To evaluate the comparative effectiveness and safety of medical interventions (defined broadly as interventions involving the administration of external substances to the body or use of external nonbehavioral procedures to treat symptoms of autism spectrum disorder [ASD]) for children with ASD.
We searched MEDLINE, Embase, the Cumulative Index of Nursing and Allied Health Literature, and PsycINFO from January 2010 through September 2016.
We included comparative studies of medical interventions that included at least 10 children with ASD. Two investigators independently screened studies and rated risk of bias. We extracted and summarized data qualitatively given significant heterogeneity. We also assessed strength of the evidence (SOE) and considered cumulative data from eligible studies included in our 2011 review of medical therapies and newly published studies.
The 76 unique comparative studies (including 12 comparative studies addressed in the 2011 review) meeting our criteria included 72 randomized controlled trials (RCTs), 2 nonrandomized trials, and 2 retrospective cohort studies. Thirty-nine studies had low, 29 had moderate, and 8 had high risk of bias. Populations, treatment approaches, and outcomes assessed varied across studies. Relative to placebo, seven studies addressing risperidone or aripiprazole reported statistically significant improvements in challenging behavior in the short term (<6 months) but also clinically significant harms. Longer term effectiveness was reported in uncontrolled extensions. Three studies comparing risperidone and aripiprazole reported few significant differences in effects on weight gain between agents. RCTs addressing methylphenidate (n=2), atomoxetine (n=2), and guanfacine (n=1) reported significant improvements in hyperactivity, with frequent harms. Omega-3 fatty acids (4 RCTs) were not associated with changes in challenging behavior. N-acetylcysteine and tetrahydrobiopterin were not associated with improvements in social skills and symptom severity, respectively. Despite the number of RCTs with low or moderate risk of bias addressing nutritional supplements or specialized diets, evidence is insufficient for all clinical efficacy and harms outcomes because few, small studies addressed each diet or supplement. Similarly, although 14 RCTs with low or moderate risk of bias compared risperidone plus an adjunct medication with risperidone plus placebo, few addressed the same adjunct agents. Studies of hyperbaric oxygen therapy versus sham treatment using differing protocols reported conflicting results. Fourteen studies addressed other interventions, most evaluated in only one study, and typically reported some positive treatment effects on sleep, ASD symptoms, or language.
Risperidone and aripiprazole ameliorated challenging behaviors in the short term, but with clinically significant side effects (high SOE). Methylphenidate and atomoxetine were also associated with improvements in hyperactivity in small short-term RCTs (low SOE), with improvements maintained over 6 months for atomoxetine (low SOE for longer term effects). Methylphenidate was associated with clinically significant harms (low SOE), while atomoxetine was associated with clinically moderate harms (low SOE). Omega-3 fatty acid supplementation, N-acetylcysteine, and tetrahydrobiopterin failed to show benefits (low SOE). Evidence for other interventions and outcomes studied was insufficient. While the conduct of studies has improved considerably over time (i.e., growing number of RCTs and use of standardized measures), data on longer term (≥6 months) results and harms of most interventions are lacking. Similarly, more research is needed to understand characteristics of the child or treatment that modify outcomes, whether effectiveness of interventions generalizes across different settings such as the home or school, and how components of interventions may drive effects.
