Neurofeedback for Attention-Deficit/Hyperactivity Disorder

Introduction

Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental condition characterized by developmentally inappropriate, persistent, pervasive, and impairing inattention and/or hyperactivity-impulsivity.¹ ADHD medications, particularly psychostimulants, reduce symptoms and impairment at least in the short term^2,3 and are recommended as part of multimodal treatment approaches, including psychosocial therapies or psychoeducation programs.^3,4

Neurofeedback has been proposed as a nonpharmacological treatment for ADHD. It deploys a brain-computer interface to train self-regulation of brain activity using real-time audiovisual feedback of brain activity.⁵ Electroencephalogram (EEG)–based neurofeedback is the most common, while newer techniques training hemodynamic brain signals for deeper and more focal regulation are being explored (eg, real-time functional magnetic resonance imaging [rt-fMRI] neurofeedback and functional near infrared spectroscopy [fNIRS] neurofeedback).^{6,7,8,9,10,11} Two EEG neurofeedback protocols have typically been used. Frequency band training aims to decrease the power of θ activity (4-7 Hz) or increase the β activity (13-30 Hz) power to promote attention. Event-related slow cortical potential regulation aims to increase or decrease negative potential shifts linked to cortical activation and cognitive preparation.^12,13,14,15 Building on evidence from learning theories, such as operant conditioning (ie, implicit learning through reinforcement) or skill acquisition (ie, explicit, goal-directed learning), neurofeedback protocols are premised on the notion that by learning to self-regulate or normalize ADHD-related brain activity, one can alleviate behavioral symptoms and related impairments.^13,16

However, the clinical efficacy of neurofeedback as a treatment for ADHD remains disputed despite nearly 5 decades of research. Randomized clinical trials (RCTs) have often been of poor quality due to inadequate blinding after treatment allocation (eg, use of unblinded parent reports or lack of clarity of blinding status), use of nonstandardized suboptimal treatment protocols, or failure to assess self-regulation or learning.^16,17,18,19

In 2013, the European ADHD Guidelines Group (EAGG) published a meta-analysis,²⁰ followed by an update²¹ published in 2016, of RCTs of EEG neurofeedback in children and adolescents with ADHD. A central feature of these meta-analyses was the exploration of the impact of blinding by contrasting 2 types of ADHD symptom outcomes: those reported by individuals most proximal to the intervention setting and so judged least likely to have been blinded (eg, parent reports on home implemented neurofeedback) and those reported by individuals confirmed as certainly blinded or who were probably unaware of the treatment allocation, either due to their remoteness from the treatment setting (eg, teacher reports for home-based treatment) or because they were independent observers (eg, clinicians not involved in treatment). The EAGG 2013 analysis, only powered to analyze ADHD total symptoms, showed a significant neurofeedback-related improvement based on most proximal reports (SMD, 0.59; 95% CI, 0.31 to 0.87) but this effect was halved and became nonsignificant with probably blinded reports (SMD, 0.29; 95% CI, −0.02 to 0.61), suggesting outcome assessor bias. The EAGG 2016 analysis confirmed and extended these findings, showing significant but small neurofeedback-related reductions across core ADHD symptoms for most proximal reports (SMDs, 0.26 to 0.36) but not for probably blinded reports (SMDs, 0.06 to 0.17). Further, for RCTs using established standard neurofeedback protocols (as defined by Arns et al 2014¹⁵ and 2020¹⁶), most proximal effects sizes increased (SMDs, 0.45 to 0.55) but probably blinded outcomes could not be analyzed due to insufficient data. Neurofeedback did not improve neuropsychological functioning. In short, EAGG meta-analytic evidence does not support neurofeedback as a stand-alone treatment for ADHD. Broadly similar results have been found in other meta-analyses of blinded RCTs of neurofeedback for ADHD.^{14,22,23,24,25,26,27,28,29}

A considerable number of neurofeedback RCTs have been published since the last search conducted by the EAGG in 2015. Many RCTs now include larger samples, longer-term follow-up periods, well-controlled designs, and more have probably blinded outcomes. Some have tested the efficacy of rt-fMRI neurofeedback or fNIRS neurofeedback. In light of this, we present an updated meta-analysis of neurofeedback RCTs for ADHD to provide a more accurate and comprehensive estimate of the value of neurofeedback as a stand-alone treatment for ADHD in terms of core symptom reduction and improved neuropsychological performance. We also conducted sensitivity analyses with probably blinded outcomes relating to whether effects were stronger when standard EEG neurofeedback protocols were followed and when there was evidence that learning or self-regulation had occurred. We also ran exploratory analyses for individuals most proximal to the intervention to see if novel neurofeedback approaches targeting hemodynamic signals of brain activity were efficacious and to compare neurofeedback to medication and other nonpharmacological interventions in head-to-head trials.

Methods

This preregistered systematic review and meta-analysis (CRD42022290005) was reported according to Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA)³⁰ and PRISMA-S³¹ reporting guidelines (eAppendix in Supplement 1).

