Table 2. Diagnostic Performance of the Risk Score in the Discovery and Validation Cohorts and as Compared with the IGRA and the Xpert MTB/RIF Assay in the Validation Cohort*.
Performance Measure | Risk Score | IGRA | Xpert MTB/RIF Assay† | ||
---|---|---|---|---|---|
Test Set in the Discovery Cohort | Validation Cohort, Excluding Children with Positive IGRA in Group with Other Diseases | Validation Cohort, Including Children with Positive IGRA in Group with Other Diseases | Validation Cohort | Validation Cohort | |
No. of patients with TB | 23 | 35 | 35 | 35 | 35 |
No. of patients with other diseases | 34 | 55 | 64 | 54 | 55 |
Area under ROC curve (95% CI) | 86.2 (77.1–94.0) | 89.0 (82.3–94.9) | 89.0 (81.9–95.3) | 71.7 (58.3–84.8) | 77.1 (69.9–85.7) |
Sensitivity — % (95% CI) | 78.3 (60.9–95.7) | 82.9 (68.6–94.3) | 82.9 (68.6–94.3) | 60.0 (33.3–86.67) | 54.3 (37.1–68.6) |
Specificity — % (95% CI) | 73.5 (55.9–88.2) | 83.6 (74.6–92.7) | 82.8 (73.4–92.2) | 83.3 (72.2–92.6) | 100.0 (100.0–100.0) |
Patients in the discovery cohort of South African and Malawian children were assigned to training and test sets (80% and 20% of the cohort, respectively). The 51-transcript signature used in the group with diseases other than TB was derived from data on the South African and Malawian patients in the training set and applied to the independent Kenyan validation cohort. All analyses included children with and those without HIV infection. Sensitivity and specificity were calculated with the use of a weighted threshold for classification. ROC denotes receiver-operating-characteristic. CI denotes confidence interval.
The Xpert MTB/RIF assay was positive for 19 of 35 patients with culture-confirmed tuberculosis and none of 55 patients with diseases other than tuberculosis.