Comparative Efficacy of Antipsychotic Medication for... : Journal of Psychiatry Spectrum (2024)

INTRODUCTION

Autism, now referred to as autism spectrum disorder (ASD), is a neurodevelopmental disorder present in approximately 1% of the world population.^[1] ASD is characterized by a set of symptoms and behaviors, with stereotypy being a core feature. Stereotypies are repetitive; purposeless behaviors are categorized as restricted and repetitive behaviors (RRBs) in autism and can manifest as either motor or vocal actions. Examples of stereotypic behavior include head banging, hand writhing, head rotating, hand flapping, and vocal utterances. These movements can have disfiguring, exhausting, and sometimes injurious effects on the individual or those around them.

Repetitive behaviors in autism are commonly classified as stereotypy and compulsions. However, distinguishing between these two, particularly in those diagnosed with autism and comorbid obsessive–compulsive disorder, can be quite challenging. Many studies examining repetitive behaviors in autism utilize the Children’s Yale-Brown Obsessive-Compulsive Scale (CY-BOCS) to assess the severity and frequency of these behaviors.

The current literature on managing stereotypy in individuals with autism highlights the importance of using functional analysis and behavioral management techniques.^[2] The functional analysis involves understanding the purpose that stereotypic behavior serves for the individual, which can then inform the development of targeted behavioral interventions. In cases where expert advice on nonpharmacological interventions is not readily available, clinicians and parents may consider initiating pharmacotherapy to reduce stereotypy.

Various pharmacological interventions have been investigated for addressing repetitive behavior in autism, including antipsychotics, antidepressants, anticonvulsants, methylphenidate, omega-3 fatty acids, oxytocin, and their combinations. A review by Mulligan et al. in 2014^[3] explored a range of interventions for repetitive behaviors in autism, encompassing function-based and nonfunction-based behavioral interventions, sensory integration therapy, mixed treatments, and pharmacological interventions. However, the review only included five studies on pharmacological interventions, none of which specifically focused on antipsychotics. Despite this limitation, the review suggested that pharmacological agents could potentially be effective in treating repetitive behavior when emphasizing the need for more substantial evidence to support their use.

Yu et al.^[4] conducted a systematic review in 2020 on the efficacy of pharmacotherapy for RRB in ASD. The review included 14 randomized controlled trials (RCTs), but only three of these studies, involving a total of 140 participants, focused on the use of antipsychotics. One study used olanzapine, whereas the other two used risperidone. The studies involving risperidone showed positive results, supporting the use of antipsychotics to reduce repetitive behaviors in autism. In addition, the review noted that selective serotonin reuptake inhibitors were not more effective than placebo and did not conduct a specific meta-analysis for antipsychotics.

In a 2021 meta-analysis, Zhou et al.^[5] on the use of antipsychotics in treating RRB in individuals with ASD, analyzed seven studies involving 651 participants, all published before November 2019. The results indicated a favorable effect of antipsychotics, with a standardized mean difference (SMD) of 0.28 (95% confidence intervals [CI]: 0.08–0.48), suggesting a modest yet statistically significant reduction in RRBs in individuals with autism. Although antipsychotics were better than placebo, the authors did not comment on which antipsychotic has better efficacy in reducing the stereotypic behaviors in ASD.

Prior meta-analyses have commonly utilized a broad classification of RRB in ASD, rather than restricting to stereotypy. Our systematic review aimed to assess the efficacy of antipsychotics specifically in reducing stereotypic movements in individuals with autism. By narrowing our focus to stereotypy, we aimed to determine which antipsychotics might offer the greatest efficacy in reducing these specific behaviors. This review is crucial for building evidence base that can guide clinical practice and also furthering research on more targeted and effective treatment strategies for managing stereotypic behaviors in ASD.

METHODS

Protocol for this study was registered with PROSPERO (CRD42024524363) before the start of the study.

