Address correspondence and reprint requests to Dr. M.J. Brodie at Epilepsy Unit, Department of Medicine and Therapeutics, Western Infirmary, Glasgow G11 6NT, Scotland. E-mail: Martin.J.Brodie@clinmed.gla.ac.uk
Summary: Purpose: Treatment with sodium valproate (VPA) may be associated with polycystic ovarian syndrome (PCOS) in some women with epilepsy. By comparing hormone profiles in young adults taking VPA or lamotrigine (LTG) as monotherapy, this study aimed to explore whether a pharmacologic effect of VPA could be responsible for this observation.
Methods: Hormone profiles in men and women taking VPA (n = 40) or LTG (n = 36) monotherapy for epilepsy were compared. None of the women were receiving hormonal contraception or replacement. Patients gave details of seizure type and frequency, menstrual cycle, and medical and drug history. Body mass index was calculated, and fasting insulin, glucose, cholesterol, triglycerides (TG), high- and low-density lipoproteins, testosterone, dihydroepiandosterone (DHEA), androstenedione, sex hormone–binding globulin (SHBG), free androgen index (FAI), luteinising hormone (LH), follicle-stimulating hormone (FSH), and antiepileptic drug (AED) concentrations were measured.
Results: There were no differences between treatment groups for both sexes in age and seizure control. Only four obese VPA-treated women were hyperinsulinaemic (p = 0.05); three with abnormal menstrual cycles; one with raised testosterone. Testosterone (p = 0.02), FAI (p = 0.03), and TG (p = 0.02) levels were higher, however, in women taking the drug. Obese patients of both sexes (p = 0.01) and VPA-treated men (p = 0.03) had higher insulin concentrations.
Conclusions: VPA therapy may be associated with subclinical elevation in fasting insulin levels. Testosterone and TG levels were higher in VPA-treated women compared with the levels in those taking LTG. However, only a minority of obese females exhibited biochemical characteristics suggestive of PCOS. Biochemical screening may allow women at risk of developing PCOS to avoid VPA.
Sodium valproate (VPA) is a broad-spectrum antiepileptic drug (AED) that has been in use as a first-line agent for >30 years (1). VPA is a branched-chain fatty acid that undergoes similar metabolism to endogenous fatty acids (2,3). Weight gain is a recognised side effect and may be a consequence of mitochondrial inhibition by oxidative metabolites (4–8). Treatment with VPA has been associated with polycystic ovarian syndrome (PCOS) (9–12). This is a heterogeneous condition for which the diagnostic criteria remain a subject of contention (13). It tends to be characterised by ovarian cysts, chronic anovulation, hyperandrogenism, hyperinsulinaemia, and dyslipidaemia (6,10–14). Affected women can also have an elevated serum luteinising hormone (LH)/follicle-stimulating hormone (FSH) ratio, but this is a less sensitive marker (12,15). Although these abnormalities are often linked with obesity, this is not always so.
Recent results from studies in Finnish women suggested that ∼60% of patients taking VPA had biochemical and clinical changes associated with PCOS (9,10). As this was not our anecdotal experience, we decided to undertake this preliminary cross-sectional study. The relationship between VPA administration and PCOS is unclear, but it is possible that a pharmacologic effect of VPA is responsible. If this is the case, metabolic abnormalities also may be present in men taking the drug. For this reason, androgen, insulin, and lipid profiles were compared in young women and men treated with VPA monotherapy. As lamotrigine (LTG) has been shown to normalise some of the metabolic abnormalities accompanying VPA-associated PCOS (12), patients receiving VPA monotherapy were matched with controls taking LTG.
Premenopausal women and men younger than 50 years were recruited from the Epilepsy Clinic at the Western Infirmary in Glasgow, Scotland. All had been treated only with VPA or LTG monotherapy for ≥2 years for partial seizures with or without secondary generalisation or idiopathic generalised epilepsy. Pregnant, postmenopausal, oophorectomised, and diabetic patients and those taking hormonal contraception or hormone replacement and glucocorticoids were excluded from the study. Details of the protocol were approved by the West Ethics Committee, and patients gave written informed consent to their participation.
Patients completed a medical questionnaire, including details of seizure frequency on unchanged AED dosage for the past 6 months. Women also were asked to keep a record of their menstrual cycles during this time, with abnormal cycles regarded as being >35 days, varying from <35 to >35 days and lasting 22–35 days, but varying >4 days (16). Height and weight were measured, and body mass index (BMI) calculated. Patients with a BMI >25 kg/m2 were categorised as being overweight, and those with a BMI >30 kg/m2, as obese (17).
A fasting blood sample was withdrawn at the same hospital visit for plasma AED concentrations, insulin, glucose, high-density and low-density lipoproteins (LDL) and serum cholesterol (CHOL), triglycerides (TG), testosterone, dihydroepiandosterone (DHEA), androstenedione, sex hormone–binding globulin (SHBG), free androgen index (FAI), LH, and FSH. Samples were stored at –80°C for batch analysis. Assays were undertaken using standard commercially available methods.
