The accurate staging of ovarian cancer using 3T magnetic resonance imaging - a realistic option

Objectives The aim of the study was to determine whether staging primary ovarian cancer using 3.0 Tesla (3T) magnetic resonance imaging (MRI) is comparable to surgical staging of the disease. Design A retrospective study consisting of a search of the pathology database to identify women with ovarian pathology from May 2004 to January 2007. Setting All women treated for suspected ovarian cancer in our cancer centre region. Sample All women suspected of ovarian pathology who underwent 3T MRI prior to primary surgical intervention between May 2004 and January 2007. Methods All women found to have ovarian pathology, both benign and malignant, were then cross checked with the magnetic resonance (MR) database to identify those who had undergone 3T MRI prior to surgery. The resulting group of women underwent comparison of the MR, surgical and histopathological findings for each individual including diagnosis of benign or malignant disease and International Federation of Gynecology and Obstetrics (FIGO) staging where appropriate. Main outcome measures Comparisons were made between the staging accuracy of 3T MRI and surgical staging compared with histopathological findings and FIGO stage using weighted kappa. Sensitivity, specificity and accuracy were calculated for diagnosing malignant ovarian disease with 3T MRI. Results A total of 191 women identified as having ovarian pathology underwent imaging with 3T MR and primary surgical intervention. In 19 of these women, the ovarian disease was an incidental finding. The group for which staging methods were compared consisted of 77 women of primary ovarian malignancy (20 of whom had borderline tumours). 3T MRI was able to detect ovarian malignancy with a sensitivity of 92% and a specificity of 76%. The overall accuracy in detecting malignancy with 3T MRI was 84%, with a positive predictive value of 80% and negative predictive value of 90%. Statistical analysis of the two methods of staging using weighted kappa, gave a K value of 0.926 (SE +/- 0.121) for surgical staging and 0.866 (SE +/- 0.119) for MR staging. A further analysis of the staging data for ovarian cancers alone, excluding borderline tumours resulted in a K value of 0.931 (SE +/- 0.136) for histopathological staging versus MR staging and 0.958 (+/- 0.140) for histopathological stage versus surgical staging. Conclusion Our study has shown that MRI can achieve staging of ovarian cancer comparable with the accuracy seen with surgical staging. No previous studies comparing different modalities have used the higher field strength 3T MRI. In addition, all other studies comparing radiological assessment of ovarian cancer have grouped the stages into I, II, III and IV rather than the more clinically appropriate a, b and c subgroups.