Meta-Analysis of 397 Pediatric Mental Health Trials Finds Women, Minoritized Groups Systematically Under-Represented
核心洞察
A systematic review of 397 randomized controlled trials in child and adolescent mental health found female participants made up only 28.3% of the 46,989 enrolled participants.
White participants accounted for 75.0% of trial samples, while Asian participants represented just 2.0% and only 53.7% of trials reported any race data.
Over half of trials excluded autistic participants and 43% excluded those with intellectual disability (搜索), while 76% of depression trials excluded youth at suicide risk.
Randomized controlled trials (RCTs) of pharmacological and nutraceutical interventions for children and adolescents with mental disorders enroll populations that poorly reflect the patients most affected by these conditions, according to a systematic review and meta-analysis of 397 trials published in Nature Mental Health.
The analysis pooled data from 46,989 participants (mean age 10.2 ± 3.1 years) across 49 countries, drawing on 22,446 deduplicated citations screened from electronic databases and 18 additional records identified through manual search of prior meta-analyses. Trials were conducted in North America (231; 58.2%, of which 218 were in the USA), Asia (66; 16.6%), Europe (58; 14.6%), Oceania (17; 4.3%), South America (2; 0.5%) and across multiple continents (23; 5.8%). Risk of bias was rated low in 169 studies (42.6%), with 170 (42.8%) presenting some concerns and 58 (14.7%) rated high risk. None of the included studies reported incorporating patient and public involvement (PPI) in their design or conduct.
Female Representation Falls Short of Disease Epidemiology
Sex was reported in 97.0% of included studies, but the pooled proportion of female participants was only 28.3% (k = 385; 95% CI 26.1–30.7), with high heterogeneity (Q = 4,711.8; I² = 91.9%; P < 0.01). Representation varied sharply by disorder, ranging from 18.1% in autism spectrum disorder (搜索) (ASD; k = 135; 95% CI 15.4–21.2) and 24.3% in attention deficit hyperactivity disorder (搜索) (ADHD; k = 112; 95% CI 21.7–27.0) to 100% in eating disorders (搜索) (k = 4; 95% CI 0.0–100.0). Only trials in eating and depressive disorders (56.5%; 95% CI 49.6–63.2) enrolled a proportion of female participants equal to or exceeding that of males.
The authors note that while male predominance may partly reflect the epidemiology of neurodevelopmental disorders, recent large meta-analyses estimate that women account for around 25% of autistic individuals and 33% of individuals with ADHD—figures still notably higher than the female representation observed in the pooled analysis. They further point out that male predominance is not observed in other disorders: bipolar disorder (搜索) shows similar sex ratios, obsessive–compulsive disorder (OCD) has a higher prevalence in females (odds ratio 1.654), and most anxiety disorders (搜索) present markedly higher risk among girls, with odds ratios ranging from 2.8 to 3.5.
Sensitivity analyses found no significant differences based on risk of bias (Q = 1.14; P = 0.77). Multicenter trials (k = 187; 30.7%; 95% CI 27.9–33.7) included more females than unicenter trials (k = 198; 25.8%; 95% CI 22.4–29.5). By continent, multicontinental trials reported the highest female representation (k = 22; 42.5%; 95% CI 35.3–50.1) and Asian trials the lowest (k = 64; 23.2%; 95% CI 18.7–28.3). Egger's test detected publication bias for most studied disorders, with studies including a higher proportion of female participants more likely to remain unpublished (P < 0.01).
Pooled analyses showed no significant change in female proportion over the past five decades (β = 0.19; P = 0.11), although a significant increase was observed for ADHD (β = 0.50; s.e.m. 0.19; P < 0.01).
Race Reporting Remains Sparse and Skewed
Only 53.7% of trials reported data on racial representation. The percentage of white participants was reported in 208 trials (52.4%), Black participants in 163 (41.1%) and Asian participants in 130 (32.7%). Hispanic representation was reported in 105 trials (26.4%); of these, 41% treated Hispanic as an ethnicity (nonexclusive with race) while 59% classified it as a race. Reporting rates did not change significantly over time (t = −0.67; P = 0.50), and only 6.30% of included trials reported information on multiracial participants.
Among trials reporting race, the pooled proportion of white participants was 75.0% (k = 208; 95% CI 71.4–78.3), ranging from 65.6% in schizophrenia (搜索) trials (k = 14; 95% CI 61.4–69.5) to 98.9% in primary insomnia (搜索) RCTs (k = 2; 95% CI 1.00–100). Black participants accounted for 10.1% (k = 162; 95% CI 8.5–11.9) and Asian participants 2.0% (k = 130; 95% CI 1.3–3.2). Hispanic participants represented 10.7% when reported as an ethnicity (k = 43; 95% CI 8.0–14.3) and 6.2% when reported as a race (k = 62; 95% CI 4.6–8.2).
