Psychedelic Therapy Cost-Effectiveness Model

Psychedelic-assisted therapy concentrates a lot of cost at the start: screening, preparation, a long dosing session, monitoring, integration, and follow-up. The economic question is whether that upfront episode buys enough durable health gain, avoided care, and wider societal value to justify the spend.

The tool is built for sensitivity work. It lets a reader change the horizon, relapse hazard, perspective, threshold, comparator, dropout handling, re-treatment, and acute course cost, then see how the value case moves.

The acute decision tree resolves each patient into remission, response without remission, non-response, or dropout, with a separate serious-adverse-event branch for cost and utility decrement.

Those acute outcomes then enter a Markov model with quarterly cycles across remission, partial response, symptomatic illness, re-treatment, and death. The model applies a half-cycle correction and keeps the compound-specific assumptions inside parameter packs.

The tool reports incremental cost, incremental QALYs, ICER, net monetary benefit at all three Dutch severity tiers, payer-versus-societal cost split, cost per additional remission, budget impact, and the value added by a longer horizon.

Net monetary benefit is the clearest headline because it remains interpretable when the incremental effect is small or when the intervention is close to cost saving.

A one-off or short-course treatment is only economically attractive if the effect lasts long enough to offset the acute service cost. The current model therefore makes the relapse hazard a front-and-centre input rather than a buried assumption.

The most useful next evidence is not another acute symptom score alone. It is a sourced relapse curve over twelve months and beyond, with re-treatment and rescue-care pathways measured prospectively.

COMPASS Phase III sources are linked because they are important current evidence, but the model does not yet treat company materials as peer-reviewed extracted endpoints. COMP005 and COMP006 week-6 MADRS effects and response proxies can inform scenario presets, while remission, dropout, relapse, retreatment, and long-run utility inputs stay labelled as scenario values until endpoints, denominators, and follow-up timing are harmonised.

That separation matters because HTA modelling often needs a longer horizon than the pivotal endpoint. A six-week or nine-week primary endpoint can support short-term effect, but the economic case still turns on what happens over months and years.

The Dutch reference case is societal, so productivity loss and informal care belong in the base case. A healthcare payer may still see a net budget increase even when society sees a larger offset.

The tool reports both views side by side because that gap is the access argument and also the political problem: the value may appear outside the budget holder's own ledger.

The country selector separates HTA method assumptions from the therapy parameter packs. The Netherlands can apply the euro-denominated societal reference case directly; the UK can apply NICE-style perspective and discounting but not GBP thresholds to euro costs; Sweden can apply societal perspective and 3% discounting while keeping WTP explanatory; the Czech Republic can apply payer perspective and 3% discounting while displaying, but not applying, CZK thresholds to euro outputs.

Germany is different: the lens stays route-first because AMNOG added benefit, the G-BA's appropriate comparator therapy, and price negotiation are more decision-relevant than a generic ICER cutoff.

The base comparator remains standard of care. Spravato/esketamine is a useful access precedent and a labelled active-comparator stress test, but it is not silently substituted into the base case.

Psychedelic trials are difficult to blind, and expectancy can inflate measured effects. The expectancy adjustment applies to incremental QALYs only, so users can test whether the case survives a smaller assumed true effect.

This is especially important for MDMA-assisted therapy in PTSD, where the 2024 FDA discussion made functional unblinding and trial conduct central to the evidence interpretation.

The interface now labels major input groups as peer-reviewed, registry-linked, company-release, economic-model, scenario, or placeholder evidence. That is intentionally less detailed than a full evidence dossier, but enough to stop placeholders from looking source-backed.

Relapse hazards, utilities, downstream healthcare costs, productivity costs, informal-care costs, eligible population, excess mortality, and proper remission rates still need source-grade replacements. The current build is useful for sensitivity work, threshold thinking, and evidence planning, not as a reimbursement submission.