Individual behavioral activation in the treatment of depression: A meta analysis

ABSTRACT

Objective

Behavioral activation (BA) is an extensively examined treatment for depression which is relatively simple to apply in comparison to other psychotherapies. BA aims to increase positive interactions between a person and the environment. All previous meta-analyses focused on BA in groups and guided self-help, but none focused on BA in individual psychotherapy. The goal of the current meta-analysis is to examine the pooled effects of trials comparing individual BA to control conditions.

Methods

We conducted systematic searches and conducted random effects meta-analyses to examine the effects of BA.

Results

We included 22 randomized controlled trials (with 819 patients) comparing individual behavioral activation with waitlist, usual care, or other control conditions on distal treatment outcomes. Nine studies were rated as low risk of bias. We found a large effect (Hedges’ g = 0.85; 95% CI: 0.57; 1.1) with high heterogeneity (75%; 95% CI: 62; 83). When only studies with low risk of bias were considered, the effect size was still significant (g = 0.56; 95% CI: 0.09; 1.03), with high heterogeneity (I² = 80%; 95% CI: 66; 89; prediction interval: −0.85; 1.98).

Conclusion

BA is an effective, relatively simple type of therapy that can be applied broadly in differing populations/.

Keywords:

• depression
• behavioral activation
• meta-analysis
• psychotherapy
• psychological treatment
• psychotherapy method
• psychotherapy outcome

Clinical or methodological significance of this article: Behavioral Activation (BA) is effective in the treatment of adult depression, with effect sizes that are comparable to those found for as other psychotherapies for depression, such as CBT and interpersonal psychotherapy. It can be used briefly in a few sessions as part of a more extensive treatment package, or more extensively as an independent treatment. BA is especially useful in more severe depression, when more insight-oriented therapies are not feasible, because it is a straightforward method. Accounting for personal preferences and cultural identities of the patient is important, because BA is flexible and can be adapted to the patients’ situation and needs.

Introduction

With about 280 million people suffering from depressive disorders, depression is the most prevalent mental disorder (World Health Organization, 2022). Furthermore, depression is known to be associated with considerable excess mortality and morbidity. Apart from the personal suffering by patients and their relatives, depression is also associated with huge economic costs. Because it affects people in their working age, the costs associated with production losses are especially high (Herrman et al., 2022). Depression is considered to be one of the main major public health challenges of the coming decades (World Health Organization, 2017).

Several psychological treatments are available for depression, which are usually considered to be first-line treatments. These include cognitive behavior therapy, interpersonal psychotherapy, non-directive counseling, problem-solving therapy, psychodynamic therapy, and behavioral activation (Cuijpers et al., 2021). Behavioral activation (BA) is one of the best examined methods for treating depression. In this study we will mostly focus on BA as an independent treatment, and only briefly on treatment packages where BA is only one of the components.

BA can be briefly defined as a psychological method aimed at increasing positive interactions between a person and the environment (Cuijpers et al., 2020). A more extensive and precise definition was given in an authoritative review of the origins and current status of BA (Dimidjian et al., 2017, pp. 3–4): a “structured, brief psychotherapeutic approach that aims to (a) increase engagement in adaptive activities (which often are those associated with the experience of pleasure or mastery), (b) decrease engagement in activities that maintain depression or increase risk for depression, and (c) solve problems that limit access to reward or that maintain or increase aversive control.”

Four types of BA can be distinguished (Cuijpers et al., 2019; Mazzucchelli et al., 2009): (1) Pleasant activity (consisting of monitoring and scheduling pleasant activities, sometimes including social skills training; Lewinsohn, 1974); (2) Self-control (monitoring of activities and mood, goal setting, self-evaluation of performance and self-administering rewards; Rehm, 1984); (3) Contextual BA (activity scheduling, self-monitoring, graded task assignment, role playing, functional analysis, mental rehearsal and mindfulness; Jacobson et al., 2001; Martell et al., 2001); and (4) Behavioral activation treatment for depression (contracting to change immediate environmental contingencies, goal setting and graduated task assignment, monitoring, and self-administering rewards; Lejuez et al., 2001).

