Reduced Activation in the Pallidal-Thalamic-Motor Pathway is Associated with Deficits in Reward-Modulated Inhibitory Control in Adults with a History of Attention Deficit Hyperactivity Disorder

Abstract

Background

Attention deficit hyperactivity disorder (ADHD) symptoms persist into adulthood and are associated with functional impairments. Neuroimaging studies of reward-modulated inhibitory control can identify potential objective markers of impairment and may deepen our understanding of why probands engage in costly behaviors leading to adverse outcomes. The study aimed to identify reward-modulated inhibitory control neural circuitries, their association with ADHD symptoms, and real-world implications of a decreased capacity to engage in reward-modulated inhibitory control.

Methods

One hundred and six adults (90% male) with rigorous childhood diagnoses of ADHD were scanned with functional MRI during the Monetary Incentive Go/NoGo task. Adulthood symptoms of inattention and hyperactivity/impulsivity based on self- and informant-report were assessed. The number of lifetime-attempts taken to quit smoking were also assessed as an exemplar real-world outcome.

Results

Hyperactivity/impulsivity was negatively associated with activation in the pallidum and primary motor cortex when inhibiting a previously rewarded Go stimulus that yielded a small immediate reward in order to obtain a larger reward later on. Reduced recruitment of the pallidal-thalamic-motor circuit mediated the negative association between hyperactivity/impulsivity and reward-modulated inhibitory control accuracy. Reduced pallidum activation, in response to reward-modulated inhibitory control, was also associated with more attempts made to successfully quit smoking.

Conclusions

Probands with persistent hyperactivity/impulsivity symptoms have alterations in brain regions that calculate the value of inhibiting an action that yields an immediate reward in order to obtain delayed larger rewards. This deficit results in poor inhibitory control on basic tasks and during real-world behaviors that rely on similar processes.

Introduction

Childhood-onset ADHD persists into adulthood for 15–35% of those affected; furthermore, 22–65% of adults with a childhood-history of ADHD (probands) continue to experience impairing symptoms that do not reach full DSM criteria (1). The related costs of ADHD in adulthood are high, and symptom persistence into adulthood is often associated with multiple functioning difficulties (2). Examining associations between persistence of the core symptom dimensions of ADHD (inattention and hyperactivity/impulsivity) and the functioning of neural systems underlying inhibitory control in the context of reward may deepen our understanding of why probands engage in costly behaviors leading to adverse outcomes (3, 4).

Inhibitory control can be characterized as reactive or proactive (5, 6). Reactive inhibition is the stopping of an action triggered by an external stimulus, e.g., one brakes when seeing a stop sign (6). Reactive inhibition is supported by the hyperdirect fronto-basal ganglia motor pathway, including, right inferior frontal gyrus (rIFG), pre-supplementary motor area (preSMA), subthalamic nucleus (STN), globus pallidus pars interna (GPi), thalamus, and primary motor cortex (M1) (5, 6). The rIFG and preSMA likely send an inhibitory command via direct inputs into the STN whose excitatory neuronal outputs innervate the GPi which inhibits the thalamus which inhibits Ml leading to inhibition of the physical action (5). Adults with ADHD show less rIFG (7, 8) and thalamus (7) activation when engaging in reactive inhibition. Notably, reduced recruitment of the rIFG appears to be only present for probands with persistent symptoms (8, 9). Within probands, reduced IFG, putamen, and pallidum activation during reactive inhibition is associated with more hyperactive-impulsive symptoms (8, 10).

Proactive inhibition, in contrast, refers to responding with restraint to attain future goals, e.g., one slows down driving in a neighborhood where they will likely be stop signs (6). Proactive inhibition is thought to be implemented by the indirect fronto-basal ganglia motor pathway comprised of the dorsolateral prefrontal cortex (DLPFC), caudate, putamen, globus pallidus pars externa (GPe), GPi, thalamus, and M1 (5, 6, 11). In the context of proactive inhibitory control, the DLPFC and putamen have been implicated in setting and prioritizing among future action goals resulting in rapid deployment of inhibitory control when needed (11). Motor inhibition occurs via GPe inhibition of the GPi which inhibits the thalamus which ultimately inhibits Ml (5). In the context of reward, the DLPFC likely drives nucleus accumbens (NAcc) and ventral tegmental area (VTA) activation in response to cues that signal reward (11–13). Spreading of VTA activation to the subtania nigra (SN) subsequently drives putamen activation to prioritize actions that facilitate obtaining the reward (12, 14). Activation along the GPe/GPi boarder (GPb) likely tracks the reward magnitude, i.e., the value, associated with engaging in a specific action (15–17). This information is fed forward to the SN/VTA which reinforces DLPFC and putamen activation (12). Thus, GPe/GPi activation may play a critical role in facilitating the inhibition of immediate reward in order to receive a larger more delayed reward. It remains unknown whether alterations with the pallidal-thalamic-motor circuit mediate behavioral deficits in reward-modulated inhibitory control in the context of ADHD. Furthermore, it is unclear to what extent the functioning of neural circuits underlying reward-modulated inhibitory control explain variability in real-world outcomes (18).

