Participants and design

In this study, we enrolled 28 children with a confirmed ADHD diagnosis (based on DSM-5 criteria), between March 2019 to February 2020. Characterization of clinical ADHD symptoms in children often relies upon parent report questionnaires for ease of administration [22–24]. As such, ADHD inclusion criteria for this study was based on the Vanderbilt Parent-Report measure as a screening tool [19]. Following the Vanderbilt criteria for inattention symptoms in ADHD (inattention subscale), parents had to report a symptom frequency of ‘often’ or ‘very often’ (2 or 3 on a scale from 0-‘never’ to 3-‘very often’) on at least six of nine items relating to inattention. In addition, we confirmed this ADHD diagnosis with the Mini-International Neuropsychiatric Interview for Children and Adolescents (MINI Kid, Section N [25]), administered by a trained clinician.

Exclusion criteria were brain malformation or injury, movement disorders, psychiatric conditions (e.g., bipolar, psychotic, and autism spectrum disorders), and hearing impairment. In addition, children had to be off antipsychotic or ADHD medications (stimulants, alpha-adrenergic medications, or atomoxetine) for the duration of the study and a washout period of at least 30 days prior to initiation. Further, children had to score ≥ 70 on the Perceptual Reasoning Index (PRI) on the Wechsler Intelligence Scale for Children–Fifth Edition and had to be able to comply with all testing and requirements. Participants who were eligible were then assessed with a battery of behavioral and neural outcome measures, both prior to and no later than 7 days after the 4-week AKL-T01 intervention.

Participants were recruited from online parent groups, the Sensory Neurodevelopment and Autism Program research registry, and Cortica Healthcare, Marin Center. The study is registered with ClinicalTrials.gov (NCT03844269) as a prospective trial and has approval through WIRB Copernicus Group (IRB Tracking Number: 20190399, Study Number: 1254952). Participants’ primary caregivers provided written consent on behalf of their child and children provided informed assent.

Primary outcome metric: Neural assessment of attentional control

Exploratory outcome metrics: Objective behavioral measures of attention

To assess objective behavioral improvements from AKL-T01, we used two computerized tasks to probe components of attentional control abilities.

Exploratory outcome metrics: Parent-reported ADHD symptoms

AKL-T01 intervention

The AKL-T01 intervention is a game-based digital therapeutic developed by Akili Interactive, which implements proprietary algorithms designed to train attentional control by enhancing interference management (multi-tasking) at an adaptive and personalized degree of difficulty. In brief, the AKL-T01 intervention is a self-guided at-home intervention that involves both a perceptual discrimination targeting task, in which users respond to the instructed stimulus targets and ignore the stimulus distractors (similar to a go/no-go task), and a sensory motor navigation task, in which users continuously adjust their location to interact with or avoid positional targets. Performance in each task is assessed during single and multitask conditions.

The AKL-T01 intervention was preloaded and administered on iPad Mini 2 tablets (Apple, USA). Children were instructed to complete 5 ‘missions’ at least 5 days per week for 4 weeks for a total of approximately 25 minutes of gameplay per day (at least 100 missions). Adherence was monitored remotely by study coordinators using an electronic dashboard created by Akili. If a participant had more than two incomplete days of the intervention, a reminder phone call was made to the parents, while also providing support and feedback to the parents and children with the intervention. Additionally, AKL-T01 generates automatic reminders. After a participant completed 20 sessions of the AKL-T01 intervention, a follow-up research appointment was scheduled with the parents. All data, except for the AKL-T01 intervention, was collected at the Cortica Neurodevelopment Centers, Marin Campus.

