Articles

A novel digital intervention for actively reducing severity of

paediatric ADHD (STARS-ADHD): a randomised controlled trial
Scott H Kollins, Denton J DeLoss, Elena Cañadas, Jacqueline Lutz, Robert L Findling, Richard S E Keefe, Jeffery N Epstein, Andrew J Cutler,
Stephen V Faraone

Summary
Background Attention-deficit hyperactivity disorder (ADHD) is a common paediatric neurodevelopmental disorder with Lancet Digital Health 2020;
substantial effect on families and society. Alternatives to traditional care, including novel digital therapeutics, have 2: e168–78

shown promise to remediate cognitive deficits associated with this disorder and may address barriers to standard Published Online
February 24, 2020
therapies, such as pharmacological interventions and behavioural therapy. AKL-T01 is an investigational digital
https://doi.org/10.1016/
therapeutic designed to target attention and cognitive control delivered through a video game-like interface via at-home S2589-7500(20)30017-0
play for 25 min per day, 5 days per week for 4 weeks. This study aimed to assess whether AKL-T01 improved attentional See Comment page e150
performance in paediatric patients with ADHD. Psychiatry and Behavioral
Sciences, Duke University
Methods The Software Treatment for Actively Reducing Severity of ADHD (STARS-ADHD) was a randomised, double- Medical Center, Durham, NC,
blind, parallel-group, controlled trial of paediatric patients (aged 8–12 years, without disorder-related medications) with USA (Prof S H Kollins PhD,
Prof R S E Keefe PhD); Duke
confirmed ADHD and Test of Variables of Attention (TOVA) Attention Performance Index (API) scores of −1·8 and Clinical Research Institute,
below done by 20 research institutions in the USA. Patients were randomly assigned 1:1 to AKL-T01 or a digital control Durham, NC, USA
intervention. The primary outcome was mean change in TOVA API from pre-intervention to post-intervention. Safety, (Prof S H Kollins); Akili
Interactive Labs, Boston, MA,
tolerability, and compliance were also assessed. Analyses were done in the intention-to-treat population. This trial is
USA (D J DeLoss PhD,
registered with ClinicalTrials.gov, NCT02674633 and is completed. E Cañadas PhD, J Lutz PhD);
Department of Psychiatry,
Findings Between July 15, 2016, and Nov 30, 2017, 857 patients were evaluated and 348 were randomly assigned to Virginia Commonwealth
University, Richmond, VA, USA
receive AKL-T01 or control. Among patients who received AKL-T01 (n=180 [52%]; mean [SD] age, 9·7 [1·3] years) or
(Prof R L Findling MD); VeraSci,
control (n=168 [48%]; mean [SD] age, 9·6 [1·3] years), the non-parametric estimate of the population median change Durham, NC, USA
from baseline TOVA API was 0·88 (95% CI 0·24–1·49; p=0·0060). The mean (SD) change from baseline on the (Prof R S E Keefe); Department
TOVA API was 0·93 (3·15) in the AKL-T01 group and 0·03 (3·16) in the control group. There were no serious adverse of Pediatrics, University of
Cincinnati College of Medicine,
events or discontinuations. Treatment-related adverse events were mild and included frustration (5 [3%] of 180)
Cincinnati, OH, USA
and headache (3 [2%] of 180). Patient compliance was a mean of 83 (83%) of 100 expected sessions played (Prof J N Epstein PhD); Meridien
(SD, 29·2 sessions). Research & Lake Erie College of
Osteopathic Medicine,
Bradenton, FL, USA
Interpretation Although future research is needed for this digital intervention, this study provides evidence that
(A J Cutler MD); and Psychiatry
AKL-T01 might be used to improve objectively measured inattention in paediatric patients with ADHD, while and Neuroscience and
presenting minimal adverse events. Physiology, SUNY Upstate
Medical University,
Syracuse, NY, USA
Funding Sponsored by Akili Interactive Labs. (Prof S V Faraone PhD)
Correspondence to:
Copyright © 2020 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Dr Scott Kollins, Psychiatry and
Behavioral Sciences, Duke
Introduction due to caregiver preferences or concerns about abuse, University Medical Center,
Durham, NC 27710, USA
Attention-deficit hyperactivity disorder (ADHD) is a misuse, and diversion. Barriers to access also limit the
[email protected]
neurodevelopmental disorder of persistent impaired use of behavioural interventions, given a shortage of
attention, hyperactivity, and impulsivity that negatively properly trained paediatric mental health specialists8 and
affects daily functioning and quality of life. ADHD is variability in insurance coverage for such services.9,10
one of the most commonly diagnosed paediatric mental Indeed, studies in both the USA and the UK have found
health disorders, with a prevalence estimated to be 5% that most children with paediatric mental health needs do
worldwide,1 and exerts a substantial burden on families not have proper access to services.11,12
and society.2 Digital therapeutics for ADHD may address these
Front-line intervention for ADHD includes pharmaco­ limitations with improved access, minimal side-effects,
logical and non-pharmacological interventions, which and low potential for abuse. Numerous studies and
have shown short-term efficacy.3–5 Existing treatments meta-analyses on digital interventions targeting specific
have side-effects that limit their acceptability,6 are only cognitive functions have attempted to assess the
effective when administered, and may not be as effective magnitude of efficacy for children and adolescents with
for reducing daily impairments versus ADHD symptoms.7 ADHD. In general, the quality of the studies is low, and
Pharmacotherapy may not be suitable for some patients many do not include a control group.3 Reported effect

www.thelancet.com/digital-health Vol 2 April 2020 e168

 Articles

Research in context
Evidence before this study objective measure of attention to a significant degree in
We searched PubMed with the search terms “ADHD,” “cognitive/ paediatric patients with ADHD, as well as patient-reported and
digital training/therapeutic,” “children/pediatric,” and “clinical parent-reported attentional functioning. Across several
trial” between Jan 1, 2010, and July 31, 2019. We found that secondary outcomes, including parent and clinician ratings of
few digital interventions were available before 2010. We also ADHD symptoms and functional impairment, AKL-T01
examined review articles and meta-analyses between significantly improved outcomes from pre-intervention to
Jan 1, 2010, and July 31, 2019. Although we did not limit the post-intervention, but not to a significantly greater degree
search to English language publications, we were not able to than the control condition. This trial represents one of a small
review non-English language publications, however, no relevant number of randomised controlled trials for digital
trial seemed to be available in a non-English language journal. interventions for paediatric patients with ADHD. The methods
Several digitally based interventions for attention deficit were modelled after randomised controlled trials for other
hyperactivity disorder (ADHD) were identified. Many trials treatment modalities (ie, pharmaceutical trials) and represent
focused on training working memory, and fewer on targeted a model approach for evaluating the effects of a digital
attention and cognitive control specifically. Further, many of the intervention.
studies contained methodological limitations, including
Implications of all the available evidence
inadequate control conditions or masking, or both, no random
This study shows that a digital intervention can significantly
assignment to intervention conditions, small sample size, and no
increase attentional functioning of children with ADHD. Future
safety or adverse event assessments. Many studies used outcome
trials are warranted to examine the durability and time course
measures similar to the training tasks and did not use US Food
of this novel intervention, as well as the appropriate dose that
and Drug Administration-approved cognitive outcomes or those
might provide optimal benefit. In addition, studies to better
commonly used in clinical settings. Indeed, meta-analyses on
characterise the clinical significance of objective attention
cognitive training for children with ADHD confirm that most
measures versus subjective symptom ratings are needed. These
current studies have inadequate methodology and cannot
findings have implications for clinical practice, as AKL-T01 is a
definitively evaluate the efficacy and clinical relevance of such
safe and easy-to-access intervention that could address various
treatments. The most comprehensive review of these studies to
intervention needs for paediatric patients with ADHD and
date concluded that digital interventions cannot be
without comorbid conditions (ie, attention deficits), but
recommended on the basis of the current body of evidence.
cannot replace current standard of care.
Added value of this study
In this randomised controlled trial, AKL-T01 (an investigational
digital therapeutic) increased attentional functioning in an

