INTRODUCTION

Autism spectrum disorder (ASD) refers to a group of developmental disorders (Autistic disorder, Asperger's syndrome and pervasive developmental disorder not otherwise specified) affecting social and communicative functions and characterized by repetitive behaviors/restricted interests (American Psychiatric Association, 1994).

Impulsivity, self-injury, and other-directed aggression are common features in patients with ASD, and have a major impact on the care of affected individuals. For this reason, this domain has been the target of clinical trials, mostly focused on the use of atypical antipsychotics, (McCracken et al, 2002), and also of early studies of anticonvulsants, such as valproate, psychostimulants, and α- and β-adrenergic agonists.

Valproate has received FDA indications for the treatment of epilepsy (10 years and older), bipolar disorder (adult), and migraine prophylaxis (adult). Its mechanism of action is not well understood but may include the following: It potentiates GABA inhibitory effects in the CNS (Soderpalm, 2002) and is likely to have epigenetic effects, as it is a histone deacetylase inhibitor (Göttlicher, 2004). These two mechanisms are of specific interest in ASD, given theories of decreased inhibitory control in autism (Casanova et al, 2003), high frequency of seizures, and epileptiform EEGs in this population, especially in individuals with lower IQs (Amiet et al, 2008), and the increasing evidence to support a role of gene expression abnormalities in the pathophysiology of multiple neuropsychiatric disorders (Szyf, 2009). Other mechanisms that may or may not be relevant to the treatment of irritability include functional blockade of voltage-sensitive sodium channels, attenuation of NMDA-mediated excitation, influences on serotonin and norepinephrine function, effects on second messenger systems, and potential neuroprotective effects (Manji and Chen, 2000; Yasuda et al, 2009; Chen et al, 1999).

In adults, valproate has shown to have some efficacy at reducing aggressive behaviors across diagnostic groups, and controlled trials have documented a reduction in irritability in cluster B personality disorders (Hollander et al, 2003, 2005). There are case reports and case series suggesting the efficacy of valproate for aggression in children and adults with mental retardation and associated comorbidities (Mattes, 1992; Kastner et al, 1993; Sovner, 1989; Donovan et al, 1997; Damore et al, 1998). In an open-label study, we found that valproate reduces irritability and aggression in children with ASD (Hollander et al, 2001). Hellings et al (2005) conducted a double-blind, placebo-controlled trial of divalproex for aggression in ASD, which did not demonstrate efficacy in reducing aggression. However, the authors felt this might reflect high intersubject variability, small sample size, and a large placebo effect, and recommended further evaluation.

Most research into maladaptive behaviors in ASD has focused on the use of atypical antipsychotics (Jesner et al, 2007), with early data also potentially supporting the use of mood stabilizers, stimulants, and α- and β-adrenergic agonists (Aman, 2004). However, neuroleptic medications are associated with side effects, such as weight gain, which may increase the likelihood of diabetes and cardiovascular disease, with sedation and extrapyramidal symptoms, and not all patients exposed are responders. Lithium has documented efficacy for the treatment of irritability and aggression in several other disorders and early data may support its use in ASD (Kerbeshian et al, 1987; Teingard and Biederman, 1987; Craft et al, 1987), although, given the narrow therapeutic index of lithium, it is unlikely to be widely used in children with ASD. Other mood-stabilizing anticonvulsants, such as lamotrigine and levetiracetam, had promising early open-label data to support their use in ASD (Vebrant and Bauziene, 1994), but follow-up randomized controlled data failed to support this claim (Belsito et al, 2001; Wasserman et al, 2006).

Given the side effect profile of atypical antipsychotics, limited data to support efficacy for other compounds, conflicting early data for the use of divalproex in ASD, as well as the frequent occurrence of epileptiform EEG in patients with ASD, a further study of this compound in this population is warranted. This study examines the effect of divalproex sodium in the treatment of irritability/aggression in children with ASD, by means of a 12-week, double-blind, placebo-controlled trial, and explores the effect of baseline epileptiform activity on treatment response for irritability.

PATIENTS AND METHODS

The study was registered at clinicaltrials.gov (NCT00211757).

