medication, read plainlylast verified 2026-09-25

when an antidepressant stops working

it worked for three years and then one spring it did not. or you stopped it, felt fine for a while, went back on, and the second time was nothing like the first. the forums call it poop-out; the journals call it tachyphylaxis, and they have been arguing about what it is since 1998. here is what has been measured, what has only been proposed, the case that it is not the drug at all, and the questions to bring to the person who prescribes for you. this page does not tell you to start, stop, restart or change anything.

five questions, answered from the record

how often does an antidepressant stop working?[1][2][3]

somewhere between 9 and 57 percent of people on maintenance treatment, depending on who was studied and for how long; the 1998 review gave that range and the 2019 review, twenty years of trials later, gave the same one. the cleanest single number is from the NIMH cohort followed up to 20 years: of 171 stretches of maintenance treatment after recovery, symptoms came back during 43, which is 25 percent, at a median of 20 weeks in. people with the melancholic subtype were at higher risk. so it is common, it is not most people, and the range is wide because the definition is.

what is actually happening?[2][4]

nobody knows, and the reviews say so. the 1998 paper listed seven candidates and none has been ruled in: the placebo part of the original response wearing off; pharmacological tolerance; the depression itself getting worse; the illness changing character; a build-up of an unhelpful metabolite; undiagnosed rapid cycling; and the drug never having been protective in the first place. the 2003 paper proposed a mechanism, the oppositional model: the body recruits processes that push back against the drug, the effect fades, and when the drug is stopped those processes run unopposed for a while. its author called it a hypothesis that needs to be tested, and it still is one.

does it work the second time?[4][5][6]

this is the question the forums ask most and the literature answers least. what has been measured: in 276 people starting sertraline, each previous antidepressant course lowered the odds of responding by about a fifth (odds ratio 0.81); in 240 people randomised to paroxetine or cognitive therapy, a history of more antidepressant courses predicted a worse response to the drug but not to the therapy. the 2003 review reports resistance on rechallenge with the same drug “in a few patients.” what has not been measured is the thing you want: the response rate for restarting the exact drug that worked before. the sertraline data is a post hoc analysis and the paroxetine finding has not, to the authors’ own knowledge, been replicated. so: the evidence leans toward each course being a little less likely to land than the last, and it is thin.

what do prescribers do about it?[1]

four things, each with what the 2019 review calls limited evidence: change the dose, switch to a different class, add a second drug, or add psychotherapy. that is the entire evidence-based menu, and the review’s conclusion is that few established strategies exist. which of the four, if any, is right for you is not knowable from a page; it depends on what happened, how long you have been on the drug, and what else is going on. take the list to your prescriber and ask which one they are reaching for and why.

can i pick the next drug from the rankings?[7]

not from the evidence you will be shown. the 21-drug network meta-analysis everyone cites ranked drugs for a first acute course in 116,477 trial participants: all of them beat placebo, the differences between them were small with wide intervals, and the certainty of evidence was moderate to very low. its abstract says nothing about which drug to try after another one has faded, and no ranking exists for that question. if a switch is on the table, that is your prescriber’s call with your history in front of them, not a league table.

why the range is 9 to 57

a range that wide is not a measurement, it is a definition problem. the trials counted different things as relapse, over different lengths of follow-up, in populations selected by strict criteria that the 2019 review says may not generalise.[1] the NIMH number is narrower because it is one cohort watched for twenty years, but it is observational: nobody controlled what those 103 people were prescribed, and 20 weeks is the median point at which symptoms returned, not a rule.[3] read the range as “this happens to a lot of people, and nobody has counted it properly.” both halves of that sentence matter.

the other direction

the case that this is not the drug: three of the seven explanations on the 1998 list are about the illness, not the pill. depression that recurs on treatment may be depression that was always going to recur, with the drug delaying it rather than failing.[2] the 25 percent in the NIMH cohort means 75 percent of maintenance periods did not see symptoms return.[3] the oppositional-tolerance model is a hypothesis its own author says has not been tested, and the same abstract states that antidepressants are crucial in treating major depressive episodes.[4] the “each course works less well” finding comes from a post hoc analysis and one trial that its authors say needs replication before it should change what anyone is offered.[5][6] and the 21-drug meta-analysis stands: every antidepressant beat placebo for a first acute course.[7] a page that used only the tolerance papers would be selling you a story. a page that used only the efficacy papers would be selling you a different one.

