NONMEM Users Network Archive

Hosted by Cognigen

Further questions about weight/CrCL correlations

From: David Nix <nix>
Date: Tue, 13 Jan 2009 10:18:27 -0700

*I'm very interested in comments from the group, since these issues are
very confused in the literature.

Schwartz equation:* CrCl (ml/min/1.73m2)= [length (cm) x k] / Scr
(Patient population: infants over 1 week old through adolescence (18
years old)) k = 0.45 for infants 1 to 52 weeks old
k = 0.55 for children 1 to 13 years old
k = 0.55 for adolescent females 13-18 years old
k = 0.7 for adolescent males 13-18 years old

When one uses this equation or others like, the CrCL is normalized to
body weight, and I agree that this is a good way to report an average
for a study population; however if you want to report the renal function
for an individual, I think that you need to get rid of the normalization

e.g. BSA = 0.5 m^2 ; CrCL = 70 ml/min/1.73 m^2, then
Renal function (CrCL) for the individual would be 20 ml/min

Likewise if CrCL were a covariate for NONMEM, you would want to have the
value for the individual.
If CrCL were 70 ml/min/1.73 m^2 for two individuals; one with 1.5 m^2
and another 2.0 m^2, I don't consider these indviduals as having the
same renal function. The covariate values would be 60.7 and 80.9
ml/min, respectively.


Another separate issue:

If estimated creatinine clearance is determined by an equation, eg:

CrCL = (140-age)/scr x WT/72 ...
WT is already considered in the equation and you would expect a
correlation between the value for CrCL and WT.
Normalizing the value of CrCL does not seem like a good solution for the
reason stated above.
The CrCL equation is really attempting to measure expected creatinine
production and then determine CL based on
       Xu (0-t)/(sCr x t) with the assumptions that 100% creatinine
produced is excreted and sCr is constant.
       Creainine production is related to body weight and theoretically
to lean body weight and muscle mass.

If the population were diverse in terms of age and sex, then the CrCL
values obtain may not correlate strongly with weight.
If a high correlation is observed between CrCL and WT, then would it be
possible to factor weight out

CrCL = (140-age)/scr would be the covariate, but then in the function
for clearance, add *weight/72 anywhere CrCL is used.
The covariate values would no longer correlate, but the same
relationship would be present.
I don't know if this would avoid the statistical problem of covariate
correlation or not.

Received on Tue Jan 13 2009 - 12:18:27 EST

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: