Abstract
Background Emotional dysregulation constitutes a serious public health problem in need of novel transdiagnostic treatments. To this aim, we developed and tested a one-time intervention that integrates behavioral skills training with concurrent repetitive transcranial magnetic stimulation (rTMS).
Methods Forty-seven adults who met criteria for at least one DSM-5 disorder and self-reported low use of cognitive restructuring (CR) were enrolled in a randomized, double-blind, placebo-controlled trial. Participants were taught CR and underwent active or sham rTMS applied at 10 Hz over the right or left dorsolateral prefrontal cortex (dlPFC) while practicing reframing and emotional distancing in response to autobiographical stressors.
Results Those who received active as opposed to sham rTMS exhibited enhanced regulation (ds = 0.20 - 0.67) and reduced distress (ds = 0.32 - 0.35), as measured by self-report and psychophysiological indices during the intervention (higher high-frequency heart rate variability, lower regulation duration) and ambulatory self-report measures during the week following the intervention. All participants improved in their use of reappraisal (d = 2.25), psychopathology (d = 0.48) and emotional dysregulation (d = 0.81) and maintained their gains at the one month follow up. The procedures were acceptable and feasible with few side effects. A secondary analysis suggested that participants with stress disorders may respond particularly well to active right rTMS.
Conclusion These findings show that engaging frontal circuits simultaneously with cognitive skills training and rTMS is clinically feasible, well-tolerated and shows promise as the next generation of treatments for transdiagnostic emotional dysregulation.
Emotional dysregulation, defined as a deficit in the ability to reduce or change intense negative emotional states (Gross, 2013), is a problem with inhibitory control that occurs across psychopathology (Kring & Werner, 2004; Wolff et al., 2019) and that is unique and critical in driving the severity and duration of a wide variety of mental health problems (Bekh Bradley et al., 2011; Fernandez et al., 2016). Because of its contribution to a variety of mental health disorders such as anxiety and depression, emotion dysregulation is a serious public health concern that causes enormous burdens on society and the economy (DuPont et al., 1996; Greenberg et al., 2003; Lépine & Briley, 2011; Schmitz & Kruse, 2002).
Psychotherapy is currently the standard of care for affective disorders (Hunsley et al., 2014), and is considered moderately effective in the treatment of emotional dysregulation (Barlow et al., 2017; Cuijpers et al., 2019). Difficulties in emotion regulation can be changed through psychotherapy (Campbell-Sills & Barlow, 2007; Gaher et al., 2013; Gratz et al., 2012; Kring & Werner, 2004) that can be conducted with multi-diagnostic individuals (Barlow et al., 2017; Neacsiu et al., 2014). Nevertheless, the course of psychotherapies for emotional dysregulation can be extensive, requiring 12 or more sessions (Neacsiu et al., 2014). For example, in a previous study using 44 transdiagnostic adults with high emotion dysregulation (42% primarily depressed, 37% GAD, 15% social anxiety, 6% other anxiety), we found significant long-term improvements in emotion regulation following an active, versus a control, behavioral treatment that took place over 16 sessions (Neacsiu et al., 2014; Neacsiu, Rompogren, et al., 2018). Barlow et al. (2017) also found that a transdiagnostic treatment targeted towards emotional dysregulation was as effective in reducing anxiety severity in 12 to 16 sessions as specialized, disorder-specific treatment protocols. These treatments have in common the approach of teaching emotion regulation skills (Cuijpers et al., 2019), with the majority of transdiagnostic treatments including cognitive restructuring, relaxation, and mindfulness as some of the skills included (Berking et al., 2010; Farchione et al., 2012; Linehan et al., 2015).
Taken together, these studies are encouraging; nevertheless they draw attention to the need for improvement to maximize gains and reduce treatment burden (Roth & Fonagy, 2005). Specifically, more work is needed to accelerate the process of learning within therapy sessions, to retain skills after therapy sessions, and to help patients generalize in-session learning to out-of-session contexts (Kazdin & Blase, 2011). The intensive burden on patients to come into an office for multiple sessions and put significant effort into learning new skills leads to frequent treatment dropout and avoidance of psychotherapies, especially for emotionally dysregulated adults (Carta et al., 2013; Ciarrochi & Deane, 2001; Vogel et al., 2008).
One approach to optimize psychological interventions is to translate findings from basic neuroscience studies on emotion regulation into the development of new interventions. Emerging neuroscientific findings highlight that across mood, anxiety and stress disorders, hypoactivation in prefrontal regions such as the dorsolateral (dlPFC), ventrolateral (vlPFC), and medial prefrontal cortex (mPFC), and hyperactivation in limbic/paralimbic brain areas, such as the left anterior insula (Zilverstand et al., 2017), characterize difficulties in changing emotional arousal. Importantly, the regulatory tasks used to highlight these neural abnormalities prominently use strategies that overlap with skills taught in cognitive behavioral therapy (CBT).
Brain stimulation is one approach to altering pathological brain circuits that has translational potential to mitigate problems with emotion regulation. Transcranial magnetic stimulation (TMS), in particular, is one type of noninvasive brain stimulation that uses brief, high-intensity magnetic fields to modify neural activity underneath a stimulating coil. When applied in a repetitive manner (rTMS), currents are rapidly turned on and off, producing a time-varying magnetic field that generates current flow in the underlying neural tissue and changes cortical excitability in a manner that depends on the frequency of stimulation (Neacsiu & Lisanby, 2015). While rTMS-induced effects on cortical excitability are quite focal, with depth penetration to about 2 cm below the scalp, this stimulation also leads to down-stream effects through network connectivity, therefore allowing modulation of broad brain circuits that underlie psychopathologies (Beynel et al., 2020). Based on the neuroscience of emotion regulation reviewed above, applying high frequency rTMS over the dlPFC could have potential to improve learning of emotion regulation skills.
Newer investigations support the premise that neurostimulation can enhance emotion regulation performance. For example, excitatory stimulation of the left dlPFC facilitates significantly quicker disengagement from angry faces (De Raedt et al., 2010), while excitatory stimulation of the right dlPFC significantly enhances cognitive emotion regulation in healthy adults (Feeser et al., 2014). Furthermore, inhibitory stimulation of the left temporoparietal junction makes cognitive emotion regulation more effortful than sham stimulation (Powers et al., 2020). These experiments have only been conducted with non-clinical samples and have not yet been translated to interventions. An exception is a study where 10 sessions of 10 Hz rTMS over the right dlPFC led to improvements in affective instability and anger dysregulation in a borderline personality disorder (BPD) sample (Hasani et al., 2013). Nevertheless, this study did not include a comparison with a control treatment and did not have a longitudinal assessment of effects, leaving gaps in what could be inferred from the results. Therefore, the effectiveness of rTMS alone on clinical emotional dysregulation is unknown. Furthermore, rTMS as a stand-alone treatment is unlikely to solve the burden barrier that psychotherapies have. Successful stand-alone rTMS treatments require 10-20 daily sessions in a medical office, which can be equally burdensome (Neacsiu & Lisanby, 2015).
Nevertheless, concurrent neurostimulation can successfully be used as an augmentation tool for behavioral interventions with apparent effects even after a one session intervention. For example, rTMS can remediate hypoactive brain circuits by enhancing spontaneous activity in targeted regions [e.g., remediating working memory performance following sleep deprivation (Luber et al., 2013)]. In addition, when combined with a behavioral prime (food, paraphernalia, cues that trigger obsessions), active neurostimulation reduces problematic behaviors (binge eating, drug use, compulsive behaviors) above and beyond sham neurostimulation (Gorelick et al., 2014; Trevizol et al., 2016; Van den Eynde et al., 2010) in the short term. Furthermore, past studies have combined rTMS with 16-20 sessions of psychotherapy, demonstrating feasibility (Neacsiu, Luber, et al., 2018) with enhanced effects over psychotherapy alone (Kozel et al., 2018) or over cognitive training alone (Cunningham et al., 2015). Taken together, these studies show promise for neurostimulation as an augmentation tool to alter emotion regulation especially in combined CBT-rTMS interventions. They also highlight the need for additional translational work to reduce, rather than increase, the burden on the patients undergoing combined treatments.
This pre-registered (NCT02573246), randomized, double-blind, placebo-controlled study, using a parallel design, aimed to establish the feasibility, acceptability and preliminary efficacy of enhancing emotion regulation skills training for transdiagnostic clinical adults with excitatory neurostimulation over the dlPFC, concurrently during an emotional regulation task. We selected cognitive restructuring (CR) as a targeted emotion regulation skill because it is generally effective at downregulating emotions (Gross & Jazaieri, 2014; Webb et al., 2012), it can be implemented across a broad range of affective contexts (Gross & Jazaieri, 2014), and it engages the same frontolimbic neural networks on which neurostimulation has already been tested (Ochsner et al., 2012; Powers & LaBar, 2019). CR involves thinking differently about a situation in order to reduce negative affect (Ochsner et al., 2002; Webb et al., 2012). Theoretical considerations and empirical findings highlight that CR can be further subdivided into specific tactics, such as reframing and distancing (Powers & LaBar, 2019); nonetheless, these tactics activate overlapping areas within the core network.
In addition to subjective reports of affect, we collected high frequency heart rate variability (HF-HRV) as an objective physiological measure of emotion regulation that could be examined in the moment without interfering with skill practice and neurostimulation. In the context of emotional arousal, HF-HRV is positively correlated with executive function and dlPFC or mPFC activation (Hansen et al., 2003; Jönsson & Sonnby-Borgström, 2003; Lane et al., 2009; Thayer & Lane, 2000) and is a sensitive measure of adaptive emotion regulation (Appelhans & Luecken, 2006; Camm et al., 1996; Geisler & Kubiak, 2009). Participants who engage in effective emotion regulation have higher changes in HF-HRV from baseline than those who do not engage in regulation (Butler et al., 2006; Demaree et al., 2004). HF-HRV has been especially connected with effective and not with maladaptive regulation (Elliot et al., 2011; Fenton-O’Creevy et al., 2012). Therefore, we computed HF-HRV and interpreted increases in this measure as indicative of an increase in effective regulation.
To establish proof-of-concept and feasibility, we limited the intervention to a single session, and participants were multi-diagnostic adults who reported difficulties with cognitive emotion regulation. Given that evidence exists for the beneficial effects of both left and right dlPFC stimulation (Balconi & Ferrari, 2012; De Raedt et al., 2010; Feeser et al., 2014; Leyman et al., 2009; Remue et al., 2016), we randomized participants to receive either right or left or sham HF-rTMS in conjunction with CR practice. We hypothesized that participants receiving active versus sham rTMS in conjunction with CR would exhibit faster recovery from emotional distress during the intervention session, reduced arousal and increased use of CR during the week after the intervention, and improved emotional dysregulation, psychopathology and functioning for up to a month following the intervention.
Materials and Methods
Participants and Procedures
We aimed to recruit 20 participants in each condition to align with other neurostimulation studies where 6-20 adults per condition were sufficient to demonstrate proof-of-concept for novel treatments (Bentwich et al., 2011; Cunningham et al., 2015). Forty seven intent to treat (ITT) community participants were 8 men and 39 women between the ages of 18 and 65 (M = 30.02; SD = 10.73) who met criteria for at least one DSM-5 disorder (excluding active substance use, psychosis, and Bipolar I) and who self-reported below average use of CR when upset (see Supplement). Participants were transdiagnostic and met criteria for an average of 2.53 (SD = 1.65) current and 4.20 (SD = 1.94) lifetime diagnoses; 15 participants (31.92%) met criteria for at least one personality disorder according to the Structural Clinical Interview for DSM-5 Disorders (SCID-5; First et al., 2015; 2016). Of note, a transdiagnostic sample is defined using a different set of criteria than DSM diagnostics and therefore not heterogeneous (Fernandez et al., 2016). This approach highly encouraged by the NIMH RDoC framework (Insel et al., 2010).
