Pharmacotherapy of Autism Spectrum Disorder: Results from the Randomized BAART Clinical Trial

Abstract

Study Objective:

The objective of this trial—Biomarkers in Autism of Aripiprazole and Risperidone Treatment (BAART)—was to provide support and guidance for an evidence-based approach for the selection and monitoring of initial pharmacotherapy in patients with autism by assessing predictors of efficacy, tolerability, and safety.

Design:

Randomized, double-blind, parallel-group study.

Setting:

Three academic medical centers and a single private pediatric practice.

Patients:

Eighty children or adolescents (aged 6–17 years) with autistic disorder were enrolled, and 61 patients were randomized to study drug. Of those patients, 51 completed the 10-week trial, and 31 completed an optional 12-week blinded extension phase.

Intervention:

All patients were treated with 2 weeks of placebo before random assignment to receive aripiprazole (31 patients) or risperidone (30 patients) for 10 weeks. Sixteen placebo responders (20%) were excluded from further analysis. Drug dosing followed United States Food and Drug Administration (FDA) labeling, and weekly dosage adjustments were allowed until week 4; patients were then maintained on a fixed dose for 6 additional weeks.

Measurements and Main Results:

Safety, physical, and psychological assessments were recorded weekly or every two weeks. No significant differences in severity of illness between the aripiprazole and risperidone groups were noted at baseline. All patients significantly improved on the Aberrant Behavior Checklist irritability subscale after one week and continued for the remaining 9 weeks and the extension phase. Improvement was greatest in the risperidone group at every assessment period and was statistically significantly better than that in the aripiprazole group at weeks 3 and 6 (p<0.05). No dose-limiting adverse events occurred during the dose-titration period. Mean weight gain in the aripiprazole group was significantly less than that in the risperidone group at week 4 (0.62 vs 1.38 kg, p=0.033) and week 10 (1.61 vs 3.31 kg, p<0.001), but the difference became nonsignificant for the 31 patients completing the 3-month extension phase (4.36 vs 5.55 kg, p=0.26).

Conclusion:

Pharmacotherapy of patients with autism spectrum disorder resulted in behavioral improvement within one week and lasted at least 22 weeks. Weight gain occurred to a greater degree with risperidone than aripiprazole initially, but the differences became nonsignificant by the end of the trial. Our trial supports previous results of drug efficacy and safety in patients with autism spectrum disorder from other trials and extends the evidence-based support for choosing an FDA-approved drug for initial pharmacotherapy for autism spectrum disorder.

Autism spectrum disorder (ASD) and autism are general terms referring to a group of developmental disorders characterized by difficulties in social interactions, restrictive and repetitive behaviors, and irritability. Patients with ASD have varying degrees of intellectual ability. The estimated prevalence of ASD was 16.8 cases per 1000 children in 2014, or 1 in 59 children.(1) With the publication of the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) in 2013, the subtypes of autism including autistic disorder (AD) and Asperger’s syndrome were grouped together into the category of ASD. Although applied behavior analysis (ABA) techniques are the mainstay of treatment, they are expensive and time consuming, require trained therapists for their provision, and are not always available or reimbursable by health insurers. Pharmacotherapy is widely used (2), but only two drugs, from the therapeutic class of atypical antipsychotics, have been approved by the United States Food and Drug Administration (FDA). Risperidone received approval in 2006 and aripiprazole received approval in 2009 for the treatment of irritability of AD.

The initial choice of drug therapy has frequently been risperidone. This is likely due to its longer availability as a generic drug despite data from the adult psychiatric literature that it possesses more liability for weight gain, metabolic effects, stimulation of prolactin release, and possibly cardiovascular and neurologic adverse events compared with aripiprazole. (3–5) The pivotal clinical trials performed to demonstrate efficacy and safety of these drugs matched the active treatment against placebo in 8-week protocols.(6–9) One meta-analysis examined efficacy and secondary effects in children with autism or intellectual disability taking either risperidone or aripiprazole and concluded that efficacy was similar, but metabolic events could not be adequately compared. (10) With the exception of a nonblinded Iranian study conducted for 8 weeks in an ASD population consisting of 65% patients with AD (11), there is a lack of extant data on efficacy and adverse events from a randomized controlled trial directly comparing aripiprazole with risperidone in patients with either AD or ASD.

