SEARCH

SEARCH BY CITATION

Keywords:

  • ovarian neoplasms;
  • pelvic neoplasms;
  • regression analysis;
  • ultrasonography

Abstract

Objectives

To determine which extrauterine pelvic masses are difficult to correctly classify as benign or malignant on the basis of ultrasound findings, and to determine if the use of logistic regression models for calculation of individual risk of malignancy would improve the diagnostic accuracy in difficult tumors.

Methods

In a prospective, international, European multicenter study involving nine centers, 1066 women with a pelvic mass judged to be of extrauterine origin underwent transvaginal ultrasound examination by an experienced ultrasound examiner before surgery. A standardized examination technique and predefined definitions of ultrasound characteristics were used. On the basis of subjective evaluation of ultrasound findings, the examiner classified each mass as being certainly benign, probably benign, unclassifiable, probably malignant or certainly malignant. Even when the examiner found the mass unclassifiable (i.e. difficult mass) he or she was obliged to state whether the mass was more likely to be benign or malignant. Borderline tumors were classified as malignant.

Results

There were 90 (8%) unclassifiable masses. Multiple logistic regression analysis showed papillary projections, >10 locules in a cyst without solid components, low-level echogenicity of cyst fluid, and moderate vascularization as assessed subjectively at color Doppler examination to be ultrasound variables independently associated with unclassifiable mass. Borderline malignant tumors (n = 55) proved to be most difficult to assess with only 47% being correctly classified (i.e. classified as malignant), 29% being incorrectly classified (i.e. classified as benign) and 24% being unclassifiable vs. 90% of non-borderline tumors being correctly classified, 3% being incorrectly classified and 8% being unclassifiable (P < 0.0001). Papillary cystadeno(fibro)mas, myomas and cases of struma ovarii were also more common among the unclassifiable masses than among the classifiable ones (5.6% vs. 1.1%, P = 0.008; 4.4% vs. 0.9%, P = 0.02; 4.4% vs. 0.2%, P = 0.0006). No ultrasound variable or clinical variable (including CA 125) entered a logistic regression model to predict malignancy in difficult masses. A model could be constructed for difficult masses containing papillary projections but this model performed no better than subjective evaluation of the ultrasound image. Sensitivity and specificity of subjective evaluation with regard to malignancy in the group of unclassifiable masses were 56% (14/25) and 77% (50/65) vs. 91% (220/241) and 97% (712/735) in the classifiable masses.

Conclusions

Borderline tumors cause great diagnostic difficulties, but so do papillary cystadeno(fibro)mas, struma ovarii and some myomas. Logistic regression models do not solve the diagnostic problem in difficult pelvic masses. Copyright © 2006 ISUOG. Published by John Wiley & Sons, Ltd.