Candidate Targets for Resilience Training to Reduce Transdiagnostic Risk for Mental Illness

Abstract

Background and Hypothesis

Stress sensitivity may represent an important target for resilience-promoting, preventive interventions. Resilience Training (RT) is a 4-session, group-based behavioral intervention, focusing on mindfulness-based skills, that leads to reductions in psychopathology. To investigate the mechanisms of RT, the current study tested whether acquiring the skills taught in RT leads to decreases in psychopathology via reductions in one manifestation of stress sensitivity, emotion reactivity, and associated changes in hippocampal-frontal connectivity.

Study Design

An open trial of RT was conducted in 103 non-help-seeking young adults with mild-to-moderate psychotic experiences (PEs) and/or symptoms of depression. Transdiagnostic symptoms, emotion reactivity, and mindfulness-related skills were measured, and, in a subset of participants (n = 41), resting-state, functional magnetic resonance imaging (fMRI) data were collected, before and after completion of RT.

Study Results

Replicating and extending findings of prior studies of RT, significant increases in mindfulness-related RT skills and significant decreases in transdiagnostic symptoms and emotion reactivity, as well as changes in hippocampal-frontal functional connectivity, were observed following RT (all P < .02). Mediation analyses revealed that associations between the acquisition of the RT skills and decreases in symptoms (P < .006) were fully mediated by the decrease in emotion reactivity, which was also correlated with a significant pre-to-post increase in hippocampal-dorsolateral prefrontal cortex connectivity (P = .032).

Conclusions

RT may lead to improvements in mental health by increasing the capacity to manage day-to-day stress. Future randomized controlled trials with extended follow-up can determine whether such improvements decrease the likelihood of developing disabling mental illnesses in transdiagnostically at-risk individuals.

Introduction

Efforts to prevent the onset of psychotic disorders such as schizophrenia have primarily focused on developing and testing potentially preventive interventions for young people experiencing subclinical psychotic symptoms, sometimes referred to as psychotic experiences (PEs). Although PEs were initially thought to be relatively specific markers of risk for psychotic illnesses, a variety of evidence has shown that they represent transdiagnostic risk factors for a range of psychiatric conditions, including (but not limited to) psychotic disorders^1–6. Individuals with PEs, including those who meet criteria for a clinical high risk for psychosis (CHR-P) syndrome, have a greater than average likelihood of either having, or later developing, mood and anxiety disorders^1–3,7,8, suicidal thoughts and behaviors^9–12 and overall mental illness-related disability^13–16, in addition to having an elevated risk (4-10-fold) for developing a psychotic disorder^17–20. Thus, based on this research, PEs have been largely re-conceptualized as representing transdiagnostic markers of co-occurring or future psychopathology^3,21–24. This rapidly evolving line of work also suggests that interventions targeting young people with PEs are likely to be beneficial for reducing the incidence or severity of a wide range of mental illnesses, with the potential for broad public health impact.

However, to date, no interventions have been consistently shown to diminish or eliminate PEs, or to prevent the onset of serious mental illnesses in individuals with PEs^25,26, although trials conducted in CHR samples have found evidence that cognitive-behavioral approaches are beneficial for at least a subset of this population²⁷. A clearer understanding of the neurocognitive mechanisms underlying PEs could inform the design of additional, novel interventions for young people with these (and related) symptoms. Although the neurobiological mechanisms of PEs are incompletely understood, there is a variety of evidence suggesting that these symptoms are associated with a heightened sensitivity to stress or an increase in “emotion reactivity” to day-to-day stressors^28,29. Emotion reactivity is assessed by measuring the frequency, intensity, and duration of emotional responses to stress.^30,31 Using self-report measures, experience sampling, or physiological data, correlations between levels of emotion reactivity and PEs have been demonstrated in the general population^32–36 and in individuals with CHR-P^37–40. Thus, one potentially beneficial strategy for reducing PEs, as well as the associated symptoms and risk for poor outcomes, is to focus on reducing stress sensitivity/emotion reactivity⁴¹.

A variety of interventions are known to reduce reactivity to stress via enhancing emotional regulation and generally increasing emotional “resilience”—which is often defined as the capacity to avoid the negative consequences of, or actively adapt to, adversity or daily stress^42,43. Training in “mindfulness,” the ability to fully focus on the experiences occurring in the present, in a non-judgmental, accepting, non-reactive manner⁴⁴, has been associated with reductions in physiological responses to stress^45–47 and emotion reactivity^46,48,49, and with related reductions in depression, anxiety, and distress associated with chronic pain^50–54.

Based on this literature, we designed a brief behavioral intervention, called Resilience Training (RT), for non-help-seeking young adults with low-level, transdiagnostic symptoms of psychopathology^55–57. RT is a 4-session, group-based intervention that teaches 3 mindfulness-based techniques, including basic mindfulness and two additional skills that rely on mindfulness as a foundation: mindful self-compassion (the ability to respond to one’s negative emotions with acceptance and kindness)⁵⁸, and mentalization (the ability to non-judgmentally represent and reflect on one’s own and others’ mental states)⁵⁹. RT was designed for college students, because the age range of most college students (~18-22 years) overlaps with a peak period of risk for the onset of many mental illnesses⁶⁰. In addition, the college campus provides a relatively controlled setting for delivering and evaluating a potentially preventive intervention in a scalable, uniform manner.

In two previous trials of RT and the current study, college students were enrolled if they were experiencing PEs and/or symptoms of depression, since PEs and depression are highly co-occurring^24,61–63 and may have shared underlying mechanisms^64–67. Like PEs, depressive symptomatology is a marker of transdiagnostic risk for mental illnesses^68–70. RT was first tested in a single-arm pilot feasibility study⁵⁵ and, subsequently, in a randomized waitlist-controlled trial⁵⁶ that showed that receiving RT was associated with significantly greater reductions in PEs, PE-related distress, depression, and anxiety, as well as significantly greater increases in mindfulness and mindful self-compassion, compared to a waitlist condition. In the current open trial, we aimed to extend this work by testing whether the acquisition of mindfulness-related skills during RT leads to reductions in emotion reactivity,^46,48,49 and, if this proves to be the case, whether this effect is linked to subsequent decreases in symptoms of psychopathology. In addition, we used resting-state functional magnetic resonance imaging (fMRI) to explore whether changes in the functional connectivity of hippocampal-frontal circuits known to be involved in emotion regulation and responses to stress⁷¹ play a role in the effects of RT.

Methods

Overall Study Design and Procedures

Mental Health Screenings

Recruitment for this project occurred in two phases. The first phase was part of a larger longitudinal study examining factors that influence the trajectory of psychopathology in college students. As described previously^55,56, mental health screenings of college students were conducted during this study either on campus, using paper or online questionnaires, or via a link to an online survey included in an email. The students of the current study were enrolled in one of 5 universities, with student body sizes ranging from approximately 4,000 to 33,000, either located in Massachusetts (n = 4) or southern California (n = 1). These included a large urban university, a small liberal arts college, a mid-sized science and technology-focused institution, a two-year community college, and a large state university. Participants were compensated $30 for completing the screening survey. The screening material included the informed consent form for the screening survey, questions about participants’ demographic characteristics, as well as a battery of questionnaires, including the two scales that were used to determine eligibility for participation in a one-arm trial of RT: (1) the Peters et al Delusions Inventory (PDI)⁷² and (2) the Beck Depression Inventory (BDI)⁷³.

Screening, Eligibility, and Recruitment for the RT Trial

The 21-item PDI was chosen for measuring a subset of PEs because it has been well-validated in college student samples^74–76 and, unlike many measures of PEs employed in clinical settings, it uses non-stigmatizing, non-clinical language to describe common delusional experiences (eg, paranoia, grandiosity, ideas of reference) and some common perceptual or experiential aberrations (eg, experiences of being controlled, thought insertion, withdrawal and echoing, but not frank hallucinations) that are frequently experienced in the general population and in those who have transdiagnostic risk for mental illnesses. The rates of endorsement of the 21 items of the PDI in the current sample are shown in Figure S1. The BDI was used to measure symptoms of depression, another indicator or expression of traits (such as negative affectivity, neuroticism) associated with transdiagnostic risk for psychiatric illness,^68,69 which often precedes or accompanies the onset of psychotic symptoms^62,63,70,77.

Those who completed the questionnaires who were (1) between the ages of 18 and 30 years old (2) with mildly elevated scores on measures of PEs and/or depressive symptoms (with a PDI score > 3 and/or BDI score > 5) and (3) who were not receiving any current psychiatric treatment (psychotherapy or psychopharmacological, other than treatments for Attention Deficit Hyperactivity Disorder (ADHD)) were considered eligible for participation in the RT trial. We based the PDI and BDI cut-off scores on mean values identified in similar cohorts ^74,75,78,79. These criteria are typically met by approximately 50% of the college students screened (ie, students who volunteer to fill out a mental health survey) and are inclusive by design, since the goal was to identify young adults with some minimal level of risk for developing, in the future, a serious, debilitating mental illness. Thus, the cohort was comprised of young people who were experiencing mild-to-moderate symptoms of psychopathology, who were functioning relatively well (enrolled as full-time college students) and not seeking clinical care.

An overview of RT was provided to eligible individuals, and those who were interested in participating in the trial completed a second informed consent form that described the RT trial portion of the study, which included an optional MRI scanning session before and after RT. All procedures of the study were approved by the Massachusetts General Brigham Institutional Review Board (IRB), as well as the IRB of each of the participating undergraduate institutions, and all procedures were conducted in accordance with the Declaration of Helsinki.

To identify a minimum recruitment target/sample size, we conducted a power analysis using the effect sizes of our previous trials of RT. We estimated that the minimum sample size that would be sufficient to detect a medium effect size (power (1-β) = .80; significance (ɑ) = .05; d = .50) was 33 participants. We exceeded this number (with a total sample size of 103 for participation in the RT trial), since we anticipated that only a portion of the participants would choose to enroll in the optional MRI component of the study.

Of the 103 students who participated in RT, 41 elected to participate in the optional MRI scan session before and after RT. Participants earned $20 for each of the two assessment visits (pre- and post-RT), and $75 for completing each scanning session. The self-report measures collected pre and post-RT were also collected 6 months following the completion of RT, and participants were paid $30 for filling out those surveys. Demographic characteristics of the sample are described in Table 1.

Table 1.

