Tolonen_manuscript_20240502_preprint-html.html

Title

Working memory related functional connectivity in adult ADHD and its amenability to training: A

randomized controlled trial

Authors and affiliations

Tuija Tolonen

1,2,3

, Sami Leppämäki

, Timo Roine

2,3

, Kimmo Alho

1,3

, Pekka Tani

, Anniina Koski

Matti Laine

, Juha Salmi

1,2,3

1. Department of Psychology and Logopedics, University of Helsinki, Helsinki, Finland.

2. Department of Neuroscience and Biomedical Engineering, Aalto University, Espoo, Finland.

3. AMI Centre, Aalto Neuroimaging, Aalto University, Espoo, Finland.

4. Terveystalo Healthcare, Helsinki, Finland.

5. Department of Psychiatry, Helsinki University Hospital, Helsinki, Finland.

6. Department of Psychology, Åbo Akademi University, Turku, Finland.

Corresponding author

Tuija Tolonen, M.A.

E-mail: tuija.tolonen@helsinki.fi

Address: Department of Psychology and Logopedics, P.O. BOX 21 (Haartmaninkatu 3), 00014

University of Helsinki, Helsinki, Finland

Keywords:

Adult ADHD, working memory, fMRI, network-based statistic, connectivity, cognitive

training

Abstract

Background:

Working memory (WM) deficits are among the most prominent cognitive impairments

in attention deficit hyperactivity disorder (ADHD). While functional brain connectivity in individuals

with ADHD has been studied, alterations in WM-related functional brain networks and their

malleability by cognitive training are not well known. We examined whole-brain functional

connectivity differences between adults with and without ADHD during

-back WM tasks and rest

at pretest, as well as the effects of WM training on functional and structural brain connectivity in the

ADHD group.

Methods:

Forty-two adults with ADHD and 36 neurotypical (NT) controls performed visuospatial

and verbal

-back tasks during functional magnetic resonance imaging (fMRI). In addition, seven-

minute resting state fMRI data and diffusion-weighted MR images were collected from all

participants. The adults with ADHD continued into a 5-week randomized controlled WM training

trial (experimental group training on a dual

-back task,

= 21; active control group training on

Bejeweled II video game,

= 21), followed by a posttraining MRI. The primary outcome, functional

brain connectivity, and the secondary outcome, structural brain connectivity, were examined with

Network-Based Statistic.

Results:

At the pretest, adults with ADHD had decreased functional connectivity compared with the

NT controls during both

-back tasks in networks encompassing fronto-parietal, temporal, occipital,

cerebellar, and subcortical brain regions. Furthermore, WM-related connectivity in widespread

networks was associated with performance accuracy in a continuous performance test. Regarding

resting state connectivity, no group differences or associations with task performance were observed.

Compared with the controls, WM training did not modulate functional or structural connectivity.

Conclusion:

Our results indicate large-scale abnormalities in functional brain networks underlying

deficits in verbal and visuospatial WM commonly faced in ADHD. Training-induced plasticity in

these networks may be limited.

Trial registration:

Retrospectively registered [registration still on-going in ISRCTN Registry].

Introduction

Attention deficit hyperactivity disorder (ADHD) is manifested as difficulties to maintain attention,

impulsive behavior, and hyperactivity (1). No reliable objective behavioral markers for ADHD have

been found, but deficits in working memory (WM) are suggested to be a key factor contributing to

the core symptoms (2). WM enables mental maintenance and manipulation of information and

supports many complex behaviors needed in everyday life (3). Poor WM is associated with severe

attention problems (4–6) and has widespread effects, for instance, on academic performance (7,8)

and organizational skills (9). Prevailing etiological theories suggest that ADHD symptoms emerge

from complex network-level aberrancies in brain structure and function (see, e.g., 10,11), but whole-

brain connectomics related to WM performance are still poorly understood. Another key issue to

address in cognitive neuroscience research is the malleability of abnormal networks to behavioral

interventions, as standard psychopharmaceutical treatments are sometimes ineffective and can have

unpleasant adverse effects (12,13).

Adults with ADHD have difficulties with both verbal and visuospatial WM (14). One of the most

common WM paradigms used in brain research (15–17) is the

-back, which requires continuous

updating of WM content by determining whether or not a current item in a stimulus stream matches

the one presented

items ago (18). For example, in a visuospatial 2-back task, a participant’s task is

to determine whether a current stimulus occurs in the same location as the stimulus presented just

before the preceding one.

-back tasks activate the prefrontal and parietal cortices, insula, cerebellum,

thalamus, and cingulate cortex (15–17), overlapping the brain areas of a more general WM network

(19). Activation patterns show considerable similarities across different WM task types (17,19).

However, the left inferior frontal gyrus has shown enhanced activity during verbal vs. nonverbal WM

and the supplementary motor area during nonverbal vs. verbal WM (19). Moreover, visuospatial WM

tasks have shown increased activity in the anterior insula and cerebellar tonsil compared with an

object identification WM task (17). The brain areas engaged in

-back tasks, especially the prefrontal

cortex and insula, are usually less active during

-back performance in individuals with ADHD than

in neurotypical (NT) participants (20,21). WM involves multiple cognitive component processes that

operate with spatial and verbal WM slave systems, relying on synchronized activation between

several different brain networks (3). Therefore, examining network-level brain systems can provide

important information about the mechanisms of WM as well as related dysfunctions.

Only a few studies have investigated WM-related network-level functional connectivity in ADHD

vs. NT individuals (22–26). Functional connectivity refers to synchronized activation patterns

between brain areas, determined as, for instance, temporal correlations between blood-oxygen level-

dependent (BOLD) signals measured in different brain areas with functional magnetic resonance

imaging (fMRI) (27). While regional activity is typically reduced in individuals with ADHD, most

WM-related functional connectivity studies conducted thus far have revealed stronger functional

connectivity in individuals with ADHD than in NT individuals in widespread networks, including the

prefrontal, temporal, parietal, occipital, insular, and cingulate cortices, as well as the cerebellum,

striatum, and globus pallidus (22–26). However, hypoconnectivity was also identified between

several brain areas in these studies (22,25,26). Moreover, all but one (25) of the studies on functional

connectivity changes in ADHD were conducted in children. Given that both functional connectivity

patterns (28) and ADHD symptoms evolve during childhood (29), it is difficult to make reliable

inferences from child studies on WM-related functional connectivity in adults with ADHD. In the

only study on adults, Wolf and colleagues (25) found reduced connectivity in ADHD in fronto-

parieto-cerebellar areas and the anterior cingulate cortex, and hyperconnectivity in many frontal areas

and the dorsal cingulate cortex. Another limitation in previous studies is that most of them have

investigated the functional connectivity of predefined areas or networks, possibly exaggerating their

significance or excluding other relevant areas from the analysis. Other approaches, such as Network-

Based Statistic (NBS) (30), a data-driven clustering method used to explore the connectivity

architecture of whole-brain networks with minimal a priori assumptions, might provide

complementary information on the complex functional connectivity changes in neurodevelopmental

disorders such as ADHD.

100

Due to the important role of WM deficits in the cognitive problems related to ADHD, WM training

101

was seen as a potential candidate for ADHD rehabilitation already almost two decades ago (31). Since

102

then, the behavioral effects of cognitive training on individuals with ADHD have been extensively

103

studied. These studies have shown cognitive training related moderate improvements in untrained

104

WM tasks, and some studies have even found evidence of transfer effects on inhibition control and

105

attention which are some of the core ADHD symptoms hampering everyday life (32–36). It is less

106

clear whether cognitive training is capable of remediating aberrant brain functioning in individuals

107

with ADHD. A few cognitive training studies on children and adolescents with ADHD have reported

108

volumetric increases from pretest to posttest in frontal and cerebellar gray matter (37) and regional

109

brain activity increases in frontal, parietal and temporal lobe regions and in the cerebellum (38,39)

110

and decreases in the insula and subcortical structures (40). Our previous results from the present

111

randomized controlled trial showed that WM training restored regional activation in the fronto-

112

parietal areas, precuneus, and supplementary motor area in adults with ADHD (41). More

113

specifically, we demonstrated that the training effects were different for the criterion task and for its

114

untrained variant: increased activity was observed for the training task, and decreased activity for the

115

near-transfer task (41). This may partly explain the mixed results found in previous studies employing

116

either a criterion task or a near-transfer task only. Besides regional effects, some studies with

117

neurotypical adults have investigated how cognitive training affects structural or functional

118

connectivity (for reviews, see 42,43). To highlight a few, studies using the NBS method have revealed

119

training-related increases in structural and functional connectivity in networks encompassing frontal,

120

temporal, and parietal areas (44,45). However, the network-level neural modulation effects of WM

121

training in individuals with ADHD have not yet been studied.