One of the most interesting aspects of BA is that it is relatively simple compared to full CBT. The basic idea of BA is that therapists help patients find which activities are related to a bettemood and then work to build more of these pleasant activities into the patients’ lives. This method does not require higher-level patterns of thoughts, which for many patients is quite complicated and an intensive learning process. Therapists do not need, therefore, necessarily to be fully trained psychotherapists or clinical psychologists. One seminal randomized trial conducted in the UK showed that BA delivered by nurses in primary care was not inferior to full CBT delivered by highly trained therapists (Richards et al., 2016).

Because BA is relatively easy for both therapists and patients, it has also been used with patients suffering from other disorders and more complicated population. These groups include caregivers of dementia patients (Teri et al., 1997), children with physical disabilities (Suh et al., 2021), adults with PTSD (Wagner et al., 2019), bereavement (Acierno et al., 2021), anxiety and depression in cancer survivors, concurrent methamphetamine dependence and sexual risk for HIV acquisition among men who have sex with men (Mimiaga et al., 2019), and as a preventive intervention in college students (Fazzino et al., 2020).

Although the effects of BA were examined in several randomized trials in the 1970s, all of these early studies were small and did not have sufficient statistical power to find small effects. Several later trials have proven important for the development of BA. The study by Jacobson et al. (1996) was relatively large compared to the previous studies and found no indication that BA was less effective than CBT. Building on these results, another influential trial was conducted and again found no significant difference between BA and CBT, although BA was more effective with more severe depression (Dimidjian et al., 2006). A third significant trial conducted in the UK with 440 randomized patients showed that BA delivered by junior mental health workers in primary care was as effective as CBT delivered by psychotherapists (Richards et al., 2016).

Several recent meta-analyses have integrated the findings from randomized trials of BA as an independent treatment package on distal, end of treatment outcomes. The largest, most recent meta-analysis included 28 trials and found a pooled effect size (Hedges’ g) of 0.83 on depression, 0.37 on anxiety, and 0.64 on activation (Stein et al., 2021). Although the quality of most of the trials was modest, this meta-analysis suggested that BA is an effective treatment for depression as well as other outcomes.

A larger network meta-analysis included 331 RCTs comparing eight major types of psychotherapy for depression with each other or with control groups (waitlist, usual care, pill placebo) on distal, end of treatment outcomes (Cuijpers et al., 2021). A total of 36 trials included a BA treatment arm. No significant difference was found between BA and other major types of therapy such as CBT and Interpersonal Psychotherapy. A more specific, but smaller network meta-analysis included studies that compared treatments including BA (as independent treatment package), cognitive restructuring (CR), or the combination of BA and CR for depression, with each other or with waitlist or usual care (Ciharova et al., 2021). This meta-analysis suggested that all three had comparable effects on depression, and no significant outcome differences were found.

In recent years, several smaller meta-analytic studies of treatments including BA have been conducted. Most of these meta-analyses focused on specific target groups, such as older adults (Orgeta et al., 2017), young people (Tindall et al., 2017), and dementia and informal caregivers (Zabihi et al., 2020). Other meta-analyses focused on specific treatment formats of BA, including internet-based BA (Huguet et al., 2018) and group BA (Simmonds-Buckley et al., 2019). Another recent meta-analysis focused on BA for patients with co-occurring depression and substance use disorders (Pott et al., 2022). The number of randomized trials included in these narrower meta-analyses was relatively small, but most found moderate to large effects when comparing BA to usual care or waitlist control groups and confirmed the results of the larger (network) meta-analyses (although the effects in co-occurring depression and substance use disorders were small and non-significant; Pott et al., 2022).

All of the previous meta-analyses focused on BA in groups and guided self-help, but none focused on BA in individual psychotherapy. We decided therefore to conduct a meta-analysis of individual BA on distal, end of treatment outcomes in adult depression.

Method

Identification and Selection of Studies

The current study is part of a larger meta-analytic project on psychological treatments of depression that was registered at the Open Science Framework (Cuijpers & Karyotaki, 2020). Supplemental materials are available at the project’s website (www.metapsy.org). This database has been used in a series of earlier published meta-analyses (Cuijpers, 2017).

The studies included in the current review were identified through the larger, already existing database of randomized trials on the psychological treatment of depression. For this database we searched four major bibliographical databases (PubMed, PsycInfo, Embase, Cochrane Library) by combining terms (both index terms and text words) indicative of depression and psychotherapies, with filters for randomized controlled trials. The full search strings can be found at the website (www.metapsy.org). Further, we checked the references of earlier meta-analyses on psychological treatments of depression. The database is continuously updated and was developed through a comprehensive literature search (from 1966 to 1 January 2022). All records were screened by two independent researchers, and all papers that could possibly meet inclusion criteria according to one of the researchers were retrieved as full-text. The decision to include or exclude a study in the database was also done by the two independent researchers, and disagreements were resolved through discussion.

