Cross-sectional and longitudinal associations of anxiety and irritability with adolescents’ neural responses to cognitive conflict

Abstract

Background.

Psychiatric symptoms are commonly comorbid in childhood. The ability to disentangle unique and shared correlates of comorbid symptoms facilitates personalized medicine. Cognitive control is implicated broadly in psychopathology, including in pediatric disorders characterized by anxiety and irritability. To disentangle cognitive control correlates of anxiety versus irritability, the current study leverages both cross-sectional and longitudinal data from early childhood into adolescence.

Methods.

89 participants were recruited from a large longitudinal research study on early life temperament. The current study investigated associations of developmental trajectories of anxiety and irritability symptoms (from ages 2 to 15), as well as associations of anxiety and irritability symptoms measured cross-sectionally at age-15, with neural substrates of conflict and error processing assessed at age-15 using the Flanker Task.

Results.

Results of whole-brain multivariate linear models revealed that age-15 anxiety was uniquely associated with decreased neural response to conflict, across multiple regions implicated in attentional control and conflict adaptation. Conversely, age-15 irritability was uniquely associated with increased neural response to conflict in regions implicated in response inhibition. Developmental trajectories of anxiety and irritability interacted in relation to neural responses to both error and conflict.

Conclusions.

Our findings suggest that neural correlates of conflict processing may relate uniquely to anxiety and irritability. Continued cross-symptom research on the neural correlates of cognitive control could stimulate advances in individualized treatment for anxiety and irritability during child and adolescent development.

INTRODUCTION

Anxiety and irritability are two common, often co-occurring symptom clusters across a wide range of pediatric psychiatric disorders (1–3). Given the comorbidity of anxiety and irritability, it is essential for personalized medicine to identify unique and shared mechanisms to guide treatment for patients who present with one or both symptoms. Deficits in cognitive control, the engagement of cognitive resources to regulate goal-directed behavior, have been implicated in psychopathology broadly (4) and anxiety and irritability specifically (5,6). The current study is novel in focusing on the specificity of cognitive control deficits, and their underlying neural mechanisms, in anxiety versus irritability. Further, studies of cognitive control and psychiatric symptoms in childhood and adolescence have relied largely on cross-sectional designs. The current study leverages both cross-sectional and longitudinal assessments of anxiety and irritability from early childhood into adolescence, a critical timeframe for the development of cognitive control (6).

Cognitive control refers to a collection of high-level cognitive processes, including the ability to inhibit prepotent responses, monitor for errors, and update goal-congruent behaviors (7). Studies in anxiety report mixed results; some find increased cognitive control (or over-control) manifesting as increased neural response to errors (8–12) and increased reactive control (13,14). Others find impaired cognitive control (e.g., increased interference from conflicting information; 15,16). This inconsistency highlights the need to focus on specific aspects of cognitive control, error and conflict processing, that may manifest differential relations with anxiety.

The literature on irritability and cognitive control is comparatively sparse. While many studies implicate cognitive control deficits in associated externalizing disorders (17), irritability is not consistently linked to impaired performance on cognitive control tasks (18–21). However, these same studies have found that cognitive control (i.e., inhibition of conflicting information and engaging attentional control), more strongly engages the lateral prefrontal cortex in irritable compared to non-irritable youth, perhaps reflecting inefficient processing.

The development of irritability and anxiety symptoms are highly associated, with evidence that high levels of both symptoms across time predict poorer outcomes including increased aggression (22) and depression (1,23,24). Furthermore, the relative timing of early-life anxiety and irritability may be critical for uncovering developmental biological risk factors. For example, neural responses to errors have been shown to differentially moderate the development of anxiety and irritability, such that heighted neural response to errors is predictive of later-life anxiety and internalizing disorders, whereas a reduced response is predictive of later-life irritability and externalizing disorders (8,25). Together, these findings stress the importance of contrasting both ongoing symptoms as well as anxiety and irritability trajectories in relation to adolescents’ neural function.

Recognition of the high co-occurrence of anxiety and irritability has generated a growing body of research aimed at identifying their shared and unique mechanisms (8,26–28). These studies have found that anxiety and irritability often have divergent effects on underlying mechanisms and critically, the presence of one symptom can moderate effects of the other. For example, anxiety and irritability were associated with distinct neural patterns of functional activation and connectivity when directing attention away from threat (26). Additionally, a significant interaction emerged between anxiety and irritability, such that the level of comorbid irritability moderated the relations between heightened anxiety and amygdala-medial prefrontal cortex connectivity when viewing angry faces (27). These studies demonstrate the importance of considering the effects of anxiety and irritability jointly.

The current study investigates specificity in cognitive control deficits related to irritability and anxiety. At age-15, youth completed a Flanker Task while undergoing functional magnetic resonance imaging (fMRI). First, we leveraged a longitudinal study to model youth’s developmental trajectories of anxiety and irritability from early childhood into adolescence (from ages 2 to 15). Second, using latent variable models we derived measures of adolescent anxiety and irritability that were assessed at age-15 with the acquisition of scan data. This approach allowed us to examine both developmental trajectories and symptoms at age-15 of anxiety and irritability in relation to cognitive control. We assessed neural substrates of two specific cognitive control processes during adolescence: conflict detection and error processing. We predicted that age-15 anxiety would be uniquely associated with alterations in neural response to error, whereas age-15 irritability would be uniquely associated with alterations in neural response to conflict. Next, we conducted exploratory analyses investigating whether, above and beyond current symptoms, developmental trajectories of anxiety and irritability would be associated with aberrant neural response to error and conflict.

METHODS

Participants

The present study recruited from two cohorts enrolled in a large longitudinal research study on early life temperament. One cohort was enrolled in the study at age four months of age based on temperamental reactivity (29) and the other cohort was enrolled at age two years based on community sampling (30,31). Data for the current study were collected at assessment timepoints at target ages 2, 3, 5, 7, 9, 12, and 15-years. The target ages for each timepoint reflect the approximate age at which participants were brought to the lab for reassessments, though actual age at assessments varied. The 15-year timepoint included 101 participants who were invited to participate in an MRI scan. Following initial screening, four participants declined and four participants we excluded due to medication use (n=3) and fMRI contraindications (i.e., metal in body; n=1). Of the 93 participants for whom imaging data were acquired, one was excluded due to scanner malfunction and three were excluded following acquisition (2 due to inability to complete task; one due to correct responses <70%). Thus, the final sample comprised 89 participants (age at scan: M[SD]=16.31[0.53]; % male=53.94%; Table 1, Supplementary Table 1). Psychiatric diagnoses were assessed using the Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime version (KSADS-PL; 32), and reviewed by a Board-certified child and adolescent psychiatrist (author DSP). Two participants reported use of stimulant medication during at least previous one timepoint. No participants were taking medications at the time of the 15-year scan. Written informed consent and assent were obtained from parents and youth, respectively. Participants received monetary compensation. All study procedures were approved by the University of Maryland-College Park and National Institute of Mental Health Institutional Review Boards.

Table 1.

Sample Characteristics

Variable	M (SD) or No. (%)
Age (years)	16.31 (0.5)
Sex (male)	48 (53.9%)
IQ	115.6 (14.4)
Recruitment Cohort
Enrollment at 4-months	50 (56.2%)
Enrollment at 2-years	39 (43.8%)
Race
Caucasian	60 (67.4%)
Black or African American	14 (15.7%)
Asian	1 (1.1%)
Other	6 (6.7%)
Unknown	8 (9.0%)
Ethnicity
Latino or Hispanic	8 (9.0%)
Not Latino or Hispanic	81 (91.0%)
Socioeconomic Status
Highest Maternal Education Level
Graduate professional degree (Masters or above)	37 (41.6%)
4-year College graduate	33 (37.1%)
High School graduate	16 (18.0%)
Other	3 (3.4%)
Diagnosis
Lifetime ADHD	13 (15.3%)
Lifetime Anxiety Disorder	26 (30.6%)
Lifetime Unipolar Depressive Disorder	13 (15.3%)
Lifetime PTSD	2 (2.4%)
Lifetime Disruptive Conduct	2 (2.4%)
Lifetime DMDD	1 (1.2%)
Lifetime Other	6 (7.1%)

Note. Diagnoses (N = 85): Comorbidities are not uniquely coded, such that the values do not represent individual cases. Abbreviations: Attention-Deficit/Hyperactivity Disorder (ADHD), Post-Traumatic Stress Disorder (PTSD), Disruptive Mood Dysregulation Disorder (DMDD).

Choice of Primary Measures

Child Behavior Checklist.

Anxiety and irritability were assessed at ages 2, 3, 4, 5, 7, 9, 12 and 15-years using the maternal-report Child Behavior Checklist (CBCL; 33) The CBCL queries a range of child symptomatology in the preceding six months, with each item rated on a scale from 0 (not true) to 2 (often true or very true). Irritability was assessed using three previously-validated items (“temper tantrums or hot temper,” “stubborn, sullen, or irritable,” “sudden changes in mood or feelings”; (8,23,34) Anxiety was assessed using the five items from the CBCL DSM Anxiety subscale that were consistent in wording across timepoints (“clings to adults or too dependent,” “fears certain animals, situations, or places,” “nervous, highstrung, or tense,” “too fearful or anxious,” “worries”). Mean scores at each timepoint were calculated separately for anxiety and irritability, such that mean scores could range from 0 to 2, with higher scores reflect greater symptoms (Table 2; Supplementary Figure 1).

Table 2.

Descriptive statistics for mean CBCL anxiety and irritability scores at each timepoint.

	Timepoint
	2 yr	3 yr	4 yr	5 yr	7 yr	9 yr	12 yr	15 yr
CBCL Anxiety
N(%)	57 (64.0%)	76 (85.4%)	76 (85.4%)	77 (86.5%)	76 (85.4%)	80 (89.9%)	80 (89.9%)	89 (100%)
Score M(SD)	0.24 (0.20)	0.29 (0.25)	0.28 (0.28)	0.26 (0.28)	0.22 (0.24)	0.27 (0.30)	0.18 (0.25)	0.21 (0.32)
CBCL Irritability
N(%)	57 (64.0%)	76 (85.4%)	76 (85.4%)	77 (86.5%)	76 (85.4%)	79 (88.8%)	80 (89.9%)	88 (98.9%)
Score M(SD)	0.53 (0.45)	0.50 (0.46)	0.48 (0.44)	0.38 (0.42)	0.31 (0.40)	0.29 (0.38)	0.31 (0.45)	0.35 (0.37)

Note. CBCL=Child Behavior Checklist

Screen for Child Anxiety Related Emotional Disorders.

At the 15-year timepoint, the Screen for Child Anxiety Related Emotional Disorders parent- and youth-report (SCARED; 35) were administered to further assess anxiety. Total scores were computed for each questionnaire, with higher scores reflecting greater symptoms (parent-report: M[SD]=8.84[8.81], 0–34; youth-report M[SD]=12.51[12.12], 0–71).

Affective Reactivity Index.

Also at the 15-year timepoint, the Affective Reactivity Index parent- and youth-report (ARI; 36) were administered to further assess irritability. Total scores were computed for each questionnaire, with higher scores reflecting greater symptoms (parent-report: M[SD]=1.03[2.01], 0–10; child-report: M[SD]=1.23[2.18], 0–12).

Symptom Data Analysis

Symptom Trajectories.

To quantify symptom trajectories across ages 2 to 15, we used multilevel modeling (MLM; HLM version 8.0; Supplementary Text 1). Results of the MLMs indicated significant group-level effects of time, in which both anxiety and irritability symptoms decreased with age (anxiety: b=−0.005, p=0.048; irritability: b=−0.013, p<.001; Supplementary Table 2). Individual participants’ intercepts and slopes for anxiety and irritability were extracted from the models. Slopes quantified linear symptom trajectories across timepoints 2 to 15-years: positive values indicate increasing symptoms, negative values indicate decreasing symptoms, and a value of zero indicates no change in (or stable) symptoms across time. Intercepts were set at age-15 in order to estimate symptoms at the time of the age-15 scan. The resulting individual-level estimates were used in the next step quantifying symptoms at age-15.

Symptoms at Age-15.

To quantify symptom levels at the 15-year scan, we used confirmatory factor analysis (CFA). Separate models estimated latent variables for anxiety and irritability, each using three observed variables with their loadings on the latent variable set as equal. For anxiety, each participant’s indices were the intercept for CBCL anxiety, parent-report SCARED score, and youth-report SCARED score at the 15-year timepoint. For irritability, indices were the intercept for CBCL irritability, parent-report ARI score, and youth-report ARI score at the 15-year timepoint (Supplementary Figure 2).

