NONMEM Users Network Archive

Hosted by Cognigen

(no subject)

From: Buclin Thierry <Thierry.Buclin>
Date: Fri, 20 Nov 2015 17:44:17 +0000

Dear NMusers,

We have a small series of some 100 blood measurements in 13 neutropenic pat=
ients for a specific intravenous antibiotic. A population PK model for this=
 antibiotic has already been published, based on about 300 patients, probab=
ly with very few neutropenic cases (neutropenia has not been tested as a co=
variate). The 2 compartment model includes 4 PK parameters (CL, V1, V2, Q),=
 each with inter-individual variability, plus 7 coefficients applying to si=
gnificant covariates and 1 residual variability.

Rather than re-inventing a pop-PK model for our 13 neutropenic patients, we=
 just wanted to check whether the model already published applied to them. =
Then we planned to use model-derived MAP values for AUC and Cmin. So we fir=
st ran the model on these 13 patients with all 16 parameters fixed/bound to=
 their published reference value, and we obtained a value of the objective =
function (say OF_ref), along with estimates of AUC and Cmin. Thereafter, we=
 estimated all parameters (except one that needed to be kept fixed) and we =
obtained another value of the objective function (say OF_fit). The paramete=
r values obtained after fitting did not differ much from their published re=
ference values, comprised in +/- 2SE for all except one; however, most SEs =
were of course large due to our small number of patients. PRED vs OBS scatt=
er plots were not suggestive of bias; PRED_ref vs PRED_fit plots were well =
correlated. The fit decreased the objective function (OF_ref - OF_fit) by a=
bout 40 points.

We have a rather elementary question here : can this decrease in objective =
function be used to test statistically whether our 13 neutropenic patients =
globally differ from the population PK profile described in the published r=
eference model? Can we simply interpret the drop from OF_ref to OF_fit as f=
ollowing a Chi-square distribution with 15 degrees of freedom under the nul=
l hypothesis (i.e. neutropenic patients belong to the same population as th=
e reference)? Or should we privilege more sophisticated approaches to answe=
r this question? We searched the NMusers archive without finding much about=
 this point. Thanks in advance

Thierry Buclin, Aziz Chaouch
University Hospital CHUV, Lausanne, Switzerland

Received on Fri Nov 20 2015 - 12:44:17 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: