By the time most patients come into my office and use the words "nothing is working," they've been trying for years. Two SSRIs. Maybe an SNRI. A round of bupropion. Therapy that helped at first and then plateaued. A small dose of an atypical antipsychotic that someone added on. Sleep that never fully came back. A version of themselves they remember being — energetic, curious, recognizable to the people who love them — that now feels far away.

If that's where you are, the first thing worth knowing is that you are not unusual, and you are not at the end of the road.

The largest real-world study of antidepressant treatment ever conducted — the STAR*D trial — found that only about 30% of patients reach remission on their first antidepressant. After a second medication trial, the cumulative remission rate climbs to roughly half. By the end of four sequential treatment steps, about 67% of patients achieve remission. In other words, most people who don't respond to their first or second medication still get well — but only by changing the question, not just the prescription.

That last part is what this article is about. When antidepressants stop working, the move that helps is rarely "try another antidepressant of the same class." It's a structured re-examination of what is actually being treated, what's been optimized, and what mechanisms haven't been tried yet. What follows is the rough decision tree I walk through with patients in this situation.

Step 1 —Before changing the treatment, change the question

The single most common reason a depression treatment isn't working is that the depression being treated isn't quite what it appears to be. This isn't a criticism of any previous clinician — depression is a category that contains many neurobiologically different conditions, and the surface picture often looks the same.

Before reaching for a new prescription or a new modality, four things are worth re-examining carefully.

Bipolar spectrum features. 

Bipolar depression — even the softer "bipolar II" variant where the elevated states are subtle and often not noticed — typically does not respond to standard SSRIs, and sometimes worsens with them. A careful history of past energy, sleep, productivity, irritability, and risk-taking (not just mood) is essential. If the answer turns out to be a bipolar spectrum condition, the entire treatment strategy changes — mood stabilizers and specific atypical antipsychotics become the foundation, and unopposed antidepressants are de-emphasized.

Trauma masquerading as depression. 

Complex trauma and PTSD often present clinically as depression — flat mood, low motivation, sleep disturbance, anhedonia — but the underlying mechanism is different, and standard antidepressants will only ever provide partial benefit. When trauma is the engine, the treatment that actually moves the needle is trauma-focused psychotherapy: EMDR, Internal Family Systems, sensorimotor approaches. Medication has a supporting role, not the lead.

Adult ADHD. 

A surprising proportion of patients diagnosed with treatment-resistant depression have unrecognized adult ADHD underneath. The picture is recognizable: lifelong difficulty starting and finishing things, chronic low-grade self-criticism around productivity, jobs and projects abandoned not for lack of intelligence but for lack of follow-through, a mood that lifts noticeably when something genuinely interesting is in front of you. When ADHD is the primary driver, treating that often resolves what looked like depression.

Medical contributors. 

Untreated thyroid dysfunction, low B12 or vitamin D, sleep apnea (a remarkably common and often-missed cause of treatment-resistant depression in middle age), chronic inflammation, undiagnosed diabetes or anemia — any of these can blunt response to antidepressants. A basic medical workup catches a meaningful number of these. They are some of the most fixable contributors to a treatment-resistant picture.If any of these four turns up, the right next step isn't a different antidepressant — it's a different framing of the problem. That alone resolves a significant fraction of cases that look "treatment-resistant" on paper.

Step 2 — Optimise what's already in place

If the diagnosis still fits and the depression is genuinely unipolar, the next question is whether the current treatment has been optimised before moving to a different category of intervention. There's a meaningful difference between "the antidepressant didn't work" and "the antidepressant didn't work at a fully therapeutic dose for an adequate duration with an evidence-based augmentation strategy added on."

Three classes of augmentation have the strongest evidence in treatment-resistant unipolar depression.

Lithium augmentation. Old, cheap, and still one of the best-studied augmenting agents we have. At sub-therapeutic levels relative to bipolar dosing, lithium added to an existing antidepressant produces a meaningful response in a substantial minority of non-responders — and it's the augmentation strategy with the strongest evidence for reduced suicidality alongside the antidepressant benefit.

Thyroid hormone augmentation (T3). Low-dose liothyronine added to an SSRI has decades of evidence for converting partial responders into responders, even in patients whose thyroid lab values are normal. It's underused.

Atypical antipsychotic augmentation. Low-dose aripiprazole, quetiapine, or olanzapine added to an antidepressant — well-established for treatment-resistant depression and a regulator-approved strategy. The trade-off is the side-effect profile, which has to be discussed honestly, but the response data is real.

Antidepressant combinations — for example, an SSRI plus bupropion, or an SSRI plus mirtazapine — also fall into this category. The right augmentation depends on the symptom pattern (sleep, weight, agitation, sexual side effects), prior trials, and what the patient can tolerate. The point is that augmentation is its own legitimate step, separate from "try a different SSRI" — which is rarely the move that breaks the deadlock.

Step 3 — Mechanism-based escalation

If the diagnosis has been reconfirmed and the current treatment has been honestly optimized, and the depression remains stuck, the most important conceptual shift comes next.

Different antidepressant treatments work through different brain systems, and failing one does not reliably predict failing another. SSRIs and SNRIs act on monoamines — serotonin, norepinephrine. Ketamine acts on the glutamate system, a different neurotransmitter network entirely. TMS acts on cortical circuits through focused magnetic pulses, bypassing the pharmacological route altogether. ECT acts more broadly on global brain activity. Each is a different door into the same room.