Low use of CR was operationally defined as a mean score lower than 4.7 on the cognitive reappraisal subscale of the Emotion Regulation Questionnaire (ERQ; Gross & John, 2003) at the intake assessment, a cutoff computed by pooling means from 18 studies published before 2014 on US samples (N = 4331 participants – see Supplement for citations). We considered adults who scored below this pooled mean to have “low use” of CR strategies and therefore to be the optimal candidates for a one-time CR intervention. Figure 1 and Supplementary materials detail study flow and inclusion/exclusion criteria. The study was approved by the Duke University Health System (DUHS) Institutional Review Board and participants were paid for the study visits that they attended. No changes to inclusion/exclusion were done after beginning enrollment. All procedures were conducted at DUHS core facilities.
This study involved a total of three sessions: an intake, the intervention, and a follow-up session one month after intervention. In addition, participants completed an ambulatory assessment for a week following the intervention (8x per day for 7 days), and a battery of self-reports remotely via an online link at the end of the ambulatory week.
Intake Session
Qualified participants completed a clinical interview, a battery of self reports, and a standardized assessment (1987) which yielded four autobiographical negative emotional arousal scripts (details in supplementary materials). Participants were then randomly assigned to active right dlPFC (n = 18), active left dlPFC (n = 19), or sham stimulation (n = 22), matching for biological sex and use of psychotropic medication using a minimization algorithm (Scott et al., 2002; Taves, 1974) and a 1:1:1 ratio. Sham participants were further randomized using a 1:1 ratio to receive TMS to either the right (N = 7) or left DLPFC (N = 8). Not all randomized participants completed treatment (see Supplement).
Intervention Session
Participants returned for the 3.5-hour intervention session within a month of intake. The first 45-60 min of this session was spent on skills training, one-on-one with the first author or a trained psychologist, and was focused on learning CR and practicing on standardized and personal examples. Skills training used established procedures blending psychotherapeutic approaches (Beck et al., 1985; Linehan, 2014) with instructions in CR that matched prior neuroimaging studies (Shurick et al., 2012). Both distancing and reframing CR tactics were presented and practiced. (See supplement.) Next, physiological data was collected during a 300 s rest period (session baseline), followed by the identification of the resting motor threshold (rMT). A 600 s habituation period followed when participants received active or sham rTMS alone while listening to white noise and being instructed to think of nothing in particular. Next, participants were instructed to sit quietly for a 300 s pre-stimulus baseline, followed by instructions to imagine as vividly as possible a negative experience while hearing a 40s recording of the autobiographical stressor (120 s total), and by instructions to reduce distress using CR (600s). The rTMS began within 10 s after the CR instructions appeared. After there was a break, followed by a second and third administration of the stressor task using the procedures outlined above, but with a different personalized stressors presented in a randomized order.
Before and after the intervention session, participants rated on a scale from 0 to 3 (absent -severe) the intensity of their headache, neck pain, scalp pain, seizure (as observed by technician), hearing impairment and any other side effect. After each baseline, stress induction, and regulation period, the participant was asked to rate subjective units of distress (SUDS; 0 to 9 scale) and dissociation using a 4-item scale (Stiglmayr et al., 2009). Psychophysiological measurements were collected continuously during the intervention using the BIOPAC MP-150 recording system (Goleta, CA). Electrodes recording heart rate (HR) were placed on the participant’s wrist and ankle. Amplified analog data recorded with a 200 Hz sampling rate were converted to digital recording and filtered using BIOPAC’s AcqKnowledge 4.1 software.
rTMS Parameters
Active or sham rTMS was performed with a figure-8 coil (A/P Cool-B65) and a MagPro X100 stimulator with MagOption (MagVenture, Denmark). Stimulation was delivered over the left or right dlPFC (depending on randomization), defined according to the 10-20 system (Herwig et al., 2003). A stereotaxic neuronavigation system was used (Brainsight, Rogue Research), and a template brain (MNI) was registered to each participant’s head using anatomical landmarks. Twenty trains of 10 Hz stimulation were delivered over 4 s with a 26 s inter-train interval at 120% of rMT (Rossi et al., 2009; Wassermann, 1998). For participants who expressed sensitivity to the initial TMS pulses, a ‘ramping up’ procedure was implemented in which the first 2-10 trains were delivered at 80% rMT intensity, and increased by 5-10% during the habituation period as tolerated until target intensity was reached. Only participants who received the majority of stimulation above rMT were included in the analyses leading to 5 participants (including the pilot participant) to be excluded. Sham stimulation was delivered using the opposite, shielded face of the same A/P coil. Sham electrodes were put near the hairline for all participants and were only activated for sham participants.
Ambulatory Assessment and Follow-Up Visit
Following the intervention participants received 8 calls/day for 7 days, starting the day after the intervention, at pseudo-random times. On each call participants were asked if they had used CR since the previous call (yes/no) and current level of distress (0 to 9; secondary outcomes). At the end of the week the battery of self-reports from intake was administered again via an online link.
Participants returned to the research office a month later to complete a stressor task without rTMS, an exit interview, and the self-reports. The stressor task included measuring HR while the participant stood still for a 300-s baseline, underwent a 120 s stress induction using a fourth autobiographical stressor, and engaged in prompted CR for 300 s. After the exit interview, the blind was broken and the experimenter and the subject debriefed about the study.
