Behavioral and Neural Sustained Attention Deficits in Disruptive Mood Dysregulation Disorder and Attention-Deficit/Hyperactivity Disorder

Abstract

Objective

Disruptive mood dysregulation disorder (DMDD), characterized by severe irritability, and attention-deficit/hyperactivity disorder (ADHD) are highly comorbid. This is the first study to characterize neural and behavioral similarities and differences in attentional functioning across these disorders.

Method

Twenty-seven healthy volunteers, 31 patients with DMDD, and 25 patients with ADHD (8–18-year-olds) completed a functional magnetic resonance imaging (fMRI) attention task. Group differences in intra-subject variability in reaction time (ISVRT) were examined. The current fMRI analytic approach precisely quantified trial-wise associations between reaction time and brain activity.

Results

Group differences manifested in the relationship between reaction time and brain activity (all regions: p<.01, F>2.54, η_p²>0.06). Patients with DMDD showed specific alterations in the right paracentral lobule, superior parietal lobule, fusiform gyrus, and cerebellar culmen. In contrast, both patients with DMDD and ADHD exhibited blunted compensatory increases in activity on long reaction time trials. Additionally, youth with DMDD exhibited increased activity in the postcentral gyrus, medial frontal gyrus, and cerebellar tonsil and declive (all regions: p<.05, F>2.46, η_p²>0.06). The groups in the imaging sample did not differ significantly in ISVRT (F[2,79]=2.664, p=.076, η_p²=0.063), though ISVRT was significantly elevated among youth with DMDD and ADHD when including those not meeting strict motion and accuracy criteria for the imaging analysis (F[2,96]=4.283, p=.017, η_p²=0.083).

Conclusion

Patients with DMDD exhibited specific alterations in the relationship between pre-stimulus brain activity and reaction time. Both patients with DMDD and ADHD exhibited similar blunting of compensatory neural activity in frontal, parietal, and other regions. Additionally, patients with DMDD demonstrated elevated reaction time variability relative to healthy youth. This work is the first to identify common and unique behavioral and neural signatures of DMDD and ADHD.

INTRODUCTION

Severe, chronic irritability in youth (operationalized in DSM-5¹ as disruptive mood dysregulation disorder [DMDD]) is highly comorbid with attention-deficit/hyperactivity disorder (ADHD) in cross-sectional and longitudinal work.² In a community sample of youth ages 9–19, 26.9% of those meeting criteria for severe mood dysregulation (SMD), a research diagnosis on which DMDD was based, also met criteria for ADHD.² In clinical samples, the comorbidity is even higher: i.e., >65% of patients with DMDD had comorbid ADHD.^3–5 Relatedly, meta-analytic work has indicated robust associations between ADHD and increased emotion reactivity/negativity/lability (weighted effect size d=0.95).⁶ Further, preschool ADHD may be a risk factor for subsequent DMDD⁷ and stimulants have been shown to decrease irritability in youth with ADHD.⁸ Nevertheless, no prior imaging study has examined the brain mechanisms mediating attentional function in ADHD vs. DMDD. Here, we utilize a sustained attention paradigm to identify the common and distinct neural signatures of DMDD and ADHD relative to healthy youth. This type of work can improve our understanding of the neurobiology of these two conditions and can help disentangle effects of attentional problems and irritability. Given the comorbidity between ADHD and DMDD observed in the literature and in the current sample, it is difficult to isolate potential effects of irritability separate from the influence of attentional problems. While identifying populations of children with severe irritability without ADHD would be highly useful, these children would not be representative of the general DMDD population. Thus, comparing children with ADHD and DMDD to healthy youth allows for investigation of both the contributions of ADHD symptomology and of severe irritability in the context of attentional problems.

In this fMRI study, we assessed neural correlates of a specific measure of attentional dysfunction, increased intra-subject variability in reaction time (ISVRT), in youth with DMDD and ADHD. ISVRT has been used as a measure of sustained attention, generally thought of as one’s ability to focus cognitive activity on specific stimuli. Elevated ISVRT is one of the most replicated behavioral correlates of ADHD; meta-analytic estimates indicate medium (Cohen’s d=0.56)⁹ to large effects (Hedges’ g = 0.76)¹⁰ on various tasks. Further, sustained attention deficits have been suggested to be a potential endophenotype or to relate to genetic risk for ADHD,^11,12 as well as potentially relevant for differentiating subtypes of ADHD.¹³ Additionally, there are mixed findings when comparing children with ADHD to other clinical populations on paradigms probing sustained attention (e.g.^14–16). As noted above, given the extensive overlap between ADHD and DMDD, it is important to compare the two groups on measures, such as ISVRT, that are pathophysiologically informative. This is the first study to examine ISVRT in DMDD and to compare to youth with ADHD. Several prior studies have compared SMD and ADHD on emotional paradigms (e.g., finding differential patterns of amygdala activity during processing of neutral faces¹⁷) and increased errors when labelling facial emotion among children with DMDD relative to those with ADHD.¹⁸ Yet no studies have compared these disorders on attentional paradigms. Examining the neural underpinnings of attentional functioning in DMDD and ADHD using analytic methods tuned to this question can yield clearer conclusions about the core differences and similarities among these disorders.

