Table 1. Parameters used to simulate C_P/d4T-TP concentrations in activated PBM cells.

Population pharmacokinetics (Panhard et al., 2007)	Estimate (%CV)	95% CI
N	39 subjects
V/F (L)	23.9	13.4-42.6
Cl/F/ (L/h)	15.9	10.4-24.4
IND	1.53	0.91-2.57
ka (h-1)	0.452	0.311-0.656
ω2CL/F	(74.0)	57.1-96.0
ω2V/F	(80.6)	50.5-128.7
ω2Ka	0	1 (fixed)
σ (%)	37.7	32.7-43.4
a (ng/ml)	110
Cellular parameters		Reference
Cell volume (pL)	0.32	(Diamond et al., 2004)
Cells stimulated
PHA stimulated	40 %	(Jacobsson et al., 1995)
in humans	8 %	(Mohri et al., 2001)
Phosphorylation in vivo		95% CI
KPP (h-1)	1.45	1 (fixed)
KDP (h-1)	0.526	1 (fixed)
σ2 (%)	58
Mol. weight of d4T	242.2

Model fitting and simulations were performed using the ADVAN9 differential equation routine of NONMEM (6.2, ICON Development Solutions, Ellicott City, MD). CL and V are systemic clearance and distribution volumes, respectively and are divided by F, the unknown fraction of drug which is orally absorbed. K_a is the first-order oral absorption rate constant. IND is the potentiating factor of CL/F. Thus, when d4T is coadministered with indinavir, CL/F was boosted by a factor of 53% (Panhard, et al., 2007). K_PP and K_DP are first-order rate constants which describe d4T-TP accumulation and decay, respectively, and were obtained by co-fitting the pooled d4T-TP concentration to the median plasma concentration versus time curve. Pharmacokinetic parameters were assumed to be log-normally distributed with variances ω². %CV of a log-normally distributed parameter = √(e^(ω2)-1) × 100. The residual variance σ² in plasma concentrations of d4T was modeled using a combined proportional and additive error structure, commonly used in population pharmacokinetics. According to this error structure residual error becomes proportional to concentration and tends to “a” at lower concentrations (Panhard et al., 2007). The residual variance of the d4T-TP values ( ${\sigma }_2^2$) was modeled using a proportional error structure. Thus, observed [d4T-TP] = “true” [d4T-TP] × (1 + ε), where ε is normally distributed with variance $={\sigma }_2^2$. The model fitted for the in vivo cellular pharmacology of d4T-TP converged to > 3 decimal places, and produced a minimum value of the NONMEM objective function (-2 × log-likelihood, -2LL) = 278, using the FOCE with interaction option.

¹These parameters were considered invariant in the model.