Results of the CFAs indicated adequate fit of the models for both anxiety (CFI=0.917, TLI=0.876, RMSEA=0.19, CI₉₀=0.066–0.326) and irritability (CFI=0.869, TLI=0.804, RMSEA=0.12, CI₉₀=0–0.267). All measured variables loaded significantly on the latent variables (ps<.001). Individual participants’ factor scores for anxiety and irritability were extracted. Factor scores thus captured the shared variance among the symptom measures at age-15. The association between the factor scores for anxiety and irritability at age-15 was significant, r(89)=0.428, p<.001.

fMRI Task

Design.

Participants completed a modified Flanker Task (Supplementary Figure 3; 12) during a single fMRI session at the 15-year timepoint. The task consisted of four 6-minute runs divided into three blocks. Participants were presented with a series of five side-by-side arrows centered on the screen, and were instructed to press a button to indicate the direction of the central arrow as quickly as possible. Trials were categorized as either congruent or incongruent based on the direction of the flanking arrows. On congruent trials, the flanking arrows pointed in the same direction as the central arrow. On incongruent trials, the flanking arrows pointed in the opposite direction from the central arrow. In between blocks, participants received feedback based on task accuracy (<75%: “be more accurate”; 75–90%: “good job”; >90%: “respond faster”). Consistent with prior work, this feedback was included to balance maintenance of accuracy with maximization of errors (12).

Procedures.

Images were collected on a 3T MR750 General Electric Scanner (Waukesha, Wisconsin, USA) using a 32-channel head coil. 170 whole brain T*2 weighted echo-planar images were collected across four functional runs (TR=2000ms, TE=25, flip angle=60, field of view=96×96, slices=42/axial/3mm). For all participants structural images were collected using a structural magnetization-prepared rapid acquisition gradient echo (MPRAGE) sequence (TR=7.66, TE=3.42, flip angle=7, field of view=256×256, slices=176/sagittal/1mm).

All imaging analyses were conducted using Analysis of Functional Neuroimages (AFNI). All images were first preprocessed using afni.proc.py. Images were de-spiked, slice time corrected, co-registered, spatial smoothed with a 6mm full-width half maximum smoothing kernel, and warped to standard spaces. For motion correction, TR pairs with a Euclidean norm motion derivative that exceeded 1mm were censored (movement post-censoring: M[SD]=0.06[0.04]). Finally, at the subject level, a general linear model with 6 regressors time-locked to the stimulus onset was used to model trial condition (congruent or incongruent trials) and response type (correct, commission error, or omission error). Nuisance regressors included motion displacement along the x, y, and z axes as well as rotational movement (roll, pitch, and yaw).

Imaging Analyses

Group-level multivariate linear models were conducted to examine whole-brain associations with symptoms at age-15 and symptom trajectories (ages 2–15) using the 3dMVM function in AFNI. Four separate models were conducted. Output maps for all analyses were masked to include only gray matter voxels where signal was present in more than 90% of the participants. A threshold for significance was determined using the 3dClustsim program in AFNI. 10,000 Monte Carlo simulations were run to determine cluster extent threshold at an uncorrected threshold of p<.005. Two-sided thresholding was used to maximize accuracy. Results indicated that clusters >89 contiguous voxels (1391 mm³) met criteria for whole-brain significance (p<.05). For each significant cluster, average BOLD activity was extracted to decompose the interactions between symptom measures and condition-related differences in BOLD signal. To address the potential influence of outliers, all post hoc analyses were repeated after winsorizing all independent and dependent variables such that values >2.5 SD from the mean were set to the respective 2.5 SD value. This affected 2.75% of all data points and did not change the significance of any effects.

Neural response to conflict.

Consistent with previous work (12), neural response to conflict was operationalized using a within-subjects factor, conflict, that contrasted correct responses on incongruent trials vs. correct responses on congruent trials. Sex and average motion following censoring were covariates in both models. Model 1a: Symptoms at Age-15. Between-subject predictors of interest were the 15-year timepoint anxiety symptoms, irritability symptoms, and their interaction. Model 1b: Symptom trajectories. Between-subject predictors of interest were anxiety symptom trajectory, irritability symptom trajectory, and their interaction. The 15-year timepoint anxiety symptoms and irritability symptoms were retained from Model 1a as covariates.

Neural response to error.

Consistent with previous work (12), neural response to error was operationalized using a within-subjects factor, error, that contrasted commission errors on incongruent trials vs. correct responses on incongruent trials. Sex, average motion following censoring, and the number of commission error trials were covariates in both models. To ensure adequate power and reliability of BOLD response (12), participants with <20 errors of commission on incongruent trials were excluded from analysis (n=8). Model 2a: Symptoms at Age-15. Between-subject predictors of interest were 15-year timepoint anxiety symptoms, irritability symptoms, and their interaction. Model 2b: Symptom trajectories. Between-subject predictors of interest were anxiety symptom trajectory, irritability symptom trajectory, and their interaction. 15-year timepoint anxiety symptoms and irritability symptoms were retained from Model 2a as covariates.

RESULTS

Neural Response to Conflict

Model 1a: Age-15 symptoms.

Significant symptom-by-conflict interactions emerged for both age-15 anxiety and irritability (Table 3). Higher anxiety was associated with widespread decreased activity in response to conflict (incongruent vs. congruent trials) across multiple regions implicated in attentional control demand and conflict adaptation (37,38). Regions included the left insula (Figure 1), left middle occipital gyrus, left postcentral gyrus, bilateral fusiform gyrus, right superior occipital gyrus, left middle frontal gyrus, right postcentral gyrus, and left supramarginal gyrus.

Table 3.

Significant associations of neural activity with symptoms at age-15 and symptom trajectories

Age–15 Anxiety by Conflict
x	y	z	Size	t	p	Location
−9	−81	36	11766	−5.516	<.001	Left Middle Occipital Gyrus, Left Superior Occipital Gyrus, Left Cuneus, Left Middle Temporal Gyrus
−14	−16	61	6859	−4.694	<.001	Left Postcentral Gyrus, Left Superior Parietal Lobule, Left SMA
−29	−61	−6	4578	−4.928	<.001	Left Fusiform Gyrus, Left Lingual Gyrus, Left Inferior Occipital Gyrus
29	−51	−6	4016	−4.667	<.001	Right Fusiform Gyrus, Right Lingual Gyrus
−19	14	4	2938	−4.558	<.001	Left Insula Lobe, Left Caudate Nucleus, Left Putamen
14	−84	26	2938	−4.284	<.001	Right Superior Occipital Gyrus, Right Cuneus
64	−14	26	2359	−4.514	<.001	Right Supramarginal Gyrus, Right Postcentral Gyrus
−34	31	21	2016	−4.233	<.001	Left Middle Frontal Gyrus, Left Inferior Frontal Gyrus
16	−19	64	2000	−4.370	<.001	Right Postcentral Gyrus, Right Superior Frontal Gyrus, Right Precentral Gyrus
−56	−36	34	1859	−4.398	<.001	Left Supramarginal Gyrus, Left Inferior Parietal Lobule
Age–15 Irritability by Conflict
x	y	z	Size	t	p	Location
−49	−59	4	17969	5.729	<.001	Left Fusiform Gyrus, Left Middle Occipital Gyrus, Left Middle Temporal Gyrus, Left Inferior Parietal
41	−49	−16	15344	6.038	<.001	Gyrus Right Fusiform Gyrus, Right Precuneus, Right Superior Occipital Gyrus
−24	−29	56	3719	4.481	<.001	Left Postcentral Gyrus, Left Precentral Gyrus
44	−1	29	1703	4.169	<.001	Right Precentral Gyrus, Right Inferior Frontal Gyrus
−41	11	24	1438	4.451	<.001	Left Inferior Frontal Gyrus, Left Precentral Gyrus
Anxiety Trajectory by Irritability Trajectory by Conflict
x	y	z	Size	t	p	Location
−6	−99	9	4125	5.493	<.001	Left Cuneus, Right Precuneus, Left Precuneus, Left Superior Occipital Gyrus
Anxiety Trajectory by Irritability Trajectory by Error
x	y	z	Size	t	p	Location
−4	−94	19	1766	−4.412	<.001	Left Cuneus, Left Superior Occipital Gyrus, Left Middle Occipital Gyrus

Note. x,y,z= peak talairach coordinates (LPI); size reported in mm³; Location lists regions in descending order based on proportion overlap with cluster.