sizes are generally small but differ widely with respect to described a game-like intervention developed to engage
interventions and study designs.13 In addition, a recent the cognitive control and attention systems in older
meta-review concluded that a range of digital inter­ adults. The digital intervention improved cognitive
ventions, including working memory training and performance and these changes were associated with
neurofeedback, could not be recommended for treatment functional electroen­cephalogram (EEG) changes in the
of ADHD owing to inconsistent findings and generally prefrontal cortex.This intervention specifically targeted
minimal effects on outcomes provided by independent the management of cognitive interference, which occurs
unmasked observers.14 Nevertheless, meta-analyses when two or more tasks compete for cognitive and
conclude that interventions more broadly targeting attentional resources. The cognitive control and attention
cognitive functions generally show larger effects. systems evaluated in these studies are similar to the
ADHD has numerous well characterised but hetero­ deficits observed in paediatric patients with ADHD.18,19
geneous neurobiological substrates underlying cognitive This overlap informed the development of a novel digital
impairments that can serve as targets for intervention therapeutic, AKL-T01 (Akili Interactive Labs, Boston, MA,
development.15 For example, impairments related to USA), which was developed to engage paediatric users
attention and cognitive control are associated with lower through video game graphics and reward loops and to
activation of frontal, frontoparietal, and ventral attention use real-time adaptive mechanisms that continuously
networks.16 Research on video games and game-based personalise intervention difficulty on the basis of the
interventions that may alter brain structures and user’s ability and progression. Specifically, AKL-T01
function, suggest that targeted digital interventions— targets attentional control to manage competing tasks
based on current models of cognitive function and which and to efficiently (flexibly) shift attention between tasks.
leverage video game design to engage patients over Further, divided and selective attention systems are
time—are promising.17 Anguera and colleagues18 required to process several tasks simultaneously. In an

e169 www.thelancet.com/digital-health Vol 2 April 2020

 Articles

unmasked, proof-of-concept study, a prototype version of and parents was done to evaluate expectation of benefit
AKL-T01 showed improvements in attention, inhibition, of both interventions. The results suggested parents
and working memory in paediatric patients with ADHD had a similar expectation of benefit from both AKL-T01
but not in patients without ADHD.20 and our control condition (see appendix p 4). Parents and
The primary objective of the present trial was to patients were discouraged from discussing their
evaluate the efficacy and tolerability of AKL-T01 in randomised intervention with anyone other than an
paediatric patients with ADHD. unmasked study coordinator. Investigators and other
masked site staff were not permitted access to source
Methods documents or case report forms.
Study design
STARS-ADHD was a randomised, double-blind, parallel- Randomisation and masking
group, controlled trial done at 20 research institutions Eligible patients were randomly assigned 1:1 to receive
(appendix p 3) in the USA from July 15, 2016, to AKL-T01 or a control. The randomisation scheme was See Online for appendix
Nov 30, 2017. During the screening phase (days −28 to −7), generated by Duke Clinical Research Institute statistics
patients were evaluated for eligibility. Children treated by means of validated computer software-generated
with medication for ADHD discontinued medication to pseudorandom numbers. Randomisation was stratified
at least 3 days before baseline. At baseline (day 0), by stimulant medication status at screening. Each site
additional eligibility criteria were assessed. The study had unmasked staff who enrolled patients through
was done in accordance with the International the clinical data management system, obtained the
Conference on Harmonisation Regulations, and was randomised intervention, and trained patients on the
approved by each site’s institutional review board assigned device. Devices were provided to the unmasked
(Copernicus Group [14 sites], Duke University Health site staff by Akili (Akili Interactive Labs, Boston, MA,
System, Cincinnati Children’s Hospital Medical Center, USA) along with the list that linked the device serial
University of California Davis, University of California number to intervention; Akili was masked as to which
San Francisco, Johns Hopkins Medical Center, and patient received which device until after database lock.
Western Institutional Review Board). Parents, patients, and investigators completing outcome
measure assessments were masked to intervention
Participants allocation (appendix p 4).
Eligible patients were aged 8–12 years with a confirmed
diagnosis of ADHD as per the Diagnostic and Statistical Procedures
Manual of Mental Disorders (5th edn) criteria and Eligible patients were instructed to use their randomised
confirmed via the Mini International Neuropsychiatric intervention for about 10 min while the unmasked
Interview for Children and Adolescents Kid Screen at coordinator monitored the session to ensure patients
screening. Other key inclusion criteria included baseline could follow the rules of their assigned intervention.
scores on the ADHD Rating Scale-IV (ADHD-RS-IV) of Study interventions were administered by means of
28 and above, the Test of Variables of Attention (TOVA) an iPad mini 2 tablet (Apple, USA). iPads either had
Attention Performance Index (API) −1·8 and below, AKL-T01 or the control preloaded, and patients accessed
indicating cognitive deficits in the attention domain, and their randomised intervention with a unique username
a baseline intelligence quotient of 80 and above. Key and password.
exclusion criteria were significant comorbid psychiatric AKL-T01 is an investigational digital therapeutic that
diagnoses and use of ADHD medications that could uses a proprietary algorithm designed to improve
not be discontinued. Parents provided written informed attention and related cognitive control processes, by
consent with patient assent at screening–baseline. training inter­ference management at an adaptive and
Complete inclusion and exclusion criteria are described personalised high degree of difficulty. Interference is
in the appendix (p 2). instantiated through a video game-like interface displaying
Children who recently used or were currently using two tasks that are to be done in parallel (multitasking): a
stimulants were eligible, provided they were not optimally perceptual discrimination targeting task in which users
managed and willing to washout between 7 and 3 days respond to the instructed stimulus targets and ignore the
before baseline. This group was of particular interest stimulus distractors (similar to a Go–No-Go task) and a
because parents and children would have recent experience sensory motor navigation task in which users continuously
with some form of pharmacological intervention. As such, adjust their location to interact with or avoid positional
randomisation was stratified by medication status at targets. Performance in each task is assessed during
screening (see below). single and interference (multitask) conditions. On the
To minimise bias, parents and patients were informed basis of the individual’s performance, the interference
that the study was evaluating the effect of two different training is adapted in real time, by means of a stair-
investigational interventions for ADHD. Previous market casing algorithm methodology. This tailors the training
research with expert interviews in a sample of 59 children specifically to each individual’s performance level to

