Database, case-flow and analytical procedures
35049 cases from database were screened and 519 cases ≤14 years of age with pulmonary disease were enrolled in this study. Of the 519 cases, 173 cases were matched with a clinical profile, AFB smear, solid/liquid culture result, Xpert and final diagnosis. 346 non-matched cases included 55 with incomplete clinical data (unable to define the case, such as dead or transferred to other hospitals), 87 with repeated hospitalization, 45 with no necessity for RTS test, 110 with no RTS test for other reasons (such as unable to acquire specimen), 49 with non-paired results for Xpert, AFB smear, and MTB culture test (only one or two results). In 173 matched cases, 28 cases were excluded for the following reason: 3 with NTM and 5 with BCG vaccine infection identified by sequencing, and 20 who were culture-positive but had no record of strain identification. The 145 matched cases were analyzed for accuracy, among them 33.1% (48/145) were culture-positive (30 smear-positive and 18 smear-negative), 66.9% (97/145) were culture-negative including 54.5% (79/145) probable TB and 12.4% (18/145) non-TB cases (FIGURE 1.).. For the 145 matched cases, the mean age was 3.65 ± 3.62 years, male sex was 37.9% and 87.6% was clinically diagnosed TB. For the 346 non-matched cases, the mean age was 3.51 ± 3.50 years, male sex was 37.6%, and 91.0% was clinically diagnosed TB, no statistically difference between these two groups. TABLE 1. shows the clinical characters of the matched cases.
Estimated prevalence at the site and the notification rate from Xpert
The TB prevalence estimation via EMR record was approximately 82%, from the available follow-up records of this site it was about 70%, the total microbiologically positive rate of the cases was 45% and the eight experienced experts’ estimation was 70%. The total prevalence was estimated to be 67% (TABLE 2.).. From these estimates, the estimated number of TB cases was 259 after exclusion of repeated cases and those without a necessity of RTS testing. Of these cases, 75 had positive Xpert results. Thus, the all-factors case notification rate by Xpert was 29.0% (75/259) and was similar by culture (32.4%, 84/259, p = 0.391).
Diagnostic accuracy of the Xpert RIF assay
The diagnostic accuracy of the Xpert RIF assay was evaluated against different reference standards separately. Based on the gold standard, 48 cases were diagnosed as pulmonary tuberculosis. On this basis, the overall sensitivity and specificity of the Xpert assay were 66.7% (32/48, 95%CI 0.52–0.79) and 87.6% (85/97, 95%CI 0.87–0.98); the PPV and NPV were 72.7% (95%CI 0.57–0.85) and 84.2% (95%CI 0.74–0.90). The positive yield was 76.7% (23/30, 95%CI 0.57–0.89) in cases with both smear-positive and culture-positive results, and 50% (9/18, 95%CI 0.27–0.73) in the smear-negative culture-positive cases (TABLE 3).. The positive rates of the Xpert assay on different sample types (IS, NA and GA) were 29.6% (8/27), 30.0% (6/20), 30.6% (30/98), respectively and showed no statistically significant differences (P >0.05). The agreement between the Xpert and the culture results was 81% (kappa = 0.56) and the inconsistencies were 12 Xpert-positive culture-negative cases and 16 Xpert-negative culture-positive cases (TABLE 4).
Based on the CCRS, 127 cases were diagnosed as pulmonary tuberculosis, including 79 culture-negative clinically diagnosed TB. The sensitivities of AFB smear, MTB culture, and Xpert MTB/RIF assay were 23.6% (30/127, 95%CI 16.7–32.1), 37.8% (48/127, 95%CI 29.5–46.9) and 34.6% (44/127, 95%CI 26.6–43.7), respectively. The specificities and positive predictive values of the three assays were all 100%. The negative predictive values of AFB smear, MTB culture, and Xpert MTB/RIF assay were 15.7% (95%CI 9.8–23.9), 18.6% (95%CI 11.7–20.0), 17.8% (95%CI 11.2–27.0), respectively (TABLE 5). There were 12 children had Xpert(+)&smear(-)&culture(-) results, 2 children had Xpert(-)&smear(+)&culture(-) results. the combined sensitivity of Xpert plus culture, Xpert plus AFB smear, Xpert plus AFB smear and culture were 47.2% (60/127), 36.2% (46/127), 48.8% (62/127), respectively. They were all statistically higher than from a single test (P <0.05). Performance of Xpert MTB/RIF assay in different age groups were also analyzed. The sensitivities in 1–5 years group, 5–10 years group, 5–14 years group were 36.4%, 15.4%, 61.5%, respectively. (TABLE 6)
In routine clinical practice, non-tuberculous Mycobacterium (NTM) and BCG vaccine infection are not always pre-excluded. Therefore, we evaluated the situation when these cases arose in the analysis. In this study, NTM cases were consistently Xpert-negative and BCG vaccine infection cases were all Xpert-positive. The sensitivity, specificity, PPV and NPV of Xpert against the gold standard were 66.1% (95%CI 0.52–0.78), 87.6% (95%CI 0.79–0.93), 75.5% (95%CI 0.61–0.87) and 81.7% (95%CI 0.73–0.89), respectively. In comparisons against the CCRS, the sensitivity, specificity, PPV and NPV of Xpert were 34.7% (95%CI 0.26–0.44), 80.8% (95%CI 0.61–0.93), 80.8% (95%CI 0.78–0.97) and 20.2% (95%CI 0.13–0.29), respectively.
Detection of RIF resistance
A total of 11 cases had recorded drug-resistance via phenotype drug sensitivity test (DST) and had matched Xpert MTB/RIF test results. 5 were resistant to HRES (isoniazid, rifampin, ethambutol and streptomycin), 3 were resistant to HR, 1 was resistant to HE, and 2 were mono-resistant to H. The 8 cases with phenotypic rifampin-resistance were also detected by Xpert assay. The consistency of culture-based DST and Xpert assay for rifampin resistance was 100%.