Psilocybin-assisted therapy for treatment-resistant depression in the US: a model-based cost-effectiveness analysis

The investigators built an individual-level, probabilistic state-transition simulation to represent a cohort of US adults with TRD. The model used weekly cycles over a 12-month time horizon and tracked four mutually exclusive health states: MDD (starting state), response, remission and death. TRD was operationalised as failure to respond to two prior MDD treatments. Individuals who failed a line of therapy were assumed to move to the next line after 8 weeks, and those failing third-line treatments were assumed to receive pharmacotherapy for fourth-line and subsequent care. Interventions compared were single-dose PAT (delivered with intensive psychosocial support, base-case total cost assumed $5,000) versus a population-weighted standard of care (SOC) consisting mostly of second-generation antidepressants with psychotherapy, plus small proportions receiving ECT or esketamine. The authors adopted two perspectives: a limited healthcare perspective (direct healthcare costs such as treatment, pharmacy, outpatient, inpatient and emergency services) and a limited societal perspective that additionally included productivity losses (absenteeism and presenteeism). Costs were expressed in 2022 US$. Because long-term PAT data were lacking, outcomes were not discounted and the analysis was restricted to 12 months. Model inputs came from published sources. Transition probabilities for pharmacotherapy and psychotherapy were informed by STAR*D results; ECT inputs came from a meta-analysis; esketamine effect estimates were assumed to be about 30% higher than standard treatments when combined with antidepressants; and PAT effectiveness used a phase 2 RCT reporting a 29% remission rate and 8% response rate for a single 25-mg dose with intensive support. Relapse probabilities were taken from the literature, and because long-term PAT relapse rates are uncertain, the authors conservatively assumed PAT relapse rates matched third-line therapies in STAR*D. Health utilities (for QALY calculation) were derived from a systematic review using time-trade-off values. To characterise uncertainty, the study performed probabilistic analyses with 1,000 model runs of 1,000 simulated individuals each, sampling parameter values from assigned distributions to generate mean costs, QALYs and 95% predicted intervals. The primary incremental outcome was the ICER (incremental cost per QALY gained) comparing PAT to SOC, evaluated against a US-relevant threshold range of $100,000–150,000 per QALY. The authors also conducted conditional probabilistic analyses that deterministically varied three policy-relevant inputs (probability of remission, probability of relapse, and cost of PAT) while modelling all other parameters probabilistically; PAT cost was varied deterministically from $3,000 to $20,000 (with a focused range of $3,000–10,000 for sensitivity scenarios). Results were summarised using cost-effectiveness acceptability curves and expected loss curves to indicate decision uncertainty and the value of additional information.

In the base-case assuming a PAT cost of $5,000, PAT increased mean costs by $3,638 per person from the limited healthcare perspective and produced an incremental benefit of 0.031 QALYs over 12 months (equivalent to 31 QALYs per 1,000 individuals), driven by a 9% higher overall remission rate versus SOC. The expected ICER was $117,517 per QALY from the limited healthcare perspective and $111,069 per QALY from the limited societal perspective. Cost sensitivity analyses showed that PAT met common US cost-effectiveness thresholds when its total cost was $5,000 or less. At a PAT cost of $10,000, expected ICERs rose to $275,426 (healthcare) and $268,978 (societal), well above the examined thresholds. Probabilistic results (base-case $5,000) indicated the probability that PAT was cost-effective rose from 31% at a $100,000 threshold to 74% at a $150,000 threshold. At the $150,000 threshold PAT had a lower expected loss than SOC ($301 versus $1,307 per person), indicating it was the preferred option on average under that threshold. By contrast, when PAT cost $10,000 the probability of being cost-effective at the $150,000 threshold fell to 1%. Conditional probabilistic analyses revealed that changes in PAT effectiveness affected its economic value more pronouncedly for remission than for relapse. For example, with PAT cost at $5,000, increasing the remission probability by 15% raised the probability of PAT being cost-effective from 74% to 89%, whereas reducing relapse probability by 15% increased that probability from 74% to 85%. Joint improvements in remission and relapse produced larger gains: at a PAT cost of $10,000, increasing remission by 15% and decreasing relapse by 15% increased the probability PAT was cost-effective from 1% to 34%. Similar patterns held under the limited societal perspective. The extracted text indicates that 95% predicted intervals were calculated from the probabilistic simulations, but specific interval bounds are not clearly reported in the provided extraction.

Avanceña and colleagues interpret their modelled results to mean that a one-time, single-dose PAT programme could be cost-effective as a third-line treatment for TRD in the US under particular conditions, principally when total per-patient PAT costs are about $5,000 or lower and when short-term effectiveness (remission and relapse) aligns with or exceeds values used in the base case. The authors note that higher PAT costs (for example, $10,000) make it unlikely to be cost-effective even if effectiveness were greater than currently reported. The study is positioned relative to prior work—most notably a UK analysis that found improved health but unfavourable costs for PAT—by emphasising methodological differences: an individual-level state-transition model here (which captures sequential and recurring events and individual characteristics), a longer 12-month horizon, focus on a defined TRD population, and probabilistic characterisation of parameter uncertainty. The authors stress that the cost of psychological support and the price of psilocybin itself are key determinants of PAT's economic value. They suggest potential cost-reduction strategies such as group therapy formats or use of alternative clinical personnel but caution that the effectiveness of lower-cost delivery models has not been established and requires testing. The investigators acknowledge that their perspectives were limited and omitted some cost categories (caregiving, transportation and patient time) that a full societal analysis would include; excluding these may affect estimated incremental costs. They highlight equity and access concerns, noting existing disparities in mental healthcare utilisation and uptake of novel therapies, and recommend future work using distributional cost-effectiveness approaches to assess how PAT's adoption might affect different subgroups. Finally, the authors call for further economic evaluations as more robust long-term data emerge, particularly to inform pricing, coverage and implementation decisions while ensuring PAT remains affordable and accessible.

The authors conclude that single-dose PAT is an effective treatment for MDD and TRD and may be cost-effective compared to standard third-line treatments in the US when the total per-patient cost is $5,000 or less. They recommend that future trials and implementation research explore ways to lower the cost of PAT and to generate longer-term effectiveness and relapse data to better inform economic evaluations and policy decisions.