www.thelancet.com/digital-health Vol 2 April 2020 e170

 Articles

achieve a consistent and optimal challenge at a predefined targets), RT variability total (both halves), and d-prime
level of difficulty that is challenging but also tolerable. Half-2 (highly frequent targets).22,23
Further, on the basis of difference between single-task and Secondary efficacy endpoints were between-group
interference performance, the user advances by reducing comparisons of pre-intervention and post-intervention
interference costs (closes the performance gap between change in scores on the Impairment Rating Scale (IRS),
interference and single-task conditions, at a specified ADHD-RS-IV (Total [ADHD-RS-T], Inattentive [ADHD-
level). Progress is signalled by earning rewards and R-I], Hyperactive/Impulsive [ADHD-RS-H] subscales),
unlocking new environments. As the user proceeds Clinical Global Impressions—Improvement (CGI-I),
through the intervention and the different environments, and the Behavior Rating Inventory of Executive Function
periodic recalibration occurs to maintain an optimal (BRIEF; Parent Inhibit, Working Memory subscales, and
See Online for video difficulty level. AKL-T01 is presented in the video. Metacognition Index [post hoc]). Descriptions of each of
The control was designed to match AKL-T01 on the measures are provided in the appendix (pp 4–5).
expectancy, engagement, and time on task in the form of a We further analysed patient-reported and parent-
challenging and engaging digital word game, targeting reported perceived benefits related to attention improve­
cognitive domains not targeted by the AKL-T01 intervention ments in real life (post hoc), assessed during a structured
and not primarily associated with ADHD.21 The user was exit questionnaire asking whether the intervention
instructed to find and connect letters on a grid to spell helped their or their child’s attention in real life, with yes
words; points are awarded on the basis of number of words or no responses.
formed, word length, and the use of unusual letters. There During the intervention period, caregivers spontaneously
is progression in difficulty to maintain engagement and reported adverse events by phone to masked investigators;
expectation of benefit from patients and their caregivers. any adverse events spontaneously reported during study
During the intervention period (days 1–28), patients visits were captured. Details about use, performance, and
were instructed to use AKL-T01 or the control at home compliance with intervention were automatically recorded
for 5 sessions per day (total time on task about 25 min), by the study devices and uploaded to central servers when
5 days per week, for 4 weeks or the control for the iPads were connected to wireless internet.
25 min per day, 5 days per week, for 4 weeks. Compliance The proportions of responders at the end of treatment
was monitored electronically by unmasked study co­ phase for primary and secondary endpoints were
ordinators, who notified parents by email if the intervention prespecified on the basis of previous studies,20,24 and
was not administered over a 48-h period. AKL-T01 and the clinical meaningfulness for these analyses was defined as:
control also generated automatic reminders. Additional API improvement greater than 1·4 points, and post-test
information regarding the protocol may be found in the API score 0 or more (normative range), ADHD-RS
appendix (pp 6–7). The post-intervention visit was scheduled improvement of 2 points or more, CGI-I post-score of 1
on day 28. Patients were reassessed for attentional (very much improved) or 2 or less (very much or much
functioning, ADHD symptoms, and impairment. improved), and any improvement in IRS. Additional
post-hoc responder definitions were also examined:
Outcomes ADHD-RS total score change 30% or more, and
The primary outcome measure was the mean change in percentage of participants who scored in the normative
the TOVA API from pre-intervention to post-intervention. range for TOVA standard score measures (>85, per the
The TOVA is a validated, computerised, continuous TOVA Clinical Manual).22 In post-hoc analyses, inter­
performance test that objectively measures attention and vention effects in the subgroup of children who washed
inhibitory control, normalised by age and sex.22 TOVA out of ADHD medication after screening (n=20) and
has been cleared by the US Food and Drug Administration children who had discontinued medication before but
(FDA) to facilitate assessment of attention deficits and to within 30 days of screening (n=45) were explored.
evaluate the effects of interventions in ADHD.
TOVA presents targets and non-targets as squares that Statistical analysis
either appear at the top or bottom of the screen. The task A prespecified analysis plan governed all analyses, unless
takes 21·6 min and consists of two halves: the first half identified as post hoc. Power analyses determined that a
has a target-to-non-target ratio of 1:3·5 (similar to sample size of 150 patients per intervention group would
sustained attention tests); the second half has a target-to- be sufficient to detect an effect size of 0·40 with 90% or
non-target ratio of 3·5:1, thus requiring more inhibitory more power on a two-tailed, between-patients t test and α
control. TOVA calculates a wide range of outcome criterion of 0·05.
measures that assess processes known to be disrupted in A single interim analysis for sample-size re-estimation
patients with ADHD, such as response time variability was prespecified to occur after half of the initial sample
(attention consistency), ex-Gaussian tau (attentional size was collected, and done by separate unmasked
lapses), and response time (processing speed).22,23 The statistical staff in order to minimise any bias in the conduct
TOVA API is a composite score of the sum of three of the study. If the conditional power of the interim sample
scores: reaction time (RT) mean Half-1 (highly infrequent indicated the need for a larger final sample, the estimated