Comparisons between groups for all parameters were made by analysis of variance (ANOVA). The χ2 test and Fisher's exact test were used for analysis of categoric data. A value of p < 0.05 (two-tailed) was considered significant, and 95% confidence intervals (CIs) for mean differences were calculated where relevant. Calculations were made with the use of a Statgraphics statistical software package (STSC Software Publishing Group, Rockville, MD, U.S.A.).
A total of 76 patients (median age, 38 years; range, 15–50 years) with epilepsy (40 partial-onset, 36 idiopathic generalised) were compared. Of these, 40 (17 men, 23 women) were taking VPA, and 36 (15 men, 21 women) LTG only for a minimum of 2 years. Ages ranged from 17 to 47 years (median, 33 years) in women taking VPA, from 17 to 50 years (median, 33 years) in women taking LTG, from 18 to 40 years (median, 29 years) in men taking VPA, and from 15 to 41 years (median, 28 years) in men taking LTG. Monthly seizure counts varied from 0 to 30 (median, <1). No significant differences in seizure control were demonstrated between treatment groups. For women taking VPA, daily doses ranged from 400 to 2,500 mg (mean, 1,467 mg), with plasma concentrations varying from 18.8 to 134 mg/L (mean, 83.5mg/L). In men taking VPA, daily doses ranged from 800 to 4,500 mg (mean, 1,929 mg), with plasma concentrations varying from 32 to 120 mg/L (mean, 72.4mg/L). For women taking LTG, daily doses ranged from 50 to 600 mg (mean, 276 mg), with plasma concentrations of 1.04–18.6 mg/L (mean, 5.3 mg/L). In men taking LTG, daily doses varied from 100 to 800 mg (mean, 313 mg), and plasma concentrations from 2.13 to 20.8mg/L (mean, 6.1 mg/L).
Hormone, lipid and BMI results are illustrated in Table 1. Hyperinsulinaemia was present in four (17%) women taking VPA (p = 0.05), all of whom had a BMI >30 kg/m2(Fig. 1). Three of these had abnormal menstrual cycles; one had an increased serum testosterone. Overall, menstrual irregularities were present in five (22%) of the 23 women taking VPA and in seven (33%) of the 21 taking LTG. There was no statistical difference in insulin concentration between the two female groups, but VPA-treated women had higher levels of testosterone (p = 0.02; mean difference, 0.79; 95% CI, 0.14–1.44) and elevated FAI (p = 0.03; mean difference, 1.91; 95% CI, 0.22–3.6) compared with their counterparts taking LTG (Table 1). Only two patients, however, had a testosterone level above the normal range (Table 2). TG were also significantly increased in this population (p = 0.02; mean difference, 0.44; 95% CI, 0.07–0.81). No differences were found between the two groups for CHOL, HDL, LDL, DHEA, SHBG, androstenedione, LH/FSH ratio, and glucose (Table 1).
Table 1. Median (range) fasting lipids, hormones, insulin, glucose, and body mass index in patients taking sodium valproate or lamotrigine
Table 2. Characteristics of female patients with elevated plasma insulin and/or testosterone
Abnormal menstrual cycle
Body mass index (kg/m2)
Men taking VPA had higher insulin concentrations (p = 0.03; mean difference, 4.16; 95% CI, 0.3–8.02) than those taking LTG. None had hyperinsulinaemia. There were no significant differences between the VPA- and LTG-treated men for any other parameter (Table 1).
VPA patients were more likely to be overweight (60% VPA vs. 42% LTG), with women (73% VPA vs. 48% LTG) being affected more often than men (41% VPA vs. 33% LTG). In patients of both sexes with a BMI >25 kg/m2, insulin concentrations were significantly higher for those taking VPA compared with those taking LTG (p = 0.01; mean difference, 6.48; 95% CI, 1.38–11.58). Insulin levels were not statistically different in patients with a BMI ≤25 kg/m2.
Hyperinsulinaemia only occurred in four obese VPA-treated women, three of whom had abnormal menstrual cycles, and one of whom had an elevated serum testosterone. Compared with those taking LTG, women taking VPA had significantly higher testosterone levels. This was reflected in their elevated FAI, which is a particularly sensitive marker of high testosterone (13). Serum TG concentrations also were higher for these patients. Male patients taking VPA had significantly higher plasma insulin levels, although none had hyperinsulinaemia. VPA patients were more likely to be overweight than were those taking LTG, with women being affected more often than men. Overweight patients were statistically more likely to have higher insulin levels regardless of AED. There were no differences between the two treatment groups for age, sex, seizure type, and frequency.
