TMS Nearby

The TMS dip: why you might feel worse before you feel better

If you are partway through a course of TMS and feeling worse, you have probably already found a dozen pages telling you this is normal, that your brain is rewiring, and that it comes right. Almost none of them cite anything. This page sets out what has actually been measured, how to tell a flat stretch from a real deterioration, and which symptoms mean you should call today rather than mention it next week.

The short answer

There is no published evidence for a "TMS dip." No trial, no cohort study and no case series describes a temporary mid-course worsening that reliably recovers. What the evidence does support is a more useful distinction. Feeling flat partway through is common, and it is a poor predictor of how the course ends. Feeling steadily worse is uncommon, and in the one study that measured it session by session, it did not reverse. Those two situations look similar from the inside and need different responses, which is what the rest of this page is about.

Read this part first

Some of this needs a call today

Most of what follows can wait until your next appointment. These cannot.

Call your clinic today

  • You are sleeping much less and do not feel tired.

    With racing thoughts, unusually elevated or irritable mood, talking faster than usual, a jump in activity, or spending and risk-taking that is out of character. That pattern is treatment-emergent mania or hypomania. It is rare, and no more common than on a sham coil, but it changes the plan: published guidance is to consider stopping TMS while continuing any mood-stabilizing medication.

  • Thoughts of harming yourself, new or worse.

    The device label warns that people with major depression can get worse or develop suicidal thoughts during treatment, whatever treatment they are on, and tells clinicians to watch for it and consider changing or stopping the course. Same-day call, not a next-session mention.

  • Facial numbness.

    Across the whole NeuroStar trial program, one serious adverse event was classified as probably related to the device: one-sided facial numbness, which fully resolved once treatment stopped.

  • New ringing, muffled hearing, or a blocked-up feeling in your ears.

    The safety guidelines say anyone reporting hearing loss, tinnitus or aural fullness after TMS should be referred for an auditory evaluation. The only documented case of permanent hearing change in the TMS literature was an earplug that slipped out mid-session. If a plug works loose, say so and stop.

  • New floaters, flashes of light, or a curtain across your vision.

    Two published case reports describe retinal tear and posterior vitreous detachment after rTMS. Causation is not established. It is an eye emergency either way.

Call 911

  • A seizure.

    TMS-associated seizures happen during stimulation or within seconds of it, are self-limited, and per the 2021 international guidelines have not resulted in a seizure disorder. A seizure hours or days after a session has never been documented as caused by TMS. A first-ever seizure is still an emergency.

  • Sudden one-sided weakness or numbness, sudden confusion or trouble speaking, sudden trouble seeing, sudden trouble walking, or a sudden severe headache with no known cause.

    These are stroke signs. They are not TMS side effects. Call 911 and do not drive yourself.

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, or text the word Ayuda. 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. In Spanish, text HOLA or AYUDA to the same number.

More crisis resources →

The evidence base

Where the phrase comes from

Search the medical literature for the TMS dip and there is almost nothing there. Structured searches of Europe PMC for mid-course worsening in TMS return no trial, no cohort, no case series and no review describing it.
Search, Europe PMCResults about mood worsening in TMS
"initial worsening" + transcranial magnetic stimulationnone
"transient worsening" + transcranial magnetic stimulation4 papers, none about mood
"early worsening" + depression + stimulationnone
"feel worse" + transcranial magnetic stimulationnone
"TMS dip"one case report, one patient

One peer-reviewed paper uses the phrase. It is a case report about a single patient, published in 2025, and it describes the dip as something "anecdotally described by patients." In that same case the week-three worsening coincided with a custody dispute, which the authors name explicitly as a probable cause. The paper is the only citation the phrase has, and it is a paper that hedges it.

The figure you will see quoted online — that around 20% of patients experience a dip, usually in weeks two or three, lasting a few days to two weeks — traces back to a clinic blog post with no source behind it. It has since been repeated across enough clinic websites that search engines now summarize it as established fact. It is not established. Nobody has counted.

