A 45-year-old woman, Carol, presents with a relapse of major depression. She responded well to escitalopram three years ago, remained stable for a couple of years, and then stopped the medication on her own. Now, despite restarting at the same dose, her symptoms are not improving. She comes in to see you and asks, “Why isn’t it working this time?”

When a patient stops an antidepressant and restarts it later, they don’t always get the same benefit back. It’s a phenomenon we have yet to fully understand, but evidence suggests it is not uncommon. For example, a recent review of 12 studies involving nearly 600 patients found that nonresponse upon restarting a previously effective antidepressant ranged as high as 57% (Majkic N and Taylor D, J Psychopharmacol 2025;39(10):1114–1119). Response rates varied widely across studies, and factors such as duration off medication or tapering strategy did not reliably predict who would fail to respond.

If discontinuing and then restarting an antidepressant may lead to nonresponse the second time around, should we just keep patients on antidepressants indefinitely? Not necessarily. Prolonged exposure has been associated not only with medical adverse effects (eg, reduced bone density, weight gain, bleeding risk, or cardiac effects, depending on the medication), but also with neuroadaptive changes that erode antidepressant response over time.

Why loss of response occurs

One explanation for this loss of effectiveness comes from Giovanni Fava’s “oppositional tolerance” hypothesis (Fava GA, J Clin Psychiatry 2003; 64:123–133). Antidepressants shift neurotransmitter activity in one direction; SSRIs, for example, increase synaptic serotonin. The brain responds by adjusting in the opposite direction to restore equilibrium, with neuroadaptive changes that begin early in treatment and persist over time, such as downregulation of postsynaptic serotonin receptors or alterations in downstream signaling. These counter-adaptations might persist after the medication is stopped, increasing vulnerability to relapse. When restarting an antidepressant later on, the adaptations may blunt its benefit, so the formerly observed response does not fully return.

This framework also helps explain the “poop-out” effect, ie, the waning of benefit during continuous treatment. You’ve probably seen this in your patients and most likely addressed it by escalating the dose, which is often a reasonable first step.

Loss of response after discontinuing and then restarting an antidepressant is a related but distinct clinical pattern. Here, it is reasonable to return to the last effective dose, but if the response does not return, switching or augmenting may be more helpful than continuing to push the same medication. Of course, not every relapse reflects tolerance; missed doses, new psychosocial stressors, or illness progression can also play a role. However, tolerance is one way that long-term exposure can complicate our patients’ outcomes. Repeated stop–start cycles may reinforce the brain’s counter-adaptations and further reduce the chance of regaining benefit, although this hypothesis has not been tested directly.

You tell Carol, “This lack of response happens more than people think. Sometimes, after a medication has been stopped, restarting it does not fully recapture the original benefit.” You explain that you’ll treat this restart like a new treatment trial and give it a fair run, but you also note that you will watch the medication’s efficacy closely and be ready to change course if she doesn’t improve.

Withdrawal masquerading as relapse

Symptoms that emerge during or shortly after discontinuation can reflect something else entirely: withdrawal. Withdrawal can resemble a depressive relapse, particularly after abrupt discontinuation of SSRIs or SNRIs with short half-lives, such as paroxetine or venlafaxine. Patients may experience symptoms that resemble depression, including fatigue, insomnia, anxiety, and irritability, along with more classic discontinuation symptoms such as dizziness, flu-like malaise, headache, and “electric shock” sensations.

In one study, patients whose antidepressants were interrupted for only five to eight days showed worsening depression scale scores; however, when the medications were resumed, there was no evidence of diminished efficacy, suggesting that the apparent “relapse” was actually withdrawal (Rosenbaum JF et al, Biol Psychiatry 1998;44(2):77–87). Misdiagnosing withdrawal as recurrence makes continuous treatment look more protective than it is and complicates decisions about long-term maintenance. Slow tapers reduce withdrawal symptoms and may help mitigate the neuroadaptive changes thought to contribute to tolerance, although evidence for this effect is still speculative.

Navigating the risks

Continuous treatment is not always benign, and discontinuation is not always safe. Here’s how you and your patient can balance these risks:

1. Educate patients about the risks of stopping and restarting antidepressants. Tell them that restarting a drug might not always bring back the same benefit, and that repeated stop–start cycles may increase the risk of nonresponse (though we lack direct studies). Talk with them about possible indicators of withdrawal, such as dizziness, flu-like symptoms, insomnia, irritability, anxiety, sensory disturbances, or brief mood worsening, so they don’t mistake them for signs of relapse.
2. Taper slowly. Rapid discontinuation is consistently associated with worse outcomes. As a ballpark, reduce the dose by 10%–25% every 2–4 weeks, slowing further if symptoms emerge. Patients on agents with short half-lives often need the lower end of that range (eg, ~10% every 2–4 weeks), and the last 25% of the taper often needs the smallest steps, using liquids, scored tablets, or smaller doses that you can ask a pharmacy to compound.
3. Don’t chase every symptom spike with a full reinstatement, since repeated off–on cycles may themselves contribute to nonresponse (Majkic and Taylor, 2025). Instead, pause at the current dose or go back one step and slow the taper.
4. Don’t default to indefinite maintenance. Individualize treatment duration with a shared decision-making approach, weighing patients’ histories, severity, comorbidities, and psychosocial stressors. Patients with more prior episodes are at higher risk of relapse after discontinuation (Geddes JR et al, Lancet 2003;361(9358):653–661), but many can discontinue their medication safely. Even patients with recurrent depression can remain well after discontinuation with clinical support (Lewis G et al, NEJM 2021;385(14):1257–1267). Adjunctive psychotherapy (eg, CBT) and lifestyle interventions (eg, regular exercise, stable sleep routines, and reduced substance use) can improve the likelihood of successful antidepressant discontinuation.
5. Think of reinstating the medication as a fresh treatment trial. Most of what we know about nonresponse comes from studies where people restarted the same antidepressant. We don’t know whether switching medication classes after a failed reinstatement reliably works, but research in treatment-resistant depression offers a clue: In the STAR*D trial, about one in four patients who didn’t respond to an SSRI went on to achieve remission after switching antidepressants, whether they switched to another SSRI, an SNRI, or bupropion (Rush AJ et al, Am J Psychiatry 2006;163(11):1905–1917).

After Carol has spent six weeks back on escitalopram at a therapeutic dose with no meaningful improvement, you treat the restart as a failed trial and switch her to venlafaxine while adding psychotherapy to help prevent relapse. Carol’s depressive symptoms then begin to improve steadily.

CARLAT VERDICT: Restarting a previously effective medication does not always restore benefit, likely due in part to neuroadaptive changes. Symptoms that emerge after stopping a medication—especially after rapid discontinuation—may reflect withdrawal rather than true relapse. When restarting, don’t assume the prior response will return. When tapering, proceed gradually to minimize withdrawal and avoid misinterpreting symptoms as recurrence.