The current study addresses three questions, making use of a longitudinal ADHD sample rigorously diagnosed in childhood and followed into adulthood: (1) Are persisting ADHD symptoms related to inhibition accuracy, behavioral markers of proactive inhibitory control, and activation in neural regions underlying reward-modulated inhibitory control? (2) To what extent do alterations in behavior and neural function indicative of/supporting proactive inhibitory control in the context of reward mediate the association between persisting ADHD symptoms and behavioral performance? (3) What are the real-world implications of a decreased capacity to engage in reward-modulated inhibitory control for probands (in this case, attempts to quit smoking)? Using a dimensional approach, we examined associations between persisting ADHD symptoms and neural substrates of reward-modulated inhibitory control in probands using a Monetary Incentive (MI) Go-NoGo task. In the MI Go-NoGo task, participants must inhibit a prepotent response to forgo an immediate reward in order to obtain a larger delayed reward (Reward-NoGo), and they must also inhibit a prepotent response under no reward conditions (Neutral-NoGo). Given the robust literature on vulnerability to daily smoking in ADHD including in this sample (19), we assessed the number of attempts needed to quit smoking as one real-world indicator of the capacity to engage in proactive inhibition.

We hypothesized that: (1) hyperactivity/impulsivity symptom severity (hereafter, “impulsivity”) would be more strongly negatively associated with inhibition accuracy, markers of proactive inhibitory control, and activation in brain regions associated with proactive inhibitory control in the Reward-NoGo condition relative to the Neutral-NoGo condition; (2) the association between impulsivity and inhibition accuracy during the reward condition would be: (a) mediated by markers of proactive inhibitory control, and (b) serially mediated by greater activation from the GPe to the GPi to the thalamus to the M1; and (3) GPb activation would be negatively associated with the number of attempts needed to successfully quit smoking.

Methods & Materials

Participants

Participants were 106 probands from the Pittsburgh ADHD Longitudinal Study – a sample diagnosed with DSM-III-R or DSM-IV ADHD as children between 1987 and 1996 using comprehensive, standardized, multi-informant diagnostic methods including clinician consensus (20): M(SD)_age=33.5(3.4), 89.6% male, 72.6% Caucasian, 32% with persisting ADHD symptoms (see Table 1). Inclusion criteria for the current study included absence of conditions that contraindicate scanning. Participants taking stimulant medication (n=3) were required to refrain for 24 hours prior to the scanning session and smokers were instructed to abstain for 2 hrs. prior to scanning (see Supplemental Methods for further details).

Table 1.

Demographic characteristics of probands

Clinical Measures	N	%
Sex, males	95	89.6
Race and Ethnicity
Caucasian	77	72.6
African American	15	14.2
Multiple	11	10.4
Asian	1	0.9
American Indian/Alaskan Native	1	0.9
Other	1	0.9
Handedness
Right	70	72.2
Left	11	11.3
Mixed	16	16.5
Clinical Diagnoses
Mooda	15	15
Anxietyb	12	12.1
Impulsive-Compulsivec	2	2.5
ADHD symptom persistence
Persistent ADHD symptoms	34	32.1
Desisting ADHD symptoms	72	67.9
Psychotropic Medications
Stimulants	3	2.8
Antidepressants/Antianxiety	13	12.3
Smoker Status
Non Smokers	38	35.9
Smokers who successfully quit	28	26.4
Smokers who failed to quit	33	31.1
Smokers who haven’t tried to quit	7	6.6
	M	SD
Age, years	33.5	3.4
Intelligence Quotient	100.9	12.9
ADHD symptom severity mean scores
Inattention	0.93	0.7
Impulsivity/Hyperactivity	0.99	0.7
Quit Attempts
Smokers who successfully quit	5.5	4
Smokers who failed to quit	6.2	3.7

Note.

^aIncludes: Bipolar (I, II, Other), Major Depressive Disorder, Dysthymic lifetime, Not Otherwise Specified (NOS), General Medical Condition (GMC), Substance-Induced.

^bIncludes: Panic, Agoraphobia, Social Phobia, Obsessive Compulsive Disorder, Post Traumatic Stress Disorder, Generalized Anxiety Disorder, GMC, Substance-Induced, NOS.

^cIncludes: Gambling, Buying, Sexual, Internet, NOS.

