How Many Anxious Kids in Community Mental Health Would Be Eligible for an RCT? And Does It Matter? Insights from a Naturalistic Sample and a Non-Systematic Review

Highlights

What are the main findings?

• Among youth seeking specialty anxiety treatment in a community mental health setting, the majority would have been ineligible for at least one published randomized controlled trial of cognitive behavioral therapy for pediatric anxiety and related disorders.
• Ineligible youth required nearly twice as many treatment sessions and were more than twice as likely to receive case management services.

What is the implication of the main finding?

• These results do not imply cognitive behavioral therapy is less effective for clinically complex youth. Rather, they suggest that youth commonly excluded from efficacy trials may require more flexible or prolonged care than is typically represented in tightly controlled RCT protocols.
• Trial exclusion status may potentially serve as a prognostic indicator in community settings, helping clinicians set realistic expectations for families and signal when treatment adaptation or added supports are warranted.

Abstract

Background: Decades of randomized controlled trials (RCTs) support cognitive behavioral therapy (CBT) for pediatric anxiety, but exclusion criteria may limit generalizability to routine settings. We examined common exclusion criteria in recent CBT RCTs for pediatric anxiety, trends in these criteria over time, and whether meeting RCT exclusion criteria affects outcomes in a naturalistic sample. Methods: We reviewed 81 RCTs from the past 25 years assessing CBT for pediatric anxiety or related disorders to identify common exclusion criteria. We examined how often youth seeking exposure-based treatment for anxiety or OCD at an urban community health center met these exclusion criteria and whether this impacted treatment response, using three-year retrospective chart review data (n = 94). Results: Common exclusion criteria in identified RCTs included psychotropic medication use (66.7%), autism spectrum disorder (63.0%), and other psychiatric comorbidities. Suicidal ideation increased as an exclusion criterion over time (p < 0.05, Cramér’s V = 0.23). Based on these criteria, 53% of participants in our naturalistic sample would have been excluded from one or more RCTs. Excluded patients did not differ in baseline characteristics. Excluded youth required nearly twice as many treatment sessions and had more than double the rate of case management utilization (all ps < 0.01). Conclusions: Youth who would have been excluded from at least one RCT had poorer prognoses. Findings support continued emphasis on pragmatic trials to advance understanding of how to augment treatments to better meet the diverse needs of youth.

1. Introduction

Cognitive behavioral therapy (CBT) is the most widely used and well-supported intervention for treating anxiety and related disorders in children [1]. Decades of research and numerous randomized controlled trials (RCTs) have established CBT’s efficacy and effectiveness, particularly for generalized anxiety disorder (GAD), social anxiety disorder (SAD), obsessive–compulsive disorder (OCD), panic disorder, and specific phobias [2,3,4,5]. As a result, CBT is considered a front-line treatment for youth suffering from anxiety and related disorders [6]. Across studies, CBT has outperformed wait-list controls, treatment as usual, and attention controls [7,8]. Efficacy trials have also shown that CBT-based interventions are superior to alternative methods such as relaxation, bibliotherapy, family-based support, and emotional disclosure in the treatment of childhood anxiety disorders [9,10,11,12,13].

RCTs are generally considered the most rigorous level of scientific evidence we have to evaluate the efficacy and effectiveness of psychotherapy [14]. Common to RCTs is the use of exclusion criteria, which are routinely employed in clinical research trials to isolate the effects of an evidence-based treatment (EBT) on target mechanisms of interest and minimize confounding variables, serving as a necessary component of rigorous science. For example, CBT efficacy trials for anxiety disorders may exclude certain comorbidities (e.g., depression, Attention-Deficit/Hyperactivity Disorder [ADHD]) or historical factors (e.g., past treatment history, certain medications) that would complicate evaluation of whether treatment was having the intended effects [15,16].

