Diagnosis of polycystic ovaries by three-dimensional transvaginal ultrasound.

Objective: To study diagnostic thresholds for polycystic ovary (PCO).

Design: Retrospective cohort study.

Setting: Academic hospital.

Patient(s): Normoandrogenic ovulatory women and patients with polycystic ovary syndrome (PCOS).

Intervention(s): Two-dimensional (2D) and three-dimensional (3D) transvaginal ultrasound.

Main Outcome Measure(s): The mean follicle number per ovary (FNPO) of both ovaries and the maximum number follicles in a single sonographic plane (FSSP) of either ovary were determined using 3D transvaginal ultrasound. Ovarian volume was determined using 2D transvaginal ultrasound.

Result(s): Twenty-nine normoandrogenic ovulatory women were compared with 10 patients with PCOS. Diagnostic thresholds for PCO with 100% specificity as determined by receiver operator characteristic (ROC) curves were > or =20 for mean FNPO, > or =10 for maximum FSSP, and > or =13 cm3 for ovarian volume. Both 2D and 3D transvaginal ultrasound were highly accurate in the diagnosis of PCO as determined by areas under the curve (AUC) that were >90% for all three measures.

Conclusion(s): Mean FNPO and maximum FSSP by 3D transvaginal ultrasound have comparable high accuracy for diagnosis of PCO. The diagnostic threshold with 100% specificity for mean FNPO is > or =20, which is greater than suggested by the Rotterdam Consensus Workshop in 2003. Use of the consensus standard, consequently, may result in overdiagnosis of PCO. A threshold of > or =20 mean FNPO using 3D transvaginal ultrasound may be appropriate to minimize false-positive diagnoses of PCO.