Casgevy and Lyfgenia, the two FDA-approved gene therapies for sickle cell disease, list at $2.2 million and $3.1 million respectively, administered once. The conventional pharmaceutical reimbursement architecture, which assumes ongoing administration of medicine over months or years, was not designed to accept a single transaction of this magnitude tied to a clinical benefit that may persist for decades or may not. Whether the existing architecture bends or breaks depends less on financial engineering than on whether the durability data the therapies generate over the next five to ten years matches what the early trials suggested.
The gene therapy reimbursement question is not new—Glybera, the first gene therapy approved in Europe in 2012, was priced at over a million euros and famously sold one dose before being withdrawn from the market in 2017, partly because the reimbursement architecture failed to evolve in time. What is new in 2024 and 2025 is the volume. There are now over 30 FDA-approved cell and gene therapies, with another wave of late-stage candidates on the horizon. Whether the industry can build the reimbursement infrastructure to support broad patient access at sustainable prices is no longer a hypothetical question.
The financial mismatch between a single-dose curative therapy and conventional health insurance is, in its abstract form, a problem of duration. Health insurance premiums are calculated annually, claims are processed annually, and the financial liability for a covered patient resets each plan year. A therapy that delivers benefit over twenty years is, from the insurer’s perspective, a twenty-year asset acquired with current-year cash. If the patient changes insurers in year three, the insurer that paid for the therapy receives no further benefit; the insurer that gains the patient receives a benefit it did not pay for. The misalignment is acute in the United States, where commercial insurance turnover is high, but it is present in every system that has not centralized either pricing or risk-pooling at a national level.
Outcomes-based contracts, which tie payment to clinical durability, have been the most-discussed solution. The mechanics are straightforward in principle: the manufacturer agrees to refund some or all of the purchase price if the therapy fails to deliver promised durability, measured against pre-specified clinical endpoints over a defined period. The mechanics are difficult in practice. Defining failure is contested—does a partial loss of benefit count? Tracking individual patients across insurance changes is administratively burdensome. The data infrastructure to verify outcomes at the patient level often does not exist. Health Affairs analysis of value-based contracts for gene therapies through 2022 documented multiple high-profile arrangements that produced very few actual rebate payments because the verification infrastructure could not support the clinical endpoint definitions in the contract.
Annuity payment models have been the second major proposal. Rather than a lump-sum payment, manufacturers would receive a stream of payments over five or ten years, contingent on continued clinical benefit. The model preserves manufacturer revenue while reducing immediate insurer cash flow demands and aligning payment timing with benefit accrual. The legal complications, however, are substantial. Medicaid’s best-price reporting requirements, which in their current form would require manufacturers to extend any annuity-priced discount to all Medicaid programs at the lowest annuity payment level, have been a structural barrier. The Centers for Medicare and Medicaid Services has issued guidance permitting some workarounds, but the workarounds remain narrower and more cumbersome than the industry would like.
The CMS Cell and Gene Therapy Access Model, launched in 2025 and focused initially on sickle cell disease, represents the most concerted federal effort to date to address the reimbursement architecture for these therapies. The model offers participating state Medicaid programs the option to enter into outcomes-based agreements negotiated by CMS with manufacturers, with shared savings flowing back to the states that participate. The early state participation has been broader than expected, including several Republican-led states whose Medicaid programs have substantial sickle cell populations. Whether the model produces durable savings, or merely shifts the cash flow timing without affecting total spending, will depend on outcomes data that will not be available for another three to five years.
There is a quieter dimension to gene therapy economics that the financial engineering conversation has largely ignored. The patient eligibility criteria for these therapies are, in most cases, narrower than the indication labels suggest. Casgevy’s approval in sickle cell disease was for patients with recurrent vaso-occlusive crises, but the actual treatment requires myeloablative conditioning, which carries its own substantial morbidity and is not appropriate for patients with significant comorbidities. The downstream reality is that the eligible population is considerably smaller than the prevalent population of the indication, which means that aggregate spending in any given year is more contained than the per-patient price would suggest. This containment is, however, not because the economic problem is solved—it is because the clinical bottleneck happens to be tighter than the financial one.
Manufacturing capacity is a separate constraint that interacts with reimbursement in ways the early policy debate underweighted. Cell therapies in particular, including the CAR-T constructs that pioneered the category, have faced manufacturing bottlenecks that have limited patient access more directly than reimbursement decisions have. A therapy whose authorized patients exceed the manufacturer’s annual production capacity will, in practice, be rationed by waiting list. Reimbursement policy is downstream of this rationing, not a binding constraint on it. As manufacturing capacity expands over the next several years, the reimbursement constraint will become more visible, and the policy conversation will reorient accordingly.
What the cell and gene therapy field has learned, somewhat painfully, is that the durability claims that justify the prices require empirical confirmation that the field has not yet had time to deliver. The longest-followed CAR-T patients have been observed for less than a decade. The first sickle cell gene therapy patients were treated in trial settings beginning around 2018, and the post-marketing population has only just begun to accumulate. If the durability holds, the per-year-of-benefit pricing will look reasonable to most analysts. If the durability falters—if the gene-edited cells revert, if the CAR-T cells exhaust, if the integrated viral payloads silence over time—the entire commercial logic of the therapies fails simultaneously. The investors funding the next generation of gene therapies are, in a sense, making a bet on a hypothesis whose answer will not arrive for another decade.
There is a separate set of considerations that arises specifically in the Medicaid context, where the gene therapy reimbursement question collides with the budgetary realities of state government. Several state Medicaid programs have begun encountering single-quarter spending spikes from gene therapy payments that disrupt their managed care actuarial models, force special legislative appropriations, and in some cases trigger downstream reductions in other Medicaid services. The resulting political dynamic—a single sickle cell patient’s gene therapy displacing routine prenatal care, school-based health services, or non-emergent dental coverage in a state’s Medicaid budget—has produced a kind of political triage that the original gene therapy advocates did not anticipate.
What the gene therapy reimbursement architecture eventually settles into is unlikely to be the elegant outcomes-based contracts that the policy literature spent the late 2010s describing. It is more likely to be a patchwork: federal and state participation in pooled-risk models like the CMS access model for the highest-priority indications; cash-flow restructuring through annuity arrangements where the legal architecture permits; manufacturer-funded patient access programs for the most acute cases; and conventional fee-for-service reimbursement at negotiated discounts for the rest. The patchwork will be inelegant, administratively expensive, and unlikely to deliver the value-based pricing that the field’s advocates promised. It will also, probably, deliver patient access at scales the original gene therapy field would have considered improbable a decade ago. Whether this counts as success depends on which problem one thought the gene therapy revolution was supposed to solve. If the problem was therapeutic—the existence of cures for previously untreatable diseases—the field has delivered. If the problem was reimbursement—the existence of a financing architecture proportionate to the therapeutic novelty—the field is still working it out, and the next several years will be more revealing than the previous five.
