Assessing Brain-Behaviour Coupling in Non-invasive Brain Stimulation Using Reliable Change Indices: Evidence from pre-Supplementary Motor Area - right Inferior Frontal Gyrus transcranial Alternating Current Stimulation

By pooling data from three studies and employing robust Reliable Change Indices, this research found no evidence of a systematic individual-level coupling between 20 Hz tACS-induced changes in preSMA-rIFG connectivity and response inhibition, suggesting that such effects are likely negligible and highlighting the need for better-powered designs and more sensitive neural markers in non-invasive brain stimulation research.

The Gist

The Big Question: Does "Tuning the Radio" Change the Song?

Imagine your brain is a complex radio station with many different channels. Scientists have developed a tool called tACS (Transcranial Alternating Current Stimulation), which is like a gentle, non-invasive "tuner" that you can place on your head. The goal is to tune specific parts of the radio (the brain) to a better frequency to improve how you think or act.

In this study, the scientists wanted to tune two specific "dials" in the brain:

1. The preSMA: Think of this as the brain's "brake pedal."
2. The rIFG: Think of this as the brain's "stop sign."

Together, these two areas help you stop yourself from doing something impulsive (like hitting the gas when you see a red light). The researchers hoped that by using their "tuner" (tACS) to make the connection between these two dials stronger, people would get better at stopping themselves.

The Problem: The "Group" vs. The "Individual"

In the past, scientists would test this on a group of people and say, "Hey, on average, the group got better at stopping!" That's like saying, "The average temperature in this city went up by 5 degrees."

But that doesn't tell you if you specifically got better. Maybe some people got much better, some got worse, and some didn't change at all, but the average looked good.

The big question this paper asks is: If a specific person's brain "tuning" gets stronger, does that specific person get better at stopping?

The Old Way vs. The New Way

The Old Way (The "Ratio" Trap):
Usually, scientists measure change by taking a "before" score and dividing it by an "after" score (like a ratio).

• The Analogy: Imagine you are trying to measure how much a plant grew. If you start with a tiny sprout (1 inch) and it grows to 2 inches, that's a 100% increase! But if a giant tree (100 inches) grows to 101 inches, that's only a 1% increase.
• The Problem: This method is mathematically messy. It creates a lot of "noise" and can make it look like there is a connection between brain changes and behavior when there isn't one, or hide a real connection. It's like trying to hear a whisper in a room full of static.

The New Way (The "Reliable Change Index" or RCI):
The authors used a smarter math tool called the Reliable Change Index (RCI).

• The Analogy: Instead of just looking at the raw numbers, RCI asks: "Is this change big enough to be real, or is it just random noise?" It accounts for the fact that our brain measurements aren't perfect (like a slightly fuzzy radio signal). It filters out the static so we can hear the true signal.

What They Did

The researchers didn't just look at one small experiment. They combined data from three different studies involving 69 young people.

1. They gave everyone the "brain tuner" (tACS) to try to strengthen the connection between the brake pedal and the stop sign.
2. They measured the brain connection (using EEG, like a brain microphone).
3. They measured how fast people could stop (using a "Stop Signal" task).
4. They used their new "RCI" math to see if the people whose brains changed the most were the ones who got the best at stopping.

The Result: The "Silent" Connection

The bad news: They found no connection.

Even though the "brain tuner" worked on the group level (the average brain connection did get stronger), it didn't predict who got better at stopping.

• Some people had huge brain changes but no change in their stopping ability.
• Some people had no brain changes but got much better at stopping.
• Some people got worse at both.

It was like tuning a radio and hearing the static get quieter, but the song playing didn't get any clearer for any specific listener. The correlation was essentially zero.

Why Does This Matter?

1. It's not a failure of the brain: The study confirms that the "brake pedal" and "stop sign" areas are important for stopping. The problem isn't that the brain doesn't work that way; it's that our current "tuning" method is too blunt to control individual results.
2. Individual brains are wild: Every person's brain is unique. Anatomy, genetics, and even what you had for breakfast can change how the "tuner" works. A one-size-fits-all approach doesn't work well for individuals.
3. Better Math is needed: The study proves that using the old "ratio" math was likely giving scientists false hope or confusing results. The new "RCI" method is a much cleaner way to look at data.

The Takeaway

Think of this study as a reality check for the field of brain stimulation.

• The Promise: We can change brain activity with electricity.
• The Reality: Just because we change the brain activity doesn't mean we can reliably predict who will get better at a task.

The authors suggest that in the future, we need "smart tuners" that adjust in real-time based on what the brain is doing right now (like a closed-loop system), rather than just setting a dial and hoping for the best. Until then, we have to be careful about promising that brain stimulation will fix specific problems for specific people.

Technical Summary

1. Problem Statement

Non-invasive brain stimulation (NiBS) research frequently reports group-level improvements in neurophysiological markers (e.g., connectivity) and behavioral outcomes (e.g., reaction time). However, a critical gap exists in establishing individual-level brain-behaviour coupling: determining whether the magnitude of neurophysiological change in a specific individual predicts the magnitude of their behavioral change.

Previous studies attempting to address this face three major limitations:

1. Underpowered Designs: Sample sizes sufficient for detecting group-level intervention effects are typically too small for robust correlation analyses of individual differences.
2. Statistical Flaws in Change Scores: Most studies use ratio-based change scores (e.g., post/pre ratios) to standardize change. These metrics are statistically problematic because they increase sampling variation, violate assumptions of normality and homoscedasticity, and can introduce spurious correlations.
3. Measurement Error: Standard change scores often fail to account for the inherent unreliability of neurophysiological and behavioral measures, which attenuates observed correlations.

