Attentional Control May Be Modifiable with Mindfulness-Based Cognitive Therapy to Prevent Suicide

Abstract

Objective:

To test whether Mindfulness-Based Cognitive Therapy to Prevent Suicide (MBCT-S) is associated with improvement in attentional control, an objective marker of suicide attempt.

Method:

In the context of a randomized clinical trial targeting suicide risk in Veterans, computerized Stroop and emotion Stroop (E-Stroop) tasks were administered 3 times over 6-months follow-up to 135 high suicide risk Veterans. Seventy were randomized to receive MBCT-S in addition to enhanced treatment as usual (eTAU), and 65 were randomized to eTAU only. E-Stroop word types included positively- and negatively-valenced emotion, suicide, and combat-related words. Interference scores and mixed effects linear regression analyses were used.

Results:

Veterans receiving MBCT-S showed a more favorable trajectory of attentional control over time, as indicated by performance on two E-Stroop tasks. Combat-stress interference scores improved over time among Veterans in MBCT-S. Interference processing time for negative affective words deteriorated over time among Veterans receiving eTAU only.

Conclusions:

MBCT-S may effectively target attentional control, and in particular reduce processing time during affective interference, in high suicide risk Veterans. Future studies to replicate these findings are warranted.

Introduction

A suicide death occurs every 11 minutes in the U.S. (Centers for Disease Control and Prevention, 2020). A key research priority towards reducing the suicide rate is identifying objective, modifiable markers of suicide attempt (SA) to target in treatment (National Action Alliance for Suicide Prevention Research Prioritization Task Force, 2014). These efforts have consistently pointed to attentional dyscontrol as an objective marker of SA risk (Mann et al., 2009).

Mindfulness-Based Cognitive Therapy (MBCT) (Segal, Williams, & Teasdale, 2013; Segal, Williams, & Teasdale, 2002) is a candidate intervention to improve this objective marker of SA risk. MBCT was designed to target attentional dyscontrol and prevent relapse to Major Depressive Disorder (Teasdale, 1988). Individuals with attentional control difficulties often struggle to direct their attention away from distressing cues, which can result in patterns, such as rumination, that decrease the use of instrumental behaviors and problem solving and ultimately lead to depressed mood (e.g., Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008; Williams et al., 2007).

In non-suicidal cohorts, attentional control is modifiable using mindfulness-based intervention (i.e., a short-term, secular treatment approach, such as MBCT, that targets mindfulness and ultimately mental health using extended meditation practices). Specifically, MBCT has been shown to improve attentional control in patients with depression, post-traumatic stress disorder (PTSD), and mild traumatic brain injury (Britton et al., 2018; Holas, Krejtz, Wisiecka, Rusanowska, & Nezlek, 2020; Jermann et al., 2013; Shahar, Britton, Sbarra, Figueredo, & Bootzin, 2010; van der Velden et al., 2015, for a review; Verhoeven, Vrijsen, van Oostrom, Speckens, & Rinck, 2014). In addition, for Veterans, as well as active duty service members readying for deployment, mindfulness-based interventions can improve attentional control (Cole et al., 2015; Jha et al., 2015; Meland et al., 2015).

Given that attentional control can be modified using mindfulness-based intervention in non-suicidal populations, and that attentional dyscontrol is an objective marker of SA, a promising line of research is testing whether MBCT improves attentional dyscontrol in suicidal populations. In an uncontrolled pilot study, depressed outpatients at high risk for suicide showed improved attentional control after receiving MBCT for suicide prevention (MBCT-S) (Chesin et al., 2015), a treatment combining MBCT and Safety Planning (Stanley & Brown, 2008). The same pilot study also showed improvements in several measures related to cognition and depression, including rumination and cognitive reactivity. While providing important preliminary evidence that MBCT-S may improve attentional control among high suicide risk patients, replication of these MBCT-S effects in comparison to a control group is needed.

