Atypical Cortical Activation during Risky Decision-making in Disruptive Behavior Disordered Youth with Histories of Suicidal Ideation

Abstract

Objective.

Suicidality is a leading cause of death among adolescents. In addition to other psychiatric conditions, youth with attention-deficit/hyperactivity disorder (ADHD) and disruptive behavior disorders (DBDs) are at heightened risk for suicide. Decision-making deficits are a hallmark symptom of ADHD/DBD and also implicated in suicidal behavior. We examined behavioral and neural differences in decision-making among ADHD/DBD youth with (SI+) and without (SI-) histories of suicidal ideation.

Method.

57 youth with ADHD/DBD (38% SI+) completed the balloon analog risk task (BART), a risky decision-making task, during fMRI scanning. Mean stop wager (mean wager at which youth bank money) was the primary measure of risk taking. We conducted whole-brain and region of interest (ROI) analyses with the anterior cingulate cortex (ACC) and the orbitofrontal cortex (OFC) during choice (win vs. inflate) and outcome (inflate vs. explode) contrasts using parametric modulators accounting for probability of balloon explosion.

Results.

There were no differences between SI+ and SI- in BART performance. SI+ showed decreasing activation in the right medial frontal gyrus when choosing inflate as explosion probability increased, compared to SI-. During explosions, SI- showed increasing activation in the L OFC as explosions became more likely. SI+ showed increasing left medial OFC activity in response to inflations as explosion probability increased.

Conclusion.

SI+ youth may show heightened sensitivity to immediate reward and decreased sensitivity to potential loss as evidenced by medial frontal gyrus activity. OFC findings suggest that SI+ youth may be drawn to reward even when there is high probability of loss.

Introduction

Suicide is the second leading cause of death among adolescents and young adults, and the steepest increase in suicide deaths occur during this developmental period (1–3). Research on suicidality, including suicide attempts and suicidal ideation (SI), has traditionally focused on samples of youth with mood disorders (4), yet, the majority of suicidal youth under age 14 have no previous psychiatric diagnoses (4–5), or have psychiatric disorders that are distinct from (6) or comorbid with mood disorders (7). Disruptive behavior disorders (DBDs) including oppositional defiant disorder and conduct disorder, as well as attention-deficit/hyperactivity disorder (ADHD), are common disorders that are also associated with suicidality (8). These impulse control disorders may have risk pathways toward suicidality that are distinct from those of mood disorders (9), yet mechanisms underlying suicidality among this population are understudied. The current study examined differences in risky decision-making behavior and its neural underpinnings among ADHD/DBD youth with (SI+) and without (SI-) lifetime reports of suicidal ideation. Despite understanding that there are different processes involved in suicidal ideation and acting on suicidal ideation, the majority of neuroimaging research has focused on neural correlates of suicidal behavior.

ADHD/DBD and Suicidality

The incidence of suicidality is high among youth with ADHD and DBDs, with estimates of suicidality (suicidal ideation or behavior) in up to 10% of youth and adolescents with ADHD (10). An analysis of the National Violent Death Reporting System found that ADHD was the most common mental health problem among elementary-aged suicide decedents (8). Additional datasets have also highlighted the increased risk of suicidal ideation and behavior among children, adolescents, and adults with ADHD (10). DBDs are comorbid with ADHD in >50% of cases (11) and themselves are associated with suicidality in children/adolescents (12–14). A number of symptoms common to both ADHD and DBDs, including aggression, emotion dysregulation, and impulsivity, are also associated with suicide attempts (15) and younger age of suicide attempt (16–18). Decision-making deficits, in particular risky decision-making, are a hallmark symptom of ADHD and DBDs (19) and have also been implicated in suicide risk models.

Decision-Making: Suicidality & ADHD/DBD

Decision-making is conceptualized as a two-phase process (20). In the choice phase, individuals weigh potential courses of action (e.g., consider potential positive or negative outcomes of different choices) and then actively complete choice-related actions, which may also be based on their motivation to avoid loss or to seek reward (21). In the outcome phase, individuals learn and process the outcome of their choices (e.g., process whether the outcome was expected or unexpected, favorable or unfavorable) and potentially use this information for further decision-making (21).

Youth with ADHD and DBD typically make riskier and more frequent risky decisions, prefer smaller immediate rewards over larger delayed rewards, and are less likely to modify behavior in response to loss or punishment (22–29). These types of decision-making deficits are also implicated in suicidality. In studies examining decision-making on gambling tasks, adolescents with a history of suicide attempt (both with and without mood disorders) demonstrated greater risk-taking propensity as measured by taking larger risks (30), an inability to predict adverse outcomes and learn from mistakes (31), and lower loss aversion (32) compared to control youth. Taken together, lower loss aversion, greater risk-taking propensity, and reduced sensitivity to punishment may be an endophenotype for suicidality which is likely overrepresented in ADHD/DBD samples (33).

Neurobiological Processes in Decision-Making: Implications for Suicidal Ideation

Multiple neural circuits are involved in both the choice and outcome phases of decision-making, as well as in risk-taking propensity, reward sensitivity, avoidance behavior, and punishment sensitivity (34–35). Further, ADHD is characterized by fronto-cortical abnormalities, including overactivation of the default mode network and underactivation of fronto-striatal, fronto-parietal, and other circuits involved in executive function and attention in addition to decision-making (36–37). However, for the current study we focused on two brain areas, the orbitofrontal cortex (OFC) and the anterior cingulate cortex (ACC), as they are implicated in decision-making in general, as well as in both suicidality and ADHD/DBD. To our knowledge, no neuroimaging studies have directly assessed neural activation in the context of suicidality in ADHD/DBD youth.

The OFC is connected with the amygdala and critical in the processing of emotional responses and behavioral inhibition (38). With respect to decision-making, the OFC has been shown to be involved in the interpretation of reward value, anticipation of outcomes, and inhibition responses to avoid unwanted or uncomfortable information (39–42). Neuroimaging studies examining suicidality in adults have attributed deficits in risk-sensitive decision-making and minimization of negative outcomes to deficits in the OFC among those with a history of suicidality (30, 33, 43). For example, adults with a history of suicide attempts (44) as well as non-depressed first-degree relatives of suicide completers (45) have both shown increased OFC activation in response to wins (44) and when making risky choices (45) as compared with healthy adults.

Among adolescents, reward-based decision-making and problems with emotion regulation and behavioral inhibition seen in ADHD and DBDs are driven, in part, by atypical OFC activity (27, 46–49); however, to date, no research has examined the potential role of OFC activity in suicidal behavior in youth with ADHD/DBD. We hypothesized that despite OFC dysfunction already documented in ADHD/DBD youth (27), those with a history of suicidality would show uniquely increased OFC activity in response to wins, as in adults.

The anterior cingulate cortex (ACC) has been shown to influence learning from previous decisions, utilizing information from both rewarding and non-rewarding outcomes to determine choices in decision-making (50). Although there is debate over the exact function of the ACC in decision-making and learning, findings have consistently shown activation of the ACC, along with the OFC and other portions of the prefrontal cortex, during the choice and outcome phases of decision-making (51). Poor decision-making seen in youth with DBDs may be related to under-activation of the ACC during reward-based risky decision-making (49). There is also evidence for a relationship between suicidality and ACC activity. One neuroimaging study found that adults with histories of suicide attempts showed increased ACC activity in response to wins compared to those without such history (44), possibly highlighting altered sensitivity to reward among those with a history of suicidality.

