Let's Talk About Non-Responders in Psychedelic Therapy

For a decade, many headlines, documentaries, and investor pitch decks have sold a singular story: that compounds like psilocybin and MDMA are capable of resetting the depressed brain and resolving trauma with unprecedented speed. 

But on August 9, 2024, that narrative collided with regulatory reality.

The FDA issued a Complete Response Letter rejecting Lykos Therapeutics' application for MDMA-assisted therapy for PTSD. This was not a mere bureaucratic delay. The letter gave a forensic dismantling of the field’s “breakthrough” data. The FDA’s Advisory Committee, which voted 9-to-2 against efficacy, cited fatal flaws that critics had whispered about for years: “functional unblinding” that invalidated placebo controls, suppressed safety signals, under-documented “euphoria”, and allegations of ethical misconduct that had been minimised in top-line reports. One of the two who voted in favour of approval had an unreported conflict of interest. 

The rejection served as a “black swan” event. We ought to take a closer look at the available data. Across modern clinical trials, a substantial minority of patients receiving psychedelic‑assisted therapy don't show a meaningful response – but the exact figure depends heavily on which drug is used, what condition is being treated, and how doctors measure “improvement.”

When MDMA‑assisted therapy is used to treat PTSD, the results look fairly encouraging. In a recent phase 3 clinical trial, about 86.5% of participants receiving MDMA showed clinically meaningful improvement (at least a 10‑point drop on the standard PTSD severity scale) after 18 weeks. This means roughly one in seven didn't respond well to the treatment at that point. Earlier pooled data from phase 2 studies painted a slightly murkier picture: about half the participants no longer met a PTSD diagnosis after two MDMA sessions, though many of the remaining half had at least some symptom improvement. In practical terms, between 10 and 25% of MDMA‑treated patients show little to no response, while 30 to 50% continue to meet PTSD criteria even if they're partially better.

Psilocybin shows less impressive numbers when treating depression, particularly in patients who've already failed multiple antidepressants (so‑called treatment‑resistant depression). In a large phase 2 study, only 37% of patients receiving the highest dose improved meaningfully at three weeks, meaning roughly two in three didn't meet response criteria. Full remission – essentially returning to normal – occurred in just 29%, so nearly three in four were still significantly depressed at that point. When psilocybin is used for standard depression (not treatment‑resistant), results improve modestly: a trial comparing psilocybin to a placebo found about 42% sustained improvement and 25% remission, translating to roughly 58% non‑responders and 75% still meeting depression criteria.

In contemporary psychiatric trials of depression, “response” is usually defined as a ≥50% reduction in score on a rating scale such as the MADRS or Hamilton scale. In this framework, anyone who does not reach that 50% threshold is classed as a “non‑responder,” even if they show smaller but still meaningful improvement. “Remission” is usually defined as ending below a fixed cut‑off (often MADRS ≤10) rather than by percentage change. 

The Johns Hopkins psilocybin trial for depression treated 24 participants with a single or double psilocybin dose. Using the standard ≥50% symptom reduction criterion, researchers reported 71% response, implying 29% non-responders. 54% met remission criteria at 1 month. The COMPASS Pathways trial, conducted in treatment-resistant depression (TRD), is harsher. At the primary 3-week timepoint with the 25 mg psilocybin dose, only 37% responded according to the ≥50% symptom reduction criterion (63% non-responders). 

Using the remission threshold (depression score ≤10), only 29% achieved remission (71% non-remitters). One recent trial, published in JAMA, found a 53% “success rate” in terms of “response” and a 24% for “remission”. 

The true proportion of non-responders in psychedelic therapy may be significantly underestimated due to chronically short follow-up periods. While most psilocybin trials assess outcomes between one and seven weeks after dosing, with only exceptional studies extending to 3 or 6 months, conventional antidepressant efficacy research typically follows patients for 12 to 24 months, with some studies tracking outcomes over multiple years to properly assess relapse and recurrence. 

In COMPASS’ Phase 2b depression trials, sustained response rates fell to ~20% by week 12 for high-dose groups. In several early psilocybin depression studies, investigators themselves noted that for a considerable portion of participants, the strong initial antidepressant effect began to fade after about four to six weeks, with symptoms ‘creeping back’ even in those who had looked like clear responders at first.In a 2025 trial of psilocybin for alcohol use disorder, there was no statistically significant difference in relapse rates between the psilocybin and placebo groups at the six-month mark. While MDMA-assisted therapy trials measure primary endpoints at 18 weeks, evidence-based psychotherapy studies for PTSD routinely employ 12- to 24-month follow-up periods, acknowledging that meaningful relapse data requires extended observation. Guidelines recommend around 10-12 months of antidepressant treatment to minimise relapse risk.

Participant attrition, or drop-outs, in naturalistic psychedelic studies can reach 71%. Clinical trials report more modest dropout rates of 5-11% in psilocybin studies, though this compares favorably to trauma-focused therapies where dropout rates approach 47-56%. It also represents a fraction of the 1.9% MDMA-assisted therapy dropout rate versus 15.7% in placebo groups. FDA briefing documents on Lykos’ MDMA trials highlighted that ~25% of participants dropped out of the long-term follow-up phase. In a trial where success is defined by sustained remission, losing a quarter of the cohort – likely the non-responders or those who relapsed – creates a survivorship bias that artificially inflates durability claims. 

Perhaps most significantly, the populations studied may not represent those most in need: trials routinely exclude individuals with active suicidal ideation with intent or plan within the past 3-12 months, those with suicidal behaviour in the preceding two years, anyone meeting criteria for alcohol use disorder or other current substance use disorders, and patients with bipolar disorder, psychotic disorders, or certain personality disorders: precisely the populations that account for the majority of treatment-resistant cases and between 76% and 99% of real-world patients with depression. 

Only 26% of treatment-seeking participants qualified for enrollment in recent phase III MDMA trials. The reported response and remission rates may reflect outcomes for a highly selected subset. 

Psychiatric trials use depression scores like the MADRS and GRID-HAMD, which are observer-rated – meaning a clinician, not the patient, assigns the scores based on the interview. A MADRS item might ask: “How much do you feel sad?” with response options from 0 (not sad) to 6 (constant sadness). But “constant” means different things to different people and raters. One interviewer might interpret “I feel sad most of the time” as scoring 4, while another gives it 5.

Depression rating scales tend to assume a unidimensional construct, that depression is a single dimension with low and high points. In reality, depression is heterogeneous: one person might have severe mood symptoms with intact appetite and sleep, another might have anhedonia with normal mood but total loss of motivation. A single total score cannot capture this diversity.

In a study of 803 patients, the Montgomery-Åsberg Depression Rating Scale (MADRS) scale correctly identified only 66% of cases who met depression criteria clinically and had a positive predictive value of just 50% – meaning half the people scoring above 14 didn't actually have depression by clinical interview. A meta-analytic review by Bagby et al. (2004) of the GRID-HAM-D scale found that its consistency across 70 studies ranged widely from 0.46 to 0.92 – an enormous spread! In eight of twelve studies reporting reliability, it was less than 0.76, below conventional thresholds for acceptable internal consistency. 

Rater bias is more pronounced with interested funding. Systematic reviews have shown that industry-funded trials report 50% higher efficacy than independent studies. With public companies relying on positive Phase 2 and 3 data to survive, the pressure to P-hack results, exclude non-responders, and minimise safety signals is more serious.

This bias is compounded by researcher expectancy effects. Meta-analyses show that researcher allegiance – “belief in the superiority of a treatment” – may inflate results. Yet only 17% of meta-analyses of psychotherapy trials even discussed or reported researcher allegiance, and only 4% actually evaluated it. 

In psychedelic research, this bias may be even more pronounced because researchers openly advocate for legalisation and have personal investment in the field's success. A New York Times investigation into work at Johns Hopkins described researchers acting in a “spiritual leader capacity,” infusing clinical interactions with religious symbols and guiding participants toward specific metaphysical conclusions. When a researcher believes they are ushering in a “paradigm shift” for humanity, the subconscious pressure on participants to report positive outcomes is immense. This creates a demand characteristic that rivals the drug itself: patients want to please their therapists, and therapists want to validate their life's work. 

In the Lykos MDMA trials, increases in suicidal ideation or emotional instability were often coded as part of a “healing process”, one understood esoterically by therapists through intuition. Rick Doblin of Lykos/MAPS has repeatedly stated that psychedelic therapy, pursued through “political science”, may help to usher in a “spiritualised humanity” of “Net Zero Trauma” by 2070. This culture underlies the Meaghan Buisson case. Video footage from a Phase 2 trial showed therapists pinning her down, an ethical breach that went unreported in the study’s initial safety publications. 

Another problem is blinding. It is an open secret that the vast majority of participants in inactive placebo-controlled trials correctly guess their treatment assignment. Participants who “trip” know they have received the active drug, triggering a massive placebo amplification: an effect that some have suggested making use of! Once a rater believes they are assessing someone in the psilocybin condition, their interpretation of ambiguous symptoms may shift; a patient's fatigue or emotional lability gets coded differently than if the rater believed them to be on placebo.

