Available via license: CC BY 4.0
Content may be subject to copyright.
Page 1/27
Perception of social inclusion/exclusion and
response inhibition in adolescents with past suicide
attempt: a multimodal task-based fMRI study
Fabrice Jollant ( fabrice.jollant@universite-paris-saclay.fr )
Paris-Saclay University https://orcid.org/0000-0001-5809-4503
Anthony Gifuni
Stanford University https://orcid.org/0000-0003-2418-5591
Fabricio Pereira
Mallar Chakravarty
Douglas Mental Health Institute, Montréal, Canada
Martin Lepage
McGill University https://orcid.org/0000-0003-4345-6502
Henry Chase
University of Pittsburgh
Marie-Claude Geoffroy
Université de Montréal
Eric Lacourse
Mary Phillips
University of Pittsburgh School of Medicine https://orcid.org/0000-0003-4958-1291
Gustavo Turecki
McGill University https://orcid.org/0000-0003-4075-2736
Johanne Renaud
Article
Keywords: Adolescent, Attempted Suicide, Functional Neuroimaging, Social Exclusion, Executive Function
Posted Date: January 5th, 2023
DOI: https://doi.org/10.21203/rs.3.rs-2271723/v1
License: This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
Page 2/27
Abstract
The occurrence of suicidal behaviors increases during adolescence. Hypersensitivity to negative social
signals and decits in cognitive control are putative mechanisms of suicidal behaviors, which necessitate
conrmation in youths. Multidomain functional neuroimaging could enhance the identication of
patients at suicidal risk beyond standard clinical measures. Three groups of adolescents (N = 96; 78%
females, age = 11.6–18.1) were included: patients with depressive disorders and previous suicide
attempts (SA,
n
= 29); patient controls with depressive disorders but without suicide attempt (PC,
n
= 35);
and healthy controls (HC,
n
= 32). We scanned participants with 3T-MRI during social inclusion/exclusion
(Cyberball Game) and response inhibition (Go-NoGo) tasks. Neural activation was indexed by the blood-
oxygenation-level dependent (BOLD) of the hemodynamic response during three conditions in the
Cyberball Game (“Control condition”, “Social Inclusion”, and “Social Exclusion”), and two conditions in Go-
NoGo task (“Go” and “NoGo” blocks). ANCOVA-style analysis identied group effects across three whole-
brain contrasts: 1) NoGo vs. Go, 2) Social inclusion vs. control condition, 3) Social inclusion vs. control
condition). Normalized contrasts in signicant clusters were used to train a support vector machine-
based classier with a stratied 5-fold cross-validation, and diagnostic performance was assessed. In
line with previous adult studies, we found that SA had lower activation in the left insula during social
inclusion vs. control condition compared to PC and HC. We also found that SA compared to PC had
higher activity in the right middle prefrontal gyrus during social exclusion vs. control condition, and in
bilateral precentral gyri during NoGo vs. Go conditions. Task-related measures (Self-reported emotional
reactivity in the Cyberball Game, response times and number of errors in the Go-NoGo Task) did not
discriminate between groups. Moreover, while clinical data (Self-reported depression and impulsivity
scores) yielded moderate accuracy (Accuracy: 70%/ Area Under Curve: 0.81), activity during Go-NoGo
(81%/0.90), Cyberball Game (89%/0.90), or a combination (88%/0.95) signicantly enhanced
identication of past suicidal behaviors. In conclusion, adolescent suicidal behaviors are likely
associated with neural alterations across multiple domains. Alterations in the processing of social
perception and response inhibition may underlie the development of suicidal crises, from onset with
social triggers to susceptibility to act out. Neuroimaging should be further tested as a tool to predict
suicidal behavior.
1. Introduction
Adolescent suicidal behaviors are increasingly recognized as a leading public health concern [1, 2].
Suicide ranks as the second or third leading cause of youth mortality worldwide [3], and rates have
tended to increase in recent years in the USA and Canada [4, 5]. Alarmingly, non-fatal suicidal ideation
and behaviors during adolescence have a prevalence of 12% and 4%, respectively [6, 7]. Adolescence is
associated with an abrupt increase in suicidal thoughts and behaviors (STB) [7, 8], which later increases
the risk for psychological diculties [9] and STB in adulthood [10]. During the recent COVID-19 pandemic,
an important increase in STB in adolescents but not in adults has been observed in several countries [11,
Page 3/27
12]. Despite the burden of adolescent STB, our understanding of adolescent STB neural basis remains
rudimentary, limiting preventive and therapeutic approaches.
Clinical evaluation to predict suicidal behavior notoriously lack accuracy, and the use of historical and
self-reported data provides marginal improvement to prognostic models [13, 14]. Psychiatric research has
emphasized the importance of "biomarkers" as a stepping stone toward developing predictive tools and
facilitating personalized care [15]. With its ability to identify brain structural and functional alterations
correlating with psychopathology, neuroimaging has emerged as a potential tool to provide biomarkers
associated with suicide risk [16]. Most neuroimaging studies have probed brain alterations associated
with suicide risk in adult populations [17–20], with studies in adolescents recently increasing in number
[21]. To date, a few neuroimaging studies reported on single functional neural modalities, identifying
disparate processes linked with STB, including facial emotional processing [22], self-referencing [23], and
social cognition [24]. However, single neuroimaging modalities might be insucient to capture fully the
underlying neural predisposition to adolescent STB. In the current study, we jointly examined two distinct
functional domains that might both contribute to adolescent suicidal risk from a painful social situation
to the emotional and behavioral response: sensitivity to social inclusion/exclusion and response
inhibition.
Several theories of suicide recognize that the feeling of social disconnection is an important precipitant
of STB [25–27]. From puberty to adulthood, social relationships become increasingly salient for
adolescents [28]. Numerous epidemiological and clinical evidence suggests that social exclusion, peer
victimization, and cyberbullying contribute to adolescent STB [29–32]. One of the most validated
paradigms to investigate the neural correlates associated with social exclusion is the fMRI Cyberball
Game [33, 34], a virtual ball-tossing game simulating inclusion and exclusion [35]. While most research
has examined the neural correlates of social exclusion among healthy subjects [34, 36], fewer studies
have addressed patients with STB. A study in adult females showed altered activity in the left insula and
supramarginal gyrus in patients with past suicide attempt vs. control patients during exclusion vs.
inclusion [37]. One study in adolescents with depression and non-suicidal self-injury, a behavior
overlapping with suicidal behaviors [38], showed elevated activation of the medial and ventrolateral
prefrontal cortex during social exclusion relative to inclusion (Groschwitz et al., 2016). Finally, in a sample
of adolescents with suicide ideation, another study [39] revealed that the subgroup of adolescents with
recent suicide attempts had elevated activity in the anterior cingulate cortex and lateral prefrontal cortex
(PFC) during any social interaction, including periods of inclusion and exclusion. Given the heterogeneity
of these samples and their small sizes, these ndings need further exploration and replication. It also
remains unclear whether these anomalies related to social processes correlate with other cognitive
processes related to suicidality, such as cognitive control.
Cognitive control, the ability to exibly orient attention to goal-relevant information and discard irrelevant
information, has been implicated in the suicidal transition from suicidal ideas to acts [40–42]. Cognitive
control is notably instrumental to ecient regulation of emotions [43]. In adolescents, however, the
association between cognitive control and suicidal behaviors is still unclear [44], which might be
Page 4/27
explained by at least two reasons. First, suicidal behaviors are notoriously heterogeneous, with some
suicidal attempts done with minimal premeditation, while others are planned over several days or weeks
[45]. Second, adolescence is characterized by rapid developmental shifts in cognitive control [46].
Adolescents’ cognitive skills are typically characterized by higher variability between assessments and
lacking the exibility seen in adulthood [47]. Substantial empirical evidence shows that immaturity in
cognitive control systems predisposes youth to impulsive sensation-seeking and riskier decision-making
[48, 49]. A common paradigm to investigate neural correlates of cognitive control is the Go-NoGo Task
[50], which is not only sensitive to age-related maturation [51] but also disorder-related alterations [52].
One previous study in a small sample [53] found that adolescents with past suicide attempts exhibited
decreased neural activation in the right anterior cingulate cortex compared to depressed controls, but not
compared to healthy controls.