The goal of the current study was to assess risky decision-making behavior and underlying neural mechanisms among ADHD/DBD youth with (SI+) and without (SI-) histories of suicidal ideation utilizing a laboratory-based risky decision-making paradigm, the balloon analog risk task (BART) (50) during functional magnetic resonance imaging (fMRI). We examine youth with comorbid ADHD/DBD given very high rates of comorbidity (11). Given recent evidence for the risk of suicide among those with ADHD/DBD, as well as the role of risky decision-making in both ADHD/DBD and suicidality, research in this area is needed to better understand mechanisms of suicide risk among this population. Further, we focus on suicidal ideation given that the majority of neuroimaging research has focused on suicidal behavior, yet there are distinct mechanisms underlying suicidal ideation and behavior (53). We hypothesized that those with a history of suicidal ideation would demonstrate increased risk-taking (i.e., make more risky decisions and make riskier decisions) on the BART. Further, we hypothesized that SI+ youth would show increased activation in the ACC and in the OFC when making a risky choice and in response to wins/reward. We also conducted whole-brain analyses to further explore other involved areas given the scant research on neural mechanisms of decision-making related to suicidality. Finally, given that emotion regulation (54–56) and impulsivity (57–58) have also been associated with both suicidality and decision-making, we also examined whether potential differences in decision-making are related to deficits in emotion regulation and impulsivity.

Method

Participants

We recruited right-handed, English-speaking 11–12-year-old children (63.2% males) as part of an ongoing longitudinal study (n=71 consented, screened, and enrolled; Table 1). After consent/assent, diagnoses were determined utilizing the K-SADS-PL (59). Individuals who met DSM-5 (60) criteria for a diagnosis of ADHD and a DBD (either oppositional defiant, conduct or other specified disruptive behavior disorders) were eligible. Individuals meeting DSM-5 criteria for a current (within the last month) mood disorder, as well as those with any lifetime history of psychotic symptoms, bipolar disorder, autism spectrum disorder, substance use, neurological problems, or debilitating medical conditions were excluded, give the possible influence of those conditions on brain activation. Other exclusionary criteria included estimated Full-Scale IQ<80 (61); routine MRI contraindications; and use of any psychopharmacologic medications (other than psychostimulants) within the last two weeks (with the exception of those noted in the results). Psychostimulant medications were held the day of participation, as is conventional in ADHD research studies. All procedures were conducted in accordance with Indiana University Institutional Review Board. Additional methods for the study and study sample have been published elsewhere (62).

Table 1.

Demographics & Study Variables

	Non-suicidal (SI−) (n=35)	Suicidal (SI+) (n=22)
Age	11.91 (0.53)	11.88 (0.48)	t=0.25
Gender			χ2=1.14
Male	24 (68.6%)	12 (54.5%)
Female	11 (31.4%)	10 (45.5%)
Race			χ2=3.22
White	16 (45.7%)	9 (40.9%)
African American	10 (28.6%)	9 (40.9%)
Hispanic	0	1 (4.5%)
Multiracial	9 (25.7%)	3 (13.6%)
Parental Education	2.03 (.9)	1.73 (0.96)	t=1.25
IQ	104.31 (12.30)	103.64 (15.77)	t=.33
Emotion regulation composite	2.89 (0.30)	2.88 (0.44)	t=0.09
Emotion regulation	2.65 (0.39)	2.71 (0.40)	t=0.46
Emotional lability	2.01 (0.59)	1.95 (0.34)	t = 0.47
UPPS Impulsivity traits
Sensation seeking	2.61 (0.56)	2.69 (0.74)	t=0.45
Lack of planning	2.09 (0.55)	2.21 (0.57)	t=0.75
Lack of perseverance	2.18 (0.34)	2.23 (0.35)	t=0.56
Negative urgency	2.36 (0.72)	2.57 (0.76)	t=1.05
Positive urgency	2.52 (0.73)	2.57 (0.68)	t=0.25
Disruptive behavior disorder type a			χ2=3.38
Conduct disorder	2 (6%)	1 (4.5%)
Oppositional defiant disorder	21 (60%)	16 (72.7%)
Disruptive mood dysregulation disorder	0	1 (4.5%)
Other DBD	12 (34.3%)	4 (18.2%)
ADHD type a			χ2=2.0
Combined type	16 (45.7%)	9 (40.9%)
Inattentive type	14 (40%)	10 (45.5%)
Hyperactive type	3 (8.6%)	0
Other specified ADHD	2 (6%)	3 (13.6%)
Lifetime psychiatric diagnoses b
Depressive disorders	2 (6%)	5 (18.2%)	χ2=3.63
Anxiety disorders	6 (17.1%)	6 (27.3%)	χ2=0.83
Traumatic stress disorders	2 (6%)	2 (9.1%)	χ2=0.24
Adjustment disorders	1 (2.9%)	2 (9.1%)	χ2=1.05
Psychotropic medications			χ2=2.24
Stimulant ADHD	18 (51.4%)	10 (45.5%)
Non-stimulant ADHD	1 (2.9%)	3 (13.6%)
Antidepressants	1 (2.9%)	3 (13.6%)
BART performance
Mean reaction time	1039.05 (347.3)	1077.14 (465.4)	t=0.35
Mean stop wager	0.77 (0.26)	0.79 (0.26)	t=−0.31
Total winnings	34.9 (5.66)	33.53 (7.15)	t=0.13