Information sources and search strategy

Two researchers (SK and NV) searched the two databases – PubMed and Cochrane CENTRAL. We restricted our search to journal articles published in the English language between 01 January 1992 to 15 March 2024. Search was done with various combinations of main keywords and MeSH words along with their synonyms with different combinations: “autism spectrum disorder” OR “autistic” OR “Asperger” OR “Pervasive developmental disorder” AND “stereotypic movement disorder” OR “stereotyped behaviour” OR “head banger” OR “body-rocking” OR “body rocking” AND antipsychotics (with individual names of antipsychotic drugs). The detailed search strategy is summarized in Supplementary Material https://links.lww.com/JOPS/A4.

Eligibility criteria

Inclusion criteria

To be included in a full-text review, the article must be a RCT published in the English language with antipsychotics monotherapy as an intervention compared with a placebo. Autism condition should have been diagnosed using standard criteria using either International Classification of Diseases 10^th revision, Diagnostic and Statistical Manual of Mental Disorders fourth edition (DSM-IV), or DSM-5. It was not restricted it to age, gender, or race. There should be a measure of stereotypy using one of the standardized instruments as an outcome, such as the Aberrant Behavior Checklist (ABC)-Stereotypy subscale or CY-BOCS-Compulsions subscale.

Exclusion criteria

Publications that are protocols, registered trials, commentaries, editorial, and open-label studies were excluded. RCTs that used antipsychotics as an add-on to behavioral therapy or a combination of antipsychotics, RCTs that did not assess stereotypy, and the participants who were intellectually impaired without a diagnosis of ASD were excluded. Trials published before 1992 and those reporting outcomes using different scales or reporting no mention of stereotypy were also excluded. The review was conducted in accordance with the PRISMA guidelines.^[6]

Procedure

After initial screening, the abstracts of the article were screened independently by two authors (SK and NV) to decide whether to include or exclude them for full-text retrieval.^[2] Any conflicts were resolved by mutual discussion until consensus was obtained. The full-text articles were downloaded for a detailed reading. Data extraction (as mentioned under data analysis) was done independently by two authors (SK and NV). The references list of the full-text articles of the included studies were manually searched to find additional relevant studies.

Our outcome of interest was a change in the score of stereotypy in the population diagnosed with autism as measured on standard scales. ABC-Stereotypy subscale and CY-BOCS-Compulsions subscale were used for rating stereotypy. Summary estimates for change in mean and standard deviation (SD)/standard error (SE) estimates for the above scales were extracted from the included articles for antipsychotic and placebo groups.

Data acquisition

The following data were extracted: author name and year of publication, the sample size in each group, participant characteristics, antipsychotic used, details of randomization, blinding, assessment of stereotypy and score values at baseline and at the end of the trial, a statistical method used, information on prior registration, and financial disclosure. These were cross-checked by another author, verifying with the original article.

If the data from the same study were presented in two separate articles, the data from the earlier published article were utilized. If a trial employed different measures of stereotypy and subsequently published two separate articles, both the articles were considered studies. For studies assessing varying doses of antipsychotics, we computed the combined mean and SD on the stereotypy scale for the antipsychotic group. The mean change in scores at the trial in both the antipsychotic and placebo groups was noted using intention-to-treat analysis. Whenever SE was provided, it was converted to SD. If a study reported an outcome with mean and SE, we converted the SE to SD using the formula SD = SE × Square root of the sample size. For studies with missing data, we tried contacting the authors^[7,8] and removed them as no response was obtained. Quality assessment of the full texts included for meta-analysis was done using the risk of bias tool version 2.0 (RoB 2).^[9] RoB 2 was used independently by two authors (SK and NV). Each study included was assessed on the following five areas: bias when randomization of participants, changes from the replanned interventions, outcome data if missing, outcome ascertainment different from preplanned, and selective reporting of the results. Risk under each domain and the overall risk were graded as either low risk, some concerns of risk, or high risk. We used a RoB 2 macro-enabled Excel sheet that has an inbuilt algorithm; if there is a discrepancy in correct risk grading in different domains and overall risk categorization between the algorithm generated and that by the reviewer, the reviewer had to give comments to justify his choice of category. If there was a discrepancy in quality assessment between two reviewers, they were analyzed for discrepancy and discrepancies were sorted out based on discussion. Those full-text articles included in the meta-analysis underwent risk of bias assessment (RoB 2) by two authors independently for the outcome measure on two scales, and a final consensus-based RoB chart was prepared for each measure of outcome. Thus, a final quality evaluation chart for all the studies included in the meta-analysis was created.

