Clinical and Electrophysiological Correlates of Hopelessness in the Context of Suicide Risk

Abstract

Hopelessness is a key risk factor for suicide. This analysis explored whether hopelessness indicates a recent suicide crisis state and is linked with magnetoencephalography (MEG) oscillatory power and effective connectivity differences. Change in hopelessness ratings and effective connectivity post-ketamine were also evaluated in a subsample of high-risk individuals to evaluate correlates of dynamic changes over time. Participants (66F;44M;1 transgender) included individuals with suicide crisis in the last two weeks (High Risk (HR), n=14), those with past suicide attempt but no recent suicide ideation (SI) (Low Risk (LR), n=37), clinical controls (CC, n=33), and healthy volunteers at minimal risk (MinR, n=27). MEG oscillatory power and clinical hopelessness ratings (via the Beck Hopelessness Scale (BHS)) were evaluated across groups. Dynamic casual modeling (DCM) evaluated connectivity within and between the anterior insula (AI) and anterior cingulate cortex (ACC). A subsample of HR individuals who received ketamine (n=10) were evaluated at Day 1 post-infusion. The HR group reported the highest levels of hopelessness, even when adjusting for SI. MEG results linked hopelessness with reduced activity across frequency bands in salience network regions, with no group or group-by-interaction effects. Using DCM, the HR group had reduced intrinsic drive from granular Layer IV stellate cells to superficial pyramidal cells in the ACC and AI. In the pilot HR study, reduced hopelessness was linked with increased drive for this same connection post-ketamine. Hopelessness is a possible proxy for suicide risk. Electrophysiological targets for hopelessness include widespread reductions in salience network activity, particularly in the ACC and AI.

Clinical Trials Identifier: NCT02543983

Introduction

Suicide continues to be a global crisis, with over 700,000 individuals dying by suicide each year (1). Better identification and treatments for suicidal thoughts and behaviors have been stymied by an absence of objective markers for the acute suicide risk state. Hopelessness, broadly defined as pessimism or a lack of optimism about the future, is a key suicide risk factor that has been linked to suicidal thoughts, attempts, and death (2–4), implicated in theoretical models for suicide (5–7), and included as a key diagnostic criterion in clinical models of the suicide crisis (8). This suggests that hopelessness is a promising proxy for suicide research, as it is associated with the crisis state, dynamic over time, and does not require the patient to report suicidal thoughts.

A burgeoning literature on the neurobiology of suicidal thoughts and behaviors has implicated the salience network, specifically the anterior cingulate cortex (ACC) and anterior insula (AI), in mediating the switch between the dorsal prefrontal cortex (DLPFC) and ventral PFC systems (9, 10); however, much less is known about the neurobiology of hopelessness in the context of suicide risk. An MRI study found a negative relationship between the severity of hopelessness and bilateral medial orbitofrontal cortex (mOFC) response in individuals with bipolar disorder, suggesting that hopelessness may be related to mOFC engagement deficits (11). In addition, an analysis of individuals with bipolar disorder found that suicide attempts were linked to reduced intrinsic connectivity distribution in the ventral medial PFC (vmPFC) and right AI, whereas hopelessness in these individuals was only linked with vmPFC intrinsic connectivity distribution (12). Furthermore, a study of intermittent accelerated theta burst stimulation (aiTBS) to treat individuals with treatment-resistant depression found that increased subgenual anterior cingulate cortex (sgACC)-mOFC functional connectivity during aiTBS significantly decreased hopelessness (13). Collectively, these studies implicate several brain areas (ACC, vmPFC, AI, and mOFC) linked to hopelessness and suicide risk that may be potentially responsive to treatment.