For this meta-analysis, we included randomized controlled trials in which BA in an individual format for adults with depression were compared with a control condition (waitlist, care-as-usual, other non-active control groups). We operationalized BA as an intervention in which the registration of pleasant activities and the increase of positive interactions between a person and his or her environment were the core elements of the treatment (Cuijpers et al., 2020). Social skills training could be a part of the intervention. Cognitive behavioral interventions in which also other components, such as cognitive restructuring were included, were not included in this meta-analysis. Depression could be defined as meeting criteria for a depressive disorder according to a diagnostic interview or as a score above the cut-off on a self-report depression measure.

Quality Assessment and Data Extraction

We assessed the validity of included studies using four criteria of the Risk of Bias (RoB) assessment tool (version 1, developed by Cochrane Collaboration; Higgins et al., 2011). The RoB tool assesses possible sources of bias in randomized trials including: the adequate generation of allocation sequence; the concealment of allocation to conditions; the prevention of knowledge of the allocated intervention (masking of assessors); and dealing with incomplete outcome data (this was assessed as positive when intention-to-treat analyses were conducted, meaning that all randomized patients were included in the analyses). Assessment of the validity of the included studies was conducted by two independent researchers, and disagreements were solved through discussion.

We coded participant characteristics (diagnostic method; recruitment method; target group; mean age; proportion of women); characteristics of BA (number of sessions) as well as general characteristics of the studies (type of control group; publication year; country). Extended definitions of the moderators are available at the website of the meta-analytic project from which these meta-analytic data were drawn (www.metapsy.org). Because we have collected these data over a period of 15 years and because the number of trials in our database is very large (currently more than 900 RCTs), reliabilities across different raters have not been collected.

Outcome Measures

For each comparison between BA and a control condition, the effect size indicating the difference between the two groups at post-test in terms of distal outcomes (depressive symptoms) was calculated (Hedges’ g; Hedges & Olkin, 1985). Effect sizes were calculated by subtracting (at post-test) the average score of the BA group from the average score of the control group and dividing the result by the pooled standard deviation. Because some studies had relatively small sample sizes, we corrected the effect size for small sample bias. When means and standard deviations were not reported we calculated the effect size using dichotomous outcomes; and if these were not available either, we used other statistics (such as t-value or p-value) to calculate the effect size.

Meta-Analyses

The meta-analyses were conducted with the meta and metafor packages in R (version 4.1.1) and R studio (version 1.1.463 for Mac). The “metapsyTools” package (Harrer et al., 2022) was specifically developed for the meta-analytic project of which this study is a part. We calculated the pooled effect sizes in several different ways so that we could explore if different methods of pooling outcomes resulted in different outcomes. We pooled all outcomes within a study, before pooling across studies. To pool effect sizes within studies, an intra-study correlation coefficient of ρ = 0.5 was assumed. This method of pooling the effect sizes was the main method of analyses.

