What it is:Major depressive disorder (MDD) is one or more major depressive episodes with no history of mania or hypomania. That last clause is doing a lot of work: it's what makes this "unipolar" depression, and it's a diagnostic checkpoint you actually have to clear before reaching for an antidepressant, because giving one to someone with unrecognized bipolar depression can flip them into mania.
The core problem:Low synaptic serotonin, norepinephrine, and dopamine is the old explanation, but it can't be the whole story. Every antidepressant raises monoamine levels within hours, yet nobody feels better for 2 to 4 weeks. What actually seems to matter is what happens downstream over those weeks: increased BDNF (brain-derived neurotrophic factor), new synapse formation, and structural remodeling, especially in the hippocampus. The monoamine bump is the trigger, not the cure.
What you do about it:SSRI first line for almost everyone who's an uncomplicated, physically healthy outpatient. All antidepressants are roughly equal in efficacy at comparable doses, so you're choosing based on side effect profile, not superior efficacy. Give it a full 6-week trial at the maximum tolerated dose before calling it a failure.
The delayed-action paradoxis the single best organizing idea for this whole chapter. If raising serotonin fixed depression, people would feel better in hours, like a stimulant. They don't. The 2 to 6 week lag is the neuroplasticity cascade catching up, and it explains why ketamine and esketamine (which act on glutamate/NMDA signaling instead) can lift mood in hours to days instead of weeks. Same eventual downstream target, faster on-ramp.
MDD isn't one flavor. The DSM-5 lets you stack "specifiers" onto the core diagnosis, and several of them change your drug choice directly.
| Specifier / subtype | What it looks like | Why it matters clinically |
|---|---|---|
| Atypical features | Hyperphagia, hypersomnia, leaden paralysis, high anxiety, mood reactivity | The opposite polarity of "classic" depression; think of it before assuming everyone has insomnia and appetite loss |
| With psychotic features | Hallucinations or delusions layered on the depressive episode | Psychotherapy alone is not adequate; usually needs an antidepressant plus an antipsychotic, or ECT |
| With anxious distress | Tension, restlessness, fear something awful will happen | 50–70% comorbidity with anxiety; favors an SSRI/SNRI started at a lowdose to avoid early activation |
| Seasonal pattern | Episodes tied to time of year, classically fall/winter onset | Light therapy is a reasonable adjunct/alternative |
| Peripartum onset | Onset during pregnancy or within 4 weeks postpartum (clinically tracked out to ~1 year) | Affects 6–20% of postpartum patients; distinct from the much rarer, more dangerous postpartum psychosis; zuranolone/brexanolone are FDA-approved only here |
| Persistent depressive disorder (dysthymia) | Meets a lower symptom bar but for >2 years | Low-grade, chronic; easy to miss because nothing looks acute |
| Agitated depression | Fear, insomnia, extreme irritability and restlessness | An activating antidepressant (bupropion, an SNRI) can make this worse; consider something calmer |
"Depressed" and "bipolar depression" are not the same disease and do not get the same first-line drug. Rule out a history of manic or hypomanic episodes before starting monotherapy with an antidepressant. If bipolar disorder is on the table, that's a mood stabilizer conversation, not this chapter.
There isn't one accepted mechanism. There are several overlapping hypotheses, and understanding how they layer on each other explains both why so many different drug classes work and why they all take weeks.
| Hypothesis | Claim |
|---|---|
| Monoamine hypothesis | Reduced synaptic NE, 5-HT, and/or DA drives symptoms; the original rationale for every reuptake-blocking or degradation-blocking drug class |
| Receptor desensitization | Chronic drug exposure downregulates NE and 5-HT1A receptors; the timingof that downregulation lines up with clinical improvement far better than the timing of the initial neurotransmitter rise does |
| Dysregulation hypothesis | The problem is a broken homeostatic feedback loop for monoamine signaling, not simply "too little" transmitter |
| Inflammatory / neurotrophic hypothesis | Chronic stress and inflammation disturb glutamate and GABA transmission and suppress BDNF, driving loss of synaptic connections and neuronal atrophy, especially in the hippocampus |
Serotonin and norepinephrine rise within 2 to 3 hours of the first antidepressant dose. Mood doesn't move for weeks. That gap is the whole clue: the acute monoamine increase over weeksdrives a rise in BDNF, which drives neuronal sprouting and synaptic remodeling, concentrated in the hippocampus. Thatstructural change, not the acute chemistry, is thought to be what actually resolves the episode. ECT independently raises BDNF too, which is one reason it works despite a totally different mechanism up front.
Ketamine and esketamine block the NMDA glutamate receptor. That's a completely different starting point from monoamine reuptake, but it converges on the same endpoint: increased BDNF and synaptic remodeling, just reached in hours instead of weeks because glutamate signaling sits more directly upstream of that remodeling machinery. That's the whole reason esketamine can lift a suicidal patient's mood same-day when an SSRI cannot.
5-HT and NE pathways from the raphe nuclei and locus ceruleus overlap heavily in the frontal cortex and limbic system, and they cross-regulate each other. That overlap explains the symptom split you see clinically: serotonergicdysfunction tracks with anxiety, social withdrawal, and appetite/sleep disturbance, while noradrenergicdysfunction tracks with low energy, poor concentration, reduced alertness, and a lowered pain threshold. If a patient's chief complaints are fatigue and brain fog rather than anxiety and appetite change, that's a hint an SNRI's added NE activity might serve them better than a pure SSRI.
Depression shows up across four domains, and no single symptom is required except one: depressed mood or anhedonia has to be present. Everything else is a supporting cast.
| Domain | Findings |
|---|---|
| Emotional | Anhedonia (loss of pleasure), persistent sadness, hopelessness, guilt or worthlessness, crying, anxiety, psychotic features in severe cases, suicidal ideation with or without a plan |
| Physical | Weight gain orloss, fatigue, pain (headache especially), sleep disturbance in either direction, appetite change in either direction, decreased libido, GI and cardiovascular complaints (palpitations) |
| Cognitive | Poor concentration, impaired recent memory, indecisiveness, confusion |
| Psychomotor | Retardation (slowed movement, thought, speech) oragitation |
10 to 15% of patients with depression will attempt suicide at some point. Distinguish passiveideation (wishing you were dead) from activeideation (thoughts of acting on it), and always ask whether there's a specific plan, since a plan sharply raises risk. Asking directly does not plant the idea; not asking just means you miss it. Women attempt more often; men complete more often. All antidepressants carry a black box warning for increased suicidal thinking and behavior in patients under 25, worst in the first few weeks of treatment and again in the 30 days after any dose change or discontinuation.
Kids often present as boredom, irritability, somatic complaints (stomach aches), and slipping grades rather than "I feel sad." Older adultsoften minimize mood complaints entirely and instead present with anhedonia, fatigue, cognitive impairment, and sleeplessness that can look like early dementia or "just getting old." That's why anhedonia and energy/appetite/sleep changes deserve just as much weight as stated mood in a geriatric workup.
Diagnosis is clinical, built on DSM-5 criteria plus a workup to rule out mimics.
| Letter | Symptom |
|---|---|
| S | Sleep disturbance (insomnia or hypersomnia) |
| I | Interest lost (anhedonia) |
| G | Guilt or worthlessness, often excessive or inappropriate |
| E | Energy decreased |
| C | Concentration impaired, indecisiveness |
| A | Appetite change (usually decreased, ~5% weight change from baseline) |
| P | Psychomotor agitation or retardation |
| S | Suicidal ideation, plan, or attempt; recurrent thoughts of death |
Need ≥5 of these 9(add "D" for depressed mood as the ninth) present nearly every day over the same 2-week period, and at least one of the five must be depressed mood or anhedonia. Symptoms must cause real distress or functional impairment, and can't be better explained by a substance or a medical condition. No history of a manic or hypomanic episode (unless that episode itself was substance- or condition-induced).
| Tool | Who fills it out | Notes |
|---|---|---|
| PHQ-9 | Patient | 9 items mirror the DSM-5 criteria, 0–27 scale; the workhorse in primary care |
| GDS(Geriatric Depression Scale) | Patient | Built for older adults, avoids over-weighting somatic symptoms that overlap with normal aging |
| HAM-D | Clinician | Structured interview, heavily used in trials |
| MADRS | Patient + clinician | Sensitive to change, good for tracking treatment response |
| Beck Depression Inventory | Patient | Behavioral symptom self-report |
| Score | Severity | Action |
|---|---|---|
| 0–4 | Minimal | Treatment usually not indicated |
| 5–9 | Mild | Clinical judgment based on duration and functional impact |
| 10–14 | Moderate | Clinical judgment; leaning toward treatment |
| 15–19 | Moderately severe | Warrants treatment: antidepressant, psychotherapy, or both |
| 20–27 | Severe | Warrants treatment: antidepressant, psychotherapy, or both |
PHQ-9 question 9 asks about thoughts of being better off dead or self-harm. It's buried inside a routine 9-item screen, but any positive answer needs direct follow-up regardless of the total score. A patient can score "mild" overall and still flag positive on item 9.
Goals:resolve current symptoms (remission), prevent relapse and recurrence, and prevent suicide, ultimately returning the patient to their prior level of functioning.
For a physically healthy outpatient with no contraindication steering you toward a specific class: start an SSRI. All antidepressant classes are roughly equal in efficacy at comparable doses, so the SSRI's edge is tolerability and safety in overdose, not superior response rates.
| Term | Definition |
|---|---|
| Non-response | <25% symptom improvement |
| Partial response | 26–49% improvement |
| Response | >50% improvement |
| Remission | No symptoms remaining |
An adequate trialis at least 6 weeks at the maximum tolerated dose. Don't judge a drug a failure before you've hit that bar, and don't forget that improvement in sleep and appetite typically shows up before mood does, often within 1 to 2 weeks, with full antidepressant effect on mood taking 2 to 6 weeks and sometimes longer.
Both the STAR*D trial and the British Association of Psychopharmacology guidelines found that switching and augmentation strategies outperform simply raising the dosein poor responders. If a reasonable dose isn't working after an adequate trial, don't just keep pushing the same drug higher, change strategy.
| Phase | Duration | Goal |
|---|---|---|
| Acute | 6–12 weeks | Remission |
| Continuation | 4–9 months after remission | Eliminate residual symptoms, prevent relapse (return of the sameepisode) |
| Maintenance | 12–36 months, sometimes lifelong | Prevent recurrence (a whole new episode) |
Lifelong maintenance is favored for higher-risk patients: younger than 40 with ≥2 prior episodes, or any age with ≥3 prior episodes.
Start at roughly half the usual adult initial dose and titrate more slowly. It can take 6 to 12 weeks, not the usual 2 to 6, to see the full antidepressant response in this population, so don't declare failure on the same timeline you'd use for a 30-year-old.
This is the part of pharmacotherapy that actually separates a good clinician from someone who just picked a drug off a list. Roughly half to 60% of patients improve on their first antidepressant. What you do with the other half is where the real skill is.
Three variables to weigh every time: cross-taper vs stop-then-start vs washout, discontinuation syndrome risk, and drug interactions during the overlap window.
| Scenario | Practical approach |
|---|---|
| SSRI (not fluoxetine) → another SSRI | Stop first SSRI, start the new one at a low dose (citalopram/escitalopram/paroxetine 10 mg, sertraline 25 mg). Cross-taper is also acceptable in most cases. |
| SSRI (not fluoxetine) → venlafaxine or duloxetine | Stop SSRI, start SNRI low (venlafaxine 37.5–75 mg total daily, duloxetine 60 mg daily). Watch for CYP2D6/1A2 inhibition from the outgoing SSRI slowing SNRI metabolism during overlap. |
| SSRI (not fluoxetine) → mirtazapine | Cross-taper, or taper the SSRI to its minimum effective dose first, then switch to mirtazapine 15 mg at bedtime. |
| Fluoxetine→ anything | Its long half-life (4–6 days, active metabolite norfluoxetine 4–16 days) means it self-tapers. Stop it, then wait a 4–7 day washout before starting most new agents at a low dose; cross-tapering is generally notrecommended. |
| Any agent → fluvoxamine, or fluvoxamine → anything | Never cross-taper. Taper and fully stop the first drug, then start fluvoxamine (or the new agent) at a low dose. Fluvoxamine's CYP1A2/3A4 inhibition makes overlap risky. |
| Venlafaxine or duloxetine → mirtazapine | Taper and stop the SNRI, then start mirtazapine low, or cross-taper starting mirtazapine at 15 mg at bedtime. |
Any SSRI or TCA combined with an MAOI can cause a potentially fatal reaction. Going from fluoxetine to an MAOI needs a 5-week washoutbecause of its long active metabolite; other SSRIs/SNRIs typically need about 2 weeks. Going the other direction, off an MAOI onto a serotonergic drug, also needs roughly a 2-week washout to let MAO activity regenerate.
Highest risk switching froma serotonergic drug toa nonserotonergic one, and worst with short half-life agents like venlafaxineand paroxetine. Symptoms: "GI flu" (nausea, cramping), paresthesias, irritability, insomnia, dizziness, vivid dreams, and the classic electric "brain zaps." Fluoxetine and bupropion often don't need tapering at all because fluoxetine's long half-life tapers itself and bupropion has low discontinuation-syndrome risk. For everything else, taper over several weeks; a conservative approach for paroxetine or venlafaxine is roughly 25% dose reduction every 4 to 6 weeks.
| Class / Drug | Initial Dose | Usual Range / Target |
|---|---|---|
| SSRIs | ||
| Citalopram (Celexa) | 20 mg/day | 20–40 mg/day (max 40; 20 mg cap if >60 y/o or CYP2C19 poor metabolizer) |
| Escitalopram (Lexapro) | 5–10 mg/day | 10–20 mg/day (max 20) |
| Fluoxetine (Prozac) | 10–20 mg/day | 20–60 mg/day (max 80, 20 mg increments) |
| Fluvoxamine (Luvox) | 25–50 mg/day | 50–300 mg/day (max 300; split BID if >100 mg) |
| Paroxetine (Paxil) | 10–20 mg/day | 20–50 mg/day IR (max 50); CR max 62.5 mg/day |
| Sertraline (Zoloft) | 25–50 mg/day | 50–200 mg/day (max 200, 25 mg weekly increments) |
| SNRIs | ||
| Desvenlafaxine (Pristiq) | 50 mg/day | 100 mg/day (tolerability drops above 50 mg) |
| Duloxetine (Cymbalta) | 30 mg/day | 30–90 mg/day (max 120; no added efficacy above 60 mg for MDD) |
| Venlafaxine (Effexor) | 37.5–75 mg/day | 75–225 mg/day (max 375 IR / 225 ER) |
| Levomilnacipran (Fetzima) | 20 mg/day | 40–120 mg/day |
| Tricyclics (TCAs) | ||
| Amitriptyline (Elavil) | 25 mg/day | 100–200 mg/day (max 300); therapeutic level 100–250 ng/mL |
| Nortriptyline (Pamelor) | 25 mg/day | 50–150 mg/day (max 150); therapeutic level 50–150 ng/mL, best-established TCA window |
| Doxepin (Sinequan) | 25 mg/day | 100–200 mg/day (max 300) |
| Imipramine (Tofranil) | 25 mg/day | 100–200 mg/day (max 300) |
| Desipramine (Norpramin) | 25 mg/day | 100–200 mg/day (max 300); cardiac death risk, avoid in cardiac hx |
| NDRI | ||
| Bupropion (Wellbutrin) | 150 mg/day (75 mg BID) | 150–450 mg/day (max 450 IR/ER, 400 SR) |
| Mixed serotonergic | ||
| Trazodone (Desyrel) | 50 mg/day | 150–300 mg/day (max 600); 25–100 mg at bedtime for sleep only, off-label |
| Nefazodone (Serzone) | 100 mg/day | 200–400 mg/day (max 600); black box hepatotoxicity |
| Vilazodone (Viibryd) | 10 mg/day, titrate over 2 weeks | 20–40 mg/day (max 40, must take with food) |
| Vortioxetine (Trintellix) | 10 mg/day | 20 mg/day (max 20) |
| Serotonin/α2 antagonist | ||
| Mirtazapine (Remeron) | 15 mg at bedtime | 15–45 mg/day (max 45) |
| MAOIs | ||
| Phenelzine (Nardil) | 15 mg | 30–90 mg/day (max 90) |
| Tranylcypromine (Parnate) | 10 mg | 20–40 mg/day (max 60) |
| Selegiline transdermal (Emsam) | 6 mg/24h | 6–12 mg/24h (max 12; diet restrictions only apply above 6 mg) |
| SGA augmentation | ||
| Aripiprazole | 2 mg | 2–15 mg; CANMAT Level 1, 1st-line augmentation |
| Quetiapine | 50 mg | 150–300 mg; CANMAT Level 1, 1st-line augmentation |
| Brexpiprazole | 1 mg | 1–3 mg; CANMAT Level 1, 2nd-line |
| Alternative augmentation | ||
| Lithium | 300 mg | 600–1200 mg; target level 0.6–1 mEq/L |
| Triiodothyronine (Cytomel) | 0.025 mg/day | 0.025–0.05 mg/day; monitor free T3 |
| Buspirone | 10 mg | 10–60 mg, divided BID–TID |
SSRIs block presynaptic SERT, increasing synaptic serotonin. Their edge over TCAs isn't efficacy, it's tolerability: essentially no affinity for muscarinic, histaminergic, α1-adrenergic, or dopaminergic receptors, so no reliable anticholinergic effects, no postural hypotension, no cardiotoxicity, minimal sedation, and they're far safer in overdose.
Common adverse effects:nausea, headache, insomnia or activation, and sexual dysfunction (patients underreport this spontaneously, so ask directly). Most GI effects fade over days to weeks; slower titration helps.
Citalopram(and to a lesser extent escitalopram) prolongs QTc in a dose-dependent way. Citalopram is capped at 40 mg/day, and 20 mg/day for patients over 60, CYP2C19 poor metabolizers, or anyone on a CYP2C19 inhibitor. Discontinue if QTc is persistently >500 msec or rises >50 msec from baseline.
Picking an SSRI by patient:sertralinehas the best cardiovascular safety data (studied post-MI and in heart failure) and is a reasonable default in CV disease. Fluoxetineis the most activating, well-tolerated, has the most pediatric/adolescent data, but carries the highest hyponatremia signal in some reports and the most CYP interactions (2D6, 2C19, 3A4). Paroxetineis more sedating, has real anticholinergic activity that makes it a poor choice in older adults, and has the roughest discontinuation syndrome of the class due to a short half-life. Fluvoxamineis a strong 1A2/3A4 inhibitor and should never be cross-tapered with anything, always taper off before starting the next drug.
Rare but real risks:hyponatremia (mainly older women, via elevated vasopressin, can appear within 72 hours of starting), a roughly 1.5-fold increase in GI bleed risk (serotonin is involved in platelet aggregation, worse combined with NSAIDs or antiplatelets), and serotonin syndrome with any other serotonergic drug.
Same serotonin reuptake blockade as an SSRI, plus norepinephrine reuptake blockade. Venlafaxineis the clearest example of dose-dependent pharmacology in this chapter: primarily serotonergic at low doses, gaining meaningful NE activity at higher doses, and DA activity only at very high doses. That's why low-dose venlafaxine behaves clinically like an SSRI and higher-dose venlafaxine starts to look and feel like an SNRI.
Why reach for this class instead of an SSRI:neuropathic pain or fibromyalgia (duloxetine is first-line for chemotherapy-induced peripheral neuropathy specifically), psychomotor slowing/low energy where the added NE activity is activating, or comorbid pain syndromes layered on the depression.
SNRIs cause a dose-related rise in blood pressure, most notably diastolic BP with venlafaxine, that can require dose reduction or discontinuation if sustained hypertension develops. Check BP at baseline and with titration. Duloxetine's rare but real adverse effect is liver toxicity, transient or sustained, so it's worth a baseline sense of hepatic function in anyone with liver risk factors.
Nausea and adverse-effect-related discontinuation rates run a bit higher with venlafaxine and duloxetine than with SSRIs. Abrupt venlafaxine discontinuation risk includes cardiovascular effects, so it deserves the same taper respect as paroxetine.
TCAs block both NE and 5-HT reuptake ("old-timey SNRIs"), but they also bind adrenergic, muscarinic, and histaminergic receptors nonselectively, and that off-target binding is where all the trouble comes from: anticholinergic effects (dry mouth, blurred vision, constipation, urinary retention, confusion, tachycardia), sedation and weight gain (H1 blockade, worst with amitriptyline and doxepin), and orthostatic hypotension plus cardiac conduction delay (α1 blockade and direct cardiac effects).
This, more than efficacy, is why TCAs lost first-line status. Cardiac conduction effects can cause fatal arrhythmias in overdose, worsened by combination with alcohol, benzodiazepines, or barbiturates. Limit dispensed quantity or route through a family member in any patient with suicidal ideation. Desipraminespecifically carries an increased sudden cardiac death signal, especially with a family history of sudden cardiac death or conduction disease; get a baseline ECG before starting a TCA in children, adolescents, or anyone over 40.
Receptor selectivity within the class(Ki values, lower = tighter binding): amitriptyline is SERT-favoring (Ki ~4 nM SERT vs 20 nM NET); nortriptyline flips to NET-favoring (Ki ~2 nM NET vs 16 nM SERT). This doesn't change antidepressant efficacy, but it explains why nortriptyline tends to be a bit better tolerated: less serotonergic GI/sexual burden. Nortriptyline has the best-established therapeutic plasma window(50–150 ng/mL) of any TCA, useful when checking for non-adherence, toxicity, or a pharmacokinetic interaction.
Abrupt TCA withdrawal, especially at high doses, can cause a cholinergic rebound: dizziness, nausea, diarrhea, insomnia, restlessness. Taper.
MAOIs (phenelzine, tranylcypromine, isocarboxazid) block monoamine oxidase, letting NE, 5-HT, and DA accumulate in the presynaptic terminal. They're effective, arguably underused for treatment-resistant or atypical depression, but the interaction burden pushed them to a reserve role.
MAO normally clears dietary tyramine. Block MAO and tyramine builds up, triggers a surge of NE release, and can precipitate a hypertensive crisis (occipital headache, stiff neck, nausea, sweating, sharply elevated BP, rarely stroke). Avoid completely:aged cheese, dry-aged/cured meats, fermented soy (soy sauce, miso), sauerkraut, kim chee, tap beer, red wine. Fresh dairy, fresh meat, and spirits are fine. Selegiline transdermal at ≤6 mg/24h is the one MAOI where the diet restriction relaxes, since it has less peripheral MAO-A inhibition at that dose.
Washout rules (know these cold):stopping fluoxetine before starting an MAOI needs 5 weeksbecause of its long-lived active metabolite; other antidepressants generally need about 2 weeks off before an MAOI starts, and 2 weeks off an MAOI before starting most serotonergic drugs. Combining an MAOI with an SSRI or TCA risks a potentially fatal reaction, chiefly serotonin syndrome.
Most common adverse effect is postural hypotension, more with phenelzine than tranylcypromine. Phenelzine tends to be sedating; tranylcypromine can be stimulating enough that the last dose is given in early afternoon.
Bupropioninhibits NE and DA reuptake and blocks central nicotinic receptors, with essentially no serotonergic activity, which is exactly why it's the go-to when sexual dysfunction or weight gain from another agent is the problem. It's activating (useful for fatigue/hypersomnia-heavy presentations, off-label used for ADHD), roughly weight-neutral to weight-losing, and doubles as smoking cessation therapy (marketed as Zyban at similar dosing). The tradeoff is dose-dependent seizure risk, about 0.4% at the 450 mg/day ceiling dose, so stick to labeled max single and daily doses, and it's contraindicated in bulimia/anorexia nervosa and active seizure disorders.
Mirtazapineworks by blocking presynaptic α2 autoreceptors and heteroreceptors, boosting both NE and 5-HT release, plus strong H1 antihistamine activity. That H1 blockade is the whole clinical story: sedation and significant weight gain (>7% over baseline is common), which makes it a good fit for an underweight, insomnia-heavy, low-appetite patient and a poor fit for someone already struggling with weight or daytime sleepiness. No meaningful anticholinergic or sexual side effects, which is its niche when those are the limiting factor on another drug.
Trazodoneblocks 5-HT2 receptors and weakly inhibits SERT, plus strong H1 and α1 blockade, so at antidepressant doses (150–300 mg) it's quite sedating; at low doses (25–100 mg at bedtime) it's commonly used off-label purely as a sleep aid, especially in patients on an SSRI or bupropion whose insomnia needs a non-habit-forming fix.
Rare (about 1 in 6,000 male patients) but a true emergency. Counsel any male patient starting trazodone to seek immediate care for an erection lasting >4 hours; delay risks permanent impotence.
Vilazodoneand vortioxetineare SSRIs with added 5-HT1A partial agonist activity. Vilazodone must be taken with foodfor adequate absorption and needs its own titration schedule (10 mg → 20 mg → 40 mg over 2 weeks); it's marketed around fewer sexual side effects but isn't considered highly effective and is sometimes added to a standard SSRI rather than used alone. Vortioxetine is metabolized by CYP2D6 and may help cognitive symptoms of depression specifically, at the cost of dose-limiting nausea.
All four exist because waiting 4–6 weeks for an antidepressant to work is not an option for a patient in acute crisis, and all four work faster because they bypass the slow monoamine route.
Esketamine (Spravato)is the s-isomer of ketamine, an NMDA receptor antagonist (also a mu opioid agonist) FDA-approved for treatment-resistant depression, always in combination with an oral antidepressant. It's intranasal, administered only in a supervised clinic setting under REMS, with mandatory 2-hour post-dose observation. Effects can appear within hours and last about 5 days at twice-weekly dosing. It is notfor psychotic depression, it can worsen psychosis, and it carries real abuse potential from its dissociative effects.
Dextromethorphan/bupropion (Auvelity)pairs a weak NMDA antagonist/SERT-NET inhibitor with bupropion, which inhibits CYP2D6 and thereby slows dextromethorphan's metabolism, extending its action. Onset is around 2 weeks, faster than a standard antidepressant, slower than esketamine, with efficacy converging with bupropion alone by 6 weeks.
Zuranolone (Zurzuvae)and brexanolone (Zulresso)are neurosteroids, positive allosteric modulators of GABA-A, FDA-approved specifically for postpartum depression. Zuranolone is oral, effects within days; brexanolone requires a 60-hour continuous IV infusion and is extremely costly. Zuranolone carries a driving restriction (wait 12 hours post-dose) due to sedation/dizziness risk.
Ketamine and esketamine work by driving fast glutamatergic/BDNF signaling changes rather than the slow monoamine cascade, which is the mechanistic reason they act in hours to days instead of weeks.
| Population | Key points |
|---|---|
| Older adults | Start at ~half the adult dose, titrate slower; SSRIs are typically first choice, bupropion/venlafaxine/mirtazapine also well tolerated; hyponatremia risk is higher; expect up to 6–12 weeks for full response |
| Pediatric / adolescent | Only fluoxetine and escitalopramare FDA-approved under age 18; every antidepressant carries the black box suicidality warning in this group; avoid desipramine given sudden cardiac death reports, obtain a baseline ECG if any TCA is used |
| Pregnancy | Stopping antidepressants during pregnancy carries ~5x the relapse risk of continuing; SSRI use has been linked to low birth weight, respiratory distress, and congenital heart defects, but absolute risk is unclear and confounded by the depression itself; sertraline and fluoxetine have the most safety data; weigh risk of untreated depression seriously, it is not a "safe default" |
| Lactation | Only small amounts of most SSRIs pass into breast milk; fluoxetine transfers more due to its long half-life; generally considered compatible but without robust long-term outcome data |
Before calling a patient "resistant," work through the checklist: is the diagnosis actually correct, is there a psychotic component, was the dose and duration adequate, are adverse effects limiting the dose, is the patient actually adherent, was a stepwise approach used, was outcome properly measured, is there a coexisting medical/psychiatric condition, is something else interfering, and could pharmacogenomics be at play. Most "treatment-resistant" patients turn out to have received an inadequate trial, not a truly resistant illness.
After 6 to 8 weeks with only a partial response, options are dose optimization, augmentation, adding psychotherapy, or ECT. With no response at all, options are switching antidepressants or adding psychotherapy/ECT. This is also where esketamine earns its place, for genuinely treatment-resistant cases after standard options are exhausted.
| Parameter | When | Watching for |
|---|---|---|
| Suicidal ideation | Every visit, especially weeks 1–4 and the 30 days after any dose change or discontinuation | Black box period for all antidepressants, sharpest in patients <25 |
| Symptom severity (PHQ-9 or equivalent) | Baseline, then periodically through acute phase | Response vs partial response vs nonresponse, to decide whether to hold, optimize, or switch |
| Adherence | Every visit | Most "treatment failures" are inadequate trials; confirm the drug was actually taken at an adequate dose long enough |
| Sodium | Within 72 hours of SSRI/SNRI initiation if at-risk (older, female) | Hyponatremia from SIADH-like effect |
| Blood pressure | Baseline and with SNRI titration | Dose-related hypertension, especially venlafaxine |
| ECG | Baseline for any TCA in children/adolescents or patients >40; before high-dose citalopram in at-risk patients | QTc prolongation, conduction delay |
| Weight | Periodically, more often on mirtazapine or TCAs | Clinically significant gain |
| TCA plasma level | Steady state, ~1 week at constant dose, drawn 12h after last dose | Nortriptyline has the clearest therapeutic window (50–150 ng/mL); also useful for suspected non-adherence or toxicity |
| Liver function | Baseline and periodically on duloxetine or nefazodone | Hepatotoxicity, black box for nefazodone |
| Sexual function | Ask directly, don't wait for spontaneous report | Underreported, common reason for silent nonadherence |