In the current study, the Cyberball Game and Go-NoGo tasks during fMRI in three groups: depressed
adolescents with past suicidal behaviors, depressed adolescents without past suicidal behaviors, and
healthy controls. We expected to nd neural disturbances in each functional domain in adolescents with
past suicidal behaviors compared to control groups, including in regions involved in emotional and
mentalizing processes, such as the anterior cingulate cortex, the striatum, the insula, or the precuneus;
and in cognitive control regions, such as the lateral prefrontal cortex [17]. Additionally, we explored the
potential diagnostic utility of neuroimaging disturbances using either linear regression or supervised
machine learning classiers. We posited that each task-based neuroimaging modality would provide
superior diagnostic accuracy compared to clinical data such as depressive symptoms or self-reported
impulsivity. Furthermore, we expected that combining functional domains would be better than single
domains and that using supervised machine learning methods would improve the identication of
adolescents with past suicidal behaviors.
2. Methods
2.1 Participants
From September 2012 to January 2019, three groups of adolescents (N = 104) aged 11 to 18 years were
recruited: 1) Adolescents with a depressive disorder and a history of at least one suicide attempt (SA;
n
=
30); 2) Adolescents with a depressive disorder without a lifetime history of suicide attempt (Patient
controls, PC;
n
= 38); and 3) Adolescents without a personal history of psychiatric disorder or a suicide
attempt (healthy controls, HC;
n
= 36).
A suicide attempt was dened as any self-injurious behavior carried out with the intent to die [54]. This
did not include aborted suicide attempts (i.e., halted by oneself), interrupted suicide attempts (i.e., halted
by someone), or non-suicidal self-injuries. Depressive disorders, dened by DSM-IV criteria [55], included
major depressive disorders, dysthymia, and depressive disorder not otherwise specied.
Page 5/27
Participants with a depressive disorder (SA and PC) were primarily recruited from a third-line psychiatric
clinic specialized in adolescent depressive disorders (at the Douglas Mental Health University Institute,
Montreal, Canada), with additional cases recruited from aliated community clinics. Access to these
specialized psychiatry services is limited to patients aged below 18 years. HC were recruited from the
community through advertisements posted in schools, local clinics, youth centers, and groups of parents
on social media. HC were initially screened through a phone interview by a trained research assistant to
exclude adolescents with a history of psychiatric disorders or a suicide attempt.
Exclusion criteria included neurological disorders (e.g., epilepsy, brain tumor), traumatic brain injury (> 1
min loss of consciousness, neuroimaging anomaly, or persistent post-concussive symptoms), autistic
spectrum disorder, bipolar disorders, psychotic disorders, intelligence quotient (IQ) less than 70,
pregnancy, and any contraindication for magnetic resonance neuroimaging (e.g., dental apparatus). A
family history of suicidal behavior was an additional exclusion criterion for the HC group (but not the SA
or PC groups) because of evidence that some neural phenotypes are transmitted within families with
suicide histories [56, 57]
Neuroimaging data were not available for 5 participants (2 PC and 3 HC) either due to claustrophobia (n
= 1), refusal (n = 2), or corrupted or irretrievable data (n = 2). Three additional participants (1 SA, 1 PC, 1
HC) were excluded from analyses due to poor data quality (see below for quality check criteria). Hence,
after exclusion and missing data, the nal analyzed sample (N = 96) included 29 SA, 35 PC, and 32 HC.
The research protocol was approved by the Douglas Institute Research Ethics Board (Protocol 12/20).
Consent was provided by at least one parent or legal guardian, and all adolescents assented to the
experiment. All participants were compensated when completing the assessment procedure (50 CAD).
2.2 Assessment
Assessments were completed in two or three visits, which included clinical and neuropsychological
assessments and a neuroimaging session performed within the same week. Suicidal behaviors were
assessed with the Suicide History Questionnaire [see Appendix in 58], a clinical interview with a child and
adolescent psychiatrist (JR), and cross-validation with notes from patients’ medical les (reviewed by
AJG). Psychiatric disorders were characterized by the Kiddie Schedule for Affective Disorders and
Schizophrenia—Present and Lifetime [59, K-SADS-PL; 60]. Current depressive symptoms were measured
with the Beck Depression Inventory-II [BDI-II, 61]. Full-scale IQ was measured using the Wechsler
Intelligence Scale for Children − 4th edition [WISC-IV; 62] in participants younger than 17 years and with
the Wechsler Adult Intelligence Scale − 4th edition [WAIS-IV; 63] for participants aged 17–18 years. Self-
reported impulsivity was assessed with the Barratt Impulsivity Scale [BIS-11; 64].
2.3 Imaging procedure
2.3.1 Description of tasks
Page 6/27
Cyberball Game.
The Cyberball Game is a computer ball-tossing game seeking to simulate social
exclusion [33, 35, 65]. Through instructions given in the scanner, participants were invited to play online
with two other players, which were presented as real, although they were pre-programmed. Participants
were represented by a pair of hands in the lower center portion of the screen, while other players were
represented by animated cartoons next to a named prole picture in each upper corner of the screen. After
receiving the virtual ball, subjects could throw it to either player by selecting one of two buttons on a
remote control in their right hand. The Cyberball Game consisted of three rounds lasting 2:30 minutes
following xed order: 1) Passive viewing, 2) Social inclusion, 3) Social exclusion. In the rst round
(“Passive viewing”), participants were told that a connection problem prevented them from joining the
game and that they were to watch the two other players exchange the ball. This rst round serves as a
baseline for the following rounds and aims to increase the realism of the task. In the second round
(“Social inclusion”), participants receive the ball randomly one-third of the time and can throw back the
ball to either player. The third round (“Social exclusion”) starts as the second one, but after 60 seconds
and without any warning, the other players stop throwing the ball toward the participant. The total
duration of the task was 10 minutes. After the scanning session, the distress induced by social exclusion
was documented with the Need-Threat Scale [NTS; 66, 67], which includes ratings of self-esteem (e.g., “I
felt liked”), belongingness (e.g., “I felt rejected”), meaningful existence (e.g., “I felt invisible”), and control
(e.g., “I felt powerful”), scaled from 1 (not at all) to 5 (very much). Items were reverse-coded when
appropriate and averaged to create a composite score. After completing the NTS, participants were
debriefed about the task.
Go-NoGo Task
. Cognitive inhibition was tested with a classical version of the Go-NoGo task [68],
implemented in E-Prime 2.0.10.182 (USA). The task comprised a block-design paradigm in which
participants were shown a total of 144 letters across 6 blocks for a total duration of 6.1 minutes. Trials
were distributed in 2 block types: 3 “Go” blocks (Block A), and 3 “NoGo” blocks (Block B), presented in an
ABBAAB order. In “Go” blocks, participants were instructed to press a button in response to visually
presented letters using their right index nger as quickly as possible. In “NoGo” blocks, participants were
instructed to avoid pressing the button to a non-target letter (letter X), while still pressing the button in
response to target letters (i.e., letters other than X). Each block (either “Go” or “NoGo”) started with a 20-
second blank-screen resting period and 5-second instructions followed by 24 trials. Each trial consisted of
a black xation cross on a white screen followed by a letter. The duration of the xation cross varied
between 700, 900, 1100, or 1300 milliseconds, randomized to prevent habituation, with 6 trials of each
duration resulting in an average of 1000 milliseconds. Letters (target or non-target) were presented for
500 milliseconds on a white screen. In each block (“Go” or “NoGo”), 12 target letters (50%) and 12 non-
target letters (50%) were presented in a pre-determined pseudorandomized order. For all trials, reaction
time (time between stimulus onset and button press response), omission errors (i.e., not responding to a
target error), and commission errors (i.e., responding to a non-target letter during a “NoGo” Block) were
recorded. The sum of omission errors is usually interpreted as indicative of attentional or speed decits,
whereas commission scores would reect inhibition decits [69]
2.3.2 Imaging Acquisition and preprocessing
Page 7/27
Magnetic resonance imaging (MRI) scans were acquired at the Douglas Cerebral Imaging Centre using a
Siemens Magnetom Trio (Tim System 3T, MR B17) scanner equipped with a 12-channel head coil. The
complete scanning procedure (see Supplements -
Scanning Protocol
and
Neuroimaging acquisition
for
full description) included a high-resolution T1 anatomical scan (repetition time [TR] = 2300 ms; echo time
[TE] = 2.98 ms; inversion time [TI] = 900 ms; ip angle [FA] = 9°; eld of view [FOV] = 256 mm; voxel size
[VS] = 1x1x1 mm; matrix =: 256x256, acquisition time [T] = 9.23 min), task-based functional MRI with a
socio-emotional paradigm (Cyberball Game; TR = 3000ms; TE = 25ms; FA = 90°, FOV = 200 mm; matrix =
64x64; VS = 3.1x3.1x4.0 mm; T = 5.5 min) and a cognitive control paradigm (Go-NoGo task; TR = 2090ms;
TE = 30 ms ; FA = 90°; FOV = 224 mm; matrix: 64x64; voxel = 3.5x3.5x3.5 mm; T = 6.27 min). The
functional images were quality controlled and preprocessed using fMRIPrep 1.4.1 [70–72].
