Dried blood spots and sparse sampling: a practical approach to estimating pharmacokinetic parameters of caffeine in preterm infants

Abstract

Aims: Dried blood spots (DBS) alongside micro-analytical techniques are a potential solution to the challenges of performing pharmacokinetic (PK) studies in children. However, DBS methods have received little formal evaluation in clinical settings relevant to children. The aim of the present study was to determine a PK model for caffeine using a 'DBS/microvolume platform' in preterm infants.

Methods: DBS samples were collected prospectively from premature babies receiving caffeine for treatment of apnoea of prematurity. A non-linear mixed effects approach was used to develop a population PK model from measured DBS caffeine concentrations. Caffeine PK parameter estimates based on DBS data were then compared with plasma estimates for agreement.

Results: Three hundred and thirty-eight DBS cards for caffeine measurement were collected from 67 preterm infants (birth weight 0.6-2.11 kg). 88% of cards obtained were of acceptable quality and no child had more than 10 DBS samples or more than 0.5 ml of blood taken over the study period. There was good agreement between PK parameters estimated using caffeine concentrations from DBS samples (CL = 7.3 ml h⁻¹ kg⁻¹; V = 593 ml kg⁻¹; t(½) = 57 h) and historical caffeine PK parameter estimates based on plasma samples (CL = 4.9-7.9 ml h⁻¹ kg⁻¹; V = 640-970 ml kg⁻¹; t(½) = 101-144 h). We also found that changes in blood haematocrit may significantly confound estimates of caffeine PK parameters based on DBS data.

Conclusions: This study demonstrates that DBS methods can be applied to PK studies in a vulnerable population group and are a practical alternative to wet matrix sampling techniques.

Figure 1

Goodness of fit plots for observed vs. population model predicted concentration (A) and observed vs. individual model predicted concentration (B)

Figure 2

Diagnostic goodness of fit plots of (A) weighted residual vs. model predicted concentration, (B) weighted residual vs. time (IWRES) after dose and (C) individual predictions against absolute individual weighted residuals. In plots (A) and (B) the dotted lines represent perfect agreement between observed and predicted concentrations. In plot (C) the line is a locally weighted polynomial regression (LOESS) smoothing line