Some studies have examined correlations between ISVRT and average brain activity across a task in healthy individuals^19,20 or those with ADHD.²¹ However, recent work has linked reaction time (RT) to blood-oxygen-level-dependent (BOLD) signal on a trial-by-trial basis. This approach captures brain-behavior relationships more precisely than traditional fMRI analyses that rely on correlations between activity and behavior averaged across a task. Prior work using this trial-by-trial approach in healthy adults implicated frontal and parietal regions in attention control.^22–24 Particularly, these findings linked attentional lapses, manifesting as relatively long RTs, to both pre-stimulus decreases in cognitive control activity and increased compensatory activity in fronto-parietal regions.²² Recently, this approach has been extended to study risk for and expressions of bipolar disorder.²⁵ Of note, these findings suggested that this analytic method, as compared to more traditional imaging analytic approaches, is better able to capture subtle differences in attentional functioning between related conditions. As such, this method might prove useful when comparing ADHD and DMDD.

Our goal was to compare the neural correlates of reaction time variability in youth with DMDD and ADHD. We compared 27 healthy volunteers, 31 individuals with DMDD, and 25 individuals with ADHD. We expected to find elevated ISVRT in both patient groups and to find commonalities as well as differences between DMDD and ADHD in the neural dysfunction associated with these deficits. While no prior neuroimaging literature examines attentional functioning in DMDD, based on prior work with this behavioral paradigm, we expected to find alterations in the relationship between trial-wise RT and BOLD activity in mainly fronto-parietal attention control regions.^22,25

METHOD

Participants

Participants enrolled in an institutional review board-approved protocol at the National Institute of Mental Health in Bethesda, MD. Youth (<18 years) provided written informed assent, and their guardians provided written consent. We examined three groups: individuals with ADHD, individuals with DMDD, and healthy volunteers (HV) with no history of any psychiatric diagnoses.

Children’s psychiatric diagnoses were assessed using parent- and child-report on the Kiddie-Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime version (KSADS).²⁶ Irritability over the prior 6 months was assessed using parent-and child-report on the Affective Reactivity Index (ARI, averaged across reporter; range 0–12).²⁷ ADHD symptomology was assessed using the Conners’ Parent Rating Scale (CPRS - total T-score, Inattentive Subscale T-score, Hyperactivity subscale T-score, and Emotion Lability subscale T-score).²⁸ General functioning was assessed using the Children’s Global Assessment Scale (CGAS; 1=extremely impaired - 100=well-functioning).²⁹

Participants were excluded for pervasive developmental disorders, schizophrenia, bipolar disorder, posttraumatic stress disorder, IQ lower than 70 on the Wechsler Abbreviated Scale of Intelligence,³⁰ substance abuse within the past 2 months, significant medical illness, head trauma, neurological disorder, contraindications to MRI, having a first-degree relative with bipolar disorder (given prior findings of altered RT-BOLD signal associations based on a familial history of bipolar disorder²⁵), or treatment with long-acting stimulants. Participants taking short-acting stimulant medications (n=9 DMDD; n=12 ADHD) withheld their stimulants for 48 hours before scanning, as stimulants have been shown to reduce reaction time variability.³¹ Patients continued on their other medications, given ethical issues precluding withdrawal from other, longer-acting agents.

Of the 99 individuals (8–18 years old) who completed the fMRI scan successfully, data were excluded from one individual for structural abnormalities, five individuals for excessive head motion (see Imaging Analysis), and 10 individuals for poor task performance (accuracy below 70%). The final sample included 83 participants: 27 healthy volunteers, 31 individuals with DMDD, and 25 individuals with ADHD. Note that data from these 27 healthy volunteers have been examined previously as part of larger samples to examine age-related differences in the relationship between RT and activity on this task³² and to examine potential neural endophenotypes of bipolar disorder.²⁵ Data from youth with DMDD and ADHD are novel and have not been examined previously.

In-Scanner Behavioral Paradigm

Participants performed a modified global-local selective attention task with 576 total trials during fMRI scanning.^22,32 Figure S1, available online, displays the four task stimuli that appeared with equal frequency. Each stimulus was a large letter made up of smaller letters (“H” or “S”). On half of trials, the large and small letters were congruent (e.g., large H comprised of small Hs), while the other half of trials were incongruent (e.g., large S comprised of small Hs). This occurred over six task runs where participants were asked to identify, by button-press, either the large global letter or the small local letters in alternating runs. Each run included 96 trials: 48 congruent trials and 48 incongruent trials. All stimuli were centered on a red fixation point. On each trial, the stimulus appeared for 200ms followed by 2300ms of fixation in addition to a variable inter-trial interval, jittered in TR increments (1250ms), ranging from 2.5s to 16.25s (mean=3.98s). Stimulus presentation and jitter orders were optimized and pseudo-randomized using AFNI’s make_random_timing.py program.

Behavioral Data Analysis

Analyses examined accuracy, mean RT, standard deviation (SD) of RT, and ISVRT, excluding trials with no response or an RT <300ms after stimulus onset. ISVRT was calculated as the coefficient of variation of reaction time (CV-RT; SD/mean RT) and was the primary behavioral measure of interest. CV-RT is a preferable outcome measure to SD of RT as it handles issues with the correlation between mean and SD of RT.³³ Additionally, reaction time data was modeled using an ex-Gaussian distribution, which convolves a normal and an exponential distribution (retimes package³⁴ in R³⁵). Three parameters were examined to characterize behavior on the task: the mu and sigma parameters indicate the mean and standard deviation of the normal distribution function and the tau parameter relates to the mean and variance of the exponential function. The tau parameter was of particular interest to assess group differences in the exponential portion of the RT distribution, generally indicating increased numbers of long RT response, noted as increased in samples with ADHD previously.^36–39

To assess group differences in behavior, repeated measures analyses of variance (ANOVAs) were conducted using SPSS v2 2⁴⁰ with condition (congruent vs. incongruent) as a within-subject factor and group (HV, DMDD, ADHD) as a between-subjects factor. Correlations between behavioral measures and ARI and CPRS scores are presented in Supplement 1, available online, as well.

Imaging Acquisition

Neuroimaging data were collected on a 3T General Electric Signa scanner and 32-channel head coil. After a sagittal localizer scan, an automated shim calibrated the magnetic field to reduce signal dropout due to susceptibility artifact. Six functional runs of 304 time points each were acquired using 24 contiguous 5mm interleaved axial slices covering the entire brain. These scans used a 64×64 matrix with echo-planar single shot gradient echo T2* weighting (TR=1250ms; TE=25ms; FOV=240mm; flip angle=35°), yielding 3.75×3.75×5mm voxels. To reach longitudinal magnetization equilibrium, the six initial images from each run were discarded. A high-resolution structural MPRAGE scan (1mm interleaved sagittal slices; TE=min full; TI=425; FOV=25.6; freq × phase=256 × 256; flip angle=7°; 1mm³ voxels) was also acquired for co-registration with the functional data.

Imaging Preprocessing

Structural images were preprocessed using FreeSurfer v5.3.0,⁴¹ including motion correction, intensity correction and normalization, and skull stripping (autoreconl). Functional data were analyzed using Analysis of Functional NeuroImages⁴² (AFNI; http://afni.nimh.nih.gov/afni/) with standard preprocessing, including slice-timing correction, alignment of all volumes to base volume and non-linear registration to a Talairach template, despiking, spatial smoothing (6-mm FWHM kernel), masking, and intensity scaling. This work utilized the computational resources of the NIH HPC Biowulf cluster (http://hpc.nih.gov).

Data were rigorously controlled for head motion. Specifically, we excluded any pair of successive TRs where the sum head displacement (Euclidean norm of the derivative of the translation and rotation parameters) between those TRs exceeded 1mm. TRs where more than 10% of voxels were outliers were also excluded. Participants were excluded if the average motion per TR after censoring was greater than 0.25 mm or if more than 15% of TRs were censored for motion/outliers. Additionally, six head motion parameters were included as nuisance regressors in individual-level models.

Imaging Analysis

Mean standardized reaction times (MSRT) were calculated to mean-center RTs and were included in our trial-by-trial imaging analyses. For each trial, we subtracted a participant’s mean RT for that trial type (congruent or incongruent) for that run from that trial’s RT.²² A lower MSRT indicated a faster RT relative to one’s run- and condition-wise mean RT, and higher MSRT indicated slower relative RT.

We analyzed brain activity using a linear regression model (a sample design matrix is presented in Figure S2, available online) with two important elements described below: an estimated shape method⁴³ and amplitude modulation by RT. The estimated shape method, i.e. a finite impulse response model, was used rather than assuming a standard BOLD response shape. This allows for estimation of BOLD signal at multiple TRs throughout a trial, though this comes at the cost of degrees of freedom in the model. Thus, the estimated shape method generates more precise estimation of the shape and timing of RT-BOLD signal relationship. Each trial was modeled with 13 basis functions (tent function), each spanning 1 TR (1250ms) (see Figure S2B, available online). Thus, trials were modeled with a 16.25s (13 TRs) window starting 2.5s before stimulus onset to examine pre-stimulus effects as in prior work.^22,25,32 Third-order Legendre polynomials modeling baseline drift and six head motion parameters were included to control for potential confounding effects.

Each participant’s model included regressors for congruent, incongruent, and error trials (each condition was modelled by 13 regressors). As in standard fMRI analyses, this allowed us to examine average stimulus-locked BOLD activity. This average BOLD activity indicates the mean percent signal change estimated at each timepoint across all trials in a condition, i.e. not examining trial-by-trial differences in RT.

Additionally, we conducted an amplitude modulation analysis that allowed us to examine associations between trial-by-trial variation in BOLD response and variation in RT. Specifically, each participant’s model included regressors for modulation by MSRT on congruent and on incongruent trials.²² Indexing the strength of the RT-BOLD association, the coefficients for the amplitude modulation effects indicate differences in percent signal change per second increase above one’s mean RT (mean centered for a given condition and run). Thirteen regressors were included for each condition to examine modulation of BOLD signal by trial-wise MSRT at each of the 13 timepoints estimated. Example regressors for trial onset and for MSRT modulation are presented in Figure S2C, available online.

Examining Group Differences in Brain Activity

Our goal was to identify common and distinct neural signatures associated with DMDD and ADHD. Given this goal and the high correlation between continuous measures of ADHD symptomology and irritability (Table S1, available online), we focused on categorical analyses comparing youth with DMDD, ADHD, and HV. This clinically relevant approach allowed us to compare youth who had been carefully phenotyped and differentiated into three groups: youth with clinically relevant attentional problems but not severe irritability, youth with elevated attentional problems and severe irritability, and youth with low levels of both symptoms. While consistent with DSM-5, this approach, as compared to a fully continuous approach, also reduces multi-collinearity by reducing variability in the independent measures. As such, group differences in both average BOLD activity and RT-BOLD activity associations were assessed using whole brain, voxel-wise multivariate models⁴⁴ (3dMVM in AFNI). The models included group (HV, ADHD, DMDD) as a three-level between-subjects factor and within-subject factors for condition (congruent, incongruent) and timepoint (TR/’tent’). We focused on group differences, particularly the group × timepoint contrast. Group differences may also be apparent as a function of condition, i.e. the group × condition × timepoint contrast, yet this contrast did not yield interpretable findings in prior work.²⁵ Therefore, we provide these in Supplement 1, available online, for completeness but do not focus on these contrasts in the main text.

Monte Carlo simulations were performed using the 3dClustSim program from AFNI to correct for multiple comparisons. All analyses were restricted to a whole brain mask with 33,350 voxels. Smoothness of the residual timeseries from the individual level models was estimated based on a Gaussian plus mono-exponential spatial autocorrelation function (3dFWHMx with acf flag) for all participants and averaged, yielding an effective smoothness of FWHM=14.38mm (ACF parameters, a=0.50, b=5.61, c=12.50). Two-sided thresholding was examined for whole-brain F-tests with first-nearest neighbor clustering (NN=1). To obtain a whole-brain family-wise error correction of p<.025 (to correct for examination of RT-BOLD associations and average BOLD activity), all results were thresholded at a voxel-wise p<.001 and a cluster extent of 25 voxels. Activity was extracted for each individual as an average from each cluster for post hoc analyses and visualization. Greenhouse-Geisser adjustment corrected for violations of sphericity. Pairwise post hoc group t-tests were run at each timepoint to parse group differences.

Follow-Up Control Analyses

As the groups differed in the mean motion remaining after censoring (Table 1), we included this as a covariate (and interactions with motion) in follow-up control analyses to rule out effects of motion. Additionally, we included age (and interactions with age) in these analyses as age has been shown to relate to RT-BOLD signal associations in some regions.³² In an additional control analysis, we excluded 1 patient with ADHD and 13 with DMDD taking medications (other than short-acting stimulants, which were withheld for the scan) to test for effects of medication status.

Table 1.

Participant Characteristics

	HV (n=27)	DMDD (n=31)	ADHD (n=25)	Comparison
Sex, n (%) female	15 (56)	15 (48.4)	10 (40)	χ2=1.26, p=.53
Age, Mean (SD)	15.17 (1.90)	13.74 (2.30)	14.24 (2.54)	F (2,80)=2.94, p=.06, ηp2=.07
IQ, Mean (SD)	110.78 (11.68)	110.48 (14.83)	117.76 (12.26)	F(2,80)=2.59, p=.08, ηp2=.061
Head Motion, Mean (SD)	0.072 (0.025)	0.100 (0.040)	0.075 (0.044)	F(2,80)=4.80, p=.01, η p2 =.11
ARI, Mean (SD)a	0.57 (0.99)	5.73 (2.56)	2.34 (1.93)	F(2,76)=46.596, p<.001, np2=.55
CPRS Total, Mean (SD)b	42.86 (2.87)	70.69 (14.50)	66.30 (9.76)	F(2,70)=44.14, p<.001, np2=.56
CPRS Inattention, Mean (SD)b	43.05 (2.60)	67.83 (13.15)	64.39 (9.54)	F(2,70)=41.34, p<.001, np2=.54
CPRS Hyperactivity, Mean (SD)b	45.10 (2.86)	69.83 (14.61)	64.52 (13.69)	F(2,70)=26.80, p<.001, np2=.43
CPRS Lability, Mean (SD)b	42.91 (2.21)	68.66 (12.21)	49.74 (6.43)	F(2,70)=61.50, p<.001, np2=.64
CGAS, Mean (SD)c	–	49.97 (8.80)	70.91 (8.24)	t(52)=8.88, p<.001, d=2.5
Lifetime diagnoses, n (%)
MDD	–	11 (36.7)	0 (0)	χ2=10.65, p=.001
ODD or CD	–	28 (90.3)	0 (0)	χ2=45.16, p<.001
Any anxiety disorder	–	19 (61.3)	3 (12)	χ2=14.10, p<.001
ADHD	–	24 (77)	25 (100)	χ2=6.45, p=.011
Inattentive type	–	11	17
Hyperactive type	–	0	1
Combined type	–	10	7
ADHD-NOS	–	3	0
Medication, n (%)
None (after withholding stimulants)	–	13 (41.9)	24 (96)	χ2=18.05, p<.001
Antipsychotic	–	11 (35.5)	0 (0)	χ2=11.04, p=.001
Antiepileptic	–	4 (12.9)	0 (0)	χ2=3.47, p=.06
Antidepressant	–	10 (32.3)	1 (4)	χ2=7.00, p=.008

Note: Demographic and clinical data are summarized for each group: healthy volunteers (HV), patients with disruptive mood dysregulation disorder (DMDD), and patients with attention-deficit/hyperactivity disorder (ADHD). Analyses of variance, independent-samples t-tests, or chi-square tests were used to compare across groups. Significant group differences are marked in bold. ARI = affective reactivity index; CD = conduct disorder; CGAS = Children’s Global Assessment Scale; CPRS = Conners Parent Rating Scale; ODD = oppositional defiant disorder; NOS = not otherwise specified.

^aARI scores were missing from 4 HVs.

^bCPRS scores were missing from 6 HV, 2 individuals with DMDD, and 2 with ADHD.

^cCGAS scores were missing from 2 patients with ADHD.

RESULTS

Demographics and Diagnostics

The three groups did not differ significantly on sex, age, or IQ (Table 1). Youth with DMDD exhibited significantly more head motion than the other groups. Thus, motion was examined as a covariate in the follow-up control analyses.

As expected based on the inclusion criteria, youth with DMDD exhibited elevated irritability (ARI) and emotion lability (CPRS) vs. the other two groups. Both youth with DMDD and youth with ADHD exhibited elevated ADHD symptomology (CPRS - subscales and total T-scores) relative to HVs. Youth with DMDD showed significantly more impairment (CGAS) than youth with ADHD, as well as higher rates of major depressive disorder (MDD), anxiety, oppositional defiant disorder (ODD)/conduct disorder (CD) diagnoses, and medication use. Pairwise group differences and correlations among all demographic and clinical factors are presented in Table S1, available online. For example, there was high correlation between ADHD symptomology and irritability (r[68]=.590, p<.001).

Behavioral Effects

Using a repeated measures ANOVA, no significant group difference in ISVRT emerged (F[2,79]=2.664, p=.076, η_p²=0.063; Figure S3a, available online); condition (p=.448) and group × condition (p=.185) effects were also not significant. Similar results were noted when examining the tau parameter, specifically a nonsignificant overall effect of group (F[2,79]=2.602, p=.081, η_p²=0.063; Figure S3b, available online). ISVRT and tau correlated positively with both irritability and ADHD symptom severity (Spearman rhos between .249 and .415; Table S1, available online). Table S2, available online, summarizes ANOVA effects on, and symptom correlations with, accuracy, mean RT, SD of RT, ISVRT, mu, sigma, and tau.

Importantly, these analyses examined the subsample of children who were able to complete the scan successfully with high accuracy and minimal head motion. When including the 16 children excluded from the imaging analysis, we identified significant group differences in ISVRT (F[2,96]=4.283, p=.017, η_p²=0.083) and tau (F[2,96]=5.250, p=.007, η_p²=0.099); both metrics were elevated among youth with DMDD and ADHD relative to HVs (ps<.05).

RT-BOLD Effects

The RT-BOLD analyses assessed group differences in the relationship between trial-wise RT and BOLD amplitude. Five regions showed a group × timepoint interaction predicting trial-by-trial RT-BOLD signal relationships (yellow regions in Figure 1, Table 2, Figure S4, available online). All five regions also showed significant main effects of timepoint (all ps<.001). These regions showed two main patterns: RT-BOLD signal patterns specifically altered in patients with DMDD only or patterns altered in both patient groups (see Figure 2A for an example time course). First, the right paracentral lobule, right superior parietal lobule, right fusiform gyrus, and right cerebellar culmen showed increased RT-BOLD signal associations at the first pre-stimulus timepoint among youth with DMDD vs. HV and youth with ADHD; the latter two groups showed zero or negative RT-BOLD signal associations at that pre-stimulus timepoint. This suggests that longer RT responses are associated with a pre-stimulus increase in activity among youth with DMDD whereas the other groups showed either no relationship or a slight negative association between trial-wise RT and pre-stimulus activity. Second, all five regions showed a blunting of RT-BOLD at the peak of the time course (or later in the trial for the culmen) in both patient groups relative to HVs, i.e. weaker positive associations between RT and BOLD amplitude. This indicates that potential compensatory increases in activity were present on long RT trials among HVs but were blunted among patients. Overall, these findings had medium-large effect sizes (group × timepoint η_p² = 0.06–0.11). For example, in the superior parietal lobule, an independent samples t-test comparing individuals with DMDD to those with ADHD and HV individuals indicated a medium-large effect (t[81]=3.11, p=.003, Cohen’s d = 0.69).

