What Is Treatment-Resistant Depression? A Clinician’s Guide to TRD

Why this matters

Treatment-resistant depression sits at the intersection of every hard problem in psychiatry: definitional ambiguity, biological heterogeneity, limited tools, and high stakes. Patients carrying this label have already failed what was supposed to work, and they now face a confusing landscape of next-line options that range from solid evidence to early-stage experiment. This guide is built for clinicians who want a working framework that respects both the science and the practical realities of seeing these patients on a Tuesday afternoon. Sources are limited to literature published between 2021 and 2026.

How TRD is defined, and why that definition is still contested

The working definition used by both the FDA and the EMA is reasonably simple. TRD is a depressive episode that has not responded to at least two adequate antidepressant trials of different classes, given at therapeutic doses for an adequate duration, usually six weeks each ^[1,2]. Most research protocols treat “non-response” as less than a 50% reduction on a validated scale (HAM-D or MADRS), although the EMA does not specify a numerical threshold and instead relies on clinically meaningful improvement ^[2].

That apparent simplicity falls apart on contact with the literature. A 2026 systematic review identified more than 150 distinct TRD definitions in published trials, varying in how many failed trials are required, what counts as adequate dose and duration, and whether psychotherapy or neuromodulation belong in the failure count ^[2]. The original Thase-Rush staging framework, still widely cited, was published in 1997 and predates ketamine, esketamine, SAINT, and modern pharmacogenomics ^[1,2]. Definitional drift makes epidemiology fuzzy and clinical-trial eligibility inconsistent ^[1].

TRD vs. difficult-to-treat depression

A clinically useful distinction has been gaining traction: TRD versus difficult-to-treat depression (DTD). TRD is defined narrowly by failed pharmacological trials. DTD is broader. It captures patients whose depression keeps causing significant burden despite reasonable treatment, including those who can’t sustain remission, who function poorly even when scores look better on paper, or who can’t tolerate medications ^[1,2]. The DTD framework reorients goals away from chasing remission and toward optimizing function and quality of life, and it brings psychotherapy and social interventions in earlier rather than treating them as last resorts after pharmacology runs out ^[2]. For patients with multiple failed modalities, DTD is often a more accurate and more humane frame.

Staging: TRD is a spectrum

Several validated staging systems exist. None is perfect, and they don’t fully agree with each other. Maudsley and Thase-Rush classifications overlap only about 60–70% on what counts as “severe TRD” ^[2]. In practice, the choice of staging tool matters less than the discipline of applying one consistently.

For routine practice, MSM and MGH-S are the most informative because they account for trial optimization and chronicity rather than just counting failed prescriptions. Whichever model is used, staging only works when each prior trial is documented properly: drug, class, dose, duration, response. Without that audit, “TRD” risks meaning “we ran out of ideas,” which is a different problem entirely ^[1].

Rule out pseudoresistance before anything else

Roughly a third to two-thirds of apparent treatment failures are not biological resistance. They are pseudoresistance, meaning the patient was never adequately treated in the first place ^[2]. Skipping this step is the most common error in TRD care, and it sends patients down expensive treatment pathways they didn’t actually need.

The clinician-side contributors are familiar but persistent: under-dosing (especially in primary care), trials that ended at four weeks instead of six, sequential trials within the same mechanism (three SSRIs is not three adequate trials), missed bipolarity, and the absence of psychotherapy. The patient-side contributors include non-adherence, which runs 30–60% in MDD and is worth confirming with plasma levels when feasible; pharmacogenomic variability in CYP2D6, CYP2C19, and CYP2B6, which can produce roughly tenfold differences in drug exposure; comorbid anxiety, ADHD, PTSD, or substance use that has gone unaddressed; and medical contributors like hypothyroidism, sleep apnea, anemia, or chronic inflammation ^[2,9]. The 2023 CPIC guideline now explicitly recommends dose or drug adjustments for poor and ultrarapid metabolizers of SSRIs ^[9].

A clean pseudoresistance workup before escalation is not a bureaucratic checkbox. It is the difference between treating disease and treating an artifact of inadequate care.

Epidemiology and burden

Roughly 30% of patients with major depression do not respond adequately to standard antidepressants and meet some definition of TRD ^[1,3]. Apply the FDA/EMA criteria more strictly, as in STAR*D, and that number climbs toward 55% for failure of two sequential adequate trials ^[1]. By rough global estimates, more than 100 million people meet at least one TRD definition ^[1].

The cost burden is concentrated. A 2024 Markov modeling study using real-world data on 25,190 newly diagnosed patients showed that fewer than 20% develop TRD or significant comorbidity, but those patients account for 31–54% of all depression-related care costs. The pattern holds across the U.S., Japan, South Korea, and Hong Kong ^[4]. The greatest collective burden falls on middle-aged and older women, while older men with comorbid medical illness carry the highest individual cost burden per patient ^[4]. TRD is also strongly associated with elevated suicidal ideation, self-harm, and completed suicide, and the mortality gap is meaningful ^[1].

Neurobiology, and why it now drives treatment choice

The biology of TRD has stopped being academic background and started shaping which patient gets which treatment.

Beyond monoamines

Serotonin, norepinephrine, and dopamine remain part of the picture, but their dysregulation in TRD appears more entrenched and less responsive to standard agents than in non-resistant depression ^[3]. Sex differences are visible early. Women with MDD show higher central serotonin synthesis despite lower plasma tryptophan, with greater mood sensitivity to tryptophan depletion. That pattern is consistent with greater serotonergic vulnerability and differential SSRI response compared to men ^[3]. The relative weakness of monoaminergic agents in TRD is what motivated the search for non-monoaminergic mechanisms.

Glutamate and synaptic plasticity

The shift toward glutamate has been the most consequential change in the field. Dysregulated NMDA receptor signaling impairs synaptic plasticity, and reduced BDNF, particularly in prefrontal cortex and hippocampus, narrows the substrate that antidepressants need in order to work ^[3]. This is the mechanism that makes ketamine and esketamine work rapidly in patients who have exhausted monoaminergic options ^[1].

HPA axis dysregulation

Chronic HPA axis hyperactivity is a recurring finding in TRD, driving cortisol-mediated neurotoxicity and impaired neurogenesis ^[3]. There is now meaningful work suggesting that distinct cortisol trajectories preceding ECT predict response with roughly 80% accuracy, a candidate biomarker that deserves more clinical traction ^[3]. Sex differences are pronounced. Women show exaggerated ACTH and cortisol responses to acute stress and weaker glucocorticoid receptor-mediated negative feedback, prolonging HPA activation ^[3].

Neuroinflammation

Pro-inflammatory cytokines (IL-6, TNF-α, IL-1β) and microglial activation are consistently elevated in TRD ^[3]. This helps explain why patients with chronic inflammatory illness are over-represented in TRD populations and supports interest in anti-inflammatory augmentation, although the clinical-trial evidence remains thinner than the mechanistic case.

Mitochondrial and metabolic dysfunction

FGF21 and GDF15, both mitochondrial stress markers, distinguish TRD from treatment-responsive depression with high accuracy in recent work, and up to roughly three-quarters of metabolic disruption in TRD involves lipid pathways including sphingolipid and phospholipid metabolism ^[3]. Estrogen and progesterone modulate mitochondrial function, and their decline through perimenopause may be a real contributor to TRD risk in women ^[3].

Clinical assessment: a systematic approach

Step 1: Confirm the diagnosis

Re-run DSM-5 criteria for MDD and screen explicitly for bipolar spectrum (MDQ, HCL-32). Order TSH, free T4, CBC, CMP, fasting glucose, B12, folate, testosterone in men, and inflammatory markers (CRP, IL-6 where available). Screen for sleep apnea (STOP-BANG). Run AUDIT and DAST.

