Replimune's cancer drug RP1 has now been rejected by the FDA twice, and it goes before an advisory committee this week for a third verdict, due August 2. The survival data behind it look genuinely good: in patients with advanced melanoma whose cancer had already progressed on the standard immunotherapy, 47.8% were alive at three years, rising to 83.5% among those who responded, with a median overall survival of about 33 months. For a group with few options, those are numbers worth taking seriously.
And yet the FDA, in its staff review, called the data package not interpretable. Understanding how good-looking survival numbers and "not interpretable" can both be true is the key to this case, and it turns on a problem that has nothing to do with whether the drug is safe and everything to do with what a single trial can prove.
The core dispute is about proof, not performance
The disagreement here is not the usual one about whether a drug's benefit outweighs its risks. On safety, RP1 has not been the sticking point; the company has said the FDA did not raise safety as the central issue, and the adverse events reported were manageable. The fight is narrower and more fundamental: whether the trial Replimune ran can actually demonstrate that RP1 works.
The trial, called IGNYTE, was single-arm. That means every patient in it received the drug, and there was no comparison group of similar patients who did not. Everyone got RP1 combined with nivolumab, an established immunotherapy, and the researchers measured how they did. There was no control arm to measure them against.
That design is the entire problem, and it is worth being precise about why, because it is not a technicality the FDA is hiding behind.
Why a single-arm trial struggles to prove anything
When every patient gets the drug and there is no control group, you can see what happened to the patients, but you cannot know what would have happened without the drug. Some melanoma patients survive a long time for reasons that have nothing to do with any particular treatment, and without a comparison group, you cannot separate the patients the drug helped from the patients who would have done well anyway.
This is why the gold standard in drug testing is the randomized controlled trial, where similar patients are split into two groups, one getting the drug and one not, so that any difference between the groups can be attributed to the drug. A single-arm trial removes that comparison, and with it much of the ability to prove cause and effect. Good outcomes in a single-arm trial are suggestive. They are not proof, because the thing that would turn them into proof, the counterfactual of what those same patients would have done otherwise, was never measured.
Single-arm trials are sometimes accepted, usually when a disease is so uniformly and rapidly fatal that any meaningful survival is self-evidently due to the treatment, since untreated patients would essentially all have died quickly. Advanced melanoma after anti-PD-1 failure is serious, but it is not that kind of uniformly fatal setting, and some patients do live a long time, which is exactly what makes a no-control-group design hard to interpret here.
The "contribution of effect" problem makes it harder
There is a second layer that compounds the first, and it is the specific objection the FDA has emphasized. RP1 was not given alone. It was given together with nivolumab, a drug that is already approved and already works in melanoma on its own.
That creates a genuine puzzle. If patients receiving two active drugs do well, how much of the benefit came from RP1, the drug seeking approval, and how much from nivolumab, the drug that was already known to help? This is called the contribution-of-effect problem, and it is the crux of the FDA's two prior rejections. When you combine a new drug with one that already works and give the pair to everyone with no comparison, you cannot cleanly separate the new drug's contribution from the established one's. The good results might reflect RP1 adding real benefit, or they might substantially reflect what nivolumab does anyway.
So the FDA is facing a dataset with two overlapping interpretive gaps: no control group to establish that the combination beat no treatment, and no arm testing nivolumab alone to establish that RP1 added anything on top of it. Either gap alone weakens the proof. Together they are why the agency reached for the word "not interpretable," which is a strong term meaning not that the drug failed, but that the study as designed cannot answer the question approval requires.
The genuine tension, both sides fairly stated
This is not a case of an obstructive agency versus a wronged company, and it is not a case of a company trying to sneak a bad drug through. Both positions are serious, and the difficulty is real.
The company's case is substantive. Advanced melanoma that has progressed on anti-PD-1 therapy has few good options, the responses seen in IGNYTE were durable, lasting a long time in those who responded, and durable responses in a hard-to-treat cancer are unlikely to be pure chance. Patients and melanoma advocates argue that when people are out of options, suggestive evidence of durable benefit should be enough for accelerated approval, especially with a confirmatory trial already running to settle the question later. That is a coherent argument, and the FDA's decision to grant an expedited review and convene the panel reflects that the agency takes the unmet need seriously.
The FDA's case is equally substantive. Approving a drug on uninterpretable data sets a precedent that weakens the evidence standard for everyone, and the standard exists because uncontrolled results have repeatedly looked promising and then failed when properly tested. If RP1 is approved on data that cannot show its specific contribution, and later turns out to add little over nivolumab alone, patients will have been exposed to an extra therapy, and its cost and side effects, for benefit that was never real. The agency's job is to be sure the drug works, and single-arm combination data cannot make it sure.
Both of these are true at once, which is what makes the advisory committee genuinely difficult rather than a formality.
What the committee is actually weighing
The panel is not being asked whether melanoma is dangerous or whether these patients deserve options; everyone agrees on both. It is being asked a narrower question: whether suggestive but uninterpretable data, in a setting of real unmet need, clears the bar for accelerated approval, a pathway explicitly designed to allow promising drugs to reach desperate patients earlier on less-than-definitive evidence, with confirmation to follow.
That framing is why the case is a close call rather than an obvious rejection. Accelerated approval exists precisely for situations where the evidence is incomplete but the need is urgent, so the question is whether RP1's data, incomplete in the specific way described, is incomplete enough to fail even that lower bar. Replimune has a confirmatory Phase 3 trial, IGNYTE-3, underway, which is the mechanism meant to resolve the interpretability problem after the fact. The committee has to decide whether to allow access now on the strength of suggestive data and the promise of confirