For around a third of people who start an antidepressant, the drug does not deliver adequate relief. Rush and colleagues (2006), reporting from the landmark STAR*D trial, found that only about a third of participants achieved remission after a first medication step, and that cumulative remission rates across four sequential steps remained below 70 percent, leaving a substantial minority without adequate relief after exhausting standard options. [rush-2006-stard] The clinical term for this pattern is treatment-resistant depression, and understanding what it means in 2026, which next steps have the best evidence, and where newer agents like esketamine and psilocybin sit in the hierarchy is the subject of this article.
The question of what to do when a first antidepressant fails is one of the most consequential in clinical psychiatry. McIntyre and colleagues (2023), writing in World Psychiatry, estimated that treatment-resistant depression affects 30 to 35 percent of patients with major depressive disorder and is associated with disproportionate disability, healthcare costs, and suicide risk. [mcintyre-2023-trd] The evidence landscape in 2026 is richer than it was five years ago, with two large randomised trials, a long-term esketamine safety study, and a phase-two psilocybin trial all publishing since 2022.
What treatment-resistant depression means in 2026
The most widely used working definition requires failure of at least two antidepressant trials, each at an adequate dose for an adequate duration, within the same depressive episode. Sforzini and colleagues (2022), using a Delphi consensus process involving international experts, formalised this threshold for clinical trial purposes, though they noted ongoing debate about whether “two adequate trials” should require trials from two different mechanistic classes. [sforzini-2022-trd-delphi] Their consensus also flagged a practical problem: many patients counted as treatment-resistant have actually received underdosed or prematurely discontinued medication, meaning true pharmacological resistance may be less prevalent than headline figures suggest.
McIntyre and colleagues (2023) reviewed the broader picture, noting that comorbid anxiety, chronic pain, and sleep disturbance each independently reduce antidepressant response rates, and that these factors should be screened for and addressed before concluding that a patient is pharmacologically resistant. The 2026 clinical picture therefore begins with a diagnostic audit: confirm the diagnosis, confirm adequate trials, and address treatable comorbidities before escalating to complex augmentation strategies.
Augmentation strategies on the current evidence
Adding a second agent to an inadequate antidepressant, rather than switching entirely, is the approach with the deepest randomised evidence base for treatment-resistant depression. Wang and colleagues (2023), in a network meta-analysis of four atypical antipsychotics used as augmenters, found that aripiprazole, quetiapine, olanzapine, and risperidone all improved response rates over antidepressant monotherapy, with aripiprazole showing the most favourable balance of efficacy and tolerability in head-to-head comparisons. [wang-2023-antipsychotics-aug]
Lithium augmentation has an older but robust evidence base, and in 2025 received a major head-to-head update. Cleare and colleagues (2025), leading the LQD trial, a pragmatic open-label randomised controlled trial conducted across public health services, randomised 212 patients with treatment-resistant depression to quetiapine or lithium and followed them for 12 months. [cleare-2025-lqd] Quetiapine demonstrated superior clinical effectiveness over lithium on the primary outcome of depressive symptom severity, and was also found to be more cost-effective.
The LQD result does not make lithium obsolete. Lithium has well-established long-term safety data, and clinicians may still prefer it for patients with particular metabolic concerns or where quetiapine is poorly tolerated. The practical takeaway is that both classes remain evidenced options, with quetiapine supported as a first-choice augmenter by the most recent and largest direct comparison. The choice between them should take into account individual factors including metabolic health, monitoring burden, and patient preference.
Esketamine, ketamine, and the rapid-acting agents
Esketamine, the S-enantiomer of ketamine, acts on the NMDA receptor rather than monoamine pathways, which gives it a distinctive clinical profile: antidepressant effects can appear within hours rather than weeks. It is licensed in the US and EU as an intranasal spray (Spravato) specifically for treatment-resistant depression, administered in clinic under supervision.
Zaki and colleagues (2025), reporting from a multi-year long-term extension study (SUSTAIN-3) of esketamine maintenance, followed participants over a mean of approximately three and a half years. [zaki-2025-esketamine-sustain3] They found that improvement in depressive symptoms achieved during induction generally persisted among patients who remained on maintenance treatment, with roughly half of participants meeting remission criteria at the longer assessment points. The study found no new safety signals beyond those already described in the short-term trials: transient dissociation, dizziness, and blood pressure elevation in the two hours post-dose were the most common adverse events, and all resolved within the clinic-supervised monitoring window.
The remaining question for esketamine is what happens after patients stop. Relapse rates on discontinuation remain high, and the optimal maintenance duration is not yet established. Jobnah and colleagues (2024), examining real-world ketamine use across a clinical cohort with chronic treatment-resistant depression, found that nearly all patients who had responded to ketamine relapsed within about a month of stopping, with post-relapse scores returning to pre-treatment baseline levels. [jobnah-2024-ketamine-trd] The data are a useful caution: esketamine is not a one-time reset but an ongoing maintenance strategy for most patients, with the cost and clinic-attendance burden that implies.
Psilocybin and the experimental pipeline
Psilocybin occupies a different position in the treatment landscape. Unlike esketamine, it does not yet have a regulatory approval for depression in most jurisdictions, and the evidence base is still building. The headline study is Goodwin and colleagues (2022), published in the New England Journal of Medicine, which randomised 233 patients with treatment-resistant depression to a single dose of psilocybin at 25mg, 10mg, or 1mg (active placebo). [goodwin-2022-psilocybin] The 25mg group showed significantly greater reductions on the MADRS depression scale at three weeks compared with the 1mg group, and the effect was maintained at 12 weeks in a proportion of responders.
The study had important limitations: it was unblinded (participants and therapists could generally identify who received the high dose), the trial was conducted with extensive psychological support that is resource-intensive and not currently available in standard clinical settings, and serious adverse events, including one case of suicidal ideation, occurred in the 25mg group. The authors and subsequent commentators have emphasised that psilocybin therapy as studied is not a pill-and-go intervention; the therapeutic container, including preparation, supervised dosing sessions, and integration, appears to be part of the treatment effect.
In 2026, psilocybin remains accessible in practice only through approved research trials, specialist compassionate-use pathways in a small number of countries, or decriminalised contexts where supervised administration may not be available. The pipeline includes several phase-three trials underway, but regulatory approval for routine clinical use is unlikely before the late 2020s in most health systems.
A schematic of the options described in this article. Not measured response rates.
A practical pathway when first-line treatment fails
For someone whose first antidepressant has not worked, the clinical recommendation framework described by McIntyre and colleagues (2023) provides a useful scaffold. The starting point is not immediately adding a second drug but auditing the first: was the dose adequate, was it taken for long enough, and were there adherence issues or drug interactions that reduced exposure? A significant minority of apparent non-responders are actually sub-therapeutic responders who improve with dose optimisation alone. A dated record of your own symptom scores makes that audit faster than recall does; the free PHQ-9 screener is the standard self-report severity measure clinicians use for exactly this purpose.
If the audit confirms a genuine adequate trial, the next decision is switch versus augment. Switching to an antidepressant from a different class, for example from an SSRI to an SNRI or mirtazapine, makes sense when the first drug produced no response at all; our guide to the newer antidepressants covers what has actually changed in that list recently. Augmentation is typically preferred when there has been a partial response: something is working, but not enough.
For augmentation, the LQD trial and the atypical antipsychotic network meta-analysis both support atypical antipsychotics or lithium as first-line additions. If these also fail, esketamine becomes a well-evidenced option, particularly in settings where clinic-based supervised administration is feasible. Psychological therapy, including CBT adapted for treatment-resistant presentations, should run alongside pharmacological steps rather than being deferred until medication is settled, and what actually helps day to day does not pause while the prescribing is being worked out. Evidence consistently shows combination approaches outperform either alone for this population.
Psilocybin sits at the far end of this pathway: appropriate for consideration within a clinical trial or specialist programme after standard and second-line options have been adequately trialled, not as an early escalation. For most people navigating treatment-resistant depression, the evidence in 2026 points toward careful sequencing of well-established agents before moving to experimental pathways.
When to seek help
Speak to a doctor if two antidepressants, each taken at an adequate dose for an adequate length of time, have not worked. That is the actual threshold the term treatment-resistant describes, and it is worth stating precisely, because the commonest reason a course is judged a failure is that it was stopped at three weeks or never reached a therapeutic dose. Take the specifics: which drugs, what dose, for how long, and what stopped each one.
Ask what comes next by name rather than asking whether anything else exists. Augmentation with lithium or an antipsychotic, a switch to a different class, and adding psychological therapy alongside are the established next steps, and a referral to a specialist mood service is reasonable to request at this point rather than after a third failed trial.
Go sooner if you have stopped being able to work or care for yourself, if you are drinking more, or if hope has narrowed to the point where the next option feels pointless before you have tried it. That last one is a symptom rather than an assessment.
Go urgently if you have thoughts of harming yourself or of being better off dead. Contact your local emergency services or a crisis helpline.