Efficacy and Safety of Escitalopram for Major Depressive Disorder in Children and Adolescents: A Systematic Review and Meta-Analysis with Age-Specific Findings

Introduction

Major depressive disorder (MDD) in children and adolescents is frequently comorbid with various psychiatric conditions, including attention-deficit/hyperactivity disorder (ADHD), anxiety disorders,^1–3 and post-traumatic stress disorder (PTSD).⁴ The duration of untreated depression serves as a critical predictor of clinical outcomes, with early intervention shown to reduce both symptom severity and the risk of recurrence significantly.⁵ Furthermore, longer duration of untreated depression has been associated with worse outcomes at follow-up.⁶

MDD affects approximately 2.8% of children under 13 years and 5.6% of adolescents aged 13–18 years, with prevalence rates continuing to increase.⁷ Early-onset MDD is associated with substantial impairment across multiple life domains,¹ including poor academic achievement, impaired peer relationships, and increased risk of substance use.^8,9 Depression represents a significant risk factor for suicide in adolescents,¹⁰ and untreated pediatric MDD frequently persists into adulthood with continued functional impairment.⁷ These significant consequences underscore the critical importance of effective treatment interventions for pediatric MDD.

In Thailand, MDD represents a significant and growing public health concern. The Thai National Mental Health Survey reported a lifetime prevalence of MDD of 1.6% in the general population,¹¹ and national registry data from the Department of Mental Health indicate that approximately 2.9 million Thais aged 15 years and above were diagnosed with depression in 2023, more than doubling from 1.3 million in 2015.¹² The burden is particularly pronounced among young people: a nationally representative survey across 13 public health regions found that 17.5% of Thai adolescents screened positive for depression, with a corresponding suicide risk rate of 22%,¹³ while a school-based study reported depressive symptoms in approximately 18.6% of Thai high-school students.¹⁴ Clinical MDD affects approximately 2.4% of Thai adolescents seeking care, and suicide has emerged as the third leading cause of death in this age group.^12,13 Despite this high burden, Thailand has fewer than 200 child and adolescent psychiatrists for a population of approximately 15 million adolescents,¹⁵ underscoring the urgent need for evidence-based pharmacological guidance to support prescribers in primary and secondary care settings.¹⁴

Treatment selection for pediatric MDD follows a stepped-care approach based on illness severity.^16,17 For mild to moderate depression, psychotherapy—including cognitive behavioral therapy, interpersonal therapy, and family-based interventions—represents the first-line treatment approach. For moderate to severe depression, either psychotherapy or pharmacotherapy may be initiated, with combined treatment often recommended for severe cases.¹⁸ Among pharmacological interventions, selective serotonin reuptake inhibitors (SSRIs) have demonstrated efficacy in pediatric MDD, whereas tricyclic antidepressants have failed to show superiority over placebo in this population.¹⁹

Among SSRIs, fluoxetine has the broadest evidence base and is the only agent with dual regulatory approval for pediatric MDD in both children (aged 8 years and above) and adolescents. Sertraline is widely used in clinical practice despite lacking formal approval for pediatric MDD, and duloxetine, a serotonin-norepinephrine reuptake inhibitor, has not demonstrated consistent superiority over placebo in pediatric MDD trials and carries no approval for this indication.^20,21 Published meta-analyses report pooled CDRS-R mean differences of −2.84 (95% CI: −4.12 to −1.56) for fluoxetine (Hetrick et al, 2021),²² consistent with earlier estimates (Cipriani et al, 2016),²³ and −3.51 (95% CI: −6.99 to −0.04) for sertraline versus placebo,²² providing a quantitative benchmark against which the present escitalopram findings can be interpreted.

Fluoxetine, approved by the U.S. Food and Drug Administration (FDA) for pediatric MDD in patients aged 8–18 years, has the strongest evidence base for treating depression in children and adolescents.²⁴ However, alternative treatment options are needed when fluoxetine is contraindicated, poorly tolerated, or ineffective. Escitalopram, which received FDA approval for adolescent MDD (ages 12–17 years) in 2009, based on one positive escitalopram trial and one positive citalopram trial,²⁵ represents an important alternative SSRI option.²⁶ To date, three randomized controlled trials have evaluated escitalopram in pediatric MDD: two acute treatment trials^27,28 and one continuation treatment trial.²⁹ While two trials demonstrated efficacy,^28,29 findings have been mixed, with one trial²⁷ failing to show a significant benefit in the primary analysis.

However, several factors limit confidence in these findings. First, only two acute treatment trials have been conducted, with relatively small sample sizes (n = 268 and n = 316), resulting in wide confidence intervals and limited statistical power. Second, one trial failed to demonstrate significant benefit in the overall pediatric sample (ages 6–17) but suggested potential efficacy in the adolescent subgroup (ages 12–17).²⁷ Third, pediatric antidepressant trials are characterized by high placebo response rates, which can obscure treatment effects.³⁰ Fourth, no prior systematic review has synthesized the available evidence or rigorously assessed certainty of evidence using contemporary methodological approaches such as the Grading of Recommendations Assessment, Development and Evaluation (GRADE). A comprehensive meta-analysis is therefore needed to provide more precise estimates of escitalopram’s efficacy, acceptability, and tolerability in pediatric MDD.

This systematic review and meta-analysis primarily aim to evaluate the efficacy, acceptability, safety, and tolerability of escitalopram for the acute treatment of MDD in adolescents aged 12–17 years—the age group for which regulatory approval exists and for which the available evidence is most robust. Data for children aged 6–11 years are insufficient to permit definitive conclusions, and age-stratified analyses are reported transparently to distinguish findings across the two subpopulations. Secondary objectives include assessing escitalopram’s efficacy in continuation treatment and relapse prevention, and contextualizing its effect size relative to other approved pharmacological agents for pediatric MDD.

Methods

Types of Studies

Eligible studies were randomized controlled trials (RCTs) comparing escitalopram with placebo in pediatric populations. Studies comparing escitalopram with active comparators (eg., other antidepressants) were also eligible, but none were identified.

Types of Participants

The participants included were children and adolescents aged 6–17 years with MDD diagnosed using standardized diagnostic criteria (eg., DSM-IV, DSM-5, ICD-10, ICD-11).

Types of Interventions

The intervention was escitalopram at any dose and duration compared with placebo or active control. Studies evaluating both acute treatment and continuation/maintenance treatment were eligible.

Types of Outcome Measures

Primary Outcome Measure

The primary outcome was the mean change from baseline to endpoint in the Children’s Depression Rating Scale-Revised (CDRS-R) total score.

Table 1 Basic Characteristics of Randomized, Controlled Trials of Escitalopram versus Placebo in the MDD Treatment

Secondary Outcome Measures

Secondary outcomes consisted of:

1. Mean change from baseline in Clinical Global Impressions-Severity (CGI-S).
2. Mean change from baseline in Children’s Global Assessment Scale (CGAS).
3. Mean endpoint score for Clinical Global Impressions-Improvement (CGI-I).
4. Response rate (using study-specific criteria as defined in Table 1: CDRS-R ≤28 for Wagner 2006; CGI-I ≤2 or ≥40% reduction in CDRS-R for Emslie 2009)
5. Remission rate, defined as CDRS-R total score ≤28.
6. Overall discontinuation rate (acceptability).
7. Discontinuation rate due to adverse events (tolerability).
8. Specific adverse events with incidence ≥5%.

Information Sources

We searched the following electronic databases from January 2001 through October 2025: Cochrane Central Register of Controlled Trials (CENTRAL), CINAHL, PubMed, ClinicalTrials.gov, and Scopus. The search was limited to January 2001 onwards, as escitalopram received its first approval for clinical use in 2002. We also hand-searched reference lists of included studies and relevant systematic reviews to identify additional eligible trials. No language restrictions were applied.

