Effects of acceptance and commitment therapy on negative emotions, automatic thoughts and psychological flexibility for depression and its acceptability: a meta-analysis

Abstract

Background

Acceptance and Commitment therapy (ACT) has been widely used in patients with depression. However, its effectiveness in improving psychological flexibility and reducing automatic thoughts remains uncertain. This meta-analysis of randomized controlled trials (RCTs) aimed to evaluate the effects of ACT on depression, anxiety, automatic thought, and psychological flexibility in patients with depression, as well as its acceptability.

Methods

RCTs were systematically searched in nine databases and gray literature, with the last update on March 25, 2025. Effect sizes were synthesized using a random-effects model, and subgroup analyses were conducted to explore potential heterogeneity. Publication bias was corrected using three methods: PET-PEESE, selection models, and robust Bayesian meta-analysis. The certainty of evidence was evaluated using the GRADE approach.

Results

A total of 13 RCTs from 1362 patients were included in this meta-analysis. Pooled results showed that ACT significantly improved depression [SMD = − 0.66 (− 0.80, − 0.52), P <.001, I² = 24% (0%, 75%), Certainty: Low], anxiety [SMD = − 0.43 (− 0.77, − 0.10), P <.05, I² = 77% (34%, 97%), Certainty: Moderate], and psychological flexibility [SMD = 0.50 (0.35, 0.66), P <.001, I² = 36% (0%, 83%), Certainty: Moderate] compared with controls at post-test. However, there was no significant positive effect on automatic thoughts [SMD = − 0.28 (− 0.69, 0.12), P =.17, I² = 65% (0%, 97%), Certainty: Very low]. Notably, the positive effects of ACT on depression, anxiety and psychological flexibility were maintained at follow-up. Furthermore, the difference in acceptability between ACT and the control condition was not statistically significant (P ≥.05). Subgroup analyses indicated that face-to-face ACT was more effective than internet-based ACT.

Conclusion

According to the GRADE assessment, the certainty of the evidence ranges from very low to moderate. ACT appears to significantly improve depressive symptoms, anxiety, and psychological flexibility in individuals with depression. However, its effects on automatic thoughts and its acceptability still require further investigation.

Meta-analysis registration on prospero

: CRD42024533794.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-025-07067-w.

Background

Depression is a common mental illness [1], characterized primarily by persistent low mood, loss of interest, and low energy, often accompanied by significant cognitive decline [2]. As one of the most prevalent mental health issues contributing to the global burden of disease, approximately 322 million people worldwide suffer from depression, accounting for 4.4% of the global population [3]. The emergence of the COVID-19 pandemic has triggered a number of mental health problems, with a 27.6% increase in new cases of major depressive disorder (MDD), which now affects more than 53 million people [4]. Individuals with depression are more prone to experiencing negative emotions and distorted cognition compared to the general population. Without timely treatment, depression can worsen, leading to more severe and persistent symptoms, significantly impairing quality of life, and increasing the risk of suicidal thoughts and behaviors [5]. This underscores the critical importance of effective treatment and prevention of depression.

Currently, the most common treatments for depression include medication and psychotherapy. However, many patients with depression do not respond well to antidepressants, leading to high rates of comorbidities and poor outcomes [6]. Given these concerns, psychotherapy is highly recommended due to its minimal side effects, with Cognitive Behavioral Therapy (CBT) being the preferred approach [7]. Acceptance and Commitment Therapy (ACT), a prominent therapy in the third generation of CBT, has gained recognition as an effective treatment option. ACT is based on functional contextualism and relational frame theory (RFT) and aims to enhance an individual’s psychological flexibility [8]. It fosters psychological flexibility through six interrelated and overlapping elements: acceptance, cognitive defusion, present-moment awareness, self-as-context, values, and committed action. The six elements of ACT are primarily applied to two core intervention processes: mindfulness acceptance and committed action [8]. During the mindfulness acceptance process, ACT guides individuals with depression to focus on the present moment and accept aspects of life that cannot be changed [9]. This approach helps reduce emotional avoidance behaviors and alleviates depressive symptoms. Additionally, through cognitive defusion, ACT helps individuals perceive negative automatic thoughts as transient mental events rather than as true definitions of the self [10]. Patients with depression often experience feelings of worthlessness and anhedonia [2]. In the committed action process, ACT encourages individuals to redefine their personal values and commit to actions aligned with these values [8]. This helps to boost the patient’s intrinsic motivation and sense of purpose in life, thereby reducing depressive symptoms.

Although growing evidence suggests that ACT can relieve anxiety and depression by enhancing psychological flexibility [11, 12], there is a lack of comprehensive evidence regarding its effects on psychological flexibility and automatic thoughts in patients with depression. Previous meta-analyses have predominantly focused on the effects of ACT in cancer patients [13] and chronic pain patients [14], with limited attention given to those with depression. A previous meta-analysis [15] focusing on individuals with depression assessed only depression and psychological flexibility, with limited follow-up data. Moreover, it included patients with physical disabilities, resulting in substantial clinical heterogeneity. Thus, the effect of ACT on psychological flexibility in depressed patients remains unclear. Negative automatic thoughts, characterized by persistent and recurrent false perceptions [16], are known to increase the risk of depression relapse [17]. Some studies have shown that ACT can effectively reduce these automatic thoughts in patients with depression [18, 19]. However, no meta-analysis has yet been conducted to explore the effects of ACT on automatic thoughts, highlighting the need for a comprehensive review of relevant studies. Furthermore, the acceptability of ACT is closely linked to treatment outcomes in depression [20], yet few studies have evaluated whether ACT demonstrates greater acceptability than control conditions in patients with depression [14, 21].

Therefore, this study aimed to systematically evaluate the effects of ACT on depression, anxiety, automatic thoughts, and psychological flexibility in patients with depression, as well as the acceptability of ACT. Additionally, it sought to determine whether the positive effects of ACT on patients with depression could be sustained at follow-up. Meanwhile, subgroup analysis was used to explore potential moderators that may influence the efficacy of ACT, including risk of bias, delivery format, measures, sample size, and number of sessions. The findings are intended to provide an evidence-based foundation for the development and refinement of ACT intervention programs.

Method

This meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [22], with a protocol registered on PROSPERO (CRD42024533794). The PRISMA Checklist was provided in Appendix 1 of the supplementary materials.

Search strategy and selection criteria

We systematically searched six English databases (PubMed, Embase, Cochrane Library, CINAHL, Web of Science, and PsycINFO) and three Chinese databases (CNKI, Wanfang, and VIP) for relevant literature published up to March 25, 2025. Additionally, grey literature from the Journal of Contextual Behavioral Science was reviewed to ensure the inclusion of more relevant randomized controlled trials (RCT). The search terms were based on the combination of Medical Subject Heading (MeSH) terms and free text words, including acceptance and commitment therapy, psychological flexib*, depression, depressive disorder, randomized controlled trial, controlled clinical trial. The search strategy was slightly modified for different databases. Detailed search strategies for each database are provided in Appendix 2 of the supplementary material.

Eligibility criteria

Inclusion criteria

RCTs were included based on PICOS (participants, interventions, comparisons, outcomes, study design) screening criteria:

(1) P (Participants): Patients diagnosed with depression via diagnostic interviews, meeting standardized diagnostic criteria for depression (e.g., DSM-IV, ICD-10, ICD-11). Patients not meeting criteria for clinical depression but scoring above thresholds on standardized scales were also eligible (e.g., BDI-II ≥ 14, CES-D ≥ 10, PHQ-9 ≥ 10, etc., indicating at least mild depression).

(2) I (Interventions): ACT delivered in various formats, including face-to-face and internet-based interventions.

(3) C (Comparisons): Passive controls (e.g., treatment as usual, no treatment, or waiting-list control) or provide minimal psychoeducation.

(4) O (Outcome): The primary outcomes were negative emotions (depression, anxiety), automatic thoughts and psychological flexibility; The secondary outcome was acceptability. Outcomes should be measured using standardized measures, including: depression (measured using the BDI-II, CES-D, various versions of the PHQ, RADS-2, MADRS, etc.), anxiety (measured using the BAI, HADS-A, etc.), automatic thoughts (measured using the ATQ-F), psychological flexibility (measured using various versions of the AAQ-II, CompACT etc.), and acceptability (Only measured by the percentage of participants who did not complete a post-test assessment for any reason).

(5) S (Study Design): Randomized controlled trials (RCT) designed based on ACT. If data from one RCT were split and used in multiple articles, we included only the one that met the inclusion criteria for this review and had the largest sample size.

Exclusion criteria

Studies were excluded if they met any of the following criteria: (1) Studies for which the full text and complete experimental data could not be obtained, even after reviewing the information and contacting the authors; (2) Ongoing or unfinished research.

Study selection and data extraction

All search results were imported into EndNote 21, and duplicates were manually removed. The remaining studies were independently screened by two researchers (YX.Z, RX.W) for titles and abstracts to exclude those that did not meet the inclusion criteria. The final studies for inclusion were then identified by reading the full text. Disagreements during screening or data extraction were resolved through discussion or by involving a third researcher (XC.X). The characteristics of the included RCTs (such as author, year, country, participant characteristics, sample size, brief descriptions of intervention and control groups, assessment points, outcomes, and measures) were extracted into tables to facilitate subsequent data integration. We used Cohen’s kappa to quantify the inter-rater reliability (IRR) of two reviewers in the screening process [23].

Risk of bias assessment

The quality of the randomized controlled trials included in this study was assessed using the risk assessment tool, version 2 of the cochrane risk-of-bias tool for randomized trials (ROB2) [24]. Two independent evaluators (YX.Z and XC.X) evaluated five domains: (1) randomization process, (2) deviations from intended interventions, (3) missing outcome data, (4) measurement of the outcome, and (5) selection of reported outcomes. In cases of disagreement, the two evaluators (YX.Z and XC.X) discussed the discrepancies, or a third evaluator (C.B) was consulted to reach a consensus. Studies were classified as having an overall high risk if any domain was rated as high risk. If no domain was rated as high risk but at least one domain was rated as some concerns, the study was categorized as having some concerns. When all five domains were rated as low risk, the study was classified as having an overall low risk. We quantified the inter-rater reliability (IRR) between the two evaluators using Cohen’s kappa [23].

