In a prospective, randomized, double-blind study, we compared cefotaxime with nafcillin plus tobramycin in the treatment of serious bacterial infections. Of 195 patients with suspected or proven infections who were not neutropenic, definite bacterial infections were identified in 81; 34 of 38 patients given cefotaxime and 26 of 43 given nafcillin plus tobramycin (p less than 0.01) responded to treatment. The difference in response rates occurred primarily in patients with rapidly fatal underlying disease or with an infection outside the urinary tract. A logistic regression analysis showed that treatment with cefotaxime was still associated with a higher response rate after adjusting for several potential confounding factors. Among patients treated for 3 days or more, our criteria for nephrotoxicity were met in 2 of 68 (2.9%) given cefotaxime and 16 of 57 (28.1%) given nafcillin plus tobramycin (p less than 0.001). Prolongation of the prothrombin time and enterococcal colonization did not occur more frequently with cefotaxime. We conclude that cefotaxime may be more effective and less toxic than nafcillin plus tobramycin for patients with serious bacterial infections.