The ratio between the concentrations of different antimicrobial agents varies widely during therapy and often bears no resemblance to the ratios assessed by in vitro methods to evaluate the effectiveness of multiple antimicrobial therapy. The authors examined an alternative method using patient serum that reflects the actual antibiotic levels achieved in the patient. Previous investigators have shown that the ratio of the concentration of free drug (drug-f) to the minimal bactericidal concentration (MBC) in broth approximates the serum bactericidal titer (SBT) when pooled normal human serum is used as the diluent. Theoretically, when two antimicrobial agents are administered, the SBT should be equivalent to (drug-f A/MBC drug A) + (drug-f B/MBC drug B). SBT and drug level determinations were performed on peak and trough serum specimens from ten patients with endocarditis or osteomyelitis who were receiving multiple antimicrobial therapy. The serum dilution synergy method predicted additive interactions twice as often as the checkerboard and kill curve, and predicted synergy less frequently than the kill curve. The checkerboard predicted antagonism four times more often than the other methods and provided equivocal results in four of ten cases. The suggested method may offer an alternative procedure to assess antimicrobial interactions, which is based on antibiotic levels actually achieved in vivo instead of the arbitrary concentrations often used in in vitro tests.