Statistical analysis

All statistical analyses were done in Revman version 5.4 using the random-effects model.^[10] The primary outcome of interest was stereotypy/compulsions symptom severity. Two separate forest plots were created: one for the ABC-Stereotypy subscale and another for the CY-BOCS-Compulsions subscale. The efficacy of the antipsychotics was pooled by a mean difference along with 95% CIs. The presence of significant heterogeneity between trials was determined by Cochran’s Q and I² tests. A subgroup analysis was also performed based on the different types of antipsychotics used in the trial. We did not anticipate this earlier, and hence, it was not mentioned in the protocol.

RESULTS

A total of 189 articles were retrieved from two databases, PubMed (n = 75) and CENTRAL (n = 114). After removing duplicates, we screened abstracts, going through their titles and texts, out of which 25 were shortlisted for full-text retrieval [Figure 1]. All these full-text articles were recovered and reviewed by two investigators independently. Manual search of articles was also done from the reference list of included articles. Finally, nine original research published articles were included in the systematic review.^[7,8,11-17]

Characteristics of the included studies

Finally, nine studies were considered for the systematic review. All of them were randomized placebo-controlled trials, either single center or multicenter based. Most of them were from the USA except for two studies: one from Japan^[15] and another from Canada.^[16] Only one study included the adult population;^[14] the sample size in different studies, which had trials of 6–12 weeks, varied from 31 to 150, including either gender [Table 1]. All of them had considered the intention-to-treat analysis, but only some studies stated prior trial registration.^{[7,8,11,12,15]} Most studies used one of the antipsychotics: aripiprazole, lurasidone, or risperidone. Three studies used both the ABC-Stereotypy subscale and the CY-BOCS-Compulsions scale to monitor changes in repetitive behavior.^[11,12,15]

Characteristics of the excluded studies

Most studies were disqualified because they were either open-label studies or had outcome measures other than stereotypy.^[18-33]

Meta-analysis of antipsychotic efficacy on stereotypy as measured on Aberrant Behavior Checklist-Stereotypy

Of these nine studies, six antipsychotic trials reported data for stereotypy scores as measured by the ABC-Stereotypy subscale. Other three studies had used different scales for assessment of stereotypy scores in their trials and hence data cannot be combined.

All the six trials showed a low risk of bias in quality assessment [Figure 2]. Loebel et al. (2016)^[11] and Marcus et al. in 2009^[12] showed a high risk of bias for missing outcome data (D3 domain in the risk of bias of individual studies), as they did not mention posttrial mean (SE) values for the groups. Whereas, Shea et al. in 2004^[16] showed a high risk of bias for not adequately explaining the randomization process (D1 – the domain of the risk of bias of individual studies). However, all these studies were scored as low risk of bias in the global assessment of their quality, justifying their inclusion in the meta-analysis and not doing sensitivity analysis.

Antipsychotics showed a significant decrease in stereotypy symptoms [Figure 3] compared to placebo (mean difference = −1.84, [95% CI: -2.84, -0.85], Z = 3.63, P = 0.004). There was evidence of modest heterogeneity between trials that did not reach statistical significance (Chi-square test for heterogeneity = 10.68, P = 0.06, I² = 53%).

Meta-analysis of antipsychotic efficacy on stereotypy as measured on Children’s Yale-Brown Obsessive Compulsive Scale-Compulsions subscale

Out of the nine studies, six studies reported data for stereotypy scores as measured on the CY-BOCS-Compulsive subscale. Other three studies had used different scales for the assessment of stereotypy scores in their trials and hence data cannot be combined. However, out of the six studies that measured stereotypy on the CY-BOCS-Compulsive subscale, two studies were dropped out as there were inadequate data and nonresponse to the data request.^[7,14]

Antipsychotic (4 trials, n = 554) did not show a significant reduction in stereotypy scores as measured on the Y-BOCS-Compulsions subscale.

