Functional neural changes associated with psychotherapy in anxiety disorders – A meta-analysis of longitudinal fMRI studies

Highlights

• We conducted a coordinate-based meta-analysis of studies assessing fMRI pre- and post-therapy in anxiety disorders.

• Our results are based on whole-brain findings and include more than 50% original statistical maps.

• From pre to post, activation decreased in the insula, the anterior cingulate cortex, and the dorsolateral prefrontal cortex.

• Subgroup-analysis for CBT and exposure revealed an additional cluster of activation decrease in the supplementary motor area.

Abstract

Successful psychotherapy for anxiety disorders is thought to be linked to functional neural changes in prefrontal control areas and fear-related limbic regions. Thus, discovering such therapy-associated neural changes might point to relevant mechanisms of action. Using AES-SDM, we conducted a coordinate-based meta-analysis of 22 whole-brain datasets (n = 419 anxiety patients) from 18 studies identified by our systematic literature search following PRISMA criteria (preregistration available at OSF: https://osf.io/dgc4p). In these studies, fMRI data was collected in response to negative stimuli during cognitive-emotional tasks before and after psychotherapy. Post-psychotherapy, activation decreased in the right insula, the anterior cingulate cortex, and the dorsolateral prefrontal cortex; no region had increased activation. A subgroup analysis for CBT revealed additional decrease in the supplementary motor area. Reduced activation in limbic and frontal regions might indicate therapy-associated normalization regarding the perception of internal and external threat, subsequent allocation of cognitive resources, and changes in cognitive control. Due to the integration of diverse treatments and experimental tasks, these changes presumably reflect global effects of successful psychotherapy.

Introduction

With a global point prevalence rate of 7.3% (Baxter et al., 2013) and a lifetime prevalence rate of 33.7% (Bandelow and Michaelis, 2015), anxiety disorders are among the most common mental disorders. According to the Diagnostic and Statistical Manual of Mental Disorders (DSM-5, American Psychiatric Association, 2013), anxiety disorders comprise specific phobia (SP), social anxiety disorder (SAD), panic disorder (PD), agoraphobia (AP), generalized anxiety disorder (GAS), separation anxiety, and mutism. Anxiety disorders are successfully treated via psychotherapy; in particular, cognitive behavioral therapy (CBT) that incorporates exposure has been found to be effective at a meta-analytical level (Bandelow et al., 2015, Carpenter et al., 2018, Cuijpers et al., 2016, Hofmann et al., 2012) and, thus, is considered the first-line treatment for anxiety disorders (e.g., Bandelow et al., 2014). Yet, other treatments have also been deemed effective, such as psychoanalytic and psychodynamic approaches as well as newer methods of treating mental disorders, such as mindfulness-based interventions and so-called “third wave” therapies (see e.g., Hayes, 2004; Keefe et al., 2014, Öst, 2008, Singh and Gorey, 2018, Twohig and Levin, 2017).

A growing body of literature shows that psychotherapy-related clinical improvements go along with changes in brain functioning (for reviews, see Barsaglini et al., 2014; Brooks and Stein, 2015). Yet, the specificity of neural activation changes that arise due to therapy have so far remained inconclusive. Learning more about therapy-associated neural changes might foster our understanding of the relevant mechanisms underlying successful therapy. Neurobiological models of anxiety disorders propose altered activation in brain regions underlying the processing of fear and the regulation of emotions: While limbic areas linked to bottom-up modulation and generation of fear seem hyperactive, top-down control circuits in prefrontal regions seem impaired (Duval et al., 2015, Etkin and Wager, 2007, Shin and Liberzon, 2010). Psychotherapy is assumed to address these dysfunctional activation patterns, e.g., by strengthening prefrontal areas that, in turn, regulate overactivation in threat-responsive limbic regions (e.g., Etkin et al., 2005; Penninx et al., 2021).

The limbic regions most regularly found to exhibit hyperactivation in patients with pathological anxiety compared to healthy controls are the amygdala, insula, and the anterior cingulate cortex (ACC) (for detailed review, see Duval et al., 2015; Shin and Liberzon, 2010). The amygdala is well established as a key “fear-excitatory” structure that is involved in the acquisition and expression of fear; it is often referred to as an “alarm system” (Davis, 1992, LeDoux, 2000, Shin and Liberzon, 2010). Reduced amygdala activation found after successful therapy (Furmark et al., 2002, Månsson et al., 2016) has been interpreted as a normalization of the “alarm system", potentially due to effects of habituation or inhibitory learning (Craske et al., 2008, Craske et al., 2018, Foa and McNally, 1996). However, many studies have failed to show reduced amygdala activation after therapy (e.g., Bomyea et al., 2020; Goldin et al., 2013; Klumpp et al., 2013), pinpointing the need for further clarification.