2.4 Statistical analyses
2.4.1 Demographic and clinical differences
Statistical analyses were performed in R v3.6.0 [73], implemented in Rstudio v1.1.383 [74]. Clinical and
behavioral continuous data were compared across groups with 3-way ANOVAs, and Tukey’s HSD Post
Hoc tests for bivariate comparisons. Categorical group data were compared with chi-square tests.
Correlations between clinical measures and brain activity extracted from signicant clusters were
computed with Spearman’s rank-order correlation. The alpha level was set a priori at p = 0.05 with
Bonferroni correction based on the number of signicant clusters in across both fMRI tasks.
2.4.2 Neuroimaging analysis
See the Supplements -
Anatomical data preprocessing
and
Functional data preprocessing
for the
complete neuroimaging analysis procedure. In brief, we derived within-subject contrasts maps of interest
using a block-designed models. For the Cyberball Game, we calculated an inclusion and an exclusion
contrast, using the “Passive viewing” condition as the control condition. For the Go-NoGo Task, the
contrast of interest was NoGo vs. Go blocks, reecting activity when the task demands cognitive
inhibition as opposed to a simple motor response. Between-participants whole-brain analyses were
conducted with 3dMVM in AFNI version 19.3.11 [75]. Given the effect of age, sex, and IQ on neural activity
and cortical anatomy [76, 77], the group analyses controlled for these covariates. Signicance thresholds
were set at voxel level p-uncorrected < 0.005. To minimize false-positives, a more stringent level at p <
0.001 was also tested. The resulting statistical maps were corrected for multiple comparisons at p-
corrected < 0.05 using AFNI’s 3dClustsim algorithm. Noise volume was simulated assuming a spatical
autocorrelation function (ACF) given by a mixed-model of the form a*exp(-r*r/(2*b*b))+(1-a)*exp(-r/c),
where a, b, c parameters were estimated by 3dFWHMx using residual timeseries leftovers post-GLM
tting. The cluster size surviving whole brain correction, using a grey-matter mask, was determined to be
k > 517 for Cyberball Game and k > 504 for Go-NoGo. Post-hoc pairwise analyses were only conducted in
clusters that were found to be signicant at the three-group comparison level in order to limit false
positives. Normalized cluster contrast were extracted for correlational analyses with clinical, behavioral,
and neuroimaging data.
Page 8/27
2.4.3 Diagnostic accuracy provided by multidomain
functional MRI
Classication models identifying participants with past suicide attempts were computed with a support
vector machine (SVM) based-classier, a supervised learning method already tested in suicide prediction
[78]. We adopted an SVM-based model using a linear kernel, which calculates high-dimension linear
decision boundaries (“hyperplanes”) using the “max-margin principle”. For comparison, we computed a
conventional classication model using a standard logistic regression. The analysis aimed to classify
suicidal (SA) vs. non-suicidal participants (PC and HC). Two baseline models were computed: 1) Only
sociodemographic data (age, sex, IQ), 2) Sociodemographic and clinical data (age, sex, IQ, BDI, and BIS
total scores). Models with extracted fMRI contrast coecients from single or combined tasks were built
on top of each baseline model for a total of 8 models. Models only used contrast coecients from
signicant clusters at the three-group level at p < 0.005. The performance of each model was estimated
via a stratied 5-fold cross-validation. Classication accuracy and receiver operating characteristic curves
(ROC) area under the curve (AUC) were averaged across each fold (k = 5) and compared with baseline
models with Welch’s t-tests. Training and testing classication models were implemented in Scikit-learn
(v0.23.1; [79]), a machine learning toolkit for python.
3. Results
3.1 Group sociodemographic and clinical characteristics
Sociodemographic and clinical characteristics of the three groups are presented in Table1. Groups were
broadly similar, notably SA and PC. See sTable 1 for medication use and dose ranges.
Page 9/27
Table 1
Comparison of sociodemographic, clinical and task-related behavioral variables between the three groups
Suicide
Attempt Patient
Controls Healthy
Controls Group comparison
(N = 29) (N = 35) (N = 32)
χ2/ F-stat p-
value Post-hoc
comparison
Sex
Female, n (%) 25 (86) 28 (80) 22 (73) 2.8 (df =
2) 0.24 -
Age (years)
Mean (sd) 16.3 (1.0) 16.0 (1.5) 15.3 (1.4) 4.5 (df =
2,93) 0.01* SA > HC
Race/Ethnicity
Asian, n (%) < 5 < 5 < 5 0.3 (df =
2) 0.9
Black, n (%) < 5 0 < 5 1.7 (df =
2) 0.4
White, n (%) 22 (76) 25 (71) 19 (69) 2.1 (df =
2) 0.4
First Nations, n (%) < 5 5 (14) < 5 6.4 (df =
2) 0.04* PC > SA,HC
Latinx/Hispanic, n
(%) < 5 0 0 4.7 (df =
2) 0.09
Multiethnic, n (%) < 5 < 5 8 (25) 6.9 (df =
2) 0.03 HC > SA,PC
Parental Education
Elementary, n (%) 8 (28) < 5 0 (0) 11.6 (df
= 2) 0.003* SA > PC,HC
High School, n (%) 5 (18) 5 (14) 7 (23) 0.8 (df =
2) 0.7 -
College, n (%) < 5 < 5 6 (19) 1.6 (df =
2) 0.4 -
University, n (%) 12 (43) 24 (69) 18 (58) 4.8 (df =
2) 0.09 -
Intelligence Quotient (WISC/WAIS)
Page 10/27
Suicide
Attempt Patient
Controls Healthy
Controls Group comparison
(N = 29) (N = 35) (N = 32)
Total, Mean (sd) 102.3
(15.9) 108.4
(13.7) 111.0
(12.7) 2.45 (df
= 2,93) 0.1 -
Beck Depression Scale - II
Total, Mean (sd) 30.2
(12.7) 25.6
(12.3) 5.9 (5.8) 46.4 (df
= 2,93) <
0.001* SA,PC > HC
Barratt Impulsivity
Scale
Total, Mean (sd) 73.4 (9.1) 65.26
(11.1) 58.9
(11.0) 10.4 (df
= 2,93) <
0.001* SA,PC > HC
Attention, Mean (sd) 20.5 (3.4) 17.4 (3.5) 14.8 (4.3) 15.5 (df
= 2,93) <
0.001* SA > PC >
HC
Motor, Mean (sd) 21.4 (4.6) 19.5 (4.3) 18.4 (3.7) 3.0 (df =
2,93) 0.06 -
Non-Planning, Mean
(sd) 30.5 (3.6) 28.4 (5.6) 25.7 (5.7) 6.0 (df =
2,93) 0.003 SA,PC > HC
NSSI Lifetime history
NSSI lifetime history,
n (%) 25 (86) 22 (63) 5 (16) 31.2 (df
= 2) <
0.001* SA,PC > HC
Psychiatric diagnosis
MDD, n (%) 14 (48) 25 (71) - 2.7 (df =
1) 0.1 -
Dysthymia, n (%) 5 (17) 8 (23) - 0.1 (df =
1) 0.8 -
DDNOS, n (%) 11 (38) < 5 - 4.8 (df =
1) 0.03* SA > PC
Anxiety
disorder/PTSD, n (%) 12 (41) 18 (51) - 0.6 (df =
1) 0.4 -
Eating Disorder, n (%) < 5 < 5 - 0.1 (df =
1) 0.8 -
ADHD, n (%) 9 (32) < 5 - 2.8 (df =
1) 0.09 -
Psychotropic medication
Antidepressant, n (%) 15 (52) 20 (57) - 6.0 (df =
1) 0.5 -
Page 11/27
Suicide
Attempt Patient
Controls Healthy
Controls Group comparison
(N = 29) (N = 35) (N = 32)
Mood Stabilizer, n (%) < 5 < 5 - 0.02 (df
= 1) 0.9 -
Low-Dose
Neuroleptic, n (%) 12 (41) 8 (23) - 1.7 (df =
1) 0.2 -
Benzodiazepine, n (%) < 5 < 5 - 0 (df = 1) 1 -
Stimulant, n (%) 8 (28) < 5 - 6.1 (df =
1) 0.01* SA > PC
Task-related variables
Go-NoGo : Response
Time (ms)
Go Blocks, Mean (sd) 269 (36) 257 (34) 278 (31) 3.4 (df =
2,92) 0.04* PC < HC
NoGo Blocks, Mean
(sd) 301 (50) 284 (33) 310 (34) 4.2 (df =
2,92) 0.02* PC < HC
Go-NoGo : Errors
(Count)
Omission Errors, Go
Blocks 9.8 (10.1) 8.1 (6.2) 10.1 (9.0) 0.5 (df =
2,92) 0.6 -
Omission Errors,
NoGo Blocks 3.5 (4.9) 2.9 (2.6) 3.0 (3.0) 0.3 (df =
2,92) 0.7 -
Commission Errors,
NoGo Blocks 6.7 (4.4) 8.4 (3.4) 6.2 (3.3) 3.7 (df =
2,92) 0.03* PC > HC
Cyberball - Need-
Threat Scale
Belonging, Mean (sd) 12.7 (4.3) 14.1 (3.2) 16.3 (3.9) 6.6 (df =
2,87) 0.002 SA,PC < HC
Self-Esteem, Mean
(sd) 13.1 (4.4) 13.6 (3.9) 18.2 (4.0) 14.5 (df
= 2,87) <
0.001* SA,PC < HC
Signicant Existence,
Mean (sd) 12.8 (5.2) 14.4 (3.5) 17.4 (3.2) 10.0 (df
= 2,87) <
0.001* SA,PC < HC
Sense of Control,
Mean (sd) 10.1 (3.7) 11.0 (2.9) 13.4 (4.0) 7.5 (df =
2,87) <
0.001* SA,PC < HC
Total, Mean (sd) 48.7
(14.8) 53.3
(10.4) 65.5
(12.5) 7.7 (df =
2,87) <
0.001* SA,PC < HC
Page 12/27
Among SA, the average time between the last suicide attempt and the scanning session was 13.5 months
(standard deviation (sd) = 12.0, range=[0.9–42.3]). The average number of suicide attempts was 2.0 (1.6,
[1–7]), and the average age at the rst suicide attempt was 14.3 years (1.3, [10.6–16.6]).