Data synthesis and statistical analysis

Computation of effect sizes

This meta-analysis was conducted using RevMan 5.4.1 and Stata/MP 18.0. Standardized mean differences (SMD) were calculated for the four primary outcomes—depression, anxiety, automatic thoughts, and psychological flexibility—using Cohen’s d. For acceptability, relative risk (RR) was calculated. Cohen’s d was interpreted as follows: d ≈ 0.2 indicates a small effect size, d ≈ 0.5 represents a medium effect size, and d ≈ 0.8 corresponds to a large effect size [25]. Additionally, since scores for different versions of AAQ-II (7-item and 10-item) represent different effect directions [26], we used a negative sign to ensure that the effect directions were consistent across all studies (i.e., higher scores indicate greater psychological flexibility). A random-effects model accounts for heterogeneity among studies, particularly unexplained heterogeneity [27]. Given the inherent heterogeneity in depression presentations, interventions, and study designs, a random-effects model was deemed more appropriate. Therefore, we applied the DerSimonian-Laird (DL) method [28] for all random-effects analyses. All variables are reported with 95% confidence interval (CIs), and P <.05 was considered statistically significant.

Subgroup analysis and sensitivity analysis

The Cochrane Q-test was used to assess statistical heterogeneity and I² was calculated as a measure of heterogeneity in percentage terms. Overall I² values around 75%, 50%, and 25% were interpreted as high, medium, and low levels of heterogeneity [29]. We selected overall risk of bias (ROB2 level), delivery format, measures, sample size and number of sessions as effect moderators in the subgroup analyses of primary outcomes to explore potential sources of heterogeneity. Subgroup analyses in meta-analyses often suffer from low statistical power and should be interpreted with extreme caution [30]. Therefore, we also analyzed the statistical power of each subgroup analysis using the pwr package in the R software. In addition, the sensitivity analysis for each outcome was conducted using the leave-one-out method.

Publication bias assessment

Contour-enhanced funnel plots and Egger’s test were used to assess publication bias, with P <.05 indicating significant publication bias [31]. The contours of this type of funnel plot display the different levels of statistical significance (P <.05, P <.01). This helps detecting publication bias due to the suppression of non-significant findings. The potential for publication bias should be interpreted with caution, due to the subjectivity of visual inspection in funnel plots and the limited robustness of Egger’s test in small-sample contexts. Therefore, we added more robust statistical methods, such as PET-PEESE, selection models, and robust Bayesian meta-analysis (RoBMA-PSMA) to test publication bias and compare different models [31, 32].

GRADE assessment

We used the GRADE approach to evaluate the certainty of the evidence and presented a GRADE Evidence Profile along with a Summary of Findings table for the primary outcomes.

Results

Search process

A total of 4444 studies were retrieved from the database, and 2093 duplicates were removed. After screening titles and abstracts, an additional 2277 studies were excluded. By reviewing the full texts of the remaining 74 studies, 61 studies were excluded for not meeting the inclusion criteria, ultimately including 13 studies. The Cohen’s kappa for the screening process was approximately 0.832, indicating a high level of inter-rater reliability between two researchers. The PRISMA flowchart is presented in Fig. 1.

Fig. 1.

PRISMA flowchart

Characteristics of the included studies

This meta-analysis included 13 studies with a total of 1362 participants, including 743 in the ACT group and 619 in the control group. The included studies were published between 2011 and 2024. The studies were conducted in various countries, three in the Netherlands [33–35], two each in America [36, 37], China [38, 39], and Finland [18, 19], and one each in Australia [40], Korea [41], Malaysia [42], and New Zealand [43]. The sample sizes of these studies ranged from 38 [40] to 376 [34] participants. One study [40] focused on adolescent depression, while the other 12 studies [18, 19, 33–39, 41–43] targeted adult depression. Seven studies [18, 33, 34, 36, 38, 39, 41] were internet-based ACT interventions, and six [19, 35, 37, 40, 42, 43] were face-to-face ACT interventions. The duration of the intervention ranged from 1 week [43] to 12 weeks [33], with the number of sessions ranged from 3 [40] to 20 sessions [38], and the duration of each session ranged from 10 min [43] to 120 min [35]. One study [33] had a 6-month follow-up, four studies [34, 35, 40, 42] had a 3-month follow-up, two studies [38, 41] had a 1-month follow-up and six studies [18, 19, 36, 37, 39, 43] did not report complete follow-up data. The main characteristics of the included studies are presented in Table 1.

Table 1.

Characteristics of included studies

Author(year)	Country	Study Setting	Type of participants/ Age (Mean ± SD, I/C)	Sample size (I/C)	Interventions		Delivery format/ Material	Duration/Sessions/Time of each session	Assessment points	Outcome (Measures)
					experimental	control
Ahmad Othman 2023	Malaysia	School and Community	Adult depression 18 to 29 years old	34/34	ACT	Minimal psycho-education	Face to face/ ACT manuals	10w/10session/ 45 to 60 min	Baseline, Post (10w), FU(3 m)	① BDI-II ② BAI
Arroll 2022	New Zealand	Community	Adult depression 50.3 ± 18.2/ 44.2 ± 17.9	28/29	ACT	No treat	Face to face/ ACT manuals	1w/10session/ 10 min	Baseline, Post (1w)	① PHQ-8 ④ AAQ-II-7
Bohlmeijer 2011	Netherlands	Community	Adult depression 48.84 ± 11.34/ 49.23 ± 10.07	49/44	ACT	WLC	Face to face/ “Living to the full”	8w/8session/ 120 min	Baseline, Post (8w), FU(3 m)	① CES-D ② HADS-A ④ AAQ-II-10
Broten 2014	America	Hospital	Adult depression 47.85 ± 10.78/ 39.36 ± 10.91	14/25	ACT	TAU	Face to face/ ACT manuals	NR/5session/ 50 min	Baseline, Post (NR)	① MADRS ③ ATQ-F ④ AAQ-II-7
Cai 2024	China	Clinic	Adult depression 18 to 60 years old	34/33	IACT	WLC	Internet-based /“Mobile app-based ACT”	3w/20session/ 10–15 min	Baseline, Post (3w), FU(1 m)	① PHQ-9 ④ CompACT
Davis 2024	America	Clinic	Adult depression 45.5 ± 15.4/ 44.3 ± 13.2	47/46	IACT	WLC	Internet-based /“LifeStories”	4w/4session/ 30 min	Baseline, Post (4w)	① PHQ-9 ④ AAQ-II-7
Fledderus 2013	Netherlands	Community	Adult depression 42.50 ± 11/ 32.47 ± 11.29	250/126	IACT	WLC	Internet-based / “Living to the full”	9w/9session /NR	Baseline, Post (9w), FU(3 m)	① CES-D ② HADS-A ④ AAQ-II-10
Hayes 2011	Australia	Clinic	Adolescent depression 14.61 ± 3.1/ 15.49 ± 1.35	22/16	ACT	TAU	Face to face/ ACT manuals	NR/3session/NR	Baseline, Post (NR), FU(3 m)	① RADS-2
Jeong 2024	Korea	School	Adult depression 23.73 ± 4.07/ 23.23 ± 4.90	41/39	IACT	WLC	Internet-based /“Mobile app-based ACT”	4w/4session/ 20–30 min	Baseline, Post (4w), FU(1 m)	① CES-D ② HADS-A ④ AAQ-II-10
Kyllönen 2018	Finland	Clinic	Adult depression 50 ± 12/49 ± 13	60/59	ACT	WLC	Face to face/ ACT manuals	6w/6session/ 60 min	Baseline, Post (6w)	① BDI-II ③ ATQ-F ④ AAQ-II-10
Lappalainen 2015	Finland	Clinic	Adult depression 50.32 ± 12.54/ 53.40 ± 13.35	19/20	IACT	WLC	Internet-based / “Good Life Compass”	6w/6session/ NR	Baseline, Post (6w)	① BDI-II ③ ATQ-F ④ AAQ-II-10
Pots 2016	Netherlands	Community	Adult depression 45.15 ± 10.78/ 48.54 ± 12.63	82/87	IACT	WLC	Internet-based / “Living to the full”	12w/9session/ NR	Baseline, Post (12w), FU(6 m)	① CES-D ② HADS-A ④ AAQ-II-10
Zhao 2022	China	School	Adult depression 22.84 ± 2.34/ 23.43 ± 2.70	63/61	IACT	WLC	Internet-based /“Unguided self-help program”	6w/6sesssion/ 30 min	Baseline, Post (6w)	① BDI-II ③ ATQ-F ④ AAQ-II-7

Abbreviations: ①Depression; ②Anxiety; ③Automatic thoughts; ④Psychological flexibility; ACT, Acceptance and commitment therapy; IACT, Internet-based acceptance and commitment therapy; TAU, Treat as usual; WLC, Waiting list control; Post, post-test; FU, Follow-up; CompACT, Comprehensive Assessment of Acceptance and Commitment Therapy Processes; AAQ-II-7, Acceptance and Action Questionnaire-II (7 items); AAQ-II-10, Acceptance and Action Questionnaire-II (10 items); ATQ-F, The Automatic Thoughts Questionnaire; BDI-II, Beck Depression Inventory-II; BAI, Beck Anxiety Inventory; CES-D, Center for Epidemiologic Studies Depression Scale; HADS-A, Hospital Anxiety and Depression Scale-Anxiety; MADRS, Montgomery-Asberg Depression Rating Scale; PHQ-8, Patient Health Questionnaire; RADS-2, Reynolds Adolescent Depression Scale-2; C, control; I, intervention; w, week; m, month; NR, No report

Risk of bias assessment

Regarding risk of bias, four studies (30.8%) [19, 37, 38, 40] were rated as having an overall high risk of bias because at least one domain was rated as high risk. Nine studies (69.2%) [18, 33–36, 39, 41–43] were classified as having some concerns, primarily due to the use of self-reported outcome measures, potentially introducing self-report bias. Specifically, four studies (30.8%) did not adequately report the randomization process. Many studies failed to use the intention-to-treat analysis to address missing data (deviations from intended interventions: 10 studies, 76.9%). The missing data observed in some studies may reflect underlying true values (missing outcome data: 5 studies, 38.5%). In addition, the majority of studies (11 studies, 84.6%) were either not prospectively registered or exhibited discrepancies between the reported outcomes and their pre-registered analysis plans. The Cohen’s kappa for the risk of bias assessment was approximately 0.806, with a percentage agreement of 92.3%, indicating a high level of inter-rater reliability between the two researchers. The detailed risk of bias assessment results are shown in Fig. 2.

Fig. 2.

Risk-of-bias graph and summary. (A) Overall risk of bias, with each category presented as percentages. (B) Risk of bias of the studies included in the meta-analysis

Outcomes

Negative emotions

Depression

Thirteen studies (n = 1310) reported a significant positive effect of ACT on depression at post-test [SMD = − 0.66 (− 0.80, − 0.52), P <.001, I² = 24% (0%, 75%), Fig. 3]. Six studies (n = 466) reported a significant positive effect of ACT on depression at follow-up [SMD = − 0.63 (− 0.98, − 0.29), P <.001, I² = 64% (11%, 98%), Fig. 3]. Overall, ACT significantly reduced depression symptoms at both post-test and follow-up compared to controls.