The Biomarkers in Autism of Aripiprazole and Risperidone Treatment (BAART) project recruited subjects in South Carolina for a clinical trial between September 2011 and June 2015 (ClinicalTrials.gov identifier NCT01333072). The BAART project was designed to provide evidence-based guidance in the selection and monitoring of drug treatment of AD. Although the FDA has approved the use of aripiprazole and risperidone for the treatment of irritability associated with AD, a moderate response rate in pivotal clinical trials and concerns over tolerability and weight gain can force clinicians to select alternative drug treatments for which evidence-based support is sparse. Factors considered important in the BAART trial included psychiatric history; symptom response; measures of tolerability, weight, serum prolactin concentration, and drug and metabolite concentrations; and DNA profiling for a variety of single-nucleotide polymorphisms related to target genes for drug disposition and transport, response, and tolerability. This initial report summarizes the most important findings for application to clinical practice. Subsequent reports will focus in depth on other aspects of the results.

Methods

Study Design and Treatment Protocol

In this randomized, double-blind, parallel-group study, we treated 80 children and adolescents, aged 6–17 years, with AD, of whom 92.5% (74/80) were antipsychotic naive, with 2 weeks of placebo before random assignment to receive aripiprazole or risperidone for 10 weeks. Using strict criteria (>25% decrease in the Aberrant Behavior Checklist irritability subscale [ABC-I] score before randomization), 16 placebo responders (20%) were excluded from further analysis. Drug dosing followed FDA labeling, and weekly dosage adjustments were allowed until week 4; patients were then maintained on a fixed dose for 6 additional weeks. Safety, physical, and psychological assessments were recorded at weekly or every two weeks clinic visits. At the completion of the 10-week trial, participants were offered the opportunity to continue blinded medication for an extension phase of up to 3 months.

All research personnel including physicians, coordinators, and anyone who had patient or family contact were blinded to study medication and dosage throughout the study, except for a research assistant who prepared medication for dispensing and the study pharmacist who checked the accuracy of all medication before dispensing. Drug supplies were prepared at the Medical University of South Carolina (MUSC) study site and sent by registered mail to the participating medical centers. Drugs were formulated into identical-looking capsules, and all subjects took one capsule in the morning and one in the evening. Capsules labeled for morning administration contained placebo for patients randomized to aripiprazole to maintain blinded conditions. Enrolled subjects were initially given 2 weeks of study medication with dosing directions. Unbeknownst to the subjects or their caregivers, the first 2 weeks of capsules contained no active medication to constitute a complete placebo period.

At baseline, all patients began treatment at a low medication dose that could be adjusted weekly, as judged clinically appropriate by a child psychiatrist or developmental pediatrician. The starting dosage for aripiprazole was 2 mg/day, which could be increased to 5 mg/day after one week, and increased further thereafter in 5-mg increments to a maximum dose of 15 mg/day. The initial dose of risperidone for children weighing 20–45 kg was 0.25 mg/day, which was increased to 0.5 mg/day on day 4. The dose could be gradually increased in 0.5 mg-increments to a maximum dose of 2.5 mg/day, administered as 1.0 mg in the morning and 1.5 mg at bedtime, by the fourth treatment week. A slightly accelerated dosage was allowed for children who weighed more than 45 kg, to a maximum dosage of 3.0 mg/day.

Physician-prescribed dosing was divided into seven dosage levels, with level 1 being placebo. Initially, randomized patients were given 4 days of dosage level 2 and 3 days of level 3 medication. At each of the next four visits (at the end of weeks 1–4), the physician could increase or decrease the dosage by 1 level, or keep it the same, depending on response and/or the presence of adverse events. After the week 4 visit, the dosage level could no longer be increased but could be decreased if adverse events occurred.

Study Patients

Recruitment was conducted at three academic sites in South Carolina (MUSC in Charleston, the University of South Carolina in Columbia, and Greenville Health Systems in Greenville) and a single private pediatric practice, McLeod Health Systems (Florence, SC). Study details were disseminated to hospital pediatric and psychiatry departments, pediatric practices, neighborhood clinics, county school districts, ABA therapists, and autism support groups. Nearly 5000 letters containing study information were sent to the parents of children diagnosed with autism who were registered with the South Carolina Department of Disabilities and Social Needs. The study was registered with the appropriate federal and state agencies. All study procedures were approved by institutional review boards at the three medical centers.