Searches

We used a comprehensive search strategy combining terms for escitalopram [(Escitalopram) OR (S-citalopram) OR (Cipralex) OR (Lexapro) OR (Lu-26-054)] with terms for major depressive disorder [(Major depressive disorder) OR (Major depression) OR (MDD)]. Searches were limited to children and adolescents and randomized controlled trials using database-specific filters. The specific search strategies for each database were as follows.

CINAHL: [(Escitalopram) OR (S-citalopram) OR (Cipralex) OR (Lexapro) OR (Lu-26-054)] AND [(Major depressive disorder) OR (Major depression) OR (MDD)] Filters: Child or Adolescent, and Randomized Controlled Trial.

Cochrane Library: [(Escitalopram) OR (S-citalopram) OR (Cipralex) OR (Lexapro) OR (Lu-26-054)] AND [(Major depressive disorder) OR (Major depression) OR (MDD)] Filters: Child or Adolescent, and Trial.

PubMed: [(Escitalopram) OR (S-citalopram) OR (Cipralex) OR (Lexapro) OR (Lu-26-054)] AND [(Major depressive disorder) OR (Major depression) OR (MDD)] Filters: Child or Adolescent, and Randomized Controlled Trial.

Scopus: [(Escitalopram) OR (S-citalopram) OR (Cipralex) OR (Lexapro) OR (Lu-26-054)] AND [(Major depressive disorder) OR (Major depression) OR (MDD)] Filters: Child or Adolescent, and Randomized Controlled Trial.

Supplementary Search Methods

In addition to electronic database searching, we employed several supplementary search strategies to ensure comprehensive study identification:

Reference List Screening

We manually screened the reference lists of all included studies and relevant systematic reviews to identify additional eligible trials that may not have been captured through database searches.

Trial Registry Searching

We searched ClinicalTrials.gov and other international trial registries to identify completed but unpublished trials and ongoing studies.

Grey Literature

We searched for conference abstracts and dissertations through relevant conference proceedings and dissertation databases.

Author Contact

We did not contact study authors or pharmaceutical manufacturers for unpublished data, as all identified eligible trials had been published in peer-reviewed journals. However, we noted this limitation in our discussion of publication bias.No language restrictions were applied to any search strategy. Non-English language articles were translated as needed for eligibility assessment.

Study Selection

Two reviewers (NM and BM) independently screened titles and abstracts of all retrieved records for potential eligibility. Full-text articles of potentially eligible studies were obtained and independently assessed by the same two reviewers. Disagreements were resolved through discussion and consensus. Reasons for exclusion at the full-text stage were documented. A complete list of excluded studies with reasons for exclusion is provided in Supplementary Tables 1 and 2.

Data Collection Process

One reviewer (NM) extracted data from included studies using a standardized data extraction form, which was independently verified by a second reviewer (BM). Disagreements were resolved through discussion and consensus. The extraction form was pilot tested on two studies before full implementation.

Data Items

Data were systematically extracted from each included study using a standardized form. The following information was collected:

Study Characteristics

Author, publication year, country, study design, treatment duration, trial registration number, funding source.

Participant Characteristics

Sample size, age range, mean age, sex distribution, diagnostic criteria, baseline severity scores (CDRS-R, CGI-S).

Intervention Details

Escitalopram dose range and titration schedule, placebo details, concomitant treatments allowed.

Efficacy outcomes

Mean changes and standard deviations for CDRS-R, CGI-S, CGAS; mean endpoint scores for CGI-I; response rates (using study-specific criteria as defined in Table 1); remission rates (CDRS-R ≤28).

Safety Outcomes

Overall discontinuation rates, discontinuation due to adverse events, specific adverse events with incidence ≥5%, serious adverse events.

Methodological Details

Randomization method, allocation concealment, blinding procedures, missing data handling, intention-to-treat analysis.

Other Information

Conflict of interest disclosures, sponsor involvement.

Risk of Bias in Individual Studies

Two reviewers (NM and BM) independently assessed risk of bias in included studies using the Cochrane Collaboration’s Risk of Bias tool.³¹ The following domains were evaluated: (1) random sequence generation (selection bias); (2) allocation concealment (selection bias); (3) blinding of participants and personnel (performance bias); (4) blinding of outcome assessment (detection bias); (5) incomplete outcome data (attrition bias); (6) selective reporting (reporting bias); and (7) other potential sources of bias (including funding source and conflicts of interest).³² Each domain was rated as low risk, high risk, or unclear risk of bias. Disagreements were resolved through discussion and consensus. Risk of bias assessments were incorporated into the GRADE certainty of evidence evaluations.

Assessment of Certainty of Evidence

The certainty of evidence for each outcome was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.³³ The GRADE framework evaluates five domains that may decrease certainty: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Evidence certainty was rated as high, moderate, low, or very low. High certainty evidence indicates high confidence that the true effect lies close to the estimated effect. Moderate certainty suggests moderate confidence in the effect estimate. Low certainty indicates limited confidence, and very low certainty means very little confidence in the effect estimate.

Two reviewers (NM and BM) independently assessed certainty for each outcome using the following criteria: Risk of bias was evaluated based on the Cochrane Risk of Bias assessment results. Inconsistency was assessed by examining statistical heterogeneity (I²) and consistency of effect direction. Indirectness was evaluated by considering whether the evidence directly addressed the review question in terms of population, intervention, comparator, and outcomes. Imprecision was assessed by examining sample size, number of events, and confidence interval width. Publication bias was considered based on the comprehensive search strategy and the number of included studies.

Disagreements in certainty ratings were resolved through discussion to reach consensus. The GRADE certainty assessments were summarized in a Summary of Findings table presenting the magnitude of effect for each outcome alongside the certainty rating.

Summary Measures

We assessed three main outcome categories following established conventions from prior meta-analyses:^34,35

Efficacy

Mean change from baseline in depression rating scales (continuous outcomes) and response/remission rates (dichotomous outcomes).

Acceptability

Overall discontinuation rate for any reason, serving as a proxy measure for overall treatment acceptability.³⁴

Tolerability

Discontinuation rate due to adverse events, serving as a proxy measure for treatment tolerability.³⁵ While these discontinuation-based measures are commonly used proxy indicators, we acknowledge they have limitations as they may not fully capture patient-reported acceptability and tolerability.

Statistical Analysis and Synthesis of Results

For continuous outcomes measured on the same scale (eg., CDRS-R, CGI-S, CGAS, CGI-I), we calculated weighted mean differences (WMD) with 95% confidence intervals (CIs). For outcomes measured on different scales, we planned to use standardized mean differences (SMD). When standard deviations were not reported, we calculated them from other available statistics using established methods.³⁶

For dichotomous outcomes (response rate, remission rate, discontinuation rates), we calculated relative risks (RR) with 95% CIs using the Mantel-Haenszel method.³² Number needed to treat (NNT) with 95% CIs was calculated for the response rate outcome.

We used random-effects models for all meta-analyses, as we expected clinical and methodological heterogeneity across studies (eg., differences in patient populations, dosing regimen, study duration). Random-effects models assume the true effect size varies across studies and provide more conservative estimates than fixed-effect models.³² All analyses were conducted using Review Manager (RevMan) version 5.1 software.

Assessment of Heterogeneity and Planned Subgroup Analyses

Statistical heterogeneity was assessed using the I² statistic and Chi² test, with I² values interpreted as follows: 0–40% might not be important, 30–60% may represent moderate heterogeneity, 50–90% may represent substantial heterogeneity, and 75–100% represents considerable heterogeneity.³²

We planned the following a priori subgroup analyses to explore potential sources of heterogeneity:

- Age groups: Children (ages 6–11 years) versus adolescents (ages 12–17 years).

- Treatment duration: Short-term (≤8 weeks) versus longer-term (>8 weeks) trials.

- Baseline severity: Moderate versus severe depression at baseline.

- Dosing: Fixed-dose versus flexible-dose regimens.

If sufficient studies were available (≥10 studies per subgroup), we planned to conduct meta-regression analyses to examine the relationship between treatment effects and continuous covariates such as mean baseline age, mean baseline CDRS-R score, and year of publication.