122

In the present study, we investigated differences in WM-related functional connectivity between

123

adults with and without ADHD, as well as the malleability of these differences in response to WM

124

training. We also examined brain connectivity during resting state to further explore whether possible

125

group differences during WM performance are explained by intrinsic brain connectivity. Functional

126

connectivity was analyzed with the NBS method in two ways, as group-level differences between

127

adults with ADHD and NT controls, and as correlations between functional connectivity, ADHD

128

symptoms, and continuous performance test (CPT) scores regardless of the participant’s diagnostic

129

status. CPT is the most widely studied objective behavioral marker of ADHD (46), and related

130

behavioral indices that are correlated with symptoms (47; however, see 48) were hence expected to

131

provide complementary information about aberrant functional connectivity patterns. Based on our

132

previous findings of ADHD-related reduced structural connectivity within the same sample (49), we

133

expected to find widespread hypoconnectivity in adults with ADHD compared with the NT controls,

134

although the findings of previous functional imaging studies were mixed (22–26). We also

135

hypothesized that higher error rates in the CPT would be associated with reduced functional

136

connectivity, also based on our previous findings of a corresponding association with structural

137

connectivity (49).

138

To examine the effects of WM training, the participants with ADHD took part in a randomized

139

controlled WM training trial, divided into an experimental group training

-back and an active control

140

group practicing a video game. Network-level training effects were analyzed by examining the

141

interaction between group (

-back training group vs. active control group) and session (pre- vs.

142

posttraining MRI). In addition to functional connectivity, WM training effects on structural

143

connectivity were also examined, building on our previous study of abnormal structural connectivity

144

in adults with ADHD (49). We hypothesized that WM training would lead to increased functional

145

connectivity, considering the training-related restoration of regional brain activity reported in the same

146

sample (41) and the previous connectivity studies in NT individuals (45,50–53). Although the

147

evidence of changes in structural connectivity resulting from WM training is a bit more limited

148

(44,54–56), the existing evidence for experience-induced neuronal plasticity (57) and its potential

149

implications prompted us to conduct pre-post comparisons also for the networks extracted from the

150

diffusion-weighted MR images.

151

Materials and Methods

152

Participants

153

Forty-two adults with ADHD and 36 NT controls participated in the study (see Supplementary Table

154

1 in Supplementary Results). The participants with ADHD were pre-screened and recruited by

155

psychiatrists at the Neuropsychiatry outpatient clinic of the Helsinki University Hospital and at

156

private clinics (Diacor Healthcare Services, Helsinki, Finland, and ProNeuron, Espoo, Finland).

157

Comorbid disorders were excluded as part of a regular clinical assessment with Interview for DSM-

158

IV Axis I Disorders (SCID-I) and the Mini-International Neuropsychiatric Interview (M.I.N.I.).

159

Exclusion criteria included any other severe psychiatric or neurological disorders apart from ADHD,

160

including head trauma requiring treatment, substance abuse or other addictions. The NT participants

161

were recruited via email lists at universities, polytechnics, adult high schools, and vocational schools,

162

and via the authors’ personal contacts. All participants were native Finnish speakers, had normal or

163

corrected-to-normal vision, had sufficient hearing, and met the MRI eligibility criteria. The NT adults

164

who participated only in the pretest (see Figure 1 for an overview of the study stages) were

165

compensated by 60 €, and the ADHD participants continuing to the training period and the posttest

166

by 240 €.

167

168

Figure 1

. Schematic overview of the study stages. The study consisted of three parts: a pretest

169

neuropsychological assessment and imaging session, a training period, and a posttest

170

neuropsychological assessment and imaging session. Pre-screening was conducted before the pretest.

171

The NT controls participated only in the pretest, while the participants with ADHD continued to the

172

training period and the posttest. Image adapted from (41), published under CC BY 4.0 license (58).

173

174

ADHD was diagnosed according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth

175

Edition (DSM-IV) and the participants’ current status was confirmed with the Conners’ Adult ADHD

176

Diagnostic Interview for DSM-IV (59). All participants with ADHD met the criteria for both

177

inattention and hyperactivity or only for inattention. Four participants had migraine, two had

178

hypothyroidism, and two had experienced mild epilepsy symptoms in childhood but had not needed

179

treatment since that time. Four participants had a prescription for migraine medicine, one for mild

180

depression (selective serotonin re-uptake inhibitor), and two for hypothyroidism. Thirty-three

181

participants with ADHD were using stimulants. Participants were requested to have a 24-h wash-out

182

period from psychostimulants before the pre- and posttests including the fMRI session.

183

Self-ratings and general abilities

184

ADHD symptoms were self-assessed with the Adult ADHD Self-Report Scale (ASRS) version 1.1

185

(60), mood with the Depression Scale (DEPS, 61), and alcohol consumption with the Alcohol Use

186

Disorders Identification Test – Consumption (62). General cognitive abilities were assessed with the

187

Matrix reasoning and Vocabulary tests of the Wechsler Adult Intelligence Scale (WAIS-III, 63).

188

fMRI conditions

189

Resting state.

During the resting state condition, participants were instructed to fixate their gaze to a

190

white cross presented in the center of a gray screen. Otherwise, there was no task for the participants

191

during the acquisition. The resting state run lasted for approximately seven minutes.

192

Single n-back tasks.

Two variants of single

-back tasks were presented during the fMRI session: a

193

version with visuospatial material (white squares presented in eight locations of a 3 x 3 matrix where

194

the fixation point occupied the center) and another with visually presented digits (from 1 to 9

195

presented at the center of the screen). In both task variants, the blocks consisted of four different

196

back levels (0-, 1-, 2-, and 3-back, one level in one block) presented in a counterbalanced order (0-

197

back, 1-back, 2-back, 3-back, 3-back, 2-back, 1-back, 0-back, etc.). All blocks began with an

198

instruction of the

-back level. Each location/digit was shown on the screen for 1500 ms with an

199

interstimulus interval (ISI) of 450 ms. Two response buttons, match and nonmatch, were used at load

200

levels higher than 0-back. In the 0-back task, the participants pressed the nonmatch button for each

201

stimulus. Each block consisted of 10 trials, of which randomly two on average were match trials

202

(targets). Each task level was presented five times, and the duration of the task was approximately 13

203

min for each task variant.

204

Offline cognitive tasks

205

Dual n-back task.

Dual

-back task was conducted in the pretest (outside of the scanner) and was

206

used as the training task for the experimental group. During the task, visuospatial material (same as

207

in the visuospatial single

-back task; see above) was presented simultaneously with phonological

208

material (spoken Finnish syllables /dy/, /ki/, /le/, /n

æ/

, /p

/, /ro/, /su/, or /ta/). The squares and the

209

syllables were presented concurrently, with 500 ms stimulus presentation and 2,500 ms ISI. The

210

participants were instructed to indicate whether the latest stimulus, either a square location, a syllable,

211

or both, matched the one

trials back. They had to press the left key if the square location matched

212

the location

trials back, the right key if the syllable matched the syllable

trials back, and both keys

213

if both matched the items

trials back. The probabilities of matches and nonmatches were equal in

214

both tasks. Total task duration was approximately 25 minutes, with 20 blocks of sequences, each

215

containing 20 stimuli pairs. Task difficulty was continuously adjusted based on the participant’s

216

performance. The level of

was between one and nine in each block. The program increased

by 1

217

in the following block when 90% accuracy was reached and decreased by 1 if the number of correct

218

answers fell below 75% for either stimulus type. The starting level of each session was 2-back. A

219

result screen showing the highest level of

reached during the session and the number of blocks

220

completed for each

level was displayed at the end of the session.

221

Continuous performance test.

Continuous performance test (CPT) (64) was performed outside of

222

the scanner. We used the implementation of the Conners Continuous Performance Test 2 (65)

223

available in the Psychology Experiment Building Language (PEBL) toolbox (66). In this test, the

224

participants were presented with a series of letters and instructed to press the space bar for each letter

225

apart from the letter X, which occurred in 9.7% of the trials. There were 360 trials (letters) with fixed

226

alternating intervals (1000 ms, 2000 ms, and 4000 ms), and the duration of the task was approximately

227

14 min. We used two scores in this study: omission errors (failure to respond to a target letter) and

228

commission errors (responding to a nontarget-letter X).

229

MRI acquisition

230

MRI data were collected at the Advanced Magnetic Imaging Centre (Aalto University) using a

231

Siemens MAGNETOM Skyra 3 T scanner (Siemens Healthcare, Erlangen, Germany) mounted with

232

a 30-channel head coil. Functional runs were acquired using gradient-echo echo planar imaging with

233

the following parameters: repetition time (TR) 1.9 s, voxel matrix 64 × 64, slice thickness 3.0 mm,

234

and in-plane resolution 3.1 mm × 3.1 mm. The two single

-back tasks were conducted in different

235

runs, each including 408 volumes. The resting state run included 225 volumes. In addition to fMRI,

236

an anatomical scan with a T1-weighted (T1w) MPRAGE sequence was obtained for registration

237

purposes with the following parameters: TR 2.5 s, voxel matrix 256 × 256, slice thickness 1 mm, and

238

in-plane resolution 1 mm × 1 mm.

239

The tasks presented during the functional runs were projected with a 3-micromirror data projector

240

(Christie X3, Christie Digital Systems, Mönchengladbach, Germany) on a semitransparent screen

241

located behind the participants’ head. The distance to the screen was approximately 34 cm via a mirror

242

located above the eyes of the participant with a binocular field of view 24 cm.

243

Preprocessing of the imaging data

244

Results included in this paper come from preprocessing performed using fMRIPrep 22.1.0 (67;

245

RRID:SCR_016216), which is based on Nipype 1.8.5 (68; RRID:SCR_002502). Many internal

246

operations of fMRIPrep use Nilearn 0.9.1 (69; RRID:SCR_001362). The boilerplate text in this

247

section (except for Denoising) was automatically generated by fMRIPrep with the express intention

248

that users should copy and paste this text into their manuscripts unchanged. It is released under the

249

CC0 license.

250

Anatomical data preprocessing.

The T1w image was corrected for intensity non-uniformity (INU)

251

with N4BiasFieldCorrection (70), distributed with ANTs 2.3.3 (71; RRID:SCR_004757), and used as

252

T1w reference throughout the workflow. The T1w reference was then skull-stripped with a Nipype

253

implementation of the antsBrainExtraction.sh workflow (from ANTs), using OASIS30ANTs as target

254

template. Brain tissue segmentation of cerebrospinal fluid (CSF), white-matter (WM) and gray-matter

255

(GM) was performed on the brain-extracted T1w using fast (72; FSL 6.0.5.1:57b01774,

256

RRID:SCR_002823). Brain surfaces were reconstructed using recon-all (73; FreeSurfer 7.2.0,

257

RRID:SCR_001847), and the brain mask estimated previously was refined with a custom variation

258

of the method to reconcile ANTs-derived and FreeSurfer-derived segmentations of the cortical gray-

259

matter of Mindboggle (74; RRID:SCR_002438).

260

Functional data preprocessing.

First, a reference volume and its skull-stripped version were

261

generated using a custom methodology of fMRIPrep. Head-motion parameters with respect to the

262

BOLD reference (transformation matrices, and six corresponding rotation and translation parameters)

263

were estimated before any spatiotemporal filtering using mcflirt (75; FSL 6.0.5.1:57b01774). BOLD

264

runs were slice-time corrected to 0.911s (0.5 of slice acquisition range 0s-1.82s) using 3dTshift from

265

AFNI (76; RRID:SCR_005927). The BOLD time-series (including slice-timing correction) were

266

resampled onto their original native space by applying the transforms to correct for head-motion. The

267

BOLD reference was then co-registered to the T1w reference using bbregister (FreeSurfer) which

268

implements boundary-based registration (77). Co-registration was configured with six degrees of

269

freedom. Several confounding time-series were calculated based on the preprocessed BOLD:

270

framewise displacement (FD), DVARS, and two region-wise global signals. FD was computed using

271

two formulations following Power et al. (absolute sum of relative motions; 78) and Jenkinson et al.

272

(relative root mean square displacement between affines; 75). FD and DVARS were calculated for

273

each functional run, both using their implementations in Nipype (following the definitions by 78).

274

The two global signals were extracted within the CSF and the WM. The confound time series derived

275

from head motion estimates and global signals were expanded with the inclusion of temporal

276

derivatives and quadratic terms for each (79).

277

Denoising.

After preprocessing, further denoising was applied to the fMRI data, including high-pass

278

filtering using a discrete cosine filter with 128 s cutoff, regression of volumes identified as motion

279

outliers (frames that exceeded 0.5 mm FD or 1.5 standardized DVARS), and regression of the CSF

280

and WM global signals and the six head motion parameters with their temporal derivatives and

281

quadratic terms (4 CSF global signals, 4 WM global signals, and 24 head motion parameters in total;

282

79).

283

Functional network construction

284

Automatic cortical parcellation was performed according to the Destrieux atlas (80) in the

FreeSurfer

285

software package (81), and subcortical segmentation was conducted using the FIRST algorithm (82)

286

of the FSL toolbox (83). This resulted in a total of 164 regions of interest (ROIs) for each participant

287

in their native (T1w) space. Functional connectivity matrices were constructed with the

Nilearn

288

toolbox (84) by calculating pairwise Pearson correlations of detrended, standardized and denoised

289

average time series of each ROI. Finally, Fisher transformation was applied for each correlation

290

matrix.

291

Processing of diffusion weighted images and construction of the structural connectome

292

The acquisition and preprocessing of diffusion weighted images, tractography, and construction of

293

structural networks are reported in detail elsewhere (49). Briefly, we collected 64 diffusion weighted

294

(

= 1000 s/mm

) images, performed motion, eddy current, and susceptibility distortion correction

295

(85,86), conducted constrained spherical deconvolution (87) based anatomically-constrained

296

tractography (88) from the white matter-gray matter interface, filtered the tractogram using SIFT2

297

(89,90), and constructed the structural connectome using the same ROIs as in the present study

298

(80,82). All processing steps were performed with the FSL (91) and MRtrix3 toolboxes (92).

299

300

Intervention procedure

301

After the pretest, the participants were assigned to either the experimental group or the active control

302

group according to a lottery-based randomization (20 tickets for each group) conducted by an

303

experimenter who was about to supervise the training period. The participants were unaware of the

304

group to which they were assigned. Different experimenters supervised the pre- and posttest and

305

training sessions, keeping also the experimenters conducting the assessments and the imaging

306

sessions blinded to the participants’ group membership. During the training period, the participants

307

trained three times a week, once in the laboratory (Institute of Behavioural Sciences, University of

308

Helsinki) and twice at home, for five weeks. One training session lasted for approximately 25 min,

309

and the participants were required to finish at least 10 training sessions to be invited to the posttest.

310

A laptop computer was given to the participants to conduct the training sessions at home, but they

311

were also allowed to use their own computer.

312

Experimental group.

The experimental group trained on a dual

-back task that was the same as in

313

the pretest, except that each training session lasted for 20 blocks instead of 10. The program generated

314

a logfile for each session that was stored on the computer. The participants also completed a

315

questionnaire in which they reported the maximum

-level they achieved and their levels of

316

motivation and arousal during the training session on a 10-point Likert scale (0 = very low and 10 =

317

very high). Each participant who continued until posttest (see the section Participant exclusion and

318

study drop-outs) finished at least 11 training sessions, and 12 participants completed all 15 training

319

sessions (mean number of sessions = 14.00,

= 1.56).

320

Active control group.

The active control group played the Bejeweled II video game by PopCap

321

Games from 2004. Their task was to score points by swapping jewels with adjacent ones, creating

322

chains of same-colored jewels. This game was selected because it has low WM demands and it is

323

broadly appealing, and because it has been successfully used in prior studies (e.g., 93). The

324

participants recorded their highest scores in a training log after each session. All participants who

325

continued until posttest (see section Participant exclusion and study drop-outs) performed at least 11

326

training sessions, and 17 participants completed all 15 sessions (mean number of sessions = 14.65,

327

= 0.99).

328

Participant exclusion and study drop-outs

329

Some participants were excluded from further analyses to ensure data quality. One NT participant

330

was excluded due to unsuccessful cortical parcellation, resulting in 163 ROIs. Two ADHD

331

participants were excluded from the resting state analyses because the imaging field during MRI was

332

misplaced. Participants were also excluded from specific task analysis if their mean connectivity

333

(mean of all pairwise correlations) exceeded 3 standard deviations from the group mean in the task.

334

Furthermore, three ADHD participants (2 from the experimental group and 1 from the active control

335

group) did not continue the training period to the posttest. The final group sizes were as follows: 41

336

participants with ADHD and 35 NT participants for the pretest visuospatial

-back analyses; 41

337

participants with ADHD and 34 NT participants for the pretest digit

-back analyses; 37 participants

338

with ADHD and 33 NT participants for the pretest resting state analyses; 18 in the experimental and

339

19 in the active control group for the visuospatial

-back training effect analyses; 18 in the

340

experimental and 20 in the active control group for the digit

-back training effect analyses; 17 in the

341

experimental and 17 in the active control group for the resting state training effect analyses; and 19

342

in the experimental and 19 in the active control group for the structural connectivity training effect

343

analyses.

344

Statistical analyses

345

Normality inspections.

The distributions of the scalar clinical and demographic variables were tested

346

with the Shapiro-Wilk normality test and by visual inspection of the Q-Q plots. Variables determined

347

to be normally distributed (age, Vocabulary, Matrix reasoning, mood, alcohol consumption,

348

inattention, hyperactivity-impulsivity, and commission errors) were studied with independent

349

samples

-tests when comparing groups or with Pearson’s correlation when studying associations

350

between variables. Some scalar variables (age, Vocabulary, Matrix reasoning, and mood) were

351

slightly curved resulting in rejection of normality in the Shapiro-Wilk test, and the group differences

352

were double-checked with a nonparametric Mann-Whitney U-test. The results remained the same.

353

The omission error rates were not normally distributed and nonparametric tests (Mann-Whitney U-

354

test for group differences and Spearman’s

for associations between variables) were used for

355

analyses. Group differences in gender, handedness, and education level were analyzed with the χ

356

test.

357

NBS.