Participants with missing neuroimaging data (N = 8) did not differ from the rest of the sample (N = 96) in
terms of age, sex, or race/ethnicity but not parental education (sTable 2).
3.2 Cyberball Game
Both patient groups had lower post-task NTS scores than HC, suggesting higher emotional impact in
patients in general unrelated to a history of suicide attempt (see Table1). Correlations between NTS
scores and sociodemographic and clinical variables are presented in sTables 3 and 4.
Neuroimaging results for the Cyberball Game are presented in Table2 and Fig.1 (at voxel-correction p <
0.005) and sTable 5 (at p < 0.001). For the inclusion vs. control condition contrast, a signicant group
effect was detected in the left insula. Post-hoc analyses showed lower insular activity in SA compared to
both PC and HC. For the exclusion vs. control condition contrast, a signicant group effect was found in
three clusters located in the left and right inferior frontal gyrus, and the right middle/superior frontal
gyrus. In SA vs. PC, lower activation was found in the right inferior frontal gyrus and higher activation in
the right middle/superior frontal gyrus. In SA vs. HC, lower activation was found in the left and right
inferior frontal gyri. Finally, in PC vs. HC a lower activation was found in the left inferior frontal gyrus and
the right middle/superior frontal gyrus. At a more stringent voxel threshold (p < 0.001), ndings remained
signicant in the left insula (Inclusion contrast) and the right superior frontal gyrus (Exclusion contrast).
Page 13/27
Table 2
Group difference in fMRI activity elicited by the Cyberball Game and Go-NoGo Task (3 contrasts, voxel
threshold p < 0.005; Cluster correction p < 0.05)
Size Peak voxel Statistics
Region Brodmann
Area Side Voxels
(n) Coordinates
(MNI) (F/t-stat) Direction of
effects
Contrast: Inclusion - Control
Insula †BA13 L 629 [-36,-6,8] F = 12.3 SA < PC, SA <
PC, PC = HC
Contrast: Exclusion - Control
Inferior Frontal Gyrus BA45 L 660 [-50,24,14] F = 11.9 SA = PC, SA
< HC, PC < HC
Middle/Superior
Frontal Gyrus †
BA9, BA10 R 593 [34, 40, 32] F = 11.5 SA > PC, SA
= HC, PC < HC
Inferior Frontal Gyrus BA10 R 520 [33, 44, 4] F = 9.9 SA < PC, SA <
HC, PC = HC
Contrast: NoGo vs. Go
Precentral Gyrus †BA6, BA9 R 3,409 [37, 1, 27] F = 15.6 SA > PC, SA >
HC, PC = HC
Precentral Gyrus †BA6 L 1,549 [-47,-6,32] F = 9.2 SA > PC, SA
= HC, PC < HC
Inferior Frontal Gyrus BA46 L 690 [-39,31,8] F = 6.9 SA = PC, SA
> HC, PC > HC
Fusiform Gyrus BA20,BA36 R 512 [46,–35,–27] F = 7.7 SA = PC, SA
> HC, PC = HC
† Cluster signicant at voxel threshold p < 0.001; Cluster correction p < 0.05, cf. sTable 5 for details.
R = right, L = left; SA: Patient with past suicide attempt; PC: Patient controls; HC: Healthy controls
Correlations between signicant brain cluster activity and sociodemographic and clinical variables are
presented in sTables 3 and 4. Left Insula activity in the inclusion-minus contrast correlated signicantly
total score on the NTS (ρ = 0.29, p corrected = 0.04), specically with Belonging (ρ = 0.33, p corrected =
0.008) and Signicant Existence (ρ = 0.34, p corrected = 0.007). Activity exclusion-related clusters did not
correlate with NTS Score (sTable 6).
3.3 Go-NoGo Task
Behavioral performance is presented in Table1. PC showed lower reaction times and more commission
errors than HC. There was no difference between SA versus PC or HC. Correlations between behavioral
performance and sociodemographic and clinical variables are presented in sTables 3 and 4.
Page 14/27
Multivariate analysis, controlling for age, sex, and IQ, indicated group differences for the NoGo vs. Go
contrast in four signicant large clusters (see Table2 and Fig.1 at voxel-correction p < 0.005 and sTable 4
at p < 0.001): the left and right precentral gyri, the left inferior frontal gyrus, and the right fusiform gyrus.
Pairwise post-hoc comparisons showed higher activity in bilateral precentral gyri in SA vs. PC. Increased
activity was also found in SA vs. HC in the left inferior frontal gyrus, the right precentral gyrus, the right
fusiform gyrus. Finally, PC vs. HC had lower activity in the left precentral gyrus and higher activity in the
left inferior frontal gyrus. Sensitivity analysis controlling for medication status, psychiatric diagnoses, or
head motion (average framewise displacement) did not change the pattern of neural activity related to
group effects. At a more stringent voxel threshold (p < 0.001), ndings remained signicant in both
precentral clusters.
Correlations between signicant brain cluster activity and sociodemographic and clinical variables are
presented in sTables 3 and 4. Activity in the left precentral gyrus correlated signicantly with response
times in NoGo Blocks. No other correlation between group-discriminating clusters in the Go-NoGo Task
correlated with behavioral outcomes on the task (sTable 7).
The correlation matrix between activation contrasts in the Cyberball Game and the GoNo-Go Task is
presented in Supplementary Fig.1. Activation in the left insula (Inclusion contrast) correlated negatively
(r=-0.32, p uncorrected = 0.00002) with activity in the right precentral gyrus (GoNo-Go contrast).
3.4 Diagnostic accuracy provided by multimodal task-based
fMRI
Classication performance is reported in Table3 and ROC curves of all compared conditions are shown
in Fig.2. Classication models attempting to discriminate adolescents with past suicidal behaviors from
the whole sample were built hierarchically using sociodemographic, clinical, single, or combined
neuroimaging data. In line with our hypothesis, combining neuroimaging data extracted from signicant
clusters (Go-NoGo: Right and left precentral gyri, left inferior frontal gyrus, fusiform gyrus; Cyberball
Game: Left insula, left inferior frontal gyrus, right middle/superior frontal Gyrus, right inferior frontal
gyrus) from both tasks improved classication performance, both in terms of classication accuracy and
AUC: Classication accuracy increased by approximately 20% and AUC increased by 0.25 compared to
baseline models, indicating a signicant enhancement in diagnostic power. The SVM classier using
combined neuroimaging data without clinical data provided the highest classication accuracy (92.6%)
and highest AUC (0.96). Classication accuracy and AUC were not signicantly different between models
using combined vs. single neuroimaging modalities. Of note, clinical variables (BDI and BIS) did not
signicantly enhance classication on their own, and their inclusion did not enhance models with
neuroimaging features. Finally, the SVM-based classier did not signicantly surpass conventional
regression models.
