Aminoglycosides (AMG) remain an important therapeutic modality for the treatment of gram-negative infections. Adequate AMG levels have been associated with a lower risk of toxicity. Because AMG levels cannot be predicted with confidence, drug concentrations need to be measured. The authors studied the effect of different anticoagulants on AMG concentrations, which were determined by enzyme immunoanalysis. Blood samples from patients treated with AMG were obtained, and were immediately distributed in five tubes containing EDTA, sodium-citrate, or heparin at concentrations of 5 and 50 U/ml; one serum aliquot was kept as a control. All AMG determinations were performed by the enzyme multiplied immunoassay technique with Cobas-Mira equipment. The average coefficient of variations was < 3%. All samples were run the same day. Analysis of variance for repeated measures was used. Twenty-four patients (21 male and 6 female) with a mean age of 50 years (95% confidence interval = 43 to 58) and mean serum creatinine concentrations of 0.87 to 0.29 +/- Standard Deviation received 89% gentamicin and 11% tobramycin. Peak levels of AMG obtained from plasma collected with sodium citrate or heparin were significantly lower (p < 0.001) than in serum or plasma collected with EDTA. The higher the level of AMG in serum, the greater the discrepancies between drug concentrations measured with different anticoagulants. The anticoagulant used was of critical importance in determining AMG blood levels, which were underestimated when citrate or heparin were present.