Table 2.
Sensitivity% (x/n); [95%CI] | Specificity% (x/n); [95%CI] | PPV% (x/n); [95%CI] | NPV% (x/n); [95%CI] | ||
---|---|---|---|---|---|
Gold standard 1 = Microbiological confirmation |
Per patient analysis (n=73) |
||||
Smear |
71.0 (22/31) [52.0; 85.8]) |
N/a |
N/a |
82.3 (42/51) [69.1; 91.6] |
|
MGIT |
96.8 (30/31) [83.3; 99.9] |
N/a |
N/a |
97.7 (42/43) [87.7; 99.9] |
|
MODS |
96.8 (30/31) [83.3; 99.9] |
97.6 (41/42) [87.4; 99.9] |
96.7 (30/31) [83.3; 99.9] |
97.6 (41/42) [87.4; 99.9] |
|
Xpert |
93.5 (29/31) [78.6; 99.2] |
100 (42/42) [91.6; 100.0] |
100.0 (29/29) [88.1; 100.0] |
95.5 (42/44) [84.5; 99.4] |
|
Per sample analysis (n=178) | |||||
Smear |
53.8 (42/78) [39.8; 67.9]) |
N/a |
N/a |
73.5 (100/136) [63.1; 83.9] |
|
MGIT |
92.3 (72/78) [87.4; 97.2] |
N/a |
N/a |
94.3 (100/106) [90.3; 98.4] |
|
MODS |
84.6 (66/78) [78.1; 91.1] |
99 (99/100) [94.3; 99.9] |
98.5 (66/67) [91.6; 99.9] |
89.2 (99/111) [83.3; 95.0] |
|
Xpert |
79.5 (62/78) [70.6; 88.4]) |
100 (100/100) [95.7; 100.0] |
100 (62/62) [93.3; 100.0] |
86.2 (100/116) [78.9; 93.5] |
|
Gold standard 2 = Clinical diagnosis |
Per patient analysis (n=73) |
||||
Smear |
37.9 (22/58) [25.5; 51.6] |
N/a |
N/a |
29.4 (15/51) [17.5; 43.8] |
|
MGIT |
51.7 (30/58) [38.2 ; 65.0] |
N/a |
N/a |
34.9 (15/43) [21.0; 50.9] |
|
MODS |
51.7 (30/58) [38.2 ; 65.0] |
93.3 (14/15) [68.0; 99.8] |
96.8 (30/31) [83.3; 99.9] |
33.3 (14/42) [19.6; 49.5] |
|
Xpert |
50.0 (29/58) [36.6 ; 63.4]) |
100.0 (15/15) [78.2; 100.0]) |
100.0 (29/29) [88.0; 100] |
34.1 (15/44) [20.5; 49.9] |
|
Per sample analysis (n=178) | |||||
Smear |
30.0 (42/140) [20.1 ; 39.9] |
N/a |
N/a |
27.9 (38/136) [17.1; 38.8] |
|
MGIT |
51.4 (72/140) [40.1; 62.7] |
N/a |
N/a |
35.8 (38/106) [22.8; 48.9] |
|
MODS |
47.1 (66/140) [36.5; 57.7] |
97.4 (37/38) [86.0; 99.7] |
98.5 (66/67) [91.6; 100.0] |
33.3 (37/111) [21.0; 45.7] |
|
Xpert | 44.3 (62/140) [33.8; 54.8] | 100 (38/38) [89.4; 100.0] | 100 (62/62) [93.3; 100.0] | 32.8 (38/116) [20.6; 44.9] |
N/a=not appropriate (as 100% by definition of the gold standard).