e171 www.thelancet.com/digital-health Vol 2 April 2020

 Articles

sample size required would have been communicated to percentage of instructed sessions use completed during
the sponsor and investigators, up to a maximum of the intervention period or the percentage of instructed
1000 total participants. The α criterion for the final analysis use time for the control (as the control did not follow a
of the primary outcome was reduced by means of the five-sessions-a-day format).
O’Brien-Fleming alpha-spending method to 0·0412. Since Post-hoc t tests for between-group and within-group
the interim analysis only looked at the primary outcome, changes were done on the primary and key secondary
and analysis of the secondary outcomes was gated by endpoints to assess sensitivity of results to statistical
success of the primary outcome, no α adjustment for the methodology. All such tests were in agreement with
secondary outcomes was necessary. The interim analysis corresponding non-parametric Wilcoxon tests with
was done at n=75/75 (AKL-T01/control). respect to significance for all results reported.
All outcomes were analysed in all randomly assigned A post-hoc Fisher’s exact test was done to compare the
patients by means of an intention-to-treat methodology. percentage of patients–parents in the two groups who
Safety data are presented for the population of patients indicated real-life improvements related to attention on
that received AKL-T01 or control for at-home intervention. the exit questionnaire (yes or no response).
For the safety analysis, patients who received an
intervention inconsistent with their original randomly Protocol amendments
assigned group (n=1 for AKL-T01 and n=1 for control) Three versions of the protocol were used throughout the
were recategorised to the intervention received. study. Under the first version (version 1.0, dated
The primary endpoint (change in TOVA API) was Nov 17, 2015), no participants were enrolled. Under the
analysed by means of a Wilcoxon rank-sum test owing to second version (version 1.01, dated April 6, 2016),
evidence of non-normality. The α criterion for the final 43 participants were enrolled. Under the third version
analysis was adjusted for the interim analysis sample-size (version 1.02, dated July 25, 2016), 305 participants were
re-estimation by means of the O’Brien-Fleming alpha- enrolled. Details regarding all of the changes made for
spending function, and the interim and post-interim each of these amendments are found in the appendix
cohorts were tested separately, with the results combined (pp 6–7). One exclusion criterion was added to version 1.01
via the Cui-Hung-Wang method.25 If the combined p value that disallowed participation for children who had
was less than the adjusted criterion, the primary endpoint previously been in a study with AKL-T01. Version 1.01
was considered successful. The α criterion for the primary also added a requirement that participants be able to spell
endpoint significance was 0·041. at least two words during the EVO: Words assessment at
Key secondary endpoints (IRS, ADHD-RS, CGI-I, and baseline. Version 1.02 added a requirement that sites
BRIEF-Parent) were analysed by Wilcoxon rank-sum tests document a discussion with participants and caregivers
due to evidence of non-normality. The between-group regarding intended and unintended use of the devices.
difference for the primary endpoint was calculated by
means of the Hodges-Lehmann estimate of location shift 857 patients assessed for eligibility
to coincide with the use of a Wilcoxon test. Type I error
was controlled with a resampling bootstrap method to
adjust p values due to correlated endpoints. Secondary 509 excluded
482 did not meet inclusion criteria
endpoints were not adjusted for sample-size re-estimation. 32 met exclusion criteria
If the adjusted p value was less than 0·050, the endpoint 4 enrolled late
was considered significant. Responder analyses used χ²
tests to compare AKL-T01 and control for the primary and 348 randomly assigned
secondary endpoints. To summarise findings across a
range of outcomes, odds ratios and CIs were calculated to
compare the efficacy of AKL-T01 versus control.
Prespecified sensitivity analyses of the primary and key 180 allocated to AKL-T01 168 allocated to control
179 received allocated intervention 167 received allocated intervention
secondary endpoints were designed to assess the effects 1 received incorrect allocation 1 received incorrect allocation
of: cohort, site, age, sex, missing data (if >10% of outcome
data were missing), and parent expectancy (secondary
3 lost to follow-up 2 lost to follow-up
endpoints only). For the primary and key secondary 1 withdrawn by parent 2 withdrawn by parent
endpoints, post-hoc non-parametric analyses for within- 1 investigator decision 4 excluded owing to invalid test
group changes (pre-intervention vs post-intervention) 6 excluded owing to invalid test

were done by means of the Wilcoxon signed-rank test

(SAS version 9.4). 169 included in intention-to-treat-analysis 160 included in intention-to-treat-analysis
Descriptive statistics summarised patient demo­
graphics, protocol deviations, intervention compliance, Figure 1: Trial profile
intervention-related adverse events, and qualitative survey Detailed information on inclusion and exclusion criteria can be found in the appendix (p 2).
results. Compliance with intervention was defined as the AKL-T01=an investigational digital therapeutic.

www.thelancet.com/digital-health Vol 2 April 2020 e172

 Articles

Other changes included more details on device

AKL-T01 Control
(n=180) (n=168) management and inventory procedures, updates to the
statistical analysis section, requirements that unmasking
Age, years 9·7 (1·3) 9·6 (1·3)
events be captured in the study database, and a
Male 125 (69%) 123 (73%)
requirement that the same clinician rater complete
Female 55 (31%) 45 (27%)
assessments at pre-study and post-study timepoints.
Baseline score
These changes are not expected to have any effect on
Test of Variables of Attention—Attention −5·1 (3·0) −4·9 (3·1)
Performance Index*
study outcomes. VeraSci Trials (formerly NeuroCog
Trials) reviewed data for select cognitive and clinical
Impairment Rating Scale 5·5 (1·1) 5·5 (1·2)
measures to determine quality and consistency within
ADHD-Rating Scale 39·0 (6·8) 38·3 (6·6)
and between measures. This trial is registered with
ADHD-Rating Scale—Inattentive 21·9 (3·5) 21·6 (3·7)
ClinicalTrials.gov, NCT02674633.
ADHD-Rating Scale—Hyperactivity 17·1 (6·0) 16·7 (5·4)
Clinical Global Impressions—Severity† 4·5 (0·7) 4·6 (0·6)
Role of the funding source
Data are n (%) or mean (SD). AKL-T01=an investigational digital therapeutic. The funder had a role in study conception and design,
*n=179 for AKL-T01. †Assessed only at baseline.
confirming data and statistical analyses, and conducting
Table 1: Baseline characteristics the study. All authors had full access to all the data in the
study and were involved in data interpretation and
writing of the report. The corresponding author had final
responsibility for the decision to submit for publication.
1·50
*
Results
1·25 Of 857 children screened for eligibility, 348 patients were
randomly assigned to receive AKL-T01 (n=180) or control
Mean (SE) change in TOVA API

1·00
(n=168) between July 15, 2016, and Nov 30, 2017 (figure 1
and appendix p 3). Demographic and clinical character­
Improvement

0·75
istics at baseline are shown in table 1.
0·50
The mean number of sessions completed by patients
in the AKL-T01 group was 83·2 out of 100 sessions
0·25 (83% instructed use; SD=29·2 sessions). Patients in the
control group used their intervention 480·7 min of
0 500 min (96% instructed use).
There was a significant difference between intervention
–0·25 groups on the primary efficacy endpoint (adjusted
AKL-T01 (n=169) Active control (n=160)
p=0·0060); non-parametric estimate of the population
Figure 2: Primary endpoint: TOVA API mean (SE) change pre-intervention to median change (Hodges-Lehmann estimate) was 0·88
post-intervention in the intention-to-treat population (95% CI 0·24–1·49). The mean (SD) change from baseline
*Adjusted p<0·050; prespecified Wilcoxon rank-sum test. Triangle represents
median change, pre-intervention to post-intervention.
on the TOVA API was 0·93 (3·15) in the AKL-T01 group
and 0·03 (3·16) in the control group (figure 2). There were
no intervention-group differences for secondary measures:
IRS, ADHD-RS, ADHD-RS-I, ADHD-RS-H, BRIEF-
AKL-T01 Control χ² test p Parent Inhibit and Working Memory and Metacognition
Test of Variables of Attention—Attention 79/169 (47%) 51/160 (32%) 7·60 0·0058 (post hoc) from pre-intervention to post-intervention or
Performance Index (type A: improvement mean CGI-I score at post-intervention (appendix pp 4–5).
>1·4 points)
Sensitivity analyses showed no evidence that site,
Attention Performance Index (type B: 18/170 (11%) 7/160 (4%) 4·54 0·033
post-intervention score ≥0)
baseline TOVA API, age, or sex attenuated the inter­
ADHD-Rating Scale (improvement ≥2 points from 128/173 (74%) 119/164 (73%) 0·088 0·77
vention effect. Because missing data did not exceed the
pre-intervention to post-intervention) prespecified limit, sensitivity analyses for missing data
ADHD-Rating Scale (≥30% reduction)* 42/173 (24%) 31/164 (19%) 1·43 0·23 were not done. Sensitivity to parent expectancy was not
Impairment Rating Scale 82/171 (48%) 60/161 (37%) 3·87 0·049 evaluated owing to lack of significant differences between
Clinical Global Impressions (≤2 at post- 29/175 (17%) 26/164 (16%) 0·032 0·86 groups on the secondary endpoints.
intervention) In post-hoc within-group analyses, change in TOVA
Clinical Global Impressions (1 at post-intervention) 1/175 (1%) 1/164 (1%) 0·0021 0·96 API score from pre-intervention to post-intervention
Data are n/N (%) unless otherwise indicated. AKL-T01=an investigational digital therapeutic. *Post-hoc analysis. significantly improved with AKL-T01 (p<0·0001) but not
ADHD=Attention-deficit hyperactivity disorder. AKL-T01=an investigational digital therapeutic. with control (p=0·67). Both AKL-T01 and patients in the
control group showed significant within-group improve­
Table 2: Clinical responder analysis intention-to-treat population
ments in all secondary endpoints (appendix pp 4–5).