Planned Sensitivity Analyses

To assess the robustness of our findings, we planned the following sensitivity analyses:

- Excluding studies at high risk of bias in any domain.

- Excluding studies with industry funding.

- Using fixed-effect models instead of random-effects models to assess the impact of model choice.

- Excluding studies with high attrition rates (>20%).

- Comparing intention-to-treat analyses with completer analyses where data permitted.

- Excluding the largest study to assess whether results were driven by a single large trial.

These sensitivity analyses were planned to evaluate whether the pooled effect estimates were robust to methodological decisions and study quality.

Risk of Bias Across Studies

We planned to assess publication bias using funnel plots if ≥10 studies were available for meta-analysis.³² However, given that only three studies were included, formal assessment of publication bias using funnel plots or statistical tests was not feasible. We qualitatively considered the potential for publication bias based on comprehensive database searches, industry sponsorship of all trials, and mixed results, with one of two acute treatment trials failing to demonstrate significant efficacy on the primary outcome.

Assessment of Heterogeneity

Statistical heterogeneity was assessed using the I² statistic, with I² ≥50% indicating substantial heterogeneity. We planned to explore sources of heterogeneity through subgroup analyses (eg., children vs. adolescents, acute vs. continuation treatment) and sensitivity analyses if sufficient studies were available. Forest plots were used to visually assess consistency of effect estimates across studies.

Protocol Deviations and Post-Hoc Decisions

This systematic review was conducted according to a prospectively registered protocol (PROSPERO CRD42019141268, registered August 2019). The following deviations from the original protocol were made:

Protocol Deviations

1. Extended search period: The original protocol specified searches through December 2024; we extended searches through October 2025 to capture the most recent evidence.
2. Age-specific subgroup analysis: Post-hoc subgroup analyses by age (children ages 6–11 vs. adolescents ages 12–17) were conducted based on findings from Wagner 2006²⁷ suggesting potential age-specific effects. This analysis was not pre-specified in the original protocol but was clinically justified, given that FDA approval was limited to adolescents.
3. Continuation treatment: The protocol focused primarily on acute treatment efficacy. We expanded the scope to include continuation treatment and relapse prevention, based on the identification of a relevant trial (Saito, 2023)²⁹ published after protocol registration.

No Protocol Deviations

- Primary and secondary outcomes remained as pre-specified.

- Inclusion and exclusion criteria were applied as originally planned.

- Statistical methods (random-effects models, GRADE) were conducted as specified.

- Risk of bias assessment methods were unchanged from the protocol.

All deviations represent expansions rather than restrictions of the originally planned review and were made to provide more comprehensive evidence synthesis. These deviations were decided before data extraction and analysis were completed.

Figure 1 Flow diagram of the study. Abbreviations: CCTR, Cochrane Central Register of Controlled Trials; CINAHL, Cumulative Index to Nursing and Allied Health Literature; CT.gov, ClinicalTrials.gov; n, number; No, number; PubMed, Publisher and Medline; SCOPUS, Scopus is an Elsevier abstract and citation database.

Results

Study Selection

According to searching those databases, a total of 1,272 citations (Cochrane Controlled Trials Register=154, CINAHL=94, ClinicalTrials.gov=26, PubMed=101, SCOPUS=897) were identified. (see Figure 1) After duplicate articles were removed, the titles and abstracts of 1,130 citations were inspected. Then, 23 citations were compatible for the full-text evaluation. Of these 18 were excluded from this study (11: reviews/secondary analysis of RCT data,^37–47 2: open-label,^48,49 1: extension trial,⁵⁰ 2: recruiting studies,^51,52 1: a retrospective study,⁵³ 1: combination therapy (trauma stabilization plus escitalopram) rather than escitalopram monotherapy vs. placebo.⁵⁴

Ultimately, five citations, representing three randomized controlled trials, were included for synthesis in this systematic review.^{27–29,55,56} However, other relevant or unpublished trials were not identified.

Study Characteristics

Three randomized, double-blind, placebo-controlled trials involving 664 participants met inclusion criteria: two acute treatment trials (total n=584 randomized, n=572 in ITT analysis) and one continuation treatment trial (n=80) (see Table 1).

Acute Treatment Trials

Wagner et al (2006)²⁷ randomized 268 children and adolescents (aged 6–17 years) across 20 United States sites. Of these, 264 participants received at least one dose of study medication (escitalopram n=131, placebo n=133) and comprised the safety population. The intention-to-treat (ITT) efficacy analysis included 261 participants (escitalopram n=129, placebo n=132) who received study medication and had at least one post-baseline assessment. Study completion rate was 81% (217/268 randomized participants), with 217 participants completing 8 weeks (escitalopram: 102/134, 76%; placebo: 115/134, 86%). The study used flexible dosing of escitalopram 10–20 mg/day based on clinical response and tolerability.

Emslie et al (2009) randomized 316 adolescents (aged 12–17 years) across 40 international sites, with 158 allocated to each treatment group.²⁸ A total of 312 participants received at least one dose of study medication (escitalopram n=155, placebo n=157) and comprised the safety population. The ITT efficacy analysis included 311 participants (escitalopram n=154, placebo n=157) who received study medication and had at least one post-baseline assessment. Study completion rate was 82% (259/316 randomized participants), with 259 participants completing 8 weeks (escitalopram: 126/158, 80%; placebo: 133/158, 84%).The study used flexible dosing of escitalopram 10–20 mg/day.

Combined ITT Population: The pooled acute treatment meta-analyses included 572 participants (escitalopram n=283, placebo n=289) (see Table 2).

Table 2 Participant Flow and Analysis Populations in Acute Treatment Trials

Continuation Treatment Trial

Saito et al (2023)²⁹ conducted a withdrawal trial in Japan. Following a 10–12 week open-label phase (n=128 enrolled), 80 responders were randomized to continue escitalopram (n=39) or switch to placebo (n=41) for 36 weeks to evaluate relapse prevention in adolescents aged 12–17 years.

Baseline demographic and clinical characteristics were generally well-balanced between treatment groups across all studies. Mean age ranged from 12.6 to 14.9 years, with 55–70% female participants. Baseline depression severity was moderate to severe, with mean CDRS-R scores ranging from 55 to 64. Participants with comorbid psychiatric conditions requiring psychotropic medication were excluded from all studies.

Primary efficacy analyses used intention-to-treat populations with last observation carried forward (LOCF) for missing data. Response definitions varied: Wagner 2006²⁷ used CDRS-R ≤28, Emslie 2009²⁸ used CGI-I ≤2 or ≥40% CDRS-R reduction, and Saito 2023²⁹ used ≥50% CDRS-R reduction plus CGI-S ≤2. Remission was consistently defined as CDRS-R ≤28.

Risk of Bias Within Studies

Risk of bias assessment revealed variable methodological quality across the three included studies, with domains evenly distributed between low risk and unclear risk ratings, the latter primarily due to insufficient reporting of allocation concealment procedures and assessor blinding details (see Figure 2).

Figure 2 Summary of risk of bias in randomized controlled trials of escitalopram versus placebo in children and adolescents.27–29 Notes: Green = low risk of bias; Yellow = unclear risk of bias. Studies included: Wagner 2006,27 Emslie 2009,28 Saito 2023.29.