To our knowledge, little to no research has been conducted into the electrophysiological correlates of hopelessness in the context of suicide risk. In this context, the literature would benefit from such electrophysiological studies given the exceptionally high temporal resolution, direct measure of neuronal populations, and burgeoning evidence of the role of gamma power in suicide ideation (SI). For instance, a recent magnetoencephalography (MEG) study of patients with treatment-resistant depression found a negative relationship between SI and gamma power in the anterior insula (AI) that was distinct from other symptoms of depression (9). Other analyses have used fronto-insular/ACC alpha power to examine the link between gamma power in the AI and the amygdala and implicit association with suicide risk (14) as well as SI response to ketamine, a rapid-acting intervention for both depression and suicidal thoughts (15). In addition, gamma power is one of the most consistent biomarkers of response to ketamine and is hypothesized to represent a proxy of glutamatergic-excitation and GABAergic-inhibition balance (16). MEG also allows for the analysis of effective connectivity through techniques such as dynamic causal modeling (DCM), which uses biologically-plausible models of neural dynamics to examine the causal architecture of interactions between brain regions.

This analysis used a transdiagnostic sample of patients across the continuum of suicide risk to explore the clinical and electrophysiological correlates of hopelessness. The clinical aim was to evaluate the relationship between hopelessness and acute suicide risk, as defined by the occurrence of a recent suicide crisis. The hypothesis was that individuals with a recent suicide crisis would have higher levels of hopelessness than individuals with more distal suicide attempt, those with mood disorders but no history of suicide, and healthy volunteers, even when adjusting for current SI. The electrophysiological aims were: 1) to evaluate the link between hopelessness and MEG oscillatory power, and 2) to examine AI-ACC connectivity differences in the recent suicide crisis group compared to individuals at lower suicide risk, given the previously mentioned link between these regions and suicide risk. Lastly, as a proof-of-concept, a pilot study evaluated changes in hopelessness and effective connectivity after administration of ketamine to a small subgroup of individuals at highest suicide risk; this agent is known to have rapid-acting antidepressant and anti-SI effects (17). Such analyses can clarify the role of hopelessness as a potential clinical and electrophysiological proxy for suicide risk.

Methods and Materials

One hundred and eleven participants were enrolled in either a study investigating biomarkers of known suicide risk or the Neurobiology of Suicide Protocol (NCT02543983) at the Intramural Program of the National Institute of Mental Health in Bethesda, MD, USA. Individuals with mood and anxiety disorders were enrolled based on suicide risk rather than psychiatric diagnosis; participants were excluded if they had active psychosis or substance dependence. These participants were separated into four groups: high risk (HR, n=14; 8F, 5M, 1 transgender), low risk (LR, n=37; 26F, 11M), clinical controls (CC, n=33; 16F, 17M), and minimal risk (MinR, n=27; 16F, 11M) according to responses to the Columbia Suicide Severity Rating Scale (C-SSRS) (18). The HR group comprised patients with a recent suicide crisis, defined as a suicide attempt or suicidal thoughts with intent, within two weeks of study enrollment. The ethical and safety procedures for conducting neurobiological research in individuals at heightened risk for suicide have been described elsewhere (19). The LR group comprised patients with a suicide attempt more than a year previously but no suicidal thoughts with intent in the last year. CC participants were patients with mood disorders with no history of suicide attempt or ideation. Finally, the MinR group consisted of healthy volunteers with no history of suicidal thoughts or behavior or psychiatric diagnosis. Psychiatric diagnoses were confirmed with the Structured Clinical Interview for DSM Disorders (SCID). The study design is depicted in Figure 1. The study comprised two phases. The first was a cross-sectional clinical assessment and MEG scan across one to three days. Of the total sample, 86 participants completed one or two eyes-closed, eight-minute, resting-state MEG scans within one to three days of baseline clinical assessment (HR, n=11; LR, n=28; CC, n=25; MinR, n=22). The second phase was a substudy of HR participants who received an open-label trial of ketamine, described below. Demographic information for the participants appears in Supplemental Table S1.

Figure 1.

Study Design. The study included two phases. The first phase was for all participant groups and included clinical assessment and magnetoencephalography (MEG) scanning within one to three days. The second phase was for the High Risk (HR) group only. These individuals received a ketamine administration and clinical assessments just before and one day after ketamine infusion. The MEG was repeated one to two days after ketamine administration.

