Endocrine Effects of Valproate in Adolescent Girls with Epilepsy


Address correspondence and reprint requests to Dr. H. Goldberg-Stern at Epilepsy Center, Schneider Children's Medical Center of Israel, 14 Kaplan Street, Petah Tiqwa, 49202 Israel. E-mail: racheln@clalit.org.il


Summary: Purpose: To investigate the effect of epilepsy and/or valproate (VPA) monotherapy on physical growth, weight gain, pubertal development, and hormonal status in adolescent girls with epilepsy.

Methods: The study group included 88 consecutive female patients with epilepsy aged 6–20 years (28 premenarche, 60 postmenarche) attending an endocrinology institute of a major tertiary center. Forty-five patients were under treatment with VPA, and 43 were before treatment initiation. The groups were compared for the relevant biochemical, anthropometric, ultrasonographic, and endocrine parameters.

Results: No statistically significant differences were found in any of the parameters studied between the groups, as a whole or by menarche status. The treated postmenarcheal subgroup had a higher mean testosterone level than the untreated postmenarcheal controls (1.83 ± 0.65 vs. 0.88 ± 0.24, p = 0.006). Body mass index–standard deviation score (BMI-SDS) was 0.75 in the treated group and 0.63 in the untreated group; rates of obesity were 16.3% and 15.5%, respectively. No between-group differences were found in menses irregularities, hirsutism, or acne. No correlation was found between duration or dosage of treatment and BMI-SDS, height–SDS, or androgen level. The treated group had higher levels of thyroid-stimulating hormone and lower levels of free thyroxine than did the untreated group, although still within normal range.

Conclusions: Long-term treatment with VPA in girls with epilepsy is associated with increased testosterone levels after menarche, without clinical hyperandrogenism, polycystic ovary syndrome, or an increase in BMI-SDS. VPA is a good treatment option in this age group but should be accompanied by careful endocrine observation.

Epilepsy is a major worldwide public health problem, with an estimated prevalence of 0.5–1% of the population in industrialized countries. Epilepsy has been associated with an increased frequency of reproductive endocrine disorders including polycystic ovary syndrome (PCOS), menstrual abnormalities, and low fertility rate. In childhood and adolescence, it may be associated with abnormalities in growth and puberty (Cook et al., 1992; Snyder and Badura, 1998; Guo et al., 2001; El-Khayat et al., 2004). The pathogenic mechanisms underlying the association between epilepsy and these endocrine disorders have not been fully elucidated. Some researchers claim that epilepsy itself plays a pathogenic role (Herzog et al., 1984, 1986; Nappi et al., 1994; Cummings et al., 1995; Herzog, 1996; Bilo et al., 1998, 2001), whereas others (Isojärvi et al. 1993, 1996) propose that the endocrine disorders may be at least partly attributable to the use of antiepileptic drugs (AEDs), particularly sodium valproate (VPA).

An association between PCOS and epilepsy was first suggested in 1984 by Herzog et al. (1984). Later studies demonstrated features of PCOS in 15% of women with primary generalized and localization-related epilepsy (Bilo et al., 1988; Murialdo et al., 1997). It is possible that the ictal and interictal discharges disrupt gonadotropin-releasing hormone (GnRH) pulsatility (Bilo et al., 1991; Meo et al., 1993), leading to an increase in the luteinizing hormone/follicle-stimulating hormone (LH/FSH) ratio and the development of immature follicles. These follicles are deficient in aromatase, which converts androgens to estrogens, thereby causing hyperandrogenism and PCOS.

VPA, a widely used AED, has broad-spectrum activity against both generalized and partial epilepsies. It is now the drug of choice in the treatment of idiopathic generalized epilepsies. VPA is discouraged as first-line therapy in females of reproductive age owing to the risk of teratogenicity. However, it is still the first choice in many cases in the pediatric age group, as determined by medical and economic considerations. Furthermore, it is an important treatment option in adolescent girls and young women because it does not have enzyme-inducing properties and therefore does not reduce the effectiveness of oral contraceptives.

Isojärvi and colleagues (1993,1996) suggested that VPA-induced weight gain can lead to insulin resistance and hyperinsulinemia and therefore PCOS and other endocrine abnormalities. However, the same group has more recently questioned the role of obesity and elevated fasting insulin levels (Vainionpää et al., 1999). These findings have not been replicated by other studies (Murialdo et al., 1998; Luef et al., 2002; Bauer et al., 2004). This issue is important, because the possible risk of increased obesity and PCOS for VPA could outweigh its benefits in adolescent girls and young women with epilepsy and justify the selection of an alternative drug.

