Diagnosis of polycystic ovary syndrome (PCOS): revisiting the threshold values of follicle count on ultrasound and of the serum AMH level for the definition of polycystic ovaries

TLDR

Polycystic ovarian morphology (PCOM) at ultrasound is currently used in the diagnosis of polycystic ovary syndrome (PCOS). The study hypothesizes that the previously proposed follicle threshold of 12 per ovary is outdated due to technological advances and revisits thresholds for follicle number and serum AMH for defining PCOM. The authors performed cluster analysis of hormonal and ultrasound variables and ROC curve analysis to identify optimal thresholds distinguishing PCOM from non‑PCOM groups. The ROC analysis yielded AUCs of 0.949 for follicle number and 0.973 for AMH, with optimal thresholds of 19 follicles and 35 pmol l⁻¹ (5 ng ml⁻¹); the former >12 follicle cut‑off is invalid, and AMH >35 pmol l⁻¹ is more sensitive and specific than FN >19, warranting its inclusion in PCOS diagnostic criteria.

Abstract

Polycystic ovarian morphology (PCOM) at ultrasound is currently used in the diagnosis of polycystic ovary syndrome (PCOS). We hypothesized that the previously proposed threshold value of 12 as an excessive number of follicles per ovary (FN) is no longer appropriate because of current technological developments. In this study, we have revisited the thresholds for FN and for the serum Anti-Müllerian hormone (AMH) level (a possible surrogate for FN) for the definition of PCOM.Clinical, hormonal and ultrasound data were consecutively recorded in 240 patients referred to our department between 2008 and 2010 for exploration of hyperandrogenism (HA), menstrual disorders and/or infertility.According to only their symptoms, patients were grouped as: non-PCOS without HA and with ovulatory cycles (group 1, n = 105), presumption of PCOS with only HA or only oligo-anovulation (group 2, n = 73) and PCOS with HA and oligo-anovulation (group 3, n = 62). By cluster analysis using androgens, LH, FSH, AMH, FN and ovarian volume, group 1 appeared to be constituted of two homogeneous clusters, most likely a non-PCOM non-PCOS subgroup (n = 66) and a PCOM, non-PCOS (i.e. asymptomatic) subgroup (n = 39). Receiver operating characteristic curve analysis was applied to distinguish the non-PCOM non-PCO members of group 1 and to group 3. For FN and serum AMH respectively, the areas under the curve were 0.949 and 0.973 and the best compromise between sensitivity (81 and 92%) and specificity (92 and 97%) was obtained with a threshold values of 19 follicles and 35 pmol/l (5 ng/ml).For the definition of PCOM, the former threshold of >12 for FN is no longer valid. A serum AMH >35 pmol/l (or >5 ng/ml) appears to be more sensitive and specific than a FN >19 and should be therefore included in the current diagnostic classifications for PCOS.