- On 24 Sep 2001 at 10:21:49, "Shemonayev" (shemonayev.aaa.farlep.net) sent the message

Back to the Top

Hello,

Present time, I'm working with coordination compound of Germanium with =

biolegandes (Nicotinic acid, Nicotinamide and Succinic acid). I have =

calculated already pharmacokinetics data for onecompatmental analysis =

and now I'm looking for program for twocompartmental analysis.

Any recommendation will much appreciated.

Katerina Shemonayeva

Odessa state medical university (Ukraine)

Department general pharmacology

Shemonayev.-at-.farlep.net - On 25 Sep 2001 at 10:40:19, Mark Lovern (mlovern.-a-.Pharsight.com) sent the message

Back to the Top

Dear Katerina,

WinNonlin is a state-of-the-art pharmacokinetics software package that

offers the functionality you describe. The software includes a pre-written

library of Pharmacokinetic and Pharmacodynamic models, including 1- and 2-

compartment PK models for IV bolus, IV infusion, and extravascular (1st

order absorption) data. Three compartment IV bolus and infusion models are

also included in the library.

If the model of interest is not included in our pre-built library of models,

you have the freedom to write your own model using WinNonlin's native

modeling language. Alternatively, models may also be written and compiled

in either FORTRAN or C++.

We offer a number of programs whereby users at academic institutions may

obtain copies of our software at substantially reduced prices, or in some

cases, at no charge at all. To inquire about these opportunities, please

contact our sales team at sales.-a-.pharsight.com.

Please let me know if you have any questions, or if there is some way that I

may be of further assistance.

With Best Regards,

Mark R. Lovern, Ph. D.

Pharsight Scientific Support

Phone: +44 (0) 208 323 8402

Mobile: +44 (0) 781 243 0555

FAX: +44 (0) 208 323 8415

Argentum Center

2 Queen Caroline Street

London, England W6 9DX - On 25 Sep 2001 at 17:19:54, "David S. Farrier" (DFarrier.-a-.SummitPK.com) sent the message

Back to the Top

The following message was posted to: PharmPK

Katerina,

An option to fitting simple compartments is to calculate pharmacokinetic

parameters using noncompartmental or model-independent methods. The best

and most comprehensive example of this approach is PK Solutions, which

produces some 75 pharmacokinetic results with a few mouse clicks.

You are invited to get a demo of PK Solutions and a free summary of

pharmacokinetic equations from http://www.SummitPK.com

Regards,

David

David S. Farrier, Ph.D. Phone: 970-249-1389

Summit Research Services Fax:: 970-249-1360

68911 Open Field Dr. Email: DFarrier.aaa.SummitPK.com

Montrose, CO 81401 Web: http://www.SummitPK.com - On 25 Sep 2001 at 21:28:58, David_Bourne (david.aaa.boomer.org) sent the message

Back to the Top

[One reply and a note regarding a virus alert false alarm - db]

From: Nick Holford

Date: Wed, 26 Sep 2001 11:15:47 +1200

To: david.-at-.boomer.org

Subject: Re: PharmPK Re: Pharmacokinetic software search

The following message was posted to: PharmPK

"David S. Farrier (by way of David Bourne)" wrote:

> An option to fitting simple compartments is to calculate pharmacokinetic

> parameters using noncompartmental or model-independent methods.

What do you mean by "model-independent" methods? The

non-compartmental approaches for estimation of clearance and Vss

model-dependent (i.e. they assume first-order elimination from a

central compartment). Estimation of terminal half-life assumes a

log-linear model which is equivalent to the first-order assumption.

--

Nick Holford, Divn Pharmacology & Clinical Pharmacology

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

email:n.holford.at.auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556

http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm

---

From: "David S. Farrier"

Date: Tue, 25 Sep 2001 18:13:16 -0600

To: david.-a-.boomer.org

Subject: Re: PharmPK Pharmacokinetic software (NOTE)

The following message was posted to: PharmPK

***False Alarm***

Earlier today, I sent the message below to the PharmPK list. Two people

called to tell me that they received a virus warning when they accessed our

www.SummitPK.com site. This was a false alarm and has been corrected. The

web site is clean and is inspected daily by web hosting personnel.

Apparently a short JavaScript section included in the source HTML code that

scrolls text in the status bar triggered the alert. The script has been

there for 4 years, but we removed it anyway.

