Data-Driven rTMS: Moving Beyond Standard Protocols

Repetitive transcranial magnetic stimulation (rTMS) is one of the most promising treatments for treatment-resistant depression. It is FDA-cleared, non-invasive, and backed by a substantial evidence base. It is also, in current practice, remarkably standardized — and that standardization is leaving therapeutic potential on the table.

This post explains the current state of rTMS protocols, why personalization matters, and how MindCODE's clinical intelligence tools support clinicians in making more informed protocol decisions.

The Current Standard

The dominant rTMS protocol for major depressive disorder has not changed substantially in over a decade:

Frequency: 10 Hz (high-frequency excitatory stimulation)
Target: Left dorsolateral prefrontal cortex (L-DLPFC)
Intensity: 120% of resting motor threshold (RMT)
Pulses per session: 3,000 (75 trains of 40 pulses, 4-second on / 26-second off)
Sessions: 30 sessions over 6 weeks (5 sessions/week)
Total pulses: 90,000

This protocol was established through pivotal trials (O'Reardon et al., 2007) and has a response rate of approximately 50-55% and a remission rate of approximately 30-35% in treatment-resistant populations. These numbers are meaningful — but they also mean that roughly half of patients who complete the full 6-week course do not achieve a clinically significant response.

The Stanford Neuromodulation Therapy (SNT) Variant

More recently, the Stanford accelerated protocol (now marketed as SNT/SAINT) demonstrated that compressed schedules can work:

Frequency: Intermittent theta burst stimulation (iTBS) at 50 Hz bursts
Target: L-DLPFC, guided by functional connectivity MRI targeting
Intensity: 90% of RMT (lower than standard due to iTBS efficiency)
Sessions: 10 sessions per day for 5 days (50 sessions total)
Intersession interval: 50 minutes between sessions
Total pulses: 90,000 (same total dose, different temporal pattern)

The SNT trial reported a remission rate of 79% in a small sample (n=29), though larger real-world implementations have shown more modest results (remission rates of 40-50%). The key insight is that the parameter space is larger than the standard protocol explores.

Why Personalization Matters

The 30-60% range in response rates across studies is not random noise. Multiple factors predict differential response:

Patient-Level Predictors

Baseline severity: Patients with moderate depression (PHQ-9 15-19) may respond differently than those with severe depression (PHQ-9 > 20)
Treatment resistance level: Number of failed adequate medication trials (the Maudsley Staging Method quantifies this from Stage 1 to Stage 5)
Age: Some evidence suggests younger patients respond better to high-frequency protocols, while older patients may benefit from different parameter combinations
Neuroanatomical variation: The "5.5 cm rule" for targeting L-DLPFC (5.5 cm anterior to the motor cortex along the scalp surface) misses the actual DLPFC in up to 30% of patients due to individual anatomical variation

Neuroimaging Predictors

Functional connectivity: The strength of anticorrelation between the DLPFC target and the subgenual anterior cingulate cortex (sgACC) predicts response. Patients with stronger baseline DLPFC-sgACC anticorrelation tend to respond better (Fox et al., 2012; Cash et al., 2021).
DMN hyperconnectivity: Patients with elevated default mode network connectivity at baseline show greater connectivity reduction after successful rTMS, suggesting this may be a treatable target.
Cortical thickness: Prefrontal cortical thickness may influence the effective depth of stimulation, potentially affecting the appropriate intensity setting.

Biomarker Predictors

Inflammatory markers: Elevated baseline CRP (>3 mg/L) and IL-6 have been associated with differential response to rTMS versus medication in some studies.
BDNF levels: Brain-derived neurotrophic factor levels may predict neuroplasticity-dependent treatment responses.
Pharmacogenomics: While rTMS itself does not involve drug metabolism, the concurrent medications a patient takes (and their CYP450 metabolism profile) affect the baseline neurochemical environment.

MindCODE's Approach: Decision Support, Not Autopilot

MindCODE does not generate rTMS protocols autonomously. The system provides structured decision support that helps clinicians evaluate options against the best available evidence and the specific patient's data. The clinician sets every parameter and makes every treatment decision.

