A multivariate approach to prognostication in experimental bacterial meningitis.

Known concentrations of type III pneumococci were inoculated into eighty-one rabbits by cisternal puncture. Antibiotic therapy was started the following day. Aliquots of cerebrospinal fluid (CSF) and plasma were sampled on day one immediately before therapy was started and at regular intervals thereafter for up to eight days. Samples were analyzed for glucose, lactate, lactic dehydrogenase, and creatine kinase in various combinations. Leukocyte counts were performed on all CSF specimens. The timing of the specimens proved critical to the prognostic utility of the analyses performed. Day two plasma glucose was the most important single measurement for prognostication. Day one values for CSF glucose, lactate, and leukocyte count were also important. Substantial gains in prognostic accuracy were achieved when clinical laboratory measurements were used in combination by discriminant function analysis.