Procedure

A one-day research visit included: saliva-based drug/alcohol screening; neuroimaging task practice; the neuroimaging protocol, and questionnaires. ADHD symptoms were assessed with the BAARS-IV (21). To address potential under-reporting, symptom severity scores made use of self- and collateral informant-report. As is typical, inattention and impulsivity were highly correlated (r = .78, P<.001) (22). To address multicollinearity and conserve power, we report on only impulsivity. Participants reported the number of lifetime-attempts to quit smoking, their current self-designation as smoker/non-smoker, and current smoking rate (“1=not smoking” to “8=two packs or more daily”). These data collectively identified successful quitters (26.4%). Performance on the Vocabulary and Matrix reasoning subtests of the Wechsler Abbreviated Scale of Intelligence were used to estimate each subject’s full-scale intelligence quotient (23). American Psychological Association ethical standards were followed in the conduct of this research, which was approved by the University of Pittsburgh institutional review board.

Neuroimaging Data Acquisition

Data were acquired on a Siemens 3T Trio with a 28-channel phased array coil. Anatomical images, field maps, and functional images during completion of the MI Go-NoGo task were acquired (24). See Supplemental Methods for further details.

MI Go-NoGo task

The MI Go-NoGo task (24) consisted of four blocks (1 Neutral, 3 Reward), executed during separate fMRI runs (see Figure 1). Each block contained three trial types (120 Go trials, 20 Go-Money trials, and 20 NoGo trials). All stimuli were double-digits numbers. Go trials consisted of white different-digit numbers (e.g. 15, 12, but not 22) in the center of a black background presented for 750 ms, followed by a 1,250 ms interstimulus interval (ISI) during which a black background was presented. Participants were asked to respond to Go trials with a single button press as quickly as possible. Go-Money trials consisted of same-digit numbers (e.g., 11 and 22) and also required participants to make a single button press as quickly as possible. Two same-digit numbers (e.g., 22 and 44) were presented as Go-Money trial stimuli for each block. Immediate monetary rewards for Go-Money trials were paid in proportion to reaction time (RT) during the trial. For example, if a participant’s reaction time was 400 ms, the monetary gain would be (1000 – 400)/50) = 12 cents. A maximum value of 20 cents was applied for immediate monetary rewards. Go-Money trials were presented for 750 ms, followed by a feedback screen for 750 ms and a blank-screen ISI (500 ms). Feedback screens for Go-Money (and for NoGo trials are described below) indicated response success or failure and the associated monetary reward.

Figure 1. The Monetary Incentive Go-NoGo task.

The Monetary Incentive Go-NoGo task requires subjects to respond as fast as possible to different-digit double-digit numbers (Go-trials), respond immediately to same-digit numbers (Go-Money trials), and inhibit responding to a prespecified double-digit number (No-Go trials). A) An example of the neutral condition, in which subjects are rewarded (e.g., $0.20) immediately only for correctly responding to Go-Money trials, commensurate with their reaction times. B) An example of the reward condition, in which subjects are rewarded immediately for responding to Go-Money trials (e.g., $0.20), rewarded at the end of the block based on the number of correct consecutive response inhibitions during NoGo trials (larger delayed reward, e.g., $3.60), which consist of withholding the button press to a NoGo target that was previously rewarded for fast and correct button press on Go-Money trials (e.g., 22, red arrow) but also rewarded immediately ($0.20) for failed inhibition during NoGo trials commensurate with their reaction times.

During NoGo trials, participants were asked to withhold responding upon presentation of a NoGo stimulus. There were two different types of NoGo trials (NoGo Neutral and NoGo Reward) that varied according to experimental block (Neutral vs. Reward). The NoGo stimulus was presented for 750 ms, followed by a 1250 ms ISI, a 1000 ms feedback screen and another 1000 ms ISI. At the beginning of each block, an instruction slide informed participants which double-digit number was the NoGo stimulus for that block. NoGo Neutral trials consisted of same-digit numbers (e.g., 55, 77 selected from Go trials in a practice version of the task administered before scanning). No monetary reward was applied during NoGo Neutral trials, such that only Go-Money trials were rewarded in the Neutral condition.

During the Reward blocks, NoGo stimuli were selected from Go-Money trials in the preceding block. These Go-Money trials were designed to prime a Go response through a learned association between responding quickly to the designated same-digit numbers and immediate reward; monetary reward was received for rapid responses (e.g., if 22 was selected as a NoGo-Reward-stimuli, 22 was a Go-Money trial in the preceding block). Rather than receiving an immediate reward for successful inhibition, inhibition of NoGo Reward trials resulted in delayed monetary reward at the end of each block based upon the highest number of consecutive successfully inhibited NoGo Reward trials during the block, multiplied by 40 cents. Consequently, immediate feedback for successful inhibitory control of individual NoGo Reward trials informed participants that they were successful in inhibiting but that no money had yet been accrued.

