Reliability and diagnostic performance of CT imaging criteria in the diagnosis of Tuberculous Meningitis
The original publication is available at http:// www.plosone.org
Publication of this article was funded by the Stellenbosch University Open Access Fund.
Introduction: Abnormalities on CT imaging may contribute to the diagnosis of tuberculous meningitis (TBM). Recently, an expert consensus case definition (CCD) and set of imaging criteria for diagnosing basal meningeal enhancement (BME) have been proposed. This study aimed to evaluate the sensitivity, specificity and reliability of these in a prospective cohort of adult meningitis patients. Methods: Initial diagnoses were based on the CCD, classifying patients into: ‘Definite TBM’ (microbiological confirmation), ‘Probable TBM’ (diagnostic score $10), ‘Possible TBM’ (diagnostic score 6–9), ‘Not TBM’ (confirmation of an alternative diagnosis) or ‘Uncertain’ (diagnostic score of ,6). CT images were evaluated independently on two occasions by four experienced reviewers. Intra-rater and inter-rater agreement were calculated using the kappa statistic. Sensitivities and specificities were calculated using both ‘Definite TBM’ and either ‘Definite TBM’ or ‘Probable TBM’ as gold standards. Results: CT scan criteria for BME had good intra-rater agreement (k range 0.35–0.78) and fair to moderate inter-rater agreement (k range 0.20–0.52). Intra- and inter-rater agreement on the CCD components were good to fair (k = ranges 0.47–0.81 and 0.21–0.63). Using ‘Definite TBM’ as a gold standard, the criteria for BME were very specific (61.5%–100%), but insensitive (5.9%–29.4%). Similarly, the imaging components of the CCD were highly specific (69.2–100%) but lacked sensitivity (0–56.7%). Similar values were found when using ‘Definite TBM’ or ‘Probable TBM’ as a gold standard. Discussion: The fair to moderate inter-rater agreement and poor sensitivities of the criteria for BME suggest that little reliance should be placed in these features in isolation. While the presence of the CCD criteria of acute infarction or tuberculoma(s) appears useful as rule-in criteria, their absence is of little help in excluding TBM. The CCD and criteria for BME, as well as any new criteria, need to be standardized and validated in prospective cohort studies.