Page 15/27
Table 3
Classication performance of sociodemographic data, clinical data, and functional neuroimaging
markers (single task or combined tasks) with either a logistic regression or a Support Vector Machine-
based learning model.
Statistical Model (5-fold Cross-Validation)
Variables Logistic Regression Support Vector Machine
Accuracy (%/sd) AUC (sd) Accuracy (%/sd) AUC (sd)
1. Baseline169.1 (6.2) 0.70 (0.09) 70.2 (2.3) 0.50 (0.23)
2. Clinical270.3 (9.57) 0.80 (0.10) 70.3 (8.3) 0.81 (0.10)*
3. GNG382.0 (5.2)* 0.87 (0.04)* 79.8 (5.0)* 0.87 (0.04)*
4. CB483.0 (6.0)* 0.90 (0.06)* 83.0 (6.0)* 0.89 (0.05)*
5 GNG + CB592.6 (2.5)** 0.95 (0.03)** 92.6 (2.5)** 0.96 (0.03)*
6. Clinical + GNG 83.0 (5.0)* 0.91 (0.04)** 80.9 (6.2)* 0.90 (0.05)*
7. Clinical + CB 87.19 (4.35)**†0.93 (0.05)** 89.4 (4.6)**†0.90 (0.06)*
8. Clinical + GNG + CB 86.3 (7.1)*†0.95 (0.05)**†88.4 (9.0)*†0.95 (0.06)*†
Footnotes
:
1. Baseline variables: Age, sex, and IQ
2. BDI and BIS Scores, with baseline variables
3. Go-No-Go fMRI BOLD extracted signals in signicant ROIs, with baseline variables
4. Cyberball fMRI BOLD extracted signals in signicant ROIs, with baseline variables
5. Combined Go-No-Go and Cyberball fMRI BOLD extracted signals in signicant ROIs, with baseline
variables
* p < 0.05 Compared to Baseline model (1)
** p < 0.005 Compared to Baseline model (1)
† p < 0.05 Compared to Clinical model (2)
Abbreviations
: CB: Cyberball Game, AUC: Area Under Curve, GNG: Go-NoGo Task; sd: Standard
Deviation,
4. Discussion
The objective of the current study was to identify task-related fMRI markers associated with past suicide
attempts, assessing two putatively relevant aspects of adolescent suicidality: perception of social
Page 16/27
inclusion/exclusion and cognitive control. In summary, our study revealed two major ndings. First, we
observed a set of neural disturbances during both social inclusion/exclusion and cognitive inhibition
tests in adolescent suicide with past suicide attempt in comparison to both control groups. These
alterations were mainly located in the left insula and prefrontal cortex for social perception and
interaction; and motor and prefrontal cortices for inhibition of action (at a more stringent level of
correction). Interestingly, behavioral performances (self-report of emotional feelings during the Cyberball
Game and reactions times and omission/commission errors at the Go-NoGo task) were similar between
both patient groups suggesting that functional MRI may yield ner properties than neuropsychological
tests and questionnaires to discriminate patients with and without a personal history of suicide attempt.
Second, activity in these signicant brain clusters enhanced the identication of a past history of suicidal
attempt, outperforming clinical and sociodemographic variables.
During the Cyberball Game, adolescent SA displayed distinct neural functional alterations in relation to
both inclusion and exclusion conditions, as compared to PC and HC. In contrast to previous studies
(Harm et al. 2019), we found that the set of anatomical regions distinguishing suicidal adolescents
depended on the valence of social circumstances (i.e., inclusion or exclusion). In the inclusion situation,
SA showed reduced neural activity in the left insula, a region involved in the salience network [80].
Decreased activity in left insula (although more posterior) during inclusion vs. control condition at the
Cyberball Game was also found in adult females with past attempts (Olié et al, 2017). In our study, lower
insular activity was associated with lower feeling of belongingness and having a less meaningful
existence during perceived social exclusion. These two factors are in line with psychological risk factors
associated with suicide [81], somewhat paralleling the concepts of “thwarted belongingness” and
“perceived burdensomeness” found in the Interpersonal Theory of Suicide [27]. Thus, it is possible that
the lower activation of the left insula during inclusion in SA may interfere with the normally reinforcing
feelings associated with inclusive social interactions [82]. These ndings may also reect a lower ability
of SA to feel connected to others. It may be hypothesized that in SA this may both limit their
pleasantness of being with others and also their will to seek help when in diculty. Reduced disclosure of
suicidal ideas has indeed been associated with increased risk of social isolation and suicide attempt in
adolescents [83, 84]. Of note, lower insular activity during inclusion was also correlated with higher
depressive state, suggesting a signicant effect of the negative mood state that may be stronger in
individuals at risk of suicide.
In contrast to inclusion, social exclusion in SA was associated with increased activity in right
middle/superior frontal gyrus (surviving at a more stringent correction level) and lower activity in the right
inferior frontal gyrus compared to PC (and in the left inferior frontal gyrus compared to HC although
activity also seems to be lower in SA than PC). These differences in prefrontal cortex activation point
toward possible diculties in regulating emotions and behaviors [85, 86]. Of note, activity in the left (but
not right) inferior frontal gyrus during exclusion correlated negatively with depressive symptoms,
suggesting that depression might again interfere with regulatory prefrontal control during social
interaction. Previous studies using the Cyberball Game also observed alterations in prefrontal cortex
activation during exclusion in adolescents with a history of suicide attempt or non-suicidal self-injury [39,
Page 17/27
87]. Social processing abnormalities in the prefrontal cortex have also previously been described in SA,
for instance while viewing angry faces in adults [88, 89] and adolescents [22]. Interestingly, while
neuroimaging during exclusion was able to discriminate patients with and without a history of suicidal
acts, this was not the case for the social threat questionnaire, questioning the limits of self-reports.
During the Go-NoGo Task, adolescent SA mainly exhibited greater bilateral activity in the precentral gyri –
primary motor brain regions - compared to PC (surviving more stringent correction). We did not observe
behavioral differences between SA and PC, only between patients and HC. Activity in these motor regions
during the Go-NoGo task did not correlate with impulsivity measured with self-questionnaires. However,
they were negatively correlated with response time in our study, suggesting that increased activity may
lead to slower response times. The Go-NoGo task specically demands inhibition of a prepotent motor
response [68]. Increased activation in SA may therefore reect excessive activity to achieve the same
outcome, and, therefore indirectly inecient functioning. Furthermore, these ndings corroborate an
earlier adolescent study (Pan et al. 2011) which also found that neural activity patterns during response
inhibition discriminated between SA and PC, but that behavioral measures did not. Hence, functional
markers might be more sensitive than neuropsychological measures.
We found that the activities of left insula during inclusion and right precentral gyrus during inhibition
were correlated, suggesting a functional connection between the networks underlying social
perception/interactions and cognitive inhibition. Overall, it could be hypothesized that in situation of
stress and emotional disturbances (e.g., a depressive episode), this inecient functioning of brain
networks encompassing both the left insula and motor and precentral regions may translate into a lower
feeling of social connectedness, lower ability for self-disclosure and help-seeking and inecient
emotional and behavioral regulation facilitating both the emergence of suicidal ideas and acting out.
Recent studies have shown that suicidal behaviors are associated with the dysfunctional connectivity of
various brain regions [90].
Finally, exploratory results from our study indicate that considering functional alterations related to
cognitive and socio-emotional processing signicantly enhanced the identication accuracy of
adolescents with past suicidal behaviors from PC and SA. According to two classication performance
metrics (ROC AUC, classication accuracy), neuroimaging modalities signicantly improved baseline
models that used sociodemographic data with or without clinical variables. It should be noted, however,
that while the most accurate model was provided by the combination of both fMRI modalities without
clinical variables (> 90% accuracy, > 95% AUC), the difference between combined models and single
modalities was marginal. An important caveat is that the diagnostic validity of these functional markers
were tested on the same cohort from which they were derived, which can lead to an overestimation of
their effect [91]. Their combination within this sample might not be as effective as their combined use in
an independent sample, where each neuroimaging modality might separately be less diagnostically
ecacious [92]. We also found that an SVM-based learning method did not outperform conventional
logistic regressions. While non-linear kernels with SVM, or other supervised learning methods such as
decision trees might have provided better t [78], logistic regressions already provided high diagnostic
Page 18/27
utility. Deep learning methods are promising methods to integrate neuroimaging and clinical data to
assess suicide risk, but they require large-size data for training [93].