This is why moving from one SSRI to another SSRI often disappoints, while moving from an SSRI to a mechanistically different intervention often doesn't.

In rough clinical order, here is how I think about the mechanism-based steps available after augmentation has been tried:

TMS — Transcranial Magnetic Stimulation. TMS uses focused magnetic pulses to stimulate the left dorsolateral prefrontal cortex (or, in some protocols, alternative targets) and modulate the underlying mood-regulating circuits. Published response rates in treatment-resistant depression run 50–60%, with remission rates around 30–40%. Because the mechanism is entirely outside the monoamine system, a meaningful proportion of patients who haven't responded to antidepressants do respond to TMS. There are no systemic side effects in the way medications cause them — no weight changes, no sexual side effects, no metabolic effects. It is, for many patients, the next correct step after pharmacology has plateaued. I've written a longer plain-English breakdown of the evidence and outcomes in this companion article.

One thing worth knowing in advance: response to TMS is not always linear. Some patients feel temporarily worse in the second or third week before turning the corner. That pattern has a name — the TMS dip — and I've described what it is and why it happens here. It catches people off guard, and knowing about it in advance changes whether they stick with the course long enough to benefit.

Ketamine. Ketamine acts on the NMDA glutamate receptor, increases BDNF (a key protein in synaptic plasticity), and tends to produce response within hours to days rather than weeks. In treatment-resistant depression, multi-session protocols produce response rates in the 50–70% range, with remission around 30–40%. The evidence base for reducing acute suicidality is the strongest of any antidepressant treatment we have. It is particularly useful when the depression is severe, when suicidality is part of the picture, or when speed of response matters. I'll be covering the broader evidence for ketamine in detail in a future article in this series.

ECT — Electroconvulsive Therapy. Modern ECT, performed under brief anesthesia with refined electrode placements, is meaningfully gentler than the historical version and has substantially better cognitive outcomes than its reputation suggests. For severe, treatment-resistant depression — especially with active suicidality, catatonia, psychotic features, or in older patients with limited tolerance for medication side effects — ECT remains the most powerful antidepressant treatment we have. It is heavier, requires anesthesia, and is appropriately reserved for situations where its strength is needed. But when it is the right tool, very little else compares.

MAOIs. A class of antidepressants that has been largely forgotten in routine practice, but for atypical depression and certain treatment-resistant pictures they can succeed where SSRIs and SNRIs have not. The dietary restrictions are real but manageable for motivated patients, and the transdermal patch formulation removes most of them.

The order is not absolute — patient preference, severity, prior trials, and clinical context all shape what comes next. But the underlying principle is constant: when you've failed one mechanism, move to a different one.

Step 4 — The psychotherapy that medication cannot replace

If trauma, attachment patterns, or long-standing relational difficulty are driving the depression, no medication and no neuromodulation will fully resolve that on its own. They can lift the floor — sometimes meaningfully — but the conditions that created the depression in the first place will keep generating it.

This is where trauma-focused therapies — EMDR, Internal Family Systems, sensorimotor psychotherapy, well-conducted attachment-based work — do what medication structurally cannot. The aim is not symptom suppression; it's repair of the underlying nervous-system patterns that are doing the suppressing. For patients whose depression has a strong relational or developmental component, this is often the decisive layer of treatment — and skipping it is a common reason that even successful medication trials feel partial.

Step 5 — Emerging options worth knowing about

Beyond established treatments, several emerging interventions are worth being aware of, even if access is limited:

Psilocybin-assisted therapy — currently in trial settings in many countries, with promising response data in treatment-resistant depression. Not yet broadly available outside research contexts.

Vagus nerve stimulation (VNS) — approved in some jurisdictions for treatment-resistant depression. Modest response rates, but with a sustained-effect profile that some other treatments don't have.

Deep brain stimulation (DBS) — investigational, reserved for the most severe and refractory cases.

Accelerated TMS protocols — including the SAINT/SNT protocol that delivers a full course of TMS in five days rather than six weeks. Early data is genuinely encouraging.These aren't first-line. But for a patient who has worked through everything above and is still searching, knowing they exist matters.

The honest summary

If you've been trying antidepressants for years and they haven't worked, the problem is rarely that depression is untreatable. The problem is more often that one of the following is true:

The diagnosis underneath the depression is something else — bipolar, trauma, ADHD, or a medical contributor.

The current treatment hasn't been fully optimized with augmentation.The treatments tried so far have all been from one mechanism — usually monoamines — and a different mechanism hasn't been tested yet.

The psychotherapy needed to address what is actually generating the depression hasn't been part of the picture.

The remission rates from STAR*D — about two-thirds of patients getting well by the fourth structured step — exist precisely because clinical care that escalates through these layers, instead of cycling through similar medications, eventually works for most people.

Failure of multiple antidepressants is not a verdict. It's a sorting problem. The next move depends on which brain systems and life factors are actually driving your depression — a question best answered through a structured re-evaluation rather than a generic next-step recommendation.

If you're at this point in your own treatment, the most useful thing you can do is bring a clinician this list of layers and ask, honestly, which of these have we actually worked through, and which haven't we? That single conversation often reveals where the next door is.