Objectives: a French national study on infectious encephalitis enrolled 253 patients in 2007. Fifty-two per cent of patients had a proven etiological diagnosis; 16% had bacterial encephalitis. We aimed to assess the predictive value of CSF lactate concentration to diagnose bacterial encephalitis. Patients and methods: patients from the 2007 cohort whose CSF lactate concentration was available were included. Clinical and biological features associated with a bacterial etiology were assessed using univariate analysis and multivariate logistic regression. The ROC curve of CSF lactate concentration was used to define the most appropriate cut-off associated with bacterial etiology. Results: fifty-seven patients (37 men, 20 women) were included. Ten patients had bacterial encephalitis, 20 patients had viral encephalitis. The mean CSF white blood cells (WBC), protein, glucose, and lactate levels were respectively 92 cells/mm3 (range: 0-450), 1.2 g/L (range: 0.2-8.2), 3.9 mmol/L (range: 0.8-8.8), and 2.8 mmol/L (range: 0-9.4). In univariate analysis, CSF protein level (P < 0.01), WBC count (P = 0.02), and lactate concentration (P < 0.01) were significantly associated with bacterial etiology. The only factor independently associated with bacterial etiology in the multivariate analysis was CSF lactate concentration. The area under the ROC curve of CSF lactate for the diagnosis of bacterial encephalitis was 0.86. The cut-off value of 3.4 mmol/L correctly classified 87.8% of patients with 70% sensitivity, 91.5% specificity, 64% positive predictive value, and 93% negative predictive value. Conclusion: a high CSF lactate level seems to be a better predictor than WBC or proteins to differentiate bacterial encephalitis from other etiologies.
Auteur : Maillet M, De Broucker T, Mailles A, Bouzat P, Stahl JP
Médecine et Maladies Infectieuses, 2018, vol. 48, n°. 6, p. 396-402