what this page will not do

it will not tell you to restart, stop, raise, lower or switch anything. the four strategies with any evidence are decisions that depend on your history, and stopping an antidepressant on your own has its own literature, which is on stopping antidepressants: withdrawal and relapse. if the question is whether you should still be on the drug at all, the review checklist is the list to take into the appointment. the questions below are the ones specific to a drug that faded.

what to ask your prescriber

“is this the drug fading, or the depression coming back through it?”[2]

the seven explanations on the 1998 list split roughly in half between the drug and the illness. a prescriber who can say which they think it is, and why, is doing the diagnosis the reviews say has not been done.

“how many antidepressant courses have i had, and does that change what you would try?”[5][6]

in the two trials that counted prior courses, more of them predicted a weaker response to the next drug and no weaker response to cognitive therapy. your prescriber should know the number and have a view on what it means.

“which of the four strategies are you thinking of, and what is the evidence for it?”[1]

dose change, class switch, augmentation, psychotherapy. each has limited evidence; none is established. the answer tells you whether the plan is a reason or a reflex.

“if we change something, how will we know in eight weeks whether it worked?”[5]

the sertraline study measured response at eight weeks with a rating scale. a number agreed in advance is the difference between a trial of something and a drift into the next thing.

“is therapy on the table, whether or not we change the medication?”[1][6]

psychotherapy is one of the four listed strategies, and in the 240-person trial it was the one that prior drug courses did not blunt. it is not either-or.

questions people ask

Why did my antidepressant stop working?

Nobody has established why. Reviews from 1998 and 2019 list the same candidates: the placebo component of the original response wearing off, pharmacological tolerance, the depression worsening or changing, an unhelpful metabolite, unrecognised rapid cycling, or the drug never having been prophylactic. Loss of effect during maintenance treatment has been reported in 9 to 57 percent of patients across trials, and in 25 percent of 171 maintenance periods in a 20-year NIMH cohort. Whether the drug or the illness explains your case is a question for your prescriber, not a page.

What is antidepressant tachyphylaxis or poop-out?

Tachyphylaxis is the clinical term for a re-emergence or worsening of depressive symptoms despite continued treatment with an antidepressant that previously worked. Poop-out is the informal name. The 2019 review found rates from 9 to 57 percent depending on the population and follow-up, and limited evidence for four management strategies: dose change, switching class, augmentation or combination, and psychotherapy.

If I go back on an antidepressant after stopping, will it work again?

The direct question, restarting the same drug, has not been measured in a trial whose abstract we could read. What has: in 276 people starting sertraline, each prior antidepressant course reduced the odds of responding by about 20 percent; in 240 people randomised to paroxetine or cognitive therapy, more prior courses predicted a worse response to paroxetine but not to therapy; and a 2003 review reports resistance on rechallenge with the same drug in a few patients. The evidence is thin and leans toward each course being somewhat less likely to work than the last. Whether to restart, and what, is a decision for you and your prescriber.

Should I increase the dose or switch when my antidepressant stops working?

This page cannot answer that and does not try. Dose change, class switch, augmentation and psychotherapy are the four strategies with limited evidence in the 2019 review, and none is established. Which of them fits your situation depends on your history and is a decision to make with the person who prescribes for you. Do not change a dose on your own.