Failure to inhibit during a NoGo Reward trial led to an immediate reward, proportional to the reaction time (using the same formula as for Go-Money trials). Thus, reward-response associations in NoGo Reward trials were cultivated in two ways: 1) NoGo Reward stimuli were selected from the Go-Money trial stimuli that had been employed in the previous block, wherein these stimuli had received monetary rewards for rapid responses. 2) Failure to exert control over NoGo Reward trials resulted in small monetary rewards for which immediate feedback was provided, whereas successful inhibitory control resulted in no immediate reward, but a larger delayed reward comprising the sum of the longest run of consecutive successful inhibitions.

At the beginning of each block, an instruction slide informed participants about the block condition (Neutral vs. Reward) and what numbers were selected as Go-Money and NoGo trials. A neutral block preceded three consecutive blocks of reward. Trial types were presented in a fixed, pseudo-random order. Jittering events of interest (Go-Money and NoGo trials) was accomplished by including a pseudo-random number of Go trials (0 – 6 Go trials; 0 – 12 s) between these trials. Total duration of the MI-Go/NoGo task was approximately 25 minutes. Participants practiced the MI-Go/NoGo task in two blocks before MRI scanning, such that they completely understood the task conditions and reward contingencies.

Increased caution in responding to the Reward Go-Money trials, i.e., slower reaction times, may reflect the use of proactive inhibitory control to increase accuracy on Reward-NoGo trials to maximize end of block rewards. Slower reaction times would enable greater discrimination of the same same-digit numbers and provide time to resist the effects of the primed Go response learned previously.

Behavioral and Neuroimaging Data Analysis

Go trials were modeled as the implicit baseline. Go-Money and NoGo events for both conditions were convolved with a single gamma hemodynamic response function. Activity related to trial errors, feedback screens, demeaned motion parameters, their temporal derivatives, and motion outliers (>5mm and >3SD shifts in intensity) identified using ArtRepair were also modeled as additional regressors to avoid contamination of baseline and event-related data. A one-sample t-test in the entire sample was conducted for the All-NoGo (Reward-NoGo + Neutral-NoGo correct trials) versus Go contrast using AFNI’s 3dttest++ using a mask to limit the analysis space to the combination of brain regions-of-interest (ROIs) implicated in proactive and reactive inhibitory control based on the reviewed literature. The ROIs were created using AFNIs whereami function and included: the thalamus, GPi, GPe, putamen, caudate, STN, Brodmann area 9 & Brodmann area 46 (DLPFC), Brodmann area 4 (Primary Motor Cortex), the precentral gyrus, postcentral gyrus, and the SMA. In addition, an ROI in the rIFG was created by placing a 10 mm sphere around coordinates x=38, y=38, z=2 (8). Type-I Error was controlled using a False Discovery Rate (FDR) correction, q =.0001. Average beta parameters from clusters identified as being significantly active from within the mask (see Table 2) were extracted for both the Reward-NoGo and Neutral-NoGo conditions to evaluate associations with ADHD symptoms and for use in mediational and regression analyses.

Table 2.

Regions-of-interest showing more activation during All NoGo vs. Go trials

							Beta Estimates
			Peak MNI Coordinates				Reward-NoGo		Neutral-NoGo
Region	BA	Volume	X	Y	Z	Max t-value	M	SD	M	SD	t	P	d
DLPFC	46	22	−44	30	18	5.15	0.86	3.22	0.65	1.49	−0.68	.500	0.06
Putamen		36	−28	−4	0	5.49	0.61	1.42	0.11	0.83	−3.50	.001	0.31
GPe/GPi		21	−24	−14	2	4.86	0.42	1.01	0.10	0.59	−3.16	.002	0.27
Thalamus		29	−16	−22	4	5.78	0.47	1.16	0.16	0.70	−2.69	.008	0.23
M1	4	1100	−36	−26	52	12.49	1.10	1.69	0.53	0.87	−3.71	<.001	0.32
OP1		75	−48	−20	22	6.26	--	--	--	--	--	--	--
IFG	9	72	−60	6	26	6.42	--	--	--	--	--	--	--
DMPFC	9	43	−10	58	28	6.37	--	--	--	--	--	--	--

Note. Type-1 Error controlled using FDR q <.0001, df=137 for all Reward-NoGo vs. Neutral-NoGo contrasts. DLPFC = Dorsolateral Prefrontal Cortex; GPe/GPi=Globus Pallidus Pars Externa/Interna; M1=Primary Motor Cortex; OP1 = Operculum 1; IFG = Inferior Frontal Gyrus; DMPFC = Dorsal Medial Prefrontal Cortex. MNI = Montreal Neurological Institute.

