Efficacy of Internet-Delivered Psychodynamic Psychotherapy

Research from Tomas Lindegaard PHD Ms.C. , Matilda Berg PHD Ms.C. and Gerhard Andersson PHD.

(2020). Psychodyn. Psi., (48)(4):437-454

Efficacy of Internet-Delivered Psychodynamic Therapy: Systematic Review and Meta-Analysis

Tomas Lindegaard PHD Ms.C. , Matilda Berg PHD Ms.C. and Gerhard Andersson PHD 

Recent years have seen an increase of internet-delivered interventions based on psychodynamic treatment models. To test the efficacy of internet-delivered psychodynamic therapy (IPDT), we conducted a systematic review and meta- analysis of randomized controlled trials. Following literature searches, we identified seven studies meeting inclusion criteria. The total number of participants was 528 in the treatment group and 552 in the control conditions. We found small effects favoring IPDT compared to inactive control conditions for main outcomes (g = 0.44), depression (g = 0.46), anxiety (g = 0.20), and quality of life (g = 0.40). There was significant heterogeneity between studies for main outcomes and depression. Within-group effects ranged from Hedges's g = 0.32-0.99. The effects of IPDT were maintained or increased at follow-up. Study quality varied but was generally high. No indications of publication bias were found. In conclusion, IPDT is a promising treatment alternative, especially for depression, although the small number of studies limits the generalizability of the findings.

Tomas Lindegaard, Ms.C., Ph.D. candidate in Clinical Psychology, Department of Behavioural Sciences and Learning, Linköping University, Linköping, Sweden

Matilda Berg, Ms.C., Ph.D. candidate in Clinical Psychology, Department of Clinical Neuroscience, Karolinska Institute, Stockholm, Sweden

Gerhard Andersson, Ph.D., Professor in Clinical Psychology, Department of Behavioural Sciences and Learning, Linköping University, Linköping, Sweden, and Department of Clinical Neuroscience, Karolinska Institute, Stockholm, Sweden

Introduction

One of the first studies on internet-delivered psychotherapy was conducted in 1996 by researchers at the University of Amsterdam who developed a website through which they treated 20 students suffering from posttraumatic stress symptoms, 19 of which went into remission at posttreatment (Lange, van de Ven, Schrieken, Bredeweg, & Emmelkamp, 2000). Since then, more than 300 controlled trials have been conducted evaluating various forms of internet- delivered interventions for a range of different disorders and target populations (Andersson, 2018; Andersson, Titov, Dear, Rozental, & Carlbring, 2019).

Although many different versions and types of internet-delivered interventions have been developed and tested, they all share some common features, such as the requirement of a treatment software platform to deliver the treatment and facilitate interaction between the client and therapist (Andersson et al., 2019). The treatment programs usually deliver treatment content in the form of text, but use of audio and video is also common (Andersson et al., 2019). Most studies have evaluated programs involving some form of therapist guidance; there are also many instances of self-guided treatments, although the former type of program tends to be more effective (Baumeister, Reichler, Munzinger, & Lin, 2014). As noted by Andersson and colleagues (2019), cognitive-behavioral therapy (CBT) has been the dominant therapy model used thus far in internet interventions, with meta-analytic evidence indicating that internet-delivered CBT (ICBT) is as effective as regular face-to-face CBT (Carlbring, Anders-son, Cuijpers, & Riper, 2018) and more efficacious than control conditions for anxiety and depression disorders (Andrews et al., 2018). Research also suggests that there are long-term effects (Andersson, Rozental, Shafran, & Carlbring, 2018).

In recent years there have also been a number of internet interventions based on other theoretical models, such as interpersonal psychotherapy (IPT; Dagöö et al., 2014) and mindfulness (Boettcher et al., 2014), but also psychodynamic approaches. The first of these was a study by Andersson and colleagues (2012) that used an adapted version of the self-help book Make the Leap (Silverberg, 2005) in the treatment of generalized anxiety disorder (GAD). Since then, a handful of RCTs have been conducted investigating the efficacy of treatment programs based on a range of different psychodynamic treatment approaches, such as Luborsky's supportive-expressive model of psychoanalytic psychotherapy (Zwerenz, Becker, Gerzymisch et al., 2017) and the affect phobia therapy model (APT; Johansson, Björklund et al., 2013).