In addition to these main analyses, we also calculated the pooled effect size in several different ways to ensure that all methods of pooling pointed in the same direction: (2) We also calculated the pooled effect size based on a multilevel (three-level) model (effect sizes nested in studies), using the three-level “correlated and hierarchical effects” (CHE) model, which was recently proposed by Pustejovsky and Tipton (2022); parameter tests and confidence intervals of which are also calculated using RVE to guard against model misspecification. We assumed an intra-study correlation of ρ = 0.5 for this model. (3) Then we calculated the pooled effect size with the smallest outcome per study and (4) the pooled effect size with the largest outcome per study. (5) The pooled effect size while excluding outliers, using the “non-overlapping confidence intervals” approach, in which a study is defined as an outlier when the 95% CI of the effect size does not overlap with the 95% CI of the pooled effect size (Harrer et al., 2022). (6) The pooled effect size with influential cases excluded according to the methods developed by Viechtbauer and Cheung (2010). (7) The pooled effect size adjusted for publication bias, using Duval and Tweedie’s trim and fill procedure (Duval & Tweedie, 2000), which yields an estimate of the effect size after correction for the funnel plot asymmetry. (8) The pooled effect size from studies with low risk of bias. (9) In addition, as another sensitivity analysis, we calculated the pooled effect size under the assumption that all effect sizes are independent. Random effects models were conducted in which the between-study heterogeneity variance was estimated using the restricted maximum likelihood method. We applied the Knapp-Hartung method to obtain robust confidence intervals and significance tests of the overall effect (IntHout et al., 2014). As a test of homogeneity of effect sizes, we calculated the I²-statistic and its 95% confidence interval, which is an indicator of heterogeneity in percentages, with 0% indicating no observed heterogeneity, 25% low, 50% moderate, and 75% high heterogeneity (Higgins et al., 2003). Because I² cannot be interpreted as an absolute measure of the between-study heterogeneity, we also added the prediction interval which indicates the range in which the true effect size of 95% of all populations will fall (Borenstein et al., 2009, 2017). In addition to Hedges’ g, we also calculated the Numbers-needed-to-treat (NNT; Furukawa, 1999), in which the control group’s event rate was set at a conservative 17% (based on the pooled response rate of 50% symptom reduction across trials in psychotherapy for depression; Cuijpers et al., 2021).

In addition to calculating the effect size after adjustment for publication bias trim and fill procedure, we tested for publication bias by inspecting the funnel plot on primary outcome measures and by Egger’s test of the intercept to quantify the bias captured by the funnel plot and to test whether it was significant.

Because the number of included studies was low and subgroup analyses usually have low statistical power (Cuijpers et al., 2021), we conducted only a small number of subgroup analyses that we considered essential: Type of control group, diagnosed mood disorder versus a cut-off on a self-report measure, recruitment method, adults in general versus a more specific target group, and risk of bias (low versus other).

Results

Selection and Inclusion of Studies

After examining a total of 30,889 records (21,563 after removal of duplicates), we retrieved 3,584 full-text papers for further consideration. We excluded 3563 of the retrieved papers. The PRISMA flowchart describing the inclusion process, including the reasons for exclusion, is presented in Figure 1. A total of 22 randomized controlled trials with 819 patients (399 in the BA conditions and 420 in the control conditions) met inclusion criteria.

Figure 1. PRISMA flow diagram.

Characteristics of Included Studies

A summary of key characteristics of the 22 included studies is presented in Table 1. In 7 trials participants met criteria for a depressive disorder according to a diagnostic interview, while the other 15 trials included participants who scored above a cut-off on a self-report depression scale. In 14 studies, usual care was used as control group, 5 used a waitlist control group and the 3 remaining studies used another control group (placebo, brief non-specific support, life style advice). Ten studies were conducted in the USA, 6 in Europe, and 6 in other countries. The number of sessions ranged from 1 to 24, with the majority (18 interventions) between 4 and 12 sessions.

Table 1. Selected characteristics of randomised trials comparing individual behavioral activation to control groups.