Patients were recruited in the age range of 6–17 years, consistent with the FDA-approved drug product labeling for risperidone and aripiprazole, and with a body weight of at least 15 kg. All patients met the DSM-IV criteria for AD, established by medical record review and clinical interview. Clinical impressions were corroborated with the Autism Diagnostic Interview–Revised (ADR-R) criteria. This semi-structured interview was conducted by a clinician with specific training in its application and systematic review for interrater reliability. Inclusion in the trial required a Clinical Global Impression–Severity (CGI-S) score of 4 or greater (moderately ill) and an ABC-I score of 18 or greater. Subjects were required to have a history of no prior exposure to either study medication or be free of medication, or to have had only a trivial exposure to either study medication with a washout period of at least 2–4 weeks prior to enrollment. Prior use of aripiprazole or risperidone for more than 2 weeks within the past 3 years constituted a non-trivial exposure and was an exclusion criterion for study participation.

An initial telephone interview was conducted to answer parents’ and caregivers’ questions, provide study details, and assess inclusion criteria. Figure 1 provides a flow diagram of the patients included at each phase of the study and reasons for exclusion. Patients who met all inclusion criteria were invited to participate in a clinic screening at their choice of four study sites. During the screening, the informed consent process was documented, and a medical history was recorded. Autistic disorder was diagnosed by trained developmental pediatricians and child psychiatrists using the Autism Diagnostic Interview–Revised (ADI-R) and the Autism Diagnostic Observation Schedule (ADOS). Intelligence was measured using the Stanford-Binet Intelligence Scale (5th edition). Behavior was assessed using the Vineland Adaptive Behavior Scales (2nd edition) and ABC. CGI-S and CGI-Improvement (CGI-I) scores were recorded at each patient assessment.

Figure 1.

Flow diagram of the numbers of patients at each phase of the study and reasons for exclusion.

Outcome Measures

During patient screening, basic vital sign and physical measurements were recorded including height, weight, blood pressure, and hip and waist circumference. The ABC and the Children’s Sleep Habits Questionnaire were administered to the patient’s caregiver. The study physicians (psychiatrists or developmental pediatricians) conducted a physical and neurological evaluation. To assess possible neurologic side effects, study physicians who had undergone interrater reliability training administered the Simpson-Angus Extrapyramidal Side Effects Scale (SAS), the Abnormal Involuntary Movement Scale (AIMS), and the Barnes Akathisia Rating Scale (BARS). A CGI score was recorded along with an electrocardiogram. Blood samples were collected and analyzed for standard biochemistry profiles, lipid concentrations, hematologic indices, prolactin concentration, study drug plasma concentration, and DNA profiling for the presence of relevant genetic polymorphisms.

At baseline and subsequent visits (weeks 1–10), vital signs were measured, and administration of the ABC, SAS, AIMS, BARS, and CGI instruments was repeated. Any adverse events and changes in nonstudy medications were noted. During the week 10 visit, a full evaluation was conducted, and blood samples were collected for all chemistry and hematology assays. Patients were given the option to continue in the study for an additional 12 weeks, with visits every 4 weeks. Study medication and dosage continued to be blinded during this optional extension phase.

Statistical Analysis

The study was originally powered to detect a mean difference of 3.5 units on the ABC-I scale equivalent to an effect size of 0.31. We planned to enroll 224 patients, with a goal of achieving an effective sample size of 100 per group after elimination of placebo responders. This would yield an 82% chance of detecting a difference of 3.5 for mean ABC-I scores between drug assignments. Recruitment did not meet expectations, as discussed below, but the final 25 subjects per group yielded a 76% chance of detecting an effect size of 0.5 (mean difference of 5.0). Descriptive statistics were used to characterize patient and study variables. Two-tailed t tests were conducted to compare mean outcomes between the treatment groups. Paired t tests were conducted to compare changes in the means from beginning to end of treatment at 10 weeks. A last observation carried forward analysis was not applied to the data.

