MRI of cervical cancer involves potential pitfalls

November 2, 2005

MRI is an established tool for the staging and preoperative assessment of cervical cancer, but a team of U.K. radiologists has warned that awareness of its pitfalls is essential to avoid errors.

MRI is an established tool for the staging and preoperative assessment of cervical cancer, but a team of U.K. radiologists has warned that awareness of its pitfalls is essential to avoid errors.

Cervical cancer often presents as early disease in young women, and accurate staging is vital to ensure appropriate selection of patients for surgery and to define the gross tumor volume for radiotherapy, according to Dr. John Hughes, a radiologist at St. Bartholomew's Hospital in London.

MRI provides information about pelvic structures and para-aortic nodal and abdominal disease, avoiding the need for a separate cystoscopy, intravenous urogram, and barium enema. It can also help in the selection of patients suitable for fertility-preserving surgery.

At St. Bartholomew's, 210 women underwent staging of cervical cancer between 1998 and 2004. Ninety-one of the women had surgery, which allowed correlation of the MRI and surgical findings.

False-positive MR appearances are a potential pitfall in stage I and II disease. Of the 91 patients who underwent surgery, four had areas of high signal intensity within the cervix that were indistinguishable from cervical cancer. These areas were histologically proven as inflammatory change secondary to cone biopsies in all patients. In two cases, inflammatory changes resulted in false-positive MRI appearances, Hughes told attendees at the 2005 European Congress of Radiology.

Eleven of the 91 patients had no tumor demonstrable on a pelvic MR examination. Eight of these women had a histological residual tumor measuring less than 1 cm in size. In one patient, the tumor was obscured by Nabothian follicles in the cervix; in another patient, the tumor was obscured by endometriosis of the cervix. Degradation of the images by motion artifact and bowel peristalsis obscured the underlying tumor in three patients.

Technique, parameter setting, and use of the correct plane in oblique images are critical.

"The correct plane for oblique T2-weighted scans through the cervix is perpendicular to the long axis of the cervix. This results in a cross-section of the cervix (doughnut shape), thereby demonstrating the parametrium around the cervical stroma," Hughes said.

Visualization of tumor surrounded by the low T2-weighted signal intensity of normal cervical stroma confers a negative predictive value of between 91% and 96% for parametrial invasion, he said.

Small exophytic tumors present another potential pitfall. The tumors may lie in the posterior vaginal fornix, where they can be missed. When large exophytic tumors extend into the vaginal lumen, the surrounding vaginal wall may be misinterpreted as preservation of cervical stroma.

Localized thickening of the vagina and loss of the normal T2-weighted signal intensity of the wall are the signs of tumor infiltration, and segmental thickening can appear as tumor invasion.

Pelvic lymph node dissection was performed in all of the 91 women undergoing surgery. The sensitivity of MRI for detection of nodal metastases was 63%, specificity was 56%, and overall accuracy was 89%, Hughes said. In four patients, tumor metastases were present in pelvic nodes that were undetectable on MRI. In another three patients, metastatic nodes were less than 7 mm on their short axis. Seven patients had lymph nodes greater than 7 mm, but only reactive changes were evident histologically.

"Recent advances in lymph-node-specific contrast agents have improved the sensitivity of MRI to 91% to 100% and the specificity to 94%," he said.

Cystic nodes are a potential pitfall because they may be misinterpreted as ovaries. One patient had concurrent endometriosis in the cervix, which replaced the entire stroma, leading to overestimation of the true tumor size.

In stage IIIa, the pitfalls that apply to interpretation of the proximal two-thirds of the vaginal wall are also relevant to the distal third. The exact delineation of the distal third of the vagina is not clear. It is arbitrarily considered to be distal to the bladder base, but definition is unreliable if the vagina or bladder base is prolapsed, Hughes said.

Stage IIIb involves invasion of the pelvic sidewall. Invasion of the uterosacral ligaments, which form part of the pelvic sidewall, should be examined carefully, he said.

In another ECR 2005 presentation, Japanese researchers reported on the correlation of MR and pathologic findings to assist diagnosis and staging of adenocarcinomas.

"Recognition of the problems and limitations in MR staging may encourage radiologists to carefully interpret findings," said Dr. Takashi Koyama, a radiologist at Kyoto University.

Sixty cases of pathologically confirmed invasive cervical adenocarcinoma were seen at Kyoto between 1986 and 2004. Adenocarcinomas account for up to 26% of primary carcinomas of the uterine cervix, and their incidence has increased, according to Koyama.

Cervical adenocarcinomas are generally more resistant to chemotherapy and radiotherapy, and the prognosis of patients with cervical adenocarcinoma is worse than that of patients with squamous cell carcinoma of the same stage.

Cervical adenocarcinomas are typically seen as infiltrative lesions of increased intensity on T2-weighted images, arising from either the cervical canal or lip. MRI may fail to demonstrate invasive cancer confined within the stroma. Although adenocarcinomas often infiltrate the vaginal wall and parametrium, the sensitivity for these invasions is not high, Koyama said.

The typical age of patients with adenocarcinoma of the cervix is between 47 and 53. Their clinical symptoms are similar to those of patients with squamous cell carcinoma (i.e., a watery discharge or abnormal genital bleeding), but they may remain asymptomatic. Early-stage adenocarcinomas are difficult to detect with cytologic smear because abnormal cells are more deeply located and frequently lack cytologic atypia.

In cervical adenocarcinomas, MRI may show preserved endocervical epithelium on postcontrast T1-weighted imaging, even in large tumors involving the entire cervix, but this finding may not be apparent on ordinary T2-weighted images. The imaging findings vary considerably, probably reflecting the variety of pathologic features in cervical adenocarcinomas, Koyama said.