Using SAS version 9.4, we examined whether impulsivity was differentially associated with outcomes of interest as a function of task condition by evaluating main effects of condition and impulsivity, along with condition x impulsivity interactions using a mixed effects framework. Simple slopes were estimated to probe the interaction. The Type-I Error rate was controlled using FDR corrections (25).

Mediation analyses were conducted using the process macro (26), which uses a bootstrapping method with bias-corrected confidence intervals to test the significance of each indirect effect, with significant mediation indicated by a confidence interval that does not contain 0. Bootstrap analyses and estimates were based on 10,000 bootstrap samples. Our model evaluates whether (a) impulsivity accounts for variation in Reward-Go-Money RT, which accounts for variation in inhibition accuracy; and (b) impulsivity accounts for variation in pallidum activation, which in turn accounts for variation in thalamic activation, which in turn accounts variation in M1 activation, which explains variation in inhibition accuracy.

Robust regression analyses were conducted to examine the association between GPb activation and the number of attempts need to successfully quit smoking. This analysis was limited to probands who successfully quit smoking because the number of quit attempts in unsuccessful quitters could reflect a multitude of processes un-related to inhibitory control (e.g., low motivation to quit).

Results

Task-related Analyses

Behavioral

There were no differences between the Reward-NoGo and the Neutral-NoGo conditions in Go accuracy, Go-Money accuracy, and NoGo accuracy (see Table S1). Participants had faster Go RTs and Go-Money RTs in the Reward-NoGo condition relative to the Neutral-NoGo condition.

Neuroimaging

All NoGo vs. Go contrasts were associated with greater activation in the DLPFC, putamen, GPe/GPi (overlapping with the GPb), thalamus, and M1, i.e., regions indicating recruitment of proactive inhibitory control. There was greater activation in the putamen, GPe/GPi, thalamus, and Ml during Reward-NoGo vs. Neutral-NoGo conditions (see Table 2, Figure 2). Subsequent analyses focused on regions that demonstrated modulation by reward.

Figure 2. Task activation associated with inhibitory control.

The top panel shows brain regions in the proactive inhibitory control network that were active during the All NoGo > Go contrast. The lower graph shows the difference in activation for the identified regions as a function of reward. DLPFC = Dorsolateral Prefrontal Cortex; GPe/GPi=Globus Pallidus Pars Externa/Interna; M1=Primary Motor Cortex; Put = Putamen; Thal = Thalamus. R=Right Hemisphere.

Primary Analyses

Hypothesis 1

See Table 3 for full statistical models. As shown in Figure 3, within probands, impulsivity was negatively correlated with inhibition accuracy across conditions, b=−5.3, SE=1.4, CI −1.7 to −9.0. Impulsivity was negatively correlated with Reward-Go-Money RT across conditions b=−13.7, SE=6.9, CI −0.1 to −27.3. The impulsivity x condition interactions were not significant.

Table 3.

Associations between impulsivity/hyperactivity symptoms and behavioral performance and brain activation as a function of Reward-NoGo vs. Neutral-NoGo conditions.

	Accuracy			Go-Money RT			Putamen
Effect	b	SE	P	b	SE	P	b	SE	P
Intercept	86.1	2.4	<.001	412.6	8.7	<.001	−0.01	0.20	0.954
Condition	−1.3	3.4	.708	20.3	5.9	<.001	0.76	0.28	0.009
Impulsivity	−4.0	2.0	.045	−10.3	7.3	0.161	0.13	0.16	0.441
Impulsivity x Condition	−2.6	2.8	.356	−6.8	4.9	0.170	−0.27	0.24	0.255
Simple Slopes
Irrespective of Condition	−5.3	1.4	<0.001	−13.7	6.9	0.049	−0.01	0.11	0.940
Reward	−6.6	2.0	0.001	−10.3	7.3	0.161	−0.14	0.16	0.382
Neutral	−4.0	2.0	0.045	−17.1	7.3	0.021	0.13	0.16	0.441
	GPe/GPi			Thalamus			M1
Effect	b	SE	P	b	SE	P	b	SE	P
Intercept	−0.02	0.13	0.885	0.17	0.16	0.290	0.35	0.22	0.110
Condition	0.77	0.18	<0.001	0.59	0.22	0.010	1.33	0.28	<0.001
Impulsivity	0.09	0.11	0.398	−0.03	0.14	0.821	0.14	0.18	0.433
Impulsivity x Condition	−0.45	0.15	0.005	−0.33	0.19	0.083	−0.69	0.23	0.004
Simple Slopes
Irrespective of Condition	−0.13	0.08	0.119	−0.19	0.10	0.048	−0.20	0.14	0.163
Reward	−0.35	0.11	0.002	−0.36	0.14	0.009	−0.54	0.18	0.004
Neutral	0.09	1^0.11	0.398	−0.03	0.14	0.821	0.14	0.18	0.433

Note. Putamen q(FDR)=0.255; GPe q(FDR)=.009; Thalamus q(FDR)=. 111; M1 q(FDR)=.009. df=104 for all tests. t-statistic = b/SE. GPe/GPi = Globus Pallidus Pars Externa/Interna; M1=Primary Motor Cortex. RT = Reaction Time.