The current study evaluated whether MBCT-S, relative to a control condition, improved attentional control in Veterans at high risk of suicide. The modifiability of attentional control in the face of emotional provocation was also examined. Whereas most studies linking attentional dyscontrol to SA have used attentional tasks with semantic interference (i.e., a color-word Stroop task involving relative slowing when naming the font color of words printed in incongruent ink, such as “red” in blue ink) (e.g., Keilp et al., 2013), other research suggests similar associations between SA risk and attentional dyscontrol to emotional stimuli (i.e., relative slowing when naming the font color of an emotionally-valenced word, such as “suicide,” compared to a neutral word, on an emotion Stroop (E-Stroop) task) (Wilson et al., 2019, for a review). Targeting this form of attentional dyscontrol was a central rationale in the development of MBCT, which aimed to reduce attentional dyscontrol in the context of negative emotions (Teasdale, 1999).

In the current study, both types of attentional control were examined. The current study is the first to our knowledge to test changes to attentional control with MBCT-S, or MBCT, in high suicide risk patients in a randomized clinical trial. Based on findings from prior work showing changes to attentional control with MBCT (e.g., Holas et al., 2020; van der Velden et al., 2015, for a review), MBCT-S in non-Veteran patients (Chesin et al., 2016), as well as with mindfulness-based intervention in Veterans (Cole et al., 2015), we hypothesized that MBCT-S would significantly improve the following outcomes over 6 months, relative to a control condition:

1. attentional control, as indicated by performance on the Stroop;

2. attentional control under negative affective provocation, as indicated by performance on the E-Stroop;

3. attentional control under positive affective provocation.

Since there are no available studies testing improvements to attentional dyscontrol to suicide-specific or combat-related stimuli with mindfulness-based intervention, no a priori hypotheses regarding the effect of MBCT-S on these types of attentional control were made.

Methods

Sample and Procedures

Participants were 135 Veterans who were recruited from 2 Veterans Health Administration (VHA) medical centers and determined to be at high risk of suicide, meaning they had suicide ideation with intent, resulting in hospitalization or suicide behavior in the past 30 days and a designation of at high-risk for suicide by the VHA Suicide Prevention Coordinator (SPC) or an actual, aborted or interrupted suicide attempt in the 12 months prior to study entry. Most were recruited during an inpatient admission. Exclusion criteria were few and limited to cognitive impairment, psychotic symptoms, behavioral dysregulation, medical instability, or past-year receipt of services that included mindfulness. Participants were randomized to treatment-as-usual enhanced for suicide prevention (eTAU) or MBCT-S in addition to eTAU. In the VHA, eTAU includes Safety Planning (Stanley & Brown, 2008), greater follow-up to monitor and address suicide risk, and a full array of mental health services. MBCT-S consisted of 1–2 individual sessions focused on elaborating the eTAU safety plan, 8 weekly group sessions focused on mindfulness practice, as well as optional monthly group booster sessions. Detailed information about the MBCT-S program, trial procedures and participants, as well as treatment effectiveness are provided in our earlier publications (Interian et al., 2021; Kline et al., 2016). While the primary aims of the trial focused on differences in suicide events, SAs, and psychiatric hospitalizations between groups, the current study focused on attentional control outcomes.

Participants completed assessments that included computerized neurocognitive tasks measuring attentional control at baseline, 3 months (corresponding to MBCT-S treatment end), and 6 months. Computerized neurocognitive tasks took approximately 20 to 25 minutes to complete. Prior to testing, a short practice was provided for each task. The inter-trial interval was 50 milliseconds, and stimuli presentation was pseudorandom, that is, word types were comingled but the same presentation order was used for all participants, in order to compare directly among individual participants. All study staff were trained on neurocognitive task administration and were blind to participant condition. This study was approved by the local Institutional Review Board. All participants gave written informed consent prior to participating.

