A peer-reviewed model published in August 2025 found that psilocybin-assisted therapy for treatment-resistant depression would cost about $3,640 more per patient than standard care over twelve months. A model published in March 2026 found it would save about $7,000 per patient over six months. Both take their efficacy from the same single 25 mg dose in the same Phase 2b trial, and both find psilocybin cost-effective at base-case assumptions. The 2025 model puts it at $117,517 per quality-adjusted life year (QALY), below the $150,000 threshold it applies, and its authors conclude that psilocybin may offer economic value at a course price of $5,000 or less. The 2026 model finds it dominant, meaning cheaper and more effective.

The sign on cost differs because of what each model puts around that dose: what a course costs, how it handles failure, how long it looks and what standard care costs. Neither base case models the course Compass Pathways tested in its second Phase 3 trial, which gave two doses, and its November 3 results release frames COMP360 as a few treatments a year.

One dose, two sets of assumptions

Avanceña and colleagues, Translational Psychiatry, August 29, 2025Ziadi and Park, Value in Health Regional Issues, online March 18, 2026
Efficacy sourcePhase 2b trial, 25 mg armSame trial; the two-dose scenario uses a 2021 psilocybin-versus-escitalopram trial
Base-case dosingOne 25 mg doseOne 25 mg dose; scenarios add two doses three weeks apart, or a repeat dose at week 12 for relapsers
Horizon12 months, undiscounted6 months; 10- and 30-year scenarios discounted at 3%
Course cost$5,000 assumed ($3,500 for therapists, $1,500 for psilocybin)$2,596 in Table 1, micro-costed (psilocybin $250; coded visits for assessment, preparation, a six-hour administration session, integration and follow-up)
After failure or relapseSwitch to pharmacotherapyNothing further modeled in the base case
Comparator99.59% pharmacotherapy and psychotherapy, 0.25% electroconvulsive therapy (ECT), 0.16% esketamine; transcranial magnetic stimulation (rTMS) and intravenous ketamine excludedPharmacotherapy lines 3 to 6 from the STAR*D trial; neuromodulation and adjuvant psychotherapy excluded; esketamine discussed but not modeled
Standard-of-care cost$11,061 over 12 months$8,335 over 6 months
Result per patient$3,639 more; 0.031 more QALYs; $117,517 per QALY$7,037 less (Table 2); 0.099 more QALYs; dominant
FundingSAMHSA task order and an NIH career award; no competing interests declaredNone received

Both papers are in the public record with their inputs on the page. The 2026 paper is open access and its Table 1 lists the base-case costs and probabilities.

One number in the 2026 paper is unexplained. Table 1 lists the psilocybin course at $2,596, and Table 2 lists the psilocybin arm’s total six-month cost as $1,298, half that. The text does not reconcile the two. The $7,037 saving in Table 2 is the difference between the comparator’s $8,335 and the $1,298.

The $10,000 tests are different tests

The two papers each report that psilocybin survives or fails at $10,000, and the figures are not comparable. In the 2025 paper $10,000 is the price of the whole course. It raises the cost per QALY from $117,517 to $275,426 and cuts the chance of psilocybin being cost-effective at a $150,000 threshold from 95 percent at $3,000 to 1 percent.

In the 2026 paper $10,000 is the acquisition price of a 50 mg dose of the drug, tested from $500 to $10,000 with the visit costs held at their base values. The authors report that psilocybin stayed cost-effective across the range. Their base case uses a 25 mg dose at $250, against $1,500 for psilocybin in the 2025 model. Neither is a Compass price, because Compass has not published one. The 2025 course price, at $5,000, is nearly twice the $2,596 in the 2026 paper’s Table 1.

Oregon offers one outside reference. Researchers who called the state’s licensed psilocybin service centers in mid-2025 were quoted a median of $1,900 for preparation, dosing and integration, with a range of $850 to $3,000. That is a non-medical program with no approved product, and the report does not say whether the psilocybin is included.

What failure costs, and how long to look

The models treat a patient who does not respond differently. In the 2025 model, patients who fail psilocybin move on to pharmacotherapy after eight weeks, as patients in the comparison arm do. In the 2026 base case, patients who do not respond or relapse after psilocybin receive no further treatment, while the standard-of-care arm moves on through further treatment lines, up to line 6. The 2026 authors flag this as an assumption that could bias results toward psilocybin, and test it with a repeat dose at week 12. Psilocybin still comes out ahead, by about $6,800 per patient in that scenario and about $6,400 with two doses.

The horizons differ as much. The 2026 authors chose six months because it is the longest period for which efficacy data exist. They then extend to 10 years, where psilocybin saves about $95,700 per patient, and 30 years, where it saves about $215,900 and gains 9.87 QALYs, and report that it stays ahead when its effect fades gradually to 30 percent of its first-year value by year 30. The 2025 authors stop at twelve months. They borrow relapse rates from third-line treatments in the STAR*D trial, which they say is likely an overestimate.