Use of JavaScript is quite common and I am sure we will all see a lot more

false alarms until the anti-virus definition for the Nimbda worm is made

more specific.

In sum, if you got a warning, it was a false alarm! The site is clean,

always was and always will be. So go have a look at the PK toys and software.

Thanks

David S. Farrier, Ph.D.

Director, Summit Research Services

http://www.SummitPK.com

DFarrier.at.SummitPK.com - On 26 Sep 2001 at 10:28:31, "David S. Farrier" (DFarrier.at.SummitPK.com) sent the message

Back to the Top

Nick,

It seems that Gibaldi and Perrier (Pharmacokinetics, 2nd ed., Marcel

Dekker, 1982) wrestled with the same semantics. On page 409 of their

classic text, they referred to "non-compartmental methods" as usually based

on AUC information. But it is evident from their treatment of the sum of

exponential terms describing conc-time data that this approach also fits

under the umbrella of "noncompartmental" approaches.

"Model-independent" and "noncompartmental" are seemingly synonymous, both

implying that the math is based on something other than curve fitting to

find a best fit for a presumed compartmental model. Yet, as you point out,

even area measurements, if they are carried to infinity, or the mere

calculation of a half life both assume linear first order kinetics, which

in turn implies a simple one-compartment model, at least.

Perhaps "model-independent" and "noncompartmental" are really used by most

people to imply calculation methods that are based on something other than

applying curve-fitting procedures to find the best compartmental match for

the data.

Again on page 409, Gibaldi and Perrier made this point: "Noncompartmental

methods do not require the assumption of a specific compartmental model for

either drug or metabolite. In fact, these methods can be applied to

virtually any compartmental model, provided that we can assume linear

pharmacokinetics [re: your observation]. Noncompartmental methods are

hardly new. However, the idea that noncompartmental methods provide a

general approach for pharmacokinetic analysis is both new and important.

During the preparation of this edition of "Pharmacokinetics" [book title],

there has been a distinct shift away from computer-based curve-fitting of

experimental data and elaboration of compartmental models and towards

noncompartmental methods of analysis."

Hence, PK Solutions has succeeded in encapsulating the breadth of Gibaldi

and Perrier's exponential terms and area calculation approaches to provide

some 75 pharmacokinetic parameters based on "noncompartmental" methods,

while leaving it to WinNonLin and other software to advance the

compartmental curve-fitting approach.

A survey of the literature shows that most papers on general

pharmacokinetics tabulate parameters arrived at by noncompartmental

methods, not to minimized the importance of curve-fitting to compartmental

models, when needed. Since PK Solutions produces more noncompartmental

parameters from a set of blood level data than any software available, we

like to quote one of our customers who put it this way:

"PK Solutions fills the gap for noncompartmental pharmacokinetic analysis."

I think the distinction between noncompartmental and compartmental

approaches is important to keep in mind. So thanks for the opportunity to

enter the fray. Perhaps others will want to make a new thread on this topic.

Warm Regards,

David

David S. Farrier, Ph.D. Phone: 970-249-1389

Summit Research Services Fax:: 970-249-1360

68911 Open Field Dr. Email: DFarrier.aaa.SummitPK.com

Montrose, CO 81401 Web: http://www.SummitPK.com - On 27 Sep 2001 at 18:28:08, Nick Holford (n.holford.-a-.auckland.ac.nz) sent the message

Back to the Top

The following message was posted to: PharmPK

"David S. Farrier (by way of David Bourne)" wrote:

> An option to fitting simple compartments is to calculate pharmacokinetic

> parameters using noncompartmental or model-independent methods.

What do you mean by "model-independent" methods? The

non-compartmental approaches for estimation of clearance and Vss

model-dependent (i.e. they assume first-order elimination from a

central compartment). Estimation of terminal half-life assumes a

log-linear model which is equivalent to the first-order assumption.

--

Nick Holford, Divn Pharmacology & Clinical Pharmacology

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

email:n.holford.at.auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556

http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm - On 28 Sep 2001 at 18:15:53, "Hans Proost" (J.H.Proost.aaa.farm.rug.nl) sent the message

Back to the Top

The following message was posted to: PharmPK

Dear Dr. Farrier,

I don't have a comment on your statements about

noncompartmental pharmacokinetics, in your reply to Dr. Holford.

