On the Association Between Valproate and Polycystic Ovary Syndrome


Address correspondence and reprint requests to Dr. P. Genton at Centre Saint Paul, 300 Boulevard de Sainte Marguerite, 13009 Marseille, France. E-mail: PIERGEN@aol.com


Summary: Recent studies by Isojärvi et al. have raised the issue of an increased incidence of polycystic ovary syndrome (PCOS) in women with epilepsy treated with valproate (VPA) and have proposed replacement with lamotrigine (LTG). Polycystic ovaries (PCO) are a common finding, with a prevalence >20% in the general population, and are easily detected by pelvic or vaginal ultrasonography, whereas PCOS is comparatively rare: few women with PCO have fully developed PCOS, which includes hirsutism, acne, obesity, hypofertility, hyperandrogenemia, and menstrual disorders. From an extensive review of the current literature, it appears that there are no reliable data on the actual prevalence of PCOS in normal women and in women with epilepsy. The pathogenesis of PCO is multifactorial, including genetic predisposition and the intervention of environmental factors, among which weight gain and hyperinsulinism with insulin resistance may play a part. The roles of central (hypothalamic/pituitary), peripheral, and local ovarian factors are still debated. PCO and PCOS appear to be more frequent in women with epilepsy, but there are no reliable data showing a greater prevalence after VPA. The recent studies by Isojärvi et al. may have been biased by the retrospective selection of patients. To date, there is no reason to contraindicate the use of VPA in women with epilepsy. However, patients should be informed about the risk of weight gain and its consequences.

Reproductive and endocrine functions are major concerns for clinicians who treat women with epilepsy. For example, menstrual and reproductive disturbances may occur. Among these, polycystic ovaries (PCO) are a fairly common finding, but fully developed polycystic ovary syndrome (PCOS) has a lower incidence (1). Isojärvi et al. (2) has raised the issue of an increased occurrence of PCO and hyperandrogenism in women treated with valproate (VPA) (2). In a more recent article, group has proposed that VPA be contraindicated in young women with epilepsy and replaced with lamotrigine (LTG) to reverse the morphologic and endocrine changes (3). However, other possible causative factors have been implicated in the genesis of PCO, PCOS, and reproductive disorders in young women with epilepsy, and the basis of these conditions remains controversial.

VPA has been used for >30 years in the treatment of various forms of epilepsy and is currently used by >2 million patients. It is particularly useful in the treatment of juvenile-onset idiopathic generalized epilepsies (4) and is, indeed, in the treatment of all forms of generalized seizures (5). It also is approved for treating mood disorders (6) and migraine (7) in North America. As VPA does not have enzyme-inducing properties, it represents a major treatment option in girls and in young women with epilepsy, both because of its efficacy and because it does not reduce the effectiveness of oral contraceptives.

In this article, we examined the nature and pathophysiology of PCOS and the impact of antiepileptic drugs (AEDs) on reproductive functions in women with epilepsy. From an extensive review of the literature, we propose practical clinical guidelines for the use of VPA in women with epilepsy.


According to Franks (8), “Few subjects have provoked such controversy in the field of reproductive endocrinology as polycystic ovary syndrome.” The controversy is due to very heterogeneous clinical and endocrine features and to major uncertainties in the definition of PCOS. Its pathogenesis is largely unknown, and probably multifactorial. The first description of PCOS by Stein and Leventhal in 1935 (9) characterized the uncommon, fully developed syndrome of amenorrhea, obesity, hirsutism, and bilateral sclerotic ovaries, in which the ovarian changes were characterized histologically (10). However, even in that first series, there was heterogeneity of clinical features among the patients. The concept of PCOS was widened in subsequent years, especially when the detection of ovarian changes was made possible by ultrasonography (US). The definition of PCO is presented in Table 1 and the symptoms associated with PCOS in Table 2.

Table 1.  Ultrasonic/anatomic definition of polycystic ovaries
  1. From ref. 11, with permission.

Multiple (≥10) cysts measuring 2–8 mm in diameter
Localization preferably peripheral, but also can be disseminated
Increased ovarian stroma and/or size
Table 2.  Main signs and symptoms of the polycystic ovarian syndrome
  1. The general consensus is of an association between polycystic ovaries, menstrual cycle disturbances, and hyperandrogenism. These features may be seen in various combinations.