Study Participants

Inclusion criteria

Subjects were children aged 5–17 years, outpatients, who met DSM-IV-TR diagnostic criteria for autistic disorder, full diagnostic criteria on the ADI-R and autism spectrum criteria on the ADOS-G. Subjects had to be at least moderately ill (CGI-Severity score of at least ‘4’) to justify exposure to this medication. The population was also stratified for significant irritability/aggression difficulties at baseline, such that children had an Overt Aggression Scale-Modified (OAS-M) score of at least 13 or an Aberrant Behavior Checklist (ABC)-Irritability score of at least 18 (raw scores) to qualify.

Exclusion criteria

We excluded sexually active and pregnant females and nursing mothers; subjects with overall adaptive behavior scores below the age of 2 years on the Vineland Adaptive Behavior Rating Scale; subjects with active or unstable epilepsy, other Axis I disorders, unstable medical illness, genetic syndromes, or congenital infections associated with autism-like syndromes, prematurity; subjects treated within the previous 30 days with any drug known to have a well-defined potential for toxicity or with any psychotropic drugs; subjects with clinically significant abnormalities in laboratory tests or physical examination; subjects with a history of hypersensitivity or severe side effects associated with the use of divalproex sodium or other ineffective previous therapeutic trial of divalproex sodium (serum levels within the range of 50–100 μg/ml for 6 weeks); and subjects who have begun any new nonmedication treatments, such as diet, vitamins, and psychosocial therapy, within the previous 3 months. A detailed clinical interview with parents by an expert clinician, followed by a physical examination and blood test, was used to ensure that subjects did not meet any of the exclusion criteria.

Study Design

This was a 12-week randomized double-blind, placebo-controlled trial. The study was approved by the institutional review board of the Mount Sinai School of Medicine. Informed consent was obtained after a complete description of the study to the subjects and as per the Helsinki agreement and local IRB guidelines. Participants responded to advertisements placed in newspapers, websites, and so on. If they passed a phone screen by the research assistant, they were invited to come in and sign the consent. Assent was obtained whenever possible. After consent was signed, inclusion/exclusion criteria were determined on the basis of diagnostic and adaptive functioning testing, clinical interview, physical examination, and blood test. Participants were randomized to divalproex vs placebo and the dose was titrated up according to body weight (see Table 1), therapeutic blood level (a minimum valproate blood level of 50 μg/ml, as is the established minimum for epilepsy), and ultimately treatment response. All clinicians involved in efficacy or safety assessments were blinded to the randomization condition. Efficacy measures were administered every 2 weeks by an independent evaluator, who was an experienced clinical psychologist blinded to side effects. Side effects were monitored by study physicians, who are experienced in treating children with ASD and using valproic acid formulations. The dose was titrated on the basis of feedback from a nonblinded physician who independently monitored blood. This clinician had no contact with the participants. All valproate levels and safety blood results were forwarded to him by the laboratory. He then instructed the study physicians to decrease, maintain, or increase the dose. Feedback on subjects randomized to placebo was based on a blocked schedule, so that all study clinicians remained blinded to the condition of randomization.

Table 1 Titration Schedule

Baseline Measures

Autism diagnostic interview-revised (ADI-R)

The ADI-R is a semi-structured psychiatric interview designed for the study of ASD and related disorders, typically administered to the subject's primary caretaker/family members (Rutter et al, 1994).

Autism diagnostic observation schedule-generic (ADOS-G)

This instrument was developed as a companion instrument for ADI-R. It is a standardized protocol for the observation of social and communicative behavior in children, adolescents, and adults (Lord et al, 1998).

Leiter international performance scale-revised (Leiter-R)

The Leiter-R is a nonverbal measure of intelligence and cognitive abilities. As it is nonverbal, it is especially suitable for children and adolescents who are cognitively delayed or autistic, as well as for those who are nonverbal, non-English speaking, ESL, or speech-, hearing-, or motor impaired. It has been used extensively in pharmacological studies of children with ASD (Roid and Miller, 1995,1997).

Primary Outcome Measures

We evaluated efficacy using the Clinical Global Impression-Improvement Scale (CGI-I) focusing on irritability, and the irritability subscale of the Aberrant Behavior Checklist (ABC).

The clinical global impression-improvement (CGI-I)

The CGI-I is a 7-point improvement scale. Ratings of 1 or 2 (responders) indicate a substantial reduction in symptoms, so that a treating clinician would be unlikely to readily change the treatment regimen. A rating of 3 (minimally improved) on the CGI is defined as a slight symptomatic improvement that is not deemed clinically significant; patients with such an improvement were not considered responders. We used two versions of this test, one focused on irritability (primary outcome measure) and a general version CGI-I-autism focused on all symptoms including core symptom domains. The CGI-I irritability took into consideration the scores from the ABC-Irritability subscale, the OAS-M aggression and irritability subscales and information from open-ended questioning related to the degree of interference, nature, and range of behavioral problems at school and at home (Guy, 1976).

