Skip to main content

Advertisement

Table 2 Diagnostic accuracy measures of Xpert MTB/RIF and the biomarkers uIFNγ and ADA using ROC-selected cut-points (definite-TB for sensitivity and non-TB for specificity calculations)

From: Diagnostic accuracy of quantitative PCR (Xpert MTB/RIF) for tuberculous pericarditis compared to adenosine deaminase and unstimulated interferon-γ in a high burden setting: a prospective study

Diagnostic testa Sensitivity (95% CI) (n/N)c Specificity (95% CI) (n/N)c Positive likelihood ratio, LR + (95% CI) Negative likelihood ratio, LR- (95% CI) Positive predictive value, PPV (95% CI) Negative predictive value, NPV (95%CI)
Xpert MTB/RIF 63.8% (52.4 to 75.1) 100% (85.6 to 100) n/c 0.36 100% 86.6%
44/69* *1 *2 26/26 *4 *5 (0.33 to 0.39) (98.0 to 100) (84.0 to 88.7)
uIFNγ (Intergam) (Youden’s, rule-in and rule-out cut-points: ≥44 pg/ml) c 95.7% (88.1 to 98.5) 96.3% (81.7 to 99.3) 25.8 0.045 91.7% 98.1%
67/70* *3 P <0.001 26/27*6 (3.6 to 184) (0.023 to 0.09) (88.1 to 94.3) (96.8 to 98.9)
ADA (rule-in cut-point: >107 IU/ml) 15.7% 96%     
(9.0 to 26.0) (80.5 to 99.3) 3.93 0.88 62.7% 72.7%
11/70 24/25 (0.21 to 72.5) (0.85 to 0.91) (51.4 to 72.8) (69.7 to 75.4)
*1 P <0.001      
ADA (cut-point in current clinical use: ≥35 IU/ml)[5] 95.7% 84%     
(88.1 to 98.5) (65.4 to 93.6) 6.0 0.051 71.9% 97.9%
67/70 21/25 (3.7 to 9.8) (0.026 to 0.10) (67.3 to 76.1) (96.4 to 98.7)
*1 *3 P < 0.001 *4 P = 0.03     
Tygerberg score ≥6 85.3% 77.3%     
(75.9 to 81) (56.6 to 89.9) 3.75 0.19 61.7% 92.5%
58/68 17/22 (2.52 to 5.59) (0.15 to 0.24) (56.9 to 66.2) (90.0 to 94.3)
*2 P = 0.004 *3 P = 0.04 *5 P = 0.01 *6 P =0.04     
Clinical predictors (rule-in cut-point: >6.1) 60.8% 96.3%     
(49.4 to 71.1) (81.7 to 99.3) 16.4 0.41 87.6% 85.1%
45/74 26/27 (2.25 to 119.9) (0.38 to 0.44) (82.4 to 91.4) (82.5 to 87.5)
*3 P <0.001      
Clinical predictors (Youden’s and rule-out cut-point: >3.5) b 91.9% 81.5% 4.96 0.10 68.0% 95.9%
(83.4 to 96.2) (63.3 to 91.8) (3.34 to 7.36) (0.07 to 0.14) (63.3 to 72.4) (94 to 97.2)
68/74 22/27     
*2 P <0.001 *5 P = 0.02     
Xpert MTB combined with uIFNγ (with uIFNγ if Xpert MTB/RIF negative) 97.1% 100% n/c    
(89.9 to 99.2) (86.7 to 100) 0.03 100% 98.8%
66/68 25/25 (0.01 to 0.08) (98.7 to 100) (97.7 to 99.4)
Xpert MTB combined with ADA (with ADA if Xpert MTB/RIF negative) 98.4% 100% n/c 0.02 100% 99.3%
(91.7 to 99.7) (85.7 to 100) (0.002 to 0.11) (98.7 to 100) (98.4 to 99.7)
63/64 23/23
  1. aDenominators other than 74 definite- and 27 non-TB patients indicate missing test results. Thus, 5/74 definite- and 1/27 non-TB patients did not have an Xpert MTB/RIF performed; 4/74 definite-TB did not have a uIFNγ (QFT-kit) test performed and 4/74 definite- and 2/27 non-TB did not have an ADA level measurement. Diagnostic accuracy measures presented in this table are for Xpert MTB/RIF performed on 1 ml unconcentrated PF, uIFNγ (both the QFT-kit and Intergam-kits) performed on the supernatants on 3 to 20 ml of centrifuged PF, and ADA performed on 1 to 8 ml of unprocessed PF.
  2. bYouden’s index (the cut-point allowing the most correctly classified) and optimal ‘rule-out’ (the cut-point with the lowest LR-) cut-points for the uIFNγ (both the QFT-kit and Intergam-kits) were the same (QFT: 0.18 IU/l, Intergam: 44 pg/ml) respectively. c P-values compare diagnostic tests for the indicated diagnostic accuracy measure, for example, sensitivity (specific proportions compared indicated by * and corresponding number symbol).
  3. ADA: adenosine deaminase; CI: confidence interval; n/c: not calculable as denominator equals zero; PF, pericardial fluid; uIFNγ: unstimulated interferon gamma. The reference standard was culture and/or histopathology pathognomonic of TB. TB prevalence = 30% used for predictive value estimates.