2. Methodology

Data Source & Sample:

• Pooled Dataset: The authors conducted a coordinated analysis of archived individual-level data from three independent studies (published 2023–2025) conducted by the same research group.
• Total N: 69 young, neurologically healthy participants (38 males, mean age ~24 years).
• Intervention: All studies utilized 20 Hz transcranial alternating current stimulation (tACS) targeting the pre-supplementary motor area (preSMA) and the right inferior frontal gyrus (rIFG). These regions form a cortical network critical for response inhibition.
• Designs: The pooled data included between-subject and within-subject (crossover) designs, with participants receiving either active tACS or sham (sham data were excluded).

Measures:

• Neurophysiological: Functional connectivity between preSMA and rIFG measured via EEG Imaginary Coherence (ImCoh) in the beta band during resting state.
• Behavioral: Response inhibition measured via Stop-Signal Reaction Time (SSRT) using a Stop-Signal Task (SST).

Statistical Approach:

• Reliable Change Indices (RCI): Instead of ratio scores, the authors calculated RCIs for both ImCoh and SSRT.
  • Formula: $RCI = (Post - Pre) / \sqrt{2(1 - r_{xx})} \times SD_{pre}$, assuming a test-retest reliability ($r_{xx}$) of 0.80.
  • Sign-Inversion: SSRT scores were sign-inverted so that positive RCI values indicated improvement (faster stopping).
  • Harmonization: RCI scores were z-standardized within each study (mean=0, SD=1) to control for study-specific variance before pooling.
• Primary Analysis: A Linear Mixed-Effects Regression model was used with the pooled z-standardized RCIs.
  • Fixed Effects: ImCoh RCI and Study (Project).
  • Random Effects: Participant ID (to account for non-independence and individual variability).
• Sensitivity Analysis: The authors re-ran the analysis using traditional post/pre ratio scores to empirically demonstrate the statistical limitations of this common method compared to RCI.

3. Key Contributions

1. Methodological Benchmark: The paper proposes and validates a rigorous framework for assessing individual-level brain-behaviour coupling in NiBS. It advocates for the use of RCI over ratio-based scores to account for measurement reliability and reduce statistical noise.
2. Pooled Power Analysis: By aggregating data from three studies ($N=69$), the authors achieved sufficient statistical power to detect small-to-medium effects (power $\approx$ 0.72 for $r=0.3$), addressing the chronic underpowering of individual-difference analyses in the field.
3. Empirical Test of Mechanism: The study directly tests the hypothesis that increased preSMA-rIFG beta synchrony (induced by 20 Hz tACS) linearly predicts improved response inhibition at the individual level.

4. Results

• Null Association: There was no meaningful association between tACS-induced changes in preSMA-rIFG connectivity (ImCoh RCI) and changes in response inhibition (SSRT RCI).
  • Correlation: $r = 0.013$ (uncorrected), $p = 0.92$.
  • Regression: The ImCoh RCI was not a significant predictor of SSRT RCI ($\beta = -0.05$, $p = 0.99$).
  • Variance Explained: The model accounted for a negligible proportion of variance ($R^2_{marginal} = 0.004$).
• Inconsistency: Project-wise correlations were small, non-significant, and inconsistent in direction (ranging from $r = -0.20$ to $r = 0.05$).
• Sensitivity Analysis (Ratio Scores):
  • Using ratio-based scores yielded a similarly negligible correlation ($r = 0.014$).
  • Distributional Violations: The ratio scores showed severe deviations from normality (Shapiro-Wilk $W = 0.16$, $p < .001$), whereas RCI distributions were approximately normal ($W = 0.975$, $p = .183$). This confirms the methodological superiority of RCI for statistical validity.
• Reliable Change Thresholds: Only 7 out of 69 participants exceeded the $\pm 1.96$ threshold for reliable change on either measure, precluding subgroup analyses.

5. Significance and Conclusion

• Decoupling Group vs. Individual Effects: The results demonstrate that group-level efficacy does not guarantee individual-level predictability. While the contributing studies reported group-level improvements in connectivity or behavior, these changes did not covary systematically across individuals.
• Challenge to Linear Models: The findings challenge the assumption that a simple linear metric of connectivity (ImCoh) serves as a robust biomarker for behavioral gain in NiBS. The relationship between preSMA-rIFG communication and inhibition may be non-linear, context-dependent, or mediated by other network factors not captured by static resting-state connectivity.
• Implications for NiBS Research:
  • Methodology: Future studies must move beyond ratio-based change scores and underpowered post-hoc correlations. The combination of RCI and mixed-effects modeling is recommended for assessing individual differences.
  • Clinical Translation: The high inter-individual variability in NiBS responses suggests that "one-size-fits-all" fixed-parameter protocols are insufficient.
  • Future Directions: The authors suggest exploring closed-loop (state-dependent) stimulation to adapt to moment-to-moment brain states, utilizing more sensitive neural markers (e.g., PCA-based change scores), and integrating multimodal imaging to better capture the complex mechanisms of brain-behaviour coupling.

In summary, this paper provides a cautionary but methodologically advanced analysis, concluding that under current 20 Hz tACS protocols, preSMA-rIFG connectivity changes are not a reliable predictor of individual improvements in response inhibition.