Results

Discussion

In this systematic review and meta-analysis of RCTs focusing on at least probably blinded neuropsychological outcomes, we found no evidence to support the use of current forms of neurofeedback as stand-alone treatment for ADHD. This generally negative picture had 2 possible exceptions. The first is that established standard protocols led to a nominally significant benefit on probably blinded measures of total ADHD symptoms. However, 4 observations temper the importance of this finding. First, it could be due to the standard nature of the protocols or the mixture of different neurofeedback approaches that were analyzed (frequency band training vs slow cortical potential regulation). Second, the small improvement (SMD, 0.21) likely falls short of clinical value and is smaller than what we found in 2016 with only 3 trials available for analysis (SMD, 0.36).²⁰ Third, the improvement was not observed in probably blinded measures of inattention or hyperactivity-impulsivity symptoms, separately. Fourth, the longer-term persistence of this improvement is unclear. Insufficient trials prevented an analysis of longer-term probably blinded outcomes while outcomes for individuals most proximal to the intervention showed no significant improvement. The second exception is that neurofeedback improved processing speed, a core neuropsychological component that may underpin higher-order cognitive impairment in ADHD^64,65—an effect that was also found in longer-term follow-up periods. The clinical value of these improvements is questionable, as they were barely significant with significant heterogeneity. Furthermore, although effect size estimates were small to moderate, they were derived from analyses with multiple dependent effect size estimates (6 of 15 in the postassessment analysis and 6 of 8 in the follow-up analysis) across 3 RCTs with overlapping comparator samples, potentially inflating the observed effects.

Several of our findings warrant discussion. First, methylphenidate was superior to neurofeedback in alleviating core ADHD symptoms based on outcomes for individuals most proximal to the intervention (but not probably blinded outcomes due to insufficient data). The estimated effect sizes (k = 7; SMD range, −0.49 to −0.74; 95% CI range, −0.73 to −0.23) were similar to those from meta-analyses³ of RCTs comparing methylphenidate with placebo in ADHD (SMD range, −0.49 to −0.82; 95% CI range, −1.16 to −0.62), indicating limited clinical benefit of neurofeedback. We also found no significant difference between neurofeedback and other nonpharmacological treatments (ie, cognitive training and exercise), suggesting a general ineffectiveness of nonpharmacological interventions for ADHD core symptoms when considering outcomes rated by blinded individuals. This aligns with our recent meta-analytical evidence showing negligible efficacy of cognitive training for ADHD.³⁹ Second, although no benefits were found with hemodynamic neurofeedback techniques (eg, fMRI and fNIRS), this is based on only 5 trials; further research is warranted given the potential to target deeper ADHD-relevant brain regions (eg, opercular right inferior frontal cortex or basal ganglia) with greater spatial (but not temporal) resolution than EEG-based approaches.⁹ Third, no significant clinical improvements were found at longer-term follow-up points (ie, 6 to 12 months after neurofeedback training). Fourth, effect sizes for probably blinded and most proximal outcomes were smaller in the current than in the 2016 meta-analysis (SMD ranges: probably blinded, 0.03 to 0.04 vs 0.06 to 0.15 and most proximal, 0.26 to 0.36 vs 0.19 to 0.26), but meta-regressions showed no significant relationship between publication year and effect size, suggesting factors other than improved trial design caused this reduction over time.

From a methodological point of view, two findings are noteworthy. First, neurofeedback effects did not vary as a function of how active the comparator arm was, as effects were not observed even in trials with less active control conditions. However, the limited number of trials with the least active and nonpharmacological alternative control conditions weakens this conclusion. Second, there was no relationship between whether an RCT established the existence of neurofeedback learning and clinical improvement in ADHD symptoms, suggesting that previous evidence of clinical improvement could in this situation potentially result from nonspecific, incidental impacts of neurofeedback protocols (eg, wraparound treatment, skill development, and improving tolerance of effortful tasks) rather than self-regulation of brain activity or that the relationship holds only for learning subgroups.⁶⁶ However, this interpretation is based only on 5 RCTs, so the mechanisms driving self-regulation and the individual clinical effects need further investigation.

Given these negative results, what is the future of neurofeedback as a treatment for ADHD? First, the limited but significant improvements with standard neurofeedback protocols and in processing speed needs further validation through methodologically sound trials, as this could help refine future approaches targeting higher-order functions.¹³ Second, although RCTs provide aggregate-level data, an individualized neurofeedback approach could be developed by tailoring neurofeedback parameters to individual characteristics. However, reliable baseline indicators (eg, demographic, neuropsychological, and brain function differences) of potential responders to specific training regimes (eg, slow cortical potential regulation training depending on baseline slow cortical potentials) are currently lacking.¹³ We only had sufficient RCTs to show that baseline medication status did not predict neurofeedback effects. Third, combining neurofeedback with noninvasive brain stimulation techniques (eg, repetitive transcranial magnetic stimulation, transcranial direct current stimulation, and external trigeminal nerve stimulation) could be potentially fruitful, but evidence supporting their efficacy when applied without neurofeedback is limited.^67,68,69 Additionally, incorporating virtual reality or transfer trials (where no immediate feedback is given) could provide a more ecologically valid means to self-regulate symptoms in everyday life, but this approach remains underexplored.⁷⁰

Conclusion

In conclusion, after decades of research, we found no group-level evidence supporting neurofeedback as a stand-alone treatment for ADHD. Future studies should therefore weigh the cost and benefits of administering neurofeedback over other treatments. Advances in precision medicine, including brain imaging techniques, might eventually identify specific individuals with ADHD for whom neurofeedback could be effective and safe.