Clinical Measures

Hopelessness was assessed using the Beck Hopelessness Scale (BHS) (20). SI was measured via the Scale for Suicide Ideation (SSI) (21) and a weighted average of SI items (EFA-SI) identified by a previous exploratory factor analysis (EFA) (22); the EFA items included two suicide items from the Beck Depression Inventory (BDI) and one from the Montgomery-Asberg Depression Rating Scale (23). Depression was assessed with the MADRS (suicide item removed) and by a weighted average of the depressed mood items (Supplemental Table S2) identified in the aforementioned EFA. Negative cognitions and tension factors were measured by the weighted sums of their corresponding EFA items (Supplemental Table S2). Anhedonia was assessed using the Snaith-Hamilton Pleasure Scale (SHAPS) (24).

MEG – Acquisition and Preprocessing

MEG data were recorded from a subset of participants (n=86) at 1200 Hz with a bandwidth of 0.61–300 Hz using a SQUID-based CTF 275 whole-head system (CTF Systems Inc, Canada) housed in a magnetically-shielded room (Vacuumschmelze, Germany). Synthetic third order balancing was used for active noise cancellation. A T1-weighted MRI scan was also acquired using a 3 Tesla GE scanner (GE Signa, Milwaukee, WI), and the MEG data were co-registered to this structural scan for subsequent source localization.

Offline, MEG data were visually inspected, and any channels exhibiting excessive sensor noise were removed from subsequent analysis. Independent components analysis (ICA) was used to identify artifactual components in the recordings. For the ICA analysis, data were bandpass filtered from 1–50 Hz, and the Fast ICA algorithm was used to identify potential cardiac, ocular, and other artifacts within each dataset. These components were visually inspected and removed prior to source localization.

Ketamine Administration

A subset of the HR group also participated in a substudy of open-label ketamine and repeated biomarker collection. Ten of the 14 HR participants (4 M; 5 F; 1 transgender) received intravenous, subanesthetic (0.5 mg/kg over 40 min) ketamine up to five times over three weeks. Clinical assessments as described above were repeated at Day 1 after each ketamine administration, and resting-state MEG scans (n=7) were collected one to two days after the first ketamine administration; the latter timepoint is associated with ketamine’s largest anti-SI effects (25).

Statistical Analyses

Clinical Measures: Statistical Approach

Differences in current hopelessness at baseline between suicide risk groups were evaluated using a linear model, with a specific focus on differences between the HR and all other patient groups. Assumptions for linear models were assessed using qq-plots and plots of residuals versus fitted values. Seven secondary models each evaluated the relationship between group membership while controlling for one of the following SI risk measures: SSI total score, SI-EFA (as described above), depression ratings (MADRS and EFA-based depressed mood and negative cognition and tension scores), and anhedonia ratings (SHAPS), due to previous literature linking these symptom clusters to suicide risk (26). These were performed individually to determine the robustness of the risk group/hopelessness relationship relative to each specific covariate. Initial power analyses were conducted with the aim of recruiting 31 participants per group, to detect a moderate to large effect (d=0.6), 80% power with α=0.05, two-tailed. Recruitment of the HR group ended early due to the COVID-19 pandemic, and initial power estimates were therefore not met.

In the pilot sample of HR participants who received ketamine, differences between baseline and Day 1 hopelessness, SI, depression, and anhedonia were described using group means, standard deviations, effect sizes, and 95% confidence intervals (CIs).

MEG – Source Analysis and “Virtual Electrode” Construction

Following data preprocessing, MEG source-level power was projected in the theta (4–8 Hz), alpha (9–14 Hz), beta (15–29 Hz), gamma (30–58 Hz), and high gamma (62–118 Hz) frequencies using synthetic aperture magnetometry (SAM), a linearly-constrained minimum variance beamforming algorithm, over the entire rest period using a 5 mm voxel resolution and a semi-realistic head model (27). Linear mixed-effects models implemented in AFNI were used to test for effects of group, hopelessness, and group-by-hopelessness interactions on source-localized power within each frequency band (28). For this source analysis, the CC and LR groups were combined into one group because both had similar BHS scores. For this reason, the three groups used in this analysis were HR, LR (consisting of the LR and CC groups), and MinR. Additionally, BHS total scores were mean-centered based on suicide risk group, given that there were differences between groups on BHS score.

