Table 3. Accuracies of seroantigen combinations to distinguish between TB and ORD, or LTBI, after general discriminant analysis.
| TB vs ORD | |||||
|---|---|---|---|---|---|
| Antigen combination | Resubstitution classification matrix | Leave-one-out cross-validation | |||
| %TB | % ORD | % Accuracy | % TB | % ORD | |
| Anti-TB-LTBI IgG | 95.23 (20/1) | 97.61 (1/41) | 96.8 | 95.23 (20/1) | 97.61 (1/41) |
| Anti-Tpx IgG | |||||
| Anti-MPT64 IgA | PPV: 0.95 (95% CI; 0.74-0.99) | ||||
| NPV: 0.97 (95% CI; 0.85-0.99) | |||||
| TB vs LTBI | |||||
|---|---|---|---|---|---|
| Antigen combination | Resubstitution classification matrix | Leave-one-out cross-validation | |||
| %TB | % LTBI | % Accuracy | % TB | % LTBI | |
| Anti-LAM IgA | 100.0 (21/0) | 100.0 (0/21) | 100.0 | 95.23 (20/1) | 95.23 (0/21) |
| Anti-TB-LTBI IgG | |||||
| Anti-Tpx IgG | |||||
| Anti-MPT64 IgA | PPV: 0.95 (95% CI; 0.74-0.99) | ||||
| NPV: 0.95 (95% CI; 0.75-0.99) | |||||
The predictive abilities of the optimal combination of serodiagnostic markers to differentiate between active TB (n=21), definite LTBI (n=21) (IGRA+) or ORD individuals (IGRA+ and IGRA- combined) (n=41) was investigated using best subsets general discriminant analysis and a leave-one-out cross-validation table was constructed using the variables that were included in the optimal classification model. PPV=Positive predictive value, NPV=Negative predictive value.