That matters for a specific reason. If everyone dips and it always passes, then feeling worse is reassuring evidence the treatment is working, and continuing is obviously right. If nobody has measured how often mid-course worsening happens or how often it resolves, then feeling worse is information, and what you do with it depends on which kind of worsening it is.

What the trajectory studies found

What actually happens to people over a course

Several large studies have tracked how symptoms move during TMS, session by session or week by week, and then sorted patients into patterns. Between them they cover several thousand people. None of them found a group that got worse and then recovered.

228 people, asked after every one of 20 sessions how they felt

PatternPeopleShareWhat it looked like
Strong improvers2712%Noticed improvement in the first week, much better by the end
Delayed improvers5424%Nothing until about halfway, then improved to the end
Plateau improvers4721%Improved at first, then stopped improving
Non-improvers8839%Noticed no improvement
Worsening125%Reported feeling worse across sessions

Two findings from that table are the reason this page exists. A quarter of people felt nothing at all until roughly the halfway point, and then got better — that is not a dip, it is a flat start, and it is the single most common thing mistaken for one. And the worsening group did not turn around: not one of those 12 people met the response threshold on the clinician-rated scale at 8 weeks, at 16 weeks, or at 26 weeks.

Two caveats. Twelve people is a small group, and the study measured how people felt relative to baseline rather than using a full depression scale — the authors note that measure has not been independently validated. In that trial, anyone reporting they felt the same or worse was reviewed by a psychiatrist with expertise in mood disorders. That is what should happen outside a trial too.

Everywhere else the picture is the same

388

THREE-D trial

Four patterns: non-response 11%, rapid response 19%, and two slower linear responses at 30% and 40%. No worsening group.

324

Up to 51 sessions

Four patterns: slowed response 49%, linear response over an extended course 23%, rapid response 21%, non-response 7%. The largest group by far was the slow one.

402

Measured at 5, 10, 20 and 30

The average curve drops quickly to session five, then keeps dropping more slowly and steadily to the end. It never turns upward.

Across every large trajectory study published to date, the groups that do badly are described as flat, minimal or non-responding. Not one is described as dipping and recovering.

The distinction that matters

Flat is common. Worse is not.

This is the distinction the folklore flattens, and it is the one that should shape what you do next.

If you are flat, the evidence is genuinely on your side

A study of 248 patients at one hospital TMS clinic asked a blunt question: if somebody has barely improved after ten sessions, how reliably does that predict how the course ends? Judged at four weeks, the answer looked decisive. Judged against a full 36-session course, it did not.

93.1%

How often a poor two-week response predicted non-response, judged at four weeks

72.3%

The same prediction, judged against a full 36-session course

15–34%

Of people who were actively worse at week two and still responded by the end

Roughly one in four people who looked like non-responders at two weeks still responded by the end. The 15% and 34% figures are the same cohort measured on the clinician-facing scale and the patient-completed one — people whose scores at week two were worse than at baseline, who responded anyway. A separate registry analysis of 7,215 patients reached the same conclusion from the other direction: early improvement strongly predicts ending up a responder, but a lack of early improvement is a poor predictor of ending up a non-responder. Its predictive value, the authors said, is too low to base treatment changes on. That same registry found the people who benefited from courses longer than 36 sessions were characterized by exactly this — less improvement early, then a slower but steady climb.

If you are steadily worse, that is a different conversation

The evidence above is about people who are flat or slightly worse and then recover. It is not about people on a downward slope. The only study that tracked perceived worsening session by session found that group did not recover, and the current US consensus contains the exception explicitly: patients should be encouraged to complete all 36 sessions unless clinical worsening or intolerance requires a change of therapy. That clause is the part the reassurance version leaves out.

Measure it, do not estimate it

How to tell which one you are in

You cannot answer this from how you feel today. Depression makes yesterday hard to remember accurately and makes today feel permanent. The answer comes from scores, and the scores need to exist.

