An assessment of nucleic acid amplification testing for active mycobacterial infection

Assessing diagnostic accuracy

To assess the diagnostic accuracy of NAAT, studies were only included if they provided data that could be extracted into a classic 2×2 table (Table ), in which the results of the index diagnostic test or the comparator were cross-classified against the results of the reference standard (Armitage, Berry & Matthews 2002; Deeks 2001), and Bayes’ Theorem was applied:

Table Diagnostic accuracy data extraction for NAAT

-	-	Reference standard	(culture ± DST)-	-
-	-	Disease +	Disease –	-
Index test (NAAT)	Test +	true positive	false positive	Total test positive
Or comparator (AFB)	Test –	false negative	true negative	Total test negative
-	-	Total with MTB or NTM	Total without MTB or NTM	-

AFB = acid-fast bacilli; DST = drug susceptibility testing; MTB = Mycobacterium tuberculosis; NAAT = nucleic acid amplification testing; NTM = non-tuberculous mycobacteria

Primary measures

Test sensitivity was calculated as the proportion of people with MTB or NTM infections (as determined by the reference standard) who had a positive test result using AFB and/or NAAT:

Sensitivity (true positive rate) = number with true positive result / total with MTB or NTM infections

Test specificity was calculated as the proportion of people without infection (as determined by reference standard) who had a normal test result using AFB and/or NAAT:

Specificity (true negative rate) = number with true negative result / total without MTB or NTM infections

The 95%CI was calculated by exact binomial methods.

Positive and negative likelihood ratios (LR+ and LR–) were also reported. These ratios measure the probability of the test result in patients with MTB or NTM infections compared with those without.

LR+ = sensitivity / 1 – specificity

LR– = 1 – sensitivity / specificity

An LR of 1 means that the test does not provide any useful diagnostic information, whereas LR+ > 5 and LR– < 0.2 can suggest strong diagnostic ability (MSAC 2005).

Summary measures

Diagnostic test accuracy meta-analysis was undertaken to assess the accuracy of NAAT compared with AFB microscopy in the diagnosis of MTB or NTM infections, compared with culture, using Stata version 12 (Stata Corporation 2013). Only studies that provided raw (2×2) data were included. Summary receiver–operator characteristic (SROC) curves, forest plots and LR scattergrams were generated using the ‘midas’ command in Stata, which requires a minimum of four studies for analysis and calculates summary operating sensitivity and specificity (with confidence and prediction contours in SROC space). Heterogeneity was calculated using the formula I² = 100% x (Q – df)/Q, where Q is Cochran's heterogeneity statistic and df is the degrees of freedom (Higgins et al. 2003). Summary estimates for sensitivity, specificity, LR+ and LR– were also calculated. Confidence intervals were computed assuming asymptotic normality after a log transformation for variance parameters and for LR+ and LR–.

Subgroup analyses were performed for results according to specimen type, incidence of TB in the study population and the presence of an HIV infection.

Where meta-analysis could not be performed, the median (range) sensitivity and specificity values were calculated.