Included studies showed a moderate level of heterogeneity but a nonsignificant level. Antipsychotics were associated with a score drop in stereotypy as measured in terms of compulsions compared to placebo but did not reach a significance level (mean difference = −0.87; 95% CI: 1.89–0.16, Z = 1.65, P = 0.10). Among the included studies Loebel et al. 2016^[11] and Marcus et al. 2009^[12] studies showed the high risk of bias in the D3-domain for missing outcome data, as they did not mention posttrial mean (SE) values for the groups [Figure 4]. However, both showed low risk of bias in the overall quality assessment.

There was an evidence of modest heterogeneity between trials that did not reach statistical significance between trials [Figure 5] (Chi-square test for heterogeneity = 6.65, df = 3 [P = 0.08], I² = 55%).

The risk of publication bias of studies included in the meta-analysis was not assessed, as the there were <10 studies. However, we have made the funnel plot for available studies [Supplementary Material – Funnel Plots https://links.lww.com/JOPS/A4]. A sensitivity analysis was carried out by removing one study at a time and its effect on the effect size and heterogeneity. This did not affect the results for the ABC-Stereotypy-based scores.

However, the removal of Loebel et al.’s study in 2016^[11] reduced the heterogeneity among the studies [Figure 6] and resulted in a significant reduction in stereotypy symptoms as measured on the CY-BOCS-Compulsions subscale.

A subgroup analysis was performed based on the type of antipsychotic used in the trial. Two trials using risperidone^[13,16] yielded a greater mean difference in the stereotypy on the ABC-Stereotypy subscale with interstudy heterogeneity to nil, thus providing robust evidence for the efficacy of risperidone in reducing stereotypy in autism [Figure 7].

DISCUSSION

Our objective of this systematic review of RCTs was to reveal antipsychotic efficacy in reducing stereotypy in autism. Because of the complexity of interpreting the phenomenology, researchers assessed repetitive behaviors using the ABC-Stereotypy subscale and CY-BOCS-Compulsions subscale. After screening journal-published articles in English from two databases and citation searches, nine studies were included in the systematic review. Our systematic review showed that there is an overall beneficial effect of antipsychotics in reducing stereotypy in autism compared to placebo. Meta-analysis helped us to quantify that antipsychotic significantly reduced stereotypy over placebo.

There are three reviews in the literature addressing the same issue as our study. Mulligan et al. 2014^[3] was not a systematic review and did not include any studies using antipsychotic agents in their review. Yu et al. in 2020^[4] reported 14 RCTs of pharmacological agents in treating RRB in autism, and among these, only three RCTs used antipsychotics. Investigators did not report effect estimates for antipsychotics, just mentioning that two trials were favoring antipsychotics. We have overcome the deficiency in the previous reviews in that we did a systematic review and also provided an effect size estimate for antipsychotic efficacy in reducing stereotypy in autism. Zhou et al. in 2021^[5] was the most recent systematic review and meta-analysis as ours; they expanded their search to include the Chinese language publications. They found seven studies (n = 651) that used antipsychotics with overall results favoring antipsychotics (SMD: 0.28 [0.08, 0.48]) in reducing repetitive behaviors. Our systematic review was restricted to English trials, and nine studies were found eligible for systematic review, and six studies were eligible to enter into meta-analysis (n = 726). Two forest plots were generated separately based on the outcome measure used and subgroup analyses based on antipsychotic were also done. Thus, our results are much more robust and explicit than those reported by Zhou et al. 2021.^[5]

In the meta-analysis, six eligible studies of low risk of bias were analyzed using the random-effects model, which showed a significant reduction in stereotypy scores in the ABC-stereotype subscale. In the meta-analysis, four eligible studies, all with low risk of bias, were analyzed using the random-effects model and showed a nonsignificant reduction in stereotypy scores as measured on the CY-BOCS-Compulsions subscale.