The psychodynamic internet-delivered treatment programs that have thus far been tested have generally been structured in a similar way as treatment programs based on CBT models, with weekly text- based modules, various homework exercises that the participants are encouraged to complete on a weekly basis, and e-mail support from a therapist. One could argue that some of these features, for example the use of homework assignments, constitute a significant difference between internet-delivered psychodynamic therapy (IPDT) and traditional psychodynamic therapy delivered face-to-face, given that use of homework assignments normally is considered a distinguishing feature between CBT and psychodynamic approaches (Blagys & Hilsenroth, 2002). However, in practice a majority of psychodynamically oriented clinicians report encouraging their clients to engage in between-session assignments (Kazantzis, Lampropoulos & Deane, 2005), perhaps making the discrepancy less important in the actual clinical reality. Another difference between IPDT and traditional psychodynamic therapy, according to Johansson, Frederick, and Andersson (2013), is the lack of explicit transference work found in the former. However, notwithstanding the differences noted above, IPDT and face-to-face psychodynamic therapy also share many similarities both regarding the treatment content and the role of the therapist in providing support and validation (Johansson, Frederick et al., 2013). For example, in the IPDT studies based on the affect phobia model, participants learn how to conceptualize their problems in terms of a psychodynamic conflict around feelings (Johansson, Björklund et al., 2013; Johansson et al., 2017). Participants are guided in how to restructure their own defenses, regulate anxiety, and face previously avoided feelings, which are purported to be the primary mechanisms of change (Johansson, Björklund et al., 2013; Johansson et al., 2017). Throughout this process, the therapist provides feedback and validation, mostly in relation to the homework assignments that the participants complete (Johansson, Frederick et al., 2013).

While recent years have seen a number of systematic reviews and meta- analyses regarding the efficacy of psychodynamic therapy in general (Abbas et al., 2014; Steinert, Munder, Rabung, Hoyer, & Leichsenring, 2017) as well as specific forms of psychodynamic therapy, such as experiential dynamic therapy (Lilliengren, Johansson, Lindqvist, Mechler, & Andersson, 2016), there is to our knowledge no previous review and meta-analysis regarding the efficacy of IPDT. Thus, the aim of the present study was to provide a first systematic review and meta-analysis regarding the efficacy of IPDT.

Results
Study Inclusion

Through our search strategies, we identified 73 articles from the PubMed/ MEDLINE, PsychINFO, and SCOPUS databases. After exclusion of duplicates, nonintervention studies, and studies of treatment approaches other than IPDT, 10 articles remained (see Table S1 in online supplemental material). These remaining articles were screened based on review of title and abstract, which led to the exclusion of three additional studies. Based on the full-text review of the remaining articles, all of the remaining seven studies met the inclusion criteria and were thus included in the study. For an overview of the study selection process, see Figure 1.

Study Characteristics

The study characteristics are summarized in Table 1. Of the seven included RCTs, one had depression as the primary target (Johansson, Björklund, et al., 2012), two targeted anxiety disorders (Andersson et al., 2012; Johansson, Björklund, et al., 2017), one targeted return to work (RTW; Zwerenz, Becker, Gerzymisch et al., 2017), and the remaining three had a transdiagnostic focus, targeting both depression and anxiety (Johansson et al., 2013; Lemma & Fonagy, 2013; Zwerenz, Becker, Johansson et al., 2017). The total number of participants was 528 in the treatment group and 552 in the control conditions. Only one of the seven studies had a bona fide intervention as a comparison (Andersson et al., 2012), meaning that we could only perform a quantitative synthesis comparing IPDT to the inactive control conditions. This study also had an inactive control condition and could thus be included in the analysis of IPDT compared to inactive controls.

With regard to theoretical orientation, three of the studies (Johansson, Björklund et al., 2013; Johansson et al., 2017; Zwerenz, Becker, Johansson et al., 2017) were based on the affect phobia model (McCullough et al., 2003) and more precisely on the self-help book Living Like You Mean It (Frederick, 2009). Two other studies (Andersson et al., 2012; Johansson et al., 2012) were based on the SUBGAP model as described in the psychodynamic self-help book Make the Leap (Silverberg, 2005). Of the remaining two studies, one (Lemma & Fonagy, 2013) was based on the brief dynamic interpersonal therapy model (BDIT; Lemma, Target & Fonagy, 2011) while the other (Zwerenz, Becker, Gerzymisch et al., 2017) was based on Luborsky's supportive-expressive model of psychoanalytic psychotherapy (Luborsky, 1984).

Regarding delivery format and structure of the treatment, in five of the included studies (Andersson et al., 2012; Johansson et al., 2012, Johansson, Björklund et al., 2013; Johansson et al., 2017; Zwerenz, Becker, Johansson et al., 2017) the treatment consisted of weekly text- based modules accessed through a web-based platform. The modules included various homework exercises that the participants completed on a weekly basis with e-mail support from a therapist. In one of the received feedback from their therapist on a weekly basis. Finally, in the study by Lemma and Fonagy (2013), participants were sent self-help material by e-mail once per week. The study made use of a group format in which participants had access to an online forum where they were encouraged to reflect on the treatment material via nonsynchronous communication with other group members. The group was facilitated by a therapist. The studies were carried out in either Sweden, Germany, and England, and the number of participants in each study ranged from 24 to 664.

Main Analysis—IPDT versus Inactive Controls

Between- and within-group results are presented in Table 2. With regard to the effect of IPDT compared to the inactive control conditions, we found small effects in favor of IPDT on all four outcomes; main outcome (g = 0.44), depression (g = 0.46), anxiety (g = 0.20), and quality of life (g = 0.40). For two of the outcomes, the main outcome and depression, we found a significant heterogeneity. After inspecting forest plots for the main outcome analysis, we decided to run a sensitivity analysis without the Johansson et al. (2012) study, which resulted in a nonsignificant heterogeneity (Q = 6.18, p = .29, I2 = 19.1%) and a lower effect size (g = 0.28, 95% CI [0.1, 0.45]).

In addition, we performed a sensitivity analysis removing the Zwerenz, Becker, Gerzymisch et al. (2017) study, since this study targeted participants with a somatic primary diagnosis. For the main outcome analysis, this also resulted in a nonsignificant heterogeneity (Q = 10.17, p = .07, I2 = 50.8%) and a moderate effect size of g = 0.52, 95% CI [0.22, 0.82]. Similarly for depression outcomes, removal of Zwerenz, Becker, Gerzymisch et al. (2017) resulted in a nonsignificant heterogeneity (Q = 8.98, p = .11, I2 = 44.3%) and a moderate effect size of g = 0.55, 95% CI [0.27, 0.83]. Finally, for anxiety outcomes, removing he Zwerenz, Becker, Gerzymisch et al. (2017) resulted in a nonsignificant heterogeneity (Q = 2.29, p = .81, I2 = 0.0%) and a small effect size of g = 0.32, 95% CI [0.12, 0.52].

With regard to follow-up, only two studies (Andersson et al., 2012; Zwerenz, Becker, Gerzymisch et al., 2017) reported follow-up data in comparison to an untreated control group.

Within-Group Effects

The analysis of within-group effects of the interventions pretreatment to posttreatment revealed effects between Hedges's g = 0.32-0.99 for the four outcomes. However, significant heterogeneity was found for the main outcome, depression, and anxiety. For the main outcome, after inspecting forest plots, we ran a sensitivity analysis without the studies by Zwerenz, Becker, Gerzymisch et al. (2017), Zwerenz, Becker, Johansson et al. (2017), and Johansson et al. (2012), which resulted in a nonsignificant heterogeneity (Q = 3.35, p = .34, I2 = 10.5%) and an effect size of g = 1.29, 95% CI [1.0, 1.58].

Similarly to the between-group analysis, we also performed a sensitivity analysis removing the Zwerenz, Becker, Gerzymisch et al. (2017) for the same reason as stated above. For the main outcome analysis, this also resulted in a significant heterogeneity (Q = 39.50, p < .01,I2 = 87.3%) and an effect size of g = 1.17, 95% CI [0.55, 1.79]. For depression outcomes, removal of Zwerenz, Becker, Gerzymisch et al. (2017) resulted in a significant heterogeneity (Q = 40.47, p < .01, I2 = 87.6%) and an effect size of g = 0.98, 95% CI [0.36, 1.61]. Finally, for anxiety outcomes, removing Zwerenz, Becker, Gerzymisch et al. (2017) resulted in a significant heterogeneity (Q = 15.5, p < .01, I2 = 67.7%) and an effect size of g = 0.83, 95% CI [0.47, 1.2].

With regard to follow-up, only five of the studies reported follow-up data for the treatment group and only two studies included measures of quality of life at follow-up; this outcome was therefore not included in this analysis. The analysis of posttreatment to follow-up revealed effects between Hedges's g = 0.06-0.3 for the three outcomes. Significant heterogeneity was found for the anxiety outcome.

