Figure 1.

Multivariate analysis was conducted using training-validation (~ 2/3 of all data) and test sets (~1/3 of all data) strategy. Final model (containing variables selected in step 1) predictive capabilities were assessed in the test set to avoid validating overfitted models. Methods used in step 1 included LASSO, then logistic regression. Best models here were selected as those yielding AUROC > 0.95 in the validation set. Variables with no null estimated coefficients were included in a logistic regression model and submitted to a backward procedure. Finally, the best model obtained was applied to the test set to confirm its performance based on the AUROC. Coefficient in the logistic regression were estimated as median of the coefficient of the best models in step 1. Legend: n_C, n_E: number of patients in the control and endometriosis group, respectively; LASSO: Least Absolute Shrinkage and Selection Operator; LR: logistic regression.