Abstract:    Ovarian cancer accounts for 4% of new cases of cancer and for 6% of cancer deaths in women. The prognosis of the disease is generally poor with an overall five year survival of approximately 30%. Approximately 85-90% of ovarian neoplasms are of epithelial origin (derived from tissues that come from the mesothelium). These tumors may be benign (50%), malignant (33%), or borderline malignant (16%). The serous histologic type is the most common epithelial tumour of the ovary (46-75%) and will be the focus of our attention here. About 30% of ovarian cancer patients are diagnosed with early-stage disease and about 10%-50% of them will have a recurrence after initial surgery. Most women with advanced disease will respond to initial (chemo)therapy but most of them will eventually relapse. Presently, no clinical parameters are available that can reliably predict chemosensitivity in FIGO stage III ovarian cancer (tumour with abdominal extension or extension to regional nodes) or the probability of recurrence after initial surgery in FIGO stage I ovarian cancer (tumour limited to one or both ovaries). Therefore we (in cooperation with Prof. I. Vergote and Prof. D. Timmerman, department of Obstetrics and Gynaecology of the University Hospitals Leuven, and Dr. P. Van Hummelen of the Microarray Facility of the Flanders Interuniversity Institute for Biotechnology (V.I.B.)) aim to develop and test models that use cDNA-microarray data and that:

1. Predict if a stage III ovarian tumour will relapse within 6 months after the last therapeutic intervention. Since standard chemotherapy for advanced ovarian cancer is usually platinum based (e.g., carboplatinum + paclitaxel), this model will be able to predict platinum resistance (or chemosensitivity of the tumour). This has mainly prognostic significance but might allow to develop new therapeutic strategies in the future for tumours that are predicted not to respond adequately to the standard chemotherapeutic regimen.
2. Predict if a stage I ovarian tumour will have a recurrence after initial surgery. The subset of women with early-stage disease and, according to our model, with a high probability of recurrence are ideal candidates that might maximally benefit from adjuvant treatment (chemotherapy and/or lymphadenectomy) while the women with early-stage disease and a low probability of recurrence might be spared the side-effects of adjuvant therapy.

In a pilot study, RNA obtained from 7 FIGO stage I without recurrence, 7 platin-sensitive advanced-stage (III or IV) and 6 platin-resistant advanced-stage ovarian tumors was hybridized on a cDNA microarray with 21372 spotted clones. The results revealed that a considerable number of genes exhibit non-accidental differential expression between the different tumor classes. Principal component analysis reflected the differences between the three tumor classes and their order of transition. Using a leave-one-out approach together with least squares support vector machines, we obtained an estimated classification test accuracy of 100% for the distinction between stage I and advanced-stage disease and 76.92% for the distinction between platin-resistant versus platin-sensitive disease in FIGO stage III/IV. These results indicate that gene expression patterns could be useful in clinical management of ovarian cancer. These results will be validated on a larger group of patients (about 100) in the near future.