Previous studies have reported a strong link between VPA administration and PCOS in women with epilepsy (9–12). Diagnosis of the condition is not always straightforward, as affected individuals can have substantial heterogeneity of signs and symptoms (18). The spectrum of changes includes polycystic ovaries, chronic anovulation, hyperinsulinaemia, hyperandrogenism, and dyslipidaemia (13,15,18). To further complicate matters, 23% of normal women have polycystic ovaries on ultrasound with none of the metabolic changes of PCOS (19). Hormone and androgen assays are, however, recognised to be reliable screening tools for the syndrome (20) and were, therefore, used in this study.
Women taking hormonal contraception or replacement were excluded from participation because the combined oral contraceptive pill increases SHBG levels (13), reduces androgen levels, and may increase insulin resistance (21). To further ensure the clinical relevance of the results, none of our patients were pregnant, oophorectomised, diabetic, or taking glucocorticoids. Because women tend to be seen in their fertile years with PCOS, the study was confined to premenopausal females (13), although there is evidence to suggest that the syndrome persists into old age (22). Men older than 50 were excluded because testosterone levels are known to decline with age (23).
Only four of our VPA-treated female patients exhibited criteria consistent with PCOS. Although other women taking VPA and LTG had menstrual irregularities with no other metabolic abnormalities, this criterion alone is not diagnostic of PCOS (13,24). A trend toward PCOS is suggested, however, by elevated testosterone, insulin, and TG levels, and FAI in the VPA cohort. The mechanisms responsible for these abnormalities are not fully understood (18). High testosterone concentrations are indicative of ovarian as opposed to adrenal hyperandrogenism, the latter resulting in overproduction of DHEA and androstenedione (25–27). An elevated FAI is a sensitive marker of heightened testosterone activity (13). Hyperinsulinaemia is thought to be secondary to insulin resistance (28), which may result from defects in insulin clearance and peripheral tissue degradation (29,30). There is evidence to suggest that insulinaemic stimulation leads to ovarian androgen secretion and cyst formation (31,32).
VPA has been associated with increased insulin levels (12), the reasons for which have yet to be determined. The drug also is thought to cause weight gain in susceptible patients (4–7). The relationship between obesity and hyperinsulinaemia is well recognised (33), although the two factors are not always associated (11). Gidal et al. (6) proposed that hyperinsulinaemia is a consequence of obesity and not a direct effect of VPA. Our data do not support this theory, as overweight patients taking VPA were significantly more likely to have higher insulin levels than were overweight LTG-treated patients. Moreover, male patients taking VPA were not statistically heavier, but despite this, had higher fasting insulins than did those taking LTG. Although this phenomenon was not so apparent in women, the mean insulin level was higher for those taking VPA than for the LTG-treated patients. The significance of this subclinical increase is unclear. It is perhaps a consequence of insulin resistance or overproduction. Four obese VPA-treated women did, however, have hyperinsulinaemia. These patients may somehow be more susceptible to the deleterious effects of VPA exposure, resulting in an exaggerated elevation of insulin levels, setting in motion the metabolic cascade of events that produces PCOS.
PCOS has a prevalence of 5–10% in the female population (13). Although 17% of our VPA-treated women were likely to be affected, this is a far smaller percentage than the 64% identified by Isojarvi et al. (9) in Finland. Our data are supported Bauer et al. (34), who found that only 11% of VPA-treated German women had evidence of PCOS. This variation may reflect genetic and/or phenotypic differences between Scottish, German, and Finnish women.
Apart from insulin, other parameters were normal in men, in keeping with previous results (35,36). Although there have been reports of different seizure types being associated with ovulatory and endocrine abnormalities (34,37–41), this was not borne out in our patient cohort.
In conclusion, VPA therapy in patients with epilepsy may be associated with a subclinical increase in plasma insulin, possibly secondary to insulin resistance. This could be the result of a pharmacologic effect of VPA. In men, this seems to be of little consequence. In women, testosterone and TG levels also were elevated, suggesting that a metabolic cascade may be triggered by insulinaemic ovarian stimulation. The problem appears to affect far fewer Scottish than Finnish women, however, with only a minority of obese women in this study exhibiting frank hyperinsulinaemia, hyperandrogenism, menstrual disturbance, and dyslipidaemia characteristic of PCOS. At-risk individuals with a BMI ≥30 kg/m2 and/or with a history of menstrual irregularities may benefit from screening of fasting insulin, testosterone, FAI, and TG levels before starting AED treatment. Those with abnormal results should be advised about weight reduction and should not be prescribed VPA. Women with biochemical evidence of PCOS who are already taking VPA may benefit from switching to an alternative AED.
Observations from this preliminary report have generated a verifiable hypothesis. This is currently being tested in a randomised comparison of VPA and LTG in newly diagnosed epilepsy with the aim of following up prospectively the hormone profiles in patients being treated with both drugs.