OBJECTIVES

To review the evidence on first- and second-generation antipsychotics (FGAs and SGAs) for the treatment of various psychiatric and behavioral conditions in children, adolescents, and young adults (ages ≤ 24 years).

DATA SOURCES

Eight electronic databases, gray literature, trial registries, and reference lists.

METHODS

Two reviewers conducted study selection and risk of bias assessment independently, and resolved discrepancies by consensus. One reviewer extracted and a second verified the data. We conducted meta-analyses when appropriate and network meta-analysis across conditions for changes to body composition. We rated strength of evidence for prespecified outcomes.

RESULTS

One hundred thirty-five studies (95 trials and 40 observational studies) were included. None of the evidence was rated as high strength of evidence; results having moderate strength (i.e., probably an accurate effect) are presented (with n studies) below. SCHIZOPHRENIA AND RELATED PSYCHOSES (N = 39): Compared with placebo, SGAs as a class probably increase response rates, decrease slightly (not clinically significant for many patients) negative and positive symptoms, and improve slightly global impressions of improvement, severity, and functioning. There is likely little or no difference between high and low doses of quetiapine for clinical severity and functioning. Many outcomes for individual drug comparisons were of low or insufficient strength of evidence. BIPOLAR DISORDER (N = 19): Compared with placebo, SGAs probably decrease mania, decrease depression symptoms slightly, and improve symptom severity and global functioning to a small extent. SGAs (and aripiprazole alone) probably increase response and remission rates versus placebo for manic/mixed phases. Quetiapine likely makes little or no difference in depression. AUTISM SPECTRUM DISORDERS (N = 23): Compared with placebo, SGAs as a class probably decrease irritability, and decrease slightly lethargy/social withdrawal, stereotypy, and inappropriate speech; they likely increase response rates and (slightly) clinical severity. It is likely that aripiprazole and risperidone decrease irritability. ATTENTION DEFICIT HYPERACVTIVITY DISORDER (ADHD) AND DISRUPTIVE, IMPULSE-CONTROL, AND CONDUCT DISORDERS (N = 13): Compared with placebo, SGAs as a class (and risperidone individually) probably decrease conduct problems and aggression. Risperidone alone likely decreases hyperactivity in children with a primary diagnosis of conduct disorders or with ADHD but not responding to stimulants. OTHER CONDITIONS: All outcomes had low or insufficient strength of evidence for tic disorders (n = 12), obsessive-compulsive disorder (n = 1), depression (n = 1), eating disorders (n = 3), and behavioral issues (n = 2). HARMS ACROSS CONDITIONS: From network meta-analysis, olanzapine was more harmful for gains in weight and body mass index (BMI) than other SGAs except for clozapine; results were most robust for relative harm over aripiprazole, quetiapine, and risperidone, and most applicable to the short term. Findings from pairwise meta-analysis between different SGAs were similar, except for showing longer term benefit for quetiapine and risperidone versus olanzapine, and little or no short-term differences between risperidone and quetiapine, or between different doses of aripiprazole, asenapine, or quetiapine. FGAs probably cause slightly less harm for weight and BMI compared with SGAs. There is probably little or no difference in risk for somnolence between different doses of asenapine or quetiapine. There is likely little or no difference in risk for mortality or prolonged QT interval in the short term for SGAs as a class. SGAs versus placebo/no treatment probably increase short-term risk for high triglyceride levels, extrapyramidal symptoms, sedation, and somnolence.

CONCLUSION

SGAs probably improve to some extent key intermediate outcomes for which they are usually prescribed, but they have a poorer harms profile than placebo or no antipsychotic treatment, particularly for body composition and somnolence. Data for head-to-head comparisons within and between classes were generally limited and rated as insufficient or low strength of evidence. Evidence was sparse for patient-important outcomes (e.g., health-related quality of life) and outcomes for young children (<8 years). Key priorities for research are long-term comparative effectiveness and development of systems for monitoring harms.