Fig. 3.

Forest plot of meta-analysis for the effect of ACT on depression

Anxiety

Five studies (n = 772) showed a significant positive effect of ACT on anxiety at post-test [SMD = − 0.43 (− 0.77, − 0.10), P <.05, I² = 77% (34%, 97%), Fig. 4] compared to control groups. Four studies (n = 396) showed a significant positive effect of ACT on anxiety at follow-up [SMD = − 0.40 (− 0.67, − 0.13), P <.001, I² = 42% (0%, 96%), Fig. 4].

Fig. 4.

Forest plot of meta-analysis for the effect of ACT on anxiety

Automatic thoughts

Four studies (n = 316) reported the effect of ACT on automatic thoughts at post-test, but follow-up data were unavailable. The results showed that ACT had no significant effect on automatic thoughts in depressed patients compared to the control group [SMD = − 0.28 (− 0.69, 0.12), P =.17, I² = 65% (0%, 97%), Fig. 5].

Fig. 5.

Forest plot of meta-analysis for the effect of ACT on automatic thoughts

Psychological flexibility

Eleven studies (n = 1226) reported a significant positive effect of ACT on psychological flexibility at post-test [SMD = 0.50 (0.35, 0.66), P <.001, I² = 36% (0%, 83%), Fig. 6]. Four studies (n = 400) reported that the positive effect of ACT on psychological flexibility was maintained in follow-up [SMD = 0.53 (0.33, 0.73), P <.001, I² = 0% (0%, 74%), Fig. 6]. Overall, ACT significantly improved psychological flexibility in depressed patients at both post-test and follow-up compared to the control group.

Fig. 6.

Forest plot of meta-analysis for the effect of ACT on psychological flexibility

Acceptability: (Only dropouts at post-test assessment)

Thirteen studies (n = 1362) reported dropout data for participants in both the ACT and control groups. The pooled results showed no statistically significant difference in acceptability between the ACT group and passive control group [RR = 1.35 (0.80, 2.28), P =.25, I² = 48% (0%, 81%), Fig. 7]. Overall, ACT demonstrated similar acceptability compared to passive control groups.

Fig. 7.

Forest plot of meta-analysis for the acceptability of ACT

Subgroup analysis

The results of subgroup analyses are listed in Tables S1 to S7 in Appendix 3 of the supplementary material. The subgroup analysis examined potential moderators influencing the efficacy of ACT in improving depression, anxiety, automatic thoughts, and psychological flexibility in patients with depression. The key findings are as follows: (1) Delivery format: Face-to-face ACT [SMD = − 0.66 (− 0.99, − 0.32), P <.001, I² = 0%, Power = 97.87%] was significantly more effective in reducing anxiety (follow-up) compared to internet-based ACT [SMD = − 0.22 (− 0.47, 0.03), P =.09, I² = 0%, Power = 41.01%]. (2) Assessment measures: The subgroup analysis based on assessment measures revealed a significant difference in depression (follow-up) across subgroups (P <.05), suggesting that variations in assessment measures may contribute to the observed heterogeneity. (3) Number of sessions: Although the subgroup analysis based on the number of sessions indicated a significant difference in anxiety (post-test) across subgroups (P <.05), the statistical power for both groups (≤ 5 sessions and ≥ 10 sessions) was extremely low (5%). Therefore, there is insufficient evidence to conclude that ACT interventions consisting of 5 to 10 sessions yield superior anxiety improvement in patients with depression. Notably, due to the limited number of studies within subgroups and the insufficient robustness of some findings (Power < 80%), these results should be interpreted with caution.

Sensitivity analysis

The sensitivity analysis showed that sequentially excluding primary outliers [36, 37, 39, 42] reduced heterogeneity in depression, follow-up anxiety, and psychological flexibility outcomes, while effect sizes remained stable. However, excluding Kyllönen (2018) [19] substantially altered the effect size for automatic thoughts, indicating limited robustness for this outcome. Notably, the removal of Fledderus (2013) [34] led to a directional reversal in post-test anxiety, suggesting a strong dependence of the combined effect size on this study. These findings highlight the need for cautious interpretation of post-test anxiety and automatic thought outcomes. Full sensitivity analysis results are presented in Table S8 (Appendix 4).

Publication bias test

Publication bias was assessed for all outcomes using contour-enhanced funnel plots, as shown in Figure S9 to S16 in Appendix 5 of the supplementary material. The funnel plots revealed that most studies concentrated on one side of the effect distribution (either positive or negative), with the opposite side underrepresented, indicating potential publication bias. To further investigate this, Egger’s tests were conducted and revealed significant small-study effects for depression at post-test (P <.05), indicating a potential risk of publication bias. To assess and adjust for this bias more robustly, we applied PET-PEESE, selection models, and RoBMA-PSMA. For post-test depression, the effect sizes (ES) estimated by PET-PEESE, selection models, and RoBMA-PSMA were − 0.885 (− 1.070, − 0.699), − 0.707 (− 0.868, − 0.546), and − 0.617 (− 0.758, − 0.418), respectively, all indicating significant negative effects (P <.001, BF₁₀ = 127.027) and a robust intervention effect, as detailed in Tables S9 to S16 of Appendix 5.