Figure 1.

Group × timepoint effects on reaction time-blood oxygen level-dependent (RT-BOLD) and average BOLD. Note: This figure presents the regions that showed a significant group × timepoint interaction predicting RT-BOLD relationships (amplitude modulation) in yellow and predicting average BOLD activity in blue. Thresholded F-statistics from the voxel-wise repeated measures analysis of variance (ANOVA) are presented. The peak coordinates, voxel extents, and further information about each region are presented in Table 2.

Table 2.

Group × Timepoint Interactions Predicting Reaction Time-Blood Oxygen Level-Dependent (RT-BOLD) Relationships

Region	x	y	z	Voxels	BA	F	ηp2	Group Differences - Pre-stimulus	Group Differences -Peak
Right Paracentral Lobule	2	−36	59	64	5	2.54	0.06	DMDD>ADHD=HV	HV=DMDD>ADHD
Left Precuneus	−23	−54	42	43	7	3.37	0.08	DMDD=ADHD=HV	HV>DMDD=ADHD
Right Superior Parietal Lobule	12	−64	56	41	7	3.30	0.08	DMDD>ADHD=HV	HV>DMDD=ADHD
Right Fusiform Gyrus	51	−40	−11	27	37	2.94	0.07	DMDD>ADHD=HV	HV>DMDD=ADHD
Right Culmen	51	−43	−32	25	–	5.06	0.11	DMDD>ADHD=HV	DMDD=ADHD=HV

Note: Regions showing significant group × timepoint effects on RT-BOLD relationships (amplitude modulation) in a whole brain analysis are presented here. Coordinates (x, y, z) are presented for the peak voxel in each region as well as the number of voxels exceeding the whole brain threshold. The Brodmann area (BA) for each region is also presented. The F-statistic and partial eta squared (n_p²) of the group × timepoint interaction were calculated on RT-BOLD averaged across the region. All group × time interactions and main effects of time were significant (p<.007). Pair-wise group differences at the pre-stimulus timepoint (2.5 seconds before stimulus onset) and at the peak timepoint (6.25 seconds after stimulus onset) are summarized here; significant effects at p<.05 are noted in the Group Differences columns. ADHD = attention-deficit/hyperactivity disorder; DMDD = disruptive mood dysregulation disorder; HV = healthy volunteer.

Figure 2.

Group differences in reaction time-blood oxygen level-dependent (RT-BOLD) and average BOLD. Note: This figure displays exemplar results illustrating group × timepoint interaction predicting A) RT-BOLD in the right superior parietal lobule and B) average BOLD results in the right postcentral gyrus. In panel A, estimated marginal means of percent signal change per second increase in mean standardized reaction times (MSRT) and their standard errors from the repeated measures analysis of variance (ANOVA) predicting RT-BOLD effects are displayed. In panel B, estimated marginal means of percent signal change and their standard errors are presented. For both plots, for healthy volunteers (HV; green circles), patients with attention-deficit/hyperactivity disorder (ADHD; blue triangles), and patients with disruptive mood dysregulation disorder (DMDD; red squares) are displayed with shading indicating timepoints that showed p<.05 significant t-test differences between any of the three groups. Stimulus onset is denoted as at 0 seconds; the two pre-stimulus TRs modelled are −2.5 and −1.25 seconds.

Additionally, as an exploratory follow-up, we tested the association between RT-BOLD in these regions showing group differences and ISVRT as a metric of overall attention variability. Specifically, RT-BOLD values at the peak timepoint negatively correlated with ISVRT across the sample in three of the five regions identified above; the left precuneus (r=−.35, t=−3.30, p=.001), the right superior parietal lobule (r=−.25, t=−2.23, p=.03), and right fusiform (r=−.41, t=−3.96, p<.001). Thus, the blunted compensatory RT-BOLD association at the peak of the time course in these regions was related to greater ISVRT. The right paracentral lobule and culmen regions showed negative but non-significant associations (ps>.13)

Though not a main focus of the analysis, we did identify 6 regions showing a group × condition × timepoint analysis (Table S3 and Figures S5–6, available online). These regions in the left and right precuneus, left supramarginal gyrus, right middle frontal gyrus, right occipital gyrus, and left inferior frontal gyrus tended to show similar effects as above, i.e. blunted compensatory RT-BOLD signal in the patients at the peak and increased pre-stimulus RT-BOLD signal associations in the youth with DMDD, specifically during congruent but not incongruent trials (See Supplement 1, available online, for details).