Two regions of interest (ROIs) within the salience network were identified from the whole-brain group analyses and from previous literature indicating their prominent role in both the pathophysiology of depression and risk for suicide: the ACC (Talairach=8, 18, 27) and right AI (Talairach=40, 10, −7). ‘Virtual electrode’ timeseries were calculated from these ROIs using wide-band beamformer weights (1–58 Hz) and the SAM algorithm over the entire rest period. These timeseries were imported into SPM12 (http://www.fil.ion.ucl.ac.uk/spm/) and epoched into two-second trial lengths. Subsequent connectivity analyses used these ROI timeseries.

MEG – Effective Connectivity

DCM, a modeling approach that uses biophysically-parameterized neural mass models to predict recorded electrophysiological timeseries, was used to examine effective connectivity between the ACC and AI using the conductance-based canonical microcircuit model with N-methyl-D-aspartate signaling (CMM_NMDA model) available in SPM12 (29). The model includes parameters mediating both fast (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)) and slow (NMDA) glutamatergic signaling dynamics between ROIs, as well as parameters governing within-region dynamics for superficial and deep pyramidal cell layers, granular Layer IV stellate cells, and inhibitory interneurons. This structure is outlined in Figure 2A. Three models varying feedforward and backward connectivity between the ROIs were constructed, and random effects Bayesian Model Selection (RFX BMS) was used to adjudicate between the differing model architectures. RFX BMS accounts for heterogeneity of model structure across subjects, which is important when considering pathophysiological mechanisms underlying psychiatric conditions. As shown in Figure 2B, Model 1 consisted of ACC to AI feedforward connectivity with reciprocal backward connectivity, Model 2 consisted of AI to ACC feedforward connectivity with reciprocal backward connectivity, and Model 3 consisted of ACC and AI lateral connectivity (i.e., full reciprocal feedforward and backward connectivity between both ROIs). Parametric empirical Bayesian (PEB) analysis was used to examine group differences in connectivity between the HR and combined CC, LR, and MinR groups (30). In the second-level design matrix, the first column represented the average effect over all participants (i.e., ones for all participants), while the second column tested for the difference between HR participants and the average effect (i.e., ones for HR participants, zeros for all other participants).

Figure 2.

A) Architecture of the canonical microcircuit model with N-methyl-D-aspartate signaling (CMM_NMDA model). Parametric Empirical Bayesian (PEB) analysis identified differences in intrinsic connections from spiny stellate cells to superficial pyramidal cells in both the anterior cingulate cortex (ACC) and anterior insula (AI) based on suicide risk group; the high-risk (HR) group had lower intrinsic connectivity in these regions compared to the average effect across all participants. B) Model architecture. C) Bayesian model selection identified Model 3 as providing the best fit for the data.

MEG—Exploratory Changes in Connectivity Following Ketamine

As an exploratory analysis of whether changes in hopelessness following ketamine administration in HR participants were related to changes in the intrinsic drive from granular input Layer IV spiny stellate cells to superficial pyramidal cells, Model 3 was fit to the post-ketamine scans in the subgroup of participants with resting-state data. PEB was again used to test for significant parameter changes following ketamine, with the first column testing for the average effect and the second column testing for the effect of ketamine (ones for ketamine sessions, zeros for baseline sessions). A third column was included with mean-centered BHS scores, before and after ketamine administration, for each participant and session. This third column, which examined the relationship between BHS and connectivity, was the covariate of interest for the pilot substudy. Subsequently, parameter estimates for the baseline and ketamine sessions were extracted and directly compared with change in BHS score from baseline to ketamine Day 1 in the HR group.

Results

Clinical Measures Across Suicide Risk Groups

Demographic information is presented in Supplemental Table S1. The HR group had higher hopelessness scores than the LR group (estimated difference=7.13 (SE=1.7), p=0.00006, Cohen’s d=1.23), CC group (estimated difference=5.2 (SE=1.7), p=0.003, Cohen’s d=0.95), and MinR group (estimated difference=14.4 (SE=1.8), p=<0.0001, Cohen’s d=6.50) (Figure 3A). When adjusting for anhedonia, negative cognition, and tension, all three HR group comparisons survived, and when adjusting for SI, the HR/LR and HR/MinR group differences survived. However, when adjusting for overall severity of depressive symptoms and the EFA scaled score for depressed mood in particular, no group differences survived (see Supplemental Table S3 for all results from all models).

