The estimated annual financial impact of gene therapy in the United States

The study addresses the question: What is the expected annual fiscal impact of gene therapy for the U.S. healthcare system over 2020–2034, given therapies approved or in late-stage development as of December 2019/2020? Gene therapy can substantially improve outcomes for conditions such as inherited retinal disease and spinal muscular atrophy, but very high prices raise affordability and access concerns for payers and patients. There is limited systematic information on patient counts and spending for gene therapies, and payer coverage (especially outside Medicare) is uncertain. The authors aim to estimate the number of approvals, treatable patient populations, prices, and resulting annual spending, using a simulation framework that incorporates uncertainty and pipeline dynamics to inform stakeholders and policy discussions around access and financing.

The paper notes a paucity of contemporaneous and systematic data on usage and spending for approved gene therapies in the U.S. Prior work by Quinn et al. (2019) estimated combined cell and gene therapy patient counts reaching roughly 12,000 in 2020 and over 340,000 by 2030 but did not isolate gene therapies or provide disease-level detail. Insurers and public payers have not routinely reported gene therapy utilization or expenditure statistics. Coverage variability and restrictive policies are common for approved gene therapies, with concerns that non-Medicare plans facing fixed budgets may delay or restrict coverage due to premium and wage implications. Drug development success rates are lower in oncology and rare diseases versus the average across indications, impacting expected approvals. The authors position their work as the first comprehensive projection of U.S. gene therapy spending by payer proxy, disease category, and time, incorporating clinical development risks and pricing tied to health gains.

Design: A simulation-based financial model estimating gene therapy approvals, patient eligibility, treatment uptake, prices, and spending from January 2020 through December 2034, using therapies approved or in late-stage (Phase 2/3 or 3) trials as of December 2019 (consistent with FDA reliance on foreign trial data when applicable). Compassionate use trials were excluded. Data sources: Citeline TrialTrove and ClinicalTrials.gov; literature and repositories (e.g., SEER, cancer.net) for incidence/prevalence; ICER reports and market prices for QALY and pricing calibration. Trial identification: Trials tagged as gene therapy, de-duplicated, filtered to Phase 2/3 or 3; included trials without U.S. sites due to possible FDA approval from foreign data. Diseases classified as oncology, rare, or general. Resulting dataset: 109 unique trials across 57 diseases. Approval simulation: Development programs modeled with correlated success (assumed 90% correlation across programs for the same disease platform; sensitivity showed low impact). Phase-3-to-approval probabilities (PoS3A) by therapeutic area from MIT LFE Project ALPHA path-by-path estimates: Autoimmune/Inflammation 48.5%, Cardiovascular 50.1%, CNS 37.0%, Metabolic/Endocrinology 45.7%, Oncology 28.5%, Ophthalmology 45.9%. Time to approval: Two lags modeled via triangular distributions (0–365 days, median 182.5 days) for (i) last trial end to BLA submission and (ii) BLA submission to FDA decision; trial end dates imputed via a fitted gamma distribution if missing. Diseases with prior approved therapies counted as approved by Dec 31, 2020; diseases with last trial ending before Jan 2017 and no approval treated as failed. Patient eligibility and counts: For each disease, the broadest trial-listed subpopulation was used; in absence of subgroup detail, all diagnosed patients assumed eligible. Incidence and prevalence collected from literature; when missing, estimated from U.S. population (assumed 327.7 million). If only incidence available, prevalence estimated using steady-state survival-based relation; if only prevalence available, incidence inferred by rearrangement (assuming steady state). Overlaps between subcategories adjusted to avoid double counting. Treatment timing and uptake: Newly diagnosed patients treated immediately. Prevalent (existing) patient stock declines exponentially with half-life chosen so that 25% seek treatment in year 1 (half-life 28.91 months), implying 95% of pre-approval prevalent patients treated within 10.5 years; sensitivity varied this assumption. Patient penetration: Modeled via a ramp with disease-category-specific parameters and 10% coefficient-of-variation for uncertainty: rare diseases θmax=40%, ramp T=6 months; cancers θmax=10%, ramp T=12 months; general chronic diseases θmax=1%, ramp T=5 years. Net treated patients each period equal the penetration rate times the sum of new and (decaying) existing eligible patients. Competition effects across therapies for the same disease and patient-type heterogeneity were not modeled; independence across treatment-disease segments assumed. Pricing and QALY: Cost per period equals treated patients multiplied by the therapy price. Prices estimated via a QALY-based approach: price = (price per ΔQALY) × expected ΔQALY. Expected ΔQALY per disease estimated with a model calibrated against available gene and CAR-T therapy data; age-group weighting followed ICER definitions (minors <18, adults 18–62, elderly >62). Calibration using U.S.-approved therapies as of Jan 2020 yielded price per expected ΔQALY of $101,663 for rare diseases (MSE 2.18×10^7; MAPE 11.2%) and $40,797 for other diseases (MSE 1.77×10^10; MAPE 44.2%); using all (including EU) data increased rare-disease price per ΔQALY to $114,781 but the former was used. This calibration reproduced real-world prices for Zolgensma (~$2.09M/patient estimated vs $2.1M) and Luxturna (~$0.47M/eye estimated vs $0.425M/eye), and ΔQALY estimates comparable to ICER ranges. Simulation: One million Monte Carlo iterations provided means and 5th–95th percentile bands. Sensitivity analysis varied by ±20%: penetration ceiling (θmax), ramp time (Tmax), ΔQALY, price per ΔQALY, PoS3A, incidence (new patients), prevalence (existing patients), time from Phase 3 to BLA, and BLA to approval. Additional analyses: varied correlation across development programs (0 to 1.0); varied proportion of existing patients treated in year 1; and explored adding future gene therapy programs via a Poisson process calibrated from historical initiation trends to compare against the baseline pipeline-only scenario.

Approvals and patients: • Expected cumulative approvals (Jan 2020–Dec 2034): 18.3 (90% CI: 14.0–23.0). • Monthly patients treated peak: 7,911 in July 2025 (CI: 3,978–12,477), declining to 5,424 by Dec 2034 (CI: 2,778–8,350). • By Dec 2034, only 7% of treated patients are preexisting (pre-approval) cases. • Disease-category mix of treated patients in Dec 2034: cancer 48.0%, general 30.0%, rare 22.0%. • Cumulative treated patients by Dec 2034: 1.09 million (CI: 0.595–1.66 million). • Annual new patients treated: 16,244 (2020) → 94,696 (2025) → 65,612 (2034). • Age distribution of patients over time: minors 17.9%, adults 35.4%, elderly 46.7%. Spending: • Monthly spending peaks at $2.11B in April 2026 (CI: $1.01B–$3.88B), then trends to a steady-state ~$1.62B/month (CI: $0.624B–$2.98B). • Cumulative discounted spending through Dec 2034: $241B (CI: $123B–$402B). • Annual spending: ~$5.15B (2020), rising to ~$25.3B (2026), then ~$21.0B (2034). Average annual spending across years: ~$20.4B. • Initial spending concentrated on existing cancer patients (>45.6% of monthly expenditure) falls to ~1% by Dec 2034; spending shares for new general and rare disease patients rise over time (to ~21.2% and ~46.2% by Dec 2034, respectively). Health gains and value: • Mean expected ΔQALY per treated patient: 5.12 life-years. • Average cost per ΔQALY at 2020 present value: ~$43,110. Payer proxy by age group: • Spending shares: minors 43.2%, adults 26.0%, elderly 30.9%. • Implied annual spending by payer type (assumptions: all elderly on Medicare; 2/5 children and 1/7 adults on Medicaid; remainder private): Medicare ~$8.1B, Medicaid ~$5.44B, private ~$12.2B. Sensitivity highlights: • Cumulative discounted spending and peak monthly spending scale approximately linearly with θmax, ΔQALY, and price per ΔQALY. • Changes in PoS3A and patient numbers (new/existing) have sublinear impacts. • Timing variables (Phase 3→BLA, BLA→approval, ramp-up time) have small effects in the opposite direction. • Peak-spend month shifts by up to ±10 months with ±20% changes in PoS3A, new patient counts, or time parameters; changes in θmax, ΔQALY, and price per ΔQALY do not shift peak timing. • Varying program correlation from 0.9 to 0 or 1.0 changes mean discounted cumulative spending by +3.4% or −0.4%, respectively. • Varying the proportion of existing patients treated in year 1 materially changes discounted cumulative spending (−32% to +0.08% across tested range); results remain within ±5% of baseline if this proportion is 8%–45%. Pipeline growth scenario: • Adding new programs (Poisson arrivals) increases expected approvals from 18.3 to 23.0 (+25.1%), cumulative patients from 1.09M to 1.26M (+15.3%), and cumulative spending from $306B to $354B (+15.7%).