Measures
Administered self-reports (secondary outcomes) included the ERQ, the Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004), the Outcome Questionnaire (OQ-45; Lambert et al., 1996) and the Work and Social Adjustment Scale (WSAS; Mundt et al., 2002). At the end of the intervention participants were asked to give their best guess whether they received real or sham rTMS (forced-choice question) and to rate their confidence in their response (1 = “I am certain I received sham stimulation” to 9 = “I am certain I received active stimulation”) after the intervention and at the 1 week and 1 month follow up assessments. A previously developed in-house interview (Neacsiu, Luber, et al., 2018) was administered at the 1-month follow up, and it included Likert-type questions about feasibility, acceptability, and overall satisfaction rated on a scale from 0 (not at all) to 9 (extremely; secondary outcomes). Satisfaction was rated on a 0 (low) to 100 (high) continuous scale.
Psychophysiological Measures
Each session period (e.g., baseline, habituation, stressor, and regulation) was divided into 120 s bins, and HF-HRV (primary measure) was extracted from cleaned raw ECG from each bin following established guidelines (Berntson et al., 1997). For each baseline, HR was averaged from the last 240 of the total 300 s excluding any disruptions. Time to return to one’s HR baseline during the regulation period (regulation duration; primary measure) was defined for each regulation period as the amount of time it took from the beginning of regulation for the continuously monitored HR to reach a value that was lower or equal to the average pre-stimulus baseline HR. A baseline value for regulation duration was computed as the time it took during habituation for the person to return to HR baseline after the rTMS driven increased arousal.
Statistical Analyses
Mixed-effects hierarchical linear models (MMANOVA) with analytically determined covariance structures were used to analyze the repeated measures data (Molenberghs & Verbeke, 2000). Estimated marginal means (EMMs) were compared using LSD corrections for significant main and interaction effects. Effect sizes for these models were computed using Feingold’s formula (Feingold, 2009) and interpreted using Cohen’s (Cohen, 1977) specifications.
To test immediate effects of the intervention, we conducted three analyses examining HF-HRV, regulation duration, and SUDS. The treatment condition (active left, active right, or sham), the experimental condition (regulation 1, 2, and 3), the time within each experimental condition (coded 0 to 4 for each 120 s segment within that period) were used to predict HF-HRV. Baseline HF-HRV was measured at the beginning of the experiment (session baseline) and right before each autobiographical stressor presentation (pre-stimulus baseline). Because active rTMS may have cumulative effects (Bäumer et al., 2003) that could influence the pre-stimulus baselines we included both session baseline and pre-stimulus baseline as covariates in the analyses. A MMANOVA examining regulation duration during each regulation period was also conducted using treatment condition, experimental condition, and baseline as predictors. All SUDS values collected after each pre-stimulus baseline, stressor, and regulation period during the intervention were included in the analysis. SUDS after habituation were covaried as the closest ‘baseline’ measurement before the intervention began. Treatment condition, time period of the experiment (post-baseline, post-habituation, post-stressor 1, post-regulation 1, etc.), were used in the MMANOVA model. Two generalized estimated equations models (GEE) (Liang & Zeger, 1986) using ordinal logistic models and an independent covariance structure examined differences between treatment conditions in side effects.
To test near-term effects of the intervention two hierarchical linear models (HLM) (Raudenbush & Bryk, 2002) were used to examine condition differences in SUDS and use of CR during the ambulatory assessments. Data were aggregated by either obtaining a mean (SUDS) or a sum (use of CR – yes/no) for that day’s data. The SUDS rating before the intervention day was covaried as baseline in the SUDS model, and intake ERQ reappraisal in the CR model.
To test the long-term effects of the intervention, six MMANOVA models were conducted: four examining between-condition differences at the 1-week and 1-month follow-up assessments in ERQ, DERS, OQ-45, and WSAS; and two examining HF-HRV and SUDS during the follow-up stressor task. Baseline was covaried in all of these models. An ANCOVA examined differences in regulation duration at follow up.
Results
Missing Data
Out of 47 ITT participants, one was a pilot subject who did not provide usable data and four could not tolerate rTMS. Therefore, 42 participants are included in the final outcome analyses (see Table 1 for demographic data by condition). Across the 11 variables examined for those included, only 2.2% of the data was missing. Of the 2133 calls placed during the ambulatory assessment, participants answered 70.50% of calls (see supplemental materials).
Demographics and Clinical Descriptives by Group
Normality Assumption
At both intervention and follow up, HF-HRV was transformed using the function lg10*(HF-HRV*1000000) and regulation duration was transformed with a lg10 function. Intervention SUDS were also transformed to normal distribution using a square root function.
Preliminary Analyses
There were no significant differences between conditions in age, gender distribution, marital status, sexual orientation, ethnicity and race, income, self-reported use of reappraisal, presence of a depressive disorder, use of psychotropic medication, or number of current or lifetime diagnoses met (ps > .05). Therefore, our randomization procedures were successful, and none of these variables were included in the outcome analyses as covariates.
Self-reported dissociation during the experimental tasks was minimal (MDissociation range: 0.09 and 0.16) and therefore was not included as a covariate. To examine if neurostimulation alone affected HF-HRV or SUDS, we conducted MMANOVA analyses using the data collected during the habituation period (i.e., when participants were administered rTMS or sham alone for 10 minutes without any other instructions) and controlling for baseline. There was no significant difference between conditions in HF-HRV or SUDS (ps > .10) during habituation. Therefore, HF-HRV or SUDs during habituation were not included as covariates.
Tolerability and Acceptability
Participants found the overall study procedures very acceptable (MAcceptability = 7.27, S.D. = 1.11) and feasible (MFeasibility = 6.67, S.D. = 0.75) with no differences between conditions (ps > .05). Participants were 60.88% (SD = 32.99) likely to recommend this treatment to a friend. Participants reported low distress induced by the neurostimulation while they were engaging in CR (M = 2.10; SD = 0.98) and moderate to low interference with concentrating from the TMS noise (M= 4.65, SD = 2.41). When asked the open-ended question “what was it like being in our study?”, all participants reported that they had a positive experience, found CR very useful, and thought it had a positive impact on their lives. There were no serious adverse events in the study.