The PHQ-9 runs 0 to 27

1–4

Minimal

5–9

Mild

10–14

Moderate

15–19

Moderately severe

20–27

Severe

Response conventionally means at least a 50% drop from where you started. Remission usually means a score under 5. Neither describes how you feel — both are administrative thresholds, and you can move a meaningful amount and still be counted a non-responder.

How much change actually counts as change

There is no agreed answer, and the disagreement is worth knowing about because it protects you from reading noise as a trend. Three defensible estimates exist for how much the PHQ-9 has to move before a person notices.

5 points

From a statistical estimate of the scale's measurement error

1.7 points

About 20%, from asking patients at moderate severity whether they felt better

0–14 points

Depending on where you started, averaging 3.7 points or 23%

All three agree on the thing that matters here: the sicker you were at the start, the bigger the move has to be before you can feel it. If you began at 22 and you are at 19, you have not deteriorated. You have barely moved, which is a different problem with a different answer. Two scales can also disagree about you — in one 56-patient study, response was 20% on the clinician scale, 16% on the self-report scale and 25% on either.

The only published PHQ-9 series that uses the phrase. One patient, 30 sessions.

26

Baseline

27

Week 3

11

 

18

 

10

 

13

 

5

 

3

Week 9

7

Final

Read left to right. The two highlighted assessments are the "dip" — a rise at week three and another at the fourth assessment. Both sit inside an overall descent from 26 to single figures. In the only published case using the term, the week-three rise coincided with a custody dispute, which the authors name as a probable cause.

Rule these out first

Nine things that are not the coil

Before concluding the treatment is doing this to you, these are worth ruling out. Several are common, several are fixable, and one of them is the single most under-communicated factor in TMS.
  • 1

    Benzodiazepines

    Two separate cohorts found markedly lower response rates in people taking them during a course: 16.4% against 35.5% in one, 18% against 38% in the other. A trial analysis found that not using them was associated with the rapid-response pattern. Consensus says the evidence is not strong enough to guide prescribing. Do not stop or reduce one on your own — abrupt reduction of a medication with anticonvulsant properties is itself a listed seizure risk factor.

  • 2

    A medication change you did not connect to this

    The consensus is to keep psychiatric medication stable through a course. Any change should trigger a re-check of your motor threshold, because it can shift how much stimulation you are actually receiving.

  • 3

    Sleep

    Cortical excitability rises with hours awake and after a night of lost sleep, and falls again after recovery sleep. That is one reason guidance says to re-check the motor threshold after a change in sleep pattern. Insomnia is also a side effect of treatment in its own right, reported by 5 to 7% of people.

  • 4

    Alcohol and caffeine around treatment

    Both are named in current guidance as reasons to re-check your dose. Alcohol also appears in the device labeling as a factor raising the risk of thermal injury from a warm coil.

  • 5

    Anxiety, and the way it is counted

    In a registry of 1,820 patients, people who started with significant anxiety improved by the same absolute amount on the depression scale as everyone else. Because they started higher and finished higher, they were significantly less likely to cross the response and remission lines. You can be improving as much as anybody and still be told the numbers do not look good.

  • 6

    Activation

    The international safety guidelines describe a sub-manic activation syndrome — new or worse insomnia, agitation or anxiety — and say it is "not uncommon" in people receiving daily high-frequency TMS in ordinary practice. It is distinct from both a dip and from hypomania, and it has responses available.

  • 7

    Your dose is not as fixed as it looks

    In a clinic that measured the motor threshold every single day across 374 patients, it moved more than 5% from baseline in about half of all sessions. Simulating a single measurement at the start of a course, roughly 5% of sessions would have run at 25% above the intended intensity. Some of "today was harder" is real.

  • 8

    What you expected

    Headache was reported by 58.2% of people on active TMS and 55.1% of people on a sham coil that does nothing. Across 93 sham-controlled TMS trials, dropout from side effects was 2.5% on active and 2.7% on sham. One useful finding for anyone who arrived full of dread: positive expectation predicted better outcomes in a 177-patient study, but negative expectation did not predict worse ones.

  • 9

    Your life

    No study has ever quantified how much mid-course mood change is attributable to what is happening outside the clinic. The one published case report that uses the phrase "TMS dip" is a case where the worsening coincided with a custody dispute. Write down what else was going on in the week your scores moved, because nobody else will.

The middle ground

What your clinic can change this week

Feeling worse is not a binary between finishing and quitting. There is a middle, and most of it is available at your next appointment.

Re-measure the motor threshold

It drifts for physiological reasons, and it has its own billing code, so it is a normal thing to ask for rather than a favor. Guidance suggests checking at baseline and then weekly, or any time your medication, sleep or clinical status changes.

Adjust the coil position and the intensity

The published management algorithm for scalp pain is to reassure, offer an over-the-counter or topical analgesic, make a subtle adjustment to coil position, and slightly reduce the field intensity. Positioning comes first: head, neck and spine support in the chair reduces the non-specific discomfort that turns into a post-treatment headache. A foam spacer produced a modest measured reduction in scalp pain in one pilot study.

Know the floor, though

Antidepressant efficacy below 110% of motor threshold is described in the consensus literature as questionable. Reducing intensity is a tolerability trade, not a free one, and an extended ramp-up can under-dose you.

Switch protocol

Standard 10 Hz stimulation and intermittent theta burst are equally recommended as first-line and were non-inferior to each other across 414 patients. Theta burst was rated slightly more painful. Nobody has run a trial on switching a specific non-responder from one to the other mid-course, so this is a judgment call rather than an evidence-based move.

Switch side

The one switching strategy with real data behind it. In the open phase of a national trial, patients who failed left-sided high-frequency stimulation were moved to low-frequency stimulation on the right, and 26% of them remitted. Across that whole open phase, 43 of 141 people who had already failed a course eventually reached remission.

Extend past 36 sessions

In a 7,215-patient registry, extending beyond 36 was associated with further improvement with no sign of a plateau. The practical obstacle is not clinical: insurers usually authorize 36.

Say something during the session, not after it

If you get unexpected large involuntary movement, or movement that carries on after the pulses stop, the consensus instructs the operator to pause treatment and notify the supervising clinician. The operator is required by the device labeling to be in the room with you at all times and to be qualified to recognize and manage a seizure. Telling them mid-session is what they are there for.

The decision

Should you keep going?

Nobody reading this page can answer that, and any page that claims to is selling something. What can be set out is what the guidance says and what the numbers support.

"Patients should generally be encouraged to receive all 36 treatment sessions unless clinical worsening or intolerance necessitates transition to a different therapy. In cases of lack of response after four weeks of treatment, a risk-benefit discussion with the patient that addresses the possibilities of delayed response and non-response should guide the decision regarding continuing treatment."

National Network of Depression Centers, Clinical TMS Society and IFCN consensus, 2025

Read that as three separate instructions, because it is three.

Finish the course, by default

Courses shorter than 30 sessions had worse outcomes than any longer group in the largest registry analysis available. Over 90% of Clinical TMS Society members report first seeing remission between weeks four and six, which is to say after most people have already had the bad two weeks.

Unless you are getting worse or cannot tolerate it

That exception is written into the guidance and it is not a formality. Steady, measurable worsening is a documented reason to change course, and a reason to say so now rather than at the end.

At four weeks with no response, have the conversation properly

Not a corridor question to the technician. A scheduled discussion with the prescribing clinician that covers both possibilities: that you are a late responder, and that you are not going to respond. Both are real, both have numbers behind them, and the point of the conversation is that you get to weigh them.

For the honest picture of how often TMS fails and what happens next:

Why some people say TMS ruined their life →

Take this with you

What to say at your next session

Six requests. All of them are things the standards already expect your clinic to be able to do.
  1. 1

    "Can I see my scores from the start of the course and from this week?"

    Documenting rating-scale results for the current course is a listed item in the consensus procedure note.

  2. 2

    "I want this recorded as a treatment-emergent adverse effect — the symptoms, how long, how bad."

    Also a listed item: if present, document symptom details, duration and intensity.

  3. 3

    "Can we re-check my motor threshold?"

    Especially if your medication, sleep or drinking has changed since the last check.

  4. 4

    "Can the doctor call me today rather than at my next review?"

    Technicians are expected under consensus practice standards to escalate concerning symptoms — including suicidal thoughts, mania, psychosis and panic — to the supervising clinician. The attending physician is medically responsible for evaluating you during the course.

  5. 5

    "Can we measure more often, ideally every session?"

    With the reason: in the trial that measured after every session, the pattern where nothing happens until halfway through disappeared entirely when the researchers modeled the data as if it had only been collected every fifth session.

  6. 6

    "What happens if I am not better at session 30?"

    Ask before you get there, while it is still a hypothetical rather than a disappointment.

FAQ

The questions people ask about this

  • How long can the TMS dip last?

    Nobody knows, because it has never been measured. There is no published incidence figure, no established onset and no established duration for mid-course worsening in TMS. The durations quoted online, usually a few days to two weeks, trace back to clinic blog posts without sources. What has been measured: in one trial that asked 228 patients after every session, 5% reported feeling worse across the course and none of them recovered to the point of responding at 8, 16 or 26 weeks.

  • How common is the TMS dip?

    No study has counted it. The figure circulating online, that around 20% of patients experience one, has no citation behind it. The nearest real numbers come from trials that sorted patients into patterns: 24% of one group felt nothing at all until roughly halfway and then improved, and 5% reported steadily worsening. Those are different things, and only the second is worrying.

  • What does the TMS dip feel like?

    The symptoms people describe are low mood, anxiety or irritability, fatigue, sleep disruption and mental fog, usually somewhere in the second to fourth week. None of those are specific to TMS. Scalp pain, headache and twitching are side effects of the treatment itself and typically fade after the first week. Increased insomnia, agitation and anxiety are documented in the safety guidelines as a recognized activation pattern in daily high-frequency treatment.

  • Should I keep going if I feel worse?

    The current US consensus is that patients should generally complete all 36 sessions unless clinical worsening or intolerance requires a change of therapy, so it depends on which of those you are in, and that is a question for the prescribing clinician rather than a forum. What the evidence supports: being flat or slightly worse at two weeks is a poor predictor of the final result, and about one in four people who look like non-responders then still respond by the end of a full course.

  • Does feeling worse mean TMS is not working?

    Not necessarily, and not automatically the other way either. In the pivotal trial, all six patients hospitalized for worsening depression were people who were not receiving active TMS. Depression fluctuates during any treatment. The way to tell is a scored scale at a fixed interval, compared against where you started, not against how you felt yesterday.

  • Is it normal to feel more anxious during TMS?

    Anxiety and irritability are listed in the current consensus among side effects occurring in under 5% of people, and the safety guidelines separately describe an activation pattern of new or worse insomnia, agitation and anxiety as "not uncommon" in daily high-frequency treatment. Separately, if you started with significant anxiety you may improve as much as anyone on the depression scale and still be less likely to cross the response threshold, because you started higher. Report it either way.

Sources

Every number on this page

Listed in the order the figures appear. Every entry links out to the study, the guideline or the device label it came from.

The phrase itself

Trajectories

Early response and what it predicts

Measurement

Guidelines and consensus

Confounders

What can be changed

Safety, red flags and escalation

Read next

The pillar, and the two closest siblings

Pillar

What is TMS therapy?

Spoke

Why some people say TMS ruined their life

Spoke

How many TMS sessions you need

Find TMS clinics near you

Listings show each clinic's phone number, website and hours. If you are choosing a clinic, three questions worth asking before you start: how often do you measure my symptoms, how often do you re-check my motor threshold, and how soon can I speak to the prescribing physician if something changes.

Browse all locations →