The aberrant behavior checklist (ABC)-community version (Irritability subscale)

ABC-community version is the community version of the original residential version. It is designed to objectively identify five behavior subscales through observation by the primary caregiver: irritability, lethargy, stereotypy, hyperactivity, and inappropriate speech. The ABC was filled out by parents (Aman et al, 1985).

Secondary Outcome Measures

The Overt Aggression Scale-Modified (OAS-M)

The OAS-M is an instrument developed on the basis of an earlier version (the original Overt Aggression Scale, OAS), as well as on the basis of the Schedule for Affective Disorders and Schizophrenia (SADS) (Coccaro et al, 1991). It has been previously used in pediatric studies (Buitelaar et al, 2001), although the psychometric data for the scale was originally determined in adults.

The Child–Yale–Brown Obsessive-Compulsive Scale (CYBOCS)

CYBOCS was used as a secondary measure to examine the effect of divalproex sodium on repetitive behaviors. The Child–Yale–Brown Obsessive-Compulsive Scale rates, on a 5-point scale, the time spent, distress, interference, resistance, and control in relation to obsessions and compulsions. It has been shown to be a reliable and valid scale in ASD populations, and in measuring change in treatment studies of ASD (McDougle et al, 1995).

Exploratory Outcome Measures

Vineland Adaptive Behavior Scale

This is a semistructured informant interview that assesses the daily functioning of subjects. The scale has been normed for the autistic population. Items are classified under four major adaptive domains: communication, daily living skills, socialization, and motor skills (Sparrow et al, 1984).

Young Mania Rating Scale (YMRS)

It is a checklist of 11 items that was designed to measure the severity of manic symptoms and to gauge the effect of treatment on mania severity. Youngstrom et al reported on the scale's psychometric properties in children. The scale was administered as a parent report (Young et al, 1978; Youngstrom et al, 2003).

Electroencephalogram (EEG)

A sleep-deprived EEG was attempted in all participants. No sedation was used. The EEGs were reviewed by an experienced neurologist and were classified as epileptiform if spike activity was noted, abnormal but not epileptiform if other nonspecific abnormalities were noted, or normal. Given the small number of EEGs, no attempt was made to discuss the localization of epileptiform abnormalities or specific patterns.

Safety Measures

A physical examination was conducted at baseline and end visits. Blood monitoring of hematopoietic, liver, and renal function was carried out at baseline, weeks 2 and 4, and at end visit. Weight, height, and BMI were recorded at baseline and at end visit and vital signs were taken at baseline, weeks 2 and 4, and at end visit. Adverse event monitoring took place every week for the first 4 weeks and every 2 weeks thereafter. Questioning was focused on known side effects of divalproex sodium, followed by open-ended questioning. The side effects specifically elicited included nausea, vomiting, stomachaches, appetite changes, dizziness, tremors, confusion, headaches, hair loss, and weight changes.

Statistical Methods

Baseline characteristics

Independent samples t-tests were used to determine whether there were baseline differences between treatment groups on the following potential covariates: age, intelligence level, and baseline severity (ABC and OAS-M irritability subscales).

Outcome measures

CGI-I (χ2analysis). Consistent with intent-to treat principles, for those subjects missing the week-12 ratings, we imputed their value on the CGI at week 12 using mixed regression models based on the available values from all subjects and all seven time points. The predicted scores were then used to classify the subjects as responders or nonresponders at week 12 on the basis of the following: CGI2 (responders) or CGI>2 (nonresponders). χ2 test was used to compare the response between groups.

ABC, OAS-M, CYBOCS, VINELAND, YMRS (Mixed Model Analysis). Data sets were evaluated for skewness and outliers, and winsorized if necessary. In these analyses, we used all available data across all time points and fit a four-parameter mixed effects regression model to evaluate the weeks x effect. We specified an unstructured covariance (‘MANOVA’) matrix to obtain the error terms in the analyses.

