Successful pharmacokinetic modeling often requires the ability of a simple model to describe a complex series of physiological processes. However, a simple model may be inappropriate. Physiologically-relevant modeling may offer a more appropriate description, but requires further support from in vitro/in vivo data. A well-stirred hepatic model with linear processes was proposed to describe in vivo disposition of acetaminophen and metabolites after a 100 mg/kg bolus of acetaminophen to vehicle- or phenobarbital-pretreated, renal-ligated rats. Model simulations underpredicted acetaminophen glucuronide (AG) concentrations at early time points in serum, and were inconsistent with AG biliary excretion-rate profiles. Intracellular binding of AG by ligandin was hypothesized, and a cytosolic compartment with reversible binding was incorporated into the model. In this second model, only AG bound in the cytosolic compartment was available for excretion into bile. Model 2 better described the AG biliary excretion rate-time profiles based on calculated Akaike's information criterion values. However, no apparent change was observed in the underprediction of AG serum concentrations. Parameter estimates derived from the two models also were different. The rate constants regulating AG formation and sinusoidal egress were increased significantly after phenobarbital pretreatment according to model 1, while the AG biliary excretion rate constant was decreased significantly. Parameter estimates based on model 2 suggested that phenobarbital pretreatment impaired the cytosolic binding of AG but increased significantly the AG biliary excretion rate constant. The physiologic relevance of model 2 was not supported by a subsequent investigation of the protein binding and subcellular distribution of acetaminophen and metabolites. Acetaminophen, AG and acetaminophen sulfate (AS) were not bound extensively in hepatic cytosol (mean +/- SD unbound fractions were 0.90 +/- 0.08, 0.97 +/- 0.08, and 0.88 +/- 0.06, respectively). Phenobarbital pretreatment did not alter significantly the unbound fractions of acetaminophen, AG or AS in hepatic cytosol. Acetaminophen was distributed to a greater extent in lysosomes than in the nuclear, mitochondrial, microsomal and cytosolic fractions. Distribution of AS predominated in cytosolic and lysosomal fractions. AG was detected only in cytosol. Phenobarbital pretreatment decreased the content of acetaminophen, AG and AS in all hepatic fractions. This study demonstrates the utility of pharmacokinetic modeling in exploring mechanistic hypotheses. However, these results underscore the importance of obtaining pivotal data from in vitro/in vivo studies to validate hypothesized mechanisms.