Statistical analysis

Changes in our primary outcome metric—neural measures of attentional control (MFT EEG power)—were assessed with a repeated-measures ANOVA with within-subjects factors of session (pre and post), composite 120 ms time-windows (early, peak, and late), and electrode (Fz, AFz, AF3, AF4, F1, and F2). Given that we aimed to replicate our previous analytical approaches and had no a priori hypotheses about unique effects of electrode, all MFT analyses presented are collapsed along the factor of electrode (i.e., we averaged electrode values after all other processing steps). In addition to this approach with a within-subjects factor of time-window, we also interrogated each 120 ms window separately, given that MFT activity at these windows may reflect different ongoing cognitive processes throughout the duration of the task trials. To do so, we conducted 3 separate ANOVAs for each time window, with a within-subjects factor of session (pre and post).

Improvements on exploratory metrics of attentional control tasks and survey-based measures were assessed with paired t-tests comparing pre- and post-intervention performance. When examining intervention-related changes, we calculated gain scores for each metric, where higher values indicate improvements in performance or neural activity. To examine relationships between neural improvements and behavioral and clinical improvements, we performed Pearson correlations between neural and cognitive and clinical gains scores. For statistically significant correlations, we also report the strength of the correlation as ‘very strong’ (r = 0.9–1.0), ‘strong’ (r = 0.70–0.89), ‘moderate’ (r = 0.40–0.69), ‘weak’ (r = 0.10–0.39), or ‘negligible’ (r < 0.10).

All statistical analyses were conducted using SPSS 24.0 (SPSS Inc.). We report significant effects at a threshold of p ≤ 0.05 and also report marginal ‘trends’ at p ≤ 0.10. Further, effect sizes for intervention-related changes are reported as Cohen’s d, where 0.2 = small effect, 0.5 = medium effect, and 0.8 = large effect [35]. We did not correct for multiple comparisons for each test conducted, as our primary goal was to examine neural markers of attentional control and our exploratory analyses aimed to replicate findings using similar metrics in patients with inattention and SPD [18].

Participant demographics

Twenty-eight participants meeting our inclusion and exclusion criteria were enrolled in the study. Twenty-five of these participants successfully completed the assessments and intervention (age mean ± SD: 10.44 ± 1.23 years old, age range: 8–12; number of males: 20, 80% of total sample). Of the 3 participants that did not complete the study, one participant withdrew during the pre-intervention baseline assessments and two participants did not complete the intervention (Fig 1). The 25 participants that completed the study had normal performance on the Wechsler Intelligence Scale for Children (WISC-V, see S1 Table in S1 File) and showed excellent adherence with the AKL-T01 intervention (percentage of ‘missions’ completed: 100 ± 20%, range = 61–136%). Note that participants could have opted to play 7 days a week (5 days were recommended) and additionally extend the gameplay until the scheduled outcome assessment post-intervention (when they returned their devices). Due to technical difficulties during data collection, outcome measurement data was not available for all 25 participants. Further, EEG data from 2 participants was excluded due to excessive noise (rejection of more than 30% of target trials due to voltage fluctuations greater than ± 100 μV deflections within an epoch [36, 37]) at the pre- or post-intervention assessment (available data: EEG N = 22, perceptual discrimination task N = 25, sustained attention task N = 24, Vanderbilt parent-report survey N = 25).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. CONSORT flow diagram.

https://doi.org/10.1371/journal.pone.0261981.g001

Primary outcome metric: Intervention-related changes in neural markers of attentional control

To interrogate effects of AKL-T01 on overall attention-related neural activity during the perceptual discrimination task, we conducted a repeated-measures ANOVA on the MFT EEG data, with within-subjects factors of session (pre and post) and 120 ms composite time-window (early, peak, and late post-stimulus onset). We observed a main effect of session, (F(1,21) = 4.76, p = 0.04, d = 0.44), suggesting that there was a general increase in MFT following the intervention (Fig 2A). We did not observe an interaction between session and composite time-window (F(2,42) = 0.78, p = 0.46, d = 0.11). However, given our a priori hypotheses that the intervention may affect the cognitive processes occurring at each window differently, we next examined each 120 ms composite window separately. To do so, we conducted 3 repeated-measures ANOVAs with a within-subjects factor of session, separately for each time window. We found increases in the early (0–120 ms: F(1,21) = 7.53, p = 0.01, d = 0.62) and late (760–880 ms: F(1,21) = 4.51, p = 0.046, d = 0.52) time windows. There was a numerical increase in MFT activity at the peak time period of 160–280 ms following stimulus onset, however in contrast to our previous work [17, 18], this change did not reach significance (F(1,21) = 0.82, p = 0.38, d = 0.19) (Fig 2B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Changes in primary outcome metrics: Neural markers of attention.

