This study was conducted to assess the utility of unbound brain EC50 (EC50,u) as a measure of in vivo potency for centrally active drugs. Seven mu-opioid agonists (alfentanil, fentanyl, loperamide, methadone, meperidine, morphine, and sufentanil) were selected as model central nervous system drugs because they elicit a readily measurable central effect (antinociception) and their clinical pharmacokinetics/pharmacodynamics are well understood. Mice received an equipotent subcutaneous dose of one of the model opioids. The time course of antinociception and the serum and brain concentrations were determined. A pharmacokinetic/pharmacodynamic model was used to estimate relevant parameters. In vitro measures of opioid binding affinity (Ki) and functional activity [EC50 for agonist stimulated guanosine 5'-O-(3-[35S]thio)triphosphate binding] and relevant clinical parameters were obtained to construct in vitro-to-preclinical and preclinical-to-clinical correlations. The strongest in vitro-to-in vivo correlation was observed between Ki and unbound brain EC50,u (r2 approximately 0.8). A strong correlation between mouse serum and human plasma EC50 was observed (r2 = 0.949); the correlation was improved when corrected for protein binding (r2 = 0.995). Clinical equipotent i.v. dose was only moderately related to Ki. However, estimates of ED50 and EC50 (total serum, unbound serum, total brain, and unbound brain) were significant predictors of clinical equipotent i.v. dose; the best correlation was observed for brain EC50,u (r2 = 0.982). For each opioid, brain equilibration half-life in mice was almost identical to the plasma effect-site equilibration half-life measured clinically. These results indicate that the mouse is a good model for opioid human brain disposition and clinical pharmacology and that superior in vitro-to-preclinical and preclinical-to-clinical correlations can be achieved with relevant unbound concentrations.