How does brain imaging predict who responds to TMS in depression and PTSD?
Measuring the brain’s actual response to a TMS pulse predicts depression improvement better than resting-state connectivity does, and fMRI-guided coil targeting produced better early symptom response than scalp-based targeting in patients with PTSD and depression. This work comes from a symposium presentation in Brain Stimulation covering two human samples: 36 depressed patients studied with interleaved TMS/fMRI, and 51 patients with PTSD and major depression treated with either a scalp-based or an fMRI-guided target. It is not a single large trial, and the second sample is reported as a preprint.
The unifying idea across both studies is subgenual anterior cingulate cortex, or sgACC. It sits too deep for a surface coil to reach directly, so the investigators used resting-state fMRI seeded on the sgACC to find cortical clusters with positive connectivity to it, then stimulated those surface sites. Interleaved single-pulse TMS/fMRI let them see whether a pulse delivered at the scalp actually evoked a response down in the sgACC.
What the data show:
- Circuit engagement: TMS-evoked sgACC responses were demonstrated in 36 depressed patients stimulated at functionally connected cortical targets
- Prediction: Pre-intervention sgACC-evoked response magnitude predicted depression improvement (adjusted for baseline depression; β=45.60, p=.008)
- Change with treatment: Change in sgACC-evoked response after a 3-day iTBS course tracked improvement (baseline depression adjusted; β=-34.66, p=.007)
- Head to head: Only TMS/fMRI remained a significant predictor when baseline resting-state connectivity between stimulation site and sgACC was entered into the same model
- Targeting: fMRI-guided targeting outperformed scalp-based targeting for depression symptoms [F(1,43.92)=5.933, p=.019] and PTSD hyperarousal [F(1,40.78)=5.076, p=.030] in 51 PTSD/MDD patients
- Brain state: The most clinically effective condition paired fMRI-guided targeting with watching a peaceful nature video during stimulation
- Trauma history: More trauma history was associated with less negative affect improvement (β=0.07, p=.027)
Taken together, the presenters argue that brain-based measures, online task manipulations during treatment, and trauma history can all be used to inform symptom outcomes from rTMS.
Dr. Kumar’s Take
The part of this I find genuinely useful is the head-to-head comparison inside a single model. Resting-state connectivity has been the standard way to pick a TMS target, but here it dropped out as a predictor once the TMS-evoked sgACC response was accounted for. Knowing that a circuit is correlated at rest is not the same as knowing that your pulse reaches it. Measuring the evoked response answers the question directly, and in these patients that direct measure carried the predictive signal.
I also take the nature video finding seriously, even though it sounds trivial. What the brain is doing at the moment of stimulation is a variable most clinics ignore entirely. Patients sit in a chair, sometimes on their phones, sometimes anxious about the coil. If pairing stimulation with a defined brain state improves response, that is a free intervention no one is currently optimizing.
How This Works (Biological Rationale)
The sgACC is the anatomical anchor for this entire line of work in depression, and it cannot be stimulated directly from the scalp. The workaround used here is a pathway approach: seed the sgACC on a resting-state scan, find cortical clusters showing positive connectivity to it, and stimulate one of those surface clusters so the effect propagates inward along the connected pathway.
Interleaved single-pulse TMS/fMRI turns that assumption into a measurement. A single pulse is delivered inside the scanner and the resulting sgACC signal is recorded, so the investigators can see whether the intended circuit was engaged in that specific patient rather than inferring it from anatomy or from a correlation map.
That distinction is what the results support. A patient whose sgACC responded more strongly to a pulse before treatment improved more, and the degree to which that evoked response shifted across a 3-day intermittent theta burst course also tracked improvement. Resting connectivity between the same two points did not add predictive value once the evoked response was in the model.
Results in Real Numbers
In the first sample of 36 depressed patients, pre-intervention TMS-evoked response magnitude in the sgACC predicted depression improvement after adjusting for baseline depression, β=45.60, p=.008. The change in sgACC-evoked response following the 3-day iTBS intervention was also associated with depression improvement, β=-34.66, p=.007.
When baseline resting-state connectivity between the stimulation site and the sgACC was entered into the same prediction model alongside the TMS/fMRI evoked response, only the TMS/fMRI measure remained significant while controlling for the other imaging measure.
In the second sample of 51 patients with PTSD and major depression, patients treated with the fMRI-guided target rather than a scalp-based target showed superior clinical response immediately after rTMS for depression symptoms, F(1,43.92)=5.933, p=.019, and for PTSD hyperarousal, F(1,40.78)=5.076, p=.030. The most clinically effective condition combined fMRI-guided targeting with a brain state manipulation, watching a peaceful nature video.