Random Sequence Generation (Selection Bias)

Two studies (Emslie 2009 and Saito 2023)^28,29 were rated as low risk of bias, with both using computer-generated randomization schedules with 1:1 allocation. Wagner 2006 received an unclear risk rating despite mentioning a computer-generated randomization schedule, as the method was not fully described.²⁷

Allocation Concealment (Selection Bias)

All three studies received unclear risk ratings due to insufficient reporting of allocation concealment procedures. While Wagner 2006 mentioned that randomization numbers were allocated in blocks of 4, the specific concealment procedure was not described.²⁷ Similarly, neither Emslie 2009 nor Saito 2023 explicitly described their allocation concealment methods.^28,29

Blinding of Participants and Personnel (Performance Bias)

All three studies were rated as low risk of bias. Wagner 2006 used identical tablets in appearance with a double-blind design.²⁷ Emslie 2009 maintained a double-blind design with placebo and escitalopram likely identical.²⁸ Saito 2023 employed a double-blind parallel-group design with medications likely identical in appearance.²⁹

Blinding of Outcome Assessment (Detection Bias)

Two studies (Emslie 2009 and Saito 2023) were rated as having a low risk of bias, as both employed independent Central Assessment Committees that reviewed outcome determinations blindly.^28,29 Wagner 2006 received an unclear risk rating as the blinding of assessors was not explicitly stated in the available reporting.²⁷

Incomplete Outcome Data (Attrition Bias)

All three studies adequately addressed incomplete outcome data and were rated as low risk of bias. Wagner (2006) reported an 81% completion rate (217/268 randomized participants) using intention-to-treat (ITT) analysis and documented reasons for discontinuation.²⁷ Emslie 2009 reported an 82% completion rate (259/316 randomized participants) using both ITT and last observation carried forward (LOCF) analyses with comprehensive reporting of discontinuation reasons.²⁸ Saito 2023 reported good completion rates (92.2% in the open-label phase, with variation in the double-blind period) and employed ITT analysis with multiple imputation methods.²⁹ All studies provided clear accounting of participant flow with balanced attrition between treatment groups.

Selective Reporting (Reporting Bias)

All three studies were rated as low risk for selective outcome reporting. Wagner 2006²⁷ and Emslie 2009²⁸ reported all pre-specified outcomes mentioned in their methods sections, though original study protocols were not publicly available for verification, as prospective trial registration was not standard practice when these studies were conducted. Saito 2023 reported all pre-specified outcomes and was prospectively registered (jRCT2080224520).^29,55 All studies reported primary and secondary efficacy outcomes, as well as safety outcomes, in sufficient detail for meta-analysis.

Other Potential Sources of Bias

All three studies received unclear risk ratings for other potential sources of bias. All three studies received financial support from pharmaceutical companies holding patents for escitalopram. Wagner 2006²⁷ and Emslie 2009²⁸ were funded by Forest Laboratories Inc., and Saito 2023²⁹ was funded by Mochida Pharmaceutical Co., Ltd. Several authors in each study were employees or consultants of sponsor companies. Industry sponsorship has been associated with more favorable outcomes in antidepressant trials and represents a potential source of bias. However, the studies employed generally rigorous methodological approaches, and this potential bias was considered in the GRADE certainty assessment.

Summary of Risk of Bias Assessment

Overall, the included studies demonstrated adequate handling of incomplete outcome data, selective reporting, and blinding of participants and personnel, but had consistent limitations in reporting of allocation concealment procedures across all three studies. Emslie 2009²⁸ and Saito 2023²⁹ provided sufficient methodological detail across most domains, each receiving low risk ratings in five of seven domains. Wagner 2006²⁷ had the most unclear risk ratings (four domains), reflecting inadequate reporting rather than definitively poor methods. Industry sponsorship of all trials represents a concern for potential bias, with all three studies receiving unclear risk ratings for other bias. These limitations were incorporated into the GRADE assessment, contributing to the downgrading of certainty for some outcomes (see Table 3).

Table 3 GRADE Summary of Findings - Escitalopram versus Placebo for Major Depressive Disorder in Children and Adolescents

Synthesis of Results

Efficacy

Acute Treatment

Statistical heterogeneity was not significant for any pooled efficacy outcome (I²=0% across all analyses).

Primary Outcome

CDRS-R Mean Change: Meta-analysis of two acute treatment trials (Wagner 2006: n=261, 129 escitalopram/132 placebo; Emslie 2009: n=311, 154 escitalopram/157 placebo; combined ITT population n=572, escitalopram n=283, placebo n=289)^27,28 demonstrated statistically significant superiority of escitalopram over placebo (WMD = −2.46, 95% CI: −4.83 to −0.09, p=0.04, I²=0%) (see Figure 3). However, the confidence interval barely excluded the null effect. Mean changes from baseline (mean baseline CDRS-R ~56-58 points) were −21.9 to −22.1 points for escitalopram versus −18.8 to −20.2 points for placebo, representing an additional 2.5-point improvement with escitalopram.

Figure 3 Forest plot comparing mean change scores from baseline on the Children’s Depression Rating Scale–Revised (CDRS-R) between escitalopram and placebo in children and adolescents with major depressive disorder.27,28 The pooled mean difference was calculated using a random-effects model. A negative mean difference indicates a greater reduction in depressive symptoms for the escitalopram group. Abbreviations: CI, confidence interval; IV, inverse variance; df, degrees of freedom; SD, standard deviation; MD, mean difference. CDRS-R, Children’s Depression Rating Scale–Revised. Notes: N represents the number of participants included in the ITT analysis with available post-baseline data.

Secondary Continuous Outcomes

CGI-S change (2 studies, n=572; escitalopram n=283, placebo n=289): Escitalopram demonstrated significantly greater improvement (WMD = −0.35, 95% CI: −0.57 to −0.12, I²=0%), representing approximately 0.35 points greater reduction in severity (see Figure 4).

Figure 4 Forest plot comparing mean change scores from baseline on the Clinical Global Impression–Severity (CGI-S) scale between escitalopram and placebo in children and adolescents with major depressive disorder.27,28 The pooled mean difference was calculated using a random-effects model. A negative mean difference indicates a greater improvement (reduction in severity) for the escitalopram group. Abbreviations: CI, confidence interval; IV, inverse variance; df, degrees of freedom; SD, standard deviation; MD, mean difference. CGI-S, Clinical Global Impression–Severity. Notes: N represents the number of participants included in the ITT analysis with available post-baseline data.

CGAS change (2 studies, n=572; escitalopram n=283, placebo n=289): Escitalopram showed significantly greater improvement in functioning (WMD = 2.56, 95% CI: 0.37 to 4.75, I²=0%), representing a 2.6-point greater improvement on the 100-point scale (see Figure 5).

Figure 5 Forest plot comparing mean change scores from baseline on the Children’s Global Assessment Scale (CGAS) between escitalopram and placebo in children and adolescents with major depressive disorder.27,28 The pooled mean difference was calculated using a random-effects model. A positive mean difference indicates a greater improvement in global functioning for the escitalopram group. Abbreviations: CI, confidence interval; IV, inverse variance; df, degrees of freedom; SD, standard deviation; MD, mean difference. CGAS, Children’s Global Assessment Scale. Notes: N represents the number of participants included in the ITT analysis with available post-baseline data.

CGI-I endpoint (2 studies, n=572; escitalopram n=283, placebo n=289): Escitalopram had significantly lower (better) endpoint scores (WMD = −0.29, 95% CI: −0.51 to −0.08, I²=0%) (see Figure 6).

Figure 6 Forest plot comparing mean endpoint scores on the Clinical Global Impression–Improvement (CGI-I) scale (95% confidence interval) between escitalopram and placebo in children and adolescents with major depressive disorder.27,28 Abbreviations: CI, confidence interval; IV, inverse variance; df, degrees of freedom; SD, standard deviation; MD, mean difference; CGI-I, Clinical Global Impression–Improvement.

Response Rate (2 studies, n=572; escitalopram n=283, placebo n=289): Clinical response (defined using study-specific criteria: CDRS-R ≤28 for Wagner 2006; CGI-I ≤2 or ≥40% CDRS-R reduction for Emslie 2009; see Table 1) was significantly higher with escitalopram (53.0%; 150/283) compared to placebo (43.6%; 126/289), yielding an RR of non-response of 0.84 (95% CI: 0.72 to 0.98, I²=0%) (see Figure 7). This corresponds to a 9.4% absolute increase in response rate and a number needed to treat of 11 (95% CI: 6 to 80).

