Multivariate assessment of cervical invasion of endometrial carcinoma: Comparison of transvaginal ultrasonography, hysteroscopy, and magnetic resonance imaging

Background: Preoperative assessment of cervical invasion of endometrial cancer is important in terms of surgical planning. It is of value to compare the usefulness of transvaginal ultrasonography, hysteroscopy, and magnetic resonance imaging (MRI) for evaluating the extent of the invasion. This study evaluated these 3 examinations by univariate and multivariate analysis, retrospectively, and investigated the factors responsible for misdiagnosis. Methods: From 1991 to 1995, 71 patients with endometrial cancer at Okayama University Medical School underwent transvaginal ultrasonography with a 6.5-MHz transducer, MRI at 1.5 T, and hysteroscopy using a rigid hysteroscope. The findings of these examinations were compared against pathologic results. Results: The sensitivity/specificity/accuracy were 0.73/0.89/0.86 for transvaginal ultrasonography, 0.73/0.88/0.85 for MRI, and 0.73/0.80/0.79 for hysteroscopy. The coefficient values for transvaginal ultrasonography/MRI/hysteroscopy were 0.69/0.62/-0.21 and 1.99/1.89/0.79 in quantification theory second family analysis and logistic regression, respectively. Misdiagnosed cases had the following factors: (1) large tumor volume; (2) tumor near the internal os; (3) tumor descending from the uterine cavity; (4) polypoid tumor in the endocervix; (5) artifacts due to the use of a cervix dilator made from stems of the seaweed Laminaria japonica;(6) endocervical glandular involvement only; and (7) atypical signals on MRI. Conclusion: For the preoperative assessment of cervical invasion, transvaginal ultrasonography and MRI were the most reliable. Hysteroscopy was least reliable, because of a large number of false-positive results. The use of these examinations in combination, especially transvaginal ultrasonography and MRI, would be of great value in surgical planning for endometrial cancer, considering the factors responsible for misdiagnosis.