Despite the demonstrated efficacy of CBT for pediatric anxiety, there are concerns about how tight exclusion criteria commonly used in randomized controlled trials (RCTs) may limit the applicability of research findings to youth served in community mental health settings [17,18]. For example, Weisz et al. [19] found that across a multi-decade review of pediatric anxiety RCT samples, only 2.1% of patients had been clinically referred clients, who typically show higher symptom severity, greater comorbidity, and more complex treatment needs; these characteristics may contribute to the underrepresentation of clinically complex patients in RCTs [20,21]. Additionally, a prior study found that 60.8% of youth in a community mental health setting were actively using psychotropic medication, a cause for exclusion from many RCTs seeking to isolate the effect of CBT [22]. Many participants in community settings also frequently present with multiple comorbid diagnoses [23,24]. Taken together, this suggests that, perhaps in part due to exclusion criteria, the clinical profiles of treatment-seeking youth may contain greater complexity than those of youth who are treated within RCTs. To our knowledge, no studies to date have examined the trends in exclusion criteria for RCTs of CBT. While it is established that a substantial proportion of clinically complex youth would be excluded from efficacy trials, it remains unknown whether the scope of these exclusions has expanded or narrowed over time, which specific criteria are driving any change, and the implications this has for the applicability of CBT findings to real-world clinical populations.

Closely related to the issue of exclusion criteria is that of sample representativeness with respect to the sociodemographic and cultural background of treated individuals. Most early CBT efficacy research relied on largely socioeconomically and demographically homogeneous samples in Europe and the United States, with limited representation of racially and ethnically diverse and economically disadvantaged individuals [25]. In other disciplines, tight exclusion criteria also can lead to disproportionately low representation of individuals of color in clinical trials [26]; limited representation of communities of color within RCTs is also exacerbated by limited targeted outreach, physical inaccessibility of research institutions (where many trials take place), and historic injustices such as the Tuskegee Syphilis Study that leads to understandable hesitancy among communities of color to participate in clinical research [27,28]. Multiple studies similarly demonstrate that youth treated in routine clinical settings differ in clinical presentation and racial, ethnic, and socioeconomic backgrounds from those enrolled in research trials [23,24]. Collectively, these patterns highlight a significant gap in the literature: relatively little is known from gold-standard RCT clinical trials about how CBT functions for clinically complex youth and those from socioeconomically, racially, and ethnically diverse backgrounds.

When CBT has been tested outside the controlled conditions of efficacy trials, the research for its effectiveness for pediatric anxiety and related disorders is less consistent. Recent Cochrane reviews have questioned whether community-delivered CBT offers significant advantages over psychoeducation, bibliotherapy, treatment-as-usual, or other supportive interventions for pediatric anxiety [29]. Similarly, studies comparing CBT with selective serotonin reuptake inhibitors (SSRIs) in real-world settings show no clear superiority [30,31]. This pattern persists across treatment modalities, where school and community-based CBT programs often yield outcomes comparable to treatment-as-usual [8,32]. As study design moves from tightly controlled designs to more pragmatic trials, treatment effects tend to diminish, a phenomenon described as the “implementation cliff” [19]. The gap between CBT’s strong efficacy in RCTs and inconsistent effectiveness findings has prompted questions about how demographic and clinical differences between trial participants and community populations may affect the implementation of CBT. Much effort has been made in the past few decades to examine how CBT protocols can be best applied in routine clinical settings to meet the diverse needs of youth who may not fit the tight mold of an efficacy RCT. For example, modular treatment designs that allow clinicians to match treatment techniques to the presenting problems of youth with complex presentations can increase the effectiveness of CBT in routine clinical settings [33,34].

That said, persistent concerns remain about the “transportability” of RCT findings to routine practice, especially for anxiety-related disorders [35,36]. Despite growing recognition and study of the “implementation cliff” in youth anxiety treatment [19], no published study has examined how exclusion criteria contribute to implementation challenges, nor whether the clinical and demographic disparities between RCT samples and community populations, a partial consequence of study exclusion criteria, influence treatment effectiveness. This study addresses this gap directly, contributing to an emerging body of scholarship aimed at bridging the science-to-practice divide in evidence-based mental health care for youth. This study pursues three primary, exploratory objectives: (1) identify the exclusion criteria employed in recent RCTs of CBT for pediatric anxiety, (2) examine how the use of these criteria has evolved over time, and (3) examine the impact of youth meeting exclusion criteria on treatment outcomes. This final objective included two related research questions: (a) What proportion of treatment-seeking youth in a naturalistic treatment sample would have been excluded from published efficacy RCTs? and (b) were excluded youth distinguishable from eligible youth in terms of baseline clinical presentation and treatment outcomes? For this final aim, we leveraged data from a naturalistic sample of racially and economically diverse youth receiving exposure-based CBT (EX-CBT) at a specialty anxiety treatment program within an urban community mental health center in the Northeastern United States.