Neurocognitive Measures of Attentional Control

A computerized Stroop Task (Stroop; Stroop, 1935; Keilp, Sackeim, & Mann, 2005) was used to measure attentional control. For each participant, two median reaction times across trials were calculated: One for congruently-colored words (e.g., the word “blue” printed in blue ink) and one for incongruently-colored color words (e.g., the word “blue” printed in red ink). The percentage increase in median reaction time between congruently-and incongruently-colored words, the interference effect, was the primary outcome measure of attentional control from this task. This scoring method is consistent with that used in previous research establishing a link between SA and attentional control (e.g., Keilp, Gorlyn, Oquendo, Burke, & Mann, 2008). Higher scores indicate greater interference from incongruent stimuli and thus poorer attentional control. Fifty-six incongruent trials and 52 congruent trials were presented. Internal consistency of reaction times within each stimulus type ranged from .84 to .96 for congruent words and .62 to .93 for incongruently-colored color words across the 3 time points, respectively.

A computerized emotion Stroop Task (E-Stroop; Moore et al., 2019) was used to measure attentional control in the context of four types of affective stimuli: positively- and negatively-valenced emotion, suicide, and combat-related words, e.g., “ecstatic,” “stressful,” “suicide,” and “explode,” respectively, as well as neutral words (e.g., “chalk”). Stimuli were selected from studies testing associations between behavioral measures and suicide behavior (e.g., Moore et al., 2019) and on the basis of general relevance to the phenomena of interest. Words across types were similar in terms of length and number of syllabi. E-Stroop stimuli by word type are listed in Supplemental Table 1. Emotion words were presented in one block, and there were nine trials of each word type. Attentional control in the context of affective provocation was measured with interference effects, which were calculated as the percentage increase in median reaction times to name ink color between emotionally-neutral and emotion word trials. These outcomes were labeled the positive affective interference effect, the negative affective interference effect, the suicide interference effect, and the combat-stress interference effect in alignment with the type of emotion word presented. Higher scores indicate greater interference from the type of emotional stimuli and thus greater attentional dyscontrol in the face of the particular type of affective provocation. The exception is interference from positive emotional stimuli, where higher scores indicate greater relative attention to positive stimuli and arguably resilience or mental well-being (Newman, Quigley, Fernandez, Dobson, & Sears, 2019). Word types showed internal consistencies ranging from .61 to .82 (negative), .72 to .76 (combat), .77 to .80 (suicide), .69 to .77 (positive), and .61 to .72 (neutral), across the 3 time points, respectively.

Statistical Analysis

Analyses were conducted using SAS software, Version 9.4 (SAS Institute Inc., Cary, NC, USA, www.sas.com). To test whether attentional control, as indicated by the 5 interference scores, improved more with MBCT-S than eTAU alone over 6-months follow-up, a series of 5 mixed-effects linear regression analyses were run. For each model, the attentional control measure was the dependent variable and the repeated factor was time of assessment. Because of baseline differences in negative affective and combat-stress interference effects, the main effect for treatment was not included in these 2 models. This is consistent with recommendations to adjust for baseline differences when estimating treatment effects over time (Twisk et al., 2018). The group by time interaction effect was the effect of interest to test whether changes in attentional control over time differed between groups. The analyses were otherwise unadjusted given that there were no significant baseline differences between groups on the prevalence rates of Major Depressive Disorder, PTSD, psychopharmacological treatment, past-year or lifetime SA (Interian et al., 2021). Further, given that this was an exploratory study, adjustments were not made for multiple comparisons, although significance levels with Bonferroni adjustments are provided to allow review in relation to type I error. To minimize the effect of extreme values, scores on each attentional control measure were bounded at the 5^th and 95^th percentile of the scores for each time point in each group using the %winsorization macro (Moussawi, 2009). Intercorrelations were modelled with random effects for each subject nested within the 2 study sites. Maximum Likelihood estimation procedures were used. The covariance structure that produced the lower Akaike’s Information Criteria (AIC) value between a compound symmetry and unstructured covariance structure was used (Pan, 2001). To confirm multivariate normality, the Q-Q plots of each multivariate analysis were visually inspected.

Results

Table 1 shows that the majority of participants had a previous SA (84%), with over half having multiple SAs (56%). Most presented with current depression (86%) and PTSD (61%).

Table 1.

Participant Characteristics

Demographic	N	%a
Age, mean (SD)	46.5	13.8
Male Sex	118	87.4
Some College Education	86	64.7
Clinical
MDD	116	85.9
PTSD	82	60.7
Past Year Severe Suicide Ideationb	120	88.9
Lifetime SA
0	21	15.6
1	38	28.2
2 or more	76	56.3
Past Year SA
0	52	38.5
1	57	42.2
2 or more	26	19.3

Note. There were no differences between treatment groups on these characteristics. Suicide behaviors are not mutually exclusive.

^aSome cell denominators <135 due to missing data.

^bSevere suicide ideation defined as suicide ideation involving suicide intent, as measured by the Columbia Suicide Severity Rating Scale (C-SSRS).

Supplemental Table 2 summarizes the performance of participants on the tasks that were administered, i.e., reaction times by stimulus type, assessment point and condition. Changes in attentional control across 6-months follow-up are summarized by treatment group in Table 2. For 2 types of attentional control, the effect of interest, i.e., the time by treatment group interaction effect, was significant. Combat-stress and negative affective interference effects significantly differed over time between groups (F [3, 161]=3.07, p =.03; F [3,133]=2.93, p =.04), respectively). Follow-up tests showed that with combat-stress interference, MBCT-S participants showed a significant decrease (i.e., improvement) over 6 months (F [2,161]=4.55, p =.01). In contrast, eTAU participants did not significantly change over time. Negative affective interference processing time significantly increased (i.e., worsened) among eTAU participants (F [2,133]=3.27, p =.04) . MBCT-S participants’ negative affective interference scores did not significantly change over time. Figures 1 and 2 provide a graph of these results. Differences in changes over time by group were not observed for the interference, positive affective and suicide interference effects (Table 2). There was, however, a strong trend for the interference effect score to improve over time across participants regardless of condition (F [2,169]=2.95, p =.06).

Table 2.

Attentional Control Over Time by Treatment Group

		Baseline			3-Months Follow-Up			6-Months Follow-Up
		N	M	SD	N	M	SD	N	M	SD	p-value (time*cond; timea)
Interference Effect											0.19
MBCT-S		70	18.7	11.3	43	16.0	9.8	44	13.7	11.1
Control		65	19.5	11.9	44	16.7	10.6	43	19.6	13.8
Negative Affective Interference Effect											0.04b
MBCT-S		67	5.9	12.5	43	5.0	14.5	43	1.2	15.0	0.27
Control		60	0.3	16.6	44	6.0	20.1	43	6.3	20.0	0.04
p-value (cond)a			0.03			0.70			0.21
Positive Affective Interference Effect											0.56
MBCT-S		67	9.1	17.0	43	7.7	18.7	43	9.9	17.9
Control		60	5.6	19.1	44	5.8	19.2	43	11.7	18.3
Suicide Interference Effect											0.62
MBCT-S		67	-2.4	12.8	43	-3.3	12.7	43	-4.6	14.4
Control		60	-6.6	15.8	44	-4.2	15.1	43	-6.8	11.1
Combat-Stress Interference Effect											0.03b
MBCT-S		67	16.1	20.2	43	10.8	15.0	43	6.0	16.6	0.01
Control		60	7.5	19.3	44	9.9	20.2	43	10.5	14.1	0.63
p-value (cond)			0.01			0.88			0.30

Note: Means and standard deviations are percentages. Negative scores represent a facilitation effect whereby responses to suicide-related words were faster than responses to emotionally-neutral words.

^aSimple effects reported only for models where time

^*condition interaction term was signficant.

^bThe Bonferroni-corrected p-value is .0125. This p-value does not meet this threshold. There is an increased risk of Type 1 error given the multiple E-stroop tests.

Figure 1.