Note. Values are M(SD) or n(%).

^*p<.05

^**p<.01.

^aRates are for current ADHD and DBD diagnoses.

^bRates are for lifetime history (current and past) of psychiatric diagnoses.

^cComparison is for rates of any comorbidity in the SI+ vs. SI- group.

Measures

Suicidality.

In order to study mechanisms related to suicidality more broadly we focus here on suicidal ideation, including plan, intent, and thoughts of suicide, as well as self-harm behaviors, as those have been shown to predict suicide attempts and death by suicide (63). Participants were assessed for history of suicidality and non-suicidal self-harm behavior during the K-SADS-PL semi-structured interview with both parent and child conducted by a trained clinician. Three suicidality items assessed lifetime history of the following: (1) ideation, defined as ever having thoughts of committing suicide; (2) intent, defined as ever having a plan to commit suicide; or (3) non-suicidal self-harm, defined as ever engaging in self-harm without intent of killing oneself. Individuals were also asked whether they had ever made an actual attempt to kill themselves (including self-injurious behavior with any suicidal intent) as well as the medical lethality of the attempt. Those who endorsed any item at a minimum subthreshold clinical level (i.e., infrequent or vague thoughts at least once a month; preparations with no actual intent; planned attempt; non-suicidal self-harm) were included in the SI+ group.

Emotion regulation.

Parents reported on their child’s emotional functioning on the Emotion Regulation Checklist (ERC) (64), a 24-item scale measuring emotion regulation and negativity/lability. Overall emotion regulation scores were calculated, with higher mean scores denoting higher levels of regulation (α=0.81).

Impulsivity.

Youth completed the UPPS-C (65) measuring 5 different facets of impulsivity (n=10 items each): negative urgency, positive urgency, sensation seeking, lack of planning, and lack of perseverance. Higher mean subscale scores denote higher levels of impulsive behavior (α’s=0.76–0.85).

Balloon analog risk task (BART).

The BART was administered to participants in the scanner (52) during three eight-minute runs. The BART is a decision-making task in which participants virtually inflate an image of a balloon and choose whether to risk cash rewards that increase with each balloon inflation or bank the amount and start a new balloon. Participants were told that they would win more money for larger unexploded balloons (actual earnings from participants’ highest run were paid in cash). Participants completed as many trials as possible per run. At the start of each trial, a balloon was displayed on the screen with a green decision cue indicating a button could be pressed (see Figure 1 for illustration). Participants then chose to inflate the balloon (choose inflate) or take the accumulated wager (choose win, i.e., “cash out”), via button pressing. Following choose inflate trials, there was a 0–6 second randomized delay. The balloon then either exploded (outcome explode) or inflated (outcome inflate). If the balloon exploded, participants started a new balloon trial, after a 2–4 second delay. If the balloon inflated, participants decided again whether to inflate the balloon or bank the money. For each balloon, explosions were possible at any inflation choice except the first, with the likelihood of explosion increasing as balloon size increases. A maximum of 12 inflations were possible for each balloon and the point of explosion was random for each balloon (Figure 1).

Figure 1.

Illustration of the Balloon Analog Risk Task (BART). At the start of each trial, a balloon is displayed on the screen along with a green decision cue indicating a button can be pressed (a). Participants then choose to inflate the balloon (Choose Inflate) or take the accumulated wager (Choose Win, i.e., “cash out”) via button pressing (b). The time between decision and outcome phases of each trial is randomly jittered (0–6 seconds) to enable differentiation of decision-making and feedback-related processes. Following Choose Win trials, participants view a screen that says “You Win!” for 1000 ms followed by a fixation screen for 2–4 seconds before starting a new balloon trial (c). Following Choose Inflate trials, the balloon either explodes or inflates (d). For explosions, participants view an exploding balloon for 1500 ms (e) and then the fixation screen, while inflate trials show an inflated balloon for 1500 −2500 ms before permitting another choice (f). For each balloon, explosions are possible at any inflation choice except the first, with the likelihood of explosion increasing as the balloon size increases..

The primary indicator of risky decision-making behavior was the mean stop wager, or the mean amount at which individuals chose to bank the money and stop inflating the balloon (66–67). Because chance for explosion increases with each wager increase, higher mean stop wagers denote riskier decisions.

Procedure

Imaging procedures.

Before the scanning session, participants completed mock scanning, pregnancy testing, and BART practice trials. We used a 3-Tesla Siemens Prisma MRI scanner with a 32-channel head coil. A high-resolution 3D magnetization-prepared rapid gradient echo (MPRAGE; 160 sagittal slices; 1.05×1.05×1.2mm voxel dimension) scan was used for co-registration and normalization of functional image volumes to Talairach space. BART runs were acquired using a T2*-weighted gradient echo-planar imaging (EPI) sequence (54 axial slices; voxel size 2.5×2.5×2.5mm; TR/TE 1200/29ms, flip angle 65°; Field-of-view:220×220mm, Matrix:88×88).

Statistical Analyses

Behavioral analyses.

We conducted t-tests to compare group differences (SI+ vs. SI-) in mean stop wager, total winnings, and mean reaction times. We also examined correlations between BART measures and emotion regulation and impulsivity.

Imaging analyses.

Structural images were reconstructed and then processed with the standard FreeSurfer pipeline (version 6.0) (68) in order to define regions of interest on an individual level. Surface maps were generated for each hemisphere, and the cortex was automatically parcellated into 128 distinct brain regions using the Desikan-Killiany atlas (69).

Image preprocessing of each blood-oxygen level-dependent (BOLD) time-series, using AFNI software (70), consisted of slice-time correction, de-spiking of time series outliers (3dDespike algorithm), motion correction via realignment to a single baseline time point, registering the functional image to the structural image, and spatial smoothing with a 6-mm full-width at half-maximum Gaussian kernel.

For noise reduction, individual time points with high motion (>0.5mm Euclidean norm of motion from previous time point) and/or noise (>10% of voxels across the brain considered time-series outliers; AFNI command 3dToutcount) were excluded from analyses. Participant runs were excluded if >10% of time points were excluded based on above criteria, motion exceeded 5mm from baseline at any time point, or >10% of reaction times >5000ms, signaling inattention.

After preprocessing and noise reduction, runs were concatenated, and a general linear regression model with random effects was created to estimate event-related responses. Six motion parameters, six motion derivatives, and detrending terms to correct for scanner drift were modeled, and an additional nuisance regressor was included for choice trials with reaction times >5000ms.

Regressors of interest were created by convolving the timing of each condition with a hemodynamic-response function to create a model BOLD time series for each condition (choose inflate, choose win), outcomes (outcome win, outcome inflate, outcome explode). In addition, balloon explosion probabilities were included at the event level as parametric modulators (e.g., Choose Inflate*P(explode), Outcome Inflate*P(explode)). These parametric modulators (ranging from 0% to 90%, though the range of probabilities experienced by each subject depends on their performance) incorporate the explosion probability at each inflation to measure the impact of the changing risk levels on brain activity risk of explosion.

Choice events were aligned to the time at which the button was pressed for a choice: inflating the balloon (choose inflate) or discontinuing inflation and banking the money (choose win). Outcome events were modeled as the time point that included balloon explosion or successful inflation. For participant level analyses, contrast maps for choice (choose inflate vs. win) and outcome (outcome inflate vs. explode) were obtained, as well as corresponding parametric modulation contrasts (e.g., choose inflate*P(explode) – choose win*P(explode); risk of explosion ranging from 0–90%). Outcome win was not examined because it involves no uncertainty. Individual activation maps were warped to a standard Talairach atlas for group whole-brain analyses. For group analyses, individual voxels were considered significant at p<.01, and a Monte Carlo simulation was conducted using the estimated spatial auto-correlation of the residuals, to determine a cluster-size correction for group-level significance (201 voxels).

In addition, activation contrasts from each participant were extracted from right and left middle ACC and medial OFC. We conducted t-tests to examine group differences in activation contrasts and correlations to examine relationships between ROI activity and impulsivity and emotion regulation.

Results

A total of 57 youth with usable data (n=4 excluded for motion in scanner, n=3 excluded for reaction time outliers) were included in analyses (44% white; 63% male, mean age=11.01, SD=0.50); 38.6% (n=22) had a lifetime history of suicidal ideation (see Table 1). No participants endorsed active/current suicidal ideation. No participants reported lifetime history of actual suicide attempt. There were no differences between SI+ (n=22) and SI- (n=35) groups across gender, race/ethnicity, age, IQ, or parental education (p’s>.10). The majority of youth met DSM5 criteria for ADHD combined (43.9%, n=25) or inattentive (42%, n=24) type and oppositional defiant disorder (64.9%, n=37); there were no differences in type of ADHD or DBD between SI+ and SI- (χ²=2.0, p>.10). Examining lifetime history of other psychiatric disorders, participants in the SI+ group were more likely to have other comorbid diagnoses (SI+ n=15, 68.2% vs. SI- n=14, 40%; χ²=4.29, p=.04); however, there were no group differences in rates of any specific lifetime psychiatric disorders (p’s>.05), emotion regulation (t=−0.76, p=.45), or any impulsivity traits (p’s>.05; Table 1). It was later discovered that three participants started psychotropic medications prior to scanning, but after the initial study visit when exclusion criteria were determined (Table 1).

BART Behavior

There were no group differences in decision-making behavior as measured by mean stop wager (t=−0.31), total winnings (t=0.13), or mean reaction time (t=0.35; p’s>.10). BART behavior measures were not related to impulsivity, emotion regulation, gender, race, or IQ (r’s −0.01–0.15, p’s>.10; Table 1).

Imaging Results

Choice contrast (Win vs. Inflate).

There were no group differences in either the standard or parametric modulator contrast for ACC or OFC ROI activation (p’s>.10). However, in whole-brain analyses using the parametric modulator (P(explosion)), both groups demonstrated relative increases in frontoparietal activity when choosing win as explosion probability increased; group differences were detected in the left precentral gyrus (Brodmann Area 6; Talairach peak (29, −7, 60); 214 voxels, Figure 2). As explosion probability increased, SI+ showed increasing activity in the left precentral gyrus when choosing win and decreasing activity when choosing inflate (Figure 2). Conversely, SI- showed similar responses during choose win and inflate, regardless of explosion probability.

Figure 2.

A significant cluster (214 voxels) was found to differ between groups in the right precentral gyrus (BA6), with peak Talairach coordinates (29,−7,60) for voxel-level p < .01 and a cluster-size threshold correcting for multiple comparisons (p < .05). In the Choose Win contrast, PCG activation increased with increasing explosion probability in the Suicidality group. In the Choose Inflate contrast, PCG activation decreased with increasing explosion probability in the Suicidality group. Lines represent group means of this estimated linear relationship between balloon explosion probability (x axis) and activation intensities of the blood oxygen dependent (BOLD) signal (y axis) of the cluster. Dashed lines indicate standard error of the mean.