Figure 7 Forest plot comparing relative risk (95% confidence interval) for clinical response rates between escitalopram and placebo in children and adolescents with major depressive disorder.27,28 Events represent non-responders (treatment failure), defined using study-specific response criteria (CDRS-R ≤28 for Wagner 2006; CGI-I ≤2 or ≥40% CDRS-R reduction for Emslie 2009; see Table 1). A risk ratio less than 1 indicates fewer non-responders (ie., higher response rate) with escitalopram, favoring escitalopram. Abbreviations: CI, confidence interval; df, degrees of freedom; M-H, Mantel–Haenszel.

Remission Rate: Although remission (CDRS-R ≤28) was pre-specified as a secondary outcome, pooled meta-analysis of remission rates was not performed. This was because Wagner 2006 defined both response and remission using the same threshold (CDRS-R ≤28), as noted in Table 1, making the two outcomes identical for that trial. Pooling remission data would therefore have substantially duplicated the response rate analysis for one of the two contributing studies, precluding a meaningful independent estimate. Individual study remission rates are reported in the original publications.

Continuation Treatment

One withdrawal trial (Saito 2023, n=80)²⁹ evaluated relapse prevention over 36 weeks in adolescent responders. The primary outcome (time to relapse) did not differ significantly between continued escitalopram (n=39) and placebo (n=41), with a hazard ratio of 2.96 (95% CI: 0.94 to 9.30, p=0.051). Relapse rates were 10.3% versus 26.8% (p=0.085).

Although the primary endpoint did not reach statistical significance (time to relapse: HR=2.96, 95% CI 0.94–9.30, p=0.051), escitalopram showed a numerical improvement suggesting possible benefit in relapse prevention. On exploratory secondary outcomes, statistically significant differences were observed: greater reductions in CDRS-R total score (mean change: −2.1 vs +3.3, p=0.048) and CGI-S (mean change: −0.1 vs +0.4, p=0.025), lower CGI-I scores (1.6 vs 2.1, p=0.030), and lower CGI-S worsening rates (7.7% vs 34.1%, p=0.005). These secondary findings should be interpreted as exploratory and hypothesis-generating only, as the primary endpoint was not met. Notably, the positive values (+3.3 for CDRS-R and +0.4 for CGI-S) in the placebo group indicate symptom worsening following escitalopram discontinuation, whereas the negative values in the escitalopram group reflect continued improvement or symptom stability. This pattern demonstrates that patients who continued escitalopram treatment maintained their treatment gains, while those switched to placebo experienced deterioration in depressive symptoms and overall severity.

Sensitivity Analysis

To assess the robustness of the primary efficacy findings, we performed a leave-one-out sensitivity analysis by sequentially excluding each study and re-estimating the pooled effect. This analysis was considered essential given the small number of included studies (n=2 for acute treatment).

Primary Outcome (CDRS-R): When Wagner 2006²⁷ was excluded (leaving only Emslie 2009,²⁸ n=311), the CDRS-R WMD was −3.30 (95% CI: −6.75 to 0.15), which did not reach statistical significance as the confidence interval crossed the null. When Emslie 2009²⁸ was excluded (leaving only Wagner 2006,²⁷ n=261), the CDRS-R WMD was −1.70 (95% CI: −4.98 to 1.58), also nonsignificant, consistent with the original report of Wagner 2006.²⁷ These results indicate that neither individual trial alone achieves statistical significance for the primary outcome; the pooled estimate reaches significance solely through the combined statistical power of both trials.

Secondary Outcomes: For CGI-S, Emslie 2009²⁸ alone showed a significant effect (WMD=−0.40, 95% CI: −0.72 to −0.08), while Wagner 2006²⁷ alone did not, as the confidence interval crossed the null (WMD=−0.30, 95% CI: −0.62 to 0.02). For CGI-I, Emslie 2009²⁸ alone was significant (WMD=−0.40, 95% CI: −0.71 to −0.09), while Wagner 2006²⁷ alone was not, as the confidence interval crossed the null (WMD=−0.20, 95% CI: −0.49 to 0.09). For CGAS and response rate, neither individual trial alone reached statistical significance, as all confidence intervals crossed the null or unity (CGAS: Wagner 2006²⁷ WMD=2.90, 95% CI: −0.16 to 5.96; Emslie 2009²⁸ WMD=2.20, 95% CI: −0.93 to 5.33; response rate: Emslie 2009 RR=0.79, 95% CI: 0.62 to 1.01; Wagner 2006²⁷ RR=0.87, 95% CI: 0.71 to 1.07).

These results confirm that the pooled primary outcome finding is substantially dependent on the combined contribution of both trials and is not robust to the exclusion of either study. The consistent significance of Emslie 2009²⁸ across secondary outcomes (CGI-S and CGI-I) provides partial supportive evidence; however, the failure of either trial alone to demonstrate significance on the primary outcome (CDRS-R) underscores the limited robustness of the pooled estimate. Clinicians and guideline developers should be aware that the overall evidence rests on a narrow statistical margin and a mixed results pattern between the two trials. The complete leave-one-out sensitivity analysis results are reported in Supplementary Table 3.

Main Adverse Events

Acute Treatment

Treatment-emergent adverse events (TEAEs) with an incidence of ≥5% occurred in similar proportions between treatment groups: 69–78% with escitalopram versus 68–75% with placebo, with no significant difference in the overall rate of patients experiencing at least one adverse event.^27,28 Common adverse events showed comparable rates between the escitalopram and placebo groups. The pooled relative risks (RRs) with 95% confidence intervals (CIs) for specific adverse events were: headache (RR=1.01, 95% CI: 0.76 to 1.36, I²=0%), abdominal pain (RR=1.57, 95% CI: 0.88 to 2.78, I²=0%), nausea (RR=1.38, 95% CI: 0.78 to 2.44, I²=0%), rhinitis (RR=0.87, 95% CI: 0.48 to 1.58, I²=0%), pharyngitis (RR=0.88, 95% CI: 0.50 to 1.57, I²=0%), upper respiratory infection (RR=0.76, 95% CI: 0.40 to 1.46, I²=0%), and vomiting (RR=1.23, 95% CI: 0.62 to 2.45, I²=0%).^27,28 All confidence intervals crossed 1.0, indicating no statistically significant differences between treatment groups for any individual adverse event. No heterogeneity was observed across studies (I²=0% for all outcomes).^27,28

Suicidality

Suicidal ideation or behavior occurred infrequently and at comparable rates between treatment groups, with no statistically significant difference. In the Emslie 2009 study, suicidal events were reported in 3.9% (6/155) of escitalopram-treated patients versus 2.5% (4/157) of placebo-treated patients.²⁸ Wagner 2006 reported one suicidal ideation event in the escitalopram group and none in the placebo group.²⁷ Importantly, no completed suicides occurred in either study during the acute treatment phase.

Laboratory Values

Mean changes in clinical laboratory values from baseline to endpoint were generally comparable between treatment groups. However, a statistically significant difference was observed in platelet count, with escitalopram-treated patients showing a greater decrease (mean change: −7.6 × 10⁹/L) compared to placebo-treated patients (mean change: −2.2 × 10⁹/L).²⁸ Despite statistical significance, this difference was considered clinically minor as values remained within normal ranges and no patients discontinued treatment due to thrombocytopenia.