Demographic Characteristics of the Sample

	Full sample (n = 103)		Scanned sample (n = 41)
	M	SD	M	SD
Age	19.5	1.7	19.4	1.6
Depressive symptoms (mean BDI score)	14.8	8.2	14.0	8.3
Psychotic experiences (mean PDI score)	5.1	3.2	6.0	3.2
	n	%	n	%
Sex (Female)	81	78.6	26	63.4
Race
American Indian or Alaskan Native	3	2.9	1	2.4
Asian	35	34.0	16	39.0
Black or African American	3	2.9	1	2.4
Native Hawaiian or Other Pacific Island	1	1.0	0	0.0
White	44	42.7	14	34.1
Other	17	16.5	9	22.0
Ethnicity
Hispanic or Latinx	23	22.3	8	19.5
Not Hispanic or Latinx	80	77.7	33	80.5
Year of College
Freshman	44	42.7	18	43.9
Sophomore	29	28.2	13	31.7
Junior	18	17.5	6	14.6
Senior	12	11.7	4	9.8
Born Outside the United States	26	25.2	12	29.3
Family Income
Less than $24,999	11	10.7	5	12.2
$25,000–$49,999	14	13.6	5	12.2
$50,000–$79,999	23	22.3	8	19.5
$80,000–$199,999	31	30.1	11	26.8
$200,000+	8	7.8	4	9.8
Unknown	16	15.5	8	19.5

Resilience Training

This study began during the COVID-19 pandemic; the 103 participants completed the RT intervention (in 11 cohorts with separate 4-session workshops) between October 2020 and May 2022. Because many students were attending college remotely during this time, the RT intervention was delivered via videoconferencing (Zoom), for the first time. RT is comprised of four 90 minute-long group sessions and focuses on developing greater emotional resilience, as well as greater self and other awareness, via acquiring an ability to practice mindfulness, self-compassion, and mentalization skills. Each 90-minute session was led by two clinician facilitators (in the current trial, these included 3 doctoral-level clinical psychologists, a psychiatrist, and an advanced clinical psychology doctoral student). Zoom-delivered RT was very similar to in-person RT, with the addition of breakout rooms for activities that were done in pairs (see Table S1 for an overview of the sessions).

Because this was the first deployment of RT via videoconferencing, acceptability ratings (rated from 1 to 5, or 0 to 2) were collected following completion of RT, to assess whether RT was experienced as beneficial and useful by participants when delivered this way. For these ratings, participants rated how beneficial (1 = “not at all beneficial,” 5 = “very beneficial”), useful (1 = “not useful at all,” 5 = “very useful”), and novel (1 = “not novel at all,” 5 = “very novel”) RT was for them, and how likely it was that they would recommend RT to a friend (1 = “definitely would not recommend,” 5 = “definitely would recommend”). In addition, the participants rated how much they enjoyed participating in RT over Zoom (0 = “did not enjoy at all,” 1 = “unsure,” 2 = “very much enjoyed”) and whether or not they would have preferred to participate in RT in-person (0 = “would not prefer in-person,” 1 = “unsure,” 2 = “would prefer in-person”).

Measures

Symptoms of Psychopathology and Emotion Reactivity

As described above, PEs and distress due to PEs were assessed using the 21-item Peters et al. Delusions inventory (PDI). On the PDI, participants indicate whether or not (“yes” or “no”) they have experienced an aberrant belief or experience. If they have (endorsing “yes”), then they rate (on a scale of 1 to 5) the degree to which the experience is distressing, preoccupying, and the extent to which they believe that the experience is true or real (the distress, preoccupation and conviction ratings of the PDI, respectively). The total PDI score is calculated by summing the number of items that a participant endorses and the scores on the subscales are calculated by summing the scores on the ratings of the endorsed items (without normalizing by the total PDI score). In the current and previous studies of RT, the PDI distress subscale, rather than the other two subscales, was an outcome examined in the analyses, since prior research has shown that PE-associated distress is predictive of poor outcomes^80,81. In addition, levels of paranoia-related PEs, ie, a “paranoia factor” identified by a prior factor analysis⁸², were also measured, given the known associations between paranoia, stress, and affective processes⁸³. This paranoia factor included 6 items of the PDI (items 1, 3, 4, 5, 13, and 15; see Figure S1).

Also, symptoms of depression were assessed using the 21-item Beck Depression Inventory (BDI)⁷³, and anxiety symptoms were assessed using the trait subscale of the 40-item Spielberger State-Trait Anxiety Inventory (STAI-T)⁸⁴.

In addition, to test the proposed model of RT, emotion reactivity was assessed using the 21-item Emotion Reactivity Scale (ERS)³⁰; examples of items of this scale include: “I tend to get very emotional very easily,” “my moods are very strong and powerful” and “when I’m angry/upset, it takes me much longer than most people to calm down.” Emotion reactivity is defined as the extent to which an individual experiences their emotions: (1) to a wide range of stimuli (sensitivity), (2) strongly (intensity), and (3) for an extended period of time before arousal levels return to baseline (persistence)³⁰. Processes related to emotion reactivity can be differentiated to some extent from those involved in emotional regulation, which has been defined as: the regulatory processes that influence the type, timing, and expression of emotional responses, and the capacity to modulate emotions to meet situational goals and demands³¹.

Lastly, a related measure of stress sensitivity, perceived stress, was assessed using the 10-item Perceived Stress Scale (PSS)⁸⁵.

Resilience-related Skills or Capacities

Mindfulness, self-compassion, and mentalization were assessed using the 39-item Five Facet Mindfulness Questionnaire (FFMQ)⁸⁶, the 26-item Self-Compassion Scale (SCS)⁸⁷, and the 28-item Mentalization Scale (MentS)⁸⁸, respectively. In addition, overall resilience was assessed using the 25-item Connors-Davidson Resilience Scale (CDRISC)⁸⁹, a well-validated, commonly used self-report measure assessing cognitions and subjective experiences that have been linked to resilience.

