An assessment of nucleic acid amplification testing for active mycobacterial infection

Meta-analysis of studies assessing the diagnostic accuracy of NAAT compared with culture in either AFB-positive or AFB-negative specimens

Forest plots showing the sensitivity and specificity from individual studies that compared NAAT with culture in either AFB-positive specimens or AFB-negative specimens from patients suspected of having TB are shown in Figure and Figure (Appendix ). Figure shows the pooled sensitivity and specificity values for NAAT compared with culture for AFB-positive and AFB-negative specimens.

Among the 28 studies that reported data for AFB-positive specimens, the sensitivity was at least 94% in all but 5 studies (pooled value 99%; 95%CI 96, 100). However, the specificity was much more variable, ranging from 0% to 100% between studies (pooled value 78%; 95%CI 53, 92). Conversely, in the 39 studies that reported data for AFB-negative specimens, the sensitivity was highly variable between studies, with a pooled value of 80% (95%CI 69, 87). The specificity was at least 82% in the studies that were conducted in countries with a low or medium incidence of TB, but was highly variable (range 18–100%) in studies conducted in countries with a high incidence of TB, especially those using in-house NAAT. These observations are reflected in the 95%CIs of the pooled sensitivity and specificity values from subgroup analyses shown in Figure A and B.

The LR scattergram shows that the summary LR+ and LR– values for NAAT compared with culture in AFB-positive specimens were within the lower left quadrant, indicating that a negative NAAT result can confidently exclude the likelihood of an MTB infection (as determined by culture) in patients who had an AFB-positive sample (Figure ). Unexpectedly, a positive NAAT result does not eliminate the possibility of AFB-positive patients not having a detectable MTB infection (i.e. being culture-negative). This can be explained because culture is an imperfect reference standard, which likely resulted in misclassification of many of the 22% false-positive results (1 – specificity) seen for NAAT when compared with culture in AFB-positive specimens (Figure ). Therefore, NAAT is likely to be more effective at confirming the presence of an MTB infection in these patients than the LR scattergram suggests.

In AFB-negative specimens the overall summary LR+ and LR– values for NAAT compared with culture were in the upper right quadrant of the scattergram or within the green shaded bands, indicating that a positive NAAT result is likely to correctly confirm the presence of MTB. However, interpretation of a negative NAAT result is dependent on the type of specimen tested. In patients with AFB-negative sputum a negative NAAT indicated that the patient may not be culture-positive but it could not be ruled out (summary values are within the green shaded area; Figure ). In patients with AFB-negative non-sputum specimens, a negative NAAT result provided no additional useful information. This is likely due to the paucibacillary nature of AFB-negative specimens. It should be noted that if few bacilli are present in the specimen, the possibility of a false-negative result would increase for all three tests.

Figure Forest plot showing the pooled sensitivity and specificity values for NAAT compared with culture for AFB-positive (A) and AFB-negative (B) specimens grouped according to NAAT methodology, specimen type and incidence of TB in the country in which the study was conducted

Green plots represent median (range) in groups for which meta-analysis could not be conducted.

Incidence of TB based on WHO estimates from 2012: high incidence = > 100 cases per 100,000 people; medium incidence = 10–100 cases per 100,000 people; low incidence = ≤ 10 cases per 100,000 people

K = the number of studies; NAAT = nucleic acid amplification testing; TB = tuberculosis

Figure LR scattergram for diagnosis of MTB infection by NAAT compared with culture for AFB-positive specimens according to NAAT methodology

LR = likelihood ratio; NAAT = nucleic acid amplification testing

The SROC curve for studies investigating NAAT compared with culture in AFB-positive specimens showed no threshold effects based on commercial or in-house NAAT (Figure ) or specimen type (not shown). However, for studies investigating NAAT compared with culture in AFB-negative specimens, a threshold effect was seen (Figure ). In-house NAAT tended to be more sensitive and less specific than commercial NAAT when compared with culture. Similarly, NAAT compared with culture tended to be less sensitive and more specific when testing sputum specimens than for non-sputum specimens. The AUC for NAAT versus culture in AFB-positive (0.98; 95%CI 0.96, 0.99) and AFB-negative (0.93; 95%CI 0.91, 0.95) specimens indicated that the NAATs perform well in predicting culture positivity (AUC > 0.9) for both types of specimen.

Figure LR scattergram for diagnosis of MTB infection by NAAT compared with culture for AFB-negative specimens according to NAAT methodology

LR = likelihood ratio; NAAT = nucleic acid amplification testing

Figure SROC curve for all studies investigating the sensitivity and specificity of NAAT versus culture in the diagnosis of TB for AFB-positive specimens based on NAAT methodology

AUC = area under curve; SROC = summary receiver–operator characteristic

Figure SROC curve for all studies investigating the sensitivity and specificity of NAAT versus culture in the diagnosis of TB for AFB-negative specimens based on NAAT methodology (A) and specimen type (B)

AUC = area under curve; NAAT = nucleic acid amplification testing; SROC = summary receiver–operator characteristic