  2. From ref. 12, with permission.

 Menstrual irregularity
 Hirsutism and/or alopecia
Endocrine and metabolic
 High luteinizing hormone concentrations
 High luteinizing hormone/follicle-stimulation hormone ratio
 High androgen production (testosterone, androstenedione,
  dehydroepiandrosterone sulfate)
 Abnormal lipid profile
 Insulin resistance, hyperinsulinism

Recent studies have focused on patients with US evidence of multiple ovarian cysts who exhibit a broad range of clinical and biologic correlates. Thus, PCOS is recognized as a heterogeneous syndrome with a spectrum of presentation (13), and the morphologic appearance of PCO can exist without any overt sign of the syndrome. According to Adams et al. (11), the identification of PCO by pelvic US relies on detecting an increased number of usually peripheral follicles measuring 2–8 mm in diameter (>10 on a single US plane) and an increase of stroma. This definition remains controversial: the ovaries may not be increased in size, which makes the identification of multiple follicles more difficult. Multiple cysts in ovaries of normal size may be found in hypothalamic amenorrhea, usually in the context of weight loss (14). A “polycystic-ovary–like” syndrome also can be found in women with major endocrinopathies [e.g., Cushing's syndrome, congenital adrenal hyperplasia, thyroid disease, or hyperprolactinemia (review in ref. 15)]. The homogeneous distribution of cysts throughout the stroma, or peripherally, and the presence of enlarged stroma, are sometimes difficult to assess.

PCOS has been defined by an NIH/NICHHD consensus conference (12), and the following criteria requires: ovulatory dysfunction, clinical evidence of hyperandrogenism and/or hyperandrogenemia, and exclusion of related disorders such as hyperprolactinemia, thyroid disorders, and nonclassic adrenal hyperplasia. This more restrictive definition excludes isolated findings of PCO, or of multifollicular ovaries, as well as isolated hyperandrogenism.

The prevalence of PCO by US evaluation in the general population is high (17–22%) (Table 3), and the prevalence of PCO is even higher among selected populations. Adams et al. (14) identified PCO in 32% of women with amenorrhea, 87% of women with oligomenorrhea, and 87% of women with idiopathic hirsutism (i.e., with regular ovulatory cycles). A study performed recently in the United States, using the 1992 NIH/NICHHD criteria, found a much lower prevalence of PCOS (i.e., of the fully developed syndrome): 4%(21).

Table 3.  Prevalence of polycystic ovaries in the general population
ReferenceStudy populationPCO
Diagnostic criteria
  1. PCO, polycystic ovaries; US, ultrasonography.

Cresswell et al. 1997 (16) (UK)235 women aged 40–42 years, from antenatal records21Abdominal or vaginal US, ≥10 cysts of diameter 2–8 mm with increase in stroma
Botsis et al. 1995 (17) (Greece)1,078 women aged 17–40 years, from routine Pap smear clinic17Pelvic US, ovarian volume >10 ml, increased ovarian stroma, cysts of diameter 5–10 mm
Farquhar et al. 1994 (18) (New Zealand)255 women aged 18–45 years, random from electoral roll21Abdominal US, ≥10 cysts of diameter 2–8 mm with increase in stroma
Clayton et al. 1992 (19) (UK)353 women born 1952–1969, randomly selected from clinical practice lists22Abdominal US, ≥10 cysts of diameter 2–8 mm with increase in stroma
Polson et al. 1988 (20) (UK)257 healthy volunteers aged 18–36 years22Abdominal US, ≥10 cysts of diameter 2–8 mm with increase in stroma

Conversely, the clinical changes associated with PCO are variable. Although most authors agree that women with PCO are more likely to have clinical problems associated with the syndrome, the prevalence of such symptoms may still be overestimated (Table 4). These data show that a minority of women with PCO are overweight (14–38%), have hirsutism (14–66%), or have fertility problems (18–41%). Menstrual disorders are the most common clinical finding associated with PCO (30–76%). It should be recognized that not all women with PCO are symptomatic, but long-term follow-up studies are not available.

Table 4.  Main clinical symptoms associated with polycystic ovaries
disorder (%)
problem (%)
  • MD, menstrual disorder; BMI, body mass index; NA, not applicable; NR, not reported; PCO, polycystic ovaries.

  • a

     p < 0.001 versus normal.

  • b

     Women seen in a gynecologic clinic.

Cresswell et al. 1997 (16)PCO: 48BMI: 25.4 kg/m24114a18
 Controls: 186BMI: 25.7 kg/m227215
Botsis et al. 1995 (17)PCO + MD: 14748%aNA45aNR
 PCO − MD: 3618% 20 
 Controls: 5010% 10 
Balen et al. 1995 (13)PCO: 1,741b38%706628
Farquhar et al. 1994 (18)PCO: 3923%46a23aNR
 Controls: 14419%204 
Clayton et al. 1992 (19)PCO: 4314%3014NR
 Controls: 165 7%272 
Franks 1989 (8)PCO: 300b35%526441
Polson et al. 1988 (20)PCO: 116NR76NRNR
 Controls: 33 <1  

The endocrine abnormalities associated with PCOS are multiple, but none is either pathognomonic or even universal. The most constant findings are elevated luteinizing hormone (LH) levels in serum and urine, elevated LH/follicle-stimulating hormone (FSH) ratio, elevated serum androgen levels [free and total testosterone (T), androstenedione, dehydroepiandrosterone sulfate], while FSH remains within normal limits, as do estrogens, which remain at normal early- or midfollicular levels. However, ≥50% of women with PCO morphology do not have elevated resting LH or T levels (13). There also is no correlation with elevated T levels and the presence of hirsutism. Furthermore, women with PCOS have lower sex hormone–binding globulin (SHBG) levels than normal (15), which leads to increased serum free T levels. There is therefore great heterogeneity in the endocrine correlates of PCO and PCOS. In practice, several of the classic hormonal changes reported in women with PCOS may be missing in any one individual, and their presence is, in part, determined by the clinical characteristics of PCOS patients.

Both the clinical and the endocrine correlates of women with PCO fluctuate widely among individuals, and, possibly, according to recruitment bias. In the words of Clayton et al. (19), “an isolated finding of polycystic ovaries may be a normal variation and should not necessarily imply altered fertility potential.”


Most authors have stressed that no single mechanism could account for all cases of PCOS, which is indeed a syndrome with many different causes, and appears to be a multifactorial condition (8). There is evidence for a role for central mechanisms, abnormalities of the hypothalamic–pituitary axis, hyperinsulinemia, and local factors within the ovary. In some individuals, environmental factors such as weight gain may be critical for the expression of the syndrome.

There appears to be a genetic basis for PCOS, although difficulties in defining the phenotype have caused problems in this area. PCO or clinical symptoms of PCOS are present in 31–87% of first-degree relatives of patients with fully developed PCOS (22–26). Acquired factors may interact with genetic factors: five of 19 homozygous twin pairs were discordant for PCO (27). Racial factors do not play a major part: there was no significant difference in the prevalence of PCOS between black (3%) and white (5%) women in a study in the Southeastern United States (21). Ethnic factors may, however, play a major part in the expression of some symptoms, notably hirsutism, which is common in Mediterranean women but rare in Japanese women (28). Androgens probably play an important part in the genesis of PCO. Although short-term administration of androgens fails to produce any significant ovarian change (29), PCO are frequently found in chronic states of hyperandrogenism of extraovarian origin (e.g., in adrenal hyperplasia, Cushing's syndrome or hyperprolactinism).

Several lines of evidence point to a primarily ovarian mechanism as the basis of PCOS. Local abnormal androgen production may be due to the presence of an aberrant 11β-hydroxylase (30), or to an abnormal regulation of the 17β-hydroxylase/17-20 lyase in intrathecal cells (31). Insulin and insulin-like growth factor 1 (IGF-1) increase synthesis of androgens by ovarian thecal cells after stimulation by LH (32,33). Insulin decreases the synthesis of IGF-binding protein 1 (IGFBP-1), which controls the ovarian action of IGF-1 (34) and also suppresses the synthesis of SHBG (35), which may lead to an increased bioavailability of T (36).

