The dosage-regimen-dependent teratogenicity as well as plasma and tissue levels of the antiepileptic drug valproic acid (VPA) were studied in the mouse by comparing various injection regimens and infusion of the drug via implanted osmotic minipumps. Concentrations of 225-248 micrograms VPA/ml maternal plasma (about 2X above the therapeutic concentration range) and 70-75 micrograms VPA/g gestational material (on gestation Day 8) resulted in a significant incidence of neural tube defects (exencephaly in the mouse). Similar effects were produced if those concentrations were reached several times after multiple injections or by steady-state application via implanted pumps. A single injection was less effective than multiple injections, although drug accumulation did not occur. The doses (or area under the concentration-time curve values) did not correlate with the teratogenic response of the different administration regimens: much higher (factor 10) doses were needed with the infusion regimen to produce exencephaly rates comparable to those obtained with the injection regimen. The pattern of embryotoxicity was also schedule dependent: steady-state concentrations produced predominantly embryolethality and fetal weight retardation, while intermittent injections produced a high incidence of exencephaly (up to 60% of live fetuses). The dose of VPA (and the areas under the concentration-time curves) correlated with the embryolethality and fetal weight retardation of the drug, while the peak or steady-state concentrations reached in mother and gestational material correlated with the incidence of neural tube defects.