Outcome contrast (Inflate vs. Explode).

Examining the ROIs, neither left (t=−1.46, p=.15) nor right (t=−1.55, p=.13) medial OFC activation differed between groups. In the parametric modulation contrast, there was a significant group difference in the left medial OFC (t=−2.86, p=.01); as the probability of explosion increased, SI+ showed relatively greater activation to inflations, whereas SI- had relatively greater activation during explosions (Figure 3). There were no group differences in the ACC or whole-brain for standard or parametric modulator contrasts.

Figure 3.

Group differences (Suicidality vs Non Suicidality) also emerged from the outcome contrast (Explode vs. Inflate). ROI analysis revealed significant differences in the left medial orbitofrontal cortex (lmOFC). During the Outcome Explode condition, lmOFC activation increased with increasing explosion probability in the Non-Suicidality group. During Outcome Inflate trials, lmOFC activation increased with increasing explosion probability in the Suicidality group. Line graphs represent group means of the estimated linear relationship between balloon explosion probability (x axis) and activation intensities of the blood oxygen dependent (BOLD) signal (y axis) of the cluster. Dashed lines indicate standard error of the mean.

Relationship to behavioral measures.

In correlational analyses, positive urgency was significantly related to left medial OFC in the standard (r=−0.32, p=.02) and parametric modulator (r=−0.28, p=.04) outcome contrasts; higher levels of positive urgency were related to decreasing activity in response to inflations and increasing activity in response to explosions as explosion probability increased. There were no significant findings for choice contrasts. Positive urgency was not related to OFC activation in the standard or parametric modulator contrasts in choice contrasts (p’s>.10). No other impulsivity or emotion regulation measures were related to ROIs in choice/outcome contrasts across standard/parametric modulator analyses.

Discussion

The current study examined behavioral differences in decision-making and related neural activation between non-depressed, pre-adolescent youth with ADHD/DBDs who have or have not experienced suicidal ideations and plans. We found that although there were no differences in risky decision-making behavior between groups during the BART, SI+ youth showed unique neural activation during both the choice and outcome phases of decision-making. Findings elucidate novel decision-making mechanisms that may underlie increased risk of suicidal ideation among a particularly risky and understudied population.

During the choice phase of decision-making, contrary to hypotheses, there were no differences in ACC activity across groups. However, whole-brain analyses revealed unique activity in the precentral gyrus, such that as explosion probability increased, SI+ showed increasing activation in the left precentral gyrus when making safer choices and decreasing activation when making riskier choices, while SI- showed similar activation as explosion probability increased. These results suggest that SI+ are more sensitive to immediate reward – especially as chances of loss increase – compared to SI-, even among youth with ADHD/DBD who are already predisposed to rewarding stimuli. At the same time, it seems that SI+ were less attuned to the possibility of loss, even with increasing probability of loss. We speculate that as stressors and potential losses from risky behaviors mount, individuals who are less able to predict those losses and who prefer immediate gratification may be more likely to get frustrated and therefore contemplate suicide.

Findings for atypical precentral gyrus activation among suicidal youth also parallels evidence for the role of the frontal gyri in the development of suicidality in adults (46, 71–72), and more specifically, evidence for structural abnormalities in the precentral gyrus among suicidal adults (73). Further, greater grey matter volume in the precentral gyrus has been related to non-planning impulsivity among adolescents with major depressive disorder (74), suggesting that precentral gyrus dysfunction impairs inhibitory control (75–76). Still, precentral gyrus activity was not correlated with non-planning impulsivity in the sample, although this is likely due to a restriction of range, given that all youth in the sample had ADHD. Nonetheless, considering the connections between this region and fronto-cortical and thalamic regions (77) and its implication in suicidality in depression and anxiety (78–80), our results contribute to an increasing body of literature suggesting that this area has multiple roles in motor planning, threat processing, and value-based decision-making (77), all of which may influence suicidality. Decreased activity in the precentral gyrus as risk for an adverse event increases may be consistent with impaired behavioral inhibition (75–76), a deficit in evidence accumulation (77), or a deficit in readiness potential driving potentially impaired decision-making, contributing to a “motor cognitive” deficit that may predispose to suicidal thinking and, ultimately, behavior.