Comparing US-based trials (58% of the sample) against the 2024 American Community Survey, the authors found white participants constituted 69.5% of trial samples versus 59.8% of the US population, Asian participants 2.0% versus 6.3%, and Hispanic participants 8.4% versus 19.4%. Sensitivity analyses revealed a higher proportion of white participants in studies rated at high risk of bias compared with low risk (86.9% versus 72.7%; P = 0.02). Temporal analyses showed no variation over time in the proportion of white, Black and Hispanic participants (all P > 0.05), while the proportion of Asian participants increased (β = 0.53; s.e.m. 0.27; P = 0.04).
Systematic Exclusion of High-Risk and Comorbid Populations
Excluding trials focused on children and adolescents with a primary ASD diagnosis, 56.5% of trials excluded participants with ASD or pervasive developmental disorder—a proportion rising to 68.0% in RCTs targeting depressive disorders and 87.5% in OCD studies. This systematic exclusion of participants with comorbid autism has become increasingly common over time (t = −2.57; P = 0.01), without a corresponding increase in trials specifically focusing on autistic individuals.
Regarding intellectual disability (搜索) (ID), 44.3% of trials systematically excluded participants with intellectual impairment; among these, 43.6% did not specify an IQ threshold. Among those that did, 34.3% excluded participants with IQ <80, 50% with IQ <70 and 15.7% with IQ <60 or lower. Exclusion exceeded 60% in trials on bipolar disorder (搜索), schizophrenia (搜索), insomnia (搜索) and OCD, compared with only 24.8% of ASD RCTs, with no significant change over time (t = −0.65; P = 0.51).
Suicide risk was an exclusion criterion in 26.2% of trials overall, but was most frequent in RCTs on depression and anxiety, where 76.0% and 59.1% of trials, respectively, excluded participants at risk. This practice has become more common over time (t = −2.37; P = 0.01). The authors highlight that depressive disorders are a key risk factor for suicide in adolescents and that suicidal ideation is present in over 50% of adolescents with depression, while National Institute for Health and Care Excellence guidelines recommend pharmacological intervention for young people with moderate-to-severe depression only—the group most likely to present suicidality.
Geographic Concentration and Evidence Gaps
Very few trials were conducted in low- and middle-income countries, and entire continents such as Africa were not represented at all among eligible studies, while South America contributed only a handful. The authors state that this lack of geographical diversity casts serious doubt on the external validity of the available evidence, since findings derive predominantly from high-income settings that differ markedly from low- and middle-income countries in health system capacity, cultural context and sociodemographic composition.
The review also notes that individuals with neurodevelopmental disorders are more likely to be prescribed medications off label—about 45.7% receive at least one medication—yet are less likely to be included in the evidence base guiding such treatment decisions. Similarly, the authors cite a US Government Accountability Office audit reporting that 80% of drugs withdrawn from the market had more adverse events in women than in men, largely owing to insufficient testing in female populations, and note documented sex differences in psychopharmacology pharmacokinetics, pharmacogenetics and pharmacodynamics.
Limitations and Recommended Actions
The authors acknowledge several limitations. Most outcomes examined showed substantial heterogeneity that could not be fully accounted for by subgroup analyses, so pooled estimates should be interpreted as average effects across diverse study contexts. Residual confounding is likely given the range of disorders, geographies and trial designs included. The analyses were restricted to RCTs of pharmacological and nutraceutical interventions, excluding psychotherapeutic or psychosocial approaches, so findings may not generalize to all trial types. Reporting practices for race and ethnicity varied widely, and definitions were inconsistent—particularly for Hispanic populations.
The review concludes that female participants, certain minoritized racial and ethnic groups, and other high-risk populations are consistently represented in proportions that do not reflect the populations affected by the studied disorders, undermining external validity. The authors recommend that researchers actively recruit representative samples, minimize systematic exclusion criteria, set explicit recruitment targets by sex and racial representation, and provide culturally appropriate materials. They further call on funders and regulators to increase investment in trials conducted in low- and middle-income countries and to mandate transparent reporting of age, sex, race and ethnicity, on journals to enforce reporting standards, and on trialists to integrate PPI into trial design and execution.
The study protocol was registered on PROSPERO (CRD42024629137) and reported in accordance with PRISMA guidelines.