Central mechanisms that result in increased gonadotrophin secretion may also play a role in the pathophysiology of PCOS. Most women with PCOS have an elevated LH pulse frequency and amplitude, and recent studies suggest that they may be relatively resistant to progestin negative feedback. It is possible that altered gonadotrophin secretion at puberty is a risk factor for later development of PCOS. Feedback of ovarian steroidal and nonsteroidal hormones may further influence pituitary and hypothalamic function (37). LH secretion also is significantly negatively associated with obesity, such that the most obese PCOS patients tend to have less elevated LH levels (38,39).

Finally, the pathophysiology and etiology of PCOS is intimately linked with insulin resistance/hyperinsulinemia. Many women with PCOS, whether lean or obese, have a hyperinsulinemic response to a glucose challenge (40,41). Both lean and obese women with PCOS may be insulin resistant (42). Recent studies demonstrate that 7–10% of women with PCOS have type 2 diabetes and that an additional 30% have impaired glucose tolerance, far above the rates expected for their age (43,44). In addition, there is a higher prevalence of PCOS in women who have a history of gestational diabetes mellitus (45,46). Genetic studies currently favor an association between PCOS and a particular polymorphism of the VNTR region of the insulin gene (47). However, not all women with PCOS are also insulin resistant.

The positive correlation between obesity and insulin resistance is well established. A rising body mass index correlates with worsening symptoms of PCOS (e.g., menstrual cycle disturbance and hyperandrogenism) (13). Although insulin resistance in women with PCOS is partly independent of obesity (42), it is significantly related to body weight, and insulin resistance in obese women with PCOS is improved with weight reduction (48). Pharmacologic suppression of insulin with diazoxide improves biochemical parameters in women with PCOS, suggesting a pathophysiologic role for insulin in the etiology of PCOS (49,50).

PCO can be recognized by transabdominal US in prepubertal girls (51,52), but expression of the syndrome does not occur until after puberty. Puberty is a hyperdynamic state involving major changes in the secretion of growth hormone, insulin, and IGF-1, gonadotrophin-releasing hormone pulsatility and sex steroids. It has been suggested that the evolution of PCOS in adolescence is a state of “hyperpuberty”(53,54). Thus, there is concern that metabolic derangements or environmental stimuli may exacerbate the development of PCOS in adolescent girls.

Thus, PCOS may be the consequence of multiple abnormalities, including both genetic and environmental factors. Metabolic changes including hyperinsulinism are particularly implicated in the pathogenesis of PCOS and may be affected by environmental factors.


Although it was accepted previously that women with epilepsy have lowered fertility rates (55), this may not be true, as shown recently in community-based studies from Iceland (56) and from the Rochester, Minnesota study of medical records (57); in women with epilepsy, both marriage and fertility rates are close to or only slightly below the average population values. In another study, however, the fertility rate was lower in women with epilepsy compared with that in their siblings without epilepsy (58,59).

Women with epilepsy may, however, be exposed to hormonal imbalance and predisposed to development of PCOS. Prolactin levels may be transiently increased in the wake of epileptic seizures (60,61), a finding that was amply confirmed by later studies, both after complex partial seizures of temporal lobe origin (62) and after generalized tonic–clonic seizures (63). However, symptomatic hyperprolactinemia seems to be a most uncommon finding in women with epilepsy. Other acute hormonal changes provoked by seizures include elevated LH and FSH levels (64). There is no evidence at present that such acute changes produce any permanent dysfunction of the hypothalamic–pituitary–gonadal axis (65).

There also are possible changes in hormone levels due to interictal discharges (66,67), which could cause permanent abnormal menstrual patterns in some women with epilepsy, as shown by several studies. It is worth bearing in mind, however, that all of these studies have been of relatively small numbers of women, usually with more severe seizure disorders, and there is no consensus about the incidence of menstrual irregularity among women who do not have epilepsy. No community-based studies of the menstrual patterns of women with epilepsy are available. A history of abnormal menstrual patterns was found in >50% of women undergoing temporal lobectomies for intractable epilepsy (68), a finding confirmed by Herzog et al. (69) who showed that 25% of women with temporal lobe epilepsy (TLE) had oligomenorrhea and hirsutism and were thus suspected of having PCOS.