GRADE assessment

The GRADE assessment indicates that Acceptance and Commitment Therapy (ACT) consistently outperforms the control group in improving depression, anxiety, automatic thoughts, and psychological flexibility in patients with depression, with the certainty of evidence ranging from very low to moderate. The GRADE Evidence Profile and Summary of Findings table can be found in Appendix 6.

For depression, ACT demonstrated a moderate effect in both post-test and follow-up (low certainty). However, due to potential issues related to blinding and publication bias, the certainty of the evidence is limited by risk of bias and inconsistency. For anxiety, ACT showed a small to moderate effect at post-test (moderate certainty) and follow-up (low certainty). The certainty of the evidence is constrained by inconsistency and imprecision due to limitations in blinding and insufficient sample size. For automatic thoughts, ACT had only a minimal effect at post-test (very low certainty). The certainty of the evidence is restricted by risk of bias, inconsistency, and imprecision, primarily due to issues related to randomization, blinding, and small sample size. For psychological flexibility, ACT demonstrated a sustained moderate effect in both post-test and follow-up (moderate certainty). However, the certainty of the evidence is affected by risk of bias due to limitations in blinding.

Discussion

This meta-analysis systematically evaluated the effects of ACT on negative emotions, automatic thoughts, and psychological flexibility in patients with depression, along with the sustainability of its positive effects at follow-up. It also analyzed differences in the acceptability of ACT compared to control groups. The findings indicate that ACT significantly improves depression, anxiety, and psychological flexibility in patients with depression, and these benefits are sustained at follow-up. However, no significant effect of ACT on automatic thoughts was observed. Additionally, we found no significant difference in the acceptability of ACT compared to passive control groups.

Depressed patients often fuse with their negative self-evaluations, making cognitive defusion essential for altering their relationship with thoughts, emotions, and bodily sensations [9]. Unlike CBT, which aims to directly modify negative cognitions, ACT does not attempt to change cognition itself. Instead, it emphasizes accepting negative thoughts and emotions, thereby enhancing emotional and behavioral regulation [44] and improving symptoms of depression and anxiety. The results of this meta-analysis indicate that ACT significantly reduces depression and anxiety following the intervention, with effects that persist at follow-up. Previous studies have similarly demonstrated that ACT improves both depression and anxiety [45, 46]. One possible explanation is that ACT promotes the development and enhancement of psychological flexibility in patients with depression, which in turn mediates the alleviation of both depression and anxiety. This positive effect appears to be sustained over the long term. Additionally, a recent three-level meta-analysis suggests that mindfulness-based interventions (e.g., ACT) improve mental health, potentially through changes in neuroticism that mediate alterations in trait mindfulness (e.g., psychological flexibility), ultimately influencing mental health outcomes such as depression and anxiety [47]. Through ACT training, individuals are guided to redefine their personal values and to commit to actions aligned with these values [8], which in turn leads to long-term, stable control over depression and anxiety.

Regarding automatic thoughts, the available evidence suggests that ACT does not have a significant effect on automatic thoughts in patients with depression. Due to cognitive dysfunction, individuals with depression exhibit impaired control over negative automatic thoughts, which may contribute to rumination—a persistent thinking pattern that further exacerbates depressive symptoms [17, 48]. Previous studies have reported that internet-based ACT does not significantly impact automatic thoughts in patients with depression, although it does exert significant effects on other mental health-related variables, such as depression [39]. Additionally, a study on iACT for depression [10] found that while the intervention altered cognitive fusion, it did not reduce the frequency of automatic thoughts. These findings are consistent with the results of the present meta-analysis. Moreover, some studies have indicated that neuroticism is a distal vulnerability factor for depression, increasing the risk of experiencing negative automatic thoughts in individuals with depression [49]. Meanwhile, ACT has been shown to reduce neuroticism in patients with mental disorders [50], and this reduction in neuroticism may have a nonspecific effect on automatic thoughts, potentially decreasing their frequency [47]. It is crucial to emphasize that although our findings suggest no significant effect of ACT on automatic thoughts in patients with depression, this conclusion remains uncertain because of the limited number of studies included (K = 4) and the very low certainty of evidence (GRADE: very low). High-quality, long-term follow-up studies are needed to further evaluate the impact of ACT on automatic thoughts in individuals with depression.

This meta-analysis demonstrates that ACT significantly enhances psychological flexibility in patients with depression, with effects persisting at follow-up. Psychological flexibility is a central concept in the ACT treatment model, encompassing the ability to engage with the present moment in a flexible and non-avoidant manner, accepting one’s internal experiences while taking actions aligned with personal values [8]. It is well established that patients with depression often experience cognitive disorders, which significantly contribute to reduced social functioning and poor clinical outcomes [51]. ACT helps patients shift their focus to the present moment and accept unchangeable circumstances [9], gradually fostering cognitive defusion and enhancing psychological flexibility. This meta-analysis found that the effect size of psychological flexibility at post-intervention (SMD = 0.50, 95% CI [0.35, 0.66]) was highly consistent with the findings of a previous three-level meta-analysis (SMD = 0.53, 95% CI [0.35, 0.73]) [52]. Notably, although the estimated effect sizes were similar, the study populations differed substantially. The earlier three-level meta-analysis included a mixed sample comprising individuals with psychiatric disorders, non-psychiatric conditions, or other medical issues, whereas the present meta-analysis focused exclusively on patients with depression. This targeted population may provide more clinically relevant and representative evidence for the effectiveness of ACT in enhancing psychological flexibility among individuals with depression. Importantly, the current meta-analysis extends previous research by systematically evaluating the long-term effects of ACT on psychological flexibility in individuals with depression. By incorporating follow-up data into the analysis, we found that ACT continues to enhance psychological flexibility at follow-up. Interestingly, earlier meta-analyses and systematic reviews involving ACT have reported similar findings. These include a moderate effect of ACT on psychological flexibility in family caregivers [53], as well as an increased effect of group ACT on psychological flexibility in patients with emotional disorders during follow-up [54]. This may be attributed to ACT’s use of metaphors and mindfulness techniques, which assist patients in clarifying their values and establishing short-term, medium-term, and long-term goals, gradually activating behavior and leading to sustained mental health benefits [9]. Therefore, as a flexible psychological intervention model, ACT is suitable for broader clinical implementation.

