If you have typed that phrase into a search box, nearly everything that comes back is written by a clinic that sells TMS. This page is not. It sets out what people are describing when they say it, which parts of it the evidence supports, which parts it does not, and what you can do if a course left you worse off than it found you.
The short answer
Most people who say TMS ruined their life are not describing an injury from the magnetic pulse. They are describing a failed course — weeks of daily appointments, thousands of dollars, a treatment introduced as the last thing left to try, and no improvement at the end of it. Serious physical harm from TMS is rare and reasonably well measured. Failure is not rare, and almost nobody quotes the real number.
The complaint
Those come from public forums, and forums are evidence of what people report — not evidence of what a treatment does. The rest of this page sets each one against what has actually been measured, in both directions. Some of these complaints are better supported than clinics admit. Some are not supported at all. Two of them have never been studied by anyone.
How often it fails
It gets narrower over the following year. Among people who do respond, about 66% are still responding at 3 months, 53% at 6 months and 46% at 12 months. In a 12-month follow-up of 257 people treated in ordinary US practices, 93 of them (36%) needed TMS reintroduced during the year.
Of 100 people who start a course
Respond to a full course (randomized evidence)
29
Reach remission
19
Still responding 12 months later
14
Would have responded to a sham coil
10
One fair caveat, in TMS's favor. Those 29 trials averaged about 13 sessions, roughly a third of a modern 36-session course, and longer courses do appear to work better: in a 7,215-patient registry analysis, courses shorter than 30 sessions had worse outcomes, and benefit was still accruing past session 36 with no sign of a plateau. The true figure for a full modern course is probably better than 29%. It is not 80%.
Why the two sets disagree
In one study of 307 patients at 42 US practices, the same people on the same day scored 37.1% remission on the clinician's one-item severity rating and 26.5% on the patient-completed symptom questionnaire. The scale alone moves the number by a third.
The high registry figures come from a "completer" sample defined as responders, plus anyone who received at least 20 sessions. People who stopped at session 12 because nothing was happening are excluded by construction.
In a registry the person scoring your severity is the clinician at the clinic being paid for the course. That is not an accusation; it is why blinding exists.
The large real-world TMS datasets are funded by device manufacturers and carry company employees or paid consultants as authors. The industry-independent trials produce the least flattering results in the literature.
The dip
What patients describe does not look like one thing either. On public forums people date their dip to session 2, session 5, session 8, session 15, session 18 and session 28. Something that can begin anywhere across a nine-week course is not a discrete clinical event. It is a name people give to a bad stretch.
5% of 228
In a trial that rated 228 people with treatment-resistant depression session by session, 12 of them followed a worsening trajectory — and none of that group met response criteria at 8, 16 or 26 weeks. Two other large trajectory studies, of 388 and 324 patients, found no worsening group at all.
24% of 228
In the same trial, 24% were delayed improvers who showed no change until around session 10 and then improved steadily to the end. A registry analysis of 7,215 patients found that the absence of early improvement is too poor a predictor of final non-response to base a decision to stop on.
So both of the things people get told are wrong. "Push through the dip, everyone gets one" is not evidence-based. Neither is "if it has not worked by week two it never will." What should happen instead is that the worsening gets measured — a scored scale at the same interval every week, not an impression — and reviewed with the clinician who prescribed the course. Not with a technician, and not with a forum.
Side effects, measured
| Effect | Active TMS | Sham |
|---|---|---|
| Headache | 58.2% | 55.1% |
| Scalp or application-site pain | 35.8% | 3.8% |
| Muscle twitching | 20.6% | 3.2% |
| Application-site discomfort | 10.9% | 1.3% |
| Toothache | 7.3% | 0.6% |
| Facial pain | 6.7% | 3.2% |
| Eye pain | 6.1% | 1.9% |
| Stopped because of a side effect | 4.5% | 3.4% |
| Stopped for any reason | 7.7% | 8.2% |
Read the first row and the last two rows together, because they are the most useful facts in the table.
Headache is not caused by the magnet. It happened to 58.2% of people on active TMS and 55.1% of people getting a sham coil. Headache is what depression, daily early appointments and six weeks in a chair do to people. Scalp pain is the one that is genuinely the device: 35.8% against 3.8%. And almost nobody stops because of side effects — the gap over sham is about one percentage point, and all-cause dropout was slightly higher in the sham group. In a separate pooled safety analysis of 325 patients across more than 10,000 sessions there were no deaths, no seizures, and no measurable change in hearing thresholds or cognitive test scores. Pain is also front-loaded: scalp pain and headache fall off sharply after the first week of treatment.
This is the risk people mean when they ask whether TMS is dangerous, and it is one of the better-quantified numbers in psychiatry. Two independent datasets land in the same place.
1 in 32,000
Sessions, across 586,656 sessions
1 in 1,400
Patients, across 25,526 patients
1 in 60,000+
Sessions, no risk factors, within guidelines
Three things about those seizures matter more than the rate. They happen during stimulation, not later — in one survey 15 of 24 seizures occurred on a person's first ever exposure and 18 of 24 within the first three sessions. According to the 2021 international safety guidelines, a TMS-associated seizure has never resulted in permanent damage in current practice, there is no evidence of kindling in humans, and no reported case has had onset hours or days after a session ended. And the coil type matters, which almost no patient-facing page mentions: in that 586,656-session survey the H-coil used for deep TMS ran at 5.56 seizures per 1,000 patients against 0.14 per 1,000 for the three most-used figure-8 devices combined. If you are being offered deep TMS, that is a fair question to ask. The only absolute contraindication in the international guidelines is metal or an active implant close to the coil. A history of seizures is a caution in the device labeling, not a bar.
As of 11 August 2026 the FDA's public device-report database held 338 reports for TMS depression devices: 324 injuries, 10 malfunctions and 2 deaths. The most frequently coded problems are headache (89), anxiety (73), depression (60), seizure (49), emotional changes (49), tinnitus (49), pain (44), suicidal ideation (37), cognitive changes (36), sleep problems (36) and memory impairment (33). About 42% were filed voluntarily rather than by a manufacturer, so a substantial share are patients and families.
The other direction
Four separate studies between 2003 and 2018, including a sham-controlled trial in people aged 60 to 85 — the group most vulnerable to cognitive harm — found no worsening on tests of attention, working memory, recall or executive function. The device's own trial program, 325 patients across more than 10,000 sessions, recorded no change in cognitive function. Two caveats: there is no clean head-to-head comparison of TMS and ECT on cognition, and a 2025 meta-analysis of 15 sham-controlled trials found TMS does not reliably improve cognition either.
People report both, during courses and after them. There is no published incidence figure of any kind — not a trial, not a cohort, not a case series. That is an absence of study, not evidence of absence, and it is a reasonable thing to be angry about.
Same position. There is no published evidence that TMS causes it. Most of the medical literature on TMS and blunting is about using TMS to treat blunting caused by antidepressants.
Across trials that looked for it, 0.84% on active TMS against 0.73% on sham. In bipolar depression specifically, across 1,709 patients, the odds ratio against sham was 1.3 with a confidence interval spanning 1. Worth screening for undiagnosed bipolar disorder before starting, and worth being on a mood stabilizer if you have it. Not worth fearing.
The device label does carry a warning about worsening depression and emergent suicidality, and that warning is class labeling applied to depression treatment generally. In the pivotal trial, hospitalizations for suicidal ideation were spread across the sham group, the active group and the pre-randomization period, and the one completed suicide attempt was in the sham arm. The evidence on the other side is weaker than clinics suggest too: pooled across randomized trials only, the reduction in suicidal ideation is not statistically significant.
None of this makes anyone's experience untrue. What it means is that for several of the harms people describe, nobody has done the study. The honest position is to say so, not to reassure.