Magnetic Resonance Imaging

Rationale and Hypotheses

In the resting-state fMRI analyses, we focused on measuring changes in the functional connectivity of the hippocampus following RT, because prior studies have shown evidence that (1) the hippocampus plays a central role in the responses of the brain to stress^90–92 and (2) abnormalities in hippocampal circuitry, particularly in hippocampal-frontal pathways, may contribute to transdiagnostic risk for psychopathology⁷¹. In the healthy human brain, the activity of the hippocampus and dorsolateral prefrontal cortex (dlPFC) are typically negatively correlated at rest, and prior studies using positron emission tomography or task-based fMRI methods have found evidence for greater or persistently anti-correlated functional coupling (ie, lower connectivity or more negative anti-correlations) between the hippocampus and the dlPFC during performance of working memory tasks in individuals with schizophrenia ⁹³ and first-degree relatives of individuals with schizophrenia⁹⁴. Based on these data and related findings, it has been proposed that, in healthy individuals, these two regions become uncoupled (their activity becomes less anti-correlated) during cognitively demanding tasks, and this uncoupling is disrupted in people with schizophrenia or psychosis risk⁹⁵. In addition, other studies, using resting-state fMRI, have found evidence for diminished deactivation or abnormally elevated connectivity between two typically positively correlated regions of the default network, the hippocampus and ventromedial prefrontal cortex (vmPFC), or within the default network as a whole, in mood and anxiety disorders^96,97. Thus, based on this prior literature, in the current study, we measured the functional connectivity of the hippocampal-dlPFC and hippocampal-vmPFC pathways, to test the prediction that RT “normalizes” the functional connectivity of these pathways.

To do this, we tested whether there are significant reductions in anti-correlations (leading to numerical increases in connectivity) between the hippocampus and dlPFC and significant reductions in positive correlations (numerical decreases in connectivity) between the hippocampal and vmPFC, following RT, using standard resting-state functional connectivity methods (see below).

In addition, we conducted several control analyses of the resting-state fMRI data, to test whether the amygdala-dlPFC and amygdala-vmPFC pathways showed similar changes (a positive control analysis) and to test whether 3 sensorimotor circuits showed no changes in functional connectivity following RT (a negative control analysis).

Acquisition

Before and after the intervention (“pre” and “post” RT), MRI scans were collected in Boston-based participants who opted to participate in the scan sessions, using one of two identical 3 Tesla Siemens Prisma MRI scanners with a 64-channel head coil (Head/Neck 64; Siemens Healthcare, Erlangen, Germany). Of the 41 participants who were scanned, 14 participants completed both scan sessions at the Massachusetts General Hospital Athinoula A. Martinos Center in Charlestown, MA, and 27 participants completed both scan sessions at the Boston University Cognitive Neuroimaging Center in Boston, MA.

One T1 anatomical scan (spatial resolution 1 mm isotropic, repetition time (TR) = 2530 ms, echo time (TE) = 1.69 ms, flip angle 7°) and two 6-minute, 40-second-long resting-state blood oxygenation level-dependent (BOLD) scans (2 mm isotropic, 488 frames, TR = 800 ms, TE = 37 ms, flip angle 52°) were collected and analyzed. All scans had minimal (< 0.67) average motion vector translation and were analyzed using FreeSurfer version 7.1.1⁹⁸.

A Priori Regions of Interest and Circuits

As described above, the primary functional connectivity analyses focused on two a priori circuits, the hippocampal-dlPFC and hippocampal-vmPFC pathways. Changes in the connectivity in the closely related amygdala-dlPFC and amygdala-vmPFC pathways and in 3 sensorimotor pathways were explored as well (see Supplementary Methods and Figure S2 for additional details).

The hippocampus and amygdala regions of interest (ROIs) were defined anatomically using the automated segmentation algorithm (“aseg”)⁹⁸ of FreeSurfer. Similarly, the vmPFC ROI was defined anatomically using the “medialorbitofrontal” region of the Desikan-Killiany atlas using FreeSurfer’s automated parcellation algorithm (“aparc”)⁹⁹. The dlPFC ROI was defined using a map of bilateral hippocampus connectivity that was generated in an independent resting-state fMRI dataset collected from a sample of college students (n = 122) with similar demographic and clinical characteristics^100–103; see Supplementary Methods for additional details.

Lastly, the functional connectivity of two sensory pathways, representing the functional coupling of primary visual and auditory cortices and their thalamic relay nuclei, plus one motor pathway, were examined. These were analyzed using 6 ROIs (3 cortical and 3 thalamic regions). V1 (BA 17), A1 (BA 41 and 42), M1 (BA 4), the lateral geniculate nucleus (LGN), medial geniculate nucleus (MGN), and the ventrolateral nucleus (VLN). V1, A1, and M1 were defined using the PALS Brodmann atlas¹⁰⁴, and LGN, MGN, and VLN were defined using FreeSurfer’s thalamic segmentation module¹⁰⁵.

Analysis

Behavioral Data: Changes Pre vs. Post RT and Correlations

All statistical analyses were conducted using R (version 4.2.2) and RStudio^106,107. Paired t-tests were conducted to compare scores of self-report measures before and after RT. Change scores were calculated by subtracting “pre” scores from “post” scores, and exploratory Pearson’s correlations were conducted to measure associations among these change scores.

Mediation Analyses

To test our prediction that acquisition of resilience-promoting skills during RT leads to reductions in transdiagnostic symptoms of psychopathology and that these associations are mediated by an RT-related reduction in emotion reactivity, mediation analyses were performed. Each mediation model utilized residualized change scores to examine the difference between pre- and post- RT scores for the measures of interest. Residualized change scores control for baseline differences and account for the regression to the mean that is assumed to occur in these measures across time¹⁰⁸. See Supplementary Methods for details about these analyses.