Study	Diagnosis	Nint	Nctr	Nsess	Control	Mean age	Prop women	Recr	Target group	SG	AC	BA	ITT	RoB	country
Choi et al., 2020	Self-report	90	94	5	Other^a	69	0.73	Other	Elderly	+	−	−	+	2	US
Chowdhary et al., 2016	Self-report	24	31	7	Cau	41	0.69	Other	Adults	+	−	sr	+	3	India
Dimidjian et al., 2006	Mood dis	43	53	24	Other^b	40	0.66	Comm	Adults	+	−	sr	+	3	US
Dimidjian et al., 2017	Self-report	86	77	10	Cau	29	1.00	Other	Perinatal	+	+	sr	+	4	US
Ekers et al., 2011	Mood dis	23	24	12	WL	45	0.62	Clin	Adults	+	+	sr	+	4	UK
Funderburk et al., 2021	Self-report	68	72	4	Cau	53	0.09	Clin	Veterans	−	+	sr	+	3	US
Gawrysiak et al., 2009	Self-report	14	16	1	Cau	18	0.80	Comm	Students	−	−	sr	+	2	US
Hemanny et al., 2020	Mood dis	24	26	12	Cau	40	0.86	Comm	Adults	+	−	+	+	3	Brasil
Jordans et al., 2019	Mood dis	60	60	4	Cau	nr	0.85	Other	Adults	+	+	sr	+	4	Nepal
Kanter et al., 2015	Mood dis	21	22	8	Cau	38	0.79	Clin	Minority	+	−	sr	+	3	US
McIndoo et al., 2016	Self-report	16	14	4	WL	19	0.62	Comm	Students	+	−	−	+	2	US
Nasrin et al., 2017	Mood dis	19	17	1	WL	36	0.59	Clin	Adults	+	+	sr	−	3	UK
Pagoto et al., 2013	Mood dis	68	76	18	Other^c	46	1.00	Comm	Overw w	+	+	+	+	4	US
Patel et al., 2017	Self-report	230	236	6	Cau	43	0.77	Clin	Adults	+	+	sr	+	4	India
Raue et al., 2019	Self-report	8	10	12	Cau	76	0.83	Other	Elderly	−	−	−	−	0	US
Richards et al., 2018	Self-report	15	14	8	Cau	65	0.48	Other	Gen. med.	+	+	sr	+	4	UK
Taylor & Marshall, 1977	Self-report	7	7	6	WL	22	0.71	Comm	Students	−	−	sr	−	1	Canada
Thomas et al., 2019	Self-report	15	12	8	Cau	66	0.40	Comm	Gen. med.	+	+	sr	+	4	UK
Turner et al., 1979	Self-report	17	17	5	Cau	25	0.50	Comm	Adults	−	−	sr	−	1	US
Vázquez et al., 2017	Self-report	22	19	4	Cau	58	0.93	Other	Caregivers	+	+	sr	+	4	Europe
Vázquez et al., 2020	Self-report	70	80	4	Cau	55	0.96	Other	Caregivers	+	+	sr	+	4	Europe
Wilson et al., 1983	Self-report	8	9	8	WL	40	0.80	Comm	Adults	−	−	sr	−	1	Australia

Abbreviations: AC: allocation concealment; BA: masking of assessors; Cau: Care-as-usual; Clin: clinical; Comm: community; Gen. med: general medical patients; ITT: intention to treat analyses; Mood dis: diagnosed mood disorder; N_ctr: Number of participants in the control group; N_int: Number of participants in the intervention group; N_sess: Number of sessions; Overw w: overweight women; RoB: risk of bias total score; SG: sequence generation; WL: waitlist.

Notes:

^a brief non-specific support.

^b Placebo.

^c Life style advice.

Sixteen of the 22 studies reported that the random numbers were generated correctly (adequate sequence generation; 73%); 11 reported allocation to conditions by an independent party (50%); 4 reported using blinded outcome assessors (18%) while 14 used only self-report outcomes (64%). In 17 studies, intent-to-treat analyses were conducted (77%). Nine studies (41%) met all criteria for low risk of bias, 9 other studies (41%) met 2 or 3 criteria, and 4 met only one criterion (18%).

Distal Outcomes for Individual BA on Depression

The results of the meta-analysis according to the different models are presented in Table 2. The forest plot is given in Figure 2. The main analyses indicated a large effect of BA on depression compared with control groups. Hedges’ g was 0.85 (95% CI: 0.57; 1.1), which corresponds to an NNT of 3.48. Heterogeneity was high (75%; 95% CI: 62; 83), and the prediction interval was broad (−0.25; 1.94).

Figure 2. Forest plot a) Abbreviations: bat: behavioral activation therapy; cau: care-as-usual; ctr: control; g: Hedges’ g; S.E.: standard error; vs: versus; wl: waiting list.

Table 2. The effects of individual Behavioral activation for depression according to different models.

Models	K	G	CI	I^2	CI	PI	NNT
Main outcomes	22	0.83	[0.55; 1.1]	76.05	[63.92; 84.1]	[−0.28; 1.93]	3.57
Three-Level Model (CHE)	30	0.82	[0.54; 1.1]	83.10	–	[−0.34; 1.99]	3.60
One effect size per study (lowest effect size)	22	0.76	[0.46; 1.06]	76.27	[64.29; 84.23]	[−0.44; 1.96]	3.95
One effect size per study (highest effect size)	22	0.87	[0.58; 1.16]	75.50	[63.01; 83.78]	[−0.26; 2.01]	3.35
Outliers removed	25	0.82	[0.62; 1.02]	64.27	[45.29; 76.67]	[0.04; 1.6]	3.59
Influence Analysis	30	0.82	[0.57; 1.07]	74.01	[62.85; 81.82]	[−0.3; 1.94]	3.62
Adjusted for publication bias	28	0.60	[0.29; 0.92]	78.60	[69.6; 84.9]	[−0.85; 2.06]	5.22
Only studies with low risk of bias	11	0.56	[0.09; 1.03]	80.46	[65.94; 88.79]	[−0.85; 1.98]	5.62
Assuming independence of effects	30	0.82	[0.57; 1.07]	74.01	[62.85; 81.82]	[−0.3; 1.94]	3.62