e173 www.thelancet.com/digital-health Vol 2 April 2020

 Articles

Additional exploratory post-hoc analyses were done Objective attention

TOVA API >1·4
to better interpret the change in objective measures (attention composite)
Subjective attention
Impairments and symptoms
of attention. The Standard Score transformations of TOVA API normative range
TOVA components related to attention were analysed for (attention composite)
between-group differences: significant between-group TOVA RT mean H1 normative range
(selective and sustained attention)
effects in favour of AKL-T01 were found for RT mean
TOVA RT variability normative range
Half-1 (p<0·0003), RT variability total (p=0·019), and (attentional consistency)
ex-Gaussian tau (p=0·0014).
RS Inattentive ≥30% improvement
Responder analyses showed that AKL-T01 resulted

Responder type
in TOVA API score improvements of greater than Parent report improved attention
1·4 points in 79 (47%) of 169 patients versus 51 (32%)
of 160 controls (p=0·0058; table 2). AKL-T01 versus Child report improved attention
control was also associated with the movement of more
IRS ≥1 point improvement
patients into the normative ranges across different
measures of attention on TOVA: API of 0 and above in RS Total ≥30% improvement
18 (11%) of 170 versus 7 (4%) of 160, RT mean Half-1 in
32% versus 16%, and RT variability total in 22% versus RS Hyperactive ≥30% improvement
13% (p values API p=0·033, RT mean Half-1 p=0·0043,
CGI-I ≤2
RT variability total p=0·030). Overall, AKL-T01 versus
controls moved significantly more patients into the 0 0·5 1·0 5·0
normative range in at least one objective measure of
Favours control Favours AKL-T01
attention (36% vs 21%, p=0·0027). IRS responder rates Intention-to-treat odds ratio (95% CI)
were significantly higher after AKL-T01 versus control
(82 [48%] of 171 vs 60 [37%] of 161, p=0·049). Remaining Figure 3: STARS-ADHD intention-to-treat responder forest plot
Odds ratio of 1·0 indicates that participants do not respond more to AKL-T01 than control. CIs in which the lower
responder comparisons did not differentiate between
bound does not cross 1·0 are significant. API=Attention Performance Index. CGI-I= Clinical Global Impressions-
groups. Improvement. IRS=Impairment Rating Scale. RT mean H1=Reaction Time Mean during First Half of the TOVA.
The percentage of patients reporting an improvement RS Inattentive=ADHD Rating Scale—Inattentive. RS Hyperactive= ADHD Rating Scale—Hyperactivity.
in attention on the exit questionnaire for AKL-T01 versus ADHD=Attention-deficit hyperactivity disorder. TOVA=Test of Variables of Attention. AKL-T01=an investigational
digital therapeutic.
control (126 [73%] of 172 vs 107 [66%] of 162) was not
significant (χ²(1)=2·054, p=0·15). However, the percentage
of parents reporting improvements in their child’s (TOVA API) in paediatric patients with ADHD
attention was significantly higher for AKL-T01 versus compared with the control condition. Across a range of
control (97 [56%] of 173 vs 71 [44%] of 162, χ²(1)=5·015, secondary outcomes, including parent and clinician
p=0·025). Comparisons of the efficacy of AKL-T01 versus ratings of ADHD symptoms and functional impairment,
control across a range of outcomes are summarised in the effects of AKL-T01 from pre-intervention to post-
figure 3. intervention were not different from the control
In post-hoc analyses of patients who discontinued condition. Additional attention-related measures from
stimulant medication, within 30 to 3 days before the start TOVA, including mean reaction time during infrequent
of the study (washout group), AKL-T01 significantly target stimuli, and response variability (ie, total RT
differentiated from control on most secondary efficacy variability and ex-Gaussian tau) showed significantly
endpoints including ADHD-RS (p=0·0092), ADHD-RS-I greater improvements in the AKL-T01 group. Globally,
(p=0·0083), and CGI-I (p=0·012). The difference between parent-reported improvement of attention, as assessed
medication washout groups on the IRS was not significant by the exit survey, was higher in the AKL-T01 group
(p=0·065; figure 4). compared with controls. A prespecified subgroup of
The proportion of patients reporting any intervention- patients who washed out of medications showed
related adverse events was 12 (7%) of 180 with AKL-T01 significant between-group effects on several secondary
and 3 (2%) of 168 with control (table 3). There were no endpoints, including ADHD symptoms.
serious intervention-related AEs or discontinuations due Both interventions were very well tolerated; only
to AEs in either group. The most common intervention- 12 (7%) of 180 and 3 (2%) of 168 patients in the AKL-T01
related AEs associated with AKL-T01 were frustration and control groups had intervention-related AEs,
(5 [3%] of 180) and headache (3 [2%] of 180). respectively. All AEs associated with AKL-T01 were
classified as mild or moderate in severity and resolved
Discussion after study discontinuation.
In this randomised controlled clinical trial of a digital The current study findings of improved attention (via
intervention for ADHD, the active intervention AKL-T01 TOVA API) following treatment with AKL-T01 are
significantly improved performance on the primary consistent with benefits reported in previous uncon­
outcome measure—an objective measure of attention trolled studies.18,20 As a digital therapeutic, AKL-T01 could

www.thelancet.com/digital-health Vol 2 April 2020 e174

 Articles

0
theoretically address several challenges faced by existing
p=0·0092 interventions. First, its risk–benefit profile is favourable,
as only 12 (7%) of 180 patients assigned to AKL-T01
ADHD Rating Scale Total

had AEs, compared with rates of 40–60% in trials

–4·0
of commonly used stimulant medications.26 Therefore,
Improvement

AKL-T01 could be added to standard of care without

substantial additional safety concerns. Second, the digital
–8·0
nature of this intervention could reduce barriers to access
that are inherent in other forms of behavioural or non-
pharmacological interventions.27 Digital interventions
–12·0
AKL-T01 (n=30) Control (n=34) have been cited as possible ways to improve otherwise
poor access to mental health services.14
0
p=0·0083 The primary outcome measure for this trial—the TOVA
ADHD Rating Scale inattentive Total