In terms of acceptability, there was no statistically significant difference between the ACT group and passive control group. Although the dropout rate in the ACT group was slightly higher (12.8% vs. 8.9%), the difference was not statistically significant (P =.25), which is likely due to the lower time and effort burden in the control group. A previous meta-analysis [55] reported similar results, showing a slightly higher dropout rate in the ACT-inclusive group compared to the inactive control group (16.0% vs. 9.5%), though this difference was also nonsignificant (P =.072). Additionally, a systematic review [56] found that third-wave CBT, including ACT, had acceptability comparable to waitlist controls and treatment as usual in managing acute depression. These findings suggest that participants in both groups were equally likely to discontinue the study for various reasons, indicating generally good acceptability. Although dropout rates in the ACT group did not significantly differ from those in inactive control condition, it is important to consider potential factors influencing psychotherapy acceptability, such as therapist experience, intervention feasibility, study design, and dropout definitions [20]. A meta-analysis of 68 studies identified therapist experience as a key moderator [55], with higher dropout rates in ACT interventions delivered by inexperienced master’s-level therapists (26.4%) compared to experienced psychologists (12.4%). Initial findings like these lay the foundation for future research to explore similar issues. Given that the dropout rate is significantly related to the effectiveness of ACT, future research should focus on optimizing the ACT intervention design to better align with participants’ preferences. Co-design, which actively involves patients in shaping psychological interventions, has been shown to improve effectiveness and acceptability [57] but is rarely applied in ACT development. This could provide a valuable perspective for optimizing the design of ACT and improving its acceptability among patients with depression. Additionally, strengthening therapist training or ensuring that experienced psychologists deliver ACT may enhance patient experiences and reduce dropout rates [55]. Another key strategy for improving acceptability is optimizing intervention procedures to minimize participants’ time and effort burden. However, it is important to note that this study assessed acceptability solely based on dropout rates. While dropout can serve as an indirect measure of acceptability, it primarily reflects extreme cases of non-acceptance and is susceptible to various external influences, including limited therapist experience, time constraints, and financial burden. Future research should employ more comprehensive measures to better assess ACT’s acceptability.

Despite the limitations of our subgroup analysis, such as small sample sizes and low statistical power in certain subgroups, we identified findings with potential clinical significance. For example, face-to-face ACT may be more effective than internet-based ACT in alleviating anxiety symptoms in patients with depression. This is consistent with a network meta-analysis [58], which ranked face-to-face ACT as the most effective intervention format. Several factors may help explain this finding. First, ACT emphasizes helping individuals experience negative thoughts, emotions, and sensations rather than avoiding them. The therapeutic process involves numerous experiential exercises [18], such as mindfulness practices, which are more effectively implemented under the direct supervision of a therapist. Face-to-face interactions also provide immediate emotional support and feedback—elements typically lacking in internet-based ACT—which may contribute to its lower efficacy. Second, face-to-face ACT allows therapists to tailor interventions to patients’ specific needs, enabling a more personalized and flexible treatment experience. Regular therapist guidance also plays a key role in motivating patients to complete treatment tasks, thereby improving adherence and overall effectiveness [58]. Furthermore, face-to-face ACT fosters a stronger therapeutic alliance and sense of trust. A secure attachment pattern not only facilitates self-exploration and self-disclosure but also enhances collaboration between the patient and therapist, ultimately improving treatment outcomes [59]. However, the widespread implementation of face-to-face ACT is constrained by factors such as high treatment costs, geographical barriers, a shortage of experienced therapists, and time limitations [60]. Given the rapid advancement of digital interventions, integrating the strengths of face-to-face delivery into internet-based ACT could enhance its effectiveness, improve cost-efficiency and accessibility, and provide broader psychological support for patients.