a Time course of midline frontal theta (MFT) pre- and post-intervention at each 40 ms time bin during the perceptual discrimination task, where 0 ms represents stimulus onset. Solid lines represent the mean and shaded areas represent 95% bootstrapped confidence intervals (1000 bootstrap iterations). b Intervention-related improvements in MFT at early, peak, and late 120 ms composite windows, also illustrated through topographic heat maps with the 6 electrodes of interest highlighted with a dotted bounding box. Data are presented as mean ± 95% bootstrapped confidence intervals (1000 bootstrap iterations), and circles represent individual participants. P-values represent repeated-measures ANOVAs comparing pre- and post-intervention MFT at each composite window.

https://doi.org/10.1371/journal.pone.0261981.g002

Exploratory outcome metrics: Intervention-related changes in objective behavioral measures of attention

Perceptual discrimination task.

For the perceptual discrimination task, we observed an improvement on response time (RT) (mean ± SE gain = 39.67 ± 11.86 ms; t(1,24) = 3.35, p = 0.003, d = 0.62), suggesting that participants performed faster on this measure following AKL-T01 (Fig 3A). However, we did not observe an improvement on response time variability (RTV; t(1,24) = -0.76, p = 0.45, d = 0.12) (see S2 Table in S1 File for full descriptive statistics for each metric).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Changes in exploratory outcome metrics: Behavioral and parent-report measures of attention.

Intervention-related changes in objective behavioral and parent-report measures of attention, specifically a response time (RT) on the perceptual discrimination task, b tau on the sustained attention task, and c points on the Vanderbilt inattention subscale (parent-report). Data are presented as mean ± 95% bootstrapped confidence intervals (1000 bootstrap iterations), and circles represent individual participants. d Pearson correlations between midline frontal theta (MFT) improvements at each composite window and the behavioral and parent-report improvements.

https://doi.org/10.1371/journal.pone.0261981.g003

Exploratory outcome metrics: Intervention-related changes in parent reports of clinical inattention symptoms

To test whether AKL-T01 led to changes in ADHD inattention symptoms, we examined changes in our clinical outcome measure–the inattention subscale of the Vanderbilt parent report. After the AKL-T01 intervention, parents observed significantly fewer ADHD inattention symptoms (mean ± SE gain = 5.00 ± 0.94 points; t(1,24) = 5.32, p < 0.001, d = 1.25), suggesting that participants had fewer clinical inattention symptoms following the intervention (Fig 3C). Furthermore, 44% (11 participants out of 25) of individuals who previously met inclusion criteria (≥ 6 out of 9 Vanderbilt inattention symptoms in the ‘often’ or ‘very often’ range) no longer met this cutoff after AKL-T01.

Relationships between neural and behavioral changes

To examine associations between training-related neural changes and the observed behavioral and clinical improvements, we performed correlations between neural gain scores at the three composite time-windows and improvements on the behavioral and survey-based assessments (Fig 3D). Notably, changes in RT on the perceptual discrimination task were related to early and peak post-stimulus MFT gains (r(21) = 0.69, p < 0.001 and r(21) = 0.68, p = 0.001, respectively; both ‘moderately’ sized correlations), but not to late MFT gains (r(21) = 0.23, p = 0.31). Further, we also found that early and peak post-stimulus MFT gains were related to changes in tau on the sustained attention task (r(20) = 0.72, p < 0.001 and r(20) = 0.39, p = 0.08, respectively; ‘strong’ and ‘weak’ correlations, respectively), but not to late MFT gains (r(20) = -0.11, p = 0.65). We did not, however, find any relationships between MFT and Vanderbilt inattention improvements.