The Protocol Planning Workflow

Here is how the system works in practice, step by step:

Step 1: Baseline Assessment Integration

The clinician initiates a protocol planning session for a specific patient. MindCODE pulls from the patient's Canonical Longitudinal Record:

Current and historical PHQ-9/MADRS scores with trajectory analysis
Treatment history (medications, psychotherapy, prior neuromodulation)
Available neuroimaging (structural MRI, fMRI connectivity if available)
Wearable data summary (sleep patterns, activity levels, circadian rhythm stability)
Relevant biomarkers (inflammatory markers, BDNF if available)
Pharmacogenomic profile (CYP450 panel if available)

This data is presented as the Virtual Human's descriptive twin output — a synthesized clinical picture, not raw data dumps.

Step 2: Cohort Comparison

The system identifies a matched comparison cohort from MindCODE's aggregated (de-identified) treatment outcome database. Matching criteria include:

Baseline depression severity (within 3 points on PHQ-9)
Number of failed medication trials (exact match or +/- 1)
Age range (within 10 years)
Sex
Available neuroimaging similarity (if fMRI data exists for both the patient and cohort members)

The clinician sees the matched cohort characteristics and can adjust matching criteria. The system reports:

Matched cohort: n=143
  Baseline PHQ-9: mean 18.4 (SD 3.2) — patient: 19
  Failed medication trials: mean 2.3 — patient: 2
  Age: mean 41.2 (SD 8.7) — patient: 38
  Sex: 62% female — patient: female
  With fMRI data: 67 patients (47%)

Cohort treatment outcomes (all protocol variants):
  Overall response rate (>50% PHQ-9 reduction): 52%
  Overall remission rate (PHQ-9 < 10): 34%
  Mean sessions to response: 18.4 (SD 7.2)

Step 3: Protocol Parameter Exploration

The system presents outcomes stratified by protocol parameters within the matched cohort. This is where the data becomes actionable.

By frequency protocol:

| Protocol | n | Response Rate | Remission Rate | Mean Sessions to Response | |----------|---|---------------|----------------|--------------------------| | 10 Hz standard | 82 | 49% | 31% | 20.1 | | iTBS | 38 | 55% | 37% | 15.3 | | 1 Hz R-DLPFC (low-frequency) | 23 | 48% | 35% | 22.7 |

By targeting method:

| Method | n | Response Rate | Remission Rate | |--------|---|---------------|----------------| | 5.5 cm rule | 71 | 46% | 28% | | Beam F3 method | 42 | 54% | 36% | | fMRI-guided (DLPFC-sgACC anticorrelation) | 30 | 63% | 43% |

By intensity:

| Intensity (% RMT) | n | Response Rate | Notes | |--------------------|---|---------------|-------| | 110% | 28 | 43% | Lower side effect burden | | 120% | 89 | 52% | Standard | | 130% | 26 | 54% | Higher discomfort reported |

Every cell in these tables links to the underlying patient-level data (de-identified), the confidence intervals, and the statistical test results. The system flags when sample sizes are too small for reliable conclusions (generally n < 20 for a subgroup).

Step 4: Protocol Suggestion

Based on the cohort analysis and the patient's specific data, the system generates a structured protocol suggestion. Crucially, this is framed as an option to consider, not a prescription:

Protocol suggestion for Patient M (review required)
─────────────────────────────────────────────────

Suggested parameters:
  Frequency:       iTBS (intermittent theta burst)
  Target:          L-DLPFC via fMRI-guided targeting
                   (patient has baseline fMRI available)
  Intensity:       120% RMT
  Pulses/session:  600 (standard iTBS)
  Sessions:        30 over 6 weeks
  Schedule:        5 sessions/week

Rationale:
  - iTBS showed higher response rate in matched cohort (55% vs 49%)
  - fMRI-guided targeting showed substantially higher remission rate (43% vs 28%)
  - Patient's baseline fMRI shows moderate DLPFC-sgACC anticorrelation
    (z = -0.34), which is associated with response in the fMRI-guided subgroup
  - Standard intensity (120% RMT) recommended; insufficient evidence
    in matched cohort to justify deviation