2. Method

2.1. Exclusion Criteria in Recent CBT RCTs

We conducted a narrative review of RCTs evaluating the efficacy of CBT for pediatric anxiety and related disorders, irrespective of study design, published between 2000 and 2024. A search was conducted on PubMed using the Advanced Search Builder. The terms “pediatric,” “youth,” “cognitive behavioral therapy,” and “meta-analysis” were combined using the AND operator across all fields, with “anxiety” OR “OCD” included to capture either condition. Results were filtered to studies published within the last decade. The final query was: (pediatric) AND (youth) AND (cognitive behavioral therapy) AND (meta-analysis) AND (anxiety OR OCD). The search yielded 40 articles, of which 30 were excluded because they did not focus on pediatric anxiety or a related disorder, did not employ a meta-analytic design, did not include randomized controlled trials (RCTs), or were duplicates. This resulted in a final sample of 10 meta-analyses. Individual RCTs were extracted through backward citation chaining of the included meta-analyses. Unlike a systematic review, this search was confined to a single database (PubMed) and was conducted without a pre-registered protocol.

RCTs were eligible for inclusion in the literature review if they included participants aged 18 years or younger who were seeking treatment for a primary anxiety disorder or school-related anxiety. RCTs also had to include an intervention group testing some type of individualized CBT intervention that compared CBT to a waitlist control, treatment-as-usual, or another psychological treatment. Group-based interventions were excluded to maintain comparability with our naturalistic data comparison sample. Trials allowing psychotropic medication use were included only when individual-based CBT was evaluated in combination with medication (e.g., SSRIs + CBT). Studies comparing CBT solely to medication, or those that required all participants to take medication, were excluded. Effectiveness trials were also excluded to focus specifically on exclusion criteria used in tightly controlled efficacy studies, where originally developed treatment protocols are most often tested [37]. We included eligible trials if they met both the population and CBT intervention-comparison inclusion criteria. Studies from any country and in any language were eligible, provided they satisfied these criteria. Those not written in English were read via web-based translators. The lead author (AE) reviewed all studies to confirm study eligibility. Co-author (MB) independently reviewed 10% of identified studies to confirm eligibility with 100% agreement. In total, 81 studies met the inclusion criteria (see Figure 1).

Appendix A includes an overview of the characteristics of the reviewed RCTs. All sourced meta-analyses were published after 2015, with more than half of the included RCTs dating from 2012 or later. Most included RCTs were conducted in the United States (n = 42; 51.85%), followed by Europe (n = 27; 33.33%) and Australia (n = 10; 12.35%). The sample sizes within the RCTs ranged from 10 to 488 participants (median = 55), totaling 6405 individuals across the 81 studies. The age ranges of recruited samples across CBT studies were highly heterogeneous and specific to each study; this sample included children aged 3–18 years. Similarly, the included diagnoses varied greatly between studies, with the most frequent being social anxiety disorder (35.8%), specific phobia (29.6%), obsessive–compulsive disorder (28.4%), generalized anxiety disorder (24.7%), or simply any primary pediatric anxiety disorder (23.5%). CBT was delivered with family involvement in over one-quarter of the RCTs reviewed (n = 21).

RCT exclusion criteria varied significantly between studies. Criteria occurring less than four times (e.g., pregnancy, history of autoimmune disease) or described too vaguely to consistently track (e.g., generally poor physical health, parent suffering from severe psychiatric disorder) were not included in our analysis of RCT criteria. We identified six criteria that were employed in at least 10% of reviewed RCTs. These common RCT exclusion criteria included current psychotropic medication use, comorbid autism spectrum disorder (ASD) diagnosis, comorbid psychosis, suicidal ideation, comorbid major depressive disorder (MDD) diagnosis, and comorbid ADHD diagnosis. Each study was coded to include binary variables (‘yes’ or ‘no’) corresponding to the presence of each of these six exclusion criteria in each trial. These were summed descriptively.

