Despite the apparent success of several new cephalosporins in the treatment of Gram-negative bacterial meningitis, four treatment failures with cefotaxime or latamoxef were encountered (two caused by Enterobacter and two by Serratia spp.). In-vitro parameters of susceptibility of these clinical isolates were compared with those of a meningeal Ent. cloacae isolate from a successfully treated patient. The MIC and MBC values, degrees of inoculum effect, and amounts of /?-lactamase produced correlated poorly with the observed clinical outcome. However, the extent to which an isolate was killed by the cephalosporin used for treatment, in a 6-h invitro incubation, showed good correlation. We suggest that such a test should be used to predict clinical outcome of therapy because the other parameters such as the MIC and MBC values are not sufficiently discriminatory.