Table 6.
Performance of a new five-protein signature derived from Generalised Discriminant Analyses (GDA).
| Training set | Test set | |
|---|---|---|
| ROC AUC | 84 (78–90) | |
| Sensitivity | 82 (73–90) | 81 (65–92) |
| Specificity | 74 (64–83) | 63 (46–78) |
| PPV | 75 (65–84) | 68 (52–81) |
| NPV | 81 (71–89) | 77 (59–90) |
| After adjustment of cut-off for increased sensitivity | ||
| Sensitivity | 89 (80–94) | 79 (63–90) |
| Specificity | 48 (37–59) | 41 (25–58) |
| PPV | 64 (55–73) | 58 (43–71) |
| NPV | 80 (67–90) | 65 (43–84) |
| After adjustment of cut-off for increased specificity | ||
| Sensitivity | 61 (50–71) | 58 (41–74) |
| Specificity | 91 (82–96) | 89 (75–97) |
| PPV | 87 (76–94) | 85 (65–96) |
| NPV | 69 (59–77) | 67 (52–80) |
This analysis was performed using data on the ILULU-TB cohort only (n = 249). Patients were randomly assigned to training and test sets at a ratio of 70:30. The signature comprised Complement FH, IP-10, CRP, SAA, and Transthyretin. Results are shown both before and after adjusting the probability threshold in the training set for diagnosing TB to increase each of sensitivity and specificity in turn to 90%.