API—differs from most pharmacological efficacy trials for

ADHD, which typically use parent-rated or clinician-rated
–2·0 symptom measures. The selection of the TOVA was based
Improvement

on several factors. First, because AKL-T01 was designed

specifically to target cognitive control and attention, we
–4·0 sought an outcome that would most precisely and validly
index these processes. The TOVA is an FDA-cleared
device23 for the objective assessment of attention and
–6·0 inhibitory control as part of an ADHD diagnosis or for
AKL-T01 (n=30) Control (n=34)
monitoring intervention outcomes and has been widely
0 used in both clinical practice and research studies. Second,
p=0·022
the TOVA measures cognitive functions that are relevant
BRIEF metacognition (t-score)

to the clinical presentation of ADHD,28 and attention

–2·5 performance metrics such as RT mean, RT variability, and
Improvement

ex-Gaussian tau are well characterised indicators of

attention-relevant cognitive processes, and are associated
–5·0 with clinically relevant outcomes including academic
behaviour29 and inattention and social problems.30
Finally, the TOVA setting has been described as
–7·5 mimicking “one component of the classroom situation in
AKL-T01 (n=31) Control (n=31) which children are required to remain seated and engaged
0 in a tedious, repetitive task,”31 suggesting ecological
p=0·065 validity of the TOVA test for real-world settings in which
Impairment Rating Score overall

children with ADHD often struggle.31 Traditional,

–0·4
symptom-based measures were included as secondary
measures. We also selected the IRS as a targeted measure
Improvement

of ADHD-related impairment because, as noted

previously, attentional processes are specifically linked to
–0·8
relevant clinical outcomes and symptom-based measures
(eg, ADHD-RS) and do not always correlate highly with
measures of functional impairment.5 For example, in a
–1·2
AKL-T01 (n=29) Control (n=34) paediatric stimulant medication trial, greater than 40% of
patients who showed a positive response on the primary
4·25
outcome measure (>30% reduction on the ADHD-RS)
Clinical Global Impressions—Improvement

p=0·012
4·00 failed to show significant functional improvement on a
validated measure of impairment.7 It has also been
3·75 reported that across four large-scale ADHD research
Improvement

3·50 samples, the average correlation between symptoms and

3·25

3·00
Figure 4: Medication washout subpopulation, subjective measures
2·75 ADHD=attention-deficit hyperactivity disorder. BRIEF=Behavior Rating Inventory
AKL-T01 (n=31) Control (n=34)
of Executive Function.

e175 www.thelancet.com/digital-health Vol 2 April 2020

 Articles

impairment accounted for less than 10% of variance.31 In

AKL-T01 (n=180) Active control
the current trial, despite there being few group differences (n=168)
on the ADHD-RS, significantly more children in the
Patients experiencing intervention- 12 (7%) 3 (2%)
AKL-T01 group were responders on the IRS, suggesting emergent adverse events
that this intervention, like other non-pharmacological Frustration 5 (3%) 0
interventions, may differentially affect impairment versus Headache 3 (2%) 2 (1%)
symptoms. Emotional reaction 2 (1%) 1 (1%)
In post-hoc analysis, children with a recent history Dizziness 1 (1%) 0
of pharmacological intervention exhibited significant
Nausea 1 (1%) 0
AKL-T01-related improvements in a range of symptom-
Aggression 1 (1%) 0
based outcomes, including ADHD-RS. This finding
Data are n (%). AKL-T01=an investigational digital therapeutic.
could be related to biological factors associated with a
recent pharmacological intervention, or psychological Table 3: Summary of intervention-emergent adverse events (intention-
factors such as parents being more attuned to symptom to-treat population)
changes in children recently treated with medication.
Further studies are warranted to explore the potential
effects of AKL-T01 in this important subgroup. Third, the study evaluated a 28-day intervention period
In the current study, there were no differences between with approximately 25-min daily sessions; it is unclear if
AKL-T01 and the control condition on secondary the benefits in attentional functioning might have been
measures, and several factors might explain these observed with a different regimen. The current study
findings. First, it is possible that parent or clinician- represents a single intervention of 1-month duration,
reported outcomes (ie, ADHD-RS) are not sensitive to the which is quite short. Additional studies with longer
effects of AKL-T01. In other words, the shown effects of intervention periods are needed. An ongoing study
the intervention on attentional processes may not be (ClinicalTrials.gov identifier: NCT03649074) is examining
as readily observable by parents and clinicians. The longer intervention periods (repeat intervention for a total
clinical implications of this possibility will be important of 2 months) and durability of effects 1 month after the
to explore in future studies. Second, expectations of intervention. In addition, that study is investigating
efficacy have been shown to moderate intervention effects whether the intervention has effects in children currently
in general, and also for digital interventions.32 In our treated with stimulant medication, which will help
study, parents of patients in both groups believed that address questions of generalisability. Fourth, power
their child received a novel intervention for ADHD; thus, analyses were calculated for our primary outcome to
the expectation of intervention effect can be assumed power our trial, but no power calculations were done for
for both interventions, and may partially explain any of our secondary outcomes or post-hoc analyses.
improvements in both groups. This design feature is Fifth, the study did not collect data (eg, EEG) that would
different from most pharmacological studies in which offer mechanistic explanation for the findings. The
patients and their caregivers are aware of a non-active, foundational study from which the intervention was
placebo condition. Finally, specific mechanisms common developed reported that effects of the AKL-T01 prototype
to AKL-T01 and the control condition may have resulted were mediated by EEG changes. Since EEG data were not
in improvements in both groups. Both interventions collected in this study, conclusions cannot be drawn
required continued perseverance, sometimes in the face about the neural mechanisms that might underlie
of failure, and may have trained coping and reappraisal intervention effects. Given these limitations, the transfer
skills or even increased the sense of self-efficacy and of benefit of the AKL-T01 intervention to real-world
mastery.33 Thus any intervention that requires the patient settings and the full clinical meaningfulness of the
to engage in a regular, structured setting that may include findings, as well as the mechanisms underlying these
repeated failure or repetitiveness can be seen as a effects, should be explored in further studies.
potential intervention for ADHD. Despite these limitations, the current trial had several
The current study has several important limitations. features that strengthen confidence in the results.
First, the inclusion criteria required that patients have a Diagnostic methods modelled after pharmacological
TOVA API up to −1·8, thus showing an objective baseline randomised controlled trials were used to establish
deficit in attentional function. This resulted in a eligibility for the study. Considerable steps were taken to
substantial number of patients with a clinical ADHD minimise potential biases or differences in the expectation
diagnosis being excluded from the trial. Second, children of benefit between AKL-T01 and control. These included
could not be taking medication for ADHD during the trial having masked raters and clear procedures for minimising
and could not have significant psychiatric comorbidity. discussion between parents and study staff about inter­
Therefore, it is unclear if these findings will generalise to vention assignment, and instructing parents and children
the broader population of patients with ADHD who have in both groups who believed that they were receiving an
comorbid conditions or patients taking medication. investigational intervention for ADHD.

www.thelancet.com/digital-health Vol 2 April 2020 e176

 Articles

The STARS-ADHD trial represented a randomised owner of VeraSci, which provided support for this trial. JNE is a
controlled trial for the evaluation of a digital intervention consultant for Akili Interactive Labs and receives grant support, research
support, or royalties from Akili Interactive Labs, the American Academy
to improve objectively measured attention in children with
of Pediatrics, American Board of Pediatrics, IXICO, Multi-Health
ADHD. It showed that compared to the control condition, Systems, and mehealth for ADHD. AJC has received research support,
AKL-T01 significantly improved objective measures of honoraria, or consulting fees from Akili Interactive Labs, Arbor,
attention, as measured by the TOVA. AKL-T01 also showed Ironshore, Neos, Otsuka, Purdue Canada, Shire, Sunovion, Supernus,
and Trisand is a member of the Neuroscience Education Institute Board.
effects that were not different from the control condition, SVF reports income, potential income, travel expenses, continuing
including the ADHD-RS. The intervention was well education support, or research support from Akili Interactive Labs,
tolerated, and this risk–benefit ratio suggests that AKL-T01 Arbor, Enzymotec, Genomind, Ironshore, Otsuka, Shire–Takeda,
could be a novel addition to the range of intervention Sunovion, and Supernus and a US patent (US20130217707 A1) for the use
of sodium–hydrogen exchange inhibitors in the treatment of ADHD.
options for ADHD. The digital nature of the intervention
could help to increase access for populations who might Data sharing
The STARS-ADHD Investigators agree to share de-identified individual
not otherwise be able to find non-pharma­ cological participant data, the study protocol, and the statistical analysis plan with
interventions. Various additional questions remain to be academic researchers 6 months after publication, and following
answered regarding the full clinical meaning­fulness of the completion of a Data Use Agreement. Proposals should be directed to
findings, the effect of different dosing schedules, and [email protected].
which patients might benefit the most from this type of Acknowledgments
intervention. Given these limitations, the results of the Duke Clinical Research Institute conducted the data and statistical
analyses. Writing and data analysis support, under the direction of the
current trial are not sufficient to suggest that AKL-T01 authors, was provided by Titiimaea Alailima, an employee of Akili
should be used as an alternative to established and Interactive Labs, Boston, MA, USA, Jeffrey Bower, and Norma Palma,
recommended treatments for ADHD. former employees of Akili Interactive Labs in accordance with Good
Publication Practice guidelines. Editorial support, under the direction of
Contributors the authors, was provided by Peloton Advantage, an OPEN Health
SHK, RLF, RSEK, AJC, and SVF had a role in the concept and design. company, Parsippany, NJ, with funding by Akili Interactive Labs, in
SHK, DJD, EC, JL, RLF, RSEK, JNE, and AJC had a role in acquisition, accordance with Good Publication Practice guidelines. The views and
analysis, or interpretation of data. SHK, DJD, EC, JL, RSEK, RLF, and opinions expressed within this manuscript are those of all the authors
JNE drafted the manuscript. SHK, DJD, EC, JL, RLF, RSEK, JNE, AJC, and do not necessarily represent those of the sponsor.
and SVF critically revised the manuscript. DJD and RSEK did the critical
analysis. SHK and RSEK obtained funding. DJD, EC, and RSEK References
provided administrative, technical and material support. SHK, DJD, 1 Polanczyk GV, Willcutt EG, Salum GA, Kieling C, Rohde LA.
RSEK, and JNE supervised. ADHD prevalence estimates across three decades: an updated
systematic review and meta-regression analysis. Int J Epidemiol
Declaration of interests 2014; 43: 434–42.
SHK is a consultant, principal investigator and owns stock options for 2 Gupte-Singh K, Singh RR, Lawson KA. Economic burden of
Akili Interactive Labs and received research support or consulting fees attention-deficit/hyperactivity disorder among pediatric patients in
from Arbor, Bose, Ironshore, Jazz, KemPharm, Neos, Otsuka, Rhodes, the United States. Value Health 2017; 20: 602–09.
Shire, Sunovion, and Tris. JL, DJD, and EC are employed by Akili 3 Catalá-López F, Hutton B, Núñez-Beltrán A, et al.
Interactive Labs and may own stock options. EC is a patent holder The pharmacological and non-pharmacological treatment of
(WO/2018/027080) for Processor Implemented Systems and Methods for attention deficit hyperactivity disorder in children and adolescents:
Measuring Cognitive Abilities. RLF receives or has received research a systematic review with network meta-analyses of randomised
support, acted as a consultant or served on a speaker’s bureau for Acadia, trials. PLoS One 2017; 12: e0180355.
Aevi, Akili, Alcobra, Allergan, Amerex, American Academy of Child & 4 Cortese S, Adamo N, Del Giovane C, et al. Comparative efficacy and
Adolescent Psychiatry, American Psychiatric Press, Arbor, Bracket, tolerability of medications for attention-deficit hyperactivity disorder
in children, adolescents, and adults: a systematic review and
Daiichi-Sankyo, Epharma SolutionfMRIs, Forest, Genentech, Insys,
network meta-analysis. Lancet Psychiatry 2018; 5: 727–38.
Ironshore, KemPharm, Luminopia, Lundbeck, Merck, the US National
5 Kollins SH. Moving beyond symptom remission to optimize
Institutes of Health, Neurim, Noven, Nuvelution, Otsuka, Patient-
long-term treatment of attention-deficit/hyperactivity disorder.
Centered Outcomes Research Institute, Pfizer, Physicians Postgraduate JAMA Pediatr 2018; 172: 901–02.
Press, Receptor Life Sciences, Roche, Sage, Shire, Sunovion, Supernus
6 Toomey SL, Sox CM, Rusinak D, Finkelstein JA. Why do children
Pharmaceuticals, Syneurx, Teva, TouchPoint, Tris, and Validus. RSEK is a with ADHD discontinue their medication? Clin Pediatr (Phila) 2012;
consultant for Akili Interactive Labs and has received research support or 51: 763–69.
consulting fees from Aeglea, Akebia, Akili Interactive Labs, Alkermes, 7 Weiss M, Childress A, Mattingly G, Nordbrock E, Kupper RJ,
Allergan, ArmaGen, Astellas, Avanir, AviNeuro/ChemRar, Axovant, Blood Adjei AL. Relationship between symptomatic and functional
Alcohol Content Testing Battery, Biogen, Boehringer Ingelheim, Cerecor, improvement and remission in a treatment response to stimulant
CoMentis, Critical Path Institute, Forum Pharmaceuticals, Gammon trial. J Child Adolesc Psychopharmacol 2018; 28: 521–29.
Howard & Zeszotarski, Global Medical Education, GW Pharmaceuticals, 8 Health Resources & Services Administration. Shortage areas.
Intracellular Therapeutics, Janssen, Kempharm, Lundbeck, Lysogene, https://data.hrsa.gov/topics/health-workforce/shortage-areas
Matrics Battery, MedScape, Mentis Cura, Merck, Merrakris Therapeutics, (accessed Dec 6, 2019).
Minerva Neurosciences, Mitsubishi, Montana State University, Monteris, 9 Bussing R, Zima BT, Gary FA, Garvan CW. Barriers to detection,
Moscow Research Institute of Psychiatry, National Institute of Mental help-seeking, and service use for children with ADHD symptoms.
Health, Neuralstem, Neuronix, Novartis, NY State Office of Mental J Behav Health Serv Res 2003; 30: 176–89.
Health, Orygen, Otsuka, Paradigm Testing, Percept Solutions, Pfizer, 10 Owens PL, Hoagwood K, Horwitz SM, et al. Barriers to children’s
Pharm-Olam, Regenix Bio, Reviva, Roche, Sangamo, Sanofi, SOBI, mental health services. J Am Acad Child Adolesc Psychiatry 2002;
Sengenix, Six Degrees Medical, Sunovion, Takeda, Targacept, Teague 41: 731–38.
Rotenstreich Stanaland Fox & Holt, Thrombosis Research Institute, 11 Danielson ML, Bitsko RH, Ghandour RM, Holbrook JR, Kogan MD,
University of Moscow, University of Southern California, University of Blumberg SJ. Prevalence of parent-reported ADHD diagnosis and
Texas Southwest Medical Center, Virtual Reality Functional Capacity associated treatment among U.S. children and adolescents, 2016.
Assessment Tool, VeraSci, WebMD, and Wilson Therapeutics and is J Clin Child Adolesc Psychol 2018; 47: 199–212.