Figure 3.

A) Differences in Beck Hopelessness Scale (BHS) total score across suicide risk group. B) Difference in BHS total score in the high-risk (HR) group pre- and post-ketamine administration (n=10). LR: low-risk group; CC: clinical controls; MinR: minimal risk group.

MEG - Source Analysis

A negative relationship was observed between BHS total score and power in various cortical regions including the right inferior frontal gyrus, right AI, right ACC, and right anterior temporal lobe in the theta (p_FDR<0.05), alpha (p_FDR<0.05), beta (p_FDR<0.05), gamma (p_FDR<0.05), and high gamma (p_FDR<0.05) frequency bands. There were no effects of group or group by hopelessness interactions (see Figure 4).

Figure 4.

Reduced power across various frequency bands in the right inferior frontal gyrus, right anterior insula, right anterior cingulate, and right anterior temporal lobe as hopelessness increased.

MEG - Effective Connectivity

Across all participants, Model 3 (i.e., ACC-AI lateral connectivity) best fit the data (Figure 2C). PEB analysis was used to test for random effects of model parameters. The average effect across all participants was modeled, as were differences between the HR group and the average effect. When evaluating DCM parameters, a posterior probability (Pp) of 0.9 or higher was used to identify parameters showing a significant effect. As shown in Table 1, significant differences were noted in intrinsic drive from granular input Layer IV stellate cells to superficial pyramidal cells within both the ACC (parameter estimate (Ep)= −.5862, Pp=1) and AI (Ep=−0.3954, Pp=0.9061) for the HR group compared to the average effect across all participants. This connection is highlighted in red in Figure 2A. Within both the ACC and the AI, the HR group had reduced excitatory drive for this connection compared to the average effect (ACC HR mean=0.803, ACC avg mean=0.8536, AI HR mean=0.884, AI Avg mean=0.9265) (Figure 5A).

Table 1.

Parameters mediating the average effect and difference in the high-risk group

Parameter	Parameter Estimate (Ep)	Posterior Probability (Pp)
Average Effects
Membrane capacitance of excitatory spiny stellate cells	0.3302	1
Membrane capacitance of superficial pyramidal cells	−0.9638	1
Membrane capacitance of inhibitory neurons	0.5001	1
Membrane capacitance of deep pyramidal cells	0.3592	1
Backward connection from the AI to the ACC	−0.7277	1
Backward connection from the ACC to the AI	−0.4888	1
High Risk Compared to Average Effects
Intrinsic connections within the ACC from excitatory spiny stellate cells to superficial pyramidal cells	−0.5862	1 *
Intrinsic connections within the AI from excitatory spiny stellate cells to superficial pyramidal cells	−0.3954	0.9061 *
Intrinsic gain on excitatory spiny stellate cells within the ACC	0.25	0.7656
Intrinsic gain on superficial pyramidal cells within the ACC	0.17	0.6326
Intrinsic gain on inhibitory interneuron within the ACC	−0.1917	0.6511
Intrinsic gain on deep pyramidal cells within the ACC	0.1265	0.5254
Membrane capacitance of excitatory spiny stellate cells	0.2003	0.5644

ACC: anterior cingulate cortex; AI: anterior insula

^*posterior probability greater than 0.9

Figure 5.

A) The high-risk (HR) group had lower granular Layer IV stellate to superficial pyramidal cell drive in both the anterior cingulate cortex (ACC) and anterior insula (AI) compared to the average effect (AE) across all participants. Ketamine increased this excitatory drive in the HR group (HR Post-Ketamine). B) No significant relationship was observed between change in Beck Hopelessness Scale (BHS) score and change in connectivity post vs. pre-ketamine within the ACC (R=0.074, p=0.81); however there was a negative relationship approaching significance within the AI (R=−0.41, p=0.16).

Exploratory Study: Ketamine’s Clinical and Neurophysiological Effects in a Subsample of HR Participants

The subsample of the HR group who received open-label ketamine had overall reductions in hopelessness (BHS: Cohen’s D: 1.13 95% CI: 0.22–2.15) (see Figure 3B) and anhedonia (SHAPS: Cohen’s D: 1.32, 95% CI: 0.41–2.34), although CIs for SI (SSI: Cohen’s D: 0.69: 95% CI: −0.12–1.55) and depression (MADRS SI item removed: Cohen’s D: 0.37: 95% CI: −0.34–1.1) crossed zero (Supplemental Table S4).

