Coverage With Evidence Development: Paying for a Treatment While the Proof Is Still Coming In

Coverage with evidence development is a conditional yes. A health payer agrees to fund a treatment before its benefit is fully proven, on the strict condition that data keep flowing from everyday care to show whether the therapy really works and for whom. Think of it as a lease rather than a purchase: access starts now, the meter of evidence runs, and a scheduled review decides whether the arrangement continues. The bet is that a few years of real-world data will either confirm the choice or overturn it.

Key points#

• A payer funds a promising treatment early, in exchange for a promise that evidence will continue to be gathered in routine practice.
• It exists because many new therapies for serious conditions reach the market on thin proof: small single-arm trials and short-term surrogate endpoints rather than long randomized outcomes.
• The same idea goes by several names, including managed entry agreements, conditional reimbursement, and performance-based or risk-sharing arrangements.
• The real difficulty is not starting the arrangement but ending it, because reversing coverage once patients depend on a therapy is disruptive and rarely happens.
• The value of any such deal rests on two things: whether the data being collected can actually answer the question, and whether the promised review has real teeth.

The gap that creates the dilemma#

Start with the problem payers are trying to solve. For a serious illness with few alternatives, delay is not neutral; it is measured in patients who wait. Regulators increasingly clear treatments through accelerated routes, and a 2024 review in the Orphanet Journal of Rare Diseases describes what that leaves behind for whoever pays the bill. Approvals often rest on small single-arm studies rather than head-to-head randomized trials, and they lean on surrogate or intermediate endpoints instead of the long-term outcomes you and your clinician actually care about. The therapy may genuinely help. Yet how long the benefit lasts, how it compares with existing care, and what it will cost the system over time can all be honestly unknown on the day it launches.

A payer staring at that gap has three moves. Say no until stronger proof exists, and accept that some patients miss a treatment that might have worked. Say yes with no strings, and accept that some money will fund therapies that turn out to be no better than cheaper options. Or say yes on conditions, opening access while requiring the maker to produce the missing evidence. Coverage with evidence development is that middle path.

What the arrangement looks like in practice#

Different health systems build the conditional yes in different ways, but the skeleton is the same everywhere: a start date, a data plan, and a clock. A 2025 rapid review from the Institute of Health Economics surveys how this plays out across countries.

In the United States, Medicare has used coverage with evidence development to limit payment to patients enrolled in approved studies, with a stated intent to revisit the decision once results arrive. The review points to a monoclonal antibody for Alzheimer's disease as an example, where continued Medicare coverage was tied to taking part in real-world evidence studies.

Beyond the United States, the same logic often travels under the label managed entry agreement, defined broadly as an arrangement that allows a technology to be introduced in a controlled way, frequently with risk-sharing, payment linked to performance, or a time-limited recommendation. The Institute of Health Economics report describes CAR-T cell cancer therapies granted conditional reimbursement in France, England, and Scotland, each requiring registry data to inform a later reassessment. England has funded treatments through its Cancer Drugs Fund with real-world follow-up attached, and Scotland's route for ultra-rare conditions builds in a review after three years. The report also notes tumor-agnostic cancer drugs that first drew negative recommendations on limited trial data and were later reimbursed on conditions once further real-world evidence supported a fresh submission.

Why paying first is a genuine trade-off, not a free lunch#

The upside writes itself. Patients reach a treatment sooner, and the system learns how it performs outside the tidy world of a trial, in the more complicated population that actually receives it. But the costs deserve equal billing.

The first cost is that the evidence being asked for is hard to produce well. The Orphanet review lays out why real-world data for these decisions is methodologically demanding. Without randomization, patients on different treatments tend to differ in systematic ways, so selection bias and confounding threaten any comparison. Rare diseases make this worse, because small and varied patient groups leave little room for the statistical adjustments that might otherwise correct for those differences. Comparisons against historical or outside control groups wobble when diagnostic standards and usual care drift over the years. Registries and claims records often carry gaps and missing fields, and the routine data-quality checks that are standard inside a trial are not standard outside one. The Institute of Health Economics review adds that long timelines, heavy administrative work, and the difficulty of lining up incentives between payers and manufacturers all slow evidence generation, which is part of why uptake of outcome-based deals has stayed low.

The second cost is fragmentation. Because each payer sets its own terms, one product can face annual review in one country, a three-year reassessment in another, and a five-year reassessment in a third, each demanding its own registry. A study built to satisfy one system may not answer another's question, so work is duplicated and results become hard to compare across borders.

The hard part is the exit#

Here is the crux, and it is where the lease analogy strains. A conditional yes is far easier to grant than to take back. Once patients are settled on a therapy, withdrawing coverage is disruptive and deeply unpopular, and the evidence that trickles in is often too weak to force the question either way. The Orphanet review points to a sobering record: among United States programs of this kind, only a handful ever retired the data requirement or reversed coverage, and a look at Cancer Drugs Fund reassessments found that real-world data played only a limited role in settling the original uncertainty.

The implication is uncomfortable. A tool justified largely by its promise to correct mistakes may rarely correct them in practice. If the scheduled review has no teeth, the arrangement drifts into permanent coverage wearing the costume of a trial, and the uncertainty it was meant to manage simply lands on the payer for good.

What separates a real conditional yes from a hopeful one#

Both reviews land on the same design lessons, and they are worth holding onto whether you are a clinician, a patient, or a policymaker.

Line up expectations early. The Institute of Health Economics report stresses agreement up front among regulators, health technology assessment bodies, and payers about exactly what the evidence must demonstrate, backed by transparent protocols, solid data infrastructure, and clear roles for everyone involved.

Build the study to actually answer the question. The Orphanet review offers a parallel checklist: state the intended use before starting, set the study length against the risk of patients dropping out, justify the chosen outcome measures with validated instruments, pick methods that adjust for confounding, and settle who funds and owns the data.

Keep a real off-ramp. Underneath all of it sits one condition that cannot be waived: a credible, enforceable moment at which the decision is genuinely reopened, with the answer "no" still on the table.