NONMEM Users Network Archive

Hosted by Cognigen

Re: Your suggestions/thoughts needed on allometric base or final model

From: Nick Holford <n.holford>
Date: Sun, 13 Jul 2008 21:32:00 -0700


As Steve Duffull has pointed out you can decide to be an empiricist and
ignore all prior biological knowledge and try to estimate empirically an
allometric coefficient or you can put trust in prior knowledge which
means PK parameters such as clearance will increase with weight.
Empiricists will often misinterpret their statistical tests to conclude
there is no association between weight and PK parameters when in fact
the weight distribution is inadequate or they have not properly
acccounted for important factors such as body composition.

If you believe in biology then it is foolish in most cases to attempt to
estimate an allometric coefficient because the estimate will be biased
unless you have a very informative weight distribution and good
estimates of PK parameters (dont bother if you are relying on sparse PK
sampling methods). See Anderson, B.J. and N.H. Holford,
Mechanism-Based Concepts of Size and Maturity in Pharmacokinetics. Annu
Rev Pharmacol Toxicol, 2008. 48: p. 303-332. for a discussion of the
problem and experimental evidence of the difficulties in assessing
allometric coefficients.

Body weight is always important (in adults and children) even if the
data set you are studying is inadequate to reject a null hypothesis
because of unsuitable design.


In my opinion all PK *base* models will include allometric weight
scaling of clearance and volume. If you ignore weight then is is like
ignoring dose in PK models. Both weight and dose are fundamental
covariates for predicting drug concentrations.


Bhattaram, Atul wrote:
> Hello Hong-Guang
> It is always a good idea to estimate the allometric coefficient if you
> have adequate (weight ranges, PK sampling etc) data collected. If
> body weight is not important (although that is rare in pediatrics),
> then it need not be included in the model.
> Atul
> Venkatesh Atul Bhattaram
> Pharmacometrics
> US Food and Drug Administration
> "The contents of this message are mine personally and do not
> necessarily reflect any position of the Government or the Food and
> Drug Administration."
> ------------------------------------------------------------------------
> *From:* owner-nmusers
> [mailto:owner-nmusers
> *Sent:* Friday, July 11, 2008 12:36 PM
> *To:* nmusers
> *Subject:* [NMusers] Your suggestions/thoughts needed on
> allometric base or final model
> Dear NMusers:
> As you know, body weight is an important covariate that is
> integrated into the final or covariate model in some cases. When
> analyzing pediatric pop PK data, body weight-based allometric
> power model is used frequently. By definition, base model is a
> model without any covariates. But, in the literature on the
> population PK in pediatrics, I noted that body weight is added to
> the structural model (following the principles of allometry)
> before starting the covariate model building in some but not in
> all studies. That means that some models are called allometric
> base models and others are not. What are their differences? For
> the allometric base model, body weight has been added into the
> base model regardless of whether it is an important covariate (in
> some cases, body weight is not). If body weight is not an
> important covariate as determined by further covariate model
> building, is there still the need to add body weight into the
> final allometric model (if its corresponding base model is one
> without a body weight-associated allometric component)? Logically,
> such a need seems to be not reasonable. How to deal with this
> conflict? Is there an almost agreeable thought on this issue in
> our community?
> Thank you,
> Hong-Guang

Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand

Received on Mon Jul 14 2008 - 00:32:00 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: