Search this question and you get a number: 50 to 70 percent of people respond, 30 to 40 percent go into remission. That number is real. It also comes from clinics counting their own patients, with nobody blinded and no comparison group. The blinded trials the device was cleared on found roughly half that. Both sets of figures are honestly reported. They answer different questions.
The short answer
Yes, for a meaningful minority, and the size of that minority depends on which question you are asking. In blinded trials against a sham coil, about 3 in 10 people respond and fewer than 2 in 10 reach remission — roughly double the sham rate, with a number needed to treat somewhere between 6 and 12. In ordinary practice, where nobody is blinded and there is no comparison group, about half respond. The best current summary, from a 2025 clinical guide in the American Journal of Psychiatry, is that someone with major depression can expect roughly a 50 percent chance of a clinically meaningful response — better if you have failed fewer medications, worse if you have failed more. What nobody can tell you is whether you will be in that half. There is no validated way to predict it.
The two answers
There are two sets of numbers, and they differ by about half
Most pages about TMS pick one set and print it without saying which. Google's own summary at the top of this search currently gives 50 to 70 percent response and 30 to 40 percent remission, sourced to a mix of academic medical centers and a Reddit thread. Those figures come from open-label clinical practice. They are not wrong. They are also not the numbers the device was cleared on, and the difference is not a detail.
Measure
Blinded trials, against a sham coil
Ordinary clinical practice
What it measures
How much of the improvement the coil itself is responsible for
How likely a person is to feel better by the end of a course, for any reason
Who is counted
Medication-free, strictly screened, four to six weeks of treatment
Whoever came in, usually on medication, often 30 or more sessions
Response
29.3% on active TMS vs 10.4% on sham
58% clinician-rated in a 307-patient cohort; 41% self-rated in 340 US veterans
Remission
18.6% on active TMS vs 5.0% on sham
37% clinician-rated; 20% in the veterans cohort
Who was blinded
Patient and rater, and in the better trials the operator too
Nobody
Comparison group
Yes — that is the point of the design
None
Neither column is a marketing number and neither one is the truth on its own. The left column tells you how much of the improvement belongs to the magnetic field. The right tells you how people who did this actually ended up. If you are deciding whether to give up six weeks of mornings, the right column answers your question. If you are deciding whether the machine is doing anything, it is the left.
The controlled number
What the coil adds, once you subtract everything else
The largest pooled analysis of double-blind sham-controlled trials — 29 studies, 1,371 patients — puts active rTMS at 29.3 percent response against 10.4 percent on a sham coil, and 18.6 percent remission against 5.0 percent. The odds ratio is 3.3 for both.
Percent of patients · pooled across 29 double-blind sham-controlled trials, n=1,371 · track ends at 40%
Response — active rTMS
29.3% · about 3 in 10
Response — sham coil
10.4% · about 1 in 10
Remission — active rTMS
18.6% · fewer than 2 in 10
Remission — sham coil
5.0% · 1 in 20
Two things are worth taking from that figure rather than from the headline. The first is that the sham bars are not zero. Roughly one person in ten improves substantially on a coil that is doing nothing, and one in twenty goes into remission on it — an effect large enough that a 2018 analysis of 61 trials measured the sham response at a Hedges' g of 0.8, which is not a rounding error. The second is that what the trials measured is the gap between the bars, not the height of the bright one.
The gap
Six reasons the practice numbers run higher
None of these make clinic data dishonest. They make it a different measurement — and the version of it that gets quoted is almost always the highest number available.
01
No comparison group
In a clinic every improvement is counted as treatment effect. In a trial the sham arm's 10 percent gets subtracted first, because some of it would have happened anyway.
02
Less treatment resistance
Response falls as the number of failed medications rises. It is the most consistent predictor in the literature, and walk-in patients have usually failed fewer than trial patients.
03
Softer instruments
Registries use clinician impressions and self-report questionnaires. Trials use blinded structured interviews. Within one 307-patient cohort the answer moved from 58 to 56 to 41 percent depending only on which scale you read.
04
Nobody is blinded
The clinician scoring the improvement is the clinician delivering the treatment, and both of them know the coil is real. That is the single largest difference between the two columns.
05
Longer courses
Registry courses commonly run 30 to 36 sessions. The pivotal trial averaged 26 by week six, and there is a real dose-response relationship, so longer courses genuinely do better.
06
Who reports
Registry sites take part voluntarily, and the largest registry was run by the device manufacturer. The one large non-industry cohort — 340 US veterans who completed an adequate course — found 41 percent response and 20 percent remission, roughly half the manufacturer figures on a comparable design.
What the label says
The trial that cleared the device missed its own primary endpoint
In 2007, 301 medication-free patients across 23 sites were randomized to active TMS or a sham coil. The pre-specified primary outcome was the change in MADRS depression score at four weeks. It came back at p = 0.057, which is a miss. Significance was reached only in a post-hoc analysis that removed six patients whose baseline scores sat below the entry threshold.
"Outcome on the primary efficacy endpoint (MADRS change from baseline at 4 weeks) favored NeuroStar TMS Therapy (P=0.057) over sham treatment."
The device's own labeling states that plainly. Almost no clinic page does. Several secondary measures did separate — the clinician-rated Hamilton scales at p = 0.006 and p = 0.012 — while the patient-rated self-report scale did not, at p = 0.058. Clearing the device on the weight of that evidence rather than on a single number is defensible. It is just not the story the marketing tells.
The trial worth knowing about was not paid for by a device maker
Three years later an NIMH-funded trial at four university sites, using an improved sham with scalp electrodes, randomized 190 patients and reported 14.1 percent remission on active TMS against 5.1 percent on sham — an adjusted odds ratio of 4.2 and a number needed to treat of 12. It also checked whether its blind had held, and it had. That is the cleanest evidence that TMS does something, and it is a much smaller effect than the number at the top of the search results.
And one large trial found nothing at all
A double-blind study of 164 US veterans with treatment-resistant depression reported 40.7 percent remission on active TMS and 37.4 percent on sham. No separation. The authors put the sham arm's unusually good result down to the trial itself: close monitoring, controlled medication, and somebody to see five days a week.
Durability
Getting out of an episode is not the same as staying out
This is the part the top results skip, and it is where the honest answer gets less comfortable. Harvard Health quotes a clinician on it directly: TMS is really good at getting people out of a depressive episode, but it is not good at preventing a future one.
46%
of people who responded were still responding at twelve months, pooled across 19 studies
36%
of one 257-patient cohort had TMS reintroduced inside the same year
6 weeks
is the longest course the device labeling claims established effectiveness for
The best pooled figure comes from a 2019 meta-analysis of 19 studies. Among people who responded to an acute course, 66.5 percent were still responding at three months, 52.9 percent at six, and 46.3 percent at twelve. Read the other way round: about half the people it worked for had lost the response inside a year. In the largest naturalistic follow-up, 62.5 percent of those who finished acute treatment as responders or remitters held it through twelve months — but they did so on continuation antidepressants, with TMS available again if they slipped, and 36.2 percent took it.
"The acute effectiveness of NeuroStar TMS Therapy has not been established beyond a six-week treatment course for MDD."
"This device has not been evaluated for durability of antidepressant effect in controlled clinical trials."
So a page promising lasting remission is making a claim the device's own label declines to make. What the evidence does support is narrower and still worth having: a real chance of getting out of this episode, and a reason to have a plan for what comes after it. Ask a clinic what its maintenance protocol is before you start, not after.
The comparison that matters
The real question is not TMS versus nothing
Almost nobody weighing up TMS is choosing between TMS and no treatment. They are choosing between TMS and trying another medication. That comparison flatters TMS more than the sham-controlled numbers do, because the alternative is worse than most people assume.
After this many failed antidepressants
Chance the next one produces remission
None — the first medication tried
36.8%
One
30.6%
Two
13.7%
Three
13.0%
Self-rated remission by treatment step, 3,671 patients, STAR*D
By the third medication the odds have collapsed to about one in seven, and they do not recover at the fourth. That is the figure a person considering TMS should be holding in their other hand. Against a 13 percent chance from the next prescription, even the pessimistic sham-controlled reading of TMS looks reasonable, and the practice numbers look good.
There is one direct comparison. A 2024 randomized trial put 278 patients with treatment-resistant depression on either added rTMS, added aripiprazole, or a switch to a different antidepressant. Adding rTMS beat the switch on depression scores at eight weeks; adding aripiprazole did not reach significance against the same comparator. The trial was open-label, with no sham and nobody blinded, which is a real limitation and the reason it is not the last word. It is still the closest thing to an answer to the question people are actually asking.
Prediction
What moves your odds, and what nobody can tell you
Associated with a better result
Fewer previous antidepressant failures — the most consistent predictor in both large trials
A shorter current episode
No comorbid anxiety disorder
Being currently employed
Completing a full course — 30 or more sessions, not 20
Claimed, but not established
Any scan, blood test or genetic panel that predicts an individual response
A best device or coil for efficacy — the largest head-to-head trial found theta-burst and standard 10 Hz equivalent
Response predicted from a symptom profile alone
A clinic's own quoted success rate, unless it says what was measured and by whom
The honest summary from a 2025 clinical guide is that no consistent response predictors have been identified. The strongest published prediction model, built on the 388 patients of the largest head-to-head trial, reached a c-statistic of 0.687 — better than a coin flip, nowhere near good enough to tell one person what to expect. If a clinic offers to predict your response, ask what the prediction is based on.
If you are reading this after a course that did not work
Roughly half the people who finish a course of TMS do not get a meaningful response. Being one of them is not a verdict on you, and it is not evidence that nothing will work. There are routes after TMS — a different protocol or target, esketamine, ECT, which outranks every form of TMS in the network analyses that have compared them, and combinations of these.
If you are in crisis right now, call or text 988 in the US to reach the Suicide and Crisis Lifeline, or text HOME to 741741 for the Crisis Text Line. Both are free and both answer 24 hours a day.
FAQ
The questions people ask about this
The first three are the questions Google is already rendering above the top result.
What is the success rate of TMS therapy?
There is no single one, and any page that gives you one number without saying what it measured is guessing. In blinded trials against a sham coil, about 29 percent respond and 19 percent reach remission. In ordinary practice, with nobody blinded and no comparison group, clinician-rated response runs around 58 percent and remission around 37 percent; a large non-industry cohort of US veterans found 41 and 20. Figures quoted as 70 or 80 percent have no traceable primary source.
Is TMS better than antidepressants?
For somebody who has already failed two or more, probably yes. The chance a third medication produces remission is about 14 percent, and the only randomized head-to-head trial found adding rTMS beat switching antidepressants — though it was open-label, so treat it as strong rather than conclusive. For a first-line patient the question does not really arise, because TMS is not cleared for that use.
Why is TMS not more popular?
Cost and time, mostly. A course is roughly 30 daily visits over six weeks, in person, during working hours, and insurers generally require documented failed medication trials first. The effect is also moderate rather than dramatic — real and replicated, but smaller than the way it tends to be sold, which makes it a harder thing to recommend casually.
Does TMS work long term?
About 46 percent of the people who respond are still responding at twelve months, and roughly a third of one large cohort needed a second course inside the same year. The device labeling states outright that durability has not been evaluated in controlled trials. Plan for maintenance rather than assuming you will not need it.
How long does TMS take to work?
Most people who respond notice a change somewhere between weeks two and four, and in the trial data response rates were still climbing at week six. Feeling nothing in the first fortnight is common and is not a sign the course has failed. Some people also feel temporarily worse partway through, which has its own page.
Is TMS a hoax?
No. It has been through independent, NIMH-funded, blinded, sham-controlled trials that no device maker paid for, and it separates from sham across pooled analyses of dozens of studies. The fair criticism is not that it does nothing. It is that the effect is smaller than the marketing implies, and that the numbers most often quoted come from uncontrolled clinic data presented as if they were trial results.
Sources
Every number on this page
Grouped by what they support. Every entry links out to the paper, the registry or the device labeling itself, not to a summary of it.
Listings show each clinic's phone number, website and hours. Before you book, the questions worth asking are what its own outcome figures measured, how many sessions a standard course runs to, and what happens if you respond and then relapse.
