A method was devised for predicting the serum bactericidal activity of new drugs. Six healthy volunteers received 2 g moxalactam, cefoperazone and cefotaxime, respectively, as 30-min infusions in a crossover manner at one-week intervals. The pharmacokinetics of each drug was characterized and the bactericidal activity of the serum 1 h after infusion was measured against panels of six strains of Pseudomonas aeruginosa, six strains of Escherichia coli, six strains of Staphylococcus aureus, and four strains of Klebsiella pneumoniae. The minimum bactericidal concentration of each antibiotic was determined for each organism by the standard NCCLS reference method and the method of Stratton and Reller. On the basis of these values and a serum concentration-time curve constructed from individual patient pharmacokinetic parameters, the bactericidal activity of the serum 1 h after infusion was predicted. These predictions showed a 90% agreement with measured values calculated according to the method of Stratton and Reller, whereas an agreement of 74% was obtained with the reference method. This difference was statistically significant (p less than 0.001).