Niclas Jonsson, Janet R. Wade and Mats O. Karlsson
Division of Pharmacokinetics, Uppsala University, BMC, Box 580, Uppsala, S-751 23, Sweden.
Population pharmacokinetic analysis is a commonly used tool to analyse routine clinical data. This type of data is characterised by a relative sparseness in the number of observations. It has been proposed that it would be beneficial if the samples were collected randomly in all the individuals who are to be studied (1). Such randomness will be hard to achieve if the patients studied are outpatients taking part in a phase III study. In such a case it is more likely that the sampling times will vary within certain sampling windows, circumscribed by the visiting hours at the study clinic. The relatively controlled circumstances during a clinical study makes it possible, though, to enforce some sampling time restrictions within the sampling windows, without causing too many practical problems to the staff.
The present study evaluates the benefits of taking two samples instead of one, during a visit to a clinic in a phase III study. The samples were drawn immediately the patient arrived at the clinic and just before leaving it. Factors influencing the time span between the two samples were the time the patient had to wait to see the clinician, the length of the examination and the number of other activities the patient had to go through, i.e. EGG, lab tests, etc. Designs with different time spans between the two samples were compared to single sample designs.
Different study designs were evaluated by the comparison of the parameter estimates, the capability of the design to find the true pharmacokinetic model and the ability to predict individual parameter estimates. Both simulated and real data were used to assess the various study designs.
The results show that the error in pharmacokinetic parameter estimation will decrease when a second sample is included, even if the samples are replicates at the same time point. Although none of the studied designs could find the true pharmacokinetic model, addition of a second sample made it possible to detect a more complex model. The two sample designs were also better at predicting the individual parameter values. One consequence of this is that it will be easier to find relationships between the parameters and demographic factors such as age and weight (2).
[1] L. B. Sheiner, S. L. Beal, J Pharmacokinet Biopharm 11, 303-319 (1983).
[2] J. W. Mandema, D. Verotta, L. B. Sheiner, J Pharmacokinet Biopharm 20, 511-28 (1992).
Reference: PAGE 3 (1994) Abstr 862 [www.page-meeting.org/?abstract=862]
Poster: poster