DECREASE) (see Anderson & Holford 2008).

Another minor technical point:

This model:

MAGE = 40 ; Median age

FAGE = (AGE-MAGE)/MAGE ; Fractional effect of age

TVCL = THETA(1)* (1- THETA(2)*FAGE)

can be written:

MAGE = 40 ; Median age

FAGE = (AGE-MAGE) ; Fractional effect of age

TVCL = THETA(1)* (1- THETA(2)*FAGE)

It will produce identical results except for THETA(2) which can be

interpreted as the fractional change per year. This is sometimes easier

to convey to people because it is not dependent on the data driven and

therefore somewhat arbitrary choice of MAGE.

This linear model:

TVCL = THETA(1)* (1- THETA(2)*FAGE)

can cause problems if THETA(2) is positive and the difference between

AGE and MAGE is positive and large. This model will predict negative

values for TVCL.

A more robust model is:

TVCL = THETA(1)* EXP(THETA(2)*FAGE)

This ensures that TVCL is always positive (for any realistic value of

AGE). FAGE can also be interpreted as the fractional change per year

when changes are small because exp(x)~=1+x.

Both the linear and exponential models are empirical so there is no a

priori reason to prefer the linear model and so I prefer to use an

exponential model to describe the usually small changes in the elderly.

For neonates and infants more complex empirical models are needed to

describe the much greater changes in clearance (see Anderson & Holford 2008)

Best wishes,

Nick

Anderson BJ, Holford NH. Mechanism-based concepts of size and maturity

in pharmacokinetics. Annu Rev Pharmacol Toxicol. 2008;48:303-32.).

Mona.Alameddine

*> Dear nmusers,
*

*>
*

*> I am getting funny results when i add age as a covariate on oral
*

*> clearance:
*

*> The OF decreases by 16 points,
*

*> but :
*

*> BSV on CL increases from 35% to 37%
*

*> BSV on V decreases from 49% to 47%
*

*> Intra-Individual variability decreases from 41% to 38%
*

*>
*

*> And the most important is that the age had a positive effect on CL
*

*> Where CL INCREASED with age (THETA2 = -0.37 ). Age is not correlated
*

*> to other covariates and remained significant in the final model
*

*> together with the body weight.
*

*> I also tried several ways to test the effect of age , but NONMEM
*

*> didn't converge.
*

*>
*

*> I also had funny results when i tried FOCE-I method, because i got a
*

*> very small variability on oral CL (0.7%) though i had much lower OF
*

*> (-55 points)
*

*> and the Covariance step failed.
*

*>
*

*>
*

*> I used the following:
*

*>
*

*> $SUB ADVAN2 TRANS2
*

*>
*

*> $PK
*

*> MAGE = 40 ; Median age
*

*> FAGE = (AGE-MAGE)/MAGE ; Fractional effect of age
*

*> TVCL = THETA(1)* (1- THETA(2)*FAGE)
*

*> CL = TVCL * EXP (ETA(1))
*

*> V = THETA(3)* EXP (ETA(2))
*

*> KA = THETA(4)
*

*> F1 = THETA(5)
*

*>
*

*> S2=V/1000
*

*>
*

*> $THETA (0,10)(0.5)(0,100)(0,0.6)(1 FIX)
*

*>
*

*> $OMEGA 0.5 0.5
*

*>
*

*> $ERROR
*

*> Y=F*EXP(ERR(1))
*

*> IPRED=F
*

*> IRES=DV-IPRED
*

*> W=F; if err prop W=F else W=1
*

*> IWRES=IRES/W
*

*>
*

*>
*

*> $SIGMA 0.5
*

*>
*

*> $ESTIMATION SIGDIGITS=3 METHOD=1 MAXEVAL=9999 NOABORT POSTHOC PRINT=1
*

*>
*

*> $TABLE UNCONDITIONAL ID TIME AMT CL V KA DV PRED IPRED IRES RES WRES
*

*> IWRES
*

*> AGE SEX BW HGT ETA1 ETA2 EVID NOPRINT ONEHEADER FILE=run08.table.txt
*

*>
*

*>
*

*> Any suggestions on how to handle these results??
*

*>
*

*> Thank you in advance for your help
*

*>
*

*> Mona
*

*>
*

--

Nick Holford, Professor Clinical Pharmacology

Dept Pharmacology & Clinical Pharmacology

University of Auckland,85 Park Rd,Private Bag 92019,Auckland,New Zealand

tel:+64(9)923-6730 fax:+64(9)373-7090 mobile:+64(21)46 23 53

email: n.holford

http://www.fmhs.auckland.ac.nz/sms/pharmacology/holford

Received on Tue Dec 15 2009 - 14:56:15 EST

Date: Wed, 16 Dec 2009 08:56:15 +1300

Mona,

Saik has replied to you about some of the technical issues related to

using NONMEM.

I would add that failure of the $COV step tells you nothing about the

suitability of your model. This has been discussed many times on

nmusers. It is in fact a failure of NONMEM which has been partially

fixed in NONMEM 7.

More generally the relationship between age and clearance( after

adjustment for differences in body size) in humans shows a very dramatic

INCREASE in utero and infancy (due to enzyme and organ maturation) to

reach adult values that PLATEAU in children and young adults with a slow

DECREASE in the elderly (perhaps due to declining renal function, poorer

nutrition, etc). Therefore you need to think about the reasons for the

age associated change when you try to use age to explain between subject

differences in clearance.

I would recommend that you use a sensible model to account for increases

in body size (weight) and then you can try empirically to see if AGE has

any influence (INCREASE, PLATEAU, DECREASE) (see Anderson & Holford 2008).

Another minor technical point:

This model:

MAGE = 40 ; Median age

FAGE = (AGE-MAGE)/MAGE ; Fractional effect of age

TVCL = THETA(1)* (1- THETA(2)*FAGE)

can be written:

MAGE = 40 ; Median age

FAGE = (AGE-MAGE) ; Fractional effect of age

TVCL = THETA(1)* (1- THETA(2)*FAGE)

It will produce identical results except for THETA(2) which can be

interpreted as the fractional change per year. This is sometimes easier

to convey to people because it is not dependent on the data driven and

therefore somewhat arbitrary choice of MAGE.

This linear model:

TVCL = THETA(1)* (1- THETA(2)*FAGE)

can cause problems if THETA(2) is positive and the difference between

AGE and MAGE is positive and large. This model will predict negative

values for TVCL.

A more robust model is:

TVCL = THETA(1)* EXP(THETA(2)*FAGE)

This ensures that TVCL is always positive (for any realistic value of

AGE). FAGE can also be interpreted as the fractional change per year

when changes are small because exp(x)~=1+x.

Both the linear and exponential models are empirical so there is no a

priori reason to prefer the linear model and so I prefer to use an

exponential model to describe the usually small changes in the elderly.

For neonates and infants more complex empirical models are needed to

describe the much greater changes in clearance (see Anderson & Holford 2008)

Best wishes,

Nick

Anderson BJ, Holford NH. Mechanism-based concepts of size and maturity

in pharmacokinetics. Annu Rev Pharmacol Toxicol. 2008;48:303-32.).

Mona.Alameddine

--

Nick Holford, Professor Clinical Pharmacology

Dept Pharmacology & Clinical Pharmacology

University of Auckland,85 Park Rd,Private Bag 92019,Auckland,New Zealand

tel:+64(9)923-6730 fax:+64(9)373-7090 mobile:+64(21)46 23 53

email: n.holford

http://www.fmhs.auckland.ac.nz/sms/pharmacology/holford

Received on Tue Dec 15 2009 - 14:56:15 EST