Valproate, polycystic ovary syndrome and the need for a prospective study

EDITOR — We welcome the publication of Best Practice Guidelines for the Management of Women with Epilepsy1. The Guidelines raise some important issues and will hopefully help to improve management standards. Unfortunately, the section discussing treatment issues and menstrual irregularities contains some important inaccuracies. Like many recent publications, the Guidelines do not differentiate between a polycystic ovarian morphology (PCO) and polycystic ovary syndrome (PCOS). PCO can be detected with ultrasound and is found in 22–33% of the adult female population in the UK2. When PCO is associated with elevated androgen levels, luteinizing hormone concentrations or any clinical symptoms of androgen excess it becomes polycystic ovary syndrome (PCOS). Estimates of the prevalence of PCOS vary from 5 to 26% of adult women3. PCO and PCOS are more common in obese women. It is untrue that the retrospective study published by Isojarvi et al. in 1996 noted an incidence of PCOS of 64% (14/22) in valproate-treated women. Isojarvi et al. combined women with hyperandrogenism without polycystic ovaries on ultrasound (PCO) and women with PCO without hyperandrogenism to reach the figure of 64%4. Whilst the definition of polycystic ovary syndrome (PCOS) may be a disputed issue3, Isojarvi et al. never indicated how many women with PCOS (by anyone’s definition) they actually found. The same is true for the 1993 study5 which describes the same valproate-treated patients as the 1996 paper. What is more, Isojarvi et al. never published the pretreatment body mass indices (BMIs) of the women described in the two studies. It is possible that the valproate-treated group had higher pre-anticonvulsant BMIs that the carbamazepine-treated group and was therefore at higher risk of PCO, PCOS or hyperandrogenism. We do not think that the situation was clarified by the (in fact) uncontrolled prospective study in which women with PCOS (PCO and hyperandrogenism) were switched from valproate to lamotrigine6. Further doubt about the findings of Isojarvi et al. were raised by two recent retrospective studies which did not find any increased incidence of PCO in valproatetreated women7, 8. We feel strongly that the issues surrounding the possible hormonal and metabolic side effects of valproate can only be resolved in an appropriately powered, prospective, randomized-controlled trial. The multicentre APOS study (anticonvulsants and polycystic ovary syndrome) has been designed to this end. This project is an ‘add-on’ protocol to the study of ‘standard and new anticonvulsant drugs’ (SANAD). APOS and SANAD together should finally settle the questions of whether and how commonly valproate causes unwanted hormonal side effects in women (if so, in which women), and whether lamotrigine or topiramate are safe and effective alternatives. We would like all physicians interested in the resolution of these important issues to join the authors in the APOS and SANAD projects.