Intervention-related changes in neural markers of attentional control

First, we demonstrate that AKL-T01 increases MFT broadly following post-stimulus presentation in children with ADHD, suggesting that these patients are better able to engage control functions and monitor ongoing cognitive processes to support attention-centric performance. Other neuromodulatory approaches, such as neurofeedback training, have also pointed to similar plasticity of MFT, by showing that targeted neurofeedback can broadly enhance MFT activity during training as well as during ‘baseline’ periods [38]. Critically, this shift in MFT concurrently improves behavioral attentional control abilities that require proactive control [39]. In addition, non-invasive neurostimulation applied in the theta frequency benefits complex cognitive functions, such as interference processing [40]. The findings of the present study expand on this work and suggest that a digital therapeutic targeting specific brain networks is an effective form of neuromodulation that can broadly enhance MFT activity and attention-related behavior.

In addition to general MFT improvements during task performance, we also observed that these neural changes were most pronounced in the early and late stages of the perceptual discrimination task trials. Examining each time window separately, MFT changes were most evident immediately following stimulus presentation (‘early’) and in preparation for the upcoming task trial (‘late’), although it should be noted that we did not observe a formal statistical interaction between session and time window for these analyses. Early processing (approximately 100 ms post-stimulus presentation) is thought to reflect initial orienting of attentional resources [41–43], while late processing (approximately 650–800 ms post-stimulus presentation) may reflect post-response processes, such as suppressing the previous stimulus and updating in preparation for the upcoming trial [26]. With respect to the present study, children with ADHD have shown impaired neural oscillations for both early and late stimulus processing [26]. The neural findings reported here demonstrate that AKL-T01 improves MFT signals supporting attentional control in children with ADHD, particularly (1) early processes related to attentional orienting and (2) late processes related to future event preparation. It is important to note that, unlike previous studies in older adults and children with SPD and inattention [17, 18], we did not observe significant changes in peak MFT. However, given that the children in this study did exhibit numerical peak MFT gains that also were related to positive changes in task performance, the lack of a significant intervention effect may be a function of statistical power or the heterogenous ADHD population in the present study. Future research with larger sample sizes may help confirm the presence of peak MFT improvements in this population.

Intervention-related changes in objective behavioral and clinical metrics of attention

In addition to the intervention-related neural improvements in MFT activity, we also replicate previous work demonstrating that AKL-T01 can improve both behavioral and clinical measures of attention in children with inattention difficulties [18, 20]. We further demonstrate that improved performance on these computer-based tasks was not simply due to general improvements in motoric speed, indicating that AKL-T01 improved indices of more complex attentional behaviors.

Critically, improvements on these computerized measures were also related to improvements on our EEG-based metric of attention (MFT activity), particularly for the ‘early’ and ‘peak’ time windows. These results suggest that there are benefits to including neuroimaging markers as an adjunct to assessing behavioral and clinical outcome measures, and further point to a potential underlying neural mechanism of the observed changes in behavior. Interestingly, these correlational results also point to individual differences in intervention-related improvements–children that exhibit the greatest neural changes are those that improve the most behaviorally. These findings signal the potential importance for developing personalized interventions that enhance improvements at the individual patient level.

Finally, we also found that AKL-T01 improved clinical measures of ADHD–the Vanderbilt parent report used for inclusion criteria of ADHD inattention symptoms. This finding is consistent with previous work using AKL-T01 interventions in children with attention difficulties and indicate that, in this study, children with ADHD had fewer parent-reported symptoms of inattention after the intervention. Notably, almost half (44%) of the children no longer met inclusion criteria for Vanderbilt inattention symptoms after the intervention. This is in line with our previous report that 33% of children with SPD and inattention showed this level of improvement following training [18]. These results provide compelling evidence that AKL-T01 has benefits that generalize beyond computerized tasks, and also reduces inattention behaviors observed by parents in real-world settings.