The aim of the present cross-sectional study was to investigate the effect of epilepsy and/or VPA on physical growth, weight gain, pubertal development, and hormonal status in adolescent girls with epilepsy.


The study sample included 88 patients with epilepsy: 43 girls who had received VPA monotherapy for ≥1 year (“treated group”), and 45 age-matched girls who had been diagnosed and evaluated but had not started AED treatment ("untreated group"). Six of the untreated patients had been given VPA or tegretol in the past, but treatment was discontinued ≥2 years before the present evaluation; the rest had never received AEDs. All patients were recruited consecutively from outpatients attending our epilepsy center. Inclusion criteria were as follows: age 6–20 years; absence of systemic or central nervous diseases (apart from epilepsy) that could interfere with hypothalamic–pituitary–ovarian function; and no treatment with hormonal and psychotropic medications, apart from AEDs (treated group). Patients given other AEDs or polytherapy were excluded.

Epilepsy was classified according to the International Classification of Epilepsies and Epileptic Syndromes (Commission on Classification and Terminology of the International League Against Epilepsy, 1989). A detailed medical and family history was obtained, including family history of epilepsy, type 2 diabetes, hirsutism, endocrine disorders, and PCOS. A detailed menstrual history was obtained from subjects with established menstrual cycles. Regular cycles were defined as those lasting from a minimum of 21 days to a maximum of 35 days, with no more than a 4-day variation in length from cycle to cycle. The duration of epilepsy was defined as the time elapsed from the first seizure.

Physical examination included measurements of height (with the Harpenden–Holtain stadiometer) and weight. Height was calculated as height-standard deviation score (Ht-SDS) for all girls and for both their parents, by using the Centers for Disease Control growth charts (Kuczmarski et al., 2000). Body weight was expressed as body mass index (weight in kilograms/height in meters squared), and the BMI–standard deviation score (SDS) was calculated according to the method of Rosner et al. (1998). Overweight was defined as a BMI >85th percentile and ≤95th percentile for age and gender, and obesity was defined as a BMI >95th percentile for age and gender. Waist and hip circumferences were measured (Lohman, 1986), and the waist-to-hip ratio was calculated. Skin folds were measured with a caliper (Pecoraro et al., 2003), and body fat mass was measured by using bioimpedance (Tanita, TBF-300GS, Arlington Heights, IL, U.S.A.) (Pecoraro et al., 2003). Girls were evaluated for the presence of acanthosis nigricans, hirsutism, and acne. Hirsutism was defined according to the criteria of Ferriman and Gallwey (1961), with values <7 considered normal. Acne was considered present if currently active and involving the face, neck, or upper trunk. Bone age was estimated according to Greulich and Pyle (1959). Pubertal stage was assessed according to the criteria of Marshall and Tanner (1969). On the basis of the findings, the treated and untreated patients were defined by pubertal stage as follows: prepubertal, no clinical signs of puberty (breast B1, no pubic hair); pubertal, Tanner stage B2–4; and postpubertal, Tanner 5. As the prepubertal groups were too small to draw conclusions (four VPA treated and 10 untreated), the treated and untreated patients were divided into two subgroups: before the first menstrual period and after the first period.