Results

Patient Characteristics and Metabolic and Psychiatric Measures

After the elimination of placebo responders, 61 patients were randomized to drug treatment groups (Table 1), and 51 completed the 10-week protocol. No significant differences in metabolic or psychiatric measures were noted between drug treatment groups at baseline (Table 2). Except for prolactin level, no significant changes in blood chemistry were observed at week 10 for either treatment group. Among patients in the aripiprazole group, average prolactin levels decreased from 9.3 to 2.8 ng/mL (p<0.001). Among patients in the risperidone group, prolactin increased from 9.8 to 40.4 ng/mL (p<0.001).

Table 1.

Demographic and Clinical Characteristics of the Study Patients

Characteristic	Aripiprazole group (n=31)		Risperidone group (n=30)
Male	25 (81%)		23 (77%)
Age (yrs), median	8.5		8.3
Age (yrs), range	6.0–15.1		6.3–17.5
Race
White	23 (74%)		15 (50%)
Black	7 (23%)		12 (40%)
White/Black	1 (3%)		1 (3%)
White/Native American	0		1 (3%)
Asian	0		1 (3%)
Ethnicity
Hispanic	0		5 (17%)a
Prior treatment with a study drug	4 (13%)		1 (3%)
Final medication dosage	Dosage (mg/day)	No. of patients	Dosage (mg/day)	No. of patients
	2	8	0.5	5
	5	9	1.0	5
	10	13	1.5	7
	15	1	2.0	9
			2.5	2
			3.0	2

Data are no. (%) of patients unless otherwise specified.

^aχ²=0.017.

Table 2.

Physical, Metabolic, and Psychiatric Measures of the Study Patients

Measure	Aripiprazole group (n=31)			Risperidone group (n=30)
	Baseline	Week 10	Week 22	Baseline	Week 10	Week 22
Weight (kg), mean	38.3	40.9	49.3	35.8	38.7	39.9
Weight (kg), median	31.6	33.2	37.6	33.1	32.9	34.3
BMI, mean	19.2	20.1	21.7	19.1	20.6	20.8
BMI, median	17.1	18.1	18.2	18.0	18.7	19.6
Total Cholesterol Level (mg/dl), mean	163.3	160.8		158.8	156.4
HDL Level (mg/dl), mean	56.1	54.5		56.6	54.1
LDL Level (mg/dl), mean	90.8	87.8		87.0	96.9
Triglyceride Level (mg/dl), mean	84.9	96.7		81.9	81.9
Prolactin Level (ng/ml), mean	9.3	2.8		9.8	40.4
ABC-I	27.3	14.1	13.6	27.8	12.7	8.1
CGI-S	4.87	3.33	3.40	4.60	3.48	3.27
CGI-I		2.04	1.60		1.84	1.67

BMI = body mass index; LDL = low-density lipoprotein cholesterol; HDL = high-density lipoprotein cholesterol; ABC-I = Aberrant Behavior Checklist Irritability subscale; CGI-S = Clinical Global Impression–Severity; CGI-I = Clinical Global Impression–Improvement.

Efficacy

Major outcome variables are shown in Table 2. At screening and pretreatment, no statistically significant differences were observed in the ABC-I values from patients who would eventually be randomized to a drug treatment group or who would be dropped from analysis as a placebo responder (Figure 2). At baseline, 16 of the 80 patients who had received 2 weeks of placebo had shown a drop in ABC-I scores > 25% and were considered to be placebo responders.

Figure 2.

Time course of clinical responses over 22 weeks as assessed by Aberrant Behavior Checklist–Irritability subscale (ABC-I) score. Pretreatment consisted of 2 weeks of continuous double-blind placebo treatment.

Beginning at week 1 and continuing to week 22, both drug groups at all assessment periods showed highly significant decreases in ABC-I scores compared to baseline (Figure 2; p<0.001 [p values not displayed]). A trend can be noted for more improvement in the risperidone group at all assessment points, but the difference compared to aripiprazole reached statistical significance only at weeks 3 and 6 (p<0.05). After 10 weeks, mean CGI-S scores were significantly lower for both the aripiprazole (3.33) and risperidone (3.48) groups compared to baseline (4.87 and 4.60, respectively; p<0.001) (Table 2). The majority of 51 patients at 10 weeks (78%) were rated as very much or much improved on the CGI-I scale (Figure 3). No significant differences in improvement were observed between treatments with respect to CGI-I scores. All 31 participants (50.8% of the randomized population) who completed the extension phase were rated on the CGI-I as much or very much improved.

