Aminoglycosides and carbenicillin are frequently co-administered to patients with serious gram-negative infections. Aminoglycosides are inactivated by carbenicillin in vitro, and a loss of antibacterial activity of both antibiotics results. Although these interactions are presumed to occur in vivo, previous studies have not used assay methodology that can distinguish inactivation occurring prior to and during microbiological assay from inactivation in vivo. To address this problem, we gave seven bilaterally nephrectomized mongrel dogs doses designed to achieve simultaneous therapeutic serum concentrations of aminoglycosides and carbenicillin. Serum samples were tested by radioimmunoassay on three occasions: immediately, to determine in vivo interactions, and at 24 h and 1 week to assess the time course of in vitro inactivation. In comparison with immediate radioimmunoassay, gentamicin and tobramycin concentrations decreased by 39 and 53%, respectively, when assayed at 24 h (P < 0.05) and by 75 and 82% when assayed at 7 days (P < 0.001). In contrast, amikacin concentrations were reduced by only 9 and 30% at 24 h and 7 days. Tobramycin concentrations were also determined by immediate microbiological assay and were found to be similar to those in samples stored for 24 h before radioimmunoassay. Immediate radioimmunoassay demonstrated that carbenicillin reduced in vivo serum half-lives of gentamicin and tobramycin by 40% (P < 0.05). The half-life of amikacin in vivo was not significantly altered. In the presence of carbenicillin, amikacin was the most stable aminoglycoside both in vivo and in vitro, and it is the aminoglycoside of choice in patients with renal failure who require this combination.