Resting-state fMRI Data Preprocessing

The whole-brain resting-state fMRI data were preprocessed using the Freesurfer Functional Analysis Stream (FS-FAST). Preprocessing steps included standard motion correction procedures, slice timing correction, and boundary-based registration to transform BOLD time-series data from native functional to anatomical volume and cortical surface space. Volumetric spatial smoothing was applied at twice the isotropic voxel resolution (FWHM = 4 mm).

Resting-state fMRI Data Analyses

Seed-based functional connectivity maps were created for the pre-RT and post-RT timepoints using bilateral dlPFC and vmPFC seeds (used in the primary analyses) and LGN, MGN, and VLN seeds (used in the control analyses). The first level univariate GLM analysis included nuisance regressors related to: head motion (6 parameters), mean global signal waveform, white-matter intensity (5 parameters), ventricular cerebral spinal fluid intensity (5 parameters), and scanner drift. Additionally, any time points identified as outliers were excluded (see data quality assurance procedures below), and the BOLD time series were subjected to high-pass (0.01 Hz) and low-pass (0.08 Hz) filtering. A whole-brain paired-difference analysis was performed by first computing difference maps (post-RT—pre-RT) for each individual in a common brain space, which were then subjected to a group-level analysis to identify regions where the mean paired-difference maps were significantly different from zero. The resulting significance map was thresholded (P < .05) and subjected to cluster correction using Monte-Carlo simulation (two-tailed; cluster-forming threshold of P < .05). Partial correlation coefficients were extracted from the hippocampus, amygdala, V1, A1, and M1 ROIs, and transformed to z-scores using a Fisher’s R-to-Z transformation. These z-scores were then used for t-tests (paired t-tests for the primary and control analyses and independent sample t-tests for scanner comparison analyses) and correlations.

Anatomical and Resting-state fMRI Data Quality Assurance

Anatomical data images and cortical surface reconstructions were assessed by eye for the presence of any major surface defects or poor gray/white matter boundary contrast. No major defects were identified for any subjects. Head motion was evaluated using the fslmotionoutliers tool and by examining the mean motion correction vector translation (MCVT) values. Any functional runs with 20% or more timepoints (TPs) identified as outliers would have been excluded; however, no runs for any subject met these criteria. Any functional runs with a mean MCVT greater than 1 mm (half the isotropic voxel dimension) would have been excluded; however, no runs for any subject met these criteria. Lastly, no significant pre vs. post differences were found for either head motion metric (TPs outliers: t(40) = −1.089, P = .283; MCVT: t(40) = 0.327, P = .745).

Results

Feasibility and Acceptability of Remotely Delivered RT

Of the 1457 college students who completed the screening survey, 791 were eligible for RT and 103 participants completed RT. Of these 103 participants, 34 (33%) met the PDI inclusion criterion only, 32 (31%) met the BDI inclusion criterion only, and 37 (36%) met both criteria. The typical reason for not participating in the study was a schedule conflict, since the RT groups were only offered at one or two times during the week at each college (determined by the RT facilitators’ schedules). See Figure S3 for the Consort diagram.

Attendance of the RT sessions was high across the 103 participants; 83 (81%) attended all 4 sessions, while the remaining 18 (17%) attended 3 out of 4 sessions, and just 2 (2%) attended 2 out of 4 sessions. Following completion of RT, the majority of the participants rated RT as beneficial (mean [SD] = 3.97 [.94]; 93% rated 3 or above) and reported that the concepts were useful (mean [SD] = 4.09 [.89]; 95% rated 3 or above) and novel (mean [SD] = 3.19 [1.07]; 74% rated 3 or above). Additionally, most participants said they were likely to recommend the intervention to a friend (mean [SD] = 4.02 [.92]; 84% rated 3 or above).

Regarding the remote delivery of the program, 26.2% reported that they “would not prefer in-person” participation, 32% were “unsure,” and 41.7% said they “would prefer in-person” participation (mean [SD] = 1.21 [.71]).

Self-report Data

Resilience-related Capacities

Levels of overall resilience (t(101) = 2.28, P = .025, d = .23), mindfulness (t(101) = 3.63, P < .001, d = .36) and self-compassion (t(101) = 4.33, P < .001, d = .43) were significantly higher following RT. As in previous RT trials, mentalization ability, as measured by the MentS scale, did not significantly change following RT (t(98) = .400, P = .690, d = .15).

Symptoms

Participants reported fewer total PEs (t(102) = 4.47, P < .001, d = .44), fewer paranoia-related PEs (t(102) = −4.31, P < .001, d = .43), less distress associated with PEs (t(101) = 5.57, P < .001, d = .42), and fewer symptoms of depression (t(102) = 6.68, P < .001, d = .66) and anxiety (t(101) = 4.16, P < .001, d = .41) following RT.

Emotion Reactivity and Perceived Stress

In addition, levels of emotion reactivity (t(101) = 3.29, P = .001, d = .32) and perceived stress (t(100) = 4.23, P < .001, d = .42) were lower following RT.

Effects 6 Months Later

Of the 103 enrolled in the study, 65 completed follow-up assessments online 6 months after participating in RT. All of the significant changes in symptoms and resilience-related capacities observed immediately following RT (compared to "pre"/baseline) were sustained at the 6-month follow-up time point (all P < .006, all d > .29), with the exception of the improvement in mindfulness which remained at a trend level (t(64) = 1.99, P = .051, d = .25). See Figure 1.