Network Based Statistic toolbox (30) was used to search for functional networks differentiating

358

ADHD and NT participants or associated with background variables, as well as to identify networks

359

with group × session interactions in the WM training trial. NBS is a data-driven method that uses

360

cluster-based thresholding for the correction of mass-univariate analyses. Briefly, subnetworks, or

361

clusters, are first constructed from interconnected links (ROI-to-ROI connections) that exceed a

362

primary threshold. Family-wise error (FWE) rate is then controlled by nonparametric permutation

363

testing. The FWE-corrected

-level is estimated as

the proportion of permutations in which the

364

random-data cluster size (extent) or magnitude of the effect (intensity) is as large or larger than in the

365

original data. Here, the group differences were studied with a one-sided

-test, associations with

366

background variables with Pearson’s correlations, and group × session interactions with repeated-

367

measures ANOVA. We used primary thresholds of 3.0, 3.5 and 4.0 and both extent and intensity as a

368

measure of the cluster size. 10 000 permutations were performed in each test round. Statistical

369

significance was set to



< .05.

370

Comparative reliability analyses

371

To ensure that our results are not bound to one denoising scheme (see the Denoising section), we

372

repeated all analyses with another denoising strategy. In this comparative analysis, movement outliers

373

were not regressed out. Instead, participants whose movement exceeded 3 mm/radians for any of the

374

six translational or rotational head movement parameters or whose mean FD exceeded 0.5 mm were

375

excluded from the analyses. Otherwise, the same denoising options were used. All main effects

376

remained similar, with small changes to the WM-related NBS networks. The results can be found in

377

the Supplementary Results (Supplementary Figures 1-2).

378

Brain network visualizations

379

The brain images were visualized with the

BrainNet Viewer

(94).

380

Results

381

Behavioral characteristics

382

There were no statistically significant differences between the ADHD and NT groups in the general

383

background variables (age, gender, handedness, general cognitive abilities, education level, mood or

384

alcohol consumption), but the ADHD group reported more inattention and hyperactivity symptoms

385

than the NT controls and made more commission errors in the CPT. See Supplementary Tables 1 and

386

2 in Supplementary Results for detailed characteristics regarding the sample used in the present study.

387

Group differences in functional connectivity

388

NBS identified a single network per WM task type (digit and visuospatial

-back) with lower

389

functional connectivity in the ADHD group compared with the NT group (Figure 2 & Supplementary

390

information). The networks were detected with both extent and intensity as a measure of network and

391

with all primary thresholds (

-values 3.0, 3.5, and 4.0), reducing in size with increasing thresholds

392

(results with

-values 3.0 and 4.0 in Supplementary Figure 3 in Supplementary Results). Both

393

networks encompassed prefrontal areas, the cingulate, temporal, parietal and occipital cortices,

394

cerebellum, and subcortical structures such as the dorsal striatum and globus pallidus. The digit

395

back task-related network also included the insula, amygdala, and thalamus. The digit

-back task-

396

related network included more edges than the visuospatial

-back task-related network. We found no

397

WM-related networks with increased connectivity in the ADHD group compared with the NT adults

398

(all

-values = 1.0). There were no networks with significant group differences in the resting state

399

condition (all

-values > .41).

400

401

Figure 2.

The brain networks with decreased functional connectivity in the ADHD group compared

402

with the NT group during working memory tasks. A) The network related to the visuospatial

-back

403

task included 25 edges and 23 nodes (primary threshold

= 3.5,

extent = .03,

intensity = .03,

404

FWE-corrected). B) The network related to the digit

-back task included 91 edges and 73 nodes

405

(primary threshold

= 3.5,

extent = .009,

intensity = .009, FWE-corrected). The networks were

406

identical with extent and intensity as a measure of network size.

407

Associations between functional connectivity and behavioral measures

408

NBS revealed a single network per WM task type (digit and visuospatial

-back) in which functional

409

connectivity was negatively correlated with the number of commission errors in the CPT (Figure 3

410

& Supplementary information). These two networks were detected with both extent and intensity as

411

a measure of the network size. Both networks were identified with primary thresholds of

= 3.0 and

412

= 3.5, and the network related to the visuospatial task also with

= 4.0, reducing in size with

413

increasing thresholds (results with

-values 3.0 and 4.0 in Supplementary Figure 4 in Supplementary

414

Results). Both networks included frontal, temporal, and parietal areas, and the ventral and dorsal

415

striatum. The visuospatial network included also occipital areas and the thalamus. There were no

416

statistically significant WM-related networks identified with the other background variables (

all

417

values > .17). We found no networks with statistically significant associations with resting state

418

connectivity (all

-values > .07).

419

420

Figure 3.

Networks in which WM-related functional connectivity was negatively associated with the

421

number of commission errors (CEs) in the CPT. A) The network related to the visuospatial

-back

422

task included 36 edges and 28 nodes (primary threshold

= 3.5,

extent = .02,

intensity = .02,

423

FWE-corrected

). B) The network related to the digit

-back task included 27 edges and 26 nodes

424

(primary threshold

= 3.5,

extent = .03,

intensity = .03, FWE-corrected). The two networks were

425

identical with extent and intensity as a measure of network size.

426

Cognitive training intervention

427

We found no networks with a statistically significant group × session interaction in either WM-related

428

or resting state functional connectivity or structural connectivity with NBS (all

-values > .09).

429

In addition to whole-brain analyses, we conducted a post-hoc analysis examining the training effects

430

on the networks differentiating the ADHD and NT groups. The training effects were analyzed using

431

a group × session design for each edge of the networks and the results were corrected for multiple

432

comparisons with FDR correction. For functional connectivity, we chose the 10% of edges with

433

strongest group differences from the

= 3.0 networks (see Supplementary information) to reduce the

434

number of separate analyses and to restrict the analyses to the most relevant edges regarding the group

435

differences. For structural connectivity, the whole

= 3.0 network was chosen for post-hoc analysis

436

(see Supplementary information; see 49 for further group results) due to its smaller size. These

437

analyses revealed no group × session interactions surviving the correction of multiple analyses at any

438

edge of the functional or structural networks.

439

Discussion

440

In the present study, we examined alterations in functional connectivity during WM performance and

441

resting state in adults with ADHD vs. NT controls and further investigated whether WM training

442

intervention influences the functional or structural network configuration in the ADHD group. In

443

accordance with our hypotheses, WM-related functional hypoconnectivity in adults with ADHD

444

compared with NT adults was observed in widespread brain networks. Furthermore, as we

445

hypothesized, we found that functional connectivity was associated with performance in an attention

446

task regardless of diagnostic status. However, contrary to our hypothesis, no effects of WM training

447

on functional connectivity were observed, and no intervention effects were detected in structural

448

networks either. These findings shed light on the neural basis of aberrant WM functioning and its

449

(limited) malleability to training in adult ADHD.

450

WM-related functional connectivity in adult ADHD: Group differences and associations with task

451

performance

452

We observed decreased functional connectivity during visuospatial and verbal n-back tasks in adults

453

with ADHD, which is in line with findings that individuals with this neurodevelopmental disorder

454

have difficulties in both WM domains (14). Hypoconnectivity during verbal WM task in adults with

455

ADHD vs. NT controls was also reported by Wolf and colleagues (25). In that study,

456

hypoconnectivity was found in areas that appeared also in the networks of the present study, namely

457

the ventral prefrontal cortex, anterior cingulate cortex, superior parietal lobule, and cerebellum,

458

although they also reported hyperconnectivity in the prefrontal cortex, dorsal cingulate cortex, and

459

cuneus. Decreased structural connectivity in overlapping networks was also observed in our previous

460

study on adult ADHD (49). In most previous studies that have focused on children with ADHD,

461

hyperconnectivity during WM performance has been a more common finding than hypoconnectivity

462

(22–24,26). However, WM performance evolves strongly across development (95), together with

463

neural maturation of important WM hubs, such as the prefrontal cortex (96). Brain maturation also

464

shapes connectivity patterns during WM performance (97), as well as during other cognitive

465

processes and rest (98), complicating comparisons between adult and child studies. Although direct

466

evidence of age-related changes in WM-related aberrant activity in individuals with ADHD is still

467

lacking, our results, together with the previous adult study (25), suggest that maturation in this

468

population can lead to the emergence of WM-related hypoconnectivity in adulthood. This might

469

reflect the parallel tendency of the symptom manifestation shifting from overregulation

470

(hyperactivity) to underregulation (inattention) from childhood to adulthood (99).

Overall, our

471

findings indicate that widespread aberrancies in functional networks underlie the challenges in verbal

472

and visuospatial WM that individuals with ADHD commonly encounter.

473

The networks showing WM-related group differences included several brain areas previously

474

associated with WM and ADHD. Brain areas regarded as core regions for WM, including the

475

prefrontal cortex and parietal lobules (15–17), figure regularly in findings of structural and functional

476

abnormalities in ADHD (20,21,100). Furthermore, fronto-striatal loops have been regarded as central

477

both in ADHD pathophysiology and WM performance (101,102), as have cerebro-cerebellar

478

pathways (103–105). Additionally, abnormal WM-related activation in the insula (20) and structural

479

alterations in the insula and cingulate cortex (100,106) in ADHD have also been found, with both

480

areas linked to WM performance (15–17). Finally, structural and functional differences in ADHD vs.

481

NT participants have been reported in the pre- and postcentral gyri and many temporal and occipital

482

areas (100,106).

483

In addition to examining differences between individuals with and without ADHD, we wanted to see

484

whether variance in the capability to sustain attention or in symptom severity is associated with WM-

485

related connectivity regardless of the participants’ diagnostic status. We found that the number of

486

commission errors in the CPT, presumably reflecting impulsivity and lapses of sustained attention,

487

was negatively associated with connectivity during both verbal and visuospatial WM. As the observed

488

group differences, these findings are in accordance with our previous study, where higher rates of

489

commission errors were associated with decreased structural connectivity (49). This link between

490

sustained attention and WM-related connectivity may be related to tight coupling of attention and

491

WM (107). The attention-related networks shared many brain areas with the ones showing group

492

differences, and, overall, the effects cannot be fully separated due to the significant group difference

493

in the measured error rate. Nevertheless, these results indicate that participants’ attention skills

494

modulate how brain areas are synchronized during WM performance.

495

In contrast to the results related to WM, we did not find connectivity differences between groups or

496

associations with background variables during resting state. Although resting-state connectivity is

497

regularly studied in ADHD, with some meta-analytic evidence of disrupted default-mode network

498

connectivity (10), convergence between individual studies appears to be limited (108). Our findings

499

of WM-related aberrant connectivity in ADHD, together with the lack of such differences in resting

500

state, suggest that task-based settings can reveal important additional aspects of the connectivity

501

divergences related to the cognitive domains with which individuals with ADHD struggle.

502

Training-related changes in WM-related functional connectivity in adult ADHD

503

Although accumulating evidence from behavioral studies suggests that the effects of WM training

504

are rather limited (109–111) and previous evidence of brain functioning malleability promoted by

505

WM training in ADHD partially comes from studies that did not have proper control groups (39), it

506

was nevertheless unexpected that we found no effects of WM training on functional or structural brain

507

connectivity. After all, regional training effects were reported in this same sample (41) and there are

508

studies in NT groups reporting modulations of brain connectivity as a response to WM training (42–

509

45). Several potential reasons could explain the discrepancy with previous findings of the malleability

510

of brain connectivity in NT groups. Many of these studies (50,51,55), including those using NBS

511

(44,45), have not used a control group or have used only a passive control group. Employing an active

512

control group has been observed to diminish cognitive training effects at the behavioral level (32),

513

and possibly the control group type might also have some effects at the neural level. Related to this,

514

it is important that the control group has similar expectations about the training effects as the

515

experimental group (112), as was the case in the present study (see 41). However, this has been rarely

516

reported in previous cognitive training studies examining brain connectivity. It is also possible that

517

the training effects could be observed at only some WM load levels, as was the case in one previous

518

study where the rather modest training effects were present only at the 2-back load level (52). Also

519

in the previous study reporting training effects in this sample (41), modulations of brain activity were

520

dependent both on the task load and the type of the n-back task (digit vs. spatial). An analysis

521

separating these factors was not applicable for the present study because functional connectivity

522

studies require a greater number of volumes (typically at least 6 min; 113) per condition to show

523

robust effects. The complexity of the underlying training effects, however, could only partially

524

explain the lack of effects in the present study as the model in the present study was similar to that in

525

the previous study. Regarding the NBS analysis, one could argue that the connectivity effects are too

526

small to be revealed by whole-brain analysis, which requires rigorous correction for multiple

527

comparisons. Again, this could only partially contribute to the limited training effects as we found no

528

training effects for single edges either. Yet another possibility is that the extent of WM training effects

529

is negatively affected by initial problems with related cognitive processes (for rich-get-richer effects

530

see, e.g., 114–116). However, there are also studies reporting the opposite effects (117). All in all,

531

our knowledge of training-induced malleability of brain functioning in ADHD remains rather limited,

532

calling for continued research.

533

Recently it has become evident that, in general, cognitive training effects are limited mostly to the

534

trained task and task settings that are close to the trained one (i.e., near transfer; 118). This was

535

apparent also in the current sample where WM training gains were observed predominantly in the

536

trained task and other WM tasks (41). Thus, the benefits of cognitive training in ADHD are most

537

probably primarily related to improved task performances in the cognitive domain that is being

538

targeted, such as WM. Nevertheless, training multiple cognitive processes simultaneously has led to

539

reduced inattention symptoms in meta-analyses (32,35), suggesting that multiprocess training might

540

be more efficient in rehabilitation (however, see 36). Also, examining the processes that lead to

541

cognitive enhancement, such as the role of self-generated or examiner-instructed strategies (e.g., 119)

542

could be a useful future direction. Given the limits of transfer, an even more beneficial approach

543

could be to directly train tasks that are difficult for individuals with ADHD in real life (120; see also

544

121,122).

545

Limitations

546

It is noteworthy that the present sample does not fully represent the ADHD population, as we

547

excluded participants with comorbid disorders that affect the majority of the population to avoid

548

potential confounding factors. Individuals with no comorbid disorders can be considered to represent

549

a subpopulation with fewer symptoms which potentially contribute to group differences in brain

550

connectivity. Additionally, the careful matching of background factors between the ADHD and NT

551

groups could diminish some of the group differences evident at the population level. Our sample was

552

not large enough to examine how phenotypic variability, including the gender and age of the

553

participants, which are known to contribute to symptom manifestation, may influence the findings. It

554

is also possible that long-term use of stimulant medication contributes to the observed results in a

555

way that cannot be disentangled due to the lack of treatment naive participants. Overall, ADHD is a

556

highly heterogeneous disorder, and it would be optimal to acquire larger samples for neuroimaging

557

studies to examine this variability. Moreover, the chosen whole-brain analysis might conceal more

558

subtle differences in smaller subnetworks, such as fronto-striatal loops, because of the need to correct

559

for multiple analyses. The cluster-based correction approach is also unable to identify connectivity

560

patterns that do not form a connected component if they are individually too small to survive multiple

561

analyses correction. Furthermore, the results could be at least partly driven by group-level differences

562

in WM performance (see 41), making it difficult to distinguish the effect of WM deficits from ADHD

563

as a whole in the present results.

564

Conclusions

565

The present study examined the neural underpinnings of WM deficits in adults with ADHD and tested

566

whether aberrant connectivity can be modified via systematic training performed at WM capacity

567

limits. As expected, functional connectivity was decreased in the adults with ADHD compared with

568

NT adults in widespread brain networks. The group differences observed in visuospatial and digit

569

WM tasks were partially overlapping. Unexpectedly, WM training that resulted in the modulation of

570

regional activity (group × session interaction) in this same sample (41), did not show up in the present

571

functional connectivity analysis, and training had no effect on structural brain networks either.

572

Altogether, the present study highlights that aberrancies in widespread networks underlie deficits

573

encountered in ADHD, in agreement with the view of ADHD as a brain dysconnectivity syndrome

574

but suggests that the possibility of ameliorating such network-level divergences by WM training can

575

be limited. Future research is called for, hopefully with more effective WM training paradigms and

576

larger samples which allow to account for the interindividual differences in training-related outcomes

577

that may be particularly prominent in heterogeneous neurodevelopmental disorders such as ADHD.

578

Declarations

579

Ethics approval and consent to participate

580

All participants gave their written consent to participate in the study. The study was reviewed and

581

approved by the Ethics Committee for Gynecology and Obstetrics, Pediatrics and Psychiatry of the

582

Helsinki and Uusimaa Hospital District.

583

Consent for publication

584

Not applicable.

585

Availability of data and materials

586

The data used in the present study cannot be made readily available due to participants’ privacy and

587

details in the ethical agreement for the study. If you wish to use the data for validation purposes,

588

please contact the corresponding author TT at tuija.tolonen@helsinki.fi.

589

Competing interests

590

The authors declare that they have no competing interests.

591

Funding

592

This study was funded by the Research Council of Finland (grant numbers for ML: 260276 & 323251;

593

for KA: 260054 & 297848; for JS: 325981 & 328954), Åbo Akademi University Endowment for the

594

BrainTrain project (for ML), Finnish Cultural Foundation (for TR), and Vilho, Yrjö, and Kalle

595

Väisälä Foundation of the Finnish Academy of Science and Letters (for TT).

596

Authors' contributions

597

JS, ML, KA, PT, and SL designed the experiment. SL, AK, and PT selected, recruited, and diagnosed

598

the patients. TT collected the data, processed the functional data, and performed all analyses. TR

599

processed the structural data. JS and TT wrote the manuscript, which was commented, complemented,

600

and agreed on by all authors.

601

Acknowledgements

602

We would like to thank the research assistants who took part in collecting the data (Katri Mikkola,

603

Elina Nakane, Iikka Yli-Kyyny, Nora Moberg), radiographers who helped with the MR imaging

604

(Marita Kattelus and Rami Kunnas), and all the participants in the study.

605

Additional materials

606

The following additional materials are provided together with the present article:

607

Supplementary Results: A PDF-format file (.pdf) containing supplementary tables and figures.

608

Supplementary information: An Excel-format file (.xls) containing node names and coordinates for

609

the parcellation, and individual edges and their test statistics for each network.

610

References

611

American Psychiatric Association. Diagnostic and statistical manual of mental disorders (5th

612

ed.). Arlington, VA: American Psychiatric Publishing; 2013.

613

Willcutt EG, Doyle AE, Nigg JT, Faraone SV, Pennington BF. Validity of the executive

614

function theory of attention-deficit/hyperactivity disorder: A meta-analytic review. Biol Psychiatry.

615

2005;57(11):1336–46.