Figure 3.

Summary of Clinical Global Impression–Improvement (CGI-I) scores over 22 weeks. PRs = placebo responders; 0th = 0 week; Arip = aripiprazole; Risp = risperidone.

Adverse Events and Early Termination

By week 10 of the study, 61% (19/31 patients) of the aripiprazole group and 77% (23/30 patients) of the risperidone group experienced one or more adverse events that were attributed to receiving study medication (Table 3). Adverse effects in four patients receiving aripiprazole (bedwetting, weight gain, stomach aches, and tremors) and two patients receiving risperidone (weight gain) prompted discontinuation of treatment. Three children stopped attending visits with no explanation provided, and one was discontinued by the investigators due to the addition of a protocol-excluded medication (stimulant) during the study. Eight patients taking aripiprazole (26%) experienced an increase of more than 7% of their baseline weight. Among patients taking risperidone, 21 (70%) experienced significant weight gain.

Table 3.

Adverse Events Possibly, Probably, or Definitely Related to Study Medication

	Aripiprazole group (n=31)		Risperidone group (n=30)
	No. of Patients	Percentage	No. of Patients	Percentage
Patients with at least 1 adverse event	19	61	23	77
Adverse event
sedation	7	23	2	7
somnolence	4	13	0	0
difficulty sleeping	3	10	0	0
nausea	1	3	0	0
constipation	1	3	2	7
increased appetite	0	0	3	10
enuresis	4	13	0	0
agitation	2	6	0	0
headache	1	3	1	3
decreased appetite	1	3	0	0
vomiting	1	3	0	0
drooling	2	6	4	13
dizziness	1	3	0	0
tachycardia	0	0	1	3
muscle rigidity	1	3	0	0
restlessness	0	0	1	3
weight gain >7%	8	26	21	70
other	3	10	7	23

During the optional 12-week extension phase, one patient taking aripiprazole terminated early due to enuresis. Among patients taking risperidone, four terminated early due to adverse events: increased aggression (1), tachycardia (1), and excessive weight gain (2). No serious adverse events occurred.

Discussion

The efficacy of aripiprazole and risperidone for reducing the irritability associated with AD was previously shown in pivotal clinical trials comparing drug to placebo. (6–9) The current trial is the first to compare the two drugs in a rigorous design to study AD under blinded conditions, with exclusion of placebo responders and an active treatment phase over 5 months (22 weeks). Rigorous criteria were used to establish a diagnosis of AD. The assumption was that a relatively homogeneous population would be best to define differences in drug response if patients were excluded with Asperger’s and other subcategories of ASD that would likely increase the variability in outcomes. In addition, as weight gain was a variable of specific interest, we sought to study a population that had not been exposed to either of the atypical antipsychotics that may have previously stimulated accelerated weight gain or predisposed the patient to this adverse event. Also, a history of a previous nonresponse to one of the study drugs would create an ethical exclusion to randomly assign a patient to a known or suspected ineffective treatment. The exclusion of 20% of the eligible population from analysis for demonstrating a placebo response ensured a design in which results would reflect a true drug response.

Both study drugs produced a statistically significant drop in irritability measured by the ABC-I scale within one week of randomization (Figure 2). At this time, all patients in the aripiprazole group were receiving 2 mg/day whereas patients in the risperidone group were receiving either 0.5 or 1.0 mg/day according to their weight. The final medication dosage (Table 1) allows some insight into the dose-response profile of the study drugs. Thus, 25.8% (8/31) of patients in the aripiprazole group and 33% (10/30) of patients in the risperidone group were treated effectively with the lowest dose and required no further dose increases to maintain or increase improvement through the remainder of the trial. Over half (17/30) of patients receiving risperidone were effectively treated with doses of 1.5 mg/day or lower. Only one patient receiving aripiprazole required more than 10 mg/day as an effective dose, whereas only 4 patients receiving risperidone required more than 2 mg/day. These findings provide some assurance that an initially tolerated dose is likely to be effective for a substantial number of patients prescribed either drug, without having to increase drug exposure at the risk of adverse events.