2.2. Trends in Exclusion Criteria over Time

Statistical Analyses

RCT data were subset into three-decade groups: 2000–2009 (n = 22), 2010–2019 (n = 52), and 2020–2024 (n = 7). Because the 2020–2024 decade group had a small sample size, we used Freeman-Halton extensions of Fisher’s exact test within a 3 × 2 design to explore whether the prevalence of each exclusion criterion varied significantly across decades. Each Fisher’s exact test was evaluated at a conventional threshold of α = 0.05. Effect sizes were calculated using Cramér’s V.

2.3. Comparison to a Naturalistic Treatment Sample

We performed a retrospective, observational study of naturalistic data collected during routine clinical care. No experimental manipulation or randomization was conducted, and this investigation was not a clinical trial. This analysis is reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines [38]. A completed checklist is provided in the Supplementary Materials.

The naturalistic sample we performed secondary analysis on was a three-year (2018 to 2021), chart review of consecutively treated youth in an exposure-based CBT treatment center embedded in a community mental health setting (n = 94). Treated diagnoses included DSM-5 recognized anxiety disorders, as well as OCD, trichotillomania, and tic disorders. Within this treatment setting, there are no comorbidity exclusions for program admittance other than active suicidality or homicidal ideation. Treatment is generally performed by master’s or doctorate-level clinicians under the close supervision of experts in CBT for OCD and anxiety.

Chart data included intake notes, session notes, and discharge summaries, all of which were deidentified prior to analysis. Trained coders abstracted baseline demographic (age, gender, racial identity, Medicaid status, and whether parents spoke English) and clinical characteristics (diagnoses, parental mental health history, and presence of suicidal ideation) from intake notes, treatment status from discharge summaries, and coded treatment techniques delivered in deidentified session notes (n = 2633 coded notes). Treatment response was operationalized in two ways: (1) full termination of treatment (i.e., conclusion of specialty anxiety treatment and graduation to no services), or (2) termination of specialty treatment (i.e., youth who completed specialty anxiety treatment, and either graduated to receive no additional services or transferred to ongoing supportive care or other therapy services). Full details about study methodology and primary outcomes from this chart review are published elsewhere [39]. The youth sample had a mean age of 12.35 (SD = 4.3, range = 4–22). Patients older than 18 were included in this sample, as in community mental health settings, patients occasionally continue to access services beyond this age. Sixty-four (68.1%) identified as female, 28 (29.8%) as male, 1 as non-binary, and 1 preferring to self-identify. Fifty-two participants (55.3%) identified as White, 14 (14.9%) as Black, 10 (10.6%) as Asian, 7 (7.4%) as Mixed Race, and 10 (10.6%) as Hispanic/Latine and White.

The study team extracted chart-documented variables on baseline characteristics, treatment-process events, and treatment outcomes. Baseline characteristic variables included gender, ethnicity, and Medicaid coverage. Treatment process variables reflected the resources utilized throughout care, including the proportion of sessions using case management services, the proportion of sessions using exposure techniques, and the total number of sessions delivered before discharge. Outcomes were represented as treatment termination or treatment termination and/or transfer to supportive care. Multiple variables extracted corresponded to common RCT exclusion criteria documented in our review, including suicidal ideation, ASD diagnosis, MDD diagnosis, medication use, ADHD diagnosis, and comorbid psychosis. All variables were drawn from the chart of each participant and coded as either ‘yes’ (diagnosis documented) or ‘no’ (diagnosis not documented). For analytic purposes, we created a binary exclusion variable, where participants were coded as ‘excluded’ if they met at least one RCT exclusion criterion (n = 50). Participants not meeting any of the identified exclusion criteria were coded as ‘not excluded’ (n = 44).

