TABLE 4

Stepwise covariate modeling and ANOVA

Parameter	Covariate^{^a}	Δdf	ΔOFV^{^b}	p (ΔOFV)^{^c}	p (ANOVA)^{^d}
CL_6αOH/fm_pac (l/h)	G2677T/A	2
Final model			−7.38	0.025	0.012
Alternative model 1^{^e}			−4.79	0.091	0.048
Alternative model 2^{^f}			−5.86	0.054	0.033
CL_pac (l/h)	CYP3A5*3	1
Final model			−5.91	0.015	0.024
Alternative model 1			−2.96	0.085	0.12
Alternative model 2			−0.37	0.54	0.57
CL_diOH/fm_met (l/h)	CYP3A4*1B	1
Final model			−3.52	0.061	0.13
Alternative model 1			−3.35	0.067	0.14
Alternative model 2			−3.13	0.077	0.15
CL_pac (l/h)	CYP2C8*1C	1
Final model			−3.42	0.064	0.070
Alternative model 1			−3.42	0.064	0.082
Alternative model 2			−0.01	0.93	0.91

↵a Covariates significant at p < 0.1 in the final model using stepwise covariate modeling.
↵b For a model with fixed estimates instead of priors for paclitaxel parameters.
↵c p value corresponding to the drop in OFV.
↵d Using the empirical Bayes estimates from the corresponding model supported by priors.
↵e Identical to the final model but also using IIV, supported by priors (Henningsson et al., 2001), on the distribution volumes V₁ (coefficient of variation 46%, df 28) and V₂ (CV 40%, df 13) for unbound paclitaxel.
↵f Identical to alternative model 1 but without IIV on the parameter B_lin for plasma protein binding. The model structure is hierarchical to alternative model 1 but not to the final model.