Neural and behavioural measures from attention testing show no support for efficacy of neurofeedback treatment for adult ADHD

This study reports that a secondary analysis of a randomized controlled trial found no evidence that neurofeedback treatment improves sustained attention, inhibitory control, or associated neural markers in adults with ADHD compared to a waitlist control group.

The Gist

Imagine your brain is like a busy radio station. For people with ADHD, the signal can sometimes be staticky, making it hard to stay tuned to one station (sustained attention) or to quickly switch channels when needed (inhibitory control).

For decades, doctors have tried to fix this "static" using a treatment called Neurofeedback (NFB). Think of NFB as a personal trainer for your brain waves. You wear a headset that listens to your brain's electrical signals (EEG). When your brain does something good (like focusing), the screen shows a green light or a happy sound. When it drifts, the screen goes red. The idea is that by practicing this "brain gym" for 40 hours, you can learn to tune your radio station to a clearer signal, reducing your ADHD symptoms.

However, while many people feel like it helps in real life, scientists have been struggling to prove it works in strict, controlled experiments.

What This Study Did
The researchers in this paper decided to take a fresh look at a specific group of adults with ADHD who had undergone this 40-hour "brain gym" training. They wanted to see if the training actually changed the brain's hardware, not just how people felt.

They used two main tools to check the results:

1. The "Attention Test" (TOVA): Imagine a long, boring game where you have to press a button when you see a specific shape and not press it when you see a different one. This tests how well you can stay focused and control your impulses.
2. The "Brain Camera" (EEG): While they played the game, the researchers recorded the adults' brain waves to see what was happening inside the "radio station" before and after the training.

They compared two groups:

• Group A: Did the 40 hours of brain training.
• Group B: Waited on a list to do the training later (they didn't get the training yet).

What They Found
The results were surprisingly quiet. After the training period, the researchers looked for any "magic changes" in the data.

• The Game Scores: Did the trained group get better at the attention game than the waiting group? No. Their scores were essentially the same. In fact, the group that didn't get the training actually got slightly better at avoiding one specific type of mistake, likely just because they had practiced the game twice (a "practice effect").
• The Brain Waves: Did the brain waves change in a way that showed better focus? No. The electrical signals (specifically the N2 and P3 waves, which are like the brain's "I see it!" and "I'm ready!" signals) looked the same for both groups. The "static" didn't clear up.
• The Radio Frequencies: Did the brain's rhythm (oscillations) change? No. The specific frequencies the training was supposed to target remained unchanged.

The Bottom Line
The researchers concluded that, for this group of adults, the neurofeedback training did not produce any measurable improvement in their ability to focus or control their impulses, nor did it change the underlying electrical activity of their brains.

Why This Matters (According to the Paper)
The paper suggests that while neurofeedback is popular and well-tolerated, the strict scientific evidence for it working on the neural level is missing. It's possible that the training helps people feel better or develop coping strategies (like using a checklist), but it doesn't seem to "rewire" the brain's attention circuits in the way the theory predicts.

The authors also note that their study had some limitations, such as a smaller number of people finishing the study and the fact that the waiting-list group wasn't "blinded" (they knew they weren't getting treatment yet), which can affect how people perform. However, based on the data they did have, the "brain gym" didn't show the specific results they were looking for.

Technical Summary

1. Problem Statement

Attention-Deficit/Hyperactivity Disorder (ADHD) in adults is characterized by deficits in sustained attention and inhibitory control. Neurofeedback (NFB), a non-pharmacological intervention using real-time EEG to train self-regulation of brain activity, is widely used but lacks consistent empirical support for efficacy. While early meta-analyses suggested moderate-to-strong effects, recent rigorous Randomized Controlled Trials (RCTs) with blinded outcomes have yielded mixed or null results.

The core problem addressed is the discrepancy between anecdotal clinical success ("in the wild") and controlled trial data. Specifically, it remains unclear whether NFB produces specific, measurable changes in the neurocognitive mechanisms underlying attention deficits (sustained attention and inhibition) in adults, independent of global symptom ratings. This study seeks to determine if NFB modulates event-related cortical dynamics and behavioral performance during a sustained attention task.

2. Methodology

Study Design & Participants

• Design: Secondary analysis of a previously conducted RCT (Cowley et al., 2016) utilizing a Waiting-List Control (WLC) design.
• Participants:
  • ADHD Group: $N=44$ adults (23 in Treatment group [ADHD-T], 21 in Waitlist Control [ADHD-W]) measured at two time points: Intake (baseline) and Outtake (post-treatment).
  • Neurotypical Control (NTC) Group: $N=18$ matched adults, measured once.
• Intervention: 40 sessions (~1 hour each) of NFB training using classic protocols (Theta-Beta Ratio and Sensorimotor Rhythm) and inverse-target versions.
• Exclusions: Participants with neurological diagnoses, IQ < 80, or extreme scores on comorbidity scales (anxiety, depression, psychosis) were excluded. Medication was withheld for 48 hours prior to testing.

