Differential diagnosis of Krukenberg tumors using multivariate analysis.

Aims and background: Krukenberg tumors are ovarian metastases from a gastroenteric malignancy in 90% of cases. At present, diagnostic imaging techniques (US, CT, MRI) do not provide any reliable diagnostic criteria to differentiate these metastases from ovarian primaries. We tried to use multivariate analysis to distinguish malignant ovarian primaries from Krukenberg tumors based on their differential natural history. Patients and methods: We retrospectively compared the CT findings of Krukenberg tumors (47 patients, 79 masses, only from gastroenteric malignancy) with CT findings of primary epithelial ovarian cancers (46 patients, 67 masses). We compared the same independent variables in the two groups: age, morphology, margins, carcinomatosis, bilateral versus unilateral involvement, ascites, lymph node involvement, metastases on the basis of multivariate analysis. Results: According to the best fitted model, clear-cut margins (OR: 3.75; 95% Cl: 1.14-9.72) and the presence of carcinomatosis (OR: 4.21; 95% Cl: 1.51-11.72) were the strongest predictors of a diagnosis of Krukenberg tumor. In contrast, the presence of ascites was more likely to be a protective factor (OR: 0.22; 95% Cl: 0.08-0.62). Conclusions: We can try to make a differential diagnosis between a metastatic lesion from the gastroenteric tract and a primary adnexal lesion based on the multivariate statistical analysis of intraperitoneal spread of the different types of cancer rather than on morphologic findings at CT.