Figure 1.

Partial regression plot for the effect of anxiety symptoms at age 15 on neural activity during conflict in the left insula.

Conversely, higher irritability was associated with increased activity in response to conflict (incongruent vs. congruent trials) across multiple regions implicated in response inhibition (39). Regions included the bilateral inferior frontal gyrus (IFG; Figure 2), bilateral fusiform gyrus, left middle temporal gyrus, and left postcentral gyrus.

Figure 2.

Partial regression plots for the effect of irritability symptoms at age 15 with neural activity during conflict in the bilateral inferior frontal gyrus.

There were no significant three-way interactions among age-15 anxiety, age-15 irritability, and conflict condition.

Model 1b: Symptom trajectories.

A significant three-way interaction of anxiety trajectory-by-irritability trajectory-by-conflict emerged in the left cuneus (Figure 3, Panel A). Post hoc analyses indicated a cross-over interaction. For youth whose irritability remained stable across time (i.e., slope=0, corresponding to 1 SD above the mean), there was a positive association between anxiety trajectory and neural response to conflict, such that increasing anxiety across time (more positive slope) was associated with greater activation in response to conflict whereas decreasing anxiety across time (more negative slope) was associated with decreased activation. In contrast, for youth whose irritability showed a relatively steep decrease across time (corresponding to 1 SD below the mean), there was a negative association between anxiety trajectory and neural response to conflict, such that increasing anxiety across time (more positive slope) was associated with decreased activation in response to conflict, whereas decreasing anxiety across time (more negative slope) was associated with increased activation. There were no significant two-way interactions between anxiety trajectory or irritability trajectory and conflict condition.

Figure 3.

Conditional effects between anxiety trajectory and BOLD response were calculated for the mean value and +1/−1 standard deviation (SD) from the mean value of irritability trajectory in the sample. (Panel A) For youth whose irritability remained stable across time (i.e., slope=0, corresponding to 1 SD above the mean), increasing anxiety across time (more positive slope) was associated with greater BOLD response to conflict while decreasing anxiety across time (more negative slope) was associated with lesser BOLD response to conflict. (Panel B) For youth whose irritability remained stable across time (i.e., slope=0, corresponding to 1 SD above the mean) increasing anxiety across time (more positive slope) was associated with lesser BOLD response to error while decreasing anxiety across time (more negative slope) was associated with greater BOLD response to error.

Neural Response to Error

Model 2a: Age-15 symptoms.

There were no significant two- or three-way interactions involving age-15 anxiety and/or age-15 irritability with error condition.

Model 2b: Symptom trajectories.

A significant three-way interaction of anxiety trajectory-by-irritability trajectory-by-error emerged in the left cuneus (Figure 3, Panel B). Again, post hoc analyses indicated a cross-over interaction. For youth whose irritability remained stable across time (i.e., slope=0, corresponding to 1 SD above the mean), there was a negative association between anxiety trajectory and neural response to error, such that increasing anxiety across time (more positive slope) was associated with decreased activation in response to error, whereas decreasing anxiety across time (more negative slope) was associated with increased activation. In contrast, for youth whose irritability showed a relatively steep decrease across time (corresponding to 1 SD below the mean), there was a positive association between anxiety trajectory and neural response to conflict, such that increasing anxiety across time (more positive slope) was associated with increased activation in response to error, whereas decreasing anxiety across time (more negative slope) was associated with decreased activation. There were no significant two-way interactions between anxiety trajectory or irritability trajectory and error condition.