We must highlight a few limitations of the current study. Firstly, even with a substantial number of
participants with past suicidal behaviors, we may have lacked the statistical power to detect effects at
more conservative levels of correction [94]. Our main results are based on a voxel-wise threshold of p <
0.005, which carries a higher likelihood of false-positives than a typically recommended voxel-wise
threshold of p < 0.001 [95]. However, supplementary analyses at a threshold of p < 0.001 yielded
signicant and concordant results across all contrasts. Besides, these analyses did not account for the
variability of cluster-size threshold across all brain regions. While our strategy to simulate noise volume
with a mixed-model ACF is deemed acceptable, newer methods using randomization and permutation are
considered more accurate at controlling false positive rates [96]. Second, participants had a wide age
range (11–18), which probably introduced developmental variability in our analysis despite controlling for
age. We previously reported structural brain differences across ages in this sample [58]. Third, despite
recruiting in a relatively homogeneous second-line psychiatric clinic, several clinical heterogeneities still
characterized our sample, including features of suicidal behaviors, psychiatric comorbidities, and
environmental factors. These heterogeneities may prevent generalization to different populations. Finally,
while controlling for medication status did not change our ndings, we did not account for more complex
pharmacological effects, such as dose effects, duration of treatment, and interactions.
In conclusion, our study revealed discrete sets of functional alterations associated with suicidal
behaviors during social interactions and cognitive inhibition, two important features that underlie the
development of a suicidal crisis. These alterations may furthermore enhance diagnostic identication of
adolescents with previous suicide attempts. These ndings highlight the complex mechanisms
underlying suicidal behaviors. The capacity of neural measures to predict suicidal behavior beyond
clinical and sociodemographic variables will have to be tested in large-scale longitudinal studies.
Abbreviations
SA: Patient with Past suicide Attempt; PC : Patient Controls; HC : Healthy Controls; df : degree of freedom;
sd : standard deviation; ADHD: Attention Decit Hyperactivity Disorder; DDNOS: Depressive Disorder Not
Otherwise Specied; MDD: Major Depressive Disorder; NSSI: Non-suicidal Self-injury; WISC: Wechsler
Intelligence Scale for Children; WAIS: Wechsler Adult Intelligence Scale.
Declarations
Acknowledgments
This research was nanced byManulife Research Fund in Teen Depression, which supports theManulife
Centre for Breakthroughs in Teen Depression and Suicide Preventionin Montréal, Canada. AJG was
supported theFonds de Recherche du Québec –Santé (FRQS/MSSS Resident Physician Health Research
Page 19/27
Career Training Program). MMC was supported by a Junior 2 Research Scholar Salary from
theFQRS.ML was supported by a Research Chair from FRQS and from a James McGill Professorship.
MCG holds a Canada Research Chair Tier-2 and both JR and MCG are supported by the Fonds de
recherche du Québec – Société et Culture research team on youth suicide prevention.The content is
solely the responsibility of the authors.
We thank Daysi Zentner, Geneviève Laurent, and Léa Perret for their assistance with data collection and
organization. Finally, we thank the participants and their families participating in this study as well as the
clinicians involved with adolescents and their families (Theodora Mikedis, Jean-Chrysostome Zanga,
Didier Blondin-Lavoie).
Conict of interest statement
The authors declare no conict of interest.The funding agencies played no role in the design and
conduct of the study; collection, management, analysis, and interpretation of the data; and preparation,
review, or approval of the manuscript.
References
1. Gordon JA, Avenevoli S, Pearson JL. Suicide Prevention Research Priorities in Health Care. JAMA
Psychiatry. 2020;77:885–886.
2. Patton GC, Sawyer SM, Santelli JS, Ross DA, A R, Allen NB, et al. Our future: a Lancet commission
on adolescent health and wellbeing. The Lancet. 2016;387:2423–2478.
3. Mokdad AH, Forouzanfar MH, Daoud F, Mokdad AA, El Bcheraoui C, Moradi-Lakeh M, et al. Global
burden of diseases, injuries, and risk factors for young people’s health during 1990–2013: a
systematic analysis for the Global Burden of Disease Study 2013. The Lancet. 2016;387:2383–2401.
4. Ruch DA, Sheftall AH, Schlagbaum P, Rausch J, Campo JV, Bridge JA. Trends in Suicide Among
Youth Aged 10 to 19 Years in the United States, 1975 to 2016. JAMA Netw Open. 2019;2:e193886–
e193886.
5. Zulyniak S, Wiens K, Bulloch AGM, Williams JVA, Lukmanji A, Dores AK, et al. Increasing Rates of
Youth and Adolescent Suicide in Canadian Women. Can J Psychiatry. 2021:07067437211017875.
. Brezo J, Paris J, Tremblay R, Vitaro F, Hébert M, Turecki G. Identifying correlates of suicide attempts
in suicidal ideators: a population-based study. Psychological Medicine. 2007;37:1551–1562.
7. Nock MK, Green JG, Hwang I, McLaughlin KA, Sampson NA, Zaslavsky AM, et al. Prevalence,
correlates, and treatment of lifetime suicidal behavior among adolescents: results from the National
Comorbidity Survey Replication Adolescent Supplement. JAMA Psychiatry. 2013;70:300–310.
. Orri M, Scardera S, Perret LC, Bolanis D, Temcheff C, Séguin JR, et al. Mental Health Problems and
Risk of Suicidal Ideation and Attempts in Adolescents. Pediatrics. 2020. 8 June 2020.
https://doi.org/10.1542/peds.2019-3823.
Page 20/27
9. Goldman-Mellor SJ, Caspi A, Harrington H, Hogan S, Nada-Raja S, Poulton R, et al. Suicide attempt in
young people: a signal for long-term health care and social needs. JAMA Psychiatry. 2014;71:119–
127.
10. Geoffroy M-C, Orri M, Girard A, Perret LC, Turecki G. Trajectories of suicide attempts from early
adolescence to emerging adulthood: prospective 11-year follow-up of a Canadian cohort. Psychol
Med. 2020:1–11.
11. Jollant F, Blanc-Brisset I, Cellier M, Ambar Akkaoui M, Tran VC, Hamel J-F, et al. Temporal trends in
calls for suicide attempts to poison control centers in France during the COVID-19 pandemic: a
nationwide study. Eur J Epidemiol. 2022. 30 August 2022. https://doi.org/10.1007/s10654-022-
00907-z.
12. Charpignon M-L, Ontiveros J, Sundaresan S, Puri A, Chandra J, Mandl KD, et al. Evaluation of
Suicides Among US Adolescents During the COVID-19 Pandemic. JAMA Pediatrics. 2022;176:724–
726.
13. Nock MK, Millner AJ, Ross EL, Kennedy CJ, Al-Suwaidi M, Barak-Corren Y, et al. Prediction of Suicide
Attempts Using Clinician Assessment, Patient Self-report, and Electronic Health Records. JAMA Netw
Open. 2022;5:e2144373.
14. Franklin JC, Ribeiro JD, Fox KR, Bentley KH, Kleiman EM, Huang X, et al. Risk factors for suicidal
thoughts and behaviors: A meta-analysis of 50 years of research. Psychological Bulletin.
2017;143:187–232.
15. Oquendo MA, Sullivan GM, Sudol K, Baca-Garcia E, Stanley BH, Sublette ME, et al. Toward a
Biosignature for Suicide. AJP. 2014;171:1259–1277.
1. Mann JJ, Rizk MM. A Brain-Centric Model of Suicidal Behavior. AJP. 2020;177:902–916.
17. Auerbach RP, Pagliaccio D, Allison GO, Alqueza KL, Alonso MF. Neural Correlates Associated With
Suicide and Nonsuicidal Self-injury in Youth. Biological Psychiatry. 2021;89:119–133.
1. Jollant F, Lawrence NL, Olié E, Guillaume S, Courtet P. The suicidal mind and brain: a review of
neuropsychological and neuroimaging studies. World J Biol Psychiatry. 2011;12:319–339.
19. Schmaal L, van Harmelen A-L, Chatzi V, Lippard ETC, Toenders YJ, Averill LA, et al. Imaging suicidal
thoughts and behaviors: a comprehensive review of 2 decades of neuroimaging studies. Mol
Psychiatry. 2020;25:408–427.