Antipsychotics used in the included studies were aripiprazole, lurasidone, and risperidone. Combining only risperidone trials resulted in a decrease in interstudy heterogeneity and an improvement in the efficacy in reducing the stereotypy scores. We noted that risperidone trials^[13,16] yielded better efficacy estimates for risperidone than other antipsychotics.

The study by Loebel et al. in 2016^[11] used lurasidone at two strengths (20 mg/day and 60 mg/day), and the removal of this study from meta-analysis resulted in decreased heterogeneity and improved significant scores in the stereotypy as measured on the CY-BOCS-Compulsions scale.

Major limitations

We limited our search strategy to the English literature, whereas Zhou et al.’s study^[5] reported meta-analysis with a larger pooled sample of antipsychotic trials (7 trials, n = 651). Zhou et al.’s study expanded its search strategy to include publications in Chinese and measured RRBs in autism rather than limiting to stereotypy, mentioning the effect size as SMD. We refined our methodology using outcomes measured by different subscales for stereotypy in ASD. We did not list adverse events during the trials, though none of the studies reported serious adverse events. Our study shows that antipsychotics are useful in reducing stereotypy, and risperidone shows consistently favorable results in the trials. As per the concern of adverse events, we suggest that clinicians use a generic dictum of starting low dose and increasing slowly, and we suggest monitoring stereotypy using a standard scale. In our review, we noticed other than ABC or CY-BOCS subscales, studies used subscales of different tools to assess stereotypy in autism: Autism Diagnostic Interview-Revised, Ritvo-Freeman Real Life Rating Scale, and Nisonger Child Behavior Rating Form. We had to exclude such studies. Other limitations are restricting two databases, not including ongoing trials, and restricting only to the English language.

The study’s strengths were the selection of at least two standard databases, the newer improved RoB 2, and the inclusion of studies in the meta-analysis that were of good quality RCTs. As we included studies with wide range of age, from children to adults, our findings are applicable to individuals with ASD having troublesome stereotypy irrespective of their age. Minor amendments were made to the PROSPERO registered protocol, which is mentioned later in this manuscript.

Implications and future research

The above evidence suggests that antipsychotic medications (such as risperidone, aripiprazole, and lurasidone) effectively reduce stereotypy and may be considered for use in treating autism. This is applicable to all age individuals with ASD having troublesome stereotypy. In the future, pragmatic trials comparing different antipsychotics and a placebo should be conducted. In addition, the efficacy of antipsychotics in treating autism with different comorbidities should be examined, and a standardized scale for measuring stereotypy in autism should be used to quantify results.

Deviations from protocol

We reduced the number of databases to two, limiting to PubMed and Cochrane databases, as these are sufficient to retrieve most of the studies under review. The gray literature search, which was mentioned, was deleted as we were committed to including only journal-published articles. We changed the data analysis part to include only continuous data and measure effect size as a change in mean and SD.

CONCLUSION

All trials included in our systematic review excluded participants with active seizures and severe intellectual impairment. Notably, none of the trials permitted coadministration of psychotropics or combined nonpharmacological or behavioral interventions. Consequently, the external validity of the findings is deemed acceptable. Clinicians can, therefore, approach the initiation of antipsychotic treatment, particularly with risperidone, aripiprazole, and lurasidone, for managing severe and distressing stereotypy in individuals with autism, with enhanced confidence. It is noteworthy that among the antipsychotic trials, risperidone demonstrated superior efficacy in ameliorating stereotypy scores in persons with autism.

Contributors

SK and NV developed the search strategy, and they searched and screened the references and retrieved data from the full text. SS acted as arbitrator in case of conflict. PG provided statistical assistance. SK wrote the first draft of the report with input from NV. All authors had full access to all the data in the study and had contributed to the drafting of the manuscript and reviewed the manuscript before submitting for publication.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

Supplementary Material

Supplementary materials available online.

Acknowledgments

We thank the Indian Psychiatric Society-Karnataka Chapter (IPS-KC) and Department of Psychiatry, NIMHANS, Bengaluru, for the training and support.

REFERENCES

Supplemental Digital Content

• JOPS_04_01_2024_11_19_KATTIMANI_2224_SDC224.pdf; [PDF] (615 KB)