There was no significant difference between conditions in headache (Wald χ2 [2] = 0.11, p = .95) and neck pain (Wald χ2 [2] = 2.61, p = .27). Four participants in the active left condition (38.5%), seven in the active right condition (46.7%) and five in the sham condition (35.7%) reported worsened headache after the intervention. One participant in the active left condition reported mild hearing problems following our procedures. One participant from the active left and three from the active right condition reported mild scalp discomfort following the intervention. No one experienced a seizure.
All participants reported being naïve to rTMS at the beginning of treatment. A chi-square analysis indicated a trend for sham participants to be more likely than active participants to guess their condition assignment (χ2 [2] = 5.59, p = .06) when using a forced-choice question. Nevertheless, the majority of participants were not confident in their choice (52.4% indicated it could have been either) or thought they more likely than not received real stimulation (47.6%). There was no significant difference between conditions in confidence about assigned condition right after the intervention (F[2,39] = 1.95, p = .16), or at the one week (F[2,36] = 1.57, p = .22), or one month follow up (F[2,37] = 0.10, p = .90). These findings suggest that the procedures are feasible and acceptable and that the blinding procedures described were successful.
The Effect of the Combined Intervention on Immediate Emotion Regulation
To test whether neurostimulation would enhance emotion regulation at the moment when it is administered, we conducted three MMANOVA analyses using HF-HRV during regulation, SUDS after regulation, and regulation duration as dependent variables.
Heart Rate Variability Results
The MMANOVA analysis of HF-HRV, using an unstructured covariance, showed a significant main effect of treatment condition (F[2, 31.46] = 4.33, p <.05), time within each regulation period (F[4, 38.92] = 5.64, p <.005), session baseline (F[1, 64.17] = 98.53), and pre-stimulus baseline F[1, 77.83] = 61.21, p <.001), as well as a significant treatment condition by time interaction (F[8, 41.19] = 2.92, p < .05). Active rTMS over the left (EMMlg_HF-HRV = 1.92, S.E. = .04) and right dlPFC (EMMlg_HF-HRV= 1.87, S.E. = .04) led to significantly higher HF-HRV (i.e., enhanced emotion regulation) when compared to sham rTMS (EMMlg_HF-HRV= 1.76, S.E. = .04), ps < .05. Effect sizes for these differences were small (ds = 0.2 - 0.3). There was no significant difference between the active rTMS conditions (p = .37). Emotion regulation improved over time across conditions. HF-HRV extracted from later time bins were significantly higher than values extracted from earlier time bins (ps < .05). LSD-corrected pairwise comparisons of the significant interaction effect showed a significant difference between active and sham stimulation for the first 6 minutes of regulation (ps < .05), and no difference for the last 4 minutes, suggesting that sham participants “caught up” with active participants in their emotion regulation by the end of the regulation period (See Figure 2). There was no significant main effect of experimental time period (p = .07; i.e., difference between regulation attempt 1, 2 or 3).
Changes in high frequency heart rate variability (HF-HRV; a marker of effective emotional regulation) summarized across the three regulation periods allotted during the intervention, and separated by condition, and by time segment. The original HF-HRV value was multiplied by 100000 and transformed using a log function to achieve normality. Each time point on the graph represents the estimate marginal mean from the LSD pairwise comparisons from the MMANOVA main analysis which accounts for covariates. Regulation periods followed autobiographical negative emotional inductions.
Self-Report Results
A MMANOVA with an unstructured covariance was the best fitted model to examine SUDS changes. The experimental instructions (i.e., try to relax, remember a stressor, use CR) successfully modulated distress across conditions (F[2, 38.74] = 78.04, p < .001). Participants increased an average of 3.10 points on the SUDS scale (S.D. = 2.05; range: -3 to 9) after the stressor induction and dropped an average of 2.49 points (S.D. = 2.07; range -2 to 9) after regulation, differences that were significant (p < .001; Figure 3).
Changes in subjective units of distress (SUDS; a marker of negative emotional arousal) summarized across the three experimental periods, and separated by condition, and by instruction provided. The original SUDS value was transformed using a square root function to achieve normality. Each time point on the graph represents the estimate marginal mean from the LSD pairwise comparisons from the MMANOVA main analysis which accounts for covariates. The graph presents data at the end of the pre-stimulus baseline, after the negative emotional induction using the autobiographical stressor, and at the end of the regulation period.
There was a trend for a significant fixed effect for treatment condition (F[2,35.60] = 2.99, p = .06) driven by significantly lower SUDS in the active left rTMS condition (EMMSR_SUDS = 1.17; S.E. = .08) across the experimental day when compared to sham (EMMSR_SUDS = 1.40; S.E. = .08; d = .32; p < .05) and active right rTMS (EMMSR_SUDS = 1.41; S.E. = .08, d = .33; p < .05) participants. LSD corrected pairwise comparisons did not find a significant difference between right and sham rTMS. The treatment condition by experimental instruction interaction was not significant (F[4, 39.12] = 1.75, p = .16; see Figure 3).
Regulation Duration Analyses
For the 36 participants who had above pre-stimulus baseline HR at the beginning of the regulation period, it took almost twice as long to return to HR baseline during regulation if receiving sham (Msham = 129.08, S.D. = 194.57, N = 11) versus active neurostimulation (Mactive_right= 69.06, S.D. = 103.14, N = 13; Mactive_left = 74.20, S.D. = 110.64, N = 12). A MMANOVA using autoregressive covariance structure found a trend for a significant main effect of treatment condition (F[2, 34.60] = 3.18, p = .054) favoring the active conditions, and no main effect for time in the experiment (p = .18). Regulation duration was significantly longer in the sham condition when compared to the active right condition (p < .05; d= .67) and near significantly longer when compared to the active left condition (p = .08; d = .50). There were no differences between active conditions (p = .51).