20. Campos AI, Thompson PM, Veltman DJ, Pozzi E, van Veltzen LS, Jahanshad N, et al. Brain Correlates
of Suicide Attempt in 18,925 Participants Across 18 International Cohorts. Biol Psychiatry.
2021:S0006-3223(21)01174-4.
21. van Velzen LS, Dauvermann MR, Colic L, Villa LM, Savage HS, Toenders YJ, et al. Structural brain
alterations associated with suicidal thoughts and behaviors in young people: results from 21
international studies from the ENIGMA Suicidal Thoughts and Behaviours consortium. Mol
Psychiatry. 2022:1–11.
22. Pan LA, Hassel S, Segreti AM, Nau SA, Brent DA, Phillips ML. Differential patterns of activity and
functional connectivity in emotion processing neural circuitry to angry and happy faces in
Page 21/27
adolescents with and without suicide attempt. Psychological Medicine. 2013;43:2129–2142.
23. Quevedo K, Ng R, Scott H, Martin J, Smyda G, Keener M, et al. The Neurobiology of Self-Face
Recognition in Depressed Adolescents with Low or High Suicidality. J Abnorm Psychol.
2016;125:1185–1200.
24. Oppenheimer CW, Silk JS, Lee KH, Dahl RE, Forbes E, Ryan N, et al. Suicidal Ideation Among Anxious
Youth: A Preliminary Investigation of the Role of Neural Processing of Social Rejection in Interaction
with Real World Negative Social Experiences. Child Psychiatry Hum Dev. 2020;51:163–173.
25. Baumeister RF. Suicide as escape from self. Psychological Review. 1990;97:90–113.
2. Shneidman ES. Suicide as psychache. J Nerv Ment Dis. 1993;181:145–147.
27. Van Orden KA, Witte TK, Cukrowicz KC, Braithwaite S, Selby EA, Joiner TE. The Interpersonal Theory
of Suicide. Psychol Rev. 2010;117:575–600.
2. Nelson EE, Leibenluft E, McClure EB, Pine DS. The social re-orientation of adolescence: a
neuroscience perspective on the process and its relation to psychopathology. Psychol Med.
2005;35:163–174.
29. Geoffroy M-C, Boivin M, Arseneault L, Renaud J, Perret LC, Turecki G, et al. Childhood trajectories of
peer victimization and prediction of mental health outcomes in midadolescence: a longitudinal
population-based study. CMAJ. 2018;190:E37–E43.
30. Klomek AB, Sourander A, Gould M. The Association of Suicide and Bullying in Childhood to Young
Adulthood: A Review of Cross-Sectional and Longitudinal Research Findings. Can J Psychiatry.
2010;55:282–288.
31. Perret LC, Orri M, Boivin M, Ouellet-Morin I, Denault A-S, Côté SM, et al. Cybervictimization in
adolescence and its association with subsequent suicidal ideation/attempt beyond face-to-face
victimization: a longitudinal population-based study. J Child Psychol Psychiatry. 2020. 3 February
2020. https://doi.org/10.1111/jcpp.13158.
32. Stewart JG, Valeri L, Esposito EC, Auerbach RP. Peer Victimization and Suicidal Thoughts and
Behaviors in Depressed Adolescents. J Abnorm Child Psychol. 2018;46:581–596.
33. Eisenberger NI, Lieberman MD, Williams KD. Does Rejection Hurt? An fMRI Study of Social Exclusion.
Science. 2003;302:290–292.
34. Cacioppo S, Frum C, Asp E, Weiss RM, Lewis JW, Cacioppo JT. A quantitative meta-analysis of
functional imaging studies of social rejection. Sci Rep. 2013;3:2027.
35. Williams KD, Jarvis B. Cyberball: A program for use in research on interpersonal ostracism and
acceptance. Behavior Research Methods. 2006;38:174–180.
3. Rotge J-Y, Lemogne C, Hinfray S, Huguet P, Grynszpan O, Tartour E, et al. A meta-analysis of the
anterior cingulate contribution to social pain. Soc Cogn Affect Neurosci. 2015;10:19–27.
37. Olié E, Jollant F, Deverdun J, Champeur NM de, Cyprien F, Bars EL, et al. The experience of social
exclusion in women with a history of suicidal acts: a neuroimaging study. Sci Rep. 2017;7:1–8.
Page 22/27
3. Klonsky ED, May AM, Glenn CR. The relationship between nonsuicidal self-injury and attempted
suicide: converging evidence from four samples. J Abnorm Psychol. 2013;122:231–237.
39. Harms MB, Casement MD, Teoh JY, Ruiz S, Scott H, Wedan R, et al. Adolescent suicide attempts and
ideation are linked to brain function during peer interactions. Psychiatry Research: Neuroimaging.
2019;289:1–9.
40. Huber RS, Hodgson R, Yurgelun-Todd DA. A qualitative systematic review of suicide behavior using
the cognitive systems domain of the research domain criteria (RDoC) framework. Psychiatry
Research. 2019;282:112589.
41. Richard-Devantoy S, Berlim MT, Jollant F. A meta-analysis of neuropsychological markers of
vulnerability to suicidal behavior in mood disorders. Psychological Medicine. 2014;44:1663–1673.
42. Saffer BY, Klonsky ED. Do neurocognitive abilities distinguish suicide attempters from suicide
ideators? A systematic review of an emerging research area. Clinical Psychology: Science and
Practice. 2018;25:e12227.
43. Ochsner KN, Gross JJ. The cognitive control of emotion. Trends in Cognitive Sciences. 2005;9:242–
249.
44. Gifuni AJ, Perret LC, Lacourse E, Geoffroy M-C, Mbekou V, Jollant F, et al. Decision-making and
cognitive control in adolescent suicidal behaviors: a qualitative systematic review of the literature.
Eur Child Adolesc Psychiatry. 2020. 9 May 2020. https://doi.org/10.1007/s00787-020-01550-3.
45. Witte TK, Merrill KA, Stellrecht NE, Bernert RA, Hollar DL, Schatschneider C, et al. “Impulsive” youth
suicide attempters are not necessarily all that impulsive. Journal of Affective Disorders.
2008;107:107–116.
4. Larsen B, Luna B. Adolescence as a neurobiological critical period for the development of higher-
order cognition. Neuroscience & Biobehavioral Reviews. 2018;94:179–195.
47. Luna B, Marek S, Larsen B, Tervo-Clemmens B, Chahal R. An Integrative Model of the Maturation of
Cognitive Control. Annu Rev Neurosci. 2015;38:151–170.
4. Casey BJ. Beyond Simple Models of Self-Control to Circuit-Based Accounts of Adolescent Behavior.
Annu Rev Psychol. 2015;66:295–319.
49. Spear LP. The adolescent brain and age-related behavioral manifestations. Neuroscience &
Biobehavioral Reviews. 2000;24:417–463.
50. Chikazoe J. Localizing performance of go/no-go tasks to prefrontal cortical subregions. Current
Opinion in Psychiatry. 2010;23:267–272.
51. Casey BJ, Trainor RJ, Orendi JL, Schubert AB, Nystrom LE, Giedd JN, et al. A Developmental
Functional MRI Study of Prefrontal Activation during Performance of a Go-No-Go Task. Journal of
Cognitive Neuroscience. 1997;9:835–847.
52. Erickson K, Drevets WC, Clark L, Cannon DM, Bain EE, Zarate CA, et al. Mood-Congruent Bias in
Affective Go/No-Go Performance of Unmedicated Patients With Major Depressive Disorder. AJP.
2005;162:2171–2173.
Page 23/27
53. Pan LA, Batezati-Alves SC, Almeida JRC, Segreti A, Akkal D, Hassel S, et al. Dissociable patterns of
neural activity during response inhibition in depressed adolescents with and without suicidal
behavior. J Am Acad Child Adolesc Psychiatry. 2011;50:602–611.e3.
54. Posner K, Oquendo MA, Gould M, Stanley B, Davies M. Columbia Classication Algorithm of Suicide
Assessment (C-CASA): Classication of Suicidal Events in the FDA’s Pediatric Suicidal Risk Analysis
of Antidepressants. Am J Psychiatry. 2007;164:1035–1043.
55. Diagnostic and statistical manual of mental disorders: DSM-IV. Fourth edition. Washington, DC :
American Psychiatric Association, [1994] ©1994; 1994.
5. Ding Y, Pereira F, Hoehne A, Beaulieu M-M, Lepage M, Turecki G, et al. Altered brain processing of
decision-making in healthy rst-degree biological relatives of suicide completers. Mol Psychiatry.
2017;22:1149–1154.