Vital Signs and Weight

Mean changes in vital signs (blood pressure, heart rate) from baseline to endpoint were small and similar between the escitalopram and placebo groups. Mean weight gain was modest and comparable in both treatment groups (approximately 1.2 lb in both groups).²⁸

Electrocardiogram (ECG)

No electrocardiograms at endpoint were considered abnormal and clinically significant by investigators in either treatment group.²⁸

Continuation Treatment

In the double-blind phase of Saito 2023,²⁹ adverse event incidence was higher in the escitalopram group (82.1%, 32/39) versus placebo (56.1%, 23/41), though treatment-related adverse drug reactions were similar (23.1%, 9/39 vs 22.0%, 9/41). Most adverse events were mild to moderate. No complete suicides occurred.²⁹

Overall Discontinuation Rate (Acceptability)

For acute treatment (Wagner 2006: safety population n=264, 131 escitalopram/133 placebo; Emslie 2009: safety population n=312, 155 escitalopram/157 placebo),^27,28 overall discontinuation rates did not differ significantly between escitalopram and placebo (RR = 1.37, 95% CI: 0.97 to 1.93, I²=0%). Discontinuation rates were 18.7–22.1% for escitalopram versus 13.5–15.3% for placebo.^27,28 For continuation treatment (Saito 2023, n=80), discontinuation rates were comparable during the 36-week double-blind phase.²⁹

Discontinuation Rate Due to Adverse Events (Tolerability)

For acute treatment (2 studies, Wagner 2006: n=264; Emslie 2009: n=312),^27,28 discontinuation due to adverse events did not differ significantly between escitalopram and placebo (RR = 1.88, 95% CI: 0.44 to 8.03, I²=0%). Discontinuation rates were 1.5–2.6% for escitalopram versus 0.6–1.5% for placebo. The extremely wide confidence interval reflects very few events (9 total: 6 in the escitalopram group, 3 in the placebo group) and substantial uncertainty.^27,28 For continuation treatment (1 study, n=80), discontinuation due to adverse events was numerically higher with escitalopram (5.1%, 2/39) versus placebo (0%, 0/41).²⁹

Risk of Bias Across Studies

Assessment of publication bias through funnel plot analysis was not feasible due to the limited number of included studies. Standard recommendations suggest that at least 10 studies are needed for a reliable interpretation of funnel plot asymmetry.⁵⁷ With only three trials available for meta-analysis, neither a visual nor a statistical assessment of publication bias could be conducted using conventional methods.

Therefore, the following observations regarding the evidence base are noted with appropriate caution. Funding and sponsorship. All three included studies received financial support from pharmaceutical companies holding patents for escitalopram. While industry sponsorship is common in pharmaceutical research and enables the conduct of well-designed trials, it is worth noting that industry-sponsored trials have been associated with more favorable outcomes in some analyses.⁵⁸ This does not necessarily indicate bias in individual studies but suggests the need for continued vigilance and independent replication.

Evidence-Based Characteristics

The acute treatment evidence base comprises two trials.^27,28 Both were conducted during the 2000s. The U.S. Food and Drug Administration approved escitalopram for adolescent depression (ages 12–17 years) in March 2009, based on one positive escitalopram trial²⁸ and one positive citalopram trial.^25,59 While this represents a more limited evidence base than for some other antidepressants, it met regulatory requirements at that time. The historical context surrounding the approval process has been discussed in the literature,⁵⁹ though this does not diminish the scientific merit of the individual trials, which employed appropriate methodological safeguards.

Recent Contributions to the Evidence Base

The publication of Saito 2023²⁹ represents an important addition to the literature. This continuation treatment study, published 14 years after the FDA approval, reported mixed results with the primary outcome (time to relapse) approaching but not reaching statistical significance (HR=2.96, p=0.051), while several secondary outcomes demonstrated benefits. The willingness to publish findings with a non-significant primary outcome reflects positively on contemporary publication practices and scientific transparency. This study also addresses an important evidence gap by providing the first data on continuation treatment efficacy in adolescent depression.²⁹ Regarding trial registration practices, two of the three included trials were prospectively registered (Emslie 2009: NCT00107120;⁵⁶ Saito 2023: jRCT2080224520),⁵⁵ which enhances transparency and allows for verification of reported outcomes. The Wagner 2006 trial predated widespread trial registration requirements, which were not yet standard practice at that time.^27–29

Limitations and Interpretation

While the evidence base for escitalopram in pediatric depression is smaller than might be ideal, it demonstrates reasonable methodological quality and consistency in findings. The low certainty ratings for efficacy outcomes and low to very low certainty ratings for safety outcomes reflect limitations inherent to a small number of studies rather than concerns about the quality of individual trials. The recent publication of Saito 2023²⁹ with transparent reporting of mixed results provides reassurance regarding contemporary publication practices, though the earlier evidence base remains limited.

Readers should interpret findings with appropriate consideration of these limitations, recognizing that additional research would strengthen confidence in treatment effects while also acknowledging that the available evidence provides a reasonable foundation for clinical decision-making when combined with other considerations.

Several Limitations Regarding the Evidence Base Warrant Consideration

Of the two acute treatment trials, one (Emslie 2009)²⁸ demonstrated significant efficacy on the primary outcome, while the other (Wagner 2006)²⁷ failed to show significant benefit in the overall population (CDRS-R: WMD = −1.7, p=0.31), though post-hoc analyses suggested potential efficacy in adolescents aged 12–17 years. An important methodological challenge in pediatric antidepressant trials is the substantial placebo response rate. Bridge et al (2009)³⁰ reported mean placebo response rates of approximately 58% (range: 26-87%) across 15 trials in pediatric major depressive disorder. In the present review, both Wagner 2006²⁷ and Emslie 2009²⁸ reported placebo response rates of approximately 52–53% (CGI-I ≤2, as reported in original publications). The pooled response rate analysis used study-specific criteria (see Table 1), yielding a lower overall placebo response rate of 43.6% (126/289; see Figure 7). The high placebo response presents a significant challenge in detecting treatment effects, as the robust improvement in placebo groups reduces the magnitude of difference between active treatment and placebo, even when the active treatment is effective. The ability of Emslie 2009²⁸ to demonstrate statistically significant efficacy despite these challenges provides support for a genuine treatment effect, albeit modest in magnitude. These limitations were considered in the GRADE certainty assessment.³³ Readers should interpret findings recognizing that the evidence base consists of one clearly positive acute treatment trial, one negative acute treatment trial with potential age-specific effects, and one continuation trial with mixed results.

Assessment of Publication Bias and Reporting Biases

Formal Assessment Limitations

Due to the small number of included studies (n=2 acute treatment trials, n=1 continuation trial), formal statistical assessment of publication bias using funnel plot asymmetry tests or Egger’s regression was not feasible or appropriate. Current methodological guidance recommends a minimum of 10 studies for reliable interpretation of funnel plot asymmetry.⁵⁷ Therefore, we conducted a qualitative assessment of potential publication bias and reporting biases.

Evidence Base Characteristics

All three included trials (Wagner 2006, Emslie 2009, Saito 2023)^27–29 received financial support from pharmaceutical companies holding patents for escitalopram (Forest Laboratories/Lundbeck, Mochida Pharmaceutical). Industry sponsorship is common in pharmaceutical research and has been associated with more favorable reported outcomes in some systematic analyses.⁵⁸ However, this does not necessarily indicate bias in individual studies, and the included trials employed rigorous methodological approaches.

Mixed Results Pattern

The acute treatment evidence demonstrates a mixed pattern: one trial (Emslie 2009)²⁸ showed significant efficacy on primary outcomes, while another (Wagner 2006)²⁷ failed to demonstrate significant benefit in the overall population (p=0.31), though post-hoc subgroup analyses suggested potential efficacy in adolescents. This pattern of mixed results, including publication of a trial with a negative primary outcome, provides some reassurance against selective publication of only positive findings.

Trial Registration and Transparency

Two of three trials were prospectively registered in public trial registries (Emslie 2009: NCT00107120; Saito 2023: jRCT2080224520), enhancing transparency and allowing verification that reported outcomes matched those pre-specified. Wagner 2006²⁷ predated widespread mandatory trial registration. Comparison of registry entries with published reports revealed no evidence of selective outcome reporting for the registered trials.

Recent Publication Practices

The most recent trial (Saito 2023)²⁹ reported mixed findings with the primary outcome (time to relapse) approaching but not reaching statistical significance (p=0.051), while secondary outcomes showed benefits. The willingness to publish results with a non-significant primary outcome demonstrates improved contemporary publication practices and scientific transparency.