In summary, we found preliminary evidence that when compared to CR with sham neurostimulation, rTMS administered in conjunction with CR enhances physiological (HF-HRV; regulation duration) and, after left rTMS, self-reported (SUDS) indices of emotion regulation.
The Near-Term Effect of the Combined Intervention on Psychopathology and Use of CR
Ambulatory Assessment Results
Participants indicated having used CR since the previous call in 745 instances (43.90% of answered calls) and reported SUDs above 0 in 1188 calls (70.34% of answered calls). Participants who received active left rTMS stimulation reported significantly more frequent use of CR (EMM = 3.32, SE = 0.33) when compared to participants who received active right rTMS (EMM = 1.94, SE = 0.32; d = .77; p < .001) or sham rTMS (EMM = 2.51, SE = 0.32; d = .45; p < .05; FTREATMENT_CONDITION[2, 190.80] = 6.40, p < .005). Participants who received sham rTMS did not significantly differ from the active right group (p = .14). Higher self-reported use of reappraisal at baseline predicted higher use of CR during the ambulatory week (Parameter estimate = .72, S.E. = .20; F(1, 190.58) = 12.69, p < .001). There was no significant change over time during the ambulatory week in CR use (p = .22).
We also found a trend for a significant main effect of treatment condition on average daily distress (FSUDS[2, 196.13] = 2.45, p = .09). LSD-controlled pairwise comparisons showed that participants who received active left rTMS reported significantly less average distress during the ambulatory week (EMM = 1.39, SE = 0.22) when compared to participants who received sham rTMS (EMM = 1.94, SE = 0.21; d = .35; p < .05). Participants who received active right rTMS (EMM = 1.79, SE = 0.21) did not significantly differ in their average daily distress when compared to either the active left (p = .11) or the sham group (p = .55). Higher self-reported SUDS before the intervention predicted higher SUDS during the ambulatory week (F[1, 197.79] = 3.98, p < .05). There was no significant main effect of time (p = .50).
In summary, active left rTMS led to more frequent use of CR and reduced daily distress when compared to sham and right rTMS. There was no near-term benefit for right rTMS when compared to sham. Figure 4 depicts these results.
Changes in subjective units of distress (SUDS; a marker of negative emotional arousal) averaged for each day (panel A) and in use of cognitive restructuring (CR) summed for each day (panel B), separated by condition, and by day. Each time point on the graph represents the estimate marginal mean which accounts for covariates.
One week and one month self-report assessment results
Paired sample t-tests indicated that all participants significantly improved in their use of reappraisal (t[40] = 10.02, d = 2.24, p < .001), in their emotional dysregulation (t[40] = -3.64, d = .81, p < .001), general psychopathology (t[40] = -2.13, d = .48, p < .05), and social and occupational functioning (t[39] = -2.41, d = .54, p < .05) from before to one week after the intervention. Using Jacobson and Truax’s method (118) and the intake reliability of the ERQ reappraisal (α = .87), we computed a reliable change index of 4.72 (SE = .29), which indicates that the change in reappraisal was meaningful and unlikely to have happened due to measurement error. For 13 participants the change in psychopathology from pre to post was of more than 14 points, which indicates reliable change (1992).
MMANOVA models using a compound symmetry covariance structure indicated no significant differences between conditions or condition by time interactions for reappraisal, overall difficulties in emotion regulation, or overall psychopathology severity (ps > .05). There was a significant time by condition interaction for work and social functioning impairment (F[2, 36.91] = 3.27, p < .05) driven by a significantly higher improvement from the 1-week (EMM = 14.47; S.E. = 2.16) to 1-month follow up (EMM = 9.89; S.E. = 2.29; d = .71) in the active left rTMS condition when compared to the other treatment conditions. All participants continued to improve from the 1-week to the 1-month follow up in their emotion dysregulation (F[1, 36.77] = 8.70, d = .22, p <.01) and psychopathology (F[1, 36.75] = 7.43, d = .21, p < .01) with no significant difference between treatment conditions (Table 2).
Means and SDs for longitudinal outcomes for all participants and broken by condition
Exploratory Secondary Analyses
A secondary longitudinal MMANOVA analysis examined differences between treatment conditions in emotional dysregulation controlling for psychopathology severity and ERQ reappraisal in addition to baseline dysregulation. In this model there was a significant time by treatment condition interaction (F[2, 34.10] = 3.32, p < .05) that was driven by a significantly larger improvement at the 1-week follow-up between participants who received active right rTMS (EMMDERS_Total = 80.21; S.E. = 4.05) when compared to sham participants (EMM = 93.85, S.E. = 4.42, d = .59, p < .05). In addition, higher psychopathology (FOQ_45[1, 70.18] = 45.68, p < .001) and lower reappraisal (FERQ_R[1, 58.80] = 5.39, p < .05) predicted emotional dysregulation.
The longitudinal outcome models were also investigated within the main diagnostic groups to identify any specific diagnostic profiles that may respond better to our intervention. We found a significant main effect of treatment condition (FOQ_45_Total [2, 11.83] = 4.18; FDERS_Total [2, 11.78] = 4.20, ps < .05), favoring stimulation over the right dlPFC within the model including only participants who met criteria (current or in the past) for a trauma or stressor-related disorder(post-traumatic stress, adjustment, acute stress, or other specified disorder, N=16). These participants (6 in each active, and 4 in the sham condition) reported significantly lower emotional dysregulation and psychopathology following active right rTMS (EMMDERS_Total = 64.42, S.E. = 9.21; EMMOQ_45_Total = 37.98, S.E. = 8.62) when compared to those who received active left rTMS (EMMDERS_Total = 100.90; S.E. = 9.53; EMMOQ_45_Total= 73.42; S.E. = 8.78; ps < .05). The difference between active right rTMS and sham rTMS (EMM = 93.30; S.E. = 11.56) was marginally significant for emotional dysregulation (p = .08) but not for psychopathology (EMM = 51.46, S.E. = 10.58, p = .14). We found no significant differences in results for participants with a history of an anxiety (N = 34), depressive (N = 35), or compulsive (N = 10) disorders (including current; p >.05).