Figure 3. Behavioral performance and brain regions associated with impulsivity/hyperactivity.

FDR=False Discovery Rate; GPe/GPi=Globus Pallidus Pars Externa/Interna; M1= Primary Motor Cortex.

Impulsivity x condition interactions predicting GPe/GPi activation, b=−0.45, SE=0.15, CI −0.14 to −0.75, and M1 activation, b=−0.69, SE=0.23, CI −0.22 to −1.15, survived multiple correction, qs(FDR)<.05 (see Figure 3). Impulsivity was more strongly negatively associated with GPe/GPi, b=−0.35, SE=0.11, CI −0.13 to −0.57, and M1, b=−0.54, SE=0.18, CI −0.18 to −0.91, activation in the Reward-NoGo versus the Neutral-NoGo condition (GPe/GPi: b=0.09, SE=0.11, CI −0.13 to 0.32; M1: b=0.14, SE=0.18, CI −0.22 to 0.51). The impulsivity x condition interaction for the putamen and thalamus were not significant. However, impulsivity was significantly negatively associated with thalamus activation in the Reward-NoGo condition, b=−0.36, SE=0.14, CI −0.09 to −0.63, and was not associated with thalamus activation in the Neutral-NoGo condition, b=−0.03, SE=0.14, CI −0.30 to 0.24.

Hypothesis 2

Impulsivity was not associated with Reward-Go Money RT, i.e., the mediator of the association between impulsivity and inhibition accuracy (see Figure 4A, Table 4). No statistical mediation was observed. Reward-Go-Money RT was positively associated with increased inhibition accuracy even after controlling for the effects of impulsivity.

Figure 4. Mediators of the association between impulsivity/hyperactivity symptoms and inhibition accuracy in the Reward-NoGo condition.

Panel A. Shows the mediation of the effect of impulsivity symptoms on inhibition accuracy through Go-Money reaction time, a behavioral marker of proactive cognitive control. Panel B. Show the serial mediation of the effect of impulsivity symptoms on inhibition accuracy through the pallidal-thalamic-motor pathway. Legend. Bolded lines reflect significant paths, numbers outside of parentheses are unstandardized beta weights, and number in parentheses are standard errors. *P< .05, **P< .005, ***p<.001, GPe/GPi=Globus Pallidus Pars Externa/Interna (GPe/GPi), Primary Motor Cortex (M1). The a paths are the associations between impulsivity and the mediator variables. The b paths are the associations between the mediator variables and inhibition accuracy. The d paths are the associations among the mediators. The c’ path is the direct effect of impulsivity on inhibition accuracy. The c path is the total effect of impulsivity symptoms on inhibition accuracy.

Table 4.

Mediation analyses of the effect of impulsivity/hyperactivity symptoms on inhibition accuracy by Go-Money reaction time

		Go-Money RT				Inhibition Accuracy
	Path	b	SE	P	Path	b	SE	P
Intercept	b0	412.6	8.53	.001	b0,1	59.5	11.0	<.001
Impulsivity	a	−10.3	7.13	.152	c’	−5.97	1.95	.003
Go-Money RT					b	0.06	0.03	.017
F, R2		2.08	0.02	.152		8.82	0.38	<.001
Indirect Path					Effect	B. SE	B. CI
Impulsivity → Go-Money RT → Accuracy					−0.66	0.59	−2.42	0.06

Note. The a paths are the associations between impulsivity and the mediator variables. The b paths are the associations between the mediator variable and inhibition accuracy. The c’ path is the direct effect of impulsivity on inhibition accuracy. RT= reaction time, B. SE = Bootstrapped Standard Error, B. CI = Bootstrapped confidence interval.

As shown in Figure 4B and Table 5, we evaluated a functionally and anatomically informed model designed to test the path by which impulsivity led to poor inhibition accuracy. The total effect of impulsivity on inhibition accuracy was b=−6.6, CI −2.7 to −10.6. We identified significant indirect effects of impulsivity on inhibition accuracy through GPe, thalamus, and M1 activation (indirect effect =−0.23, CI −0.04 to −0.76) and a shorter path from GPe/GPi to M1 (indirect effect =−0.45, CI −0.11 to −1.36). Impulsivity was associated with reduced GPe/GPi activation, which in turn was associated with greater activation of the thalamus, and subsequently the M1, resulting in greater inhibition accuracy.

