Impact of pharmacokinetic information reported as being below limit of quantification on the prediction of important response endpoints.
Elba Romero1, Josep-María Cendrós 2, Concepción Peraire2, Rosendo Obach2, and Iñaki F. Trocóniz.1
1, Department of Pharmacy and Pharmaceutical Technology; School of Pharmacy; University of Navarra; Pamplona 31080; Spain.
Objectives: Pharmacokinetic (PK) and pharmacodynamic (PD) information are in occasions reported as being below limit of quantification (BQL). Most of the modelling experience with BQL information come from PK analysis where the potential impact on the model parameters estimation has been discussed.
In the current analysis we provide an example in the context of hormone-related tumours, where BQL values in pharmacokinetics, without major effect on PK model estimates, have a clear impact on important pharmacodynamic endpoints.
Methods: Fifteen male healthy subjects and eleven male cancer patients received a single subcutaneous injection of an GnRH agonist at three different dose levels and two type of formulations. Blood samples were taken over a period of six months and plasma concentrations of the GnRH agonist and the hormone testosterone (TST) were measured at the same sampling time. PKPD modelling was done sequentially using NONMEM VI.
Results: The percentage of BQL observations for the current analysis were 12.5% and 1.4% from the total of GnRH agonist (515) and TST (561) data, respectively. All of them were located after 20 days after drug injection.
PK and PKPD models were developed excluding BQLs from the analysis or treating BQLs as censored observations according to method 3 in Beal 2001[1].
PK of GnRH agonist was best described using a one compartment disposition model and an absorption model characterized by a simultaneous zero- and first-order absorption which included a delay absorption compartment.
The time profiles of TST were described using a variant of the pool-precursor model[2], where the GnRH agonist elicited a dual effect: (i) increasing the release of TST from the precursor compartment and (ii) blocking the synthesis of new precursors.
Regarding the most critical treatment endpoint, time of TST below 0.5 ng/mL (T<0.5), the models excluding BQLs of GnRH predicted a median T<0.5 of 120, 180, and 190 days for the low, middle, and high dose levels. When BQLs were treated as censored observations the corresponding values of T<0.5 were 65, 110, and 130 days, respectively.
Conclusions: The current analysis shows the importance of considering BQL samples in pharmacokinetics, even in cases where the impact on populations PK parameters is marginal. This methodology is especially indicated in those cases where there is a strong correlation between plasma drug concentrations and pharmacodynamic endpoints.
References:
[1] Beal SL. Ways to fit a PK model with some data below the quantification limit. Journal of Pharmacokinetics and Pharmacodynamics 28: 481-504 (2001).
[2] Movin-Osswald G and Hammarlund-Udenaes M. Prolactin release after remoxipride by an integrated pharmacokinetic-pharmacodynamic model with intra- and interindividual aspects. Journal of Pharmacology and Experimental Therapeutics 274: 921-927 (1995).
