What's in a number? The polycystic ovary revisited.

So 19 is the new 12! This is certainly a long overdue change, but is the cut-off value proposed by Dewailly et al., in a well-designed study and elegant cluster analysis presented in this edition of Human Reproduction, the right one? Only time and external validation will tell but what is certain is that the old figures, proposed by the same group in 2003 (Jonard et al., 2003) and subsequently adopted by the Rotterdam consensus (2004a), need to be reconsidered urgently (2004b). We also see the introduction of anti-Müllerian hormone (AMH) as a potential new diagnostic criterion. Whether these variables can be used in conjunction or as separate entities remains to be seen, but the known close correlation between AMH and the number of antral follicles (Andersen et al., 2010; Jayaprakasan et al., 2010) suggests that they will not offer mutually exclusive information. Dewailly et al. suggest that AMH is the better of the two markers. The data are welcome as the current ultrasound criteria identify a morphological feature that has been shown to be present in up to 30% of the general population (Duijkers and Klipping, 2010; Johnstone et al., 2010; Kristensen et al., 2010). Can something that common truly be considered a ‘disease’? We will undoubtedly see a reduction in the apparent prevalence of polycystic ovaries (PCO) if the criteria are modified on these new data from Dewailly et al., but are the new thresholds enough or do we need to reconsider the diagnosis in more detail? In this study, Dewailly et al. used cluster analysis to reconsider the threshold values suggested by the Rotterdam consensus. This was based on the hypothesis that women with PCO in isolation, or as the authors say ‘polycystic ovarian morphology’ (PCOM), and therefore no features suggestive of anovulation or hyperandrogenism, are normal and no different from the control population. In essence they would simply represent the upper centile of patients as reflected by their own antral follicle and AMH reference ranges. Their decision to exclude these patients, subsequently defined after cluster analysis as ‘Group 1b’, from the control group means they have potentially excluded normal women. Are these not the women we need to consider in more detail, namely those with isolated PCO? The authors validate their exclusion by saying that these women tend to have slightly higher mean serum androgens (Adams et al., 2004; Mortensen et al., 2009) or AMH levels (Johnstone et al., 2010). This will depend on the criteria one uses to define normality and any respective cut-off levels applied. We are using percentiles to define disease rather than clinically relevant features. The authors go on to suggest that PCOM in women without oligoamenorrhoea and/or hyperandrogenism may still ‘ . . . represent a functional entity that may be considered as a silent form of PCOS’—a concept originally suggested by Franks et al. (2008). They also suggest that serum AMH might be the best marker to define such a group. This brings us back to the application of thresholds, which are based on the premise that a disease is present or not: a dichotomy rather than a continuum. The cut-off, as readers will know, simply represents the best compromise between the sensitivity and specificity of the test, which in this study was 81 and 92%, respectively for a threshold of 19 follicles per ovary. It is a statistical entity as opposed to a clinically relevant discriminator. The introduction of AMH makes sense as it has become a stable assay and provides a reliable measure of ovarian reserve that can be ascertained at any time in the menstrual cycle (La Marca et al., 2004). It is this combination of predictive power and simplicity, which makes AMH so attractive. Counting antral follicles is less straightforward and more time-consuming. The inherent intraand inter-observer variation seen in measuring the number of antral follicles can be reduced but not eliminated by the use of threedimensional (3D) ultrasound and, more recently, automated ultrasound. AMH more closely correlates with the number of small antral follicles (Pigny et al., 2003), which are harder to identify separately from larger follicles with ultrasound. The cohort of small follicles measuring 1–5 mm do seem to more accurately reflect ovarian reserve and are increased in women with polycystic ovarian syndrome as Dewailly’s group showed in their original 2003 paper (Jonard et al., 2003). These smaller follicles do generally contribute most to the overall pool of antral follicles, which is why counts that include larger follicles, such as those measuring 9 or 10 mm, are still clinically valid, as they are not significantly diluted by the cohort of antral follicles measuring 6 mm and more. Such counts are also easier to perform, and therefore more practical, as the observer needs to exclude only follicles above 9 or 10 mm. Automated 3D measures do provide reliable information on follicle number and size and can be used to individually identify each follicle (Deb et al., 2010) and appear to be more reflective of ovarian reserve and response (Franks et al., 2008; Deb et al., 2009) but their accuracy remains to be determined and requires further work. The authors suggest that AMH is more specific than AFC for the diagnosis of the polycystic ovary and go on to say that ultrasound is also ‘ . . . less sensitive than