Beyond the delivery format, we found that study heterogeneity may partly stem from differences in measurement tools. Using different measurements to assess the same psychological construct can reduce the comparability of results. To address this, we conducted subgroup analyses based on measurement tools and applied a random-effects model to account for inter-study heterogeneity. Additionally, we used standardized mean differences (SMD) to adjust for outcomes across different scales, thereby mitigating the impact of measurement diversity on pooled effect sizes. For depression assessment, commonly used scales include the Beck Depression Inventory-II (BDI-II) [61], the Center for Epidemiologic Studies Depression Scale (CES-D) [62], the Patient Health Questionnaire-9 (PHQ-9) [63], and the Hospital Anxiety and Depression Scale (HADS) [64]. A systematic review [65] following COSMIN guidelines found that BDI-II and PHQ-9 demonstrated better internal consistency than other depression measures, making them the most promising tools for assessing depression severity. For anxiety assessment, frequently used scales include the Beck Anxiety Inventory (BAI) [66], the Anxiety Subscale of the Hospital Anxiety and Depression Scale (HADS-A) [64], and the Generalized Anxiety Disorder-7 (GAD-7) [67]. A COSMIN-guided systematic review [68] compared 17 instruments for assessing anxiety symptoms in patients with mental disorders and recommended BAI as the preferred tool for evaluating anxiety symptoms. Regarding the assessment of automatic thoughts, the most frequently used instrument is the Automatic Thoughts Questionnaire (ATQ-F) [69], which demonstrated high reliability (Cronbach’s α = 0.80 ~ 0.94) in the studies included in this meta-analysis. For psychological flexibility, a scoping review [70] identified commonly used measures, including various versions of the Acceptance and Action Questionnaire (especially AAQ-II) [26], the Comprehensive Assessment of Acceptance and Commitment Therapy Processes (CompACT) [71], the Multidimensional Psychological Flexibility Inventory (MPFI) [72], and the Personalized Psychological Flexibility Index (PPFI) [73]. Despite AAQ-II’s strong content validity and internal consistency, concerns have emerged about its unidimensional structure and its overlap with neuroticism and psychological distress measures [74]. Kashdan et al. (2020) [73] found that PPFI, validated across seven studies, had better discriminant validity than AAQ-II, particularly in distinguishing negative emotions, psychopathology, and distress. PPFI also aligned more closely with the conceptual definition of psychological flexibility.

Based on a review of existing studies, we propose the following recommendations for standardizing measurement tools in future ACT intervention research: the BDI-II and PHQ-9 are recommended for depression measurement, the BAI for anxiety evaluation, the ATQ-F for automatic thoughts due to its strong reliability and validity in depression, and the PPFI for psychological flexibility. These recommendations aim to reduce measurement variability and improve study comparability. However, they are based on current research and require further validation in future clinical practice and studies.

Strengths and limitations

This meta-analysis has several strengths. First, it examined both immediate post-test and follow-up outcomes, providing valuable insights into the persistence of ACT’s effects over time. Second, it included automatic thoughts as an outcome, which had not been previously investigated. Third, key factors, such as overall risk of bias (ROB2), delivery format, measures, sample size, and the number of sessions, were selected as influential moderators for subgroup analyses. Additionally, power analyses were conducted for each subgroup. Fourth, more robust statistical methods, including PET-PEESE, selection models, and RoBMA-PSMA, were employed to assess publication bias and correct for effect sizes. Fifth, the GRADE approach was used to evaluate the certainty of the evidence. These analyses were conducted to enhance the robustness and certainty of the meta-analysis findings.

However, this meta-analysis also has limitations. First, due to the limited number of included studies, some outcomes, such as automatic thoughts, were derived from only a few studies and exhibited considerable heterogeneity. Although we conducted subgroup analyses to explore potential sources of heterogeneity, these findings should be interpreted with caution. Second, the inclusion of both clinically diagnosed and subclinical depression patients introduced heterogeneity in the sample, which may have contributed to ambiguity about its representativeness. Third, the measure of acceptability was based solely on dropout rates, which may not fully reflect intervention acceptability. Dropout may also be influenced by confounding factors, including time constraints, financial burden, and therapist experience.

Conclusion

The findings of this meta-analysis support the efficacy of ACT-based interventions in improving depression, anxiety, automatic thoughts, and psychological flexibility in individuals with depression. Notably, the acceptability of ACT was found to be comparable to that of passive control groups, suggesting good patient acceptance of ACT. Future research should consider employing a co-design approach to tailor ACT interventions to better align with patient preferences, thereby enhancing acceptability. Additionally, confounding factors such as time constraints, financial burden, and therapist experience that affect ACT acceptability should be considered.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ACT

Acceptance and Commitment Therapy

AAQ-II

Acceptance and Action Questionnaire-II

ATQ-F

The Automatic Thoughts Questionnaire

BAI

Beck Anxiety Inventory

BDI-II

Beck Depression Inventory-II

CBT

Cognitive Behavioral Therapy

CES-D

Center for Epidemiologic Studies Depression Scale

CI

Confidence Interval

CompACT

Comprehensive Assessment of Acceptance and Commitment Therapy Processes

GAD-7

Generalized Anxiety Disorder-7

HADS

Hospital Anxiety and Depression Scale

HADS-A

Hospital Anxiety and Depression Scale-Anxiety

MADRS

Montgomery-Asberg Depression Rating Scale

MPFI

Multidimensional Psychological Flexibility Inventory

PHQ

Patient Health Questionnaire

PPFI

Personalized Psychological Flexibility Index

RADS-2

Reynolds Adolescent Depression Scale-2

RCTs

Randomized Controlled Trials

RR

Relative Risk

SMD

Standardized Mean Difference

Author contributions

YX. Z contributed to the study concept and design. YX.Z, RX.W and XC.X jointly screened the literature and extracted the relevant content of the literature. YX.Z, XC.X, and C.B reviewed and analyzed the results. YX.Z and SR.Y wrote original draft, and YH.Z revised it. All authors contributed to revising and approving the final version of the manuscript.

Funding

This work was funded by Nanjing Health Science and Technology Development Special Fund Project (grant number: YKK24192, YKK23135 and YKK23142). The funder had no involvement in the design of the study, collection, analysis, and interpretation of data, and the writing of the manuscript.

Data availability

The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.