Predictive value of EEG/ECG Biomarkers for Treatment Response in Depression

This retrospective study of 153 adults with depression found that treatment choices aligned with individualized EEG/ECG biomarker reports were associated with significantly higher response rates compared to non-concordant treatments, suggesting that such electrophysiological decision support tools could improve clinical outcomes.

The Gist

The Big Problem: Guessing the Right Medicine

Imagine you have a broken leg. You go to the doctor, and they put a cast on it. But sometimes, the cast doesn't fit right, or the bone is broken in a weird way, and the leg doesn't heal. You have to take the cast off, try a different one, maybe get surgery, and wait months to see if it works.

This is exactly what happens with depression. Millions of people take antidepressants or try therapies like brain stimulation (rTMS) or shock therapy (ECT). But often, the first treatment doesn't work. Doctors have to play a game of "guess and check," switching medications or therapies for months or even years while the patient suffers. It's expensive, exhausting, and frustrating.

The New Idea: A "Weather Forecast" for Your Brain

The researchers in this study wanted to stop the guessing game. They asked: Can we look at a patient's brain and heart before starting treatment to predict which "key" will fit the "lock"?

They used a special tool called DeepPsy. Think of this tool as a high-tech weather forecast for your nervous system.

• The Scan: Before treatment, patients sat quietly while the machine recorded their brainwaves (EEG) and heart rhythms (ECG).
• The Report: The computer analyzed these signals and generated a report. It didn't diagnose the disease; instead, it said, "Based on your brain's current 'weather,' you are most likely to get sunshine (recovery) if you try Treatment A, but you might get a storm if you try Treatment B."

The Experiment: Did the Forecast Help?

The researchers looked back at 153 patients who had already been treated in a real-world hospital setting. They divided the patients into two groups:

1. The "Followed the Map" Group: The doctor chose the treatment that the DeepPsy report said was the best match.
2. The "Ignored the Map" Group: The doctor chose a different treatment, either because they didn't use the report or because they decided to go against the recommendation.

The Results:
The results were like finding a shortcut to the top of a mountain.

• When doctors followed the map, about 46% of patients got significantly better.
• When doctors ignored the map, only 26% got better.

That's a huge difference! It means that for every 5 people who followed the biomarker report, one extra person got better who otherwise might have stayed depressed.

Breaking It Down by Treatment Type

The study looked at four different "keys" to unlock recovery:

1. ECT (Electroconvulsive Therapy): Think of this as a "hard reset" for the brain. When the report said "use this," 70% of people improved. When they ignored it, only 50% improved.
2. rTMS (Magnetic Stimulation): This uses magnets to gently wake up specific brain areas. Following the report boosted success from 13% to 30%.
3. Ketamine: A fast-acting treatment. Following the report boosted success from 10% to 31%.
4. SSRIs (Standard Antidepressants): The most common pills. Here, the difference was massive (100% vs 11%), but the group was very small, so we have to be careful about this specific number.

The Catch: Why Isn't Everyone Doing This Yet?

Even though the results look great, the authors are being very honest about the limitations:

• It's a Retrospective Study: They looked at past data, not a controlled experiment where they forced doctors to follow the map. It's possible that doctors who followed the map were just more careful or experienced doctors in the first place.
• Small Numbers: Some groups (like the pill group) were tiny. It's like flipping a coin 3 times and getting heads every time; it looks amazing, but you need to flip it 1,000 times to be sure.
• The "Black Box": The computer algorithm that makes the predictions is proprietary (owned by a company). We know it works, but we don't know the exact secret recipe inside the machine.

The Bottom Line

Imagine you are trying to tune a radio to find a clear station. Right now, most doctors are turning the dial randomly until they find a signal. This study suggests that if we use a biomarker report (like a digital tuner), we can find the clear station much faster.

The study shows that using brain and heart data to guide treatment decisions could significantly improve recovery rates for depression. It's not a magic cure-all yet, but it's a very promising step toward "Precision Psychiatry"—treating the person, not just the disease.

In short: If you have a broken leg, you don't want a cast that might fit. You want the one that is guaranteed to fit. This study suggests we are finally learning how to measure the leg before we put the cast on.

Technical Summary

1. Problem Statement

Major Depressive Disorder (MDD) and bipolar depression frequently exhibit limited response to first-line treatments, with remission rates for the first adequate antidepressant trial typically ranging between 30–40%. Current clinical practice often relies on a "trial-and-error" approach involving successive medication switches, dose titrations, and augmentation strategies based on clinician experience rather than objective data. While clinical and questionnaire-based prediction models exist, their discriminative performance at the individual level remains moderate. There is a critical need for objective, physiological biomarkers (specifically EEG and ECG) to guide personalized treatment selection and improve response rates in routine clinical practice.

2. Methodology

Study Design and Setting:

• Type: Retrospective, observational analysis conducted at the University Hospital of Psychiatry in Zürich, Switzerland (2022–2024).
• Population: 153 adult patients (≥18 years) with ICD-10 diagnosed MDD or bipolar depression.
• Interventions: Patients were treated with one of four guideline-recommended modalities:
  1. Electroconvulsive Therapy (ECT) ($n=53$)
  2. Repetitive Transcranial Magnetic Stimulation (rTMS) ($n=48$)
  3. (Es)ketamine ($n=36$)
  4. Selective Serotonin Reuptake Inhibitors (SSRIs) ($n=16$)
• Inclusion Criteria: Patients required baseline resting-state EEG/ECG recordings and complete pre- and post-treatment outcome ratings.