Figure 1.

Changes in symptoms, emotion reactivity, perceived stress, and resilience-related capacities following RT and 6 months later. Levels of psychotic experiences (PEs), distress related to PEs, paranoia-related PEs, depression, anxiety, emotion reactivity, and perceived stress were significantly lower, and levels of overall resilience, self-compassion, and mindfulness were significantly higher following completion of RT, compared to pre-RT/baseline (n = 103; all P < .025). These significant effects (differences from pre-RT/baseline) were sustained 6 months later in a subset of the original participants (n = 65), except for the improvement in mindfulness which remained a trend (P = .05). *P < .05; **P < .01; ***P < .001. Error bars represent standard errors of the mean. PEs, psychotic experiences; PE-distress, distress related to psychotic experiences; PE-paranoia, paranoia-related PEs.

Participants with PEs

When the above analyses were repeated excluding the 32 participants who were eligible for the trial due to having symptoms of depression only, all of the significant effects above remained significant (N = 71; all P < .009; all d > .28).

Mediation Model

Exploratory correlational analyses revealed findings that were consistent with those of prior trials of RT, ie, many of the above changes that followed RT were significantly correlated with each other (see Table S3 and Figure S4). We then tested our a priori mechanistic model of RT using mediation analyses, testing the prediction that the acquisition of resilience-promoting skills during RT decreases emotion reactivity, which then leads to decreases in transdiagnostic psychopathology (see Figure 2 and Supplementary Results).

Figure 2.

The reduction in emotion reactivity fully mediated associations between the acquisition of resilience-promoting skills and the reductions in symptoms, following RT. Mediation analyses (see text and Supplementary Materials for details) provided evidence for the main hypothesis (the model of the mechanisms of RT) of the study for 3 of the symptom outcomes: paranoia-related psychotic experiences (A), anxiety (B,C) and depression (D), showing that the significant associations between the improvements in self-compassion or mindfulness and the reductions in paranoia-related PEs, anxiety, and depression, following RT, were fully mediated by the reduction in emotion reactivity. PE, psychotic experiences.

First, we found that the associations between the significant improvements in self-compassion and the reductions in (1) paranoia-related PEs and (2) anxiety were each fully mediated by the reductions in emotion reactivity following RT (ab = −.227, 95% CI = [−.508, −.014], prop = .358; ab = −2.738 [−4.613, −1.309], prop = .50, respectively).

Second, the improvement in mindfulness following RT was significantly associated with the reductions in anxiety and depression; these two effects were also fully mediated by the significant reduction in emotion reactivity following RT (ab = −.078 [−.131, −.034], prop = .445; ab = −.037 [−.079, −.007], prop = .291, respectively).

In contrast, the associations between improvements in resilience-related capacities (mindfulness, self-compassion) and the reductions in the total number of PEs and PE-related distress were not mediated by the reduction in emotion reactivity.

Controlling for sex did not alter the models (all confidence intervals of the sex-controlled models overlapped those of the original analyses).

Functional Connectivity Analyses

The resting state functional connectivity analyses revealed that the magnitude of hippocampal-dlPFC connectivity significantly increased (t(40) = 2.81, P = .008, d = .44; due to a reduction in the anti-correlation between the levels of resting-state BOLD activity of the two regions) and hippocampal-vmPFC positive connectivity decreased (t(40) = 4.36, P < .001, d = .68), following RT (Figure 3). Similarly, amygdala-dlPFC connectivity significantly increased (t(40) = 2.41, P = .021, d = .38) and amygdala-vmPFC connectivity significantly decreased (t(40) = 2.85, P = .007, d = .45) following RT (Figure S5). These effects were not influenced by the scanner used to acquire the data (all P > .27).

Figure 3.

Changes in hippocampal-frontal functional connectivity following RT. Region-of-interest analyses of resting-state fMRI data showed that hippocampal-dorsolateral prefrontal cortex (dlPFC) functional connectivity was significantly greater, and hippocampal-ventromedial prefrontal cortex (vmPFC) functional connectivity was significantly lower, following RT, compared to pre-RT/baseline (n = 41). Plots of hippocampal-dlPFC connectivity at the two time points, “Pre-RT” and “Post-RT” (A) and a coronal view of the results of a voxelwise paired t-test analysis (comparing Pre-RT vs. Post-RT dlPFC connectivity; P < .05, uncorrected, for illustration purposes) (B) are displayed. Similarly, plots of hippocampal-vmPFC connectivity at the two time points (C) and a coronal view of results of a voxelwise paired t-test analysis (Pre-RT vs. Post-RT vmPFC connectivity; P < .05, uncorrected) are displayed. **P < .01; ***P < .001. Error bars represent standard errors of the mean. dlPFC, dorsolateral prefrontal cortex; vmPFC, ventromedial prefrontal cortex. In both A and C, the average partial correlation coefficient (pcc) is plotted on the y axis.

In contrast, control analyses revealed that the functional connectivity of two early sensory pathways and one motor pathway did not significantly change following RT (LGN-V1: t(40) = 1.15, P = .259; MGN-A1: t(40) = .58, P = .566; VLN-M1: t(40) = 1.70, P = .097).