RESULTS

Participant Disposition

A total of 55 children signed consent for this trial. Of them, 27 were randomized and were included in the safety and efficacy analysis. Nonrandomized subjects either did not meet the criteria for ASD or for adaptive functioning more than 2 years (n=23) or withdrew their consent before randomization (n=13). A total of 16 subjects were randomized to active treatment and 11 to placebo. Three subjects withdrew from the study before week 12 (two on active compound and one on placebo). Only one patient on active compound discontinued because of side effects (Figure 1).

Figure 1
figure 1

Consort diagram. *One subject withdrew because of side effects, one subject withdrew for nonefficacy. **One subject withdrew because of side effects.

Baseline Characteristics

All children met ASD criteria on both ADI-R and ADOS-G. Four of the 27 children had no phrase speech delay and were therefore classified as having Asperger's disorder (see Table 2 for subject characteristics). There were no significant differences in baseline characteristics, except for IQ: the placebo group had a significantly higher mean full-scale IQ (t=2.57, df=23, p=0.017). Thus, we evaluated the effects of treatment using full-scale IQ score as a covariate in the analyses. No subjects had epilepsy and none were on anticonvulsant medications. Eight subjects met our a priori OAS-M aggression entry criteria but not the ABC-Irritability entry criteria, likely reflecting children with significant aggressive outbursts but low levels of irritability throughout the day. In addition, a total of seven children had previous exposure to risperidone.

Table 2 Subject Characteristics

Efficacy Analysis

Primary

CGI-I for Irritability. On the basis of intent-to treat analyses, 10 of the 16 active treatment subjects (62.5%) showed a response tn irritability, whereas only one of the placebo subjects (9.09%) showed a response (OR=16.66). This effect is significant by Fisher's exact test (p=0.008). The odds ratio indicates that subjects receiving treatment with divalproex sodium are over 16 times more likely to respond to treatment than subjects receiving placebo. To control for the IQ differences between groups, we reanalyzed the data using a logistic regression model, with IQ as a continuous variable. The effect remains significant (p=0.045).

ABC-Irritability subscale. There is a significant weeks x condition interaction (t=−2.09, df=22.71, p=0.048), suggesting that the active group showed a drop of more than 0.53 points per week compared with the placebo group on the ABC parent irritability ratings (Figure 2). The significant condition x weeks interaction remains significant and indeed the effects seem somewhat stronger after controlling for IQ differences (t=−2.28, df=20.38, p=0.033). The mean irritability scores at baseline and at end point were the following: divalproexbaseline 22 (7.81), divalproexend 14.5 (6.67), placebobaseline 20.30 (7.36), placeboend 17.70 (7.94), effect size d=0.44 (moderate effect size).

Figure 2
figure 2

Improvements in ABC-Irritability subscale in divalproex vs placebo-randomized subjects over 12 weeks. There is a significant weeks x condition interaction (t=−2.09, df=22.71, p=0.048), suggesting that the active group showed a drop of more than 0.53 points per week compared with the placebo group. Note that the scatter plot includes all subjects, but the symbols correspond to ABC values. Some subjects had identical scores and are therefore reflected in a single overlapping symbol.

Secondary

OAS-M Irritability subscale. For two of the time points (weeks 0 and 8), the scale has a negative skew >1.0. On the basis of both winsorized (t=−1.09, df=23.44, p=0.28) and nonwinsorized (t=−1.37, df=24.01, p=0.181) data, no statistically significant improvement in this measure was noted in subjects receiving divalproex vs placebo. Controlling for the IQ differences did not change these conclusions. The mean irritability scores at baseline and end point were the following: divalproexbaseline 6.43 (1.41), divalproexend 5.42 (2.17), placebobaseline 5.36 (2.2), placeboend 6.25 (1.28).

CYBOCS. Mixed model analysis was used to examine the effect of divalproex sodium vs placebo on repetitive behaviors as measured by CYBOCS. There were no statistically significant differences between groups (p=0.748).

Exploratory Analysis

Vineland adaptive behavioral scale, YMRS

There were no statistically significant differences between groups in either the Vineland domains (communication: p=0.865, daily living: p=0.77, socialization: p=0.119) or the YMRS (p=0.987). Controlling for IQ did not change these findings.

CGI-I autism

There were only two responders, both from the divalproex group. As such, 12.5% of divalproex subjects responded at week 12, whereas 0% of the placebo group responded. The small number of responders, with none in the placebo group, only allowed for the computation of a Fisher's exact probability to test the relationship of response to treatment. This relationship is not significant (p=0.499).