However, I have some comments on what you did not say.

What is the use of noncompartmental PK? You are right that it can

produce a lot of numbers, similarly to compartmental PK, but with

less assumptions. That's OK. But PK is more than producing

numbers. In many cases we want to make a prediction of the time

course of plasma concentration profile, eg. for dosage regimen

calculation. Data from noncompartmental modeling can be used to

determine the dosing rate for maintaining a steady state

concentration (in case of iv infusion) or an average concentration (in

case of multiple dosing with equal intervals). But what more can be

predicted from noncompartmental PK data? How about the dosing

interval and prediction of peak and trough levels?

In my view, noncompartmental PK is useful only in cases where

the data do not allow compartmental PK, and in such cases the

credibility of the data may be questionable.

Any comments to these somewhat provoking statements are

welcomed.

Best regards,

Hans Proost

Johannes H. Proost

Dept. of Pharmacokinetics and Drug Delivery

University Centre for Pharmacy

Antonius Deusinglaan 1

9713 AV Groningen, The Netherlands

tel. 31-50 363 3292

fax 31-50 363 3247

Email: j.h.proost.at.farm.rug.nl - On 1 Oct 2001 at 13:33:06, David Bourne (david.-a-.boomer.org) sent the message

Back to the Top

[A few replies - db]

From: "Gobburu, Jogarao V"

Date: Sat, 29 Sep 2001 13:18:32 -0400

To: david.aaa.boomer.org

Subject: RE: PharmPK Re: Pharmacokinetic software search

Status: R

The following message was posted to: PharmPK

Dear Dr. Proost,

In my opinion, both noncompartmental (NCA) and compartmental (CA) methods

are almost inevitable for any given problem. I would not take the stand that

NCA should ONLY be used when CA fails. Firstly, if we have observational

data, it is impossible to analyse the data without some prior knowledge. Let

us focus on experimental data, which I think is more relevant to this

discussion. As we all might appreciate, for nonlinear regression initial

estimates are critical. The only method to obtain reasonable initial

estimates is indeed NCA. One might do it formally or informally by

eye-balling, nevertheless we all do it. At the same time, we all, probably

appreciate the fact that CA offers a very powerful advantage over NCA -

generalizability.

Regarding dosing interval calculations, I think we should also take into

account the pharmacodynamics and disease progression. Hence it is a more

complicated issue and CA is inevitable.

My previous comment about initial estimates holds good for PD too.

Regards,

Joga Gobburu,

Pharmacometrics,

CDER, FDA

---

From: Roger Jelliffe

Date: Sat, 29 Sep 2001 12:49:23 -0700

To: david.-a-.boomer.org

Subject: Re: PharmPK Re: Pharmacokinetic software search

Status: R

Dear Dr. Proost:

Thanks for your comments on noncompartmental analysis. I

agree with you 100%.

Very best regards,

Roger Jelliffe

Roger W. Jelliffe, M.D. Professor of Medicine, USC

USC Laboratory of Applied Pharmacokinetics

2250 Alcazar St, Los Angeles CA 90033, USA

Phone (323)442-1300, fax (323)442-1302, email= jelliffe.aaa.hsc.usc.edu

Our web site= http://www.usc.edu/hsc/lab_apk

---

From: "David S. Farrier"

Date: Sun, 30 Sep 2001 00:54:36 -0600

To: david.aaa.boomer.org

Subject: Re: PharmPK Re: Pharmacokinetic software search

Status: R

Johannes,

Thank you for your viewpoint. I did not mean to become the

protagonist for "noncompartmental" pharmacokinetics, but I will, if

elected. I sometimes ponder how significant it is to define a

compartment, the physiological significance of which is astutely

denied. Compartments seem forever "theoretical".

Our goal at SummitPK.com was not to take sides, but to help the sides

taken. Once, I tabulated the types of parameters reported and the

calculation methods employed by authors off "pharmacokinetics" papers

in the Journal of Pharmacy and Drug Metabolism and Disposition during

the five year period of 1992-1997. I discovered that 99% of the

papers used "noncompartmental" methods to report a set of some 10-20

common PK parameters deemed sufficient (by authors and peer reviews

alike) to described the pharmacokinetics of a drug. Interestingly,

there was no consensus of software employed. Most authors admitted to

using ad hoc spreadsheet calculation.

