The aim of this study is to evaluate a new tumour marker, HE4, in comparison with CA 125 and the Risk of Ovarian Malignancy Algorithm (ROMA) in healthy women and in patients with benign and malignant gynaecological diseases. CA 125 and HE4 serum levels were determined in 66 healthy women, 285 patients with benign gynaecological diseases (68 endometriosis, 56 myomas, 137 ovarian cysts and 24 with other diseases), 33 patients with non-active gynaecological cancer and 143 with active gynaecological cancer (111 ovarian cancers). CA 125 and HE4 cut-offs were 35 U/mL and 150 pmol/L, respectively. ROMA algorithm cut-off was 13.1 and 27.7 for premenopausal or postmenopausal women, respectively. HE4, CA 125 and ROMA results were abnormal in 1.5%, 13.6% and 25.8% of healthy women and in 1.1%, 30.2% and 12.3% of patients with benign diseases, respectively. Among patients with cancer, HE4 (in contrast to CA 125) had significantly higher concentrations in ovarian cancer than in other malignancies (p < 0.001). Tumour marker sensitivity in ovarian cancer was 79.3% for HE4, 82.9% for CA 125 and 90.1% for ROMA. Both tumour markers, HE4 and CA 125 were related to tumour stage and histological type, with the lowest concentrations in mucinous tumours. A significantly higher area under the ROC curve was obtained with ROMA and HE4 than with CA 125 in the differential diagnosis of benign gynaecological diseases versus malignant ovarian cancer (0.952, 0.936 and 0.853, respectively). Data from our population indicate that ROMA algorithm might be further improved if it is used only in patients with normal HE4 and abnormal CA 125 serum levels (cancer risk for this profile is 44.4%). ROMA algorithm in HE4 positive had a similar sensitivity and only increases the specificity by 3.2% compared to HE4 alone.