Functional Connectivity Correlations

We then explored whether the above changes in functional connectivity were linked to the improvements in symptoms or resilience-related capacities following RT. Increases in hippocampal-dlPFC connectivity were significantly correlated with the decrease in emotion reactivity (r(39) = −.335, P = .032) following RT, but not with any other change in symptom or resilience-related capacity that followed RT (all P > .103). In addition, the increase in amygdala-dlPFC connectivity was correlated with the decrease in PEs following RT (r(39) = −.322, P = .040; Figure S6). There were no significant correlations between changes in symptoms and changes in amygdala-vmPFC or hippocampal-vmPFC connectivity (all r < .27; P > .05). Also, changes in resilience-related capacities (mindfulness, self-compassion, overall resilience) did not correlate with any changes in functional connectivity (all P > .103).

Discussion

Summary of Findings

The current study consisted of a mechanism-focused, remotely-delivered, open trial of RT in young people with some transdiagnostic risk for developing a serious mental illness, due to the presence of mild psychopathology (PEs and/or symptoms of depression). The overall goal of RT is to prevent the worsening of such subthreshold symptoms and associated poor outcomes, including the onset of a serious mental illness and the related disability. In the current trial of RT, levels of overall PEs, paranoia-related PEs, and PE-related distress, as well as depression and anxiety, were significantly lower, and measures of the acquisition of two of the RT skills, mindfulness and self-compassion, were significantly higher, following RT when compared to baseline (pre-RT), in line with the results of the two previous trials of RT^55,56. Moreover, these effects were sustained 6 months after the completion of RT. These findings suggest that remote delivery of RT has benefits that are similar to those observed with the in-person (original) version of RT.

The primary goal of this trial, however, was to investigate potential mechanisms of RT. First, we tested whether reductions in emotion reactivity mediated associations between the improvements in mindfulness and self-compassion and the reductions in transdiagnostic symptoms. This model was confirmed for 3 of the outcomes: paranoia-related PEs, anxiety and depression. In addition, in a subset of the participants, significant changes in hippocampal-frontal connectivity (as well as amygdala-frontal connectivity) followed RT, whereas connectivity of several control, sensorimotor circuits did not change. Lastly, the increase in hippocampal-dlPFC connectivity was significantly correlated with the reduction in emotion reactivity following RT.

Overall, the findings of this study are consistent with prior evidence demonstrating that acquiring mindfulness-related skills can reduce emotion reactivity and a wide range of symptoms of psychopathology. These data also provide preliminary evidence for the possibility that changes in frontotemporal circuitry may play a role in these effects. These results, taken together with those of the previous trials of Resilience Training^55,56 and other mindfulness-based^109,110 and compassion-based¹¹¹ interventions, suggest that even fairly brief behavioral interventions can improve the ability of vulnerable young people to manage stress and may prevent the worsening of psychopathology.

These findings also support prior evidence that symptoms on the dimensional psychosis spectrum (from mild PEs to clinical psychosis) and affective symptoms are closely linked, and that heightened emotion reactivity may be etiologically related to some PEs, as well as to depression and anxiety^28,29. Moreover, these findings provide additional evidence that the acquisition of putatively resilience-boosting skills leads to simultaneous reductions in these affective-psychotic spectrum symptoms, which could arrest or slow a process of clinical deterioration or progression to a full-blown disorder in young people who are at some risk for such progression.

However, we did not observe a mediating effect of the change in emotion reactivity following RT on the reductions in the total number of PEs or the distress associated with the total number of PEs. These negative findings suggest that the effects of RT on these two outcomes may occur via a different mechanism, which is less directly or strongly linked to the RT-related changes in emotion reactivity than the effects of RT on paranoia-related PEs, depression, and anxiety. In addition, as in previous trials of RT, we did not observe improvements in the third skill taught in RT, mentalization (despite the high level of participant interest in this skill). This consistent negative result may be related to the brevity of the exposure to mentalization during RT, or to limitations of the self-report scale used to assess the acquisition of this skill. Future studies of RT will investigate these effects further, using additional, objective tools to measure the intervention’s potential mechanisms and acquisition of the skills of RT.

Hippocampal-frontal Circuits as Potential Targets of Resilience-enhancing Interventions

It has been well-established that the hippocampus is involved in responses to stress in many mammalian species, including humans, mediated via the rich connections of the hippocampus with the hypothalamus and the high level of expression of glucocorticoid receptors in the hippocampus⁹⁰. One study conducted in healthy individuals found that the functional connectivity of the hippocampus with specific brain regions increases during stress induction^91,92; some of these connectivity increases (eg, with the hypothalamus) were associated with increases in arousal, whereas others (eg, with the dlPFC) were linked with decreases in arousal. These and related findings suggest that the increase in hippocampal-dlPFC connectivity observed following RT in the current study may have been related to an improved ability to modulate and decrease arousal responses to stress in the participants. The significant correlation between the change in hippocampal-dlPFC connectivity and the reduction in emotion reactivity following RT is consistent with this interpretation and with other evidence that the dlPFC is involved in emotional appraisals and other aspects of emotion regulation¹¹², and that mindfulness practices influence these processes¹¹³.

The hippocampus and dlPFC have only sparse connections with each other in primates¹¹⁴; thus the two regions may be functionally connected via indirect pathways between them involving the parahippocampal gyrus, presubiculum, and subiculum¹¹⁵. Functional connectivity studies conducted in humans have found that the dlPFC and hippocampus are components of distinct networks—a “task positive” (executive control/attention) and a “task negative” (default mode) networks, respectively—and thus typically show anti-correlated resting-state activity with each other¹¹⁶. In contrast, the hippocampus and vmPFC have robust direct connections in rodents and non-human primates⁷¹; consistent with this, in humans, these two regions are highly positively connected (i.e., with positively correlated activity) nodes of the default network¹¹⁷.