The majority of the other models indicated comparable pooled outcomes for BA, with only small differences from the main model. The overall outcomes of individual BA on distal, end-of-treatment outcomes were robust. However, the effect size was considerably smaller when only studies with low risk of bias were considered, although the effect size was still significant (g = 0.56; 95% CI: 0.09; 1.03). Heterogeneity was still high in these analyses (I²= 80%; 95% CI: 66; 89; prediction interval: −0.85; 1.98). Excluding outliers and influential cases did not reduce heterogeneity much.

We also found significant indications for publication bias. Egger’s test was significant (p = 0.02), and the effect size was considerably reduced after adjusting through trim and fill procedure (g = 0.60; 95% CI: 0.29; 0.92). Heterogeneity was still high after this adjustment (I²= 79%; 95% CI: 70; 85; prediction interval: −0.85; 2.06).

The results of the subgroup analyses are reported in Table 3. We found a significant difference between studies using different types of control groups (p < 0.001). Studies that used a waitlist control had a considerably larger effect size than studies using usual care or other groups. The other subgroup analyses examined the type of control group, diagnosis (mood disorder versus scoring above a cut-off), recruitment, target group, and risk of bias. None of these analyses pointed at significant differences between subgroups.

Table 3. Subgroup analyses.

		K	G	CI	I^2	CI	NNT
Control group	Usual care	13	0.69	[0.31; 1.08]	78.7	[64.1; 87.3]	4.43	<0.001
	Waitlist	5	1.47	[1.11; 1.83]	0	[0; 79.2]	1.90
	Other	4	0.70	[0.06; 1.34]	58.2	[0; 86.1]	4.35
Diagnosis	Mood	7	0.81	[0.29; 1.33]	73.6	[43.4; 87.7]	3.67	0.81
	Cut-off	15	0.87	[0.5; 1.25]	77.2	[62.6; 86]	3.37
Recruitment	Community	9	1.11	[0.71; 1.51]	62	[21.5; 81.6]	2.55	0.25
	Clinical	5	0.70	[−0.09; 1.49]	76.5	[42.8; 90.4]	4.35
	Other	8	0.68	[0.1; 1.25]	81.2	[64; 90.2]	4.50
Target group	Adults in general	9	0.95	[0.55; 1.35]	69	[37.9; 84.5]	3.05	0.46
	Other	13	0.76	[0.35; 1.18]	79.3	[65.3; 87.7]	3.95
Risk of bias	Low	9	0.68	[0.17; 1.18]	81.4	[65.7; 89.9]	4.50	0.27
	Other	13	0.98	[0.63; 1.34]	68.9	[44.8; 82.5]	2.94

Discussion

This meta-analysis shows that individual BA has significant end-of-treatment effects on depression. Although these effects may be somewhat overestimated because of publication bias and the suboptimal quality of several trials, the effects are still statistically significant and clinically meaningful after adjusting for these problems. The type of control group also affected the effect size, with waiting lists possibly overestimating the actual effects, which is in line with previous research on psychotherapies for depression (Furukawa et al., 2014; Cuijpers et al., 2021). However, several sensitivity analyses indicated that the effects found for BA are robust.

Few studies have focused on in-session outcomes of BA. One randomised trial indicated that 66% of patients receiving BA responded (their BDI score after treatment was at least 50% lower compared to baseline), whereas this proportion was 45% in the control group (Pagoto et al., 2013). Other trials have rarely reported response and remission rates of BA. Several studies did find high completion rates, treatment fidelity and treatment satisfaction (Funderburk et al., 2021; Gros et al., 2019).