Confidence: MODERATE
  - Cohort size for iTBS + fMRI-guided combination: n=18
    (below preferred threshold of n=30)
  - iTBS total pulse count differs from standard 10Hz protocol;
    dosing equivalence is an active area of research

Alternative to consider:
  - 10 Hz standard + fMRI-guided targeting (n=12 in cohort, response 58%)
    if clinician prefers more established frequency protocol

⚠ This is a decision support output, not a treatment recommendation.
  Clinical judgment must determine the final protocol.

Step 5: Session Tracking

Once the clinician sets the protocol and treatment begins, MindCODE tracks session-by-session data:

PHQ-9 or MADRS scores at defined intervals (typically every 5 sessions)
Wearable data trends (sleep, activity, circadian rhythm)
Treatment tolerability notes
Stimulation parameters actually delivered (intensity adjustments, missed sessions)

The system compares the patient's trajectory against the matched cohort's trajectory at each time point:

Session 15 checkpoint (Week 3):
  PHQ-9: 19 → 15 (21% reduction)
  Cohort comparison at session 15:
    - Patients who ultimately responded: mean 28% reduction at this point
    - Patients who ultimately did not respond: mean 11% reduction at this point
  Assessment: Patient's trajectory is between responder and non-responder
    cohort means. Insufficient data to predict final outcome at this point.
    Recommend continuing current protocol to session 20 checkpoint.
  Sleep efficiency: 64% → 69% (improving trend consistent with
    cohort responders, who showed mean 7% sleep improvement by session 15)

Step 6: Outcome Analysis

After treatment completion, the system captures the full outcome and adds this patient's data (with consent) to the outcome database, strengthening future cohort analyses:

Final PHQ-9/MADRS scores and response/remission classification
Time to response if achieved
Durability assessment at 1, 3, and 6-month follow-ups
Side effects and tolerability summary
Protocol parameters as actually delivered

Important Boundaries

We want to be unambiguous about what this system does and does not do:

What the system does:

Presents relevant outcome data from comparable patients
Structures the parameter decision space with evidence
Tracks treatment progress against empirical benchmarks
Logs all data, decisions, and outcomes for clinical quality improvement

What the system does NOT do:

Control the rTMS device
Change treatment parameters without clinician action
Override a clinician's protocol decision
Guarantee outcomes
Replace the clinical assessment that determines whether rTMS is appropriate in the first place

The Role of the Clinician at Every Step

Indication assessment: The clinician determines whether rTMS is appropriate for this patient based on their full clinical picture, including factors the system may not have access to (patient preferences, logistical constraints, comorbidities).
Parameter selection: The clinician reviews the system's cohort analysis and protocol suggestion, applies their clinical expertise, and sets the final protocol parameters.
Session delivery: A trained rTMS technician or the clinician operates the device. The system does not interface with the device.
Progress assessment: The clinician reviews the tracking data, conducts clinical assessments, and decides whether to continue, modify, or discontinue treatment.
Outcome evaluation: The clinician evaluates the overall treatment outcome and determines next steps.

Where This Is Going

The rTMS parameter space is larger than current clinical practice explores. Beyond the parameters discussed above, emerging research is investigating:

Bilateral protocols: Simultaneous or sequential L-DLPFC excitatory and R-DLPFC inhibitory stimulation
Connectivity-guided multi-target protocols: Stimulating multiple nodes in a depression-relevant circuit rather than a single target
Adaptive protocols: Adjusting parameters mid-course based on early response indicators (this is where session tracking data becomes directly actionable)
Maintenance protocols: Optimizing post-acute session frequency to prevent relapse

As the evidence base grows and MindCODE's outcome database expands, the precision of cohort matching and outcome prediction will improve. The goal is not to automate the clinician away — it is to make every clinician as informed as the world's leading rTMS researcher when they set protocol parameters for their next patient.