For laboratory workup, venous blood samples were drawn between 8 and 10 a.m. after an overnight fast. In menstruating girls, blood samples were obtained at the early follicular phase (days 2 to 6 after the menstrual period). Serum blood levels of follicle-stimulating hormone (FSH), luteinizing hormone (LH), prolactin (PRL), thyroid-stimulating hormone (TSH), insulin-like growth factor (IGF-1), sex hormone–binding globulin (SHBG), and insulin were measured with an immunometric assay; estradiol, with a radioimmunoassay (RIA); free thyroxine (FT4), with a competitive analogue immunoassay; testosterone (T) dehydroepiandrosterone sulfate (DHEA-S), with a competitive enzyme immunoassay; the Immulite 2000 analyzer (DPC, Los Angeles, CA, U.S.A.) was used in all cases. Serum androstenedione levels were determined by RIA (Diagnostic Systems Laboratories, Webster, TX, U.S.A.); 17-hydroxyprogesterone, by RIA (MP Biomedicals, Orangeburg, NY, U.S.A.); and leptin, by immunoradiometric assay (IRMA) (DSL). Assays were run as batched samples. The free androgen index (FAI), which is known to correlate well with free testosterone (Miller et al., 2004), was calculated (100xT/SHBG). Serum glucose was measured by the glucose oxidase colorimetric method by using an automated analyzer (Hitachi 917; Roche Diagnostics, Mannheim, Germany), and total cholesterol, triglycerides, and HDL-cholesterol concentrations were measured by an enzymatic colorimetric method on an automated analyzer (Hitachi 904; Roche Diagnostics). Insulin resistance was estimated by the homeostatic model assessment (HOMA-IR), and insulin sensitivity, by the ratio of fasting glucose (FG) to fasting insulin (FI) (FG/FI) and the quantitative insulin sensitivity check index (QUICKI). All three values were derived from the fasting measurements. HOMA-IR was calculated according to the following formula: HOMA-IR = (fasting insulin (μU/ml) × fasting glucose (mM)/22.5). HOMA-IR was compared both to the 95th percentile for age and sex (Allard et al., 2003) and to the mean values for pubertal and obesity status (Gungor et al., 2004). FG/FI was measured in milligrams/deciliter (mg/dl) for glucose and microunits/milliliter (μU/ml) for insulin. QUICKI was calculated according the formula: QUICKI = 1/[log (fasting insulin (μU/ml) + log(fasting glucose (mg/dl)]. As no cutoff values are known for QUICKI in children and adolescents, we compared our findings with mean values according to pubertal stage (Gungor et al., 2004) and with mean values minus 2 standard deviations.

All patients underwent transabdominal pelvic ultrasound with a conventional full-bladder, 5-MHz, real-time sector scanner (Sonoline Prima; Siemens, Issaquah, WA, U.S.A.). Menstruating patients were scanned in the early follicular phase. All scans were performed by the same investigator (L. de V.). The following ultrasonographic parameters were considered for analysis: (a) uterus: length, transverse diameter (width), and fundal anteroposterior diameter; and (b) ovaries: height, width, length, number of follicles, and maximal diameter of largest follicle observed. Uterine and ovarian volumes were calculated according to the formula for ellipsoid bodies: V = longitudinal diameter × anteroposterior diameter × transverse diameter × 0.5233. PCO morphology was established by the detection of ≥12 follicles measuring 2–9 mm in diameter and/or increased ovarian volume (>10 ml) (Rotterdam ESHRE/ASRM-Sponsored PCOS Consensus Workshop Group, 2004). We used the revised 2003 criteria of PCOS (Gungor et al., 2004), which require the presence of two of the following three parameters for diagnosis: (a) oligo- or anovulation; (b) clinical and/or biochemical signs of hyperandrogenism; and (c) polycystic ovaries.

The clinicians who performed the anthropometric measurements and pelvic ultrasound were blinded to the patients' status (treated or untreated).

Written informed consent was obtained from all families. The institutional Human Research Committee approved the study.

Statistical analysis

All statistical analyses were done with BMDP software (New System version, Statistical Solutions, Cork, Ireland). The results are given as mean ± SD. One-way analysis of variance (ANOVA) was used to compare anthropometric, laboratory, and ultrasonographic parameters between treated and untreated girls. Two-way ANOVA was used to compare variables between the subgroups according to menarche status, and between girls with generalized or partial epilepsy and treated or untreated groups. Those variables that did not have a gaussian distribution were transformed by using square-root transformation to achieve normal distribution. Multiple comparisons were made by using Bonferroni corrections. Spearman rank correlation was used to evaluate correlations between various parameters. A p value of ≤0.05 was considered significant.


Clinical characteristics

The clinical and anthropometric characteristics of the study sample are presented in Table 1. The VPA-treated group included 13 girls before menarche and 30 girls after menarche; the untreated group included 15 girls before menarche and 30 girls after menarche.

Table 1. Clinical characteristics of valproate-treated and untreated girls with epilepsy
 Valproate-treated (n = 43)Untreated (n = 45)p Value
  1. Data are expressed as mean ± SD.