That is when we decided that the goal of PK Solutions was to provide

a set of noncompartmental results that would have served all those

authors over 5 years of published articles in two journals as the

only software needed to publish their papers.

Strict compartmental analysis is definitely useful, but by the looks

of it, only to a few.

Respectfully,

David S. Farrier

http://www.SummitPK.com

The place where you can find a great noncompartmental PK analysis

program for pros and novices alike.

David S. Farrier, Ph.D. Phone: 970-249-1389

Summit Research Services Fax:: 970-249-1360

68911 Open Field Dr. Email: DFarrier.aaa.SummitPK.com

Montrose, CO 81401 Web: http://www.SummitPK.com

---

From: "Stephen Duffull"

Date: Mon, 1 Oct 2001 12:53:15 +1000

To: david.-a-.boomer.org

Subject: RE: PharmPK Re: Pharmacokinetic software search

Status: R

The following message was posted to: PharmPK

Hi

Just a comment to agree with Hans and to comment on one of his points...

> say. What is the use of noncompartmental PK? You are right

> that it can produce a lot of numbers, similarly to

> compartmental PK, but with less assumptions.

It is a common assumption that non-compartmental PK has less

assumptions. I'm not convinced that this is true at all. It is not

really "non-compartmental" since it requires the underlying

"compartmental" model to fulfil certain characteristics for AUC and CL

to retain their meaningful relationship (see Nicks previous comment)

[this is a circular problem]. I prefer to use the term "non-parametric"

PK. Non-parametric PK provides a summary of the outputs of a PK system

(similar to non-parametric statistics).

However if the observations are not close to the time of the "true"

Cmax, or the Cmin was poorly quantified, or the log-linear portion of

conc-time curve poorly characterized, or there are non-linearities in

the system, or the actual sampling time differs significantly from the

nominal sampling time then our estimates of Tmax, Cmax, AUC, AUMC,

tz(1/2), MRT etc etc will have significant error in them. Since these

non-parametric PK metrics are only of use for descriptive (rather than

predictive) purposes then do they really have less assumptions than a

model used for descriptive purposes only? Indeed one could go so far as

to develop an empirical model (eg sum of exponentials) and the

non-parametric PK metrics calculated from the model predicted

obervations rather than the actual observations [now this is getting

very circular]...

I wonder, if non-parametric PK were compared to parametric PK for

descriptive purposes that the "lack of assumptions" would indeed be

superior... [I personally doubt it]

Kind regards

Steve

Stephen Duffull

School of Pharmacy

University of Queensland

Brisbane, QLD 4072

Australia

Ph +61 7 3365 8808

Fax +61 7 3365 1688

http://www.uq.edu.au/pharmacy/duffull.htm

---

From: exfamadu.-a-.savba.sk

Date: Mon, 1 Oct 2001 15:08:48 +0200

To: david.-at-.boomer.org

Subject: Re: PharmPK Re: Pharmacokinetic software search

The following message was posted to: PharmPK

Data from noncompartmental modeling can be used to

> determine the dosing rate for maintaining a steady state

> concentration (in case of iv infusion) or an average concentration (in

> case of multiple dosing with equal intervals). But what more can be

> predicted from noncompartmental PK data? How about the dosing

> interval and prediction of peak and trough levels?

Dear Dr. Proost,

In our study:

Dedik L., Durisova M., Batorova M.: Weighting function used for

adjustment of multiple-bolus drug dosing, Meth. Find. Exp. Clin.

Pharmacol., 2000, 22, 543-549,

we presented utilization of results of a specific type of noncompartmental

modeling for the adjustment of the multiple-bolus dosing of a drug. This

method allows to determine patient-specific loading and maintenance

bolus doses of the drug, necessary to reach and maintain prescribed

trough levels of the drug in a patient at desired time-points, both (the levels

and time-points) specified by treatment requirements. The method is

exemplified by the adjustment of the multiple-bolus dosing of factor VIII in

treatment of hemophilia A.

