Bioequivalence and Bioavailability Forum

Dear Lena!

❝ We got the data from our pilot clinical trials: data sets of plasma concentration for:

❝ 1) IV bolus 1000 mg (female);

❝ 2) IR tablets 500 mg (2 times/day, 0-12, 12-24, steady state 14 days, male);

❝ 3) CR tablets 1000 mg (once a day, steady state 14 days, male);

❝ 4) CR tablets 1000 mg (steady state 5 days, female).

❝ The data are very inconsistent.

Inconsistent in which respect – can you give us more information? Do you have so called ‘rich datasets’ (frequent sampling at least on profile days) or ‘sparse data’ (just a few samples per subject)? In the former case I would start with NCA in order to get preliminary estimates of PK metrics (AUC, V/F, λ_z, t_lag,…). In the latter case WinNonlin offers a sparse sample NCA option. You can use these estimates as initial estimates in a PK model.
You will get more insight on the behavior of the drug if you try to fit a Population PK model. There you can include covariates (body weight/surface area, sex, creatinine clearance for renaly excreted unchanged drugs, …). Generally you start with the same model as in ‘classical’ PK and subsequently add covariates until the explained variance does not improve any more. Involves a lot of trial-and-error; patience mandatory.

❝ Can WinNonlin help me …

Which version of WinNonlin do you have? If you have one of the classical versions of WinNonlin I would strongly suggest to update to Phoenix for free as soon as possible. Much easier there.
I would also recommend to register at Pharsight’s Extranet. There you can even post your datasets and might get help in setting up models by Pharsight’s staff and experienced users.

❝ … to understand the behavior of our drug? I mean, perhaps I can get any model or any calculations to understand where the active substance is absorbed.

Models tell you little about the site(s) of absorption. You have to know your API and formulation: ionisation state (absorption window), enteric coating (t_lag), release properties of the CR (zero order resulting in flip-flop PK: k_a < k_el)…