  2. SDS, standard deviation score; BMI, body mass index; W/H, waist/hip ratio.

Age (yr)14.9 ± 3.313.3 ± 3.20.02
Age at onset (yr)10.3 ± 3.510.6 ± 4.00.72
Epilepsy duration (yr) 4.6 ± 2.8 2.9 ± 2.8<0.003  
Height SDS −0.18 ± 0.9  0.01 ± 0.90.53
BMI SDS   0.6 ± 1.8 0.7 ± 1.30.74
Triceps thickness (mm)19.4 ± 8.417.3 ± 8.50.28
Waist circumference (cm)71.7 ± 9.2 68.7 ± 11.90.25
W/H ratio  0.8 ± 0.06  0.8 ± 0.050.82
Body fat mass (%)26.8 ± 8.227.0 ± 9.80.95
Bone age SDS0.51 ± 1.10.69 ± 1.10.61

At admission, the untreated girls were slightly younger than the VPA-treated girls (mean age, 13.3 ± 3.2 vs. 14.9 ± 3.4 years). Epilepsy appeared at the same age in both groups (mean age, ∼10.5 years). Mean disease duration was 4.69 ± 2.85 years for the VPA-treated girls and 2.89 ± 2.81 for the untreated girls (p < 0.003). Twenty-two (49%) of the untreated group had epilepsy for <1 year. Mean treatment duration was 3.5 years (range, 1–9.5 years), and 77% of the patients treated for >2 years. Mean dose at evaluation was 12.20 ± 4.54 mg/kg/day (range, 3–25). The mean plasma level of VPA was 60.6 ± 22.4 μg/ml.

The classification of epilepsy showed that the VPA group included nine patients with partial epilepsy and 34 with generalized epilepsy, and the untreated group included 16 patients with partial epilepsy, 27 patients with generalized epilepsy, and two with undetermined type. The rates of epilepsy by type were similar in the two groups, excluding the two patients with undetermined type (p = 0.094).

BMI-SDS was similar in the two groups (mean, 0.63 ± 1.82 vs. 0.75 ± 1.34, respectively; p = 0.74). This also held true on comparison of the subgroups by menarche status: mean BMI-SDS was 0.51 ± 1.47 in the postmenarcheal VPA-treated girls and 0.88 ± 1.52 in the postmenarcheal untreated girls (p = 0.34). A BMI-SDS of ≥2.0 was found in six (13.9%) treated girls and six (13.3%) untreated girls, and overweight, in 11 (25.6%) VPA-treated girls and seven (15.5%) untreated girls (p = 0.3), and obesity, in seven (16.2%) VPA-treated girls and four (8.8%) untreated girls (p = 0.5). Overall prevalence of obesity and overweight was 33%: 41.8% in the VPA group and 24.4% in the untreated group (p = 0.1). The average age of the girls with either obesity or overweight was similar in the two groups (14.88 years and 14.38 years).

No difference was found between the groups in triceps thickness, waist or hip circumference, fat percentage, or bone-age SDS.

Ht-SDS was within normal range for both groups and was well correlated both with maternal and paternal height.

No correlation was found between dosage or duration of treatment and BMI-SDS, Ht-SDS, or androgen levels.

No significant differences were found between girls with partial epilepsy and girls with generalized epilepsy in any of the clinical, metabolic, or hormonal parameters.

Metabolic evaluation

Fasting serum levels of total cholesterol, triglycerides, HDL-cholesterol, LDL-cholesterol did not differ between the groups and were within the normal range (Table 2). Serum total cholesterol exceeded 199 mg/dl in two girls, and serum LDL-cholesterol was >129 mg/dl in two girls; all four patients were in the untreated group. No significant correlation was noted in the VPA-treated patients between serum lipid levels and age, duration of therapy, or serum drug levels.

Table 2. Fasting levels of metabolic parameters in valproate-treated and untreated girls with epilepsy
 Valproate-treated (n = 43)Untreated (n = 45)p Value
  1. Data are expressed as mean ± SD.

  2. HOMA-IR, homeostatic model assessment of insulin resistance; QUICKI, quantitative insulin sensitivity check index; FG/FI, fasting glucose/fasting insulin; IGF-1, insulin-like growth factor 1; TSH, thyroid-stimulating hormone; FT4, free thyroxine.

Glucose (mg/dl)78.5 ± 4.5 80.2 ± 6.7 0.20
Insulin (μ/ml)12.0 ± 9.2 10.7 ± 10.00.57
Total cholesterol (mg/dl)159.8 ± 19.3 160.3 ± 28.8 0.92
HDL cholesterol (mg/dl)52.9 ± 11.352.8 ± 12.00.96
LDL cholesterol (mg/dl)75.2 ± 36.274.5 ± 42.40.92
HOMA—IR2.31 ± 1.742.16 ± 2.260.76
QUICKI0.35 ± 0.040.35 ± 0.030.52
FG/FI9.5 ± 6.311.1 ± 6.5 0.32
IGF-1 (ng/ml)227.4 ± 100.0288.4 ± 139.00.13
Leptin (ng/ml)26.3 ± 14.417.3 ± 12.50.20
TSH (mIU/L)2.48 ± 1.441.68 ± 0.9 <0.01  
FT4 (pM)15.9 ± 2.9 17.3 ± 2.2 0.02

Fasting serum levels of glucose were similar in the two groups and within the normal range. Mean fasting insulin level was also similar in the two groups, measuring 12.0 ± 9.2 μU/ml (range, 2–41) in the VPA group and 10.7 ± 10.0 μU/ml (range, 2–62) in the untreated group. No differences were seen in HOMA-IR, QUICKI, and FG/IF between the groups or among the pubertal subgroups.

Mean HOMA-IR was similar in both groups (2.31 ± 1.74 vs. 2.16 ± 2.26, p = 0.76). An HOMA-IR value >95th percentile for age and sex was found in eight (25%) of the 32 in the treated group and five (12.5%) of 40 in the untreated group; this difference was not statistically significant (p = 0.21). However, only three patients in the treated group and four in the untreated group had HOMA-IR values above the normal mean values according to pubertal stage and for obese and nonobese individuals. Among the 25 patients with an HOMA-IR >2.0, only three in the treated group and four in the untreated group had a BMI-SDS of ≥2.

BMI-SDS was negatively correlated with SHBG (r=−0.64; p < 0.001) but did not correlate with insulin levels.

Hormonal evaluation

The treated group had a significantly higher TSH level (2.48 ± 1.48 vs. 1.68 ± 0.91 mIU/L; p < 0.01) and lower FT4 level (15.90 ± 2.92 vs. 17.39 ± 2.26pmol/l; p = 0.02) than the untreated group, although values were within normal range in all cases.

Mean testosterone level was significantly higher in the VPA-treated group than in the untreated group (1.64 ± 1.45 vs. 0.81 ± 0.21 nM; p < 0.003), and significantly higher in the postmenarcheal VPA-treated subgroup than in their untreated counterparts (1.83 ± 0.65 vs. 0.88 ± 0.24 nM; p = 0.006) (Table 3). This difference was even more prominent when the comparison was confined to postpubertal girls (2.12 ± 1.74 vs. 0.91 ± 0.25; p = 0.007; n = 25 treated and 24 untreated, respectively). The data required for calculating FAI were available in 19 postmenarcheal girls, 11 in the VPA-treated group and eight in the untreated group; no difference in mean values was found between the groups (2.96 ± 2.6 vs. 3.23 ± 2.6; p = 0.82, respectively).

Table 3. Serum hormone levels in valproate-treated and untreated girls with epilepsy
 Before menarcheAfter menarche
Valproate-treated (n = 13)Untreated (n = 15)Valproate-treated (n = 30)Untreated (n = 30)
  1. Data are expressed as mean ± SD.

  2. T, testosterone; DHEA-S, dehydroepiandrosterone sulfate; A, androstenedione; E2, estradiol; LH, luteinizing hormone; FSH, follicle-stimulating hormone.

  3. ap = 0.006.

Age (yr)11.19.616.615.1
T (nM)1.2 ± 0.70.7 ± 0.0 1.83 ± 0.65a 0.9 ± 0.2a
DHEA-S (μg/ml)0.7 ± 0.31.7 ± 1.52.7 ± 2.14.0 ± 2.5
A (nM)3.9 ± 2.71.6 ± 1.26.0 ± 4.64.6 ± 2.2
E2 (pM)42 ± 4916.9 ± 22.0154 ± 276110 ± 162
LH (U/L)1.7 ± 1.50.4 ± 0.55.2 ± 3.53.9 ± 2.7
FSH (U/L)3.4 ± 1.82.9 ± 1.65.0 ± 2.15.0 ± 1.9

Mean androstenedione level was also higher in the treated group, but the difference did not reach statistical significance when analyzed by menarcheal subgroups. No significant differences were found between the groups for LH, FSH, DHEA-S, estradiol, PRL, SHBG, IGF-1, or leptin levels (data for the last four are not shown). No significant correlations were found between testosterone levels and BMI-SDS, insulin levels, duration of epilepsy, or duration or dose of VPA treatment.

The prevalence of irregular menses, acne, and hirsutism was similar in the two groups (24% vs. 25% for irregular menses and 26.8% vs. 28.6% for acne, respectively) (Table 4).

Table 4. Prevalence of clinical endocrine abnormalities in valproate-treated and untreated girls with epilepsy
 Valproate-treatedUntreatedp Value
  1. PCO, polycystic ovary; PCOS, polycystic ovary syndrome.

PCO morphology44% (11/25)31% (4/13)0.30
Hirsutism23% (10/43)13% (6/45)0.28
Irregular menses24% (6/25) 29% (7/24)0.75
Acne27% (12/25) 29% (10/24)0.77
Overweight26% (11/43)16% (7/45)0.30
Obesity16% (7/43)  9% (4/45)0.50
PCOS16% (4/25) 16% (4/24)0.92

No difference was found between the two groups in uterine or ovarian measurements or PCO morphology. PCOS was found in four girls in the VPA group and four girls in the untreated group. Only one of the eight girls with PCOS was obese, with a BMI-SDS of 2.5. This girl was not treated with VPA.


In this study of girls with epilepsy, long-term VPA treatment was not associated with obesity, PCOS, or shorter height. Our data showed no difference between the VPA-treated and untreated girls with epilepsy in mean BMI-SDS, fat percentage measured by bioimpedance, triceps skinfold thickness, or waist-to-hip ratio. Although a trend of a higher prevalence of overweight and obesity was noted in the VPA group, it did not reach statistical significance. The rates of both overweight and obesity were similar to those reported recently in healthy 13-year-old schoolgirls in Israel (Lissau et al., 2004). These data confirm the findings presented in a summary based on 16 clinical trials involving a total of 1140 patients treated with VPA, wherein only 3% of the patients had a gain in body weight (Schmidt, 1984). However, the reported rate of weight gain in the consecutive studies of the different patient populations treated with VPA varied from 4 to 71% (Jallon and Picard, 2001). Egger and Brett (1981) noted a body-weight gain in 44% of 100 children with epilepsy treated with VPA at a daily dose of 30–50 mg/kg. They found that “Small doses of VPA did not affect body weight, whereas doses higher than 30 mg/kg did.” No relation was found with the type of epilepsy or the degree to which seizures were controlled by the drug. Thus the absence of an association of VPA with a higher prevalence of obesity in the present study can be explained by the dose used, which did not exceed 30 mg/kg/day in any of the patients. However, in view of previous findings that an increase in obesity in VPA-treated children first became evident after 4 years of treatment, it is also possible that our follow-up period was too short to draw conclusions regarding this factor (Rättyä et al., 1999).

VPA treatment was not associated with a higher prevalence of acne, menses irregularities, PCOS, or PCO morphology on ultrasonography. No difference was found in DHEAS levels between the groups, even when analyzed by pubertal subgroups, in accordance with the results of two previous studies (Vainionpää et al., 1999; Stephen et al., 2001), but in disagreement with a recent one (El-Khayat et al., 2004). However, treatment was associated with higher testosterone levels in the girls after menarche, although not in the girls before menarche, implying that VPA may induce clinical and/or laboratory hyperandrogenism in postmenarcheal girls, as in adult women (Isojärvi et al., 1996; Murialdo et al., 1997). These data do not support the study of Vainionpää et al. (1999), which found increased testosterone levels in the prepubertal and pubertal stages. It is possible that the sensitivity to VPA-induced hyperandrogenism is a function of sexual maturation. The presence of hyperandrogenemia without hyperandrogenism may reflect the increase in testosterone that precedes the appearance of clinical signs. Others have suggested that in women treated with VPA, an increase in insulin may be responsible for insulinemic ovarian stimulation, triggering a metabolic cascade affecting several pathways and including excessive testosterone production (Vainionpää et al., 1999). However, according to our data, no correlation occurred between testosterone levels and BMI-SDS or insulin.

Alternatively, because testosterone is metabolized in the liver and VPA is an enzyme inhibitor, treatment with VPA could elevate serum testosterone levels by inhibiting its metabolism (Herzog, 1996). In addition, VPA may inhibit the conversion of testosterone to estradiol (Taubøll et al., 2003) or induce ovarian androgen biosynthesis (Nelson-DeGrave et al., 2004). Although we cannot rule out the possibility that the increased testosterone levels in the VPA-treated group were related to disease duration, we found no correlation between the duration of the epilepsy or of treatment and testosterone levels. These data also confirm previous studies showing an increase in serum testosterone levels after only 1 month of VPA therapy in women with epilepsy (Rättyä et al., 2001).

Our study did not support the association between PCOS and the use of VPA that was previously reported in adults (Isojärvi et al., 1996; Murialdo et al., 1997). A few plausible explanations exist for our results. First, the younger population may respond differently from the adults. Second, if obesity is indeed a major cause of PCOS in VPA-treated girls, then the patients in the present study would not be expected to manifest PCOS, as they were not obese. Third, both obesity and PCOS may be associated with higher doses of VPA. Finally, these results may be in line with those of Bilo et al. (2001) and Luef et al. (2002), who found that the prevalence of PCOS is increased in epilepsy, independent of AED use or seizure type.

It is noteworthy that previous studies investigating the effects of AEDs in children with epilepsy compared monotherapy with polytherapy (El-Kyayat et al., 2004) and included mixed populations of boys and girls (Guo et al., 2001; Jallon and Picard, 2001) or patients with epilepsy caused by different pathologies (El-Kyayat et al., 2004). Others compared treated patients with epilepsy with healthy controls (Rättyä et al., 1999; Vainionpää et al., 1999). To the best of our knowledge, ours is the first study to compare the effect of VPA in girls with epilepsy with the effect of epilepsy per se. Because it included a homogeneous group of girls treated with VPA monotherapy, any difference between the groups should be attributable to the treatment and not to the disease. Although the VPA-treated group had a longer disease duration than the untreated group, which could have contributed to the differences between them, considering that we failed to find a between-group difference for most of the parameters studied, this factor may actually support the absence of an endocrine effect of VPA. Furthermore, even if the untreated patients represent a group with less-frequent or less-severe seizures, this too would strengthen the data demonstrating an absence of hazardous endocrine effects. The small size of the prepubertal groups in the present study precluded a significant conclusion. Thus the study groups were redivided into subgroups according to menarche status. This method not only yielded more significant results, but also has an advantage in terms of practical clinical management, as it is easier to determine menarche status than to assess precise pubertal stage in a busy clinic.

Serum TSH levels were increased and FT4 levels slightly decreased in the girls treated with VPA, although they remained within normal range in all patients. The mechanism and clinical significance of these alterations remains unknown. Verrotti et al. (2001) did not find any change in thyroid hormone metabolism, and the TSH response to thyroid-releasing hormone was normal. A VPA-induced increase in TSH was previously described (Vainionpää et al., 2004), but not in association with a decrease in FT4. As the present study is a cross-sectional one, we cannot determine whether the between-group differences were a consequence of the treatment.

The absence of a significant association of long-term VPA therapy with mean levels of serum lipids in our study is in line with previous data (Berlit et al., 1982; Reddy, 1985).

Ht-SDS was comparable in both groups to that of the general population and to the Ht-SDS of the parents. Thus growth does not seem to be affected by either epilepsy or VPA treatment. These results support those studies showing that patients with epilepsy have normal linear growth and final height (Macardle et al., 1987; Kurlemann et al., 1997; Jallon and Picard, 2001; Mikkonen et al., 2005) and disagree with some reports by others (Guo et al., 2001; El-Khayat et al., 2004).

In our study, neither the type nor the duration of epilepsy was associated with the prevalence of clinical or laboratory endocrine abnormalities. These data do not support the statement by Herzog (1996) that women with temporal lobe epilepsy more frequently have disorders of the pituitary–gonadal axis.

In conclusion, long-term treatment with ≤25 mg/kg/d of VPA in girls is not associated with growth disturbances, obesity, or PCOS in most cases. Treatment may lead to increased testosterone levels in the postpubertal period, without clinical hyperandrogenism and without an increase in the prevalence of menstrual irregularities. Thus VPA seems to be a good choice for the treatment of epilepsy in girls and young female adolescents. Careful endocrine observation is recommended during treatment, especially in the postpubertal age group. Further prospective studies are required to corroborate our findings.