Relation of EEG, Blood Levels and Treatment Response

Relation of IQ to EEG findings

In our sample, the mean nonverbal IQ was higher for subjects with epileptiform abnormalities than for those with no epileptiform abnormalities (92.75 vs 65.9).

Relation of response to EEG findings

We were able to obtain a sleep-deprived EEG at baseline for 19 of the 27 children. Of these, 17 EEGs included interpretable records and 10 of those were within the active group. Given the small sample size, the data are exploratory but intriguing, given the paucity of data regarding the effect of epileptiform abnormalities on treatment response in ASD. Table 3 suggests that subjects with abnormal/epileptiform EEGs (2/3), especially those with epileptiform EEGs (2/2), may be more likely to respond to divalproex sodium than subjects with normal EEG records (4/7).

Table 3 CGI-Irritability Treatment Response Based on Drug Status and EEG Results

Relation of response to valproate levels and mean dose

We examined whether valproate blood levels or dose correlated with improved outcomes in this sample. This analysis includes only the active group (n=16). Responders tended to have higher mean valproate blood levels compared with nonresponders: 89.77 (31.7) vs 64.33 (59.3), respectively. Subjects with therapeutic valproate levels between 87 and 110 mcg/ml had a 100% response rate on the CGI-Irritability Scale, whereas subjects with levels <87 had a 60% response rate and subjects with levels >110 had a response rate of 33%. Valproate dose had a moderate effect on improvement scores: responders 25.5 (8.58) mg/kg vs nonresponders 22.7 (0.83) mg/kg.

Safety Analysis

Divalproex sodium was well tolerated within this group. Most side effects were mild to moderate in severity and most resolved with small changes in dosing and did not require a discontinuation of medication. Table 4 lists all reported side effects that were different from baseline. The only treatment emergent side effect that was cited as a reason for discontinuation was a paradoxical increase in irritability associated with insomnia on just 250 mg of active compound in one participant. The side effect resolved with tapering off of the medication. No serious adverse events were reported in this cohort. Increased aggression was reported as a side effect in two subjects in the active group and in one subject in the placebo group. No cases of altered mental status were reported and no abnormalities in systolic/diastolic blood pressure or heart rate were noted at any visits for any of the subjects. There were no clinically meaningful elevations of liver function tests, no suppression of blood lines, and no cases of pancreatitis. There were no significant differences in weight gain between groups (weight gainplacebo=2.95±3.37 lbs, weight gainactive=3.02±6.41 lbs), but one subject in the active group had a clinically significant weight gain (>7% of starting weight). This subject had a trough valproate level of 104 μg/ml. Given that we had only one subject with such weight gain in the active group, we cannot reach any conclusion regarding the relationship between significant weight gain and blood levels. Unfortunately, height measurements proved to be very difficult and were tolerated by only a small number of children.

Table 4 Adverse Eventsa

DISCUSSION

This study suggests that valproate may be effective in the treatment of irritability in ASD. There are several reasons why this may be the case. First, the GABA-enhancing mechanism of valproate may be relevant to both the pathophysiology of aggression and that of ASD (Bjork et al, 2001; Casanova et al, 2003). Second, the documented ability of valproate to inhibit kindling has been proposed as an additional mechanism that may explain its effectiveness in treating mood lability, and as such, may be particularly important in the treatment of irritability (Soderpalm, 2002). Third, the treatment of underlying epileptiform abnormalities may contribute to behavioral response. This theory, although controversial, is supported by our very preliminary data that showed that children randomized to divalproex sodium with epileptiform EEGs were classified as responders. This hypothesis is further supported by a report by Stoll et al (1994), who reviewed 115 bipolar and schizoaffective lithium-refractory patients and found that those with a seizure or head injury history and abnormal EEG findings were much more likely to have a robust response to valproate (70%). Of interest, our sample had a mean nonverbal IQ below 70 and a systematic review of the literature has suggested that children with IQs under 70 are more likely to have seizures and epileptiform EEGs (Amiet et al, 2008). However, the usual pattern did not hold true, with the mean IQ actually being higher in children with epileptiform abnormalities than in children with nonepileptiform EEGs. Thus, our small sample does not support this well-documented phenomenon, but is limited by a small sample size.

