A Systematic Review and Meta-analysis of the Diagnostic Accuracy of Nucleic Acid Amplification Tests for Tuberculous Meningitis

Abstract

The diagnosis of tuberculous meningitis (TBM) is difficult and poses a significant challenge to physicians worldwide. Recently, nucleic acid amplification (NAA) tests have shown promise for the diagnosis of TBM, although their performance has been variable. We undertook a systematic review and meta-analysis to evaluate the diagnostic accuracy of NAA tests with cerebrospinal fluid (CSF) samples against that of culture as the reference standard or a combined reference standard (CRS) for TBM. We searched the Embase, PubMed, Web of Science, and Cochrane Library databases for the relevant records. The QUADAS-2 tool was used to assess the quality of the studies. Diagnostic accuracy measures (i.e., sensitivity and specificity) were pooled with a random-effects model. All statistical analyses were performed with STATA (version 14 IC; Stata Corporation, College Station, TX, USA), Meta-DiSc (version 1.4 for Windows; Cochrane Colloquium, Barcelona, Spain), and RevMan (version 5.3; The Nordic Cochrane Centre, the Cochrane Collaboration, Copenhagen, Denmark) software. Sixty-three studies comprising 1,381 cases of confirmed TBM and 5,712 non-TBM controls were included in the final analysis. These 63 studies were divided into two groups comprising 71 data sets (43 in-house tests and 28 commercial tests) that used culture as the reference standard and 24 data sets (21 in-house tests and 3 commercial tests) that used a CRS. Studies which used a culture reference standard had better pooled summary estimates than studies which used CRS. The overall pooled estimates of sensitivity, specificity, positive likelihood ratio (PLR), and negative likelihood ratio (NLR) of the NAA tests against culture were 82% (95% confidence interval [CI], 75 to 87%), 99% (95% CI, 98 to 99%), 58.6 (95% CI, 35.3 to 97.3), and 0.19 (95% CI, 0.14 to 0.25), respectively. The pooled sensitivity, specificity, PLR, and NLR of NAA tests against CRS were 68% (95% CI, 41 to 87%), 98% (95% CI, 95 to 99%), 36.5 (95% CI, 15.6 to 85.3), and 0.32 (95% CI, 0.15 to 0.70), respectively. The analysis has demonstrated that the diagnostic accuracy of NAA tests is currently insufficient for them to replace culture as a lone diagnostic test. NAA tests may be used in combination with culture due to the advantage of time to result and in scenarios where culture tests are not feasible. Further work to improve NAA tests would benefit from the availability of standardized reference standards and improvements to the methodology.

Keywords: meta-analysis; test accuracy; tuberculosis; tuberculous meningitis.

FIG 1

Flow diagram of literature search and study selection.

FIG 2

QUADAS-2 assessments of the included studies. Patient Selection, describe the methods of patient selection; Index Text, describe the index test and how it was conducted and interpreted; Reference Standard, describe the reference standard (gold standard test) and how it was conducted and interpreted; Flow and Timing, describe any patients who did not receive the index tests or reference standard or who were excluded from the 2-by-2 table and describe the interval and any interventions between the index tests and the reference standard (26).

FIG 3

Paired forest plots of pooled sensitivity and specificity of the NAA tests against culture.

FIG 4

Summary receiver operating characteristic (SROC) plot for NAA tests against culture. The SROC plot shows a summary of test performance, visual assessment of the threshold effect, and the heterogeneity of the data in ROC space between sensitivity (SENS) and specificity (SPEC); each circle in the SROC plot represents a single study, and the summary operating sensitivity, specificity, and SROC curve with both confidence and prediction regions are shown. The dashed line that is around the pooled point estimate shows the 95% confidence region. The area under the curve (AUC) acts as an overall measure of test performance. In particular, when AUC is between 0.8 and 1, the accuracy is relatively high. If the SROC curve were in the upper left corner, it would show the best combination of sensitivity and specificity for the diagnostic test. The number of studies which used NAA tests against culture is shown within each circle.

FIG 5

Paired forest plots of pooled sensitivity and specificity of NAA tests against CRS.

FIG 6

Summary receiver operating characteristic (SROC) plot for NAA tests against CRS. The number of studies which used NAA tests against CRS is shown within each circle.