2.4. Statistical Analyses

Naturalistic sample data, including baseline characteristics, RCT exclusion status, and treatment process and outcome variables, were analyzed in IBM SPSS Statistics (Version 31). To compare dichotomous baseline characteristics between participants who would versus would not have been excluded from RCTs, we used χ² tests of association. To evaluate differences in continuous treatment-process variables, we used descriptive statistics and conducted independent-samples t-tests between the RCT exclusion groups. We calculated effect sizes for these comparisons using Cohen’s d, providing a standardized measure of the observed group differences. To examine the impact of RCT exclusion status on treatment outcomes, we fit two binary logistic regression models, controlling for baseline demographic covariates. Each model included RCT exclusion status as the independent variable and either treatment termination and termination or transfer to supportive care as the dependent variable. Pre-analysis χ² tests revealed that race was significantly associated with both Medicaid status (χ² = 4.3, p = 0.038) and ethnicity (χ² = 14.2, p < 0.001), suggesting multicollinearity that could distort model estimates; race was therefore excluded as a covariate. Final covariates included ethnicity, gender, and Medicaid status. Multicollinearity among retained predictors was further evaluated using variance inflation factors (VIF) and tolerance statistics; VIFs ranged from 1.01 to 1.06 across both models, well within acceptable thresholds (VIF < 5; tolerance > 0.20). As all predictors were binary, the assumption of linearity of the logit was not applicable and was not assessed. While covariate screening used a conventional threshold of α = 0.05, a conservative threshold of α = 0.01 was applied across all primary analyses to reduce the elevated risk of type I error associated with multiple comparisons.

3. Results

3.1. Exclusion Criteria in Recent CBT RCTs

Across the literature, RCTs frequently excluded participants based on factors such as comorbid psychiatric conditions, safety concerns, and concurrent treatment use. In order of frequency, the six most common exclusion criteria were current psychotropic medication use (66.7%), ASD diagnosis (63.0%), comorbid psychosis (60.5%), suicidal ideation (45.7%), MDD diagnosis (22.2%), and ADHD diagnosis (16.1%). In studies permitting an ASD diagnosis (n = 30), comorbidity was allowed if an anxiety disorder was the primary presenting concern and CBT was delivered to target that condition. In some RCTs (n = 23), psychotropic medication use was prohibited unless participants had maintained stable dosages for several weeks or months prior to the pretreatment assessment (see

Table 1. Common Exclusion Criteria by Trial Publication Date.

Decade (n)	Suicidal Ideation (%) *	Autism (%)	Depression (%)	Medication (%) a	ADHD (%)	Comorbid Psychosis (%)
2000–2009 (22)	7 (31.8%)	14 (63.6%)	5 (22.7%)	14 (63.6%)	6 (27.3%)	14 (63.6%)
2010–2019 (52)	24 (46.2%)	32 (61.5%)	12 (23.08%)	34 (65.4%)	7 (13.5%)	33 (63.5%)
2020–2024 (7)	6 (85.7%)	5 (71.4%)	1 (14.3%)	4 (57.1%)	0 (0%)	2 (28.6%)

Note. Percentages reflect the proportion of studies for that decade which included the given criteria. ^a Trials excluded for psychotropic medication use in various ways. Most (n = 28) excluded for any medication use at all, while some (n = 23) required stable use for several weeks, medication use to be unrelated to OCD treatment (n = 2), or complete cessation for the inclusion in the trial (n = 1). * p < 0.05.

).

3.2. Trends in Exclusion Criteria over Time

When the data were examined within subsets of publication date, we found minimal change in identified exclusion criteria across decades, except for suicidal ideation. Fisher’s exact test revealed a significant association between study decade and the use of suicidal ideation as an exclusion criterion (p = 0.049, V = 0.23), with prevalence increasing across decades (2000–2009: 31.8% of studies; 2010–2019: 46.2%; 2020–2024: 85.7%). The breakdown of the common exclusion criteria by publication date is shown in Table 1.

3.3. Comparison to a Naturalistic Treatment Sample

We determined the proportion of participants in our naturalistic sample who would have met typical RCT inclusion criteria and whether RCT exclusion status was associated with certain characteristics and treatment outcomes. Racial and ethnic data were missing for one participant; all other demographic and baseline characteristics were fully reported. Data were missing for 3 participants (3%) on exposure use in sessions, case management use, and the total number of sessions attended. Analyses used available-case (listwise) deletion, resulting in varying sample sizes across models with these variables. No imputation procedures were used.

Table 2. Differences in Baseline Characteristics and Treatment Process between Excluded and Non-Excluded Participants (n = 94).