The model-based projections indicate that U.S. annual spending on gene therapies could average roughly $20.4B over 2020–2034, peaking mid-decade and stabilizing thereafter, given the late-stage pipeline as of end-2019/2020. These estimates directly address the policy question of affordability and access by decomposing patients and spending over time, by disease categories, age groups, and inferred payer types. Under conservative assumptions (excluding post-2019 program entries in the baseline), the projected spending suggests meaningful but manageable budget impacts relative to overall U.S. health and tax revenues. The estimated average ΔQALY gain (5.12) and cost per ΔQALY (~$43,110) support the notion that gene therapy delivers substantial health value, even as aggregate expenditures are significant. The analysis highlights that initial spending is driven by clearing the backlog of existing (especially cancer) patients, with long-run spending dominated by incident cases in rare and general diseases. Sensitivity analyses identify key drivers (penetration ceiling, ΔQALY, and price per ΔQALY), providing levers for policy and negotiation. The study discusses financing mechanisms (outcomes-based contracts, annuities, pooled reinsurance models, and programs like Cigna’s Embarc and CMS’s Cell and Gene Therapy Access Model) that could mitigate payer risk under uncertainty about durability and real-world effectiveness, thereby facilitating access while managing budget shocks. Overall, results are relevant for designing coverage, payment, and risk-pooling strategies for durable gene therapies.

The authors developed a simulation framework that integrates clinical development risk, disease epidemiology, patient uptake, and QALY-based pricing to estimate U.S. gene therapy approvals, treated patients, and spending through 2034. They project average annual spending of ~$20.4B with substantial health gains, noting that their baseline likely understates long-run impact because it excludes post-2019 pipeline growth. The framework can be updated as data evolve and used to evaluate payment models and policy options to ensure access while maintaining fiscal sustainability. Future research should incorporate: (1) real-world effectiveness and durability data to refine ΔQALY and price assumptions; (2) clinical delivery costs and care pathway changes; (3) competition dynamics within disease-population pairs; (4) broader pipelines including accelerated approvals; and (5) payer behavior and coverage policies affecting penetration and timing.

Key limitations include: (1) Conservative pipeline scope: baseline excludes programs entering clinical trials after December 2019, potentially understating long-run approvals, patients, and spending (a separate pipeline-growth scenario indicates ~16% higher cumulative patients and spending). (2) Uncertain inputs: multiple parameters (incidence/prevalence, ΔQALY, price per ΔQALY, penetration, timing lags, PoS3A, program correlation) rely on limited empirical evidence; sensitivity analyses provided but residual uncertainty remains. (3) Pricing approach: QALY-based pricing may over- or underestimate realized prices; does not incorporate potential offsets (e.g., avoided chronic therapy) or higher clinical delivery costs of gene therapies. (4) Competition not modeled: assumes independence across treatment-disease segments; does not capture price or share effects from multiple therapies targeting the same disease-population pair. (5) Patient uptake assumptions: exponential decline of prevalent pool and fixed penetration ramps may not generalize across indications; results are sensitive to the proportion of existing patients treated early. (6) U.S.-only approvals: trials conducted solely outside the U.S. were excluded to align with FDA pathways and data constraints; may bias estimates downward if they target unique diseases/populations not captured in U.S.-registered trials. (7) Data lags and imputation: trial end dates and timelines were sometimes imputed, introducing model risk.