PEB was used to test whether ketamine altered the intrinsic drive from granular input Layer IV spiny stellate cells to superficial pyramidal cells, with mean-centered BHS total score included as the covariate of interest in the analysis. When evaluating DCM parameters, a posterior probability of 0.9 or higher was again used to identify significant effects. As shown in Figure 5A, increased excitatory drive from granular Layer IV stellate cells to superficial pyramidal cells within both the ACC (Ep=−0.0398, Pp=1) and AI (Ep=−0.0234, Pp=1) were associated with reduced hopelessness in the HR group before and after ketamine administration.

Finally, in order to directly examine whether change in spiny stellate-to-superficial pyramidal cell drive was associated with hopelessness ratings before and after ketamine for the HR participants, parameter change values were extracted and correlated with BHS change scores. Although not statistically significant, there was a negative linear trend between changes in connectivity (post-pre) and changes in BHS (post-pre) (R=−0.409, p=0.16) in the AI but not the ACC (R=−0.074, p=0.81) (Figure 5B).

Discussion

In a sample of individuals across the continuum of suicide risk, increased hopelessness was associated with recent suicidal crisis; differences between the highest and lower risk groups persisted even when adjusting for suicidal thoughts. Electrophysiologically, MEG results linked hopelessness with a wide swath of reduced activity across frequency bands in regions linked to the salience network. Specifically, there was a negative relationship between hopelessness and power in the right inferior frontal gyrus, right AI, right ACC, and right anterior temporal lobe. However, no group or group by hopelessness interaction effects were observed. Differences in effective connectivity were also noted between the HR group and the average effect across all participants. In particular, the HR group had reduced intrinsic drive from granular input Layer IV stellate cells to superficial pyramidal cells in both the ACC and AI.

Together, these results suggest that hopelessness may be a key indicator of recent suicide crisis, particularly in the absence of differences between the LR and CC groups. Whereas covarying for SI did not fully attenuate differences between suicide risk groups, adjusting for depression—specifically a depressed mood scale score that did not include any items assessing pessimism or hopelessness—negated differences by suicide risk groups. Therefore, in this context, hopelessness was most closely aligned with depressive symptoms, which supports theories asserting that hopeless depression is a critical risk factor for suicidal behavior (6). In addition, in a pilot subsample, open-label ketamine was linked to large reductions in hopelessness (Cohen’s D: 1.13), along with anhedonia (Cohen’s D: 1.32), SI (Cohen’s D: 0.69), and depression (Cohen’s D: 0.37), suggesting that hopelessness can have a dynamic response to intervention. Though preliminary, these results support the strategy of studying hopelessness as a specific construct within the larger scope of suicide risk, which is multifactorial and contains many potential subtypes (31). Further evaluation of the relationship between hopelessness and other suicide risk factors is indicated, as it is likely that hopelessness functions in tandem with other related constructs such as depressed mood and psychological pain in the development of suicide risk (6, 7).