616

D’Esposito M, Postle BR. The cognitive neuroscience of working memory. Annu Rev

617

Psychol. 2015;66:115–42.

618

Alloway TP, Gathercole SE, Kirkwood H, Elliott J. The cognitive and behavioral

619

characteristics of children with low working memory. Child Dev. 2009;80(2):606–21.

620

Kofler MJ, Rapport MD, Bolden J, Sarver DE, Raiker JS. ADHD and working memory: The

621

impact of central executive deficits and exceeding storage/rehearsal capacity on observed

622

inattentive behavior. J Abnorm Child Psychol. 2010;38(2):149–61.

623

Tillman C, Eninger L, Forssman L, Bohlin G. The relation between working memory

624

components and ADHD symptoms from a developmental perspective. Dev Neuropsychol.

625

2011;36(2):181–98.

626

Rogers M, Hwang H, Toplak M, Weiss M, Tannock R. Inattention, working memory, and

627

academic achievement in adolescents referred for attention deficit/hyperactivity disorder (ADHD).

628

Child Neuropsychol. 2011;17(5):444–58.

629

Alloway TP, Alloway RG. Investigating the predictive roles of working memory and IQ in

630

academic attainment. J Exp Child Psychol. 2010;106(1):20–9.

631

Kofler MJ, Sarver DE, Harmon SL, Moltisanti A, Aduen PA, Soto EF, et al. Working

632

memory and organizational skills problems in ADHD. J Child Psychol Psychiatry. 2018;59(1):57–

633

67.

634

10.

Sutcubasi B, Metin B, Kurban MK, Metin ZE, Beser B, Sonuga-Barke E. Resting-state

635

network dysconnectivity in ADHD: A system-neuroscience-based meta-analysis. World J Biol

636

Psychiatry. 2020;21(9):662–72.

637

11.

Faraone SV, Asherson P, Banaschewski T, Biederman J, Buitelaar JK, Ramos-Quiroga JA,

638

et al. Attention-deficit/hyperactivity disorder. Nat Rev Dis Primer. 2015;1:15020.

639

12.

Brinkman WB, Simon JO, Epstein JN. Reasons why children and adolescents with ADHD

640

stop and restart taking medicine. Acad Pediatr. 2018;18(3):273–80.

641

13.

Banaschewski T, Roessner V, Dittmann RW, Santosh PJ, Rothenberger A. Non-stimulant

642

medications in the treatment of ADHD. Eur Child Adolesc Psychiatry. 2004;13 Suppl 1:I102-I116.

643

14.

Alderson RM, Kasper LJ, Hudec KL, Patros CHG. Attention-deficit/hyperactivity disorder

644

(ADHD) and working memory in adults: A meta-analytic review. Neuropsychology.

645

2013;27(3):287–302.

646

15.

Yaple Z, Arsalidou M. N-back working memory task: Meta-analysis of normative fMRI

647

studies with children. Child Dev. 2018;89(6):2010–22.

648

16.

Yaple Z, Stevens WD, Arsalidou M. Meta-analyses of the n-back working memory task:

649

fMRI evidence of age-related changes in prefrontal cortex involvement across the adult lifespan.

650

NeuroImage. 2019;196:16–31.

651

17.

Wang H, He W, Wu J, Zhang J, Jin Z, Li L. A coordinate-based meta-analysis of the n-back

652

working memory paradigm using activation likelihood estimation. Brain Cogn. 2019;132:1–12.

653

18.

Kirchner WK. Age differences in short-term retention of rapidly changing information. J

654

Exp Psychol. 1958;55(4):352–8.

655

19.

Rottschy C, Langner R, Dogan I, Reetz K, Laird AR, Schulz JB, et al. Modelling neural

656

correlates of working memory: A coordinate-based meta-analysis. NeuroImage. 2012;60(1):830–

657

46.

658

20.

Cortese S, Kelly C, Chabernaud C, Proal E, Di Martino A, Milham MP, et al. Toward

659

systems neuroscience of ADHD: A meta-analysis of 55 fMRI studies. Am J Psychiatry.

660

2012;169(10):1038–55.

661

21.

McCarthy H, Skokauskas N, Frodl T. Identifying a consistent pattern of neural function in

662

attention deficit hyperactivity disorder: A meta-analysis. Psychol Med. 2014;44(4):869–80.

663

22.

Bédard ACV, Newcorn JH, Clerkin SM, Krone B, Fan J, Halperin JM, et al. Reduced

664

prefrontal efficiency for visuospatial working memory in attention-deficit/hyperactivity disorder. J

665

Am Acad Child Adolesc Psychiatry. 2014;53(9):1020-1030.e6.

666

23.

Kowalczyk OS, Cubillo AI, Criaud M, Giampietro V, O’Daly OG, Mehta MA, et al. Single-

667

dose effects of methylphenidate and atomoxetine on functional connectivity during an n-back task

668

in boys with ADHD. Psychopharmacology (Berl). 2023;240(10):2045–60.

669

24.

Massat I, Slama H, Kavec M, Linotte S, Mary A, Baleriaux D, et al. Working memory-

670

related functional brain patterns in never medicated children with ADHD. PloS One.

671

2012;7(11):e49392.

672

25.

Wolf RC, Plichta MM, Sambataro F, Fallgatter AJ, Jacob C, Lesch KP, et al. Regional brain

673

activation changes and abnormal functional connectivity of the ventrolateral prefrontal cortex

674

during working memory processing in adults with attention-deficit/hyperactivity disorder. Hum

675

Brain Mapp. 2009;30(7):2252–66.

676

26.

Wu ZM, Bralten J, An L, Cao QJ, Cao XH, Sun L, et al. Verbal working memory-related

677

functional connectivity alterations in boys with attention-deficit/hyperactivity disorder and the

678

effects of methylphenidate. J Psychopharmacol Oxf Engl. 2017;31(8):1061–9.

679

27.

Rogers BP, Morgan VL, Newton AT, Gore JC. Assessing functional connectivity in the

680

human brain by FMRI. Magn Reson Imaging. 2007;25(10):1347–57.

681

28.

Grayson DS, Fair DA. Development of large-scale functional networks from birth to

682

adulthood: A guide to neuroimaging literature. NeuroImage. 2017;160:15–31.

683

29.

Franke B, Michelini G, Asherson P, Banaschewski T, Bilbow A, Buitelaar JK, et al. Live

684

fast, die young? A review on the developmental trajectories of ADHD across the lifespan. Eur

685

Neuropsychopharmacol. 2018;28(10):1059–88.

686

30.

Zalesky A, Fornito A, Bullmore ET. Network-based statistic: Identifying differences in

687

brain networks. NeuroImage. 2010;53(4):1197–207.

688

31.

Klingberg T, Fernell E, Olesen PJ, Johnson M, Gustafsson P, Dahlström K, et al.

689

Computerized training of working memory in children with ADHD – A randomized, controlled

690

trial. J Am Acad Child Adolesc Psychiatry. 2005;44(2):177–86.

691

32.

Cortese S, Ferrin M, Brandeis D, Buitelaar J, Daley D, Dittmann RW, et al. Cognitive

692

training for attention-deficit/hyperactivity disorder: Meta-analysis of clinical and

693

neuropsychological outcomes from randomized controlled trials. J Am Acad Child Adolesc

694

Psychiatry. 2015;54(3):164–74.

695

33.

Pauli-Pott U, Mann C, Becker K. Do cognitive interventions for preschoolers improve

696

executive functions and reduce ADHD and externalizing symptoms? A meta-analysis of

697

randomized controlled trials. Eur Child Adolesc Psychiatry. 2021;30(10):1503–21.

698

34.

Robledo-Castro C, Lerma-Castaño PR, Bonilla-Santos G. Effect of cognitive training

699

programs based on computer systems on executive functions in children with ADHD: A systematic

700

review. J Atten Disord. 2023;27(13):1467–87.

701

35.

Zou X, Yu F, Huang Q, Huang Y. The effect of cognitive training on children with attention

702

deficit and hyperactivity disorder: A meta-analysis. Appl Neuropsychol Child. 2024;0(0):1–10.

703

36.

Westwood SJ, Parlatini V, Rubia K, Cortese S, Sonuga-Barke EJS. Computerized cognitive

704

training in attention-deficit/hyperactivity disorder (ADHD): A meta-analysis of randomized

705

controlled trials with blinded and objective outcomes. Mol Psychiatry. 2023;28(4):1402–14.

706

37.

Hoekzema E, Carmona S, Ramos-Quiroga JA, Barba E, Bielsa A, Tremols V, et al.

707

Training-induced neuroanatomical plasticity in ADHD: A tensor-based morphometric study. Hum

708

Brain Mapp. 2011;32(10):1741–9.

709

38.

Hoekzema E, Carmona S, Tremols V, Gispert JD, Guitart M, Fauquet J, et al. Enhanced

710

neural activity in frontal and cerebellar circuits after cognitive training in children with attention-

711

deficit/hyperactivity disorder. Hum Brain Mapp. 2010;31(12):1942–50.

712

39.

Stevens MC, Gaynor A, Bessette KL, Pearlson GD. A preliminary study of the effects of

713

working memory training on brain function. Brain Imaging Behav. 2016;10(2):387–407.