Associations with Task Behavior

No significant associations between task accuracy and clinical symptoms were found. A significant symptom-by-conflict interaction emerged for age-15 irritability symptoms, F(1,84)=5.09, p=0.03, such that participants with higher age-15 irritability showed a greater RT difference between incongruent and congruent correct response trials (i.e., flanker congruency effect). Follow-up analyses examining associations between irritability-linked alterations in neural response to conflict and RT difference indicated that greater neural response to conflict in the left IFG, right IFG, left Fusiform Gyrus, and right Fusiform Gyrus, but not the left Postcentral Gyrus, was associated with greater RT difference (all ps< .01). Further, neural response to conflict in the left IFG, right IFG, Left Fusiform Gyrus, and Right Fusiform Gyrus all significantly mediated the relations between age-15 irritability and RT difference (see Supplementary Text 2 for additional details).

DISCUSSION

The present study is the first to integrate symptoms measured cross-sectionally at age-15 with developmental trajectories of anxiety and irritability, in investigating neural correlates of cognitive control in developmental psychopathology. Results revealed that anxiety and irritability at age-15 related to distinct patterns of neural response to conflict: anxiety was uniquely associated with decreased neural response, while irritability was uniquely associated with increased neural response. With respect to developmental trajectories of anxiety and irritability, these trajectories interacted in relation to neural responses to both error and conflict.

The distinct relations of anxiety and irritability with neural response to conflict adds to a growing body of research emphasizing the importance of parsing unique vs. shared correlates of these two commonly co-occurring sets of symptoms (26–28). Our finding relating irritability to increased IFG activity in response to conflict is consistent with prior work. Two other studies related increased neural response in lateral prefrontal regions to elevated irritability, in the context of intact behavioral performance (20,21). Importantly, while this study was designed to parse differential associations with neural response to error and conflict, other aspects of cognitive control may connect to these findings. For example, the IFG is implicated in both response to conflict and response inhibition. Thus, future work employing paradigms that parse response to conflict and response inhibition is needed to further delineate the precise role of the IFG in relation to irritability. While we found intact overall task performance, age-15 irritability was associated with a greater reaction time difference between incongruent and congruent correct responses indicating a greater interference effect of conflict on cognitive control. Further, increased neural response to conflict was associated with greater reaction time difference, as indicated by a greater flanker congruency effect. Moreover, neural response to conflict statistically mediated the relations between age-15 irritability and greater interference effect of conflict on cognitive control (Supplementary Text 1). These results suggest that increased neural response to conflict may reflect compensatory processing or inefficient recruitment of neural resources in youth with elevated irritability.

In contrast, we found that anxiety was associated with decreased activity in response to conflict across a number of regions including the insula. The insula is implicated across a wide range of processes including integrating sensory and affective information with cognitive processes (40). Specifically in the context of cognitive control the insula is theorized to exert a causal influence on the deployment of attentional control and cognitive control circuitry when greater cognitive control is required (38,40). Furthermore, prior meta-analytic work has linked abnormal activation in the insula during cognitive control to transdiagnostic psychiatric symptoms (4). As predicted by attentional control theory (15,16), decreased activation might reflect impaired recruitment of neural resources underlying cognitive control, supporting the theory that inhibition of conflicting information is impaired in anxiety. However, this interpretation is difficult to reconcile with the intact behavioral performance observed in the current study. The over-control theory of anxiety posits that anxiety is associated with enhanced cognitive control (13,14). As noted above, follow-up analyses identified that increased neural response to conflict mediated the association between age-15 irritability and slowed behavioral responses in the presence of conflict. No such associations emerged between age-15 anxiety and task behavior. Therefore, anxiety-linked reduced neural activation to conflict may reflect a highly efficient neural response to conflict that manifests as no significant slowing in behavioral responses. Additional work clarifying the nature of the relations between anxiety and inhibitory control deficits is needed to identify whether anxiety may be related to impaired recruitment or greater neural efficiency during cognitive conflict.

