Skip to content
Prev 413 / 20628 Next

nlme and NONMEM

Hi All,

This thread reminds me of an experience using nlme about 10 years ago. I was
remodeling a previously analyzed (and published) pharmacokinetic data set on
the drug remifentanil; NONMEM had been used to estimate a 3 compartment (6
parameter) model. The data included multiple plasma concentration values for
each subject.

Using nlme, no convergence was possible for a three compartment model
despite various choices of the control language and covariance structure. A
two compartment model converged.

Doug provided very useful tips to me at the time. For example, a visual
inspection of the raw data (time course of remifentanil concentration decay)
revealed only one inflection point in the decay curves for most subjects,
whereas two inflection points would be consistent with a three compartment
model. The data was not sufficient to fit a three compartment model.

I have never used NONMEM. Speaking to associates using NONMEM in the 90s,
they assured me that NONMEM could always be tweaked to converge. This was
considered a virtue.

This is an example NONMEM allowing overparametrized models.

Nathan