With best regards,

Maria Durisova

Maria Durisova, Ph.D., D.Sc.,

Senior Research Worker and

Scientific Secretary

Institute of Experimental Pharmacology

Slovak Academy of Sciences

842 16 Bratislava

Dubravska cesta 9

Slovak Republic

Phone/Fax: 00421 2 5477 5928

http://nic.savba.sk/sav/inst/exfa/durisova.htm - On 1 Oct 2001 at 20:43:11, Nick Holford (n.holford.-at-.auckland.ac.nz) sent the message

Back to the Top

The following message was posted to: PharmPK

> From: "Stephen Duffull"

> I prefer to use the term "non-parametric"

> PK. Non-parametric PK provides a summary of the outputs of a PK system

> (similar to non-parametric statistics).

I don't agree that non-parametric is a useful term in this context.

The term non-parametric in a hypothesis testing context refers to not

requiring to assume a specific *parametric* distribution. The term

non-compartmental in a pharmacokinetic context refers to not

requiring to assume a specific *compartmental* structure.

The assumptions of NCA are similar, but more restrictive (e.g. only

linear systems make sense to estimate Vss), than compartmental

analysis.

NCA is used to predict parameters (CL, Vss, etc) so what makes it

non-parametric?

--

Nick Holford, Divn Pharmacology & Clinical Pharmacology

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

email:n.holford.-a-.auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556

http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm - On 2 Oct 2001 at 12:57:53, David_Bourne (david.at.boomer.org) sent the message

Back to the Top

[Two replies - db]

From: "Hans Proost"

Date: Tue, 2 Oct 2001 16:42:21 MET

To: david.-a-.boomer.org

Subject: Re: Pharmacokinetic software search

The following message was posted to: PharmPK

Dear colleagues,

Thank you for the many replies to my provoking comments to

noncompartmental analysis (NCA). Actually I agree with each of

these replies, and I think that the position of NCA is somewhat

clarified.

IMHO there is at least one useful application of NCA. The

estimation of clearance (or CL/F for extravascular dosing) from

Dose/AUC is the most simple and robust method (unless a

considerable part of AUC must be obtained by extrapolation). If this

estimate of clearance differs markedly from one obtained from a

very sophisticated CA, I probably would prefer the NCA estimate.

Or perhaps better, I would prefer none of them.

With respect to terminology, the term NCA is actually broader than

I had in mind. The approach described by Dr. Maria Durisova and

colleagues is indeed NCA. As I understand, they describe drug

behavior using mathematical equations which cannot be translated

to compartmental models (i.e. not polyexponentials). This may

work, as may be concluded from their examples, but it remains

difficult, if not impossible, to understand these functions in a

physiological context. The attractive feature of compartmental

modeling is that it helps understanding what happens in the body.

For example, in case of impaired renal function, clearance may be

expected to decrease parallel with creatinine clearance, and dose

corrections can be predicted quantitatively. Such predictions under

various conditions are allowed since the compartmental modeling

has at least some physiological meaning. Of course, we should not

overinterpret the model, and we should always be aware of its

limitations. However, it makes sense.

Best regards,

Hans Proost

Johannes H. Proost

Dept. of Pharmacokinetics and Drug Delivery

University Centre for Pharmacy

Antonius Deusinglaan 1

9713 AV Groningen, The Netherlands

tel. 31-50 363 3292

fax 31-50 363 3247

Email: j.h.proost.-at-.farm.rug.nl

---

From: "Stephen Duffull"

Date: Tue, 2 Oct 2001 17:00:46 +1000

To: david.-at-.boomer.org

Subject: RE: PharmPK Re: Pharmacokinetic software search

The following message was posted to: PharmPK

Nick

> > From: "Stephen Duffull"

> > I prefer to use the term "non-parametric"

> > PK. Non-parametric PK provides a summary of the outputs of a PK

> > system (similar to non-parametric statistics).

>

> I don't agree that non-parametric is a useful term in this

> context.

That's fine - this is just terminology. We know that NCA does require

compartmental assumptions (eg linearity is an obvious one - but also it

is not appropriate to assume the AUC-CL relationship for a 2 cpt Rowland

model either...)

> The term non-parametric in a hypothesis testing

> context refers to not

> requiring to assume a specific *parametric* distribution. The term

> non-compartmental in a pharmacokinetic context refers to not

> requiring to assume a specific *compartmental* structure.

Non-parametric simply means no parameters. Whether this is for

statistics or not. NCA is a method for summarising data (ie

concentrations) into various metrics (eg Tmax, Cmax, AUC, AUMC etc).

Whether the user then wants to convert these to "parameters" from some

unknown model is their own choice. We also know that compartmental

structure needs to be assumed for NCA derived "parameters" to have

meaning. This obviously differs from the IO model where we try and

estimate models and parameters that match the inputs to the outputs (ie

a parametric PK analysis).

> NCA is used to predict parameters (CL, Vss, etc) so what makes it

> non-parametric?

It is non-parametric simply because to get the summary "statistics" of

the concentrations you do not have to have parameters. Similarly you do

not need parameters to describe a distribution of (any) data using

median or percentiles.

Cheers

Steve

Stephen Duffull

School of Pharmacy

University of Queensland

Brisbane, QLD 4072

Australia

Ph +61 7 3365 8808

Fax +61 7 3365 1688

http://www.uq.edu.au/pharmacy/duffull.htm - On 4 Oct 2001 at 14:46:51, David Bourne (david.-a-.boomer.org) sent the message

Back to the Top

[A few more replies - db]

From: "Hans Proost"

Date: Tue, 2 Oct 2001 16:42:21 MET

To: david.aaa.boomer.org

Subject: Re: Pharmacokinetic software search

Status: R

The following message was posted to: PharmPK

Dear colleagues,

Thank you for the many replies to my provoking comments to

noncompartmental analysis (NCA). Actually I agree with each of

these replies, and I think that the position of NCA is somewhat

clarified.

IMHO there is at least one useful application of NCA. The

estimation of clearance (or CL/F for extravascular dosing) from

Dose/AUC is the most simple and robust method (unless a

considerable part of AUC must be obtained by extrapolation). If this

estimate of clearance differs markedly from one obtained from a

very sophisticated CA, I probably would prefer the NCA estimate.

Or perhaps better, I would prefer none of them.

With respect to terminology, the term NCA is actually broader than

I had in mind. The approach described by Dr. Maria Durisova and

colleagues is indeed NCA. As I understand, they describe drug

behavior using mathematical equations which cannot be translated

to compartmental models (i.e. not polyexponentials). This may

work, as may be concluded from their examples, but it remains

difficult, if not impossible, to understand these functions in a

physiological context. The attractive feature of compartmental

modeling is that it helps understanding what happens in the body.

For example, in case of impaired renal function, clearance may be

expected to decrease parallel with creatinine clearance, and dose

corrections can be predicted quantitatively. Such predictions under

various conditions are allowed since the compartmental modeling

has at least some physiological meaning. Of course, we should not

overinterpret the model, and we should always be aware of its

limitations. However, it makes sense.

Best regards,

Hans Proost

Johannes H. Proost

Dept. of Pharmacokinetics and Drug Delivery

University Centre for Pharmacy

Antonius Deusinglaan 1

9713 AV Groningen, The Netherlands

tel. 31-50 363 3292

fax 31-50 363 3247

Email: j.h.proost.aaa.farm.rug.nl

---

From: Roger Jelliffe

Subject: Re: PharmPK Re: Pharmacokinetic software search

Mime-Version: 1.0

Dear All:

Let's talk a bit about the term "nonparametric". Often this

is used in place of the word "noncompartmental". However, to my

knowledge, the parameters in noncompartmental models are usually

described in terms of means and variances, with the usual implied

assumption of Gaussian or lognormal parameter distributions. Because

of this, it may nor be useful or correct to describe noncompartmental

models as being nonparametric.

The word nonparametric (NP) is also used in connection with

describing the parameter distributions in conventional compartmental

models, either linear or nonlinear. In this regard, the NP approaches

make no parametric assumptions about the shape of the distribution of

the parameters, as do the usual parametric methods, which usually

assume that the parameter distributions are either normal or

lognormal, and therefore are fully described by the parameter means

and variances that one gets. In contrast, the nonparametric approach,

first done by Alain Mallet, (nonparametric maximum likelihood - NPML)

makes no such assumptions about the shape of the parameter

distributions - the shape, whatever it is, is determined solely by

the data of doses and responses in the population. The approach finds

the distribution having the maximum likelihood given the raw data and

the weighting scheme used. Alan Schumitzky has also developed the

nonparametric expectation - maximization (NPEM) approach. They both

get basically the same results from the same raw data. More recently,

Bob Leary at the San Diego Supercomputer Center has developed the

nonparametric adaptive grid (NPAG) approach, which is considerably

more efficient and much faster, with results having greater

likelihood. All these methods are examples of nonparametric analysis

of models which rave definite structure, usually compartmental

models. They are designed to find the most likely parameter

distributions given the raw data. Usually, the likelihood obtained by

the nonparametric methods is greater that that seen with parametric

methods, as they are designed to exactly that, unconstrained by any

assumptions of normality or lognormality or anything like that. In

addition, the NP methods are mathematically consistent. They have the

property that the more you sample from a population, the more closely

the results approach the true results in the population. Parametric

methods may have smaller population coefficients of variation about

the population parameter means. Usually the results are less likely,

though, as they are constrained by the assumed shape.

David Bourne, if you will accept an attachment to this

message, I will attach a powerpoint file for a recent roundtable

discussion, which describes this a bit more for those who are

interested, in the context of optimal individualized drug therapy for

which the population models provide the framework (the Bayesian

prior) for planning the optimal initial dosage regimen.

[Sorry Roger no attachments...in this case especially it seems to be

too big - db]

Notice also, that when one uses Bayes' theorem to get

Bayesian posteriors from the measured responses or concentrations,

that you then get a nonparametric Bayesian posterior joint density.

The new method of "multiple model" dosage design is then specifically

designed to achieve target goals with maximum precision, whether this

based on the Bayesian prior ( the NP population model) or the

posterior joint parameter density.

I look forward to discussing this more with those who are

interested. To my knowledge, these methods, based on nonparametric

models as the Bayesian prior, can develop the most precise dosage

regimens I know, for the optimal attainment of target goals, for

drugs having narrow margins of safety, where we need to have such

precision.

Best regards to all,

Roger Jelliffe

---

From: "Dr. Ibrahim Wasfi"

Subject: Re: PharmPK Re: Pharmacokinetic software search

To: PharmPK.at.boomer.org

MIME-version: 1.0

X-Priority: 3

Hans Proost wrote

"The

estimation of clearance (or CL/F for extravascular dosing) from

Dose/AUC is the most simple and robust method (unless a

considerable part of AUC must be obtained by extrapolation"

Could you please be specific ( percentage ! ) about " unless a considerable

...."

Regards,

Ibrahim A. Wasfi

Forensic Science Laboratory

P O Box 253, Abu Dhabi

United Arab Emirates

Tel + 4092522

Fax + 4463470

---

From: exfamadu.-a-.savba.sk

Date: Thu, 4 Oct 2001 11:40:18 +0200

To: david.-a-.boomer.org

Subject: Re: PharmPK Re: Pharmacokinetic software search

The following message was posted to: PharmPK

Dear Dr. Hans Proost,

> With respect to terminology, the term NCA is actually broader than

> I had in mind.

There is a good paper:

K. H. Norwich: Noncompartment models of whole-body clearance of

tracers: A review, Annals of Biomedical Engineering, 25, 1997, 421-439.

> The approach described by Dr. Maria Durisova and

> colleagues is indeed NCA. As I understand, they describe drug

> behavior using mathematical equations which cannot be translated

> to compartmental models (i.e. not polyexponentials). This may

> work, as may be concluded from their examples, but it remains

> difficult, if not impossible, to understand these functions in a

> physiological context.

The models used in our work are called the transfer-function models.

Time-domain responses of these models to mathematically described

drug inputs may have the form of polyexponentials.

All the linear compartment models can be written in the form of the

transfer-function models. An example of a relationship between a

compartment model and a transfer-function model is described in detail in

our study:

M.Durisova, L.Dedik, et al., Pharmacokinetics of Factor VIII in hemophilia

A patients assessed by frequency response method, Methods and

Findings in Experimental and Clinical Pharmacology, 20, 1998, 217-

226. This study also shows examples of bio-medical purport of parameters

of transfer-function models. Other examples can be found in our other

studies.

With best regards,

Maria Durisova

Maria Durisova, Ph.D., D.Sc.,

Senior Research Worker and

Scientific Secretary

Institute of Experimental Pharmacology

Slovak Academy of Sciences

842 16 Bratislava

Dubravska cesta 9

Slovak Republic

Phone/Fax: 00421 2 5477 5928

http://nic.savba.sk/sav/inst/exfa/durisova.htm

---

From: Prah.James.aaa.epamail.epa.gov

Subject: Re: PharmPK Re: Pharmacokinetic software search

To: david.-at-.boomer.org

MIME-version: 1.0

Dr. Bourne,

Given clearance and dose by one route can dose, given AUC and clearance, be

calculated for a different route such as dermal in the same subject?

Thanks,

Jim

James D. Prah, PhD

US EPA

Human Studies Division MD (58B)

Research Triangle Park, NC, 27711

919 966 6244

919 966 6367 FAX - On 9 Oct 2001 at 14:30:01, David Bourne (david.-a-.boomer.org) sent the message

Back to the Top

[Two replies - db]

From: "Hans Proost"

Date: Mon, 8 Oct 2001 09:41:55 MET

To: david.at.boomer.org

Subject: Re: Pharmacokinetic software search

The following message was posted to: PharmPK

Dear Dr. Wasfi,

In reply to my comment:

> "The

> estimation of clearance (or CL/F for extravascular dosing) from

> Dose/AUC is the most simple and robust method (unless a

> considerable part of AUC must be obtained by extrapolation"

you wrote:

> Could you please be specific ( percentage ! ) about " unless a considerable

> ..."

Your question is clear, but the answer is more complicated.

Consider the following example:

The AUC from time zero until the last sampling time is 80 (arbitrary

units). If this AUC is determined using linear or log-linear

trapezoidal rule, if it is calculated from a reasonable number of

measurement (say 10 or more) that are well-spread over the time

scale, this value may be considered rather accurate. The precision

of the calculated AUC may be estimated from the measurement

error.

If the concentration at the last sampling point is not zero (if fact, it

will never be zero, but let's stay practical for this moment), then the

AUC must be extrapolated from:

AUC(t_last to infinitity) = C_last / k_terminal

where k_terminal is the estimated elimination rate constant

obtained over, say, the last four measurements.

Let us assume that the extrapolated AUC is 20. One can be sure

that this value is less precise than the AUC(0 to t_last), since it

includes the error in C_last and in k_terminal. The precision of the

extrapolated AUC may be estimated from the measurement error in

C_last and the precision (SD) of the estimated k_terminal.

Thus, the total AUC is 100, and the extrapolated part is 20% of

total AUC. In general, one may expect that the accuracy in the

total AUC is 'reasonable'. However, it should be noted that the

value '20%' is dependent on the precision of the extrapolation.

If the extrapolated part of the AUC is considerably larger than 20%,

the situation becomes much worse. Therefore, as a rule of thumb,

one should not use AUC values if the extrapolated part exceeds

20%. However, this value of 20% is rather arbitrary.

The precision of the total AUC can be estimated from the precision

of both parts of the AUC. This procedure allows a fair judgement of

the impact of the extrapolation on the reliability of the estimated

total AUC. This procedure provides a better basis for judgment than

whatever 'rule of thumb'.

I don't know the exact regulatory guidelines at this moment, but the

aforementioned reasoning may be helpful in estimating the

precision of total AUC.

Sincerely,

Hans Proost

Johannes H. Proost

Dept. of Pharmacokinetics and Drug Delivery

University Centre for Pharmacy

Antonius Deusinglaan 1

9713 AV Groningen, The Netherlands

tel. 31-50 363 3292

fax 31-50 363 3247

Email: j.h.proost.aaa.farm.rug.nl

---

From: Roger Jelliffe

Date: Mon, 08 Oct 2001 10:20:33 -0700

To: david.aaa.boomer.org

Subject: Re: PharmPK Re: Pharmacokinetic software search

Dear David:

Thanks for including my comments. Sorry you cannot include

the attachment. For those who are interested in nonparametric

population modeling, there is more material on our web site.

www.lapk.org

Most is under the heading of teaching topics.

Very best regards to all,

Roger Jelliffe

Want to post a follow-up message on this topic? If this link does not work with your browser send a follow-up message to PharmPK@boomer.org with "Pharmacokinetic software search" as the subject

PharmPK Discussion List Archive Index page

Copyright 1995-2010 David W. A. Bourne (david@boomer.org)