For both anxiety and irritability, our longitudinal analyses revealed a significant decrease in symptoms from early childhood to early adolescence. The relatively steep decrease in irritability over development is consistent with normative trajectories of relatively high irritable mood, temper outbursts, and reactive aggression in early childhood, which decreases with time into adolescence as the prefrontal cortex matures (23,34). Consistent with prior work examining development of normative fears and anxiety (41), the trajectory for anxiety exhibited a slight, but significant decrease as well. With respect to associations of trajectories with brain response, a significant interaction emerged in the left cuneus for neural correlates of response to both error and conflict. Specifically, in youth with relatively stable irritability, decreasing anxiety across time was associated with decreased neural response to conflict (lesser BOLD response to incongruent vs. congruent stimuli) and increased neural response to error (greater BOLD response to error vs. correct response). Conversely, in these youth with relatively stable irritability, increasing anxiety across time was associated with increased neural response to conflict (greater BOLD response to incongruent vs. congruent stimuli) and decreased neural response to error (lesser BOLD response to error vs. correct response). Prior research implicates the cuneus in complex aspects of visual processing, as occur during visual conflict. While further research should be conducted, the associations demonstrated here link developmental trajectories of anxiety and irritability to altered cuneus response to conflict and error (42). Such alterations may reflect aspects of evolving function in this region; longitudinal imaging data are needed to consider this possibility. The current associations may reflect the visual nature of the conflict stimulus present in the task. Critically, these findings suggest that considering joint symptom trajectories may be important and additional information can be gleaned by considering their effects above and beyond current symptoms. Indeed, the effects for symptom trajectories emerged in a region that was not implicated in the effects of age-15 symptoms. However, this is the first study to investigate this effect. Future work would benefit from larger samples that could model joint trajectories (i.e., latent growth curve analyses) and samples with higher levels of clinical symptoms.

Contrary to expectations, there was no significant association between age-15 anxiety and neural response to error. We had predicted such an effect based on the literature (43). Anxiety has been linked to larger error-related negativity (ERN; 43) reflecting activation in the anterior cingulate cortex (ACC) and medial prefrontal cortex (mPFC; 44,45). Furthermore, associations between anxiety reduced BOLD response to error have been found in the dorsal lateral prefrontal cortex (dlPFC; 46). Contrary to these prior findings, we found no significant associations between anxiety and neural response to error. There are several possible explanations for the current finding. First, some past work has found that the association between anxiety and error responding is moderated by early life measures of infant temperament, such that no main effect of anxiety is observed (8,12). The association between anxiety and response to error has also been found to vary with age (12,47). It may be that our measurement of brain response at age-15 is being affected by an inflection point at which the association between anxiety and error response changes (43). Future work extending the current findings with longitudinal measures of neural response may better capture the developmental changes in neural response to error as they relate to anxiety.

The current study has several limitations. The CBCL was used as the longitudinal measure of symptoms as it was collected across all time points. CBCL items to measure anxiety and irritability were selected that were consistent across time and with prior work (8,23,34). This approach relied on only five and three items to measure anxiety and irritability, respectively. While prior work has noted the consistency and validity of the composite irritability score (48–50), recent work has found that the CBCL provides reliable information in adolescence but less so in childhood (51). Future longitudinal work employing other established measures of irritability would be ideal to extend the findings of the current study. Finally, overall levels of anxiety and irritability were relatively low. However, other literature suggests that findings across the full spectrum of symptoms contribute meaningfully to the understanding of neural correlates of these dimensions. For example, individual differences in anxiety and irritability symptoms, including at sub-clinical levels, predicts risk for later clinical psychopathology (52,53). Given the significant findings demonstrated here, future work in clinical samples is warranted.

Despite these limitations, the current study is the first to identify independent associations of anxiety and irritability symptoms at age-15, and interactive effects of longitudinal symptoms, in relation to the neural substrates of cognitive control in adolescence. The study provides initial support for the importance of studying the shared and unique contributions of mood and anxiety symptoms to neural mechanisms underlying cognitive control. The distinct neural correlates in response to conflict suggest that cognitive mechanisms underlying conflict processing may be critical for disambiguating mechanisms underlying anxiety and irritability. It may be that in the context of clinical irritability, inefficient cognitive control processing leads to the under-control of behavior resulting in outbursts. Conversely, in the context of clinical anxiety, aberrant cognitive control may manifest as the over-control of behavior resulting in excessive avoidance or inaction. Thus, consistent with our findings that irritability and anxiety symptoms were differentially associated with neural response to conflict, one might consider the utility of targeting cognitive control to address these clinical problems. Such efforts may require unique approaches in the two symptom dimensions, given distinct patterns of neural correlates for irritability and anxiety. By further probing cognitive control as a potential target for intervention, future work could lead to critical advances in the development of more individualized approaches to the treatment of these two highly prevalent sets of symptoms in youth.