Discussion

This systematic review and meta-analysis most likely constitute the first review of the efficacy of IPDT. In total, seven studies were identified that met all inclusion criteria. Overall, we found small effects in favor of IPDT compared to inactive control conditions regarding overall effects of treatment (g = 0.44), as well as on depression (g = 0.46), anxiety (g = 0.20), and quality of life (g = 0.40) outcomes. No comparison in relation to active control conditions was possible, since only one of the eligible studies included an active control (Andersson et al., 2012). With regard to study quality, four of seven studies were rated as having low risk of bias across all rated criteria. We did not find evidence of publication bias.

The effect sizes obtained in the present study appear to be lower than those found in meta-analysis of internet interventions based on CBT models. For example, Andrews et al. (2018) found an overall effect of ICBT of Hedges's g = 0.80 compared to control conditions. They are also somewhat lower than those found in meta-analyses of face-to-face psychodynamic therapy where, for example, Abbas et al. (2014) found effects in the medium range for general psychiatric symptoms (SMD, 0.71), anxiety (SMD, 0.64), and depression (SMD, 0.61). However, a more recent meta-analysis by Cuijpers, Karyotaki, de Wit & Ebert (2019) found an effect of Hedges's g = 0.39 for psychodynamic therapy for adult depression, which is more in line with the findings of the present study. Due to the limited number of studies and the large heterogeneity between studies, the findings of the present study have to be viewed as highly tentative. In addition, the sensitivity analysis, which excluded Zwerenz, Becker, Gerzymisch et al. (2017), resulted in moderate effect sizes for both overall outcomes and depression, more akin to the findings by Abbas et al. (2014).

With regard to within-group effects, the effect sizes found in the present study are again somewhat lower than those found in face-to-face studies of psychodynamic therapy. For example, in their meta-analysis of experiential dynamic therapy, Lilliengren et al. (2015) found within- group effects ranging from d = 1.11 for general psychiatric symptoms to d = 0.86 for quality of life/ global functioning, which is also similar to within-group effects found in the meta-analysis by Town et al. (2012). Overall, the findings of the present study do not exclude the possibility that IPDT might result in slightly lower treatment effects than face-to- face psychodynamic therapy. However, similarly to the between-groups effect mentioned above, removal of Zwerenz, Becker, Gerzymisch et al. (2017) resulted in within-group effect sizes more similar to those found by Lilliengren et al. (2015) and Town et al. (2012). Regarding follow- up, we found evidence for continued improvement on anxiety symptoms (g = 0.3) while the other outcomes remained stable.

Concerning study quality, there was a considerable variance between studies with regard to their risk of bias. All of the four studies conducted by the research group led by Professor Andersson were rated as having low risk of bias across all rated domains, whereas the other three studies were rated as having unclear or high risk of bias in at least one of the rated domains. Due to the small number of studies, it was not possible to investigate whether and how study quality related to the reported effects of the treatment under study.

With regard to the limitations of the present study, one major limitation concerns the limited number of eligible studies, which limits the generalizability of the findings. Due to this limitation, it was not possible to examine the effects of IPDT for specific disorders; instead, we calculated an overall effect on the main outcome measures of the studies. The secondary analysis of depression, anxiety, and quality of life mixed studies targeting the same outcome as either a primary or a secondary outcome, which could potentially lead to lower estimated effect sizes. Also, due to the small sample size, it was not possible to conduct subgroup analysis based on, for example, study quality or treatment model. Another limitation concerns the fact that we calculated within-group effects using the formula for independent groups. This procedure was chosen because the included studies did not report correlations between pre- and postmeasurements, however this might have led to a slight overestimation of effect sizes (Dunlap et al., 1996).

Conclusion

The results of the meta-analysis suggest that IPDT is more efficacious than inactive control conditions, with small effect sizes in favor of IPDT with regard to main outcomes, depression, anxiety, and quality of life. Moreover, the effects of IPDT seem to be maintained or increase at follow-up. Existing IPDT studies are based on a variety of theoretical models, with Leigh McCullough's affect phobia model being the most common. Study quality was generally high and we found no indications of publication bias. However, due to the small number of eligible studies, the findings in this present review and meta-analysis has to be considered highly tentative; further research is needed.

Acknowledgement

The present review was funded in part by a professors grant to the last author.