But one review “found that no study… assessed for expectancy in participants”. The intense subjective effects of the drug make the “blind” effectively nonexistent. Those in the control group, feeling no effects, may realise they are on the placebo and experience a “nocebo” crash of disappointment. Some control group participants drop out, meaning we don’t know what happened to them in the long term. One Johns Hopkins participant who had a history of previous suicidal thoughts was enrolled in the study and received the low dose of psilocybin during his first session. Following this session, the participant's behaviour became concerning. According to Griffiths, the man reported feeling “bored” and refused to submit to the researchers' post-session questions. He soon committed suicide. Some have speculated that the act resulted from a dramatic ‘disappointment effect’ induced by the low dose. 

When this expectancy gap is closed, the “miracle” often reduces. The 2023 Stanford Anesthesia Study provided one definitive test: researchers administered ketamine or a saline placebo to patients under surgical anaesthesia, ensuring no one experienced the psychedelic trip or knew which drug they received. Ketamine performed no better than the placebo. Both groups improved dramatically, solely due to the belief that they might have been treated. This finding was reinforced by the 2025 KARMA-Dep 2 Trial, where ketamine failed to outperform an active placebo (midazolam) in a rigorous hospital setting. In psilocybin depression trials, for example, niacin has been used as an active placebo in large RCTs of major depressive disorder. 25 mg psilocybin produced around 42–53% sustained response versus 11–24% in the niacin group at six weeks, a clear advantage but much less dramatic than early open-label or inert-placebo work that suggested 60–70%+ response or remission rates. 

It is worth putting “response” in context. Psychedelics do not clearly and robustly outperform SSRIs, their main alternative, on primary outcomes. The much‑publicised head‑to‑head trial of psilocybin versus escitalopram for depression found no statistically significant superiority of psilocybin over the SSRI. While the psilocybin arm showed larger improvements on several secondary measures, the “disappointed” author conceded that these “outcomes lacked correction for multiple comparisons” and “no clinical conclusions [could] be drawn from these data”. And since SSRIs don’t do much for the average individual, the clinically meaningful ‘response’ rate of psilocybin is perhaps much lower than assumed. Irving Kirsch's meta-analyses of FDA-submitted data found that 82% of the “drug response” to SSRIs was duplicated by placebo, leaving only an 18% true drug effect. Examining all antidepressants approved by the FDA between 1983 and 2008, only 43% of trials showed a statistically significant benefit of the drug over the placebo; 57% were failed trials. The real-world burden of SSRI side effects is also substantial. Patients frequently report sexual dysfunction, weight gain, emotional blunting, and withdrawal or “discontinuation” syndromes that can be severe and protracted, especially with paroxetine and venlafaxine. 

When psychedelic therapy fails, it rarely does so neutrally. Perhaps most disturbing is the potential for Prolonged Adverse Effects (PAEs). 16% of users in naturalistic settings reported enduring declines in well-being. These include Hallucinogen Persisting Perception Disorder (HPPD), depersonalisation, hypomania, and “ontological shock” – a destabilising collapse of worldview. A harrowing 2025 case report detailed a psychology trainee who developed severe suicidality and anhedonia after mandated training sessions, a condition her supervisors ignored until she required hospitalisation. After local decriminalisation, psilocybin‑related poison‑centre calls in US youth more than tripled and hallucinogen‑related emergency department visits in California rose by over 50% from 2016 to 2022.

Psilocybin trials have also shown flickers of a “worsening” signal that is rarely discussed. In COMPASS Pathways' COMP360 Phase 2b trial, three participants in the high-dose group exhibited suicidal behaviour, compared to zero in the

It is worth considering why some would respond to psychedelic treatment better than others. 

One problem might be persistent “blunting” from SSRIs. In a 2023 study, Gukasyan et al. demonstrated that receptor “blunting” can last for 3 to 6 months following discontinuation, rendering standard two-week washout protocols insufficient for restoring neuroplastic potential. The “emotional breakthrough” coined by Leor Roseman may be more difficult to realise. While Erritzoe et al. (2024) found no differences in the types of experiences the two groups had, they reported that participants who discontinued antidepressants specifically to join a psilocybin trial showed significantly reduced treatment responses compared to unmedicated cohorts. 

For a subset of non-responders, the failure of therapy is driven by an overactive “psychological immune system.” Research by Zeifman et al. (2020) identifies experiential avoidance – the rigid suppression of difficult internal states – as a primary predictor of therapeutic failure. In these cases, the psychedelic does not induce a breakthrough because the patient’s defensive structure is too robust to be dismantled by the drug alone. This phenomenon was recently expanded upon by the hypothesis of Psychedelic Iatrogenic Structural Dissociation (PISD) proposed by Elfrink (2025). Elfrink suggests that in trauma-exposed individuals, psychedelics can trigger a severe defensive reaction where the psyche 'compartmentalises' the drug’s effects to protect against overwhelm. Rather than “surrendering,” these patients subconsciously deploy dissociation or intellectualisation to neutralise the experience, resulting in a “constrained” session that yields no therapeutic change but often leaves the patient with a profound sense of defectiveness and failure.