Average BOLD Analyses

Ten regions showed a group × timepoint interaction predicting BOLD activity, i.e. average activity rather than relationships with RT (blue regions in Figure 1, Table 3, Figure S7, available online). Nine of these regions showed significant main effects of time (ps<.001); the remaining inferior temporal gyrus region (peak = −47, −19, −25; 40 voxels) did not show a meaningful hemodynamic response function and thus is not discussed further. Several regions showed increased activity among youth with DMDD relative to both youth with ADHD and HV youth (see Figure 2B for example time course). These regions included the left and right postcentral gyrus, right cerebellar declive, right medial frontal gyrus, and right cerebellar tonsil. The parahippocampal gyrus and posterior cingulate showed blunting among patients vs. HVs. Group differences in the thalamus were not consistent across the modelled time course. No regions showed significant group × condition × timepoint interactions.

Table 3.

Group × Timepoint Interactions Predicting Average Blood Oxygen Level-Dependent (BOLD) Activity

Region	x	y	z	Voxels	BA	F	ηp2	Group Differences - Peak
Right Postcentral Gyrus	26	−33	52	162	3	4.45	0.10	DMDD>ADHD=HV
Right Declive	37	−75	−22	85	–	4.10	0.09	DMDD=HV>ADHD
Left Postcentral Gyrus	−30	−33	52	73	3	3.85	0.09	DMDD>ADHD=HV
Right Medial Frontal Gyrus	9	−19	52	72	6	3.69	0.09	DMDD>ADHD=HV
Left Thalamus	−2	−22	3	48	–	2.46	0.06	DMDD>ADHD=HV
Right Parahippocampal Gyrus	30	−36	−8	39	36	3.61	0.08	HV>DMDD>ADHD
Left Posterior Cingulate	−5	−54	24	39	31	2.77	0.07	DMDD=HV=ADHD
Right Cerebellar Tonsil	23	−47	−46	29	–	3.61	0.08	DMDD>HV>ADHD
Right Superior Frontal Gyrus	2	13	49	25	6	3.00	0.07	DMDD=HV>ADHD

Note: Regions showing significant group × timepoint effects on average BOLD activity in a whole brain analysis are presented here. Coordinates (x, y, z) are presented for the peak voxel in each region as well as the number of voxels exceeding the whole brain threshold. The Brodmann area (BA) for each region is also presented. The F-statistic and partial eta squared (n_p²) of the group × timepoint interaction were calculated on BOLD activity averaged across the region. All group × time interactions and main effects of time were significant (p<.05). Pair-wise group differences at the peak timepoint (3.75 seconds after stimulus onset) are summarized here; significant effects at p<.05 are noted in the Group Differences column. ADHD = attention-deficit/hyperactivity disorder; DMDD = disruptive mood dysregulation disorder; HV = healthy volunteer.

Follow-Up Analyses

To control for potential effects of head motion or age, we included these additional covariates in follow-up repeated measures analyses of covariance predicting RT-BOLD and average BOLD in the regions described above. All group × timepoint effects predicting RT-BOLD and average BOLD remained significant (all ps<.023) when controlling for these additional factors. This suggests that group differences in brain activity were not accounted for by group differences in age or head motion. Excluding patients on medication reduced our power to observe group × timepoint effects. Thus, while the RT-BOLD effects retained their effect sizes, most became only trend-level significant due to the drop in sample size (Table S4, available online).

DISCUSSION

The current study is the first to compare neural and behavioral alterations in attentional functioning in ADHD and DMDD, with evidence of both specific and shared dysfunctions. Specifically, when quantifying precisely trial-wise associations between RT and BOLD activity, we identified increases in pre-stimulus activity associated with long RT trials among DMDD youth relative to both ADHD and healthy youth. However, in patients with ADHD and those with DMDD relative to healthy youth, we identified blunting of the peak activation in trials with long RTs. This peak may represent compensatory activity that occurred in response to the long RT trials in healthy youth, but failed to manifest in either those with ADHD or DMDD. In an exploratory follow-up, we note that this blunting was related to elevated ISVRT across the sample in most of the identified regions. Finally, when examining average BOLD activity, as in typical fMRI analyses, we identified increases in activity specific to DMDD during this attentional paradigm.

Thus, some neural dysfunction was similar among patients with ADHD and those with DMDD, particularly blunted RT-BOLD signal associations at the peak of the time course of activity. Again, these RT-BOLD signal associations indicate trial-by-trial associations between activity and reaction time. In particular, this analytic approach allows the identification of neural activity associated with attentional lapses on long RT trials.²² Here, we see that healthy individuals mobilize parietal and related brain regions on long RT trials, but patients fail to deploy this compensatory increase. Additionally, we see altered average BOLD activity across all trials, rather than as a function of RT differences, among patients with ADHD and those with DMDD in the parahippocampal gyrus, posterior cingulate, and superior frontal gyrus, for example. Similar regions have shown altered activity in prior sustained attention studies of ADHD, e.g.⁴⁵. Additionally, prior work in healthy participants identifies these frontal, parietal, cingulate, and other regions as components of circuits underpinning sustained attention,⁴⁶ with convergence among neuroimaging and lesion studies highlighting particular contributions from parietal attention control regions. Thus, blunting among both patients with ADHD and those with DMDD may reflect underlying attentional dysfunction shared across the two groups. Blunted activity in frontal, parietal, cingulate, fusiform, and other regions has been identified in meta-analyses examining patients with ADHD relative to healthy individuals across a variety of attentional, cognitive control, and working memory tasks.^47,48