57. Jollant F, Wagner G, Richard-Devantoy S, Köhler S, Bär K-J, Turecki G, et al. Neuroimaging-informed
phenotypes of suicidal behavior: a family history of suicide and the use of a violent suicidal means.
Translational Psychiatry. 2018;8:1–10.
5. Gifuni AJ, Chakravarty MM, Lepage M, Ho TC, Geoffroy M-C, Lacourse E, et al. Brain cortical and
subcortical morphology in adolescents with depression and a history of suicide attempt. J
Psychiatry Neurosci. 2021;46:E347–E357.
59. Kaufman J, Birmaher B, Brent DA, Ryan ND, Rao U. K-SADS-PL. J Am Acad Child Adolesc Psychiatry.
2000;39:1208.
0. Kaufman J, Birmaher B, Brent D, Rao U, et al. Schedule for Affective Disorders and Schizophrenia for
School-Age Children-Present and Lifetime version (K-SADS-PL): Initial reliability and validity data.
Journal of the American Academy of Child & Adolescent Psychiatry. 1997;36:980–988.
1. Beck AT, Steer RA, Brown G. Manual for the Beck Depression Inventory-II. San Antonio, TX:
Psychological Corporation; 1996.
2. Wechsler D. Wechsler Intelligence Scale for Children, Fourth Edition (WISC-IV). San Antonio, TX: The
Psychological Corporation; 2003.
3. Wechsler D. Wechsler Adult Intelligence Scale. 4th edition. San Antonio, TX: Pearson Assessment;
2008.
4. Patton JH, Stanford MS, Barratt ES. Factor structure of the Barratt Impulsiveness Scale. Journal of
Clinical Psychology. 1995;51:768–774.
5. Williams KD, Cheung CKT, Choi W. Cyberostracism: Effects of being ignored over the Internet. Journal
of Personality and Social Psychology. 2000;79:748–762.
. Gerber JP, Chang S-H, Reimel H. Construct validity of Williams’ ostracism needs threat scale.
Personality and Individual Differences. 2017;115:50–53.
7. Zadro L, Williams KD, Richardson R. How low can you go? Ostracism by a computer is sucient to
lower self-reported levels of belonging, control, self-esteem, and meaningful existence. Journal of
Experimental Social Psychology. 2004;40:560–567.
Page 24/27
. Simmonds DJ, Pekar JJ, Mostofsky SH. Meta-analysis of Go/No-go tasks demonstrating that fMRI
activation associated with response inhibition is task-dependent. Neuropsychologia. 2008;46:224–
232.
9. Criaud M, Boulinguez P. Have we been asking the right questions when assessing response inhibition
in go/no-go tasks with fMRI? A meta-analysis and critical review. Neurosci Biobehav Rev.
2013;37:11–23.
70. Esteban O, Markiewicz CJ, Goncalves M, DuPre E, Kent JD, Salo T, et al. FMRIPrep. Zenodo; 2018.
71. Esteban O, Markiewicz CJ, Blair RW, Moodie CA, Isik AI, Erramuzpe A, et al. fMRIPrep: a robust
preprocessing pipeline for functional MRI. Nature Methods. 2019;16:111–116.
72. Esteban O, Ciric R, Finc K, Blair RW, Markiewicz CJ, Moodie CA, et al. Analysis of task-based
functional MRI data preprocessed with fMRIPrep. Nat Protoc. 2020;15:2186–2202.
73. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R
Foundation for Statistical Computing; 2019.
74. RStudio Team. RStudio: Integrated Development Environment for R. Boston, MA: RStudio, Inc.; 2016.
75. Chen G, Adleman NE, Saad ZS, Leibenluft E, Cox RW. Applications of multivariate modeling to
neuroimaging group analysis: a comprehensive alternative to univariate general linear model.
Neuroimage. 2014;99:571–588.
7. Gong G, Rosa-Neto P, Carbonell F, Chen ZJ, He Y, Evans AC. Age- and Gender-Related Differences in
the Cortical Anatomical Network. J Neurosci. 2009;29:15684–15693.
77. Graham S, Jiang J, Manning V, Nejad AB, Zhisheng K, Salleh SR, et al. IQ-related fMRI differences
during cognitive set shifting. Cereb Cortex. 2010;20:641–649.
7. Boudreaux ED, Rundensteiner E, Liu F, Wang B, Larkin C, Agu E, et al. Applying Machine Learning
Approaches to Suicide Prediction Using Healthcare Data: Overview and Future Directions. Frontiers in
Psychiatry. 2021;12.
79. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine
Learning in Python. Journal of Machine Learning Research. 2011;12:2825–2830.
0. Menon V, Uddin LQ. Saliency, switching, attention and control: a network model of insula function.
Brain Struct Funct. 2010;214:655–667.
1. O’Connor RC, Nock MK. The psychology of suicidal behaviour. Lancet Psychiatry. 2014;1:73–85.
2. Eisenberger NI. The pain of social disconnection: examining the shared neural underpinnings of
physical and social pain. Nature Reviews Neuroscience. 2012;13:421–434.
3. Apter A, Horesh N, Gothelf D, Gra H, Lepkifker E. Relationship between self-disclosure and serious
suicidal behavior. Compr Psychiatry. 2001;42:70–75.
4. Husky MM, Zablith I, Alvarez Fernandez V, Kovess-Masfety V. Factors associated with suicidal
ideation disclosure: Results from a large population-based study. J Affect Disord. 2016;205:36–43.
5. Dixon ML, Thiruchselvam R, Todd R, Christoff K. Emotion and the prefrontal cortex: An integrative
review. Psychol Bull. 2017;143:1033–1081.
Page 25/27
. Mitchell DGV. The nexus between decision making and emotion regulation: A review of convergent
neurocognitive substrates. Behavioural Brain Research. 2011;217:215–231.
7. Groschwitz RC, Plener PL, Groen G, Bonenberger M, Abler B. Differential neural processing of social
exclusion in adolescents with non-suicidal self-injury: An fMRI study. Psychiatry Res Neuroimaging.
2016;255:43–49.
. Jollant F, Lawrence NS, Giampietro V, Brammer MJ, Fullana MA, Drapier D, et al. Orbitofrontal Cortex
Response to Angry Faces in Men With Histories of Suicide Attempts. AJP. 2008;165:740–748.
9. Olié E, Ding Y, Le Bars E, de Champeur NM, Mura T, Bonafé A, et al. Processing of decision-making
and social threat in patients with history of suicidal attempt: A neuroimaging replication study.
Psychiatry Res. 2015;234:369–377.
90. Wagner G, Li M, Sacchet MD, Richard-Devantoy S, Turecki G, Bär K-J, et al. Functional network
alterations differently associated with suicidal ideas and acts in depressed patients: an indirect
support to the transition model. Transl Psychiatry. 2021;11:100.
91. Poldrack RA, Huckins G, Varoquaux G. Establishment of Best Practices for Evidence for Prediction: A
Review. JAMA Psychiatry. 2020;77:534–540.
92. Dinga R, Schmaal L, Penninx BWJH, van Tol MJ, Veltman DJ, van Velzen L, et al. Evaluating the
evidence for biotypes of depression: Methodological replication and extension of Drysdale et al.
(2017). NeuroImage: Clinical. 2019;22:101796.
93. Roy A, Nikolitch K, McGinn R, Jinah S, Klement W, Kaminsky ZA. A machine learning approach
predicts future risk to suicidal ideation from social media data. Npj Digit Med. 2020;3:1–12.
94. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2 edition. Hillsdale, N.J: Routledge;
1988.
95. Cox RW, Chen G, Glen DR, Reynolds RC, Taylor PA. FMRI Clustering in AFNI: False-Positive Rates
Redux. Brain Connect. 2017;7:152–171.
9. Cox RW. Equitable Thresholding and Clustering: A Novel Method for Functional Magnetic Resonance
Imaging Clustering in AFNI. Brain Connect. 2019;9:529–538.
Figures
Page 26/27
Figure 1
Group Differences on the Go-NoGo Task and Cyberball Game fMRI (voxel threshold p<0.005, Cluster
corrected p<0.05)
Caption:
Page 27/27
A. Go-NoGo Task (NoGo vs Go Contrast)
B. Cyberball Game (Inclusion vs Control Contrast)
C. Cyberball Game (Exclusion vs Control Contrast)
Figure 2
Diagnostic accuracy provided by Socio-Demographic Data, Clinical Variables, and Neural Contrast
Extracted from Signicant Clusters in the Go-NoGo Task and the Cyberball Game fMRI - Receiver Operant
Curves (ROC) Based on k=5 Stratied Cross-Validation
Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.
Supplements.docx