	Excluded from ≥1 RCT (n = 50)	Not Excluded (n = 44)	Test Statistic
Male Gender	12/50 (24.0%)	16/44 (36.4%)	χ2(1) = 1.17
Ethnicity (Hispanic vs. non-Hispanic)	5/50 (10.0%)	5/43 (11.6%)	χ2(1) = 0.00
Race (White vs. non-White)	28/50 (56.0%)	24/43 (55.8%)	χ2(1) = 0.00
Race (Black vs. non-Black)	9/50 (18.0%)	5/43 (11.6%)	χ2(1) = 0.32
Race (Asian vs. non-Asian)	6/50 (12.0%)	4/43 (9.3%)	χ2(1) = 0.01
Enrolled in Medicaid	40/50 (80.0%)	28/44 (63.6%)	χ2(1) = 2.37
Proportion of sessions using exposure	41% (SD = 26.8%)	42.3% (SD = 29.2%)	t(85.7) = 0.22
Proportion of sessions using case management	26.1% (SD = 26.8%)	11.8% (SD = 14.1%)	t(72.8) = −3.23 *,a
Sessions attended	37.56 (SD = 27.78)	19.28 (SD = 13.66)	t(70.1) = −4.05 **,a

Note. For χ² analyses, percentages reflect the proportion of participants within each group who possessed the given characteristic. Race and ethnicity data were missing for one participant. Due to missing data values, degrees of freedom were not constant across t-tests. ^a Significant Levene’s test for equality of variance; test statistic for equal variances not assumed reported. * p < 0.01; ** p < 0.001.

compares the baseline characteristics and treatment process variables between youth who were excluded from more than one RCT and those not excluded. We did not observe any significant differences in demographics between excluded and non-excluded youth. Youth who have been excluded from at least one RCT had a higher proportion of sessions involving case management (M = 26.1%, SD = 26.8%, t_(72.8) = −3.23, p = 0.002, d = −0.66) than those not excluded (M = 11.8%, SD = 14.1%). Excluded youth also had a significantly longer mean number of sessions attended (M = 37.56, SD = 27.78, t_(70.1) = −4.05, p < 0.001, d = −0.82) than those not excluded (M = 19.28, SD = 13.66).

Table 3. Logistic Regression Models examining the Relationship between RCT Exclusion Status and Treatment Outcomes.

	β	S.E. (β)	Wald X2(1)	Odds Ratio (95% CI)	Nagelkerke R2	p
Full treatment termination	−0.90	0.46	3.82	0.41 (0.17, 1.00)	0.17	0.051
Specialty treatment termination (either full termination or transfer to supportive care	−0.34	0.48	0.50	0.71 (0.28, 1.83)	0.07	0.478

Note. CI = confidence interval. Covariates used in these models included ethnicity, gender, and Medicaid status.

presents the results of the binary logistic regression model, with odds ratios used as an index of effect size. RCT exclusion status did not relate to the likelihood of full treatment termination (i.e., ending specialty anxiety treatment and graduating to no therapy services) although trended in the direction of exclusion status being associated with lower likelihood of successful termination, (OR = 0.41, p = 0.051), with the model showing modest fit compared to the null model (Nagelkerke R² = 0.17). Similarly, RCT exclusion status did not predict the likelihood of successfully concluding specialty anxiety treatment only, when including youth who ended treatment and may have gone on to additional non-anxiety focused therapy services (OR = 0.71, p = 0.478). This model demonstrated poor fit relative to the null model (Nagelkerke R² = 0.07).

4. Discussion

CBT is a highly efficacious treatment for pediatric anxiety and related disorders, with multiple decades of controlled trials demonstrating its superiority over waitlist and attention controls, and in some cases, other psychotherapies [7,8,9,10,11,12,13]. Yet, data about the success of CBT when delivered in community settings has yielded mixed results, which some have theorized may be partly due to greater complexity and diversity of youth treated in community settings compared to those treated in closely controlled trial settings. To our knowledge, this is the first study to specifically catalogue common exclusion criteria used in pediatric anxiety CBT efficacy trials, examine how exclusion criteria have changed over time, and evaluate how this may influence treatment outcomes in a naturalistic sample. We found that the most common exclusion criteria in recently published RCTs were psychotropic medication use, ASD diagnosis, comorbid psychosis, suicidal ideation, MDD diagnosis, and comorbid ADHD, with exclusion based on suicidal ideation having possibly increased over the past two decades. In our chart review sample, most youth would have been ineligible for at least one RCT, consistent with prior work highlighting differences between youth treated in naturalistic vs. research settings [23,24]. Moreover, youth who would have been excluded from RCTs attended significantly more treatment sessions and required a higher proportion of sessions involving case management.