714

40.

de Oliveira Rosa V, Rosa Franco A, Abrahão Salum Júnior G, Moreira-Maia CR, Wagner F,

715

Simioni A, et al. Effects of computerized cognitive training as add-on treatment to stimulants in

716

ADHD: A pilot fMRI study. Brain Imaging Behav. 2020;14(5):1933–44.

717

41.

Salmi J, Soveri A, Salmela V, Alho K, Leppämäki S, Tani P, et al. Working memory

718

training restores aberrant brain activity in adult attention‐deficit hyperactivity disorder. Hum Brain

719

Mapp. 2020;41(17):4876–91.

720

42.

Brooks SJ, Mackenzie-Phelan R, Tully J, Schiöth HB. Review of the neural processes of

721

working memory training: Controlling the impulse to throw the baby out with the bathwater. Front

722

Psychiatry. 2020;11:512761.

723

43.

Constantinidis C, Klingberg T. The neuroscience of working memory capacity and training.

724

Nat Rev Neurosci. 2016;17(7):438–49.

725

44.

Román FJ, Iturria-Medina Y, Martínez K, Karama S, Burgaleta M, Evans AC, et al.

726

Enhanced structural connectivity within a brain sub-network supporting working memory and

727

engagement processes after cognitive training. Neurobiol Learn Mem. 2017;141:33–43.

728

45.

Zuber P, Gaetano L, Griffa A, Huerbin M, Pedullà L, Bonzano L, et al. Additive and

729

interaction effects of working memory and motor sequence training on brain functional

730

connectivity. Sci Rep. 2021;11(1):23089.

731

46.

Gualtieri CT, Johnson LG. ADHD: Is objective diagnosis possible? Psychiatry Edgmont.

732

2005;2(11):44–53.

733

47.

Albrecht B, Uebel-von Sandersleben H, Wiedmann K, Rothenberger A. ADHD history of

734

the concept: The case of the continuous performance test. Curr Dev Disord Rep. 2015;2(1):10–22.

735

48.

Arrondo G, Mulraney M, Iturmendi-Sabater I, Musullulu H, Gambra L, Niculcea T, et al.

736

Systematic review and meta-Analysis: Clinical utility of continuous performance tests for the

737

identification of attention-deficit/hyperactivity disorder. J Am Acad Child Adolesc Psychiatry.

738

2024;63(2):154–71.

739

49.

Tolonen T, Roine T, Alho K, Leppämäki S, Tani P, Koski A, et al. Abnormal wiring of the

740

structural connectome in adults with ADHD. Netw Neurosci. 2023;7(4):1302–25.

741

50.

Takeuchi H, Taki Y, Nouchi R, Hashizume H, Sekiguchi A, Kotozaki Y, et al. Effects of

742

working memory training on functional connectivity and cerebral blood flow during rest. Cortex J

743

Devoted Study Nerv Syst Behav. 2013;49(8):2106–25.

744

51.

Jolles DD, van Buchem MA, Crone EA, Rombouts SARB. Functional brain connectivity at

745

rest changes after working memory training. Hum Brain Mapp. 2013;34(2):396–406.

746

52.

Thompson TW, Waskom ML, Gabrieli JDE. Intensive working memory training produces

747

functional changes in large-scale frontoparietal networks. J Cogn Neurosci. 2016;28(4):575–88.

748

53.

Sánchez-Pérez N, Inuggi A, Castillo A, Campoy G, García-Santos JM, González-Salinas C,

749

et al. Computer-based cognitive training improves brain functional connectivity in the attentional

750

networks: A study with primary school-aged children. Front Behav Neurosci. 2019;13:247.

751

54.

Salminen T, Mårtensson J, Schubert T, Kühn S. Increased integrity of white matter

752

pathways after dual n-back training. NeuroImage. 2016;133:244–50.

753

55.

Takeuchi H, Sekiguchi A, Taki Y, Yokoyama S, Yomogida Y, Komuro N, et al. Training of

754

working memory impacts structural connectivity. J Neurosci. 2010;30(9):3297–303.

755

56.

Metzler-Baddeley C, Foley S, de Santis S, Charron C, Hampshire A, Caeyenberghs K, et al.

756

Dynamics of white matter plasticity underlying working memory training: Multimodal evidence

757

from diffusion MRI and relaxometry. J Cogn Neurosci. 2017;29(9):1509–20.

758

57.

Sampaio-Baptista C, Johansen-Berg H. White matter plasticity in the adult brain. Neuron.

759

2017;96(6):1239–51.

760

58.

CC BY 4.0 Deed. https://creativecommons.org/licenses/by/4.0/deed.en. Accessed 29 April

761

2024.

762

59.

Epstein J, Johnson DE, Conners CK. Conners’ Adult ADHD Diagnostic Interview for DSM-

763

IV™ (CAADID™) [Database record]. APA PsycTests.

764

60.

Kessler RC, Adler L, Ames M, Demler O, Faraone S, Hiripi E, et al. The World Health

765

Organization adult ADHD self-report scale (ASRS): A short screening scale for use in the general

766

population. Psychol Med. 2005;35(2):245–56.

767

61.

Salokangas RK, Poutanen O, Stengård E. Screening for depression in primary care.

768

Development and validation of the Depression Scale, a screening instrument for depression. Acta

769

Psychiatr Scand. 1995;92(1):10–6.

770

62.

Bush K, Kivlahan DR, McDonell MB, Fihn SD, Bradley KA. The AUDIT alcohol

771

consumption questions (AUDIT-C): An effective brief screening test for problem drinking.

772

Ambulatory Care Quality Improvement Project (ACQUIP). Alcohol Use Disorders Identification

773

Test. Arch Intern Med. 1998;158(16):1789–95.

774

63.

Wechsler D. WAIS-III-käsikirja [The handbook of the Finnish version of the 3rd edition of

775

the Wechsler Adult Intelligence Scale]. Helsinki, Finland: Psykologien Kustannus Oy; 2005.

776

64.

Rosvold HE, Mirsky AF, Sarason I, Bransome ED, Beck LH. A continuous performance test

777

of brain damage. J Consult Psychol. 1956;20(5):343–50.

778

65.

Conners CK & MHS Staff (Eds.). Conners’ Continuous Performance Test II: Computer

779

program for Windows. Technical guide and software manual. North Tonawanda, NY: Multi-Health

780

Systems; 2000.

781

66.

Mueller ST, Piper BJ. The Psychology Experiment Building Language (PEBL) and PEBL

782

Test Battery. J Neurosci Methods. 2014;222:250–9.

783

67.

Esteban O, Markiewicz CJ, Blair RW, Moodie CA, Isik AI, Erramuzpe A, et al. fMRIPrep:

784

A robust preprocessing pipeline for functional MRI. Nat Methods. 2019;16(1):111–6.

785

68.

Gorgolewski K, Burns C, Madison C, Clark D, Halchenko Y, Waskom M, et al. Nipype: A

786

flexible, lightweight and extensible neuroimaging data processing framework in Python. Front

787

Neuroinform. 2011;5:13.

788

69.

Abraham A, Pedregosa F, Eickenberg M, Gervais P, Mueller A, Kossaifi J, et al. Machine

789

learning for neuroimaging with scikit-learn. Front Neuroinform. 2014;8:14.

790

70.

Tustison NJ, Avants BB, Cook PA, Zheng Y, Egan A, Yushkevich PA, et al. N4ITK:

791

Improved N3 bias correction. IEEE Trans Med Imaging. 2010;29(6):1310–20.

792

71.

Avants BB, Epstein CL, Grossman M, Gee JC. Symmetric diffeomorphic image registration

793

with cross-correlation: Evaluating automated labeling of elderly and neurodegenerative brain. Med

794

Image Anal. 2008;12(1):26–41.

795

72.

Zhang Y, Brady M, Smith S. Segmentation of brain MR images through a hidden Markov

796

random field model and the expectation-maximization algorithm. IEEE Trans Med Imaging.

797

2001;20(1):45–57.

798

73.

Dale AM, Fischl B, Sereno MI. Cortical surface-based analysis. I. Segmentation and surface

799

reconstruction. NeuroImage. 1999;9(2):179–94.

800

74.

Klein A, Ghosh SS, Bao FS, Giard J, Häme Y, Stavsky E, et al. Mindboggling morphometry

801

of human brains. PLoS Comput Biol. 2017;13(2):e1005350.

802

75.

Jenkinson M, Bannister P, Brady M, Smith S. Improved optimization for the robust and

803

accurate linear registration and motion correction of brain images. NeuroImage. 2002;17(2):825–

804

41.

805

76.

Cox RW, Hyde JS. Software tools for analysis and visualization of fMRI data. NMR

806

Biomed. 1997;10(4–5):171–8.

807

77.

Greve DN, Fischl B. Accurate and robust brain image alignment using boundary-based

808

registration. NeuroImage. 2009;48(1):63–72.

809

78.

Power JD, Mitra A, Laumann TO, Snyder AZ, Schlaggar BL, Petersen SE. Methods to

810

detect, characterize, and remove motion artifact in resting state fMRI. NeuroImage. 2014;84:320–

811

41.

812

79.

