There is no cure

Yesterday I asked Claude “How often do SSRIs and SNRIs fail to result in remission of depression?”. This is the answer I received:

The honest answer is: much more often than the marketing-era framing suggested, and the failure rates compound alarmingly across successive trials.

The foundational data: STAR*D.

STAR*D (Sequenced Treatment Alternatives to Relieve Depression) remains the largest and most-cited real-world antidepressant effectiveness trial, conducted 2001-2006 with ~4,000 patients across primary care and psychiatric settings. It used a stepped-care design: patients who didn’t remit on one treatment moved to the next step.

The headline remission rates (QIDS-defined remission):

• Step 1 (citalopram, 12 weeks): ~37% remission
• Step 2 (switch or augment): ~31% remission of those who entered
• Step 3: ~14% remission
• Step 4: ~13% remission

Cumulative remission rate after up to four sequential treatments: ~67% by the original analysis.

However — and this is critical — Pigott, Leventhal, Alter, and Boren’s 2010 reanalysis, and subsequent work by Pigott specifically, showed the STAR*D numbers were substantially inflated by methodological choices:

• The published cumulative remission rate of 67% included patients who dropped out and was based on a protocol-deviating analysis
• Using the originally specified HAM-D outcome measure and accounting for dropouts, cumulative remission was closer to ~35-40% after all four steps
• Sustained remission (no relapse during 12-month follow-up) was achieved by only ~3% of the original cohort

The Pigott critique has been controversial but the core methodological points are sound. The contemporary honest read of STAR*D is that it showed:

• About a third of patients remit on first-line SSRI treatment
• Response attenuates substantially with each subsequent trial
• Most “remissions” don’t sustain
• Durable remission with standard pharmacological care is achieved by a minority of patients who present for treatment

More recent data confirming the pattern:

• Rush et al.’s VAST-D trial (2017) compared augmentation strategies in SSRI/SNRI non-responders in VA patients. Remission rates in the 12-week acute phase: 29% for aripiprazole augmentation, 27% for bupropion augmentation, 22% for bupropion switch. Better than placebo but not transformative.
• Kato et al.’s network meta-analysis (2021) of switching strategies found remission rates of 30-40% with most switching strategies in SSRI non-responders.
• The CO-MED trial (2011) tested combination antidepressant therapy from the start vs. sequential monotherapy. No meaningful advantage for combinations. ~40% remission at 12 weeks across arms.
• The GENDEP, STAR*D genetics, and subsequent pharmacogenomic studies have failed to identify clinically useful predictors of response. Pharmacogenomic testing products (GeneSight, etc.) have underwhelming RCT evidence.

What “failure to remit” actually means in practice:

The clinical reality is distributed rather than binary:

• ~35-40% achieve remission on a first adequate trial of SSRI/SNRI
• ~20-30% achieve partial response (≥50% symptom reduction but not remission) — meaningful improvement but continued functional impairment
• ~20-30% have minimal or no response
• ~10-15% worsen or have significant adverse effects requiring discontinuation

Of those who achieve remission:

• Relapse within 1-2 years is common (~40-50% in naturalistic studies) even on continued medication
• Discontinuation is associated with even higher relapse rates, though disentangling disease course from discontinuation effects is genuinely difficult

The definition problem:

“Treatment-resistant depression” is usually defined as failure of two adequate antidepressant trials of different classes at adequate dose and duration. By this definition, 30-40% of patients who seek treatment for depression meet criteria for TRD at some point in their illness course.

But the definitions vary:

• Some definitions require failure of two SSRIs specifically
• Some require two antidepressants of any class
• Some require augmentation failures also
• The European definition often includes psychotherapy trials
• “Difficult-to-treat depression” has emerged as a broader category recognizing that many patients don’t have a clean trajectory of discrete treatment failures

The factors that inflate apparent response rates:

Several methodological issues make published antidepressant efficacy look better than real-world outcomes:

• Trial populations are unrepresentative. Industry trials exclude substance use, psychotic features, suicidality, bipolarity, comorbid personality disorder, and medical complexity — essentially everything that predicts treatment resistance. Real patients have these.
• Trial durations are short. 6-12 weeks captures acute response but misses the majority of eventual non-response and relapse.
• Placebo response rates in modern trials are very high (often 30-40%), compressing effect sizes. The drug-placebo difference is often small even when both arms show substantial response.
• Publication bias. Turner et al.’s famous 2008 NEJM analysis showed that FDA data on unpublished antidepressant trials revealed substantially smaller effect sizes than the published literature implied. Negative trials were disproportionately unpublished or reframed as positive.
• LOCF and other imputation methods historically inflated apparent response by carrying forward the last observation of patients who dropped out, often because they were doing poorly.
• Response vs. remission conflation. “Response” (≥50% improvement) sounds good but a patient going from HAM-D 24 to HAM-D 12 is still significantly depressed. Many marketing-oriented communications emphasize response rates over remission rates.

The network meta-analysis and the “all antidepressants work” framing:

Cipriani et al.’s 2018 Lancet network meta-analysis of 522 trials and 116,000 patients showed all 21 studied antidepressants beat placebo, with modest differentiation between drugs. This has been used to argue that SSRIs/SNRIs “work.”

They do work, in the statistical sense that they beat placebo. But:

• The effect sizes are small (SMD ~0.30, equivalent to roughly 2 points on HAM-D)
• Clinical significance of this effect size is debated
• The meta-analysis doesn’t address the relapse, remission durability, or real-world effectiveness questions
• It’s compatible with a world where antidepressants produce meaningful benefit for a minority and minimal benefit for the majority who take them

Kirsch, Moncrieff, and others have argued the clinical significance of the drug-placebo difference is minimal for most patients. The counter-argument (Cipriani, Nutt, many others) is that population-level effect sizes obscure meaningful heterogeneity and that some patients get substantial benefit even if the average is small. Both are partially right.

