Morphological criteria for the diagnosis of Polycystic Ovarian Syndrome (PCOS) by 3-dimensional (3D) transvaginal ultrasound (TVUS)

Abstract

The TVUS characteristics of polycystic ovaries (PCO) remain controversial. A recent consensus opinion revised the recommended criteria for PCO from the previous criteria of ≥ 10 follicles in a single sonographic plane, to a follicle number per ovary (FNPO) ≥ 12, and/or increased ovarian volume >10 mL. This study compares ovarian volume, maximum number of follicles in a single sonographic plane (FSSP), and FNPO between normoandrogenic ovulatory (NO) women and PCOS patients to determine the accuracy of diagnosing PCO by TVUS. Retrospective cohort study. Thirty-one normoandrogenic women (luteal serum progesterone >3.0 ng/mL, modified Ferriman-Gallwey score ≤ 7) and 10 PCOS patients (hyperandrogenic anovulation excluding other endocrinopathies) were studied. All subjects underwent two-dimensional (2D) TVUS to assess mean ovarian volume of both ovaries (if present) and to rule out follicles ≥ 10mm in size. 3D TVUS also was performed and stored on CD-ROM to determine the maximum FSSP of either ovary, and the mean FNPO of both ovaries, as assessed by 2–4 independent observers, blinded to each other’s results. TVUS was performed in NO women and PCOS patients on cycle day 5 of the follicular phase and during a period of amenorrhea, respectively. Ovarian volume, FSSP, and FNPO were compared between the two groups using t-test and receiver operator characteristic curves (ROC). Intraclass coefficients of correlation (ICC) were calculated to analyze interobserver variation (N=2 observers) for FSSP and FNPO estimates. Results are expressed as means ± SD (P < 0.05 considered significant). The mean ovarian volume for the NO women was 7.15 ± 2.14 cm3 (95% CI 6.36–7.94) and for PCOS patients was 13.56 ± 3.52 cm3 (95% CI 11.04–16.08), P<0.0001. The average maximum FSSP was greater in PCOS patients (13.3 ± 3.21[95% CI 11.01–15.60]) than in NO women (6.5 ± 1.43[95% CI 5.94–7.06)], P<0.0001). The mean FNPO was also greater in PCOS patients (29.80 ± 11.53[95% CI 21.56–38.05]) than in NO women (9.89 ± 3.59[95% CI 8.55–11.23], P<0.0001). The ICC for the maximum FSSP and mean FNPO were 0.72 and 0.82, respectively. The diagnostic value determined by ROC as characterizing PCOS by ovarian volume was 12.9 cm3, with a specificity of 100% and a sensitivity of 50% (area under the curve (AUC) 94.8%). A maximum FSSP of 10.0 predicted PCOS with 100% specificity and 80% sensitivity (AUC 97.7.0%), while an average FNPO of 19.8 predicted PCOS with 100% specificity and 80% sensitivity (AUC 98.7%). Our data demonstrate the maximum FSSP and mean FNPO are comparable in characterizing PCO by 3D TVUS, followed by ovarian volume by 2D TVUS. The ICC for the maximum FSSP indicates strong agreement between the two analyzed observers, while that for the mean FNPO shows an excellent correlation. The maximum FSSP of 10.0 agrees with the previously accepted TVUS diagnostic criteria for PCO (i.e., the NIH Consensus Group, 1990), while the FNPO value of 19.8 is greater than the newly recommended FNPO value of ≥ 12 (i.e., the Rotterdam PCOS Consensus Group). Consequently, 3D-TVUS can be used to accurately identify PCO but also will overdiagnose this condition based upon current recommendations.