Comprehensive Search Strategy

Our search strategy was comprehensive, including five major databases, trial registries, and handsearching of reference lists. We identified no unpublished studies through trial registry searches, though the absence of identified unpublished trials does not definitively rule out their existence.

Overall Assessment

While formal quantitative assessment of publication bias was not feasible, several factors provide partial reassurance: (1) publication of trials with negative or mixed primary outcomes, (2) prospective trial registration for recent studies, (3) comprehensive search strategy, and (4) transparent reporting in recent trials. However, the universal industry sponsorship, small evidence base, and impossibility of formal publication bias testing remain important limitations. The true magnitude of publication bias, if present, cannot be determined from the available evidence. These limitations were incorporated into our GRADE certainty assessments, contributing to low rather than high certainty ratings for efficacy outcomes.

Certainty of Evidence

The certainty of evidence was assessed using GRADE criteria³³ and is summarized in Table 3. For efficacy outcomes, certainty was rated as low for all measures. The primary outcome of CDRS-R mean change was downgraded from high to low certainty for two reasons: (1) Risk of Bias — universal industry sponsorship of all included trials (Forest Laboratories and Mochida Pharmaceutical) introduces potential commercial bias, as industry-funded trials have been associated with more favorable outcomes; and (2) Imprecision — the confidence interval barely excluded the null effect (WMD=−2.46, 95% CI: −4.83 to −0.09) and the combined sample size was modest (n=572). Similarly, response rate, CGI-S change, CGAS change, and CGI-I endpoint scores were all rated as low certainty for the same reasons.

For safety and tolerability outcomes, certainty was lower. Overall discontinuation rate was rated as low certainty, downgraded for both indirectness (8-week trials may not adequately capture long-term acceptability) and imprecision (confidence interval includes both potential benefit and harm). Discontinuation due to adverse events was rated as very low certainty, downgraded one level for indirectness and two levels for very serious imprecision due to very few events and an extremely wide confidence interval (0.44 to 8.03).

Evidence was downgraded for risk of bias due to the potential for commercial bias, as all included trials received funding from pharmaceutical companies with patent interests in escitalopram. Although the trials employed adequate randomization and blinding, industry sponsorship represents a systematic concern that may inflate reported treatment effects. Evidence was also downgraded for imprecision given the modest sample size (n=572) and confidence intervals that barely excluded the null. Inconsistency was not a concern, with I²=0% across all outcomes. Publication bias could not be formally assessed due to the small number of studies (n=2 for acute treatment); however, the potential for selective publication is acknowledged given universal industry sponsorship. The continuation treatment study (Saito 2023)²⁹ could not be formally rated using GRADE as it was a single study without comparators for meta-analysis.

Discussion

The present systematic review and meta-analysis evaluated the efficacy, acceptability, and tolerability of escitalopram for major depressive disorder in children and adolescents. Meta-analysis of two acute treatment trials (n=572) demonstrated statistically significant but modest improvements favoring escitalopram over placebo on the primary outcome (CDRS-R: WMD=−2.46, 95% CI: −4.83 to −0.09, p=0.04). However, the confidence interval barely excluded the null effect. Significant benefits were also observed on secondary outcomes, including CGI-S, CGI-I, CGAS, and response rates (NNT=11). However, findings were mixed: while one trial (Emslie 2009)²⁸ demonstrated clear efficacy, another trial (Wagner 2006)²⁷ failed to show a significant benefit in the overall population (p=0.31), although post-hoc analyses suggested potential efficacy in adolescents aged 12–17 years. Overall discontinuation rates and discontinuation due to adverse events did not differ significantly between escitalopram and placebo, indicating acceptable tolerability and acceptability profiles in short-term (8-week) treatment. For continuation treatment, a single trial (Saito 2023, n=80)²⁹ showed mixed results: the primary confirmatory outcome (time to relapse) did not reach statistical significance (p=0.051) and should not be interpreted as evidence of efficacy. Although exploratory secondary outcomes showed statistically significant differences numerically favoring escitalopram continuation, these findings are hypothesis-generating only and cannot be used to support definitive conclusions regarding relapse prevention. A clear distinction must be maintained between the non-significant primary endpoint and these exploratory secondary findings.

These findings add to the evidence base for antidepressant treatment in pediatric depression. While tricyclic antidepressants have not demonstrated efficacy in treating MDD in children and adolescents,¹⁹ several selective serotonin reuptake inhibitors (SSRIs) have shown benefits in this population.^60,61 Fluoxetine, approved by the U.S. Food and Drug Administration for pediatric MDD in patients aged 8–18 years, has the strongest evidence base with multiple positive trials demonstrating superiority to placebo.²⁴ A Cochrane network meta-analysis reported pooled CDRS-R mean differences of −2.84 (95% CI: −4.12 to −1.56) for fluoxetine and −3.51 (95% CI: −6.99 to −0.04) for sertraline versus placebo,²² providing a quantitative benchmark against which the present escitalopram findings (WMD=−2.46, 95% CI: −4.83 to −0.09) can be interpreted. These comparisons confirm that escitalopram’s benefits are more modest and less consistent than those for fluoxetine, with only one of two acute trials showing primary-outcome efficacy. An important contextual consideration is the high placebo response rate characteristic of pediatric antidepressant trials: Bridge et al (2009)³⁰ reported mean placebo response rates of approximately 58% across 15 trials, and both trials in this review reported placebo response rates of approximately 52–53%.(CGI-I ≤2, as reported in original publications). However, the pooled analysis in this review used study-specific response criteria (see Table 1), yielding a lower overall placebo response rate of 43.6% (126/289; see Figure 7). This discrepancy reflects the more stringent response criterion (CDRS-R ≤28) applied to Wagner 2006,²⁷ compared with CGI-I ≤2 reported in the original publication. This substantially reduces the detectable difference between active treatment and placebo, and may partially explain why one trial failed to demonstrate significant efficacy.

The safety and tolerability profile of escitalopram was generally favorable. The incidence of treatment-emergent adverse events was comparable between escitalopram and placebo (69–78% vs 68–75%), with no significant differences for any specific adverse event. Regarding suicidality, events occurred infrequently and at comparable rates between groups, with no completed suicides in any included trial. These findings are consistent with broader evidence suggesting that SSRIs do not increase suicidality compared to placebo in controlled trials and may reduce it when depression is adequately treated.⁶² Discontinuation rates due to adverse events were low and similar between groups (1.5–2.6% for escitalopram vs 0.6–1.5% for placebo), indicating good tolerability. Overall discontinuation rates were also comparable between groups (18.7–22.1% vs 13.5–15.3%). These findings confirm that short-term tolerability is acceptable; however, longer-term safety data remain limited, and no definitive conclusion on long-term safety can be drawn.

Several important limitations warrant consideration. First, the evidence base is limited, with only two acute treatment trials included in the meta-analysis, one of which failed to demonstrate significant efficacy on the primary outcome; furthermore, the leave-one-out sensitivity analysis confirmed that neither individual trial alone achieves statistical significance for the primary outcome (CDRS-R), with the pooled estimate reaching significance solely through the combined statistical power of both trials. This limited evidence base, combined with universal industry sponsorship of all included trials, raises concerns about commercial bias, as reflected in the downgrading of GRADE certainty ratings. Second, the evidence applies primarily to adolescents aged 12–17 years. Third, the magnitude of benefit is modest (approximately 2.5 CDRS-R points; NNT=11). Fourth, all trials were short (8 weeks). Fifth, continuation treatment evidence comes from a single trial with mixed results. Sixth, universal industry sponsorship of all included trials raises the possibility of selective outcome reporting, outcome switching, and analytic flexibility in trial conduct, even in prospectively registered studies. Such concerns have been documented in prior pediatric SSRI trials (eg., citalopram, paroxetine). While our assessment of included trials found no evidence of selective outcome reporting, this possibility cannot be fully excluded and represents an additional limitation when interpreting the evidence presented in this review. Seventh, pooled remission rates could not be analyzed independently because the response and remission definitions in Wagner 2006 were identical (both CDRS-R ≤28).