In summary, these findings suggest that CR training and practice during a one-time session is successful in reducing psychopathology, emotional dysregulation and in improving reappraisal. Active left rTMS accelerated improvements between the one week and the one month follow up in functioning. Active right rTMS was superior to sham in exploratory analyses where the severity of psychopathology and acquisition of reappraisal were controlled for. Furthermore, right rTMS was especially effective in augmenting CR training for participants who meet criteria for trauma and stress disorders.
One Month Follow-Up Behavioral and Physiological Results
Participants were successful during the follow-up stressor task to increase their subjective distress from baseline (ΔSUDSstressor-baseline = 2.30, S.E. = .36, p < .001) and then regulate it back to baseline (ΔSUDS_stressor-regulation = 2.35, S.E. = .36, p < .001; ΔSUDSregulation-baseline = -.05, S.E. = .29, p < .001) using a fourth autobiographical stressor and CR as a regulation technique (Ftime [2, 78.48] = 25.12, p < .001). There were no significant differences between treatment conditions in the success of regulation when examined via either HF-HRV or SUDS (FHF-HRV [2, 34] = 1.62; FSUDS [2, 99.71] = 1.00; ps > .05). A univariate GLM test of regulation duration (M = 40.31s, SD =59.10) showed no significant difference between treatment conditions (p = .71; N = 27). In summary, participants maintained their regulation ability at follow up, but no effect of rTMS augmentation could be seen one month after its administration on these measures.
Discussion
This proof-of-concept study demonstrated the feasibility of augmenting emotion regulation skills training and practice with rTMS in a one-time intervention designed to target a dysregulated neural circuit. Neurostimulation enhanced emotion regulation in the moment according to behavioral, physiological, and self-reported indices. The augmented gains could be seen up to one week later and all participants maintained their gains a month later. Participants tolerated the procedures well, found the approach acceptable, and reported having a positive experience. These results are encouraging of continued efforts to combine neurostimulation with skills training in brief, transdiagnostic interventions.
All participants learned CR in a 45-minute session following a standardized scripted protocol and then practiced CR on their own autobiographical stressors. This training by itself, which can be easily integrated in any practice, proved valuable and led to behavioral improvements. The increase in use of reappraisal from before to after the intervention was significant (corresponding to a large effect size, 2.25), and the participants reported improved emotional dysregulation, psychopathology, and overall functioning (corresponding to moderate effect sizes). This improvement was maintained at a one-month follow-up assessment. For 30.9% of our sample, the improvement in psychopathology was clinically significant.
The effect size of the pre-post change in reappraisal (d = 2.25) is larger than found in traditional 16-week CBT trials (ds = 1.06-1.07; Klumpp et al., 2017; O’Toole et al., 2019). Changes in psychopathology (d = 0.48) were lower than traditional 16-week emotion regulation interventions (d = 1.46) but higher than treatment as usual (d = 0.34; Neacsiu, Rompogren, et al., 2018). Changes in emotional dysregulation (d = 0.81) were lower than 16 to 20 week emotion regulation interventions. For example, a 20 week-long emotion regulation skills training intervention for GAD yielded a pre-post effect of 1.02 in changes in difficulties with emotion regulation (Mennin et al., 2018). An 18-session transdiagnostic treatment for anxiety disorders (Barlow et al., 2017; Farchione et al., 2012) found a pre-post effect size ranging from 1.08 and 1.30 for the primary self-report outcome. In our prior study, a 16-week long skills training intervention for transdiagnostic emotional dysregulation yielded a 2.79 pre-post effect size on the DERS (Neacsiu et al., 2014). While there is room for improvement [e.g., the rate of clinical response found by longer interventions is almost double (Farchione et al., 2012; Neacsiu et al., 2014)], finding half the response benefit while administering only one CR session is worthy of future investigation.
Our goal was to enhance this training with neurostimulation over the dlPFC, a region that is critical for successful implementation of cognitively-mediated emotion regulation strategies (Ochsner et al., 2012; Powers & LaBar, 2019). The addition of active rTMS enhanced emotion regulation as measured by HF-HRV, regulation duration, and self-reported distress. Active stimulation got participants to calm down faster. Sham participants achieved the same regulation effect over a longer (almost double) time frame. Therefore, neurostimulation played an important role in enhancing the utility of the emotion regulation skill in the moment.
In the week following the intervention, having received active stimulation meant more use of reappraisal and lower daily distress. Furthermore, secondary analyses also found a significantly higher reduction in emotional dysregulation following active right neurostimulation if the analysis included moderators (i.e., increasing power to detect differences), or if the group examined included those with a diagnosis of trauma and stress disorders. A month after the intervention, participants who received active left stimulation continued to improve in their social and occupational functioning. Behavioral and physiological indices of emotion regulation a month after the intervention were not different between treatment groups. Taken together these findings highlight that neurostimulation also enhances acquisition of behavioral skills with noticeable effects up to a week after a one time session. Furthermore, for those who meet criteria for trauma or stress disorders, a one-time combined intervention using right dlPFC neurostimulation may provide clinically significant improvement in their psychopathology and emotional dysregulation. Given that existing protocols that combine psychotherapy and neurostimulation are 16 to 20 weeks long (Kozel et al., 2018; 2019), our approach provides a promising avenue for future research. Future research should investigate if a higher dose of rTMS or more sessions enhance the acquisition augmentation beyond one month..