Step 2: Audit the medication history

For every prior antidepressant, document name, class, maximum dose reached, whether that dose was actually therapeutic, duration at therapeutic dose, reason for discontinuation, and degree of response (use MADRS or HAM-D retrospectively where you can). This audit is what separates real TRD from pseudoresistance, and it’s also what lets you stage the patient honestly ^[1,2].

Step 3: Severity and suicide risk

MADRS is preferred for TRD because it tracks change well and is the dominant outcome measure in modern TRD trials ^[1]. HAM-D-17 remains common in research and is what the GSRD anchors its <50% non-response definition to ^[1]. PHQ-9 is appropriate for screening and ongoing monitoring. C-SSRS belongs in every TRD evaluation, not just the first one.

Step 4: Pharmacogenomics

Pharmacogenomic testing (CYP2D6, CYP2C19, CYP2B6, SLC6A4, HTR2A) should be on the table for any patient with two or more antidepressant failures. CYP2D6 and CYP2C19 polymorphisms can shift exposure roughly tenfold, which makes “standard” dosing pharmacokinetically inappropriate for sizable patient subgroups. The 2023 CPIC guideline now formally recommends dose or drug adjustments for poor and ultrarapid metabolizers of SSRIs ^[9].

Evidence-based treatment strategies

Tier 1: Pharmacological augmentation

When two adequate trials have failed, augmentation, meaning adding a second agent rather than switching, is usually the first escalation. The current evidence base supports lithium, aripiprazole, brexpiprazole, quetiapine XR, cariprazine, olanzapine-fluoxetine, and T3 (liothyronine) ^[1].

Tier 2: Glutamatergic therapies (ketamine and esketamine)

Ketamine and esketamine were the most consequential pharmacological advance in TRD in two decades, and they remain the fastest-acting interventions available outside ECT. They restore synaptic connectivity and BDNF signaling through NMDA-receptor antagonism, which is why the response window is hours to days rather than weeks ^[1].

IV racemic ketamine. A typical regimen is 0.5 mg/kg infused over 40 minutes, repeated several times over two to three weeks. Response is often visible within 24 hours, including in suicidality. The major limitation is regulatory rather than clinical: ketamine is not FDA-approved for TRD, which creates real barriers around insurance coverage and protocol standardization ^[1].

Esketamine (SPRAVATO). In January 2025, the FDA approved SPRAVATO as the first monotherapy option for adults with TRD, expanding the previous indication that required concurrent oral antidepressant therapy ^[10]. The pivotal trial showed superior MADRS reduction versus placebo as early as 24 hours, with a 22.5% remission rate at week 4 versus 7.6% on placebo ^[10]. SPRAVATO is administered in a certified setting under REMS, with two hours of post-dose monitoring required ^[10].

Tier 3: Neuromodulation

rTMS. FDA-approved for TRD since 2008, rTMS targets the dorsolateral prefrontal cortex with repetitive magnetic pulses. A 2025 prospective study using an accelerated high-frequency protocol over six days reported a 52% response rate and 24% remission rate in 25 patients with treatment-resistant depression ^[5]. A 2024 randomized trial directly comparing rTMS to a medication switch in moderate TRD favored rTMS on response rate, remission rate, and HAM-D reduction over an eight-week course ^[8].

SAINT (Stanford Accelerated Intelligent Neuromodulation Therapy). SAINT combines fMRI-guided individualized DLPFC targeting based on functional connectivity to the subgenual anterior cingulate, an accelerated protocol of 10 iTBS sessions per day over five days, and a higher cumulative pulse load than standard rTMS. The 2020 open-label work and the 2022 sham-controlled double-blind RCT in the American Journal of Psychiatry reported MADRS reductions and remission rates substantially above standard rTMS, with effects maintained at one month ^[6,7]. A 2026 confirmatory study extended these findings and added EEG biomarker work ^[6].