The results also indicate that hopelessness in the context of suicide risk is associated with reduced power in brain regions comprising the salience network, in line with previous findings linking salience network dysregulation with SI (9) and supporting the theoretical framework that the salience network is crucial in the transition from suicidal thoughts to actions (10). Our source-level analysis identified no significant differences in power between the HR group and the other groups, which indicates that the relationship between hopelessness and electrophysiology did not change by suicide risk group; this finding should also be interpreted in the context of a low sample size for the HR group. However, our connectivity findings demonstrated that HR participants, who had the highest levels of hopelessness at baseline, had significantly reduced intrinsic drive from granular Layer IV stellate cells to superficial pyramidal cells in both the ACC and AI, key nodes of the salience network, which complemented the source analysis results and permitted more focused considerations of group differences. Evidence suggests that glutamatergic drive from stellate cells to superficial pyramidal cells gate the lateral spread of activation in superficial cortical layers (32). Because superficial pyramidal cells are thought to be the primary generators of the MEG signal (33), these findings suggest that the negative relationship between wide-band power and hopelessness might be partially explained by reduced glutamatergic drive from spiny stellate cells to superficial pyramidal cells in both the ACC and AI for the HR participants. Interestingly, ketamine administration increased spiny stellate to superficial pyramidal cell drive in our subsample of HR participants with pre- and post-ketamine scans, suggesting that ketamine might normalize intrinsic excitatory connectivity—and thus increase cortical excitability—in patients at high risk for suicide. Given the observed trend toward a relationship between change in BHS scores and spiny stellate-to-superficial pyramidal cell drive in the insula pre-to-post ketamine administration, the results further underscore the importance of the salience network more broadly, and the insula in particular, in suicide risk. Because the insula has been implicated in appraisal of negative emotional stimuli (34) and interoception (35), which are also dysregulated in suicide risk (36, 37), further work on the role of the insula in recovery from the suicide crisis is indicated. In particular, it will be critical to evaluate intrinsic connectivity in the AI as a potential mediator of improvements in hopelessness and suicidal thoughts in response to ketamine or other rapid-acting interventions.

Hopelessness, particularly as assessed by the BHS, is defined as negative predictions about the future, for instance, “My future seems dark to me.” Researchers have further expanded conceptualizations of hopelessness by using cognitive and task-based assessments to evaluate episodic future thinking more broadly. For example, individuals at risk for suicide appear to have difficulty generating possible positive future experiences, suggesting a potential fluency deficit for imagining future events (38). One longitudinal follow-up study of individuals hospitalized for self-harm found that decreased positive future thinking fluency was associated with risk for repeated self-harm, even when adjusting for more traditional hopelessness scores (39). In-depth analyses of the types of future events imagined by at-risk patients found that individuals with suicidal thoughts imagine future events as having a longer duration, suggesting potentially less specificity in thinking about the future (40). Intriguingly, an MRI study that used similar paradigms of future episodic thinking found that the salience network—specifically insula activity—was associated with increased complexity of future thinking in a healthy volunteer sample (41). Therefore, future directions for this work include conducting similar tasks in individuals at risk for suicide to identify potential underlying neural deficits in future thinking.

Limitations of the analysis include, first, the small sample size, particularly of the HR group, due to the difficulties associated with recruiting participants into neurobiological research just after a suicide attempt and the COVID-19 pandemic. Larger sample sizes would also be needed for more in-depth evaluation of the impact of variables such as sex, which is critically important for suicide risk (42). Second, the clinical assessments were conducted within days of the MEG scan, but not necessarily in real time; because ecological momentary assessment research is revealing fluctuations in clinical symptoms, particularly SI (43), future studies should assess hopelessness just before the time of scanning. Third, the clinical assessments were linked with resting-state scans rather than task-based data interrogating hopelessness. Useful paradigms exist that probe episodic future thinking and could be used in future research (44), but to our knowledge, no such studies have yet been conducted with MEG. Lastly, the subsample of HR participants who received ketamine points to intriguing suggestions of hopelessness as a biomarker of ketamine response, but this sample was very underpowered and without a placebo control; thus, results should be interpreted with caution. The study is also associated with several notable strengths. First, the study included a transdiagnostic approach to suicide research in which biological and clinical assessments were conducted more proximately in time to the suicide crisis. Second, control groups at differing levels of suicide risk were included as comparisons. Finally, a rapid-acting intervention (ketamine) was used to probe the utility of the biomarker under study.

Conclusions

To our knowledge, this study is the first to use MEG to investigate hopelessness in the context of suicide risk. The findings underscore that hopelessness may represent a key proxy for suicide risk specifically because it is associated with the suicide crisis, does not require the direct reporting of suicidal thoughts, and responds to rapid-acting interventions. Our results also point to the role of the salience network, specifically the ACC and AI, in suicide risk and hopelessness, and our pilot analysis in a subsample of HR patients suggest that ketamine may reduce hopelessness and increase effectivity connectivity in these regions, suggesting potential mediators for ketamine’s anti-suicidal effects. Further research should seek to evaluate whether direct interventions that target hopelessness and its electrophysiological correlates can reduce suicide risk.