We would also like to make note of our preliminary findings suggesting that therapeutic blood levels of valproate are associated with better response. Such results are congruent with data from a large study of valproate in adults with acute mania, in which it was reported that subjects with valproate blood levels of 87 mcg/ml or higher had improvement that was twice the effect size compared with subjects with blood levels less than 87 mcg/ml (Allen et al, 2006). Although our sample size did not allow for proper dose response and blood level response analysis, it is suggestive enough to highlight the need for a larger study designed and powered to address this issue.

The results of this study are not congruent with the report from the previous small randomized study of this drug in ASD (Hellings et al, 2005). As previously mentioned, the authors reported high intersubject variability and a large placebo effect, and recommended further evaluation. We would like to suggest that our strategy of stratifying for irritability severity at baseline and enrolling only those subjects with significant difficulties in this domain may be responsible for decreased intersubject variability and increased power in this study.

The safety profile of valproate in this study was very good. One should not assume though that the safety profile of a medication in a short-term study would be reflective of a long-term safety with this medication. Reported side effects of this medication include abdominal discomfort, nausea/vomiting, ataxia and tremor, hyperamonnemic encephalopathy, headaches, and weight gain. Serious side effects such as hepatic insufficiency and agranulocytosis have been rarely reported. One should consider the fact that the frequency of the rare side effects is much higher in children under the age of 2 years, and as such, the use of this medication in that age group remains controversial. There is also much controversy on the prevalence of polycystic ovarian syndrome, which is of concern to parents of high-functioning children in terms of its effects on fertility. Finally, although only one child in our cohort showed significant weight gain, such a side effect has been reported extensively in studies on other disorders, and follow-up studies will be critical in assessing the risk for cardiovascular and metabolic outcomes.

Limitations of our study include the relatively small sample size, which did not allow for a complete analysis of EEG and valproate blood level data. In addition, the absence of an EEG record at the end of the study makes it impossible for investigators to determine whether an improvement in EEG patterns correlated with treatment response. The choice of the ABC-Irritability subscale, although a validated measure in ASD, precludes us from making recommendations regarding specific types of aggression that may be responsive to divalproex. The fact that only seven children had previous exposure to an atypical antipsychotic also did not allow us to explore whether those with previous risperidone treatment were less responsive to valproate vs those without previous risperidone, and this remains a question for a future trial.

Of interest, there was no change in Vineland scores in this acute trial. The effect sizes of improvement noted in Vineland in the RUPP risperidone studies were small to medium and documented mostly at 6 months of treatment (Williams et al, 2006). In the acute phase and with our sample size, we did not have the power to detect such effects. Qualitative reports of relief due to decreased aggression did not seem to affect the scores on Vineland. There was also no change in CGI-I autism, which is not surprising, given that core symptom domain severity was scored within this measure. Finally, OAS-M did not seem to be sensitive to the changes detected by ABC. OAS-M is not a validated measure in ASD and does not have pediatric psychometric data, and we identified conceptual issues related to how this scale is scored that make it difficult to obtain reliable information in this population. For example, a child with relatively benign repetitive hitting of her chest that was continuous throughout the day was likely to get worse scores than children with more rare but severe self-injury.

Although it is hard to compare the effect size of a pilot study with that of a multicenter trial, the effect size for improvement on the ABC-Irritability subscale is moderate, but less than what was reported in the RUPP risperidone trial (very large). A follow-up of larger trials will be required for an appropriate comparison between the two drugs in terms of effect size of response and safety. In addition, a larger sample study powered to address the question of whether the presence of epileptiform abnormalities, while controlling for IQ, affects differential treatment response to anticonvulsants compared with atypicals, and whether treatment response is mediated by improvements in epileptiform abnormalities is required.

DISCLOSURE

Eric Hollander received consultation fees from Abbott, Neuropharm, Nastech, BMS, and Forest; received research grants from Abbott, and UBS Pharma; and has intellectual property related to oxytocin and memantine and ASD. Evdokia Anagnostou received a consultation fee by InteGragen. The rest of the authors have nothing to disclose.

CLINICAL TRIAL REGISTRATION

Clinicaltrials.gov

Title: Divalproex Sodium vs Placebo in Childhood/Adolescent ASD; ClinicalTrials.gov Identifier: NCT00211757; URL: http://clinicaltrials.gov/ct2/show/NCT00211757?term=ASD+and+divalproex&rank=1.