Cerebrospinal fluid lactate: Is it a reliable and valid marker to distinguish between acute bacterial meningitis and aseptic meningitis?

Cerebrospinal fluid (CSF) lactate assay has been a subject of research since 1925. A systematic review by Huy and colleagues in the previous issue of Critical Care summarizes data from 25 studies evaluating the role of CSF lactate in the differential diagnosis between acute bacterial and aseptic meningitis. The authors concluded that CSF lactate is a good single indicator and a better marker compared with conventional markers. But concerns remain because of poor quality of included studies, lack of proper 'gold standard', and limited applicability. More studies with a rigorous design are needed to determine definitively whether CSF lactate assay is a reliable and valid marker to distinguish between acute bacterial meningitis and aseptic meningitis.

Distinguishing bacterial from viral meningitis/aseptic meningitis (VM/AM) is an age-old problem. Th e distinction is important because bacterial meningitis (BM) requires urgent intravenous antibiotic administration in the hospital whereas AM is self-limiting. A reliable and valid marker is necessary to make this distinction.
In the previous issue of Critical Care, Huy and colleagues [1] summarize data from 25 studies describing the test characteristics of cerebrospinal fl uid (CSF) lactate in the diff erential diagnosis between acute BM and AM. Th e methodology used by the authors is reasonably sound and inspires confi dence in the results. Reviewer bias was controlled with the use of two independent reviewers for selection of studies, data extraction, and quality assessment. Th e chance-corrected agreement (kappa = 0.898) between the reviewers was very high for study selection, although the same was not reported for data extraction and quality assessment.
Huy and colleagues used well-established criteria to assess the quality of the selected studies, and examined and explored heterogeneity using standard methods and summarized the results using a summary receiver operator characteristic (SROC) curve. Th e authors presented Q value and area under the curve (AUC) as measures of accuracy. Th e AUC of CSF lactate concentration was 0.9840, indicating excellent accuracy. Th e authors concluded that CSF lactate is a good single indicator to distinguish BM from AM and a better marker than other conventional markers. However, comments on threats to internal validity and generalizability of the fi ndings are warranted.

Threats to internal validity
Th e quality of a meta-analysis can be only as good as the included studies ('GIGO' principle: garbage in, garbage out). Th e quality of studies included in the review by Huy and colleagues is somewhat unsettling. Specifi cally, reported blinding of lactate assay in only 3 (13%) studies and consecutive or random recruitment of participants in 12 (50%) is a matter of concern. It is possible that other studies blinded the lactate assay without reporting the fact, but this matter remains speculative in the absence of confi rmation from the study authors. Compromised quality of original studies threatens the validity of the conclusions.
Lack of a proper 'gold standard' for AM or VM is a vexing problem in this area of research. When an imperfect standard is used to evaluate a diagnostic test, distortions occur in the commonly used measures of test performance, like sensitivity/specifi city [2]. Distorted measures carry the error in their meta-analysis. Th is review suff ers from this error.
Two comments on the comparison of CSF lactate assay with conventional CSF markers are warranted. First, clinicians diagnose AM on the basis of a pattern of a combination of fi ndings on conventional CSF markers (CSF total number of leukocytes, CSF glucose, CSF/ plasma glucose quotient, and CSF protein), not on individual markers. Th us, it is clinically relevant to compare this pattern and the CSF lactate assay and to determine whether the assay adds signifi cantly to the pattern. However, the authors did not address this question; this was probably because they did not have access to the individual patient data that are necessary to perform this comparison. Th e review, therefore, fails to answer this question.
Second, the authors assert a lower accuracy of individual CSF markers compared with the CSF lactate test based on point estimates of AUC. It is not clear whether the observed diff erences in AUC are statistically signifi cant. An objective assessment of this is possible by using the Hanley test [3] and by calculating the confi dence interval around the estimates.

Applicability
CSF lactate assay is not available in most centers in developing countries and rural settings. Th is and the fact that many patients receive antibiotics before lumbar puncture compromise the applicability of the fi ndings.

Summary
Th e review is a worthwhile contribution to the fi eld, has a sound methodology, and provides a summary of the results from published data. Clearly, a meta-analysis of individual patient data and more studies are required to determine defi nitively whether CSF lactate assay is a reliable and valid marker to distinguish between BM and AM.