The pattern of ABC-I scores over time demonstrates that improvement increased with both drugs throughout the study (Figure 2). Scores for both drugs were statistically significant compared to baseline for all assessment intervals (p values not shown). The mean ABC-I score for risperidone was lower at all assessments than that for aripiprazole and statistically significantly lower at weeks 3 and 6 (p<0.05, Figure 2). These differences cannot be accounted for by a difference in baseline scores. Drug dosage was held constant after week 4, but clinical improvement can be seen to continue. By week 10, at the conclusion of the main study, and at week 22, at the conclusion of the blinded extension phase, patients had improved while maintaining the same dosage as week 4. These results imply that time on drug therapy was more important than increasing the dosage for continued clinical improvement.

Placebo responders (those who showed >25% decrease in the ABC-I at 2 weeks) were excluded from further analysis. Twenty percent of patients showed improvement in blinded ratings without drug exposure. This increases the reliability of the results from the patients in the trial who subsequently received active medication. Placebo responders reflect a difficulty in clinical practice of knowing when a response is a true drug effect versus a positive expectation. The degree of improvement in the ABC-I of placebo responders was greater over the 2-week placebo period than the degree of improvement seen in the first 2 weeks of drug in either active medication group. Randomized patients continuing in the trial did not show a similar degree of improvement as the excluded placebo responders until they had received active drug for 10 weeks. Even though statistically significant improvement was present in patients receiving an active medication for one week, clinicians may view with some skepticism the patient who shows dramatic improvement at the outset of treatment. Whether clinical improvement for patients showing a placebo response after 2 weeks continues with subsequent treatment cannot be currently addressed on the basis of these data.

Adverse events were responsible for few dropouts from the trial, and none were considered serious adverse events (Table 3). As dose titration ended at week 4, the appearance of adverse events that resulted in discontinuation persisted for a much longer period, with several patients concluding participation during the extension phase. Of interest was the occurrence of four episodes of enuresis occurring with aripiprazole (Table 3). This adverse event is normally associated with the use of risperidone. (14) Also apparent was a differential adverse event profile, weight gain, and an increase in plasma prolactin level with risperidone.

A limitation of this trial was the difficulty in recruitment to prestudy goals. We used resources from four sites across South Carolina. On three occasions, we used the mailing address of all children in the Department of Disabilities and Social Needs of South Carolina, the government agency responsible for determining eligibility for receipt of state support of children with a diagnosis of autism. Anecdotal data from school officials, pediatricians, psychiatrists, and social workers, along with objective data from medical records of potential patients, leads to a partial explanation for low recruitment. By the age of 6 years—the FDA lower limit for prescribing of the study drugs to children within label guidelines—a large proportion of eligible subjects were already being treated off-label with one of the study drugs or with alternative pharmacotherapy. The difficulty imposed by this situation for conducting evidenced-based research of comparative treatments is that a previous off-label exposure may influence trial eligibility and/or drug effects.

One additional study has reported results of a trial in which aripiprazole was compared to risperidone in a predominant population of patients with ASD.(11) Important differences exist in the study design of that trial and our trial. Our trial was conducted in a rigorously defined population of patients with AD, rigorous efforts were made to maintain double-blind conditions, and 92.5% of our patients were antipsychotic naïve.

Several compounds from other drug classes, including antidepressants, have failed to demonstrate comparative effectiveness (2,13–16). Atypical antipsychotics produce adverse events in most patients, regardless of indication. Until better tolerated or more effective drugs are developed for the symptoms of autism, risperidone and aripiprazole are likely to remain the major drugs of choice for behavioral symptoms in children and adolescents with ASD.

Conclusion

These results from this rigorous clinical trial document the outcomes, using widely accepted measures of efficacy, and benefits provided to children and adolescents with AD who were treated for 10 weeks with aripiprazole or risperidone after exclusion of placebo responders. Our trial supports previous results of drug efficacy and safety in patients with ASD from other trials (6–11) and extends the evidence-based support for choosing an FDA-approved drug for initial pharmacotherapy for ASD.