These authors subsequently completed their work and showed that 28 of 50 consecutive patients with TLE had menstrual dysfunction. Nineteen of these had endocrine anomalies: 10 had a hormonal profile suggestive of PCOS, two had premature menopause, one had hyperprolactinemia, and six had hypogonadotrophic hypogonadism. The same group reported on 30 women with TLE and reproductive endocrine disorder (70) and showed significant association between PCO and left temporal lobe lateralization: of 16 patients with TLE and PCO morphology, 15 had a left epileptic focus. The interictal pulse frequency and amplitude of LH secretion were examined in 14 women with TLE treated with AEDs and eight age-matched healthy controls (71): epilepsy patients had more variable and lower LH pulse frequencies than controls, and there was a trend toward higher pulse frequencies in women with left temporal foci compared with those with right-sided foci. Three women were diagnosed as having PCOS, and all had a left-sided focus. The clinical relevance of these data is not established, and the definitions of PCO and PCOS in these studies remain controversial.

Menstrual disturbances have been reported in other forms of epilepsy besides TLE. In their study of 20 women with idiopathic generalized epilepsy (IGE), Bilo et al. (72) showed a 25% incidence of reproductive endocrine disorders (five patients overall; three with PCOS and two with hypothalamic ovarian failure). All but three of their 20 patients were receiving AED therapy. Of the five women with a reproductive endocrine disorder, one was untreated, and three had developed menstrual problems before initiation of AED therapy. Endocrine dysfunction was studied in 10 women with untreated epilepsy (partial or generalized), normal body mass index, and normal menstrual cycles, and compared with data from five matched controls (73,74). Ovulation was documented in all women. The only difference in reproductive hormone profiles was a significantly higher LH pulse frequency (but not amplitude) in the women with epilepsy. A group of 101 women with epilepsy (36 with IGE, 65 with partial epilepsy) treated with various AEDs was studied by Murialdo et al. (75); 83 of these patients had pelvic US. PCO were found in 17% of the patients (21% of patients with IGE, 14% of patients with partial epilepsy), whereas 12 other patients had multifollicular ovaries without any clinical correlate. For these authors, the incidence of PCO or multifollicular ovaries was not higher than that in the general population.

Thus, there are considerable data pertaining to changes in the hormonal milieu of women with epilepsy, and a suggestion that these individuals are more likely to have derangements of the menstrual cycle, PCO, or PCOS. This has been hypothesized for patients with TLE in particular, because epileptic discharges from the limbic system may propagate to the hypothalamic–pituitary axis (76,77), but the evidence is less convincing in IGE. However, as yet the significance of these findings is uncertain.


The basic sex hormone changes produced by AED treatment can be summarized as follows (78): hepatic enzyme–inducing drugs [phenytoin (PHT), phenobarbital (PB), primidone (PRM), carbamazepine (CBZ)] elevate SHBG, whereas other drugs do not. During prolonged AED treatment using enzyme-inducing drugs, free hormone levels may be lowered (especially T), whereas LH and FSH levels may rise. However, Duncan et al. (79) reported no significant change of free T or gonadotrophins with AEDs in a large series of patients with epilepsy. Detailed changes in serum reproductive hormones linked to specific AEDs are reported in Table 5. Contrary to the enzyme-inducing AEDs, VPA has been shown to have little effect on reproductive hormone levels. More specifically, it was shown that VPA did not exert any long-term influence on the hypothalamic–pituitary axis in children and adolescents (89). There is only a single case report of pubertal arrest under VPA (90). In clinical practice, VPA is often indicated in juvenileonset epilepsies, especially in juvenile myoclonic epilepsy, which starts around puberty, where its efficacy is unchallenged (91). This coincides with a period of life when early reproductive disorders such as PCOS or eating disorders (e.g., anorexia, bulimia) and weight gain also become manifest (92,93).

Table 5.  Changes in reproductive hormones observed in patients taking antiepileptic treatment
  1. DHEAS, dehydroepiandrosterone sulfate; LH, luteinizing hormone; FSH, follicle-stimulating hormone; Prl, prolactin; SHBG, sex hormone–binding globulin.