Findings align with and extend prior literature documenting a substantial gap between research samples and the youth treated in community settings. In our narrative review of recent RCTs, we found that the use of exclusion criteria in efficacy trials has remained relatively stable over time, except for a possible increase in exclusion due to suicidal ideation. While this trend should be interpreted cautiously in the absence of a systematic search, it may highlight an important and growing area of divergence between research and community practice. Furthermore, the observation that most participants in our naturalistic sample would have been excluded from at least one CBT efficacy trial raises questions about whether treatment-seeking populations may be clinically distinct from those represented in tightly controlled clinical trials. Finally, in our analysis of treatment processes in the community sample, we found that RCT exclusion status was associated with a 2.2-fold increase in the proportion of sessions involving case management and a 1.9-fold increase in session dosage. Exclusion status did not relate to the likelihood of full treatment termination or successfully concluding anxiety specialty treatment and transferring to supportive care only. Collectively, findings are consistent with the idea that youth who meet common exclusion criteria may benefit from standard CBT processes for anxiety and related disorders; however, such youth may also benefit from more complex, intensive, or prolonged care than what is typically represented in efficacy trials. This aligns with broader trends in the field that are moving away from singular-protocol approaches toward those that can more flexibly meet youth needs [33,34,40,41,42]. Overall, these results provide continual support for an “implementation cliff” in community mental health, in which intervention effects drop as EBTs move away from laboratory-controlled conditions [19]. While the observational nature of our study design restricts causal inferences, these results do suggest that a significant amount of youth in clinically representative conditions (1) would not meet criteria for inclusion in trials (53%), and (2) those who would have been excluded have more complicated and resource-intensive treatment courses.

In addition, the demographic characteristics of our naturalistic sample differed notably from those historically represented in RCTs. In our sample, 55.3% of clients identified as White, 14.9% as Black, 10.6% as Asian, and 7.4% as mixed race. By contrast, a systematic review of CBT efficacy trials reported that, among participants whose race is documented, nearly 75% are White [43]. However, we found no racial or ethnic differences between youth who would have been excluded versus included in RCTs (see Table 2), suggesting that RCT exclusion status was primarily driven by clinical characteristics rather than demographic factors. This suggests that underrepresentation of racial and ethnic minorities in pediatric anxiety RCTs may be primarily driven by recruitment and access barriers rather than exclusion criteria themselves. Furthermore, observed differences in treatment outcomes between those who would have or would not have been excluded from an RCT in the naturalistic sample may be more driven by clinical differences, rather than racial or ethnic identity. However, race was partially confounded with Medicaid status in this sample, which may reflect the association between race and adverse social determinants of health rooted in historical systemic inequities [44,45].

Findings point to important directions for future research. First, the results underscore the importance of pragmatic treatment trials with broader inclusion criteria, consistent with an ongoing shift in the literature toward enhancing the generalizability of trial findings to routine clinical populations [46,47]. Second, this study found that those who would have been excluded from RCTs required significantly more treatment sessions and case management, a result that requires further study. Instead of viewing this as indicative of decreased effectiveness of CBT for anxiety and related disorders (especially as RCT exclusion did not relate to the likelihood of successfully concluding specialty treatment), it may be reflective of a need for treatment adaptation or augmentation in this population. Third, the findings of this study could be used to help set expectations for what treatment progress might look like for youth and families. We found that RCT exclusion status was associated with longer duration of treatment and higher rates of case management use, findings that may be of prognostic use. Sharing the ‘RCT match status’ of their child may help families in understanding the prognosis of treatment with CBT and in shared decision-making about treatment adaptations. For example, explaining exposure therapy as something that should be completed within “12–16 weeks” (aligned with traditional RCT protocols) may give some families false hope of how soon they can expect treatment success to occur. Similarly, such information may help clinicians identify the extent to which a traditional CBT protocol may need to be adapted (e.g., incorporation of case management services, extended intervention length, needing to modify CBT content and delivery to address comorbid conditions) and develop appropriate expectations for how and when to expect to see treatment progress. Finally, our findings suggest that racial and ethnic background were not related to RCT exclusion status; yet the naturalistic treatment sample was much more racially and ethnically diverse than clinical trials in existing CBT literature [25,43]. To our knowledge, no meta-analyses have examined the racial and ethnic characteristics reported in CBT trials specifically for pediatric anxiety. Future research should examine disparities in representation within CBT trials for youth anxiety and identify the structural and cultural barriers that limit participation among minoritized youth.

