TABLE 2.
PK model building comparisona
| Model | –2LL | AIC | Bayesian bias (μ/mL) (central) | Bayesian imp (μg/mL2) (central) | R2 Bayesian (central) | Bayesian bias (μg/mL) (pup kidneys) | Bayesian imp (μg/mL2) (pup kidneys) | R2 Bayesian (pup kidneys) |
|---|---|---|---|---|---|---|---|---|
| Two-compartment base model | 712.2 | 723.3 | –0.278 | 1.04 | 0.626 | 0.135 | 0.134 | 0.906 |
| Three-compartment base model | 543.2 | 558.8 | –0.576 | 1.39 | 0.965 | 0.791 | 1.2 | 0.97 |
| Three-compartment model adjusted for trimesterb,d (as depicted in Fig. S1) | 534.0 | 559.9 | –0.019 | 1.25 | 0.978 | 0.021 | 0.011 | 0.999 |
| Three-compartment model adjusted for TBW (VDdams × TBW/0.3075c)d | 540.3 | 566.2 | –0.157 | 1.16 | 0.952 | −0.0724 | 0.0352 | 0.998 |
a
PK, pharmacokinetic; −2LL, −2 log-likelihood; AIC, Akaike information criterion; imp, imprecision; VD, volume of distribution; VD1, volume term 1; VD2, volume term 2; TBW, total body weight; dam, female mother; pup, offspring.
b
Final model based on overall performance on regression of observed versus predicted concentrations, visual plots of parameter estimates, −2LL/AIC, and rule of parsimony. The clinical significance of the variable was also considered in evaluation.
c
Median weight of the dams.
d
Models were not statistically different.