We also reported atypical OFC activity among SI+ during the outcome phase as explosion probability increased. SI- showed increasing activation in the left medial OFC in response to loss (explosion) and no change in activity in response to wins (inflate) with increasing explosion probability. Conversely, SI+ showed no change in OFC activity in response to losses and increasing OFC activity in response to wins as explosion probability increased. Results parallel findings for greater right OFC responsivity to wins vs. losses in suicidal adults compared to non-suicidal depressed adults and healthy control adults (44). The lack of change in activation in the OFC in response to loss despite increasing probability of loss may highlight a deficit in SI+ youth’s ability to process and learn from loss, consistent with findings among lower loss aversion and inability to learn from mistakes among youth with history of suicide attempt (31).

The OFC is also involved in emotion regulation; however, our measure of emotion regulation was not correlated with OFC activity. Again, given that the sample consisted of youth with ADHD and DBD, and there were no group differences emotion regulation, lack of findings are likely due to restriction of range. There was a relationship between OFC activity and positive urgency, or the tendency to engage in impulsive action when experiencing positive emotions. These findings are consistent with previous work that has shown a role of the OFC in emotion-based impulsivity (80).

BART Performance

Contrary to hypotheses, there were no differences between SI+ and SI- decision-making behavior on the BART. Lack of differences in risk-taking or risky decision-making behavior is inconsistent with previous findings for increased risk-taking propensity among youth with a history of suicidality (30–31). There are a few possible reasons for this. First, all participants in our sample met criteria for ADHD and DBD, which is associated with risk-taking behavior. Thus, lack of differences could be due to restriction of range, if all youth were already prone to risky decision-making (28). Second, previous studies examined youth with actual suicidal attempts (30–31), while in the study sample, participants only endorsed history of suicidal ideation. Thus, there may be a difference between those with suicidal ideation versus those who act on intentions. Third, it may be that the BART, at least delivered over 24 minutes, is insensitive to behavioral differences, an observation we reported previously when comparing high-risk and healthy youth (81).

Limitations

The study is not without limitations. All SI+ reported a history of suicidal ideation only; there were no youth with a history of actual suicide attempt or behavior, thus, it is unknown whether there are differences in those with actual suicidal behavior. Future research should asses this given evidence for differences in adolescents with suicidal ideation and adolescents who act on ideation (82). Research should also assess suicidality utilizing other validated measures that assess more detailed information on frequency and intensity of suicidality (83). Second, there was some evidence for greater psychiatric comorbidity among SI+; additional research should further explore the role of other comorbidity in suicidality and decision-making, especially given high comorbidity with ADHD. Third, a small number of participants took medications on or around the time of brain scanning, thus we cannot exclude the possibility that this influenced our findings. Finally, the cross-sectional design limited evaluation of developmental trajectories associated with results.

Conclusion

To our knowledge, this is the first imaging study examining suicidality and neural mechanisms of decision-making among a high-risk group of non-depressed youth with ADHD/DBD. This study is particularly important given recent findings that youth with ADHD/DBD are at higher risk of suicide (8). Further, our sample represents a developmental stage (aged 11–12) during which ADHD and DBDs are more prevalent than mood disorders (82), thus, the importance of examining potential risk factors for suicidality among this largely prepubertal group is important. Even beyond decision-making deficits seen among youth with ADHD/DBD, youth with a history of suicidality showed unique decision-making driven by the MFG, as well as unique responsivity to reward and lack of responsivity to loss / learning from losses as driven by atypical OFC activity. Prospective studies are needed to test whether neural activation differences observed here are robust biomarkers of later suicidal behavior among youth with DBDs and whether neural processing of decisions is malleable to intervention among this highly vulnerable group.