Effect of MBCT-S and eTAU on Combat-Stress Interference Scores

Note: Modelled means shown, and error bars represent standard errors.

Figure 2.

Effect of MBCT-S and eTAU on Negative Affective Interference Scores

Note: Modelled means shown, and error bars represent standard errors.

Discussion

In this randomized, prospective, assessor-blinded clinical trial with high suicide risk Veterans, we found significant differences between conditions in two attentional control tasks over time, both of which pointed to an advantage for MBCT-S for either improving or protecting against worsening of attentional control in the face of affective provocation. Specifically, improvement in attentional control in the context of combat-related stimuli over 6-month follow-up was observed among participants receiving MBCT-S. ETAU participants evidenced significant deterioration in negative affective interference processing over 6-month follow-up, a pattern that was not seen among MBCT-S participants. Confidence in these results, and thus the benefit of MBCT-S to attentional control, is buoyed by methodological aspects of the current study. We measured attentional control using objective, computerized tasks. Also, these observations were demonstrated in relation to a control group that had high rates of mental health treatment access, including high rates of residential treatment (Interian et al., 2021). Therefore, results are less likely attributable to general treatment effects and more likely specific to MBCT-S. Taken together with an understanding of attentional dyscontrol as an endophenotype for SA (Mann et al., 2009), the current study shows that MBCT-S may improve this objective marker of SA risk. However, replication studies are needed to confirm these results, given the lack of significant findings when adjusting for multiple comparisons.

Only a few cognitive training programs (e.g., attention bias modification or attention control training) have been tested on combat-stress interference processing (Kangaslampi & Peltonen, 2019). These training programs improved attentional control in the context of combat-related stimuli among Veterans with PTSD (Badura-Brack et al., 2015; Khanna et al., 2016; Kuckertz et al., 2014). Prior to this study, calls were made to further examine mindfulness-based interventions for PTSD. This was in part due to the role of attentional dyscontrol in maintaining PTSD, along with an understanding of the positive impact of mindfulness-based intervention on attentional control (Boyd, Lanius, & McKinnon, 2018, for a review). Findings from this study thus extend available information and show MBCT-S may reduce attentional dyscontrol to combat-related stimuli in Veterans who are at high risk for suicide and, for the most part, had co-morbid PTSD.

Somewhat consistent with our second hypothesis, that MBCT-S would improve negative affective interference, we found significant differences between treatment groups over time on this outcome. Consistent with a protective effect, our findings were driven by a significant increase in interference processing time for negative affective stimuli among eTAU participants, but not those receiving MBCT-S. Importantly, differences in attentional dyscontrol to negative affective stimuli between conditions were likely not due to a relative worsening of depression among eTAU participants. Rather, participants in both conditions realized improvements in depression that were sustained over 6 months follow-up in the current study (Interian et al., 2021). Whereas previous work has demonstrated that mindfulness-based interventions improve negative affective interference effects (De Raedt et al., 2012; Holas et al., 2020; Ortner, Kilner, & Zelazo, 2007), current results instead show a prophylactic effect for MBCT-S against the worsening of this type of attentional control. MBCT treatment developers proposed that attentional control in the face of negative affective stimuli was the mechanism by which MBCT prevented relapse to depression (Segal et al., 2013). Thus, our findings also offer preliminary insight into the applicability of the theory of MBCT to the area of suicide prevention.

Inconsistent with our first hypothesis, significant attentional control improvements, as indicated by standard Stroop performance, were not observed with MBCT-S. This is despite the fact that interference effect scores declined meaningfully in the MBCT-S group, and by 6 months follow-up, were equivalent to those seen in depressed patients without a history of SA, which in turn are lower than those of suicide attempters (Keilp et al., 2013). The lack of MBCT-S advantage may be due to a trend for interference processing to improve over time across both conditions (p=.06). As noted earlier, engagement in a range of treatments available as part of VHA eTAU was high (citation removed for blind review). It may be that this form of attentional control improves generally with treatment and is not specific to MBCT-S. Such a conclusion is consistent with evidence that attentional control improves with antidepressant medication in high suicide risk patients (Gorlyn et al., 2015).