Several limitations should be considered when interpreting the findings of this study. First, our review of RCT exclusion criteria did not rely on a full systematic review; instead, we relied on recently published meta-analyses (which did employ systematic review procedures) and backward citation chaining. Although this strategy captured a substantial number of efficacy trials, it may have resulted in the omission of eligible studies not represented in the selected meta-analyses. A notable limitation of backward citation chaining is that identified references will necessarily predate the seeded work. Given that half of the included meta-analyses were published in 2020, with two published in that year itself, the search strategy yielded only seven RCTs within the 2020–2024 timeframe. Due to this limited sample size and resulting low cell counts, Fisher’s exact test was employed to examine trends in exclusion criteria, a test that is inherently conservative and statistically underpowered [48]. This may account for the modest significance level and effect size (p = 0.049, V = 0.23) observed between study decade and the use of suicidal ideation as an exclusion criterion; thus, these findings should be interpreted with caution. Related, this review is best characterized as a narrative rather than a systematic review, and our literature sample should not be assumed to fully represent the complete body of RCTs of CBT for pediatric anxiety. Accordingly, our findings should be interpreted as preliminary and suggestive, rather than definitive. In addition, the clarity and specificity of exclusion criteria varied considerably across RCTs; some provided only broad descriptions (e.g., excluding youth with a “parent with a serious psychiatric disorder”), which introduced ambiguity during data extraction. Furthermore, because our sample comprises a heterogeneous set of RCTs, our method cannot determine how exclusion criteria may vary across geographic regions, participant age ranges (e.g., young children versus older adolescents), or diagnostic focus (e.g., trials targeting OCD vs. SAD). Furthermore, our review of RCTs of CBT for pediatric anxiety excluded effectiveness trials by design, which are designed to more accurately model routine clinical practice. This study sought to evaluate how the research conditions in tightly controlled trials, designed to minimize confounding variables to maximize internal validity and establish treatment efficacy, translate to treatment outcomes in a truly naturalistic clinical setting. To fully characterize the depth of any possible ‘implementation cliff,’ we restricted our focus to efficacy trials, as effectiveness studies with minimal exclusion criteria were unlikely to meaningfully inform this aim. Limitations also apply to the naturalistic chart review. The retrospective observational design of this study precludes causal interpretation of the associations reported here. Moreover, clinical and demographic variables were abstracted from medical records, which are subject to potential input errors and inconsistencies, and thus may not fully capture participants’ clinical presentations. Finally, the sample size was relatively small (n = 94), which has particular implications for interpreting logistic regression analyses. While the sample size met key assumptions for recommended events per variable [49,50], there may still have been reduced statistical power. The results of our logistic regression should not be viewed as definitive until replication in a future study with a larger sample size.

This study also had several notable strengths. First, our narrative review encompassed RCTs published between 2000 and 2024, providing a preliminary look at how exclusion criteria appear to have evolved over time. To our knowledge, this is also the first study to link RCT exclusion criteria to clinical outcomes in a community sample of racially, economically, and culturally diverse youth in an urban U.S. setting. Overall, findings suggest that youth who meet common CBT trial exclusion criteria may require longer and more complex treatment courses when receiving care in naturalistic settings. Future research broadening trial eligibility, particularly with respect to medication use and comorbid diagnoses, may help improve the translation of findings from CBT efficacy trials to community clinical settings.