CarbamazepineDHEASDecreasedIsojärvi et al. 1990 (in males) (80)
Geisler et al. 1997 (81)
Stoffel-Wagner et al. 1998 (82)
 LH, FSH, PrlIncreasedFranceschi et al. 1984 (83)
Bauer 1996 (61)
 Follicular-phase LHDecreasedStoffel-Wagner et al. 1998 (82)
 Luteal-phase E2DecreasedStoffel-Wagner et al. 1998 (82)
 SHBGIncreasedBeastall et al. 1985 (84)
PhenytoinDHEASDecreasedIsojärvi et al. 1990 (in males) (80)
 LH, FSH, PrlIncreasedFranceschi et al. 1984 (83)
Elwes et al. 1985 (85)
 SHBGIncreasedBeastall et al. 1985 (84)
 E2Increased in malesHerzog et al. 1991 (86)
PhenobarbitalSHBGIncreasedBeastall et al. 1985 (84)
ValproateDHEASNo changeIsojärvi et al. 1990 (in males) (80)
 LH, FSH, PrlNo changeFranceschi et al. 1984 (83)
 PrlNo changeInvitti et al. 1988 (87)
 FSHSlight increase (in males)Geisler et al. 1997 (81)
 LH, mid-cycle pulsesNo changeLado Abeal et al. 1994 (88)
 SHBGNo changeRamsay and Slater 1991 (78)

The impact of AEDs on reproductive endocrine disorders is debated. In their study of 50 women with TLE, Herzog et al. (1) included 20 untreated and 30 treated patients; the incidence of menstrual disorders was 60% in untreated women and 53% in treated women, and PCO or hyperandrogenism was diagnosed in 30% of untreated women and in 13% of treated women. They commented that it is unlikely that the endocrine disorders observed in the study were attributable to the effects of AEDs. In their study of 65 women with epilepsy, including 21 women treated with VPA, 21 with PB, and 23 with CBZ, and 20 healthy controls, Murialdo et al. (94) showed that there was no difference in hirsutism score, in ovary volume, or in the prevalence of PCO, although the VPA group had higher body weight and body mass index. In a previous study, the same group stressed that the incidence of PCO in women treated for epilepsy was not higher than that in the general population, but noted that VPA was more often associated with ovulatory dysfunction and PCO than CBZ or PB; however, patients with IGE were preferably treated with VPA and overall had the highest incidence of PCO (75).

The selective responsibility of specific AEDs in the pathogenesis of PCOS was not discussed in detail until the recent reports by Isojärvi et al. (2,95,96). These authors presented prevalence data on the occurrence of PCO and hyperandrogenism in women treated for epilepsy, proposed a physiopathologic explanation of the VPA-induced changes (97,98), and finally also proposed a therapeutic change to reverse VPA-induced PCOS (3).

In their first retrospective study (2), 238 patients with epilepsy of mean age 33 years (range, 18–45 years) and with a mean duration of treatment of 9 years (range, 0–31 years), and 51 healthy controls with a mean age of 35 years (range, 22–45 years) were included (2). Twenty-nine (12%) were being treated with VPA, 120 (50%) with CBZ, 12 (5%) with a combination of these two drugs, and 62 (26%) with other AEDs; 15 (6%) were untreated. Menstrual disturbances were present in 45% of the VPA group, 19% of the CBZ group, 25% of the combination group, 13% of the group receiving other drugs, 0 of the untreated group, and 16% of controls. Vaginal US examination was performed in all patients with menstrual disturbances and in some of the others, as well as in the controls. US revealed PCO in 43, 22, 50, and 11%, respectively, in the four treatment groups and 5% in regularly menstruating controls; elevated T occurred at an incidence of 17%, 0, 38%, and 0, respectively, in the four treatment groups. Eighty percent of women treated with VPA before the age of 20 years had PCO or elevated T; this compares with 27% of women treated with other AEDs. Menstrual disturbances were experienced by 20% of the epilepsy population and 16% of healthy controls. Women with menstrual disturbances were more likely to have PCO than those with regular cycles.