The comparators overlap more than the two results suggest. Both are almost entirely pharmacotherapy. The 2026 authors say leaving out esketamine and neuromodulation gives a conservative, lower-bound estimate of psilocybin’s value, because costlier comparators would raise its savings. The 2025 authors say excluding rTMS and ketamine will likely affect psilocybin’s relative value, without giving a direction.

Neither base case uses Phase 3 data. COMP005 reported a 3.6-point advantage on the Montgomery-Åsberg Depression Rating Scale over placebo at week 6 for a single dose. COMP006 reported 3.8 points for two doses three weeks apart against a 1 mg dose.

Assumed, trialed and labeled

Doses

  • Models assume: One; the 2026 paper adds two-dose and week-12 repeat scenarios
  • Trial protocols: COMP005: one 25 mg dose, with a blinded retreatment dose and an open-label added dose. COMP006: two doses three weeks apart, up to four over 52 weeks
  • Esketamine label: Twice weekly in weeks 1 to 4, weekly in weeks 5 to 8, then weekly or every two weeks

Time in the room

  • Models assume: 2026: a six-hour session billed as an office visit with prolonged service. 2025: not itemized
  • Trial protocols: Approximately six to eight hours in Compass’s trials, with at least one healthcare professional present throughout
  • Esketamine label: At least two hours of observation each session

Staff

  • Models assume: 2025: two trained professionals described; $3,500 for therapists
  • Trial protocols: Usona’s registered Phase 3 protocol: two facilitators for a 7 to 10 hour session, plus 4 hours of integration
  • Esketamine label: Under a healthcare provider’s supervision

Preparation and integration

  • Models assume: 2026: coded psychotherapy visits
  • Trial protocols: At least one and up to three preparatory sessions in Compass’s Phase 2b trial; follow-up sessions
  • Esketamine label: None

BW has set out how far the leading psilocybin-class programs diverge on dosing days, doses and staffing in its October comparison of treatment models. The cost models use none of that spread. The trial protocols also describe what trials required, not what a label will. No label exists yet, and what it will require of monitoring or staffing is unsettled.

The comparison ICER proposes

The Institute for Clinical and Economic Review (ICER) posted a draft scope for COMP360 on October 21, and public comment on it closed on November 10. The scope proposes a network meta-analysis against esketamine, with intravenous ketamine if data permit, and rTMS. Its base case would run five years from a health care system perspective, with costs that include drug administration and monitoring, and would account for retreatment after relapse or loss of response as data permit. A revised scope is due November 18, the model analysis plan is scheduled for January 22, 2027, and the New England CEPAC public meeting for May 20.

Esketamine gives that comparison a priced benchmark. Medicare pays the visit under two billing codes, G2082 for up to 56 mg and G2083 for more, whose descriptors include the drug and the observation. In CMS’s final 2027 fee schedule the national non-facility totals are 29.12 and 40.93 relative value units. At the $32.63 conversion factor for non-qualifying participants in alternative payment models ($32.95 for qualifying ones), that is about $950 and $1,336 a visit before geographic adjustment. Twelve visits, the first eight weeks of the label, come to about $11,400 at the lower dose and $16,000 at the higher. That is a yardstick, not a forecast. Commercial plans negotiate their own rates, esketamine continues past week 8, and the American Medical Association’s Category III codes for monitoring psychedelic sessions, 0820T to 0822T, are contractor-priced, with no national Medicare rate.

What a payer would still need

  • A payment for the dosing day. Hours, staff mix and who may bill all feed the price. The 2026 paper prices it as an office visit with prolonged service, which is one choice among several.
  • Durability and retreatment. Compass’s two Phase 3 trials are blinded through week 26. Its November 3 release reports that an added 25 mg dose in the open-label COMP005 extension extended durability to one year. That is company-reported, open-label data.
  • Evidence against the treatments patients use now. Neither Phase 3 trial compares psilocybin with esketamine or rTMS.
  • A budget. ICER’s scope puts treatment-resistant depression at an estimated 2.8 million Americans among those treated for major depression, and Compass cites 4 million living with the condition. Every 1 percent treated is 28,000 to 40,000 patients, or $140 million to $400 million at $5,000 to $10,000 a course, before retreatment. That is arithmetic on the models’ price range, not a forecast of uptake.
  • A legal path. Compass said on November 9 that it had completed its new drug application, expects a launch in the first half of 2027 subject to FDA approval, and says COMP360 must be rescheduled before it can be marketed in the United States. As of November 14 it has not announced that FDA accepted the application. Executive Order 14401, signed April 18, directs the Attorney General, in consultation with HHS, to initiate and complete review of Schedule I products that have finished Phase 3 trials for a serious mental health disorder, so that rescheduling, if appropriate, can proceed as quickly as practicable once they are approved. It sets no deadline.

The Academy of Managed Care Pharmacy reported on September 8 on a March roundtable of health plan leaders, delivery networks, pharmacy benefit managers and clinical experts. The participants identified open questions on comparative effectiveness, patient selection, durability and cost-effectiveness. They map onto the rows above.

The two models share one assumption, a single dose, at the point where Compass’s plans are least settled: how many doses make a course, and who is paid to sit with the patient for a day. ICER’s model plan is the first document likely to put numbers on both.