e177 www.thelancet.com/digital-health Vol 2 April 2020

 Articles

12 Davies SC. Annual Report of the Chief Medical Officer 2013, Public 24 Gordon HA, Rucklidge JJ, Blampied NM, Johnstone JM.
Mental Health Priorities: Investing in the Evidence. London, UK: Clinically significant symptom reduction in children with attention-
Department of Health, 2014. deficit/hyperactivity disorder treated with micronutrients: an open-
13 Cortese S, Ferrin M, Brandeis D, et al. Cognitive training for label reversal design study. J Child Adolesc Psychopharmacol 2015;
attention-deficit/hyperactivity disorder: meta-analysis of clinical and 25: 783–98.
neuropsychological outcomes from randomized controlled trials. 25 Cui L, Hung HM, Wang SJ. Modification of sample size in group
J Am Acad Child Adolesc Psychiatry 2015; 54: 164–74. sequential clinical trials. Biometrics 1999; 55: 853–57.
14 Hollis C, Falconer CJ, Martin JL, et al. Annual research review: 26 Wolraich ML, Greenhill LL, Pelham W, et al. Randomized,
digital health interventions for children and young people with controlled trial of oros methylphenidate once a day in children with
mental health problems—a systematic and meta-review. attention-deficit/hyperactivity disorder. Pediatrics 2001; 108: 883–92.
J Child Psychol Psychiatry 2017; 58: 474–503. 27 Koerting J, Smith E, Knowles MM, et al. Barriers to, and facilitators
15 Faraone SV, Asherson P, Banaschewski T, et al. Attention-deficit/ of, parenting programmes for childhood behaviour problems:
hyperactivity disorder. Nat Rev Dis Primers 2015; 1: 15020. a qualitative synthesis of studies of parents’ and professionals’
16 Cortese S, Kelly C, Chabernaud C, et al. Toward systems perceptions. Eur Child Adolesc Psychiatry 2013; 22: 653–70.
neuroscience of ADHD: a meta-analysis of 55 fMRI studies. 28 Diamond G, Badir M, Sevilla P, Inbar D, Gadoth N.
Am J Psychiatry 2012; 169: 1038–55. Comparison between neurological examination and computerized
17 Shams TA, Foussias G, Zawadzki JA, et al. The effects of video test of attention for suspected ADHD: implications for assessment of
games on cognition and brain structure: potential implications for a common childhood disability. Int J Disabil Hum Dev 2013;
neuropsychiatric disorders. Curr Psychiatry Rep 2015; 17: 71. 12: 289–95.
18 Anguera JA, Boccanfuso J, Rintoul JL, et al. Video game training 29 Antonini TN, Narad ME, Langberg JM, Epstein JN.
enhances cognitive control in older adults. Nature 2013; 501: 97–101. Behavioral correlates of reaction time variability in children with
19 Lenartowicz A, Delorme A, Walshaw PD, et al. and without ADHD. Neuropsychology 2013; 27: 201–09.
Electroencephalography correlates of spatial working memory 30 Abdel-Kader AA, Mohamed NA, El Sayed BB, Amin OR, Halawa IF.
deficits in attention-deficit/hyperactivity disorder: vigilance, Continuous performance task in attention deficit hyperactivity
encoding, and maintenance. J Neurosci 2014; 34: 1171–82. disorder children. Egypt J Neurol Psychiat Neurosurg 2016; 53: 19–22.
20 Davis NO, Bower J, Kollins SH. Proof-of-concept study of an at-home, 31 Teicher MH, Ito Y, Glod CA, Barber NI. Objective measurement of
engaging, digital intervention for pediatric ADHD. PLoS One 2018; hyperactivity and attentional problems in ADHD.
13: e0189749. J Am Acad Child Adolesc Psychiatry 1996; 35: 334–42.
21 Hardy JL, Nelson RA, Thomason ME, et al. Enhancing cognitive 32 Boot WR, Simons DJ, Stothart C, Stutts C. The pervasive problem
abilities with comprehensive training: a large, online, randomized, with placebos in psychology: why active control groups are not
active-controlled trial. PLoS One 2015; 10: e0134467. sufficient to rule out placebo effects. Perspect Psychol Sci 2013;
22 Leark RA, Dupuy TR, Greenberg LM, Kindschi CL, Hughes SJTOVA. 8: 445–54.
Professional manual. Los Alamitos, CA: The TOVA Company, 2018. 33 Granic I, Lobel A, Engels RC. The benefits of playing video games.
23 Forbes GB. Clinical utility of the Test of Variables of Attention Am Psychol 2014; 69: 66–78.
(TOVA) in the diagnosis of attention-deficit/hyperactivity disorder.
J Clin Psychol 1998; 54: 461–76.

www.thelancet.com/digital-health Vol 2 April 2020 e178