Trauma history did not predict positive affect improvement, β=0.04, p=.2. More trauma history was associated with significantly less negative affect improvement, β=0.07, p=.027.
What the Research Shows
The first study enrolled 36 depressed patients (ClinicalTrials.gov ID NCT04014959) and used interleaved single-pulse TMS/fMRI to demonstrate circuit engagement, measuring TMS-evoked sgACC responses to stimulation of functionally connected cortical targets. Targets were identified from resting-state fMRI scans seeding the sgACC and selecting positive connectivity clusters. TMS/fMRI was performed before and after a 3-day intermittent theta burst rTMS intervention aimed at the sgACC pathway. Additional circuit and symptom specificity analyses are reported in the full manuscript.
The second study enrolled 51 patients with PTSD and major depression, treated with either a scalp-based target or an fMRI-guided target defined by positive functional connectivity with the sgACC, and compared clinical response immediately after treatment. A separate analysis in that line of work examined whether trauma history predicted affect outcomes.
Who Benefits Most
Patients whose stimulation site demonstrably engages the sgACC are the ones this data supports. In the first sample, a stronger pre-treatment evoked sgACC response predicted greater depression improvement, which makes the evoked response a candidate selection measure rather than a purely mechanistic one.
For patients carrying both PTSD and depression, the fMRI-guided target outperformed scalp-based targeting on depression symptoms and on hyperarousal specifically, so those are the symptom domains where individualized targeting showed its advantage here.
Trauma history matters for expectations. More trauma history was associated with less improvement in negative affect, so patients with heavier trauma burden may need that dimension tracked separately rather than assumed to move with everything else.
Safety, Limits, and Caveats
This is a symposium abstract, and it reports mechanism and prediction rather than safety outcomes. The samples are small, 36 and 51 patients, and the second study is available as a preprint while the trauma history analysis is listed as in preparation. Findings at this stage should be treated as directional.
The clinical response advantage in the second study was measured immediately after the rTMS course, so durability is not established by these results. The first study used a 3-day intervention, which is far shorter than a standard clinical rTMS course.
The bigger practical limit is access. Everything predictive here required an MRI scanner, and the strongest predictor required TMS delivered inside the scanner. That is a research capability, not a community clinic capability.
Practical Takeaways
- Treat fMRI-guided targeting as the better-supported option in these samples compared with scalp-based targeting
- Recognize that resting-state connectivity lost significance as a predictor when the TMS-evoked sgACC response was in the same model
- Ask whether the circuit is actually engaged, not just whether the coil is in the standard position
- Consider what the patient is doing during stimulation, since the most effective condition here paired fMRI-guided targeting with a peaceful nature video
- Ask about trauma history, which was associated with less negative affect improvement
- Track depression symptoms and hyperarousal separately in patients carrying both diagnoses
Related Studies and Research
Episode 31: Depression Explained, The Biology Behind the Darkness
Episode 32: Depression Recovery Roadmap: A Step-by-Step, Evidence-Based Plan
Default Mode Network Mechanisms of Transcranial Magnetic Stimulation in Depression
Psychiatric Applications of Repetitive Transcranial Magnetic Stimulation
Accelerated TMS - moving quickly into the future of depression treatment
FAQs
Why use fMRI to pick the TMS target?
The sgACC cannot be stimulated directly from the scalp, so targets were chosen from resting-state scans seeded on the sgACC by selecting clusters with positive connectivity to it. In the 51-patient study, that fMRI-guided target produced superior clinical response immediately after treatment compared with a scalp-based target.
What is TMS/fMRI and why does it matter?
It means delivering single TMS pulses while the patient is in the scanner so the evoked response in a deep region can be measured directly. In these 36 depressed patients, the magnitude of that evoked sgACC response predicted depression improvement, and it remained the significant predictor when tested against resting-state connectivity in the same model.
Does what the patient does during stimulation matter?
In this work, the most clinically effective rTMS condition paired fMRI-guided targeting with a brain state manipulation, specifically watching a peaceful nature video during treatment.
Does trauma history change what to expect from rTMS?
More trauma history was associated with significantly less improvement in negative affect, β=0.07, p=.027. Trauma history did not predict improvement in positive affect, β=0.04, p=.2.
Bottom Line
Directly measuring whether a TMS pulse reaches the sgACC predicted depression improvement in 36 depressed patients, and it outperformed resting-state connectivity when both were tested together. In 51 patients with PTSD and depression, choosing the target with fMRI beat scalp-based targeting for depression symptoms and hyperarousal immediately after treatment, and the strongest condition added a defined brain state during stimulation. These are small samples presented in abstract form, but they point at a version of rTMS where the target is verified in the individual patient rather than assumed.
