Table 4. Performance of the Four-Marker Panel and EGF, sCD26, CAL and CEA in the Diagnosis of Lung Cancer.
| Training Set | Cut-off | Sn (%) | Sp (%) | PPVa (%) | NPVa (%) | AUC (95% CI)b |
|---|---|---|---|---|---|---|
| Multivariate Algorithm: EGF, sCD26, CAL, CEA | >0.266 | 97 | 43 | 57.6 | 94.7 | 0.873 (0.811–0.925) |
| EGF | >178.48 pg/mL | 95 | 22.2 | 49.4 | 84.8 | 0.698 (0.615–0.773) |
| sCD26 | ≤637.2 ng/mL | 95 | 13.9 | 46.8 | 77.7 | 0.711(0.629–0.785) |
| CAL | >96.37 ng/mL | 95 | 30.6 | 52.2 | 88.4 | 0.759 (0.679–0.827) |
| CEA | >258.2 pg/mL | 95 | 11.1 | 46 | 73.6 | 0.744 (0.663–0.814) |
| Clinical Modelc | >0.237 | 95 | 26.4 | 50.8 | 86.9 | 0.717 (0.637-0.799) |
| Validation Set | ||||||
| Multivariate Algorithm: EGF, sCD26, CAL, CEA | >0.266 | 91.7 | 45.4 | 57.3 | 87.3 | 0.837 (0.718-0.936) |
| Clinical Modelc | >0.237 | 91.7 | 27.3 | 50.2 | 80.5 | 0.659 (0.488-0.816) |
Abbreviations: Sn = Sensitivity, Sp = Specificity, PPV = Predictive Positive Value, NPV = Negative Predictive Value.
aPositive and negative predictive values were estimated assuming a prevalence of lung cancer of 44.4% (QDU of the Pneumology Service of Hospital Álvaro Cunqueiro EOXI Vigo).
bAUC and 95% CI evaluated in the training test is not protected against overfitting.
cClinical model includes gender, age and smoking.