Regarding insula and ACC, a recent meta-analysis has found that enhanced activation in those regions before treatment predicted better treatment response to CBT in anxiety patients (Picó-Pérez et al., 2022). Furthermore, the (anterior) insula and the (dorsal) ACC both seem to be involved in implicit and explicit emotion regulation (Picó-Pérez et al., 2019), a key process presumably enhanced by psychotherapy. Yet, the specific role of the insula and the ACC within the complex process of emotion regulation is not entirely understood. As both are regarded as key nodes of the salience network, which detects relevant internal and external cues to guide an adequate response (Menon, 2015), researchers have proposed that activation in the insula and ACC found during emotion regulation tasks indicate the processing of emotional information and the mediation of regulatory attempts instead of reflecting the regulatory process itself (Picó-Pérez et al., 2019). Despite inconsistencies, longitudinal studies indicate that activation in the insula and ACC decreases following treatment (for review, see Brooks and Stein, 2015; Brown et al., 2019; Klumpp et al., 2013).

Interacting with the “fear-excitatory” systems, the prefrontal cortex (PFC) plays a crucial role in regulating the (limbic) fear response (Phillips et al., 2008, Sotres-Bayon and Quirk, 2010). Subregions of the PFC (especially ventromedial and dorsolateral) have been found critical to extinction learning and highly relevant in emotion regulation, e.g., cognitive reappraisal (Ochsner and Gross, 2005, Sotres-Bayon et al., 2006), potentially via exerting inhibitory control over the amygdala (Quirk et al., 2003, Wager et al., 2008). In line with this, hypoactivation of the PFC in anxiety disorders might be linked to deficits in emotion regulation (Ball et al., 2013, Goldin et al., 2009). However, PFC hypoactivation cannot be seen as a general characteristic of anxiety disorders, as several studies have found hyperactivation or no activation differences compared to healthy controls in the PFC (Duval et al., 2015, Etkin and Wager, 2007). Similarly, enhanced as well as dampened PFC activation has been reported following therapy (Brooks and Stein, 2015), and it is proposed that treatment might enhance a more flexible and adaptive utilization of the PFC, not greater activation per se (Hofmann et al., 2012b). Taken together, there is accumulating evidence for treatment-associated changes in limbic as well as frontal brain regions; however, the exact brain regions involved as well as the direction of effects remain elusive. This heterogeneity is probably caused by multiple factors, including content-related differences (such as the specific interventions applied) but also MRI-related problems, such as the small sample sizes common in this field, the uniqueness of the individual human brain, and differences in MRI scanners used (see e.g., Marek et al., 2022).

Previous meta-analyses on neural effects of psychotherapy in mood and anxiety disorders, which also included depression, obsessive-compulsive disorder (OCD) and post-traumatic stress disorder (PTSD), hint that psychotherapy reduces activation in limbic regions (Marwood et al., 2018, Messina et al., 2013) and the inferior frontal gyrus (Marwood et al., 2018). However, conclusions drawn specifically for anxiety are limited: Despite accumulating evidence for transdiagnostic psychotherapeutic effects on the brain, psychotherapy also seems to yield anxiety-specific neuronal changes (e.g., Craske et al., 2011), and the therapeutic elements (such as behavioral exposure) applied in practice vary depending on the patient’s diagnosis. Furthermore, only a small number of studies on anxiety disorder and no original data have been included in existing meta-analyses, diminishing their generalizability.

To address the issues outlined above, the present coordinate-based meta-analysis aims to identify consistent patterns of neural changes associated with psychotherapy for anxiety disorders, focusing on studies assessing fMRI before and after psychological treatment. To our knowledge, this is the first meta-analysis on task-based functional neural changes associated with psychotherapy focusing on anxiety disorders and using best practice meta-analytic methods by including original data (Radua et al., 2012). Based on the literature described above, we expect psychotherapy to be associated with a reduction of activation in limbic regions and potentially an increase of activation in the PFC.