sources

  1. Kinrys G, Gold AK, Pisano VD, et al. Tachyphylaxis in major depressive disorder: a review of the current state of research. Journal of Affective Disorders 2019;245:488-497. Review of clinical trials and meta-analyses to January 2017. “Rates of tachyphylaxis varied from 9% to 57% depending on the patient population and duration of follow-up. Limited evidence suggests potentially beneficial strategies for managing tachyphylaxis, including change in antidepressant dosing, switch of class of antidepressant medication, augmentation or combination pharmacotherapy, and psychotherapy.” Conclusion: “Few established treatment strategies exist.” Studies “largely heterogeneous” with strict inclusion criteria. PMID 30439676. https://doi.org/10.1016/j.jad.2018.10.357
  2. Byrne SE, Rothschild AJ. Loss of antidepressant efficacy during maintenance therapy: possible mechanisms and treatments. Journal of Clinical Psychiatry 1998;59(6):279-288. MEDLINE review from 1966. “The return of depressive symptoms during maintenance antidepressant treatment has occurred in 9% to 57% of patients in published trials. Possible explanations include loss of placebo effect, pharmacologic tolerance, increase in disease severity, change in disease pathogenesis, the accumulation of a detrimental metabolite, unrecognized rapid cycling, and prophylactic inefficacy.” “Double-blind controlled studies are needed to ascertain the optimal strategy.” PMID 9671339. https://doi.org/10.4088/jcp.v59n0602
  3. Solomon DA, Leon AC, Mueller TI, et al. Tachyphylaxis in unipolar major depressive disorder. Journal of Clinical Psychiatry 2005;66(3):283-290. NIMH Collaborative Depression Study, 103 subjects with unipolar major depression enrolled 1978 to 1981 and followed up to 20 years; treatment observed, not controlled. 171 maintenance-treatment intervals after recovery, median 20 weeks. “Tachyphylaxis occurred during 43 (25%) of these 171 maintenance treatment intervals.” The melancholic (endogenous) subtype significantly raised the risk. PMID 15766292. https://doi.org/10.4088/jcp.v66n0302
  4. Fava GA. Can long-term treatment with antidepressant drugs worsen the course of depression? Journal of Clinical Psychiatry 2003;64(2):123-133. Literature review. Reported clinical findings point to “occurrence of tolerance to the effects of antidepressants during long-term treatment, onset of resistance upon rechallenge with the same antidepressant drug in a few patients, and withdrawal syndromes following discontinuation.” Proposes the oppositional model of tolerance: “continued drug treatment may recruit processes that oppose the initial acute effects of a drug and may result in loss of clinical effect. When drug treatment ends, these processes may operate unopposed, at least for some time, and increase vulnerability to relapse.” The author states the hypothesis “needs to be tested” and that “antidepressant drugs are crucial in the treatment of major depressive episodes.” PMID 12633120. https://doi.org/10.4088/jcp.v64n0204
  5. Amsterdam JD, Williams D, Michelson D, et al. Tachyphylaxis after repeated antidepressant drug exposure in patients with recurrent major depressive disorder. Neuropsychobiology 2009;59(4):227-233. Post hoc analysis: 276 patients with major depressive disorder treated with sertraline 150 to 200 mg daily for 8 weeks; non-responders then randomised to sertraline plus atomoxetine (n = 72) or sertraline plus placebo (n = 74) for 8 more weeks. “The number of prior antidepressant drug exposures was negatively associated with response to initial sertraline therapy (odds ratio = 0.81, p = 0.0035). The odds ratio indicates a 19.9% reduced likelihood of response with each prior antidepressant treatment trial.” Prior trials were not associated with response in the continuation phase. PMID 19571597. https://doi.org/10.1159/000226611
  6. Leykin Y, Amsterdam JD, DeRubeis RJ, Gallop R, Shelton RC, Hollon SD. Progressive resistance to a selective serotonin reuptake inhibitor but not to cognitive therapy in the treatment of major depression. Journal of Consulting and Clinical Psychology 2007;75(2):267-276. 240 patients with moderate-to-severe major depression randomised to paroxetine or cognitive therapy for 16 weeks; prior antidepressant exposure assessed by structured interview, self-report and medical records. “More prior AD exposures predicted poor response to paroxetine therapy but not to CT”; the authors call for replication before the finding changes practice. PMID 17469884. https://doi.org/10.1037/0022-006X.75.2.267
  7. Cipriani A, Furukawa TA, Salanti G, et al. Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis. Lancet 2018;391(10128):1357-1366. 522 trials, 116,477 participants. “All antidepressants were more effective than placebo,” with odds ratios from 2.13 (amitriptyline) to 1.37 (reboxetine); differences between drugs ranged from 1.15 to 1.55 for efficacy “with wide CrIs on most of the comparative analyses”; certainty of evidence “moderate to very low.” It is an acute-treatment ranking. The abstract says nothing about second courses, rechallenge or switching after loss of effect, and it is cited here only for what it does say. PMID 29477251. https://doi.org/10.1016/S0140-6736(17)32802-7

related: antidepressants and emotional blunting · stopping antidepressants: withdrawal and relapse · the review checklist · does TMS work for depression

adam