Corrections for leukocytes and percent of neutrophils do not match observations in blood-contaminated cerebrospinal fluid and have no value over uncorrected cells for diagnosis

Background: The utility of published adjustments for leukocytes in blood-contaminated cerebrospinal fluid (CSF) is controversial. In this study, we assess how these adjustments and a novel rule for predicting percent of neutrophils match observations in CSF. We also determine the incremental value of corrected over uncorrected cells in CSF for discriminating between bacterial and viral meningitis. Methods: We analyzed blood-contaminated CSF specimens collected from 1-month to 18-year-old children presenting to an urban academic pediatric emergency department between 1993 and 2003. Predictions of leukocytes (total) and neutrophils in CSF were derived from a standard rule and from an alternative rule based on a regression between neutrophils in peripheral blood and CSF. The match between observed and predicted cell counts was estimated by the coefficient of determination (R-2). The value of corrected over uncorrected cells for diagnosing bacterial meningitis was evaluated by comparing the areas under respective receiver operator characteristic curves (AUC). Results: At an R-2 of 0.11, predicted leukocytes matched observed leukocytes poorly for 682 CSF specimens that met study criteria. The percent of neutrophils in CSF predicted by the regression 7% + (0.5 x percent of neutrophils in peripheral blood) also fit observed neutrophils only modestly (R-2 0.27). For diagnosing bacterial meningitis, there was no difference between AUC values for corrected and uncorrected leukocytes and percent of neutrophils. Conclusion: In blood-contaminated CSF, there is poor to modest correlation between observed and predicted counts of leukocytes and of neutrophils. Adjusted blood counts in CSF have no advantage over uncorrected counts for predicting bacterial meningitis.