Discovery of non-invasive biomarkers for the diagnosis of endometriosis

Table 5 Performance of cross-validated multi-marker models for discriminating endometriosis from control groups

Models	AUC	Sensitivity	Specificity
E versus C (all phases)
CA125, sICAM1, CPM	0.768	0.667	0.8
CA125, sICAM1, VEGF	0.777	0.644	0.8
CA125, sICAM1, FST	0.77	0.644	0.8
CA125, sICAM1	0.778	0.6	0.9
CA125, sICAM1, IL1R2	0.758	0.6	0.9
CA125, sICAM1, MCP1	0.757	0.6	0.9
E versus P (all phases)
sICAM1, FST, TNC	0.679	0.667	0.8
sICAM1, TNC	0.708	0.622	0.8
sICAM1, TNC, Oestradiol	0.68	0.622	0.8
sICAM1, PAEP, TNC	0.695	0.622	0.8
sICAM1, MIF, PAEP	0.697	0.622	0.8
sICAM1, LUM	0.665	0.444	0.9
E versus C + P (all phases)
CA125, sICAM1, FST, CPM	0.706	0.578	0.8
CA125, sICAM1, VEGF, PAEP	0.71	0.578	0.8
CA125, sICAM1, PAEP	0.719	0.578	0.8
CA125, sICAM1, MIF, PAEP	0.704	0.578	0.8
CA125, MIF	0.621	0.467	0.9
E versus C + P (Proliferative)
sICAM1, FST, Oestradiol	0.769	0.769	0.8
sICAM1, MIF, FST	0.781	0.692	0.8
sICAM1, FST	0.802	0.692	0.8
CRP, sICAM1, FST	0.802	0.692	0.8
CA125, sICAM1, FST	0.814	0.692	0.8
sICAM1, MIF, FST	0.781	0.615	0.9
E versus C + P (Secretory)
CA125, MIF, PAEP	0.705	0.654	0.8
CA125, sICAM1, MIF	0.725	0.615	0.8
CA125, MIF, TNC	0.683	0.615	0.8
CA125, MIF, PAEP	0.705	0.538	0.9
CA125, MIF, TNC	0.683	0.577	0.9

Models were generated by logistic regression using up to 4 candidates with cross-validation by leave-one-out. The best performing models (by sensitivity) and area under the ROC curve (AUC) are reported for each comparison at fixed specificities of 0.90 or 0.80. E = endometriosis; C = no pain controls; P pain controls. Control groups were pooled (C + P) for some of the analyses

ISSN: 1559-0275