• Best method for predicting malignancy in pelvic mass in HK [ Time Frame: 3 months after last subject enrolled ]
  The sensitivity, specificity and predictive powers for 4 different predicting methods - (i) IOTA simple rules followed by ROMA (if IOTA inconclusive), (ii) IOTA simple rules followed by expert ultrasound assessments (if IOTA inconclusive), (iii) ROMA, or (iv) RMI will be compared.
• Accuracy of these 4 different prediction methods in different hospital settings [ Time Frame: 3 months after last subject enrolled ]
  The sensitivity, specificity and predictive powers for these 4 different prediction methods for the 3 hospitals will be compared and factors influencing the accuracy will be explored.
• Performance of IOTA simple rules followed by ROMA (if IOTA inconclusive) in different histological subtypes [ Time Frame: 3 months after last subject enrolled ]
  Since the tumour markers assessed by ROMA may be more commonly raised in some subtypes (eg serous adenocarcinoma) compared to other subtypes (eg mucinous adenocarcinoma), it is possible that the performance of IOTA followed by ROMA may differ in different histological subtypes. The final histological subtypes will be obtained from the pathology reports from the surgery. The sensitivity and specificity of IOTA followed by ROMA will be compared amongst the different major histological subtypes.

Our objective is to determine if the combination of simple ultrasound features (IOTA simple rules) and a new biomarker (HE4) together with a common tumour marker (CA 125) can accurately predict ovarian cancer in women found to have a pelvic mass on ultrasound. The investigators hypothesize that the use of two biomarkers (HE4 and CA 125) in a mathematical algorithm (Risk of Malignancy Algorithm, ROMA) can be used to predict malignancy in a pelvic mass which has indeterminate ultrasound features. This is a prospective cohort study involving women undergoing operation for a pelvic mass. 720 women scheduled to have an operation to remove a pelvic mass would be recruited from 3 hospitals (QMH, UCH and PYNEH). Pre-operatively, each woman will have an ultrasound assessment using the IOTA simple rules criteria and have blood taken for tumour markers HE4 and CA 125. In women where IOTA ultrasound rules are inconclusive, 2 strategies for prediction will be compared - calculation of risk by ROMA (Strategy A) vs referral for an expert ultrasound (Strategy B). These pre-operative risk predictions will be correlated with the final pathology found at the operations.

Main outcome measures include the sensitivity, specificity, positive and negative predictive powers for Strategy A compared to Strategy B. Sensitivity and specificity will be compared using the McNemar test. Area Under the ROC Curve (AUC) will be calculated and compared using the Delong method for the 2 strategies.

The investigators expect AUC of both strategies will be similar. This would suggest that ROMA can replace expert ultrasound in the pre-operative prediction of ovarian cancer.

Inclusion Criteria:

• Women over the age of 18
• Found to have a pelvic mass on ultrasound, MRI, CT or PET scan
• Scheduled for operation (laparoscopic or open) for a pelvic mass (including ovarian cystectomy and oophorectomy)
• Women who would understand the informed consent

Exclusion Criteria:

• Women who refused a transvaginal scan
• Pregnant women
• Surgical removal is delayed for more than 120 days from the date of the ultrasound examination
• Previous history of ovarian, peritoneal or fallopian tube cancer or unknown malignancy
• History of bilateral oophorectomy

• United Christian Hospital
• Pamela Youde Nethersole Eastern Hospital, Hong Kong