In a study examining in-session outcomes of BA (Santos et al., 2017), it was found that in 79% of the patients participating in BA, changes in activation preceded or co-occurred with changes in depression, whereas this was not the case in patients receiving usual care. In another study among women with postpartum depression, it was found that sudden gains predicted lower depression scores after treatment (O'Mahen et al., 2017) and that participants with higher social functioning at baseline, did not experience a stressful life event during therapy, and completed more online modules were more likely to have a sudden gain. Sudden gains were defined as a large and stable drop in symptoms from one session to the next, and typically occurring early in treatment (Tang & DeRubeis, 1999). It was also found in this study that therapists focussed on specific, concrete content more in the session preceding the sudden gain than in the yoked sessions with women who did not have sudden gains. However, a more concrete therapist focus was only associated with having a sudden gain, not with the outcome on depressive symptoms at post-test. In another study of BA it was found that enactment of activities was significantly related to the likelihood of remission at post-treatment, and of maintaining improvement at 3-month follow-up (Meeks et al., 2019). More research is clearly needed to understand in-session processes and outcomes of BA.

Unfortunately, potential negative effects of BA have hardly been examined. Some of the trials in our meta-analysis report negative effects in patients receiving BA. Ekers et al. (2011) reported that none of the patients receiving BA deteriorated, whereas 16% in usual care deteriorated. In another study (Funderburk et al., 2021), the symptoms worsened in 10% of participants receiving BA, whereas 7% worsened in the control group. Other studies also reported serious adverse events (Patel et al., 2017; Raue et al., 2019; Richards et al., 2018; Thomas et al., 2019) but these numbers are very low (<5%) and in most cases not related to the treatment, such as hospital admissions, or equally divided across the treatment and control groups. Because of the small number of trials reporting negative effects, it is unfortunately not possible to make a reliable estimate of the proportion of patients who deteriorate during BA. Meta-analyses of psychotherapies in general show that this proportion is low and significantly smaller in treatment than in control conditions (Cuijpers et al., 2018, p. 2021).

The studies included in our meta-analysis varied considerably in terms of target samples. Only six of the 22 studies were aimed at adults in general, whereeas one study was specifically aimed at Latinos in the US (Kanter et al., 2015), one at veterans (Funderburk et al., 2021), one at women with obesity (Pagoto et al., 2013), two were aimed at older adults (Choi et al., 2020; Raue et al., 2019), two at college students (Gawrysiak et al., 2009; McIndoo et al., 2016), two at non-professional caregivers (Vázquez et al., 2017, 2020), and two at patients with comorbid general medical disorders (Richards et al., 2018; Thomas et al., 2019). While most studies were conducted in the US and Europe, four studies were conducted in low- and middle-income countries, two in India (Chowdhary et al., 2016; Patel et al., 2017), one in Nepal (Jordans et al., 2019) and one in Brasil (Hemanny et al., 2020). There were not enough studies in each of these specific groups to examine whether the effect sizes differed from those in other groups.

The number of trials examining the effects of individual BA is relatively small, and the quality of many of these trials is suboptimal. Although our meta-analysis indicated significant effects of individual BA treatment packages on distal outcomes, these results should be considered with caution because of the modest number and the low quality of these trials. However, overall individual BA is effective in the treatment of depression and previous research has shown that there are few indications that BA is less effective than other methods. We know that the entire BA treatment package works on distal outcomes, but we know little of its effects on immediate session or intermediate outcomes. Virtually all of the research conducted has focused on distal, end-of-treatment outcomes. It should be noted that this meta-analysis only focused on BA for depression. Future meta-analytic research should also focus on other mental health problems that can be treated with BA, as well as other outcomes next to depression.

We also do not we know which of the components of BA are responsible for its salutatory impact. The science has not made enough progress to say which of the components prove necessary. BA is often used as a stand-alone intervention for depression, but it is also often combined with cognitive restructuring and other components, such as problem-solving. It is not known what the relative contribution of BA is compared with other components of such broader cognitive behavioral packages. Research on working mechanisms and mediators of therapies has mostly been correlational and currently, no mechanism of BA or any other therapy or component has been sufficiently empirically validated (Cuijpers et al., 2019). Therefore, it is still unknown how therapies work. Additional research using additive or dismantling designs are probably necessary.

Author Note: This article is adapted, by special permission of Oxford University Press, from a chapter by the same authors in C. E. Hill & J. C. Norcross (Eds.) (2023), Psychotherapy skills and methods that work. The interorganizational Task Force on Psychotherapy Methods and Skills was cosponsored by the APA Division of Psychotherapy/Society for the Advancement of Psychotherapy.

Disclosure Statement

No potential conflict of interest was reported by the author(s).