The Biomarker Tool (DeepPsy):

• System: The study utilized the DeepPsy system, a decision-support tool that generates individualized biomarker reports.
• Data Acquisition: Resting-state EEG (22-channel, 10-20 system) and simultaneous ECG were recorded (2–6 minutes, eyes-closed).
• Processing: Data underwent preprocessing (artifact rejection, ICA, filtering) and were analyzed using DeepPsy algorithms (v0.3.167MD).
• Biomarkers: The system integrates specific EEG features (vigilance regulation, alpha peak frequency, frontal alpha asymmetry) and ECG features (heart rate, heart rate variability, LF/HF ratio) derived from a systematic review of literature linking these markers to treatment outcomes.
• Output: The system generates modality-specific likelihoods of response (e.g., "increased likelihood," "normal," "reduced").

Classification of Concordance:

• Concordant Treatment: The clinical team chose a treatment modality (and specific protocol, e.g., 1Hz rTMS) that aligned with the DeepPsy report's recommendation of an "increased likelihood of response."
• Non-Concordant Treatment: The clinical team chose a modality where the report did not indicate an increased likelihood of response.
• Note: Non-concordance does not imply the report was ignored; it reflects clinical decisions where other factors (patient preference, contraindications, prior history) overrode the biomarker suggestion.

Outcome Measures:
Response was defined differently per modality based on standard clinical scales:

• ECT: Clinical Global Impression-Efficacy Index (CGI-E) score of 1 or 2.
• rTMS & (Es)ketamine: ≥50% reduction in Beck Depression Inventory II (BDI-II) score.
• SSRIs: ≥50% reduction in Hamilton Depression Rating Scale (HAMD) score.

Statistical Analysis:

• Comparison of response rates between concordant and non-concordant groups using Fisher's exact test (for small samples) and two-proportion z-tests.
• Calculation of Odds Ratios (OR), 95% Confidence Intervals (CI), and Number Needed to Treat (NNT).

3. Key Contributions

• Real-World Validation: Unlike many studies confined to controlled trials, this study evaluates the predictive utility of an integrated EEG/ECG biomarker report in a routine, naturalistic clinical setting.
• Multi-Modal Assessment: It is one of the few studies to simultaneously evaluate biomarker concordance across four distinct treatment classes (pharmacotherapy, neuromodulation, and ketamine-based therapy).
• Decision Support Framework: It demonstrates a practical workflow where a biomarker report acts as a "horizontal" decision aid (choosing which guideline-recommended option to use) rather than a "vertical" diagnostic tool.

4. Results

Overall Findings:
Across all modalities, patients receiving concordant treatments showed significantly higher response rates than those receiving non-concordant treatments.

• Total Cohort: 46% (42/92) response rate for concordant vs. 26% (14/54) for non-concordant.
• Absolute Difference: +20 percentage points.
• Relative Increase: ≈77%.
• Number Needed to Treat (NNT): ≈5.
• Statistical Significance: The combined analysis showed statistical significance (Fisher $p=0.01–0.02$; z-test $p=0.01–0.02$; OR 2.40, 95% CI 1.15–5.00).

Modality-Specific Breakdown:

Treatment	Concordant Response	Non-Concordant Response	Absolute Diff	NNT	Statistical Trend
ECT	70% (21/30)	50% (10/20)	+20%	5	Trend (p=0.08)
rTMS	30% (10/33)	13% (2/15)	+17%	6	Trend (p=0.10)
(Es)ketamine	31% (8/26)	10% (1/10)	+21%	5	Trend (p=0.10)
SSRIs	100% (3/3)	11% (1/9)	+89%	1	Significant (p<0.01)*

*Note: The SSRI result is based on very small sample sizes (n=12 total), resulting in wide confidence intervals and should be interpreted with caution.

5. Significance and Limitations

Significance:

• Clinical Impact: The findings suggest that integrating EEG/ECG biomarker reports into clinical decision-making can substantially improve treatment response rates in routine practice. An NNT of ~5 is comparable to or better than many established psychiatric interventions.
• Precision Psychiatry: The study supports the shift from "one-size-fits-all" algorithms to stratified medicine, where physiological markers guide the selection of the most likely effective intervention for a specific patient.
• Feasibility: It confirms that such biomarker tools can be successfully implemented and utilized in standard hospital settings without disrupting routine workflows.

Limitations:

• Retrospective & Observational: The study was not a randomized controlled trial (RCT). Treatment decisions were not experimentally controlled, introducing potential selection bias and confounding by indication (e.g., clinicians may have been more thorough in monitoring concordant cases).
• Sample Size: Subgroup sizes were modest, particularly for SSRIs, leading to wide confidence intervals and limiting the statistical power for individual modality analyses.
• Heterogeneity of Outcomes: Different scales (CGI, BDI-II, HAMD) were used for different modalities, introducing variability in how "response" was defined and measured.
• Proprietary Algorithm: The specific weighting and structure of the DeepPsy algorithm are proprietary, limiting independent verification of the model's internal mechanics.

Conclusion:
The study provides hypothesis-generating evidence that concordance with EEG/ECG-based biomarker reports is associated with higher antidepressant treatment response rates across multiple modalities. The authors conclude that these findings warrant confirmation in prospective, randomized trials to establish causal efficacy and refine the clinical utility of such decision-support tools.