The money
| Service | Average billed | Medicare allowed |
|---|---|---|
| Initial session with motor threshold (CPT 90867) | $881.79 | $278.93 |
| Each subsequent session (CPT 90868) | $603.06 | $180.94 |
~$22,000
Billed for a 36-session course
~$6,600
Allowed by Medicare for the same course
3.3x
What providers submit against what Medicare pays
Course totals are our arithmetic on the CMS figures, not a published number. If you are paying cash or going out of network, the figure you are quoted will sit closer to the billed column than the allowed one. Ask for both.
Insurers authorize a course, and patients report that dropping out partway can complicate getting a later course approved. That creates a financial reason to finish that has nothing to do with whether the treatment is working. It should not be the reason you finish. If you are being encouraged to continue, ask to see your scored assessments.
In March 2023 a psychiatrist and three affiliated entities paid $3 million to settle federal False Claims Act allegations that included pressuring patients into unnecessary treatments, falsifying records, and billing Medicare for TMS sessions that were never performed. One settlement is not an industry. It is a documented instance of the thing patients describe.
The evidence gap
There is no study of that. Nobody has measured what a failed TMS course does to a person's hope.
The nearest evidence comes from adjacent places. Interviews with people who had deep brain stimulation for depression found that the decision was often made from a position of desperation strong enough to compromise informed consent, and that the result disappointed against the expectation more than it did against the starting point. Interviews with participants in a large TMS trial found that hope was the main reason people signed up, and that the most useful thing clinicians did was set a realistic expectation at the start.
If that describes you, three things are true and none of them are consolation prizes.
The failure is common, and the number of people it happens to is further up this page.
"Treatment-resistant" is a description of which medications you have already tried. It is not a verdict on you.
TMS is not the last option, whatever you were told when you signed up. The next two sections are the routes that are actually open.
If you are thinking about harming yourself, use one of these now.
988 Suicide & Crisis Lifeline
Call or text 988, or chat at chat.988lifeline.org. Free, confidential, 24 hours a day. Press 2 for Spanish. Press 1 if you are a veteran or service member. Deaf and hard-of-hearing callers can reach 988 by videophone, or dial 711 then 988 through a relay service.
Crisis Text Line
Text HOME to 741741. Free, confidential, 24 hours a day, in English and Spanish.
Recourse
Under HIPAA you can ask for the whole file — consent forms, motor threshold determinations, the parameters used at each session, the attendance log, every symptom score they recorded, and the prior-authorization paperwork they sent your insurer. The clinic has 30 days, plus one 30-day extension it has to tell you about in writing. Fees have to be reasonable and cost-based, and there is no per-page fee for records held electronically. They cannot refuse because you owe them money. If they refuse anyway, file with the HHS Office for Civil Rights within 180 days of when you knew about it.
Anyone can file. You do not need the clinic's cooperation or a doctor's signature. Use the consumer form, Form 3500B, through the MedWatch portal. Have ready the device brand and manufacturer, the model or serial or UDI number, your treatment dates, and what happened. Once you start an online report you have three days to finish it, and you cannot edit it after it is submitted. Be clear about what this is: it feeds the public safety database. It does not discipline anyone, it does not get you a refund, and the FDA makes no commitment to respond to you.
Licensing is state by state and profession by profession. For an MD or DO, start from the Federation of State Medical Boards directory, which links to each state board's own complaint portal. If a nurse practitioner treated you, the medical board has no jurisdiction — use the state board of nursing through the NCSBN directory. A board complaint is about the licensee's conduct. It is not a route to compensation.
You have 180 days from a denial to file an internal appeal, and four months from the final internal denial to request an independent external review. The external review is free, it takes 45 days at standard speed or 72 hours expedited, and the insurer is legally required to accept the result. Which regulator you escalate to depends on your plan, and people lose months getting this wrong: a fully insured plan goes to your state department of insurance, a self-funded employer plan goes to the US Department of Labor. Ask HR which one you have. The plan also has to give you, free, the medical-necessity criteria it applied to your claim.