ECT. Despite its history, ECT remains the most reliable acute treatment for severe TRD, particularly in patients with psychotic features, active suicidality, food refusal, or catatonia where speed is everything. Response rates in severe TRD populations consistently clear 70%, and continuation or maintenance ECT reduces rehospitalization risk in real-world data ^[1]. The trade-off is cognitive, primarily anterograde memory effects, and right unilateral ultra-brief pulse ECT meaningfully reduces that burden while preserving efficacy ^[1].

VNS. FDA-approved for chronic or recurrent TRD, vagus nerve stimulation delivers calibrated electrical pulses to the left vagus nerve via a surgically implanted device. It is generally considered after four or more failed trials. Long-term registry and open-label data show enduring benefit in patients who have not responded to anything else, though randomized data are thinner than for other modalities ^[1].

Tier 4: Emerging therapies

Psilocybin-assisted therapy. Psilocybin has produced meaningful antidepressant effects versus placebo in recent randomized trials, with effects detectable within roughly a week of dosing ^[1]. It is not yet FDA-approved for MDD or TRD. Multiple Phase II and III trials are ongoing, and the regulatory picture in 2026 remains in flux.

Neurosteroids. Brexanolone (allopregnanolone) and related neurosteroids act on GABA-A and NMDA receptors to restore inhibitory signaling and neuroplasticity ^[3]. The mechanistic case is strong, the rapid onset is real, and the differential utility in women during postpartum and perimenopausal windows is increasingly hard to ignore ^[3].

Deep brain stimulation. DBS targeting the subgenual anterior cingulate or ventral capsule/ventral striatum is reserved for the most refractory patients, typically after six or more failed trials, and remains a specialty-center procedure. Recent cost-effectiveness work suggests earlier consideration may be both clinically and economically more rational than indefinite pharmacological continuation in qualifying patients ^[1].

Precision psychiatry: where the field is going

TRD management is moving from sequential empiricism toward biomarker-guided treatment matching ^[1,3].

Machine-learning models that integrate multiple biomarker domains are starting to generate probabilistic treatment-matching outputs. The signal is real, but no model is yet ready to displace clinical judgment. The right framing is augmentation, not replacement ^[1].

Special populations

Adolescents and young adults

TRD prevalence in youth runs as high as 40% of MDD cases, driven by comorbidity (anxiety, ADHD, substance use) and an under-recognized rate of bipolar diathesis that needs to be ruled out before more antidepressant trials are added ^[1]. Suicidality risk is particularly elevated and warrants C-SSRS at every contact. Among neuromodulation options, SAINT and standard rTMS are generally preferred over ECT in adolescents because tolerability is better.

Older adults

Late-life TRD has its own biology: cerebrovascular disease, white matter hyperintensities, HPA hypercortisolemia. ECT is particularly effective here, with response and remission rates that exceed those in younger adults ^[1]. Cognitive monitoring is non-negotiable, and right unilateral ultra-brief pulse ECT is preferred. Older men with comorbid medical illness carry the highest individualized cost burden across the depression treatment course ^[4].

Women with hormonal vulnerabilities

Perimenopause, postpartum, and PMDD-related depression overlap biologically with TRD through estrogen-BDNF-HPA interactions ^[3]. Neurosteroid therapies may have differential utility in these populations, and clinical decision-making should incorporate hormonal status rather than treating it as a footnote. Women aged 41–65 carry the largest collective burden across the depression care pathway ^[4].

Conclusion

TRD is not a dead end. It is a signal that the diagnosis, the workup, or the treatment plan needs to be sharper. The clinicians who do best with these patients are the ones who take the pseudoresistance workup seriously, who stage carefully, who match treatment to mechanism rather than habit, and who keep moving when the obvious next step doesn’t work. The convergence of rapid-acting agents, sex-informed biology, and biomarker-guided matching is the most usable progress the field has had in a generation, and it is happening fast enough that staying current is now part of the standard of care.