With regards to which target was optimal to enhance emotion regulation, our results were mixed. Both left and right targeting increased HF-HRV significantly above sham stimulation. Right rTMS was significantly better than sham in terms of reducing the regulation duration, while left rTMS was only marginally significant. During the week after the intervention only left rTMS showed a significant improvements above sham in skill acquisition and reduced distress. At the end of the ambulatory week, right stimulation alone showed significant benefits in self-reported emotion regulation in secondary analyses.
The literature is also inconsistent with benefits to emotion regulation being shown in studies that employ either left or right rTMS (De Raedt et al., 2010; Feeser et al., 2014; Fonzo et al., 2017). Neuroscientific meta-analyses point to increases in both right and left dlPFC recruitment for downregulation when compared to upregulation of emotion (Kohn et al., 2014; Ochsner et al., 2012). In addition, the left versus right recruitment may depend on the tactic employed. We taught participants both reframing and distancing and encouraged practice of both tactics. Ochsner et al. (2012) found that reframing recruits primarily left prefrontal and temporal cortices, while distancing is more lateralized in the right prefrontal cortex, a finding also supported by Powers and LaBar (2019). Therefore, response to left or right dlPFC stimulation may depend on the tactic that was easier for each participant to apply. A meta-analysis using clinical samples found equal numbers of different studies that showed activation in the right, left, or bilateral dlPFC in emotion regulation tasks (Zilverstand et al., 2017). Taken together these studies and our findings suggest that engaging either the right or the left dlPFC with emotion regulation or neurostimulation is effective in reducing emotional arousal. Furthermore, excitatory neurostimulation administered over the left or right dlPFC may appear to have different effects depending on the outcome measure administered. Further work is needed to better understand differences between left and right neurostimulation.
This study fills a gap in knowledge about how to combine emotion regulation skills training with neuromodulation as an intervention for a transdiagnostic clinical sample. We used the standard TMS frequency of 10 Hz to enhance cortical excitability in the DLPFC, in line with the findings discussed in the introduction that point to transdiagnostic hypoactivity in this region during emotion regulation tasks. We also employed neuronavigation using the 10-20 system. This approach provides a more precise localization of the dlPFC when compared to the more traditional 5 cm rule (McClintock et al., 2018). We opted to divide the TMS session into 3 epochs in order to make engagement in autobiographical emotion regulation feasible. This decision was aligned with prior research highlighting that breaks may enhance facilitatory after-effects (Rothkegel et al., 2010). Furthermore, having participants spend more than 10 minutes in effortful emotion regulation would be difficult to accomplish and receiving rTMS during emotional induction would dampen the emotional experience (Remue et al., 2016), reducing the opportunity for regulation. Our results demonstrate that these decisions did not alter the potential of neurostimulation to enhance behavioral training and that therapeutic activities, like skills practice, could take place during rTMS successfully.
Sixteen participants dropped after their intake appointment (25.39%). This session mirrored an initial intake session for psychotherapy where problems to be discussed are visited and a diagnostic assessment is conducted. In traditional psychotherapy, 35 to 50% of participants drop after this first session (Barrett et al., 2008). Therefore, the study dropout rate after this first session was below the norm. Five participants could not tolerate the neurostimulation procedures (8.51%). In large trials using HF rTMS for depression, dropout rates related to tolerating rTMS range from 6-8% (Berlim et al., 2013). In previous studies conducted at Duke University Medical Center with healthy adults, the dropout rate related to rTMS tolerance was 7.9% (Beynel et al., 2019). Therefore, the dropout rate from rTMS here was comparable with other studies, supporting that this concurrent approach is feasible and acceptable.
The study had some important limitations including the sample size and lack of matched control group for the skills training component. With regard to the sample size, this proof-of-concept study tested 15 adults per condition with the primary goals of establishing feasibility. While the study was powered to detect large effect sizes, it should be noted that a recent meta-analysis of 126 online rTMS studies (Beynel et al., 2019) found an average sample size of nine individuals per treatment group. By this standard, the current study was larger than typical in the field, however, future studies should perform specific efficacy power analyses and adhere to resulting sample size criteria. Multiple comparisons were not accounted for given the small sample; nevertheless effect sizes were computed and reported to offset this limitation. Targeting for rTMS in this study used exterior head measurements and, therefore, may have not resulted in excitation of a relevant node of the emotion regulation neural network. E-field modeling (Deng et al., 2013), which is recommended to optimize rTMS administration, was not employed. Future studies are needed to replicate findings and refine procedures with larger samples.
In summary, our goal was to begin the development of a time-limited combined intervention for transdiagnostic emotional dysregulation. We found promising evidence that a rTMS augments behavioral skills training in the moment, and that the onetime session of combined rTMS-CR intervention can have effects that last as long as a week. Emotional dysregulation is a pervasive and serious public health problem in need of novel interventions. Our findings provide a promising first step in the development of novel neuroscience-driven treatment paradigms that address this hard-to-treat, transdiagnostic clinical population.
Data Availability
No external datasets or supplemental materials are available at this time. Data sets and protocol materials can be obtained by e-mailing the first author (andrada.neacsiu{at}duke.edu)
Footnotes
Data on the present paper were also presented as part of several conference talks. This research and the completion of the manuscript were supported by a Brain and Behavior Young Investigator Award, a Duke internal award, and KL2 award granted to the first author and by the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number 5KL2TR001115. The authors have no conflicts of interest to declare. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The authors would like to thank the participants and to acknowledge Lisalynn Kelley, Zach Rosenthal, PhD, Caitlin Fang, PhD, Kevin Haworth, PhD, Megan Renna, PhD, Paul Geiger, PhD, Simon Davis, PhD, Moria Smoski, PhD, Tim Strauman, PhD, Michael Babyak, PhD our research assistants and DUHS staff for their contributions.
References