Table 5.

Paths for serial mediation analyses of the effect of impulsivity/hyperactivity symptoms on inhibition accuracy via the pallidal-thalamic-motor pathway

	GPe/GPi				Thalamus				Motor				Accuracy
	b		SE	P	b		SE	P	b		SE	P	b		SE	P
Intercept	b0	0.75	0.16	<.001	b0,1	0.26	0.18	<.001	b0,2	1.05	0.27	<.001	b0,3	83.4	2.8	<.001
Impulsivity	a1	−0.35	0.13	.010	a2	−0.12	0.14	.376	a3	−0.25	0.21	.238	c1	−6.0	2.0	.004
GPe/GPi					d1,2	0.67	0.10	.008	d1,3	0.47	0.18	.008	b1	−1.8	1.8	.319
Thalamus									d2,3	0.37	0.14	.013	b2	−0.7	1.4	.618
M1													b3	2.7	1.0	.005
F, R2		6.87	0.06	.010		26.04	0.34	<.001		12.59	0.27	<.001		5.01	0.17	.001
Indirect Path											Effect	B. SE			B. CI
Total Indirect Effect											−0.61	0.91		−2.62	1.07
Impulsivity → GPe/GPi → Accuracy											0.62	0.77		−0.57	2.59
Impulsivity → GPe/GPi → Thalamus → Accuracy											0.17	0.27		−0.21	0.96
Impulsivity → GPe/GPi → M1 → Accuracy											−0.45	0.28		−1.36	−0.11	x
Impulsivity → GPe/GPi → Thalamus → M1 → Accuracy											−0.23	0.16		−0.76	−0.04	x
Impulsivity → Thalamus → Accuracy											0.09	0.20		−0.10	0.89
Impulsivity → Thalamus → M1 → Accuracy											−0.12	0.16		−0.64	0.06
Impulsivity → M1 → Accuracy											−0.67	0.69		−2.47	0.36

Note. x = P<.05; GPe/GPi=Globus Pallidus Pars Externa/Interna; M1=Primary Motor Cortex. t-statistic = b/SE. CI = confidence interval. B.SE = Bootstrapped Standard Error, B. CI = Bootstrapped confidence interval. The a paths are the associations between impulsivity and the mediator variables. The b paths are the associations between the mediator variable and inhibition accuracy. The c’ path is the direct effect of impulsivity on inhibition accuracy. The d path reflects the associations between the serial mediators.

Hypothesis 3

Reduced activation in the GPe/GPi was associated with more attempts to quit, robust b=−3.1; CI−1.80 to −4.46; χ²=21.22, P<.001, R² = 0.38, q(FDR)<.001 (see Figure S1; this effect held when examining all proband smokers who attempted to quit). No associations were observed between thalamus, M1, or putamen activation and the number of quit attempts (see Supplemental Results).

Covariate Analyses

Controlling for the presence of current anxiety disorder, antidepressant/anti-anxiety medication and stimulant use did not change the interpretation of the main findings (see Supplemental Results).

Secondary Analyses

Whole brain analyses and analyses incorporating inattention yielded similar results to main analyses (see Table S2, Figure S2 & S3). Categorical comparisons between a demographically similar comparison group from PALS without ADHD histories indicated that probands with persistent ADHD demonstrated decreased inhibition accuracy, as well as decreased GPe/GPi, and thalamus activation in the Reward-NoGo condition (see Supplemental Results).

Discussion

Several novel findings emerged from this study. First, adulthood impulsivity was associated with poorer inhibitory accuracy overall and reduced recruitment of brain regions comprising the indirect pallidal-thalamic-motor pathway when inhibiting a behavioral response primed by proximal reward for a longer-term payoff. Second, the negative association between impulsivity and successful inhibitory control was mediated by reduced recruitment of the pallidal-thalamic-motor pathway. Third, reduced GPe/GPi activation was associated with more attempts taken to quit smoking. Collectively, these findings provide new insights into possible reasons for continuing symptom-related impairments into adulthood, and their underlying neural basis.