Task & Data Acquisition

• Task: The Test of Variables of Attention (TOVA), a continuous performance test (CPT) lasting ~22 minutes.
  • H1 (Vigilance): Infrequent targets (22.5% probability), challenging inattention.
  • H2 (Inhibition): Frequent targets (77.5% probability), challenging hyperactivity/impulsivity.
• EEG Recording: 128-channel EEG recorded during TOVA.
• Preprocessing:
  • Filtering (2–45 Hz), re-referencing to mastoids.
  • Artifact removal using ICA (FastICA) for ocular/muscular artifacts and automated statistical thresholding (FASTER).
  • Epoching: 2000 ms windows (-1000 ms to +1000 ms relative to stimulus).
  • Trial Selection: Due to low error rates, analysis focused on correct responses and correct inhibitions only. Error trials were excluded due to insufficient numbers for statistical power.

Statistical Analysis

• Approach: Bayesian Linear Mixed Models (LMMs) using the brms package in R.
• Key Interaction: Group (ADHD-T vs. ADHD-W) × Time (Intake vs. Outtake) to test for treatment-specific effects (difference-in-differences).
• Variables Analyzed:
  1. Behavioral: Mean Response Time (MRT), Response Time Variability (RTV), Commission Errors (COM), Omission Errors (OM), and $d'$ (sensitivity).
  2. Event-Related Potentials (ERPs): Mean amplitude and peak latency of N2 (150–250 ms) and P3 (330–430 ms) components in Frontal and Parietal Regions of Interest (ROIs).
  3. Spectral Power: Log-PSD at 4 Hz and 8 Hz (baseline and post-stimulus) and pre-stimulus $\alpha$-band (8–12 Hz) power.

3. Key Contributions

• Rigorous Re-evaluation: Provides a high-powered, Bayesian re-analysis of behavioral and neural data from a controlled NFB trial, moving beyond simple pre-post comparisons to isolate treatment-specific effects.
• Multi-modal Assessment: Simultaneously evaluates behavioral performance, time-locked neural dynamics (ERPs), and oscillatory spectral power, offering a comprehensive view of NFB's impact on neurocognitive processes.
• Mechanistic Insight: Specifically tests whether NFB alters the neural correlates of sustained attention (N2/P3 amplitudes, spectral dynamics) in adults, addressing the gap in understanding how NFB might work (or fail to work) mechanistically.

4. Results

Behavioral Performance (TOVA)

• Primary Finding: No evidence of NFB-specific improvement. The Group × Time interaction was small and uncertain (95% Credible Intervals [CrI] spanned zero) for all variables (MRT, RTV, OM, $d'$).
• Exception: A significant Group × Time × Condition interaction was found for Commission Errors in the H2 (frequent target) condition. However, follow-up analysis revealed this was driven by an improvement in the Waitlist Control (ADHD-W) group, while the Treatment group remained stable. This suggests a practice effect or regression to the mean rather than treatment efficacy.
• $d'$: Showed modest improvement over time across both groups, likely due to practice effects.

Neural Dynamics (ERPs)

• N2 and P3 Amplitudes/Latencies: No evidence that NFB modulated ERP components.
  • Visual inspection showed a general increase in P3 amplitude from Intake to Outtake in the ADHD-T group, but a comparable increase was observed in the ADHD-W group.
  • Bayesian LMMs confirmed no credible Group × Time interaction for N2 or P3 mean amplitudes or peak latencies in either Frontal or Parietal ROIs.
• Comparison to NTC: Both ADHD groups remained distinct from Neurotypical Controls at Outtake, with no convergence toward normal values attributable to treatment.

Spectral Analysis

• Oscillatory Power: No evidence that NFB modulated oscillatory activity.
  • No differential treatment effects were found for 4 Hz or 8 Hz power during baseline or post-stimulus periods.
  • Pre-stimulus parietal $\alpha$-band (8–12 Hz) power showed no treatment-related changes (Group × Time interaction $\beta = -1.58$, 95% CrI [-5.05, 1.89]).

5. Significance and Conclusion

Conclusion
The study concludes that NFB treatment did not produce meaningful, specific improvements in sustained attention or inhibitory control for adults with ADHD. The lack of differential change between the treatment and waitlist groups across behavioral, ERP, and spectral measures suggests that NFB does not engage or modify the specific neurocognitive mechanisms targeted by the training in this population.

Significance

• Challenging Efficacy: These findings contribute to the growing body of evidence questioning the efficacy of NFB for adult ADHD, particularly when measured against objective neurophysiological markers.
• Mechanism Clarification: The results suggest that the "anecdotal" success of NFB may not stem from direct neurocognitive modulation of attention networks (as measured by TOVA/EEG) but potentially from non-specific factors (e.g., placebo, therapeutic alliance, or compensatory strategies) that do not translate into measurable task performance or neural changes.
• Future Directions: The authors highlight limitations such as the lack of a sham control (blinding) and the potential for individual heterogeneity. They suggest future research should focus on individual-level prediction, biomarker-treatment interactions, and the use of active/sham controls to better isolate specific treatment effects.

Limitations Noted

• Blinding: The WLC design precluded blinded assessment, potentially introducing expectancy effects.
• Statistical Power: Attrition and data quality exclusions reduced sample sizes, limiting the ability to detect small effect sizes or heterogeneous responder subgroups.
• Task Constraints: The fixed TOVA structure and low error rates limited the analysis of error-related neural dynamics.
• Individual Alpha Frequency (IAF): The study did not individualize frequency bands based on each participant's IAF, which may have reduced sensitivity to $\alpha$-related effects.