Additionally, we identified areas of neural dysfunction present in DMDD, but not ADHD. In patients with DMDD only, long RT trials were associated with increased activity before trial onset in the paracentral lobule, superior parietal lobule, fusiform, and cerebellum. This is in contrast to patterns observed in healthy individuals here and in prior work²² where long RT trials, likely due to lapses in attention, are typically preceded by dips in control activity. With regard to average BOLD activity (averaged across all trials, rather than relationships with trial-wise RT), we see robustly elevated activity among patients with DMDD relative to both youth with ADHD and healthy youth in the postcentral gyrus, medial frontal gyrus, as well as regions of the cerebellum. As the neural underpinnings of attention regulation have not been examined in patients with DMDD previously, this potentially represents a novel insight into neural mechanisms associated with severe irritability in youth. While recent work has begun to suggest differences between patients with DMDD and ADHD in emotion processing,¹⁷ this is the first study to explore differences between these groups in attention.

As detailed in Supplement 1, available online, we identified several group differences in RT-BOLD associations that were present only on congruent trials. Here, we again saw associations between long RT trials and increased pre-stimulus activity in youth with DMDD, along with decreased compensatory activity in both patient groups. As in these analyses identifying group differences specifically on less attentionally demanding trials, prior work has also found that children with ADHD exhibited efficient attentional filtering when task demands are high, but exhibit deficient distractor filtering under low task demands.⁴⁹ This prior study suggested that ADHD deficits relate to failures to efficiently engage top-down control as opposed to difficulties with filtering in sensory processing regions. Similarly, we note alterations in a less demanding condition in predominantly fronto-parietal attention regions.

Finally, the current results provide the first demonstration of increased intra-subject reaction time variability, as well as elevated tau ex-Gaussian parameters, among youth with DMDD or with ADHD relative to healthy youth (within the full sample of children with behavioral data; though not significant within the sample meeting strict motion and accuracy criteria for neuroimaging analyses). Increased ISVRT indicates more variable RT (higher SD), controlling for one’s mean RT. A higher tau parameter indicates a stronger exponential component to the RT distribution, i.e. more long RT trials. Both indices are generally elevated when participants have more responses with long RTs, potentially indicating more lapses in attention over the course of the paradigm.

We found both common and distinct neural signatures of attentional functioning in ADHD and DMDD in mainly fronto-parietal attention control regions. Given the high levels of ADHD symptomology among the youth with DMDD, common neural alterations in ADHD and DMDD, relative to healthy youth, may be attributable to ADHD symptomology. On the other hand, the distinct neural alterations that we found in DMDD, relative to both ADHD and healthy youth, are likely to be associated with the severe irritability that is present only in the DMDD group. It will be helpful in the future to study youth with DMDD with less ADHD comorbidity, acknowledging that this may be a relatively unrepresentative sample of youth with DMDD.

While this is the first study to compare behavioral and neural alterations in attention between youth with DMDD and ADHD, there are limitations. First, we were unable to disentangle fully relationships between ADHD symptomology and irritability, given the high comorbidity between DMDD and ADHD. The high correlation between continuous measures of irritability and ADHD symptomology restricted our ability to consider the dimensional effects of these factors. While this sample does provide a unique opportunity to compare patients with ADHD alone vs. those with both ADHD and severe, chronic irritability, ascertaining a comparison population of highly irritable youth without elevated ADHD symptomology would be difficult yet potentially informative. Such a sample might be unrepresentative of the general population of severely irritable youth, and therefore the data would have to be interpreted cautiously. However, a study that included youth with severe irritability but without ADHD symptoms could be helpful in differentiating the relative contributions of irritability and attentional problems in cognitive control deficits, such as detected using the paradigm here.

Additionally, given prior work indicating that stimulants mitigate ISVRT deficits,³¹ patients taking short-acting stimulants withheld their medications, and children taking longer-acting stimulants were excluded from the study. However, for ethical reasons, other medications could not be withheld. While we present an analysis of data from the subsample that was medication-free at the time of scan, this subsample is quite small. Ideally, future work would replicate these findings in a larger sample of medication-free or medication-naive patients.

This is the first study to identify elevated reaction time variability (ISVRT and tau) among youth diagnosed with DMDD. Further, this is the first demonstration of both neural alterations common across patients with DMDD and ADHD, as well as neural signatures distinct to patients with DMDD relative to both youth with ADHD and healthy youth. These distinct patterns were particularly apparent using sophisticated analyses quantifying the trial-by-trial association between brain activity and RT. Future work can further parse these overlapping and distinct neural mechanisms and work to further disentangle the contribution and correlations of attention problems and irritability in youth, particularly given robust associations between ADHD and increased emotion reactivity/negativity/lability (meta-analysis⁶; Table S1, available online).