This study raised a significant point, but had several weaknesses. The major criticism that can be raised is its retrospective nature, which may have entailed selection bias with overrepresentation of women with VPA-related dysfunction. A significant selection among patients treated with VPA is most probable, as the prevalence of VPA monotherapy (12%) is far below the expected prevalence of patients with IGE, who are likely to be aggravated by CBZ. Moreover, pretreatment data were not available, the incidence of fertility problems was not stated, PCO morphology was used as a surrogate for PCOS, and the prevalence of PCO and menstrual disorders in healthy controls and untreated patients was very low. No explanation for the increased incidence of PCO in women treated with VPA was given. The higher incidence of PCO in women treated with VPA was confirmed in other centers: 67% in Norway versus 59% in Finland (95), and 72% PCO and/or hyperandrogenism in VPA-treated women (84% in obese patients vs. 62% in lean ones) among patients seen in Holland, Norway, and Finland (96). The latter study stressed the important part played by hyperinsulinemia. Yet none of these studies was prospective, and recruitment bias (i.e., selection of VPA-treated patients with major weight gain and/or menstrual problems) cannot be excluded, as the original group of VPA-treated women from Finland was included in all subsequent studies.

In a follow-up communication (97), the same group compared the levels of insulin and IGFBP-1 in women with epilepsy treated with VPA (n = 18) or CBZ (n = 43) and in healthy controls (n = 43). Fasting insulin levels were higher in the VPA group (28.1 mU/L) than in the CBZ group (15.4 mU/L) or controls (9.6 mU/L). Fasting insulin levels were high in the VPA group irrespective of their PCO and hyperandrogenism status. However, the women with PCO or hyperandrogenism had lower IGFBP-1 levels than the controls (1.4 vs. 4.6 μg/L). These data pointed to the existence of hyperinsulinism and a deficiency in IGFBP-1 in women treated with VPA, both factors concurring to provoke increased androgen secretion by the ovaries and PCOS. In their next article (98), Isojärvi and colleagues confirmed these data, presumably in the same patients, but added new information that showed that weight gain was the major factor leading to hyperinsulinemia, decreased IGFBP-1 levels and PCO; weight gain occurred in 50% of the patients studied.

This study introduced new and very interesting insights into the pathogenesis of PCOS in women on VPA, but was again biased by its retrospective nature. The conclusions are drawn from the study of a small group of 22 patients receiving VPA monotherapy, among whom 59% were obese and 50% had gained weight with VPA, and no evidence is given showing that these findings could be applied to the VPA-treated population at large. In his comment on this article, Herzog (99) noted that the article produced evidence that weight gain is associated with PCOS in women treated with VPA, but suggested that PCOS may be induced by epilepsy; unlike VPA, enzyme-inducing drugs may prevent, in part, the occurrence of PCOS by reducing biologically active T (99).

In a more recent article (3), Isojärvi et al. studied the effect of substituting LTG in a group of 16 VPA-treated patients who had PCO. Twelve women completed the study; two with juvenile myoclonic epilepsy were returned to VPA, one had a rash, and one became pregnant. During the first year of LTG therapy, the patients lost weight and experienced significant decreases in body mass index (from mean 30.9 to 28.3 kg/m2), waist circumference (98.2 to 90.1 cm), hip circumference (112.0 to 106.7 cm), and waist/hip ratio (0.87 to 0.84). The mean serum T level also decreased. The mean total number of ovarian follicles that were visualized in the epilepsy patients decreased from 20 on VPA therapy to 11 at 1 year after the switch to LTG. The incidence of menstrual disturbances decreased from 44 to 13%. Fasting insulin levels decreased significantly, and both high-density liposome cholesterol and high-density liposome/total cholesterol ratios increased significantly. The authors thus showed that substituting LTG for VPA in VPA-treated women with obesity, PCOS, and hyperinsulinism brings about a reduction in body weight and corrects hyperinsulinism and its consequences. However, the respective parts played by discontinuing VPA, weight reduction, and the effect of LTG were not studied.