If you were charged for sessions you did not attend, request an itemized bill with CPT codes and dates of service and set it against the attendance log from step 1. If you paid cash, were given a good faith estimate at least three business days before starting, and the final bill came in at least $400 over it, you can use the federal patient-provider dispute resolution process: $25, and the provider cannot send you to collections while it is open. One thing to know first — the No Surprises Act's balance-billing protections generally do not cover a freestanding outpatient TMS clinic, so starting there will not work. If Medicare or Medicaid was billed for sessions that were not delivered, that is a fraud report to the HHS Office of Inspector General.
What is actually left
Including augmentation and combinations rather than another single switch. Worth a psychiatrist who specializes in treatment-resistant depression rather than the one who referred you for TMS.
Approved by the FDA for treatment-resistant depression in adults. Given only in a certified healthcare setting under a restricted program, with at least two hours of monitoring after every dose, and never dispensed for home use. It carries a boxed warning for sedation, dissociation, abuse potential and suicidal thoughts.
Still more effective than TMS for severe depression, per the National Institute of Mental Health, with a different and more serious cognitive trade-off. It is the option TMS is most often positioned as an alternative to, and it is worth reopening rather than ruling out on the strength of how it was described to you.
On its own or alongside any of the above.
There are actively recruiting treatment-resistant depression studies listed on ClinicalTrials.gov, and trial participation is one of the few routes to a treatment that is not yet commercially available.
To find a different clinician: SAMHSA's treatment locator at findtreatment.gov, or their free 24-hour helpline on 1-800-662-4357.
FAQ
It can, and for a small number of people it does not reverse. In a trial that rated 228 patients session by session, 12 of them (5%) followed a worsening trajectory, and none of that group met response criteria at 8, 16 or 26 weeks. That is different from the mid-course "dip" people describe online, which has never been documented in a study. If your scores are going up, ask the prescribing clinician to review them — a scored scale, not an impression.
It is real vocabulary. There is no trial, cohort or case series describing a transient mid-course worsening in TMS, no published incidence figure, and no established timing. The only peer-reviewed mention of the phrase calls it something anecdotally described by patients. On forums people date their dip anywhere from session 2 to session 28. Getting worse during a course does happen, and a lack of early improvement is a poor predictor of final failure, so neither "push through it" nor "stop now" is evidence-based advice.
Across 29 randomized sham-controlled trials covering 1,371 people, 29.3% responded and 18.6% reached remission, so roughly 7 in 10 did not respond and roughly 8 in 10 did not remit. Real-world registry figures are much higher, up to 83% response, because they use clinician-rated scales, exclude people who stopped early, and have no sham arm. Both sets of numbers are real; they are measuring different things.
No case of permanent damage from a TMS-induced seizure has been reported in current practice, according to the 2021 international safety guidelines, which also find no evidence of kindling and no delayed-onset seizures. Seizure itself is rare: about 1 in 32,000 sessions and 1 in 1,400 patients across 586,656 sessions, and fewer than 1 in 60,000 sessions when TMS is given within guidelines to someone with no risk factors. The H-coil used for deep TMS carries a higher rate than standard figure-8 coils.
No study has found it. Four studies between 2003 and 2018, including a sham-controlled trial in people aged 60 to 85, found no worsening in attention, working memory, recall or executive function, and the device's own trial program found cognitive function unchanged across more than 10,000 sessions. Patients do report brain fog and word-finding trouble, and no published incidence figure for either exists. Nobody has studied it.
Start by requesting your complete records under HIPAA; the clinic has 30 days and cannot refuse because you owe money. From there: report a device-related adverse event to the FDA through MedWatch, file a complaint about the clinician with your state medical board, appeal an insurance denial (180 days for the internal appeal, four months for a free and binding external review), and dispute the bill by setting the itemized charges against the attendance log.
Sources
Read next
Pillar
What is TMS therapy?
Spoke
The TMS dip: why you might feel worse before you feel better
Spoke
Does TMS therapy work?
Listings show each clinic's phone number, website and hours. Call ahead and ask which coil they use, how they measure your response, and what they bill an insurer for a session.
Browse all locations →