We found evidence that impulsivity was associated with reduced recruitment of the indirect pallidal-thalamic-motor pathway in the context of reward-modulated inhibitory control. In particular, we observed less recruitment of the GPe/GPi extending into the GPb, the thalamus, and M1 among probands with elevated ADHD symptoms. GPe activation tracks the expected monetary reward value of engaging in a motor response (17); and sub-populations of GPb neurons increase their firing rate in response to cues signaling a reward and decrease their firing rate to non-reward predicting cues when rewards are obtained based on a specific motor movement (15, 16). Thus, one possible explanation for reduced indirect pallidum-motor pathway recruitment during reward-modulated inhibitory control may be that probands assigned a lower reward value to inhibition in response to the NoGo cue that signaled a larger delayed reward—resulting in a Go response and the receipt of an immediate reward. These results are consistent with evidence indicating that ADHD is associated with reduced activation in the NAcc during the anticipation of delayed rewards (27, 28) and with decreased activation in reward valuation regions (i.e., the caudate and thalamus) when choosing between delayed versus immediate rewards (29, 30). Comparisons to nonADHD controls provided evidence that GPe/GPi activation in the Reward-NoGo condition was below what would be expected in a nonADHD sample. Alternately, a failure to appropriately recruit the indirect pallidum-motor pathway during reward-modulated inhibitory control could also stem from deficits in setting and prioritizing among future action goals. We observed evidence that impulsivity was negatively associated with behavioral markers of proactive inhibitory control across conditions, and modest evidence that left DLPFC activation was reduced during reward-modulated inhibitory control (see Supplemental Results). These findings are consistent with research indicating that it is possible to ameliorate inhibitory control deficits in ADHD by increasing the excitability of the left DLPFC with transcranial direct current stimulation (31). We did not observe right hemisphere dominant DLPFC activation typically reported in Go/NoGo paradigms (32, 33). While right DLPFC activation has been implicated in monitoring ongoing goals and selective inhibition, left DLPFC activation has been implicated in task/criterion setting and goal internalization (33–35). The MI Go-NoGo task places strong demands on task/criterion setting and goal internalization as it requires participants to internalize multiple task criteria (i.e., Go vs. Go-Money vs. NoGo) in order to map them to the correct responses. This could explain why we observed greater left DLPFC on the MI Go-NoGo task than what has typically been reported in studies using the standard Go/NoGo task. In summary, our results suggest that brain regions involved in setting future action goals and the process of calculating the value of engaging in an action required to obtain delayed larger rewards over shorter-term rewards are both altered in those with persistent ADHD symptoms. However, a note of caution is warranted interpreting DLPFC-related findings as they were observed in secondary uncorrected analyses.

The results of our mediation analyses demonstrated that failure to recruit the indirect pallidal-thalamic-motor pathway is associated with decreased ability to inhibit behavior during reward-modulated inhibitory control. This is the first investigation to demonstrate that activation in the indirect basal ganglia motor pathway mediates inhibitory control success in probands with elevated ADHD symptoms. Importantly, the directionality of our proposed pallidal-thalamic-motor circuit is well constrained by the directionality of these pathways based on histology and probabilistic tractography (36, 37). The GPe is directly connected to the GPi and the GPi connects to the ventromedial (VM) and ventral lateral (VL) nuclei of the thalamus via the fasciculus thalamicus (36). Axonal projections from VM/VL nucleus of the thalamus terminate in layer 1 connecting to the premotor and M1 regions via the superior thalamic radiation (38). Our results are also consistent with research demonstrating greater connectivity within the indirect pathway is associated with successful inhibition of motor responses relative to inhibitory failures (39).

Given the robust literature on vulnerability to daily smoking in ADHD (19), we examined whether our neural findings on this task related to real-world behaviors that implicate similar inhibitory control processes. We found that reduced GPe/GPi recruitment during reward-modulated inhibitory control is associated with reporting more relapse occurrences prior to successfully quitting smoking. Given the putative role of the GPe/GPi our results suggest that failures of reward-modulated inhibitory control in the context of nicotine addiction in ADHD may stem from cues to abstain being imbued with lower reward-value relative to the immediate desire to curb the symptoms of withdrawal. This is consistent with findings indicating that individuals with low levels of basal ganglia activation to NoGo vs. Go trials are more likely to smoke when experiencing cravings (40). Future work should address whether difficulty recruiting the GPe/GPi in those with persistent ADHD symptoms is associated with the difficulties in engaging in other real-world activities that require inhibiting behaviors that yield immediate gratification for longer-term payoffs.

Limitations

Some probands were treated for ADHD symptoms for varying number of years, which could have affected the function of inhibitory control circuitry. The effects of the duration of treatment with psychostimulants on basal ganglia structure and function remains an unsettled question (41–47). The clinical impact of long-term medication exposure will be examined in future work. Last, our ability to generalize results to female probands is limited.

In conclusion, impulsivity symptom severity in adulthood, for individuals with ADHD rigorously diagnosed in childhood, are associated with deficits in pallidum activation that likely supports the computation of expected reward value signals used to facilitate or inhibit motor actions and to a lesser extent frontal regions implicated in setting and prioritizing among future action goals. Disruptions in the valuation process is associated with altered indirect pallidal-thalamic-motor pathway activation resulting in poor inhibitory control on basic tasks and during real-world behaviors. Devising interventions that can increase the reward value of cues that signal the need for proactive cognitive control may help improve inhibitory control and reduce engagement in costly behaviors.