PEDIATRICS Vol. 107 No. 2 February 2001, pp. 451-451

Aseptic Meningitis

To the Editor.

We read with interest the article by Negrini et al in which they showed that, during the peak months of enteroviral meningitis, cerebrospinal fluid (CSF) polymorphonuclear (PMN) cell predominance cannot discriminate between aseptic and bacterial meningitis.¹ A PMN predominance (>50%) had a positive predictive value (PPV) of 19% and a negative predictive value (NPV) of 97% for bacterial meningitis, and a PPV of 81% with a NPV of 3% for aseptic meningitis. For a sensitivity of 90% and specificity of 43% for PMN over 50% to detect bacterial meningitis, the PPV and NPV they calculated are only valid for situations where the prevalence of enteroviral meningitis is high, as in their study, where it constituted 87.5% of all cases of meningitis. A test with identical sensitivity and specificity will give very different PPV and NPV when the prevalence of aseptic meningitis is different than in their study.

A previous study illustrates this point.² In a cohort study of 98 cases of meningitis in children under 14 years old, over a few years, the prevalence of aseptic meningitis we found was 64.3% and bacterial 35.7%. Using the same cutoff value of PMN of 50%, the sensitivity of the test in our study was 69% and specificity 63% to diagnose bacterial meningitis, with a calculated PPV of 50% (compared with 19%) and a NPV of 79% (compared with 97%) to detect bacterial meningitis, a PPV of 50% (compared with 81%) and NPV of 21% (compared with 3%) to detect aseptic meningitis. Even when using the same sensitivity and specificity for PMN as in Negrini's study, the calculated PPV was 42% (compared with 19%) and a NPV of 90% (compared with 97%) to detect bacterial meningitis, with a PPV of 50% (compared with 81%) and NPV of 10% (compared with 3%) to detect aseptic meningitis.

Regardless of a test specificity and sensitivity, even if replicated with 100% reliability, the calculated PPV and NPV cannot be generalized to other populations, as they depend heavily on the prevalence of the disease in each population.^3,4 If Negrini's study was not heavily weighed by an artificially higher prevalence of aseptic meningitis than normal, the PPV and NPV of PMN certainly would have been different. More important is the fact that during an epidemic of bacterial meningitis, such as with Neisseria meningitidis, the PPV and NPV for that same test will be very different and the recommendations given by the authors would then be invalid and, indeed, potentially dangerous. Regardless of the CSF white cell count results, the presumptive diagnosis of aseptic or bacterial meningitis will always need, in addition, to rely on the clinical findings and other CSF studies.

Hassib Narchi
Sandwell District General Hospital
Department of Paediatrics
Lyndon, West Bromwich
West Midland B71 4HJ, United Kingdom

REFERENCES

1. Negrini B, Kelleher KJ, Wald ER Cerebrospinal fluid findings in aseptic versus bacterial meningitis. Pediatrics. 2000; 105:316-319 [Abstract/Free Full Text]
2. Narchi H CSF bacterial antigen testing in the diagnosis of meningitis. Ann Saudi Med. 1997; 17:101-103 [Medline]
3. Hennekens CH, Buring JE. Epidemiology in Medicine. Boston, MA: Little Brown & Co; 1987
4. Gordis L. Epidemiology. Philadelphia, PA: WB Saunders & Co; 1996

In Reply.

We appreciate Dr Narchi's comments regarding our manuscript. We agree entirely with his comment that our results are affected substantially by the prevalence of aseptic meningitis in our study. In fact, the study was designed to maximize the number of cases of aseptic meningitis by collecting data only during the time period when the peak incidence of enteroviral disease is expected. The time period of data collection represented the typical seasonality of disease at our institution, not an unusual epidemic. By reporting the high positive predictive value of polymorphonuclear (PMN) cell predominance for aseptic disease, we did not intend clinicians to use PMN percentages to diagnose aseptic meningitis. Rather, we wanted to emphasize the point that both aseptic and bacterial disease can have a majority of PMNs in the cerebrospinal fluid (CSF). Accordingly, the differential leukocyte count alone cannot be used to predict the etiologic agent. During the winter months when the prevalence of viral disease is low, the clinician must have a much higher index of suspicion that a child with meningitis has bacterial disease. Based on the epidemiology of meningitis, a child with a PMN predominance in the CSF during the winter would have a high probability of a bacterial infection.

Therefore, as Dr Narchi notes, the diagnosis of meningitis must rely on clinical findings and other CSF laboratory results. However, the results of our study suggest that practitioners can comfortably diagnose aseptic disease during enteroviral season, even if there is a PMN predominance in the CSF, as long as all other clinical and laboratory markers support this decision.

Barbara Negrini, Kelly Kelleher, and Ellen Wald
Children's Hospital of Pittsburgh
Pittsburgh, PA 15213