A cross-over study design was used to determine the pharmacokinetics of ampicillin in swine. Each of eight pigs was subjected to all of the following three treatments: (1) intramuscular (i.m.) injection of 17.6 mg/kg of ampicillin trihydrate; (2) injection of a mean dose of 17.6 mg/kg of ampicillin trihydrate using a needle-free (NF) injection device; and (3) intravenous injection of 17.6 mg/kg of sodium ampicillin administered as a bolus. Ampicillin trihydrate administered by NF injection in this study was not statistically different from i.m. injection as measured by AUC(0-infinity), MRT, MAT, or Cmax. However, the 90% confidence limits about the difference in NF to i.m. mean Cmax and AUC(0-infinity) values, expressed relative to the i.m. treatment mean, exceeded the traditional bioequivalence limits of +/-20%. In part, failure to demonstrate bioequivalence was attributable to small study size and the large within-subject variability associated with this drug. Therefore the power of this study was not sufficient to definitively prove or disprove bioequivalence and additional studies to describe appropriate dosage regimens for ampicillin trihydrate when administered by NF injection to pigs are warranted.