Satterthwaite TD, Elliott MA, Gerraty RT, Ruparel K, Loughead J, Calkins ME, et al. An

813

improved framework for confound regression and filtering for control of motion artifact in the

814

preprocessing of resting-state functional connectivity data. NeuroImage. 2013;64:240–56.

815

80.

Destrieux C, Fischl B, Dale A, Halgren E. Automatic parcellation of human cortical gyri and

816

sulci using standard anatomical nomenclature. Neuroimage. 2010;53(1):1–15.

817

81.

Fischl B. FreeSurfer. NeuroImage. 2012;62(2):774–81.

818

82.

Patenaude B, Smith SM, Kennedy DN, Jenkinson M. A Bayesian model of shape and

819

appearance for subcortical brain segmentation. NeuroImage. 2011;56(3):907–22.

820

83.

Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TEJ, Johansen-Berg H, et

821

al. Advances in functional and structural MR image analysis and implementation as FSL.

822

NeuroImage. 2004;23 Suppl 1:S208-219.

823

84.

Nilearn. 2024. https://doi.org/10.5281/zenodo.8397156. Accessed April 29 2024.

824

85.

Andersson JLR, Sotiropoulos SN. An integrated approach to correction for off-resonance

825

effects and subject movement in diffusion MR imaging. NeuroImage. 2016;125:1063–78.

826

86.

Irfanoglu MO, Walker L, Sarlls J, Marenco S, Pierpaoli C. Effects of image distortions

827

originating from susceptibility variations and concomitant fields on diffusion MRI tractography

828

results. NeuroImage. 2012;61(1):275–88.

829

87.

Tournier JD, Calamante F, Connelly A. Robust determination of the fibre orientation

830

distribution in diffusion MRI: Non-negativity constrained super-resolved spherical deconvolution.

831

NeuroImage. 2007;35(4):1459–72.

832

88.

Smith RE, Tournier JD, Calamante F, Connelly A. Anatomically-constrained tractography:

833

Improved diffusion MRI streamlines tractography through effective use of anatomical information.

834

NeuroImage. 2012;62(3):1924–38.

835

89.

Smith RE, Tournier JD, Calamante F, Connelly A. SIFT: Spherical-deconvolution informed

836

filtering of tractograms. NeuroImage. 2013;67:298–312.

837

90.

Smith RE, Tournier JD, Calamante F, Connelly A. SIFT2: Enabling dense quantitative

838

assessment of brain white matter connectivity using streamlines tractography. NeuroImage.

839

2015;119:338–51.

840

91.

Jenkinson M, Beckmann CF, Behrens TEJ, Woolrich MW, Smith SM. FSL. NeuroImage.

841

2012;62(2):782–90.

842

92.

Tournier JD, Smith R, Raffelt D, Tabbara R, Dhollander T, Pietsch M, et al. MRtrix3: A

843

fast, flexible and open software framework for medical image processing and visualisation.

844

NeuroImage. 2019;202:116137.

845

93.

Soveri A, Karlsson EPA, Waris O, Grönholm-Nyman P, Laine M. Pattern of near transfer

846

effects following working memory training with a dual n-back task. Exp Psychol. 2017;64(4):240–

847

52.

848

94.

Xia M, Wang J, He Y. BrainNet Viewer: A network visualization tool for human brain

849

connectomics. PLoS ONE. 2013;8(7):e68910.

850

95.

Gathercole SE, Pickering SJ, Ambridge B, Wearing H. The structure of working memory

851

from 4 to 15 years of age. Dev Psychol. 2004;40(2):177–90.

852

96.

Fuster JM. The prefrontal cortex – An update: Time is of the essence. Neuron.

853

2001;30(2):319–33.

854

97.

van den Bosch GE, El Marroun H, Schmidt MN, Tibboel D, Manoach DS, Calhoun VD, et

855

al. Brain connectivity during verbal working memory in children and adolescents. Hum Brain

856

Mapp. 2014;35(2):698–711.

857

98.

Stevens MC. The contributions of resting state and task-based functional connectivity

858

studies to our understanding of adolescent brain network maturation. Neurosci Biobehav Rev.

859

2016;70:13–32.

860

99.

Vos M, Rommelse NNJ, Franke B, Oosterlaan J, Heslenfeld DJ, Hoekstra PJ, et al.

861

Characterizing the heterogeneous course of inattention and hyperactivity-impulsivity from

862

childhood to young adulthood. Eur Child Adolesc Psychiatry. 2022;31(8):1–11.

863

100.

Yu M, Gao X, Niu X, Zhang M, Yang Z, Han S, et al. Meta-analysis of structural and

864

functional alterations of brain in patients with attention-deficit/hyperactivity disorder. Front

865

Psychiatry. 2023;13:1070142.

866

101.

O’Reilly RC, Frank MJ. Making working memory work: A computational model of learning

867

in the prefrontal cortex and basal ganglia. Neural Comput. 2006;18(2):283–328.

868

102.

Cubillo A, Halari R, Smith A, Taylor E, Rubia K. A review of fronto-striatal and fronto-

869

cortical brain abnormalities in children and adults with attention deficit hyperactivity disorder

870

(ADHD) and new evidence for dysfunction in adults with ADHD during motivation and attention.

871

Cortex J Devoted Study Nerv Syst Behav. 2012;48(2):194–215.

872

103.

Stoodley CJ. The cerebellum and cognition: Evidence from functional imaging studies.

873

Cerebellum Lond Engl. 2012;11(2):352–65.

874

104.

Salmi J, Pallesen KJ, Neuvonen T, Brattico E, Korvenoja A, Salonen O, et al. Cognitive and

875

motor loops of the human cerebro-cerebellar system. J Cogn Neurosci. 2010;22(11):2663–76.

876

105.

Stoodley CJ. The cerebellum and neurodevelopmental disorders. Cerebellum Lond Engl.

877

2016;15(1):34–7.

878

106.

Norman LJ, Carlisi C, Lukito S, Hart H, Mataix-Cols D, Radua J, et al. Structural and

879

functional brain abnormalities in attention-deficit/hyperactivity disorder and obsessive-compulsive

880

disorder: A comparative meta-analysis. JAMA Psychiatry. 2016;73(8):815–25.

881

107.

Oberauer K. Working memory and attention – A conceptual analysis and review.

J Cogn

882

2019;2(1):36.

883

108.

Cortese S, Aoki YY, Itahashi T, Castellanos FX, Eickhoff SB. Systematic review and meta-

884

analysis: Resting-state functional magnetic resonance imaging studies of attention-

885

deficit/hyperactivity disorder. J Am Acad Child Adolesc Psychiatry. 2021;60(1):61–75.

886

109.

Melby-Lervåg M, Redick TS, Hulme C. Working memory training does not improve

887

performance on measures of intelligence or other measures of “far transfer”: Evidence from a meta-

888

analytic review. Perspect Psychol Sci J Assoc Psychol Sci. 2016;11(4):512–34.

889

110.

Sala G, Aksayli ND, Tatlidil KS, Tatsumi T, Gondo Y, Gobet F. Near and far transfer in

890

cognitive training: A second-order meta-analysis. Collabra Psychol. 2019;5(1):18.

891

111.

Soveri A, Antfolk J, Karlsson L, Salo B, Laine M. Working memory training revisited: A

892

multi-level meta-analysis of n-back training studies. Psychon Bull Rev. 2017;24(4):1077–96.

893

112.

Boot WR, Simons DJ, Stothart C, Stutts C. The pervasive problem with placebos in

894

psychology: Why active control groups are not sufficient to rule out placebo effects. Perspect

895

Psychol Sci. 2013;8(4):445–54.

896

113.

Van Dijk KRA, Hedden T, Venkataraman A, Evans KC, Lazar SW, Buckner RL. Intrinsic

897

functional connectivity as a tool for human connectomics: Theory, properties, and optimization. J

898

Neurophysiol. 2010;103(1):297–321.

899

114.

Jones KT, Berryhill ME. Parietal contributions to visual working memory depend on task

900

difficulty. Front Psychiatry. 2012;3:81.

901

115.

Wiemers EA, Redick TS, Morrison AB. The influence of individual differences in cognitive

902

ability on working memory training gains. J Cogn Enhanc. 2019;3(2):174–85.

903

116.

Foster JL, Harrison TL, Hicks KL, Draheim C, Redick TS, Engle RW. Do the effects of

904

working memory training depend on baseline ability level? J Exp Psychol Learn Mem Cogn.

905

2017;43(11):1677–89.

906

117.

Traut HJ, Guild RM, Munakata Y. Why does cognitive training yield inconsistent benefits?

907

A meta-analysis of individual differences in baseline cognitive abilities and training outcomes.

908

Front Psychol. 2021;12:662139.

909

118.

Gobet F, Sala G. Cognitive training: A field in search of a phenomenon. Perspect Psychol

910

Sci J Assoc Psychol Sci. 2023;18(1):125–41.

911

119.

Fellman D, Jylkkä J, Waris O, Soveri A, Ritakallio L, Haga S, et al. The role of strategy use

912

in working memory training outcomes. J Mem Lang. 2020;110:104064.

913

120.

Corrigan N, Păsărelu CR, Voinescu A. Immersive virtual reality for improving cognitive

914

deficits in children with ADHD: A systematic review and meta-analysis. Virtual Real. 2023;1–20.

915