Thus, the reduction in hippocampus-vmPFC connectivity following RT observed in the current study is consistent with prior evidence indicating that mindfulness-focused interventions lead to decreases in activity and connectivity within the default network^118–120—a network which may be “hyper-connected” in individuals with mood and anxiety disorders^96,97,121. A similar reduction in the functional connectivity of the amygdala-vmPFC pathway following RT is consistent with the known reciprocal connections and interactions among the amygdala, hippocampus, and vmPFC during emotional processing and stress responses^122–124.

However, follow-up mechanistic studies of RT using a randomized controlled design are needed to clearly demonstrate a specific link between the beneficial psychological effects of RT and changes in the connectivity of the hippocampus with its network.

Transdiagnostic Prevention

RT was designed to reduce the risk for clinical worsening in young adults with vulnerability for developing mental illnesses based on the presence of mild-to-moderate subclinical symptoms of psychopathology (PEs and/or depression), which are associated with an increased risk for developing a range of poor outcomes^{1–3,7,8,68,125}. It is important to highlight that the participants of the current trial were generally healthy and functioning relatively well; they were all full-time college students and were not receiving any type of mental health-related treatment. Given the “prevention paradox”¹²⁶,¹²⁷—that the majority of cases of an illness occur in people at low or moderate risk, rather than those at high risk—focusing on such a population may prove to be particularly cost-effective, since reducing the prevalence of a common but weak risk factor has a larger effect on illness incidence than targeting a rare but strong risk factor.

Most previously tested transdiagnostic interventions designed for individuals with symptoms of anxiety and/or depression have excluded individuals with PEs¹²⁸, despite the established high levels of comorbidity of PEs and affective symptoms, and the evidence that PEs represent a marker of a more severe expression of affective illness, as well as greater risk for suicide and disability^{9,10,13,14,16}. More recently developed transdiagnostic intervention approaches have focused on a larger spectrum of psychopathology^{111,129–132}. For example, based on the transdiagnostic “staging” approach for mental health care¹²⁹, one recent randomized controlled trial of a compassion-focused, ecological momentary, hybrid intervention enrolled young adults at early, pre-illness stages (presenting with psychological distress or an at-risk mental state), as well as those at an early stage of a psychotic disorder, and observed beneficial effects of the intervention on momentary stress reactivity, resilience and quality of life¹¹¹. Also, several interventions targeting transdiagnostically at-risk adolescents have shown promising effects; for example, a randomized controlled trial of a two-session psychoeducational, cognitive behavioral intervention delivered in schools enrolled adolescents with “high risk” personality characteristics and found that the active intervention slowed the increase in general psychopathology over the 3 year follow-up period to a greater extent than a control health education intervention¹³³. Although these are early studies, it is possible that these types of broadly transdiagnostic approaches may have greater public health impact than interventions focused on the prevention of a narrow category of disorders¹²⁹. However, larger, well-controlled trials, with extended follow-up periods, are needed to fully validate this model.

Remote Delivery of Resilience Training

During this study, RT was delivered using videoconferencing (Zoom), primarily because the study began during the COVID-19 pandemic, and most colleges campuses were closed at that time, with classes occurring remotely while students lived at home. Given that the effects of the remotely delivered version of RT were essentially identical to those observed previously in the randomized, waitlist-controlled trial of in-person RT⁵⁶, the current findings suggest that remote delivery of RT is feasible and efficacious. However, 42% of participants in the current trial reported that they would have preferred the in-person version, suggesting that in-person RT should be used when feasible. Other types of remote delivery or hybrid approaches^111,134,135 may provide the benefits of both remote (eg, increased access) and in-person (eg, more social connection, support) intervention modalities.

Limitations and Future Directions

The findings of this study must be considered in light of its limitations, which include the absence of a control condition, the focus on college students (which limit the findings’ generalizability, given that more than half of U.S. adults do not attend college), and the reliance on self-report scales for measuring its outcomes. Despite the known potential pitfalls associated with self-report measures (eg, under-reporting, or a bias to report improvements), we weighed these limitations against the low participant burden associated with self-reported outcomes, which can be assessed using online questionnaires. The validity of the self-reported data in the current study is supported by several features of the results, ie, that the observed changes were in the expected directions, correlated with each other, and replicated the findings of two prior trials of RT. Also, the increase in hippocampal-dlPFC connectivity, a circuit known to be involved in emotion regulation-related processes, following RT correlated with a key finding that was measured via a self-report scale, the reduction in emotion reactivity following RT. However, future randomized controlled studies could measure, in addition to the outcomes assessed here, other aspects of emotion reactivity and responses to stress, using experience sampling or additional neuroimaging approaches, such as task-based fMRI paradigms measuring neural responses to aversive stimuli, which may be disrupted in young people with PEs¹³⁶.

Also, broadening the sample of participants, in future implementation studies of RT, to include young people who are not enrolled in college can shed light on whether RT is beneficial to those with a range of lifestyles, goals, and socioeconomic circumstances.

Taken together, the initial studies of RT provide further support for the testing and eventual dissemination of brief^137–140, evidence-based, behavioral interventions for transdiagnostically at-risk young people as one component of the larger effort to identify early intervention strategies that can reduce the incidence of disabling mental illnesses.

Supplementary material

Supplementary material is available at https://academic.oup.com/schizophreniabulletin.

Funding

This work was supported by funding from the Sidney R. Baer, Jr Foundation.

Conflicts of Interest

None declared.