Two very recent articles from the same group further studied the relation between antiepileptic drug treatment and gonadal function in young women. The first article (100) studied the effect of VPA, CBZ, and oxcarbazepine (OCBZ) in girls aged 8–18 years: the treatment groups included 40 girls taking VPA, 17 with CBZ, and 18 with OCBZ, with 49 healthy, untreated controls. There was no adverse effect with any of the drugs on growth or sexual maturation, and no statistically significant difference between the treatment groups and controls in weight or metabolic markers (insulin, IGFBP-1 and -3), except for an elevated IGF-1 in patients on CBZ and OCBZ compared with controls. The second article (101), which dealt with presumably the same population (41 girls aged 8–18 years taking VPA, compared with 54 healthy, untreated controls), found evidence of biological hyperandrogenism in some VPA-treated girls: elevated T levels were found in 38% of prepubertal, 36% of pubertal, and 57% of postpubertal girls in the treatment group (there were no data on the prevalence of such changes in the control group). None of the other endocrine parameters (including dehydroepiandrosterone sulfate, SHBG, estradiol, LH and FSH, ovarian volume, menstrual irregularities) was significantly affected. Comparing the VPA-treated girls with and without hyperandrogenism and the controls according to pubertal stage, there was no difference in body mass index, resting insulin, IGF-1, and IGFBP-1 and -3, except for a significant elevation of IGF-1 in pubertal girls with hyperandrogenism, and a lowering of IGFBP-3 in postpubertal girls with hyperandogenism. Despite the scarcity of abnormal findings in VPA-treated girls, the authors advised again against the use of VPA in girls with epilepsy. Contrary to their previous studies, these two recent studies apparently included all the willing patients in the selected age and treatment classes. It must be stressed that their findings were less spectacular than those previously reported, which confirms the mostly retrospective nature of their previous findings in older women.

Taken together, the recent studies by the Finnish group have made an important point: VPA may induce weight gain, hyperinsulinism, and insulin-resistance related symptoms, but these changes appear to be reversible. There are no data yet on the prevalence of such changes in an unselected population of women treated with VPA for epilepsy, or on the dose relation between VPA treatment and the endocrine and metabolic effects. These studies are interesting but as yet provide no compelling evidence that the use of VPA in adolescent girls or young women will have deleterious effects on their future reproductive life.


The prevalence of menstrual disturbances in the general population and in unselected epilepsy cases is not known. PCO and PCOS may or may not be more frequent in women with epilepsy than in the general population. When such changes are found in women with epilepsy, it seems reasonable to assume that hyperinsulinism and related disorders have played a significant part, and obesity is probably an additional risk factor. Women with PCO might then be precipitated into expressing PCOS. The prevalence of VPA-induced changes is not known and is low according to all evidence besides the studies of Isojärvi and his co-workers. The relative contribution of the epilepsy syndrome itself has not been defined.

There are no data to contraindicate use of VPA in women, and VPA remains a first-line option for the treatment of epilepsy in young women. In practice, literature data stress the importance of preventing weight gain in young women with epilepsy. However, the prevalence and mechanisms of weight gain associated with VPA are not well understood (93,102) and should be studied further.

Thus, we recommend that before initiating VPA treatment, baseline body weight should be measured, and the patient should be warned about the potential risk of weight gain and its consequences for her general well-being. The patient's weight should be monitored at subsequent consultations. If significant weight gain is observed, an evaluation of the cause of weight gain is indicated, measures should be proposed to reduce weight, and in some cases, the therapeutic options for her epilepsy should be reviewed.


In response to editorial inquiry, Dr. Genton provided the following information on behalf of the authors:  None of the authors has a particular or permanent relationship with Sanofi or Abbott Laboratories, and all work as independent clinicians. The group gathered in Paris in February 1999, invited by Sanofi, to produce a statement on VPA and PCO. Individuals were chosen by myself and by Sanofi staff from France and the United Kingdom according to their specific fields of competence and to their availability. They were asked to produce a balanced view on this subject. The resulting paper was circulated among all authors for approval before being forwarded to this journal. It reflects the personal opinions of all co-authors and not those necessarily of Sanofi or Abbott Laboratories (documents for internal circulation have been produced earlier by these companies).

Pierre Genton October, 1999

Acknowledgment: This publication was supported by an unrestricted educational grant from Sanofi.