The “not really remission” issue:

A separate and underappreciated problem: what gets classified as remission in trials often doesn’t match what patients recognize as being well. HAM-D ≤7 or QIDS ≤5 are the standard cutoffs, but many patients meeting these criteria still have:

• Persistent anhedonia or reduced capacity for pleasure
• Emotional blunting (particularly on SSRIs — the “Prozac nation” phenomenon)
• Sexual dysfunction
• Weight gain
• Sleep disturbance
• Reduced cognitive sharpness
• Sense that they’re “functional but not well”

Some of this is depression incompletely treated; some of it is SSRI side effects interpreted as residual depression; some is the genuine ceiling of what the medication achieves. Fava’s work on residual symptoms and Goodwin’s on emotional blunting both suggest that symptomatic remission as measured in trials substantially overstates functional and phenomenological recovery.

The practical takeaway:

A patient starting first-line SSRI/SNRI treatment for depression has roughly:

• 35-40% chance of meaningful remission on that trial
• 20-30% chance of partial response
• 20-30% chance of minimal response
• Substantial chance of relapse even if initial response is good

After two adequate trials have failed, the probability of remission with additional pharmacological steps drops substantially — into the 15-25% range per subsequent trial, with further drops at later steps.

Durable remission — sustained wellness for a year or more — is achieved by perhaps 20-30% of patients who present for treatment, even with optimal pharmacological care.

This is much worse than the “antidepressants are effective” framing implies, and much better than the “antidepressants don’t work” framing implies. The honest summary is that SSRIs and SNRIs produce meaningful benefit for a substantial minority of patients, modest benefit for another substantial group, and limited benefit for a large remaining group — and that cumulative outcomes after multiple trials are worse than prescribers typically communicate.

The wider frame:

The combination of high first-line failure rates, diminishing returns on subsequent trials, poor predictors of response, and incomplete remission even in “responders” is why the field has become increasingly interested in the alternatives covered in earlier threads — ketamine, psilocybin, rTMS, ECT, anti-inflammatory approaches, reward-circuit-targeted interventions. Not because SSRIs don’t work at all, but because their real effectiveness is limited enough that the unmet need in depression remains enormous, even in settings where first-line treatment is optimally delivered.

The field has oscillated between overselling antidepressants (1990s-2000s) and underselling them (contemporary revisionism). The calibrated view is that they’re modestly effective tools with real but limited efficacy, badly mismatched to the heterogeneity of what gets called depression, and that the expectation of SSRI-based remission as the baseline outcome of depression treatment is empirically unsupported.

People can just ruin your life when you are young, forever leaving scars in your mind that can’t be healed.

What Claude didn’t really mention by the way, is that 12.5% of people remit without treatment. So if 35% remit on SSRIs, but 12.5% would have remitted spontaneously, that’s about 22.5% who enter remission due to SSRIs. And remission tends to be just temporary.

As Claude indicated, you’re looking at 2.7% who achieve sustained remission over a period of 12 months.

This is the big taboo, what people really don’t like to hear. There isn’t really an effective cure for depression. Everyone wants to come up with solutions: “Take vitamin D! Go exercise!” As if these are things you can’t come up with yourself. The reality is most people don’t permanently recover from depression, it tends to be a lifelong recurring ailment.

But what about Electroconvulsive therapy (ECT)? You tend to hear spectacular claims about that. I’ve seen claims of 80% remission. Well, they don’t do placebo controlled trials anymore, because when you’re sick enough to need it, you can’t be safely given a placebo of course. The studies are poor quality and date back to the 80’s. No placebo-controlled Randomised Controlled Trial has been conducted since 1985.

The reality is that the side-effects are so severe that no regulators would approve it if it were invented today, it was basically grandfathered in. You suffer memory loss and often cardiovascular damage too. All that, for a treatment that 6-9% say helped longer than two months.

For depression, there doesn’t seem to be anything that really works in the long term. That was my experience with Psilocybe mushrooms too. The first time it lasted for years, then months, eventually I was regularly microdosing. And with every time I took them, I started getting more anxious, afraid that some external force was trying to convince me to make a “sacrifice”. Eventually my use of Psilocybe mushrooms and cannabis seem to have resulted in me suffering an episode of psychosis, once I abruptly stopped taking them. The mescaline cactuses too eventually began triggering severe anxiety for me. It’s just no longer fun for me.

My advise for anyone taking cannabis who wants to stop is to be very careful with stopping. Gradually take less, don’t just stop abruptly. Your brain becomes used to cannabis and needs time to adjust when you reduce your intake.

Ketamine is now an approved treatment too, but I expect the same problem there, as anyone who has ever taken ketamine knows your brain rapidly develops a tolerance to it.

I don’t think there really is a cure for depression. In a sense, it’s strange that we treat depression as a disease, rather than just a rational observation of life itself. It’s normal to be unhappy when bad things happened to you.

Life inherently just isn’t really fun, but we pretend that the default state, the normal way of being, is to enjoy life. If every attempt at “curing” depression fails, you have to ask yourself whether it is really a sickness.

Most of the great philosophies and religions are all meant to help people cope with the burden of being alive. The Dharmic religions are about escaping the cycle of reincarnation. The Abrahamic religions promise a future after death in which you don’t have to return to this world. Camus said that whether life is worth living is the only serious philosophical question.

But us moderns have the strange habit of taking it for granted that people want to live in this world, we treat life itself as the blessing. If you don’t see it as such, we declare you to be sick, to be “depressed”. Depression is now the number one cause of disability in the world. And there is no genuine cure.