Perhaps the true measure of “response” happens not in the acute psychedelic experience itself, but in how patients integrate and make meaning of that experience afterwards. This seems sensible on its face. Clinical protocols typically include post-session integration sessions, though the number of hours administered varies enormously. But the field has notably failed to conduct the most basic test of this claim: a randomised trial comparing psychedelic treatment with intensive integration against psychedelic treatment with minimal or no integration. Without such direct comparison, we cannot distinguish between integration actually causing improvement and integration simply being a marker of patients who were already likely to improve.All of this echoes a long-standing debate in mainstream psychotherapy research: the dodo bird verdict. The evidence consistently shows that therapeutic alliance, or the quality of the relationship between therapist and client, predicts outcomes far more powerfully than the specific technique being used, with alliance effects (d = 0.57) sometimes exceeding the impact of the treatment modality itself. This means that specialised ‘integration’ may not be necessary for sustained response. 

This uncertainty deepens when we look at what happens when the psychotherapy component is removed entirely. Recent trials of 5-MeO-DMT for depression showed some possible benefits with no accompanying psychotherapy, while naturalistic ketamine studies found similar improvements whether patients received additional psychotherapy or not. Is integration doing the heavy lifting, or is it simply what happens when intelligent, motivated people reflect on a powerful experience?  The evidence becomes even more complicated when we add personality. Higher openness predicts both psychedelic use and lower baseline depression, as well as the intensity of mystical experiences during sessions. 

It is also possible that labelling patients as “non-responders” risks confusing the nature of suffering. Depression often has rational, structural causes: poverty, trauma, unemployment, and social isolation. When someone remains depressed despite psilocybin, whilst their social conditions remain unchanged, calling them a “non-responder” might misframe the problem on both sides of the clinical arguments around psychedelics.  

Consider what actually predicts who engages meaningfully with integration work. Access to regular, skilled therapists requires money. The flexibility to attend weekly appointments demands employment stability. The cognitive sophistication to process and apply psychological insights correlates with education. The patience and hope to sustain effort through integration work depend on social support and life stability. All of these factors independently predict better mental health outcomes regardless of whether someone receives psychedelic therapy.

A systematic review of modern psychedelic trials revealed that 85% of participants are white, and nearly all are recruited from academic centres or urban hubs. A 2025 systematic review published in Nature Mental Health examined 49 psychedelic clinical trials and found that 93% had some college education or more, compared to 62% of the US population. The two largest US psilocybin trials (Bogenschutz et al. 2022 for alcohol dependence, and Raison et al. 2023 for major depression) reported median participant incomes of ≥$100,000. Only 12% of trials reported income data, and 31% reported educational attainment – meaning the vast majority of psychedelic trials simply don't disclose participants' class backgrounds at all. 

Current cost models for MDMA-assisted therapy estimate a price tag of $12,000, driven largely by the dozens of hours of therapist time required. Insurance coverage for such intensive protocols is speculative at best. For a patient paying out-of-pocket, a failed treatment is a financial disaster. 

This means that reported “non-response rates” tell us nothing about how these treatments work for people living in social deprivation.

To understand the true state of psychedelic research, we must subject it to the rigorous statistical framework established by John Ioannidis (2005). Ioannidis famously argued, in his 2005 paper ‘Why Most Published Research Findings Are False’, that in “hot” research areas with small studies, flexible endpoints, and strong interests at stake, most positive findings will eventually turn out to be wrong. Psychedelic therapy currently ticks every one of those boxes. In other domains matching this profile, like early hormone‑replacement studies, nutritional epidemiology, much of psychology, and candidate gene research, whole literatures looked compelling for a decade, only to collapse once larger, cleaner trials were run.

To restore credibility and protect patients, the field must adopt three urgent reforms. First, mandatory active placebos or anaesthesia protocols must become the standard for Phase 3 trials. As the Stanford and KARMA-Dep 2 studies proved, without a structurally enforced blind, “efficacy” is statistically indistinguishable from expectancy. Second, we must end the file-drawer problem. Journals and sponsors must commit to publishing negative results – like the recent ketamine and psilocybin failures – to correct the massive asymmetry in the funnel plot. 

A literature that only reflects success is not science. It is marketing.

Ed is one of our community bloggers here at Chemical Collective. If you’re interested in joining our blogging team and getting paid to write about subjects you’re passionate about, please reach out to Sam via email at [email protected]