Based on the available evidence, escitalopram should not be considered a first-line treatment, in alignment with NICE, AACAP, and FDA risk–benefit principles. It may be considered an alternative treatment option for adolescent MDD (aged 12–17 years) when fluoxetine is contraindicated, poorly tolerated, or ineffective. Treatment should be combined with appropriate psychosocial interventions, and close monitoring for suicidality is essential, particularly during the initial weeks of treatment.

Future research should address several key gaps. Additional well-designed RCTs are needed, including independent studies not sponsored by pharmaceutical manufacturers. Studies should specifically examine efficacy in younger children (ages 6–11 years). Longer-term studies are needed to evaluate sustained efficacy and optimal duration of continuation treatment. Head-to-head comparative trials with fluoxetine provide valuable information for treatment selection. Research should also identify predictors of treatment response and develop interventions to address the high placebo response rates in pediatric depression trials.

Comparison with Previous Reviews

The escitalopram effect size observed in this review (WMD=−2.46; NNT=11) is numerically smaller than the corresponding estimates for fluoxetine (WMD=−2.84) and sertraline (WMD=−3.51) reported in the Cochrane network meta-analysis by Hetrick et al (2021).²² While direct head-to-head comparisons are not available, this indirect comparison suggests that escitalopram occupies a relatively modest position within the SSRI class for pediatric MDD. The consistency of evidence also differs: fluoxetine has demonstrated superiority to placebo across multiple independent trials, whereas the present escitalopram evidence rests on only two trials, one of which failed to demonstrate primary-outcome efficacy. These comparative considerations reinforce the clinical guideline position that fluoxetine should remain the preferred first-line pharmacological agent for adolescent MDD.

Safety and Tolerability

The favorable short-term tolerability profile of escitalopram is consistent with its FDA approval for adolescent MDD and with broader SSRI class data. Of particular clinical importance is the suicidality signal: the FDA black-box warning reflects findings from earlier meta-analyses, but the trials in this review found suicidality events to be infrequent and balanced between groups. Clinicians should nonetheless maintain vigilance during the first four weeks of treatment, consistent with FDA labeling requirements. The absence of significant differences in specific adverse event rates is a clinically reassuring finding, though it must be interpreted cautiously given the short trial durations and the limited ability to detect rare events across a total sample of 572 participants.

Limitations

The findings of this review should be interpreted in light of an evidence base that currently rests on only two available trials, both of which were industry-sponsored, necessitating cautious interpretation of the overall conclusions. The inability to formally assess publication bias is a particularly important gap, as favorable trial results are more likely to be published than null results, and registered pediatric escitalopram trials without full publication cannot be excluded. These methodological constraints mean that the true efficacy of escitalopram in this population may be either overestimated (if unpublished trials were negative) or underestimated (if high placebo response rates suppressed the detectable treatment signal). Both possibilities warrant the cautious interpretation endorsed by the GRADE certainty ratings applied in this review.

Clinical Implications

Global Clinical Implications

The findings of this systematic review and meta-analysis have several important implications for the clinical management of adolescent MDD in global practice. First, the statistically significant but modest treatment effect observed (WMD=−2.46; NNT=11), combined with low-to-moderate certainty of evidence, reinforces the position established by major international guidelines — including NICE, AACAP, and the FDA — that escitalopram should not be considered a first-line pharmacological treatment for adolescent MDD. Fluoxetine, which has demonstrated larger and more consistent effects across multiple independent trials, remains the preferred first-line agent where pharmacotherapy is indicated.²² Second, the evidence from this review supports escitalopram as a reasonable second-line option for adolescents aged 12–17 years in whom fluoxetine has proven ineffective, poorly tolerated, or contraindicated. Its acceptable short-term tolerability profile — with low rates of treatment discontinuation due to adverse events (1.5–2.6%) and no significant increase in suicidality compared to placebo — supports its cautious use in clinical practice within this age group, with appropriate monitoring in the initial weeks of treatment consistent with FDA labeling requirements. Third, the high placebo response rates observed across both included trials (52–53%) are a clinically meaningful finding for practitioners worldwide. These rates are consistent with the broader literature on pediatric antidepressant trials³⁰ and suggest that structured clinical contact itself carries substantial therapeutic value, independent of pharmacological mechanisms. This reinforces international guideline recommendations that pharmacotherapy should always be embedded within a broader treatment framework — including cognitive behavioural therapy (CBT), interpersonal therapy, and psychoeducation — rather than deployed as a standalone intervention. Combined treatment approaches have consistently demonstrated superior outcomes compared to pharmacotherapy alone in adolescent MDD.⁶³ Finally, the evidence from this review applies exclusively to adolescents aged 12–17 years.

Clinicians and health systems globally should exercise caution in extrapolating these findings to children aged 6–11 years, for whom the current evidence base is insufficient to support escitalopram use.

Clinical Implications for Thailand

The findings of this review carry particular relevance for Thailand, where adolescent MDD represents a growing public health challenge. National surveillance data indicate that approximately 37.8% of Thai adolescents are at risk for depression, and the incidence rate has increased substantially over the past decade.⁶⁴ Key risk factors documented in Thai adolescents aged 12–17 years include academic-related stress, bullying and cyberbullying, family dysfunction, social isolation, and problematic smartphone use — all of which are known to converge during this developmental period and contribute to depressive onset.^64–66 This constellation of stressors underscores why the 12–17 age group represents a critical window for early identification and intervention in the Thai context.

Given the modest effect size observed in this review (WMD=−2.46; NNT=11), pharmacotherapy with escitalopram should not be considered a standalone intervention. Rather, it should be integrated within a broader care model incorporating psychosocial support, family-based interventions, and school-based mental health programs — combined approaches that have demonstrated superior outcomes compared to pharmacotherapy alone in adolescent MDD.^46,63 This integrated approach is especially critical in Thailand, where a documented shortage of child and adolescent psychiatrists relative to the adolescent population means that primary care physicians frequently serve as first-contact providers for young people with depressive symptoms.⁶⁷ The acceptable tolerability profile of escitalopram demonstrated in this review supports its cautious use in primary care settings for appropriately selected adolescents aged 12–17 years, provided that close clinical monitoring is maintained.

The high placebo response rates observed in the included trials (52–53%) may reflect, in part, the therapeutic benefit of structured clinical contact itself — consistent with evidence from the broader pediatric antidepressant trial literature.³⁰ This observation underscores the value of regular clinical engagement as an integral component of treatment, independent of pharmacological effects, and is particularly relevant in Thailand where the therapeutic relationship may itself serve as an important protective factor for adolescents with limited access to specialist care. Efforts to reduce mental health stigma, which remains a significant barrier to help-seeking among Thai adolescents,⁶⁵ will be equally essential to ensuring that effective pharmacological and psychosocial treatments reach those who need them. Future research specific to Thai adolescents is needed to evaluate locally adapted psychosocial interventions, to assess the feasibility of integrated care models within Thailand’s primary care infrastructure, and to examine whether risk factors prevalent in this population — including cyberbullying and digital-related stressors — moderate treatment response to escitalopram and other antidepressants.

Conclusions

Based on low certainty evidence, escitalopram demonstrates modest efficacy in treating major depressive disorder in adolescents aged 12–17 years, though one of two acute treatment trials failed to show a significant benefit. The observed effect size is small (WMD=−2.46 on CDRS-R, NNT=11) with the confidence interval barely excluding the null effect. Evidence for continuation treatment and relapse prevention is insufficient from a single trial with mixed results. Tolerability and acceptability were comparable to placebo, with similar rates of adverse events, including suicidality.

Escitalopram may be considered as an alternative when fluoxetine is unsuitable, though evidence is less robust than for fluoxetine. Treatment should be combined with psychosocial interventions and close monitoring. The limited evidence base, mixed results, modest effects, and universal industry sponsorship warrant cautious interpretation. Further independent, longer-term studies are needed to confirm these findings and establish efficacy in younger children.