Inconsistent with our third hypothesis, changes to attentional control under positive affective provocation did not differ between treatment groups. This finding is inconsistent with prior studies where improved positive affective interference effects were shown in previously depressed patients receiving MBCT (De Raedt et al., 2012; Holas et al., 2020). Neither of these prior studies, however, had an active control condition, which could explain the difference in findings. In the current study, neither those receiving MBCT-S nor eTAU spent relatively more time processing positive affective stimuli over time.

Suicide-interference effects did not improve with MBCT-S. This finding aligns with another study showing no improvement to this type of attentional control with treatment in patients and community members with suicide ideation (Cha et al., 2017). While there were early indications that the suicide interference effect predicted near-term SA (e.g. Cha, Najmi, Park, Finn, & Nock, 2010), the role of attentional dyscontrol to suicide-specific stimuli in suicide behavior risk is increasingly questioned (Wilson et al., 2019). Specific to this sample of high suicide risk Veterans, we similarly failed to show an association between the suicide interference effect and SA at baseline (Interian et al., 2020), despite later showing some promising effects for MBCT-S for reducing suicide behavior in this sample (Interian et al., 2021). Thus, the import of the suicide interference effect as a marker for SA risk and ultimately a target for treatment remains unclear, and more research is needed.

Limitations

Despite a randomization strategy that proved effective in preventing baseline group differences across all other clinical, suicide history and demographic variables (Interian et al., 2021), MBCT-S participants displayed poorer attentional control on two measures at baseline (combat-stress and negative affective interference effects). Although an analytic strategy to mitigate concerns about estimation accuracy in light of these baseline group differences was used (i.e., the treatment variable was not included as a conditional effect in the repeated measures analysis) (Twisk et al., 2018), replicating these results in a future study where these types of attentional control are equivalent between groups at baseline would increase confidence in the findings. Internal consistency for reaction times was borderline for certain word types at some time points. This may reflect that emotion word trials were few, i.e., n=9 trials by emotion word type at each time point. Future research may use a greater number of trials per word type or different measures of attentional control, such as accuracy measures or adaptive tasks, to overcome concerns with respect to the reliability of reaction time, and thus derived interference score, measures (Draheim, Mashburn, Martin, & Engle, 2019). Also, multiple tests were conducted and p-values were unadjusted, which increases the likelihood of a type 1 error. In fact, a p-value less than or equal to .013 would have been needed to maintain the familywise error rate at .05 given the 4 E-Stroop tests. Therefore, these results require replication in future studies with a confirmatory analytic approach. Specifically, a study focused on testing only the effects shown here, i.e., limited to testing the prophylactic effect of MBCT-S on negative affective interference scores and improvements in combat-related interference with MBCT-S, or a study with greater power given more participants, which could support adjustments for multiple outcomes, is indicated. Finally, attentional dyscontrol in the face of combat-related stimuli may be specific to PTSD (Vyas, Murphy, & Greenberg, 2020, for a review), and the slight majority of our sample had PTSD. We conducted a post-hoc analysis to test whether the association between MBCT-S and improvement in combat-stress interference over time was specific to those with baseline PTSD. The three-way interaction term was not significant, suggesting the effect of MBCT-S on the combat-stress interference effect did not differ by PTSD status. Nonetheless, the generalizability of improvements to combat-stress related interference processing with MBCT-S to high suicide risk populations without PTSD requires further testing.

Conclusions

In sum, improvement in attentional control in the context of two affective provocation cues, combat-related and negative affective words, were found over six-months follow-up with MBCT-S in high suicide risk Veterans. This study adds to a very limited literature on attentional control outcomes with mindfulness-based interventions in Veterans (Boyd et al., 2018), and high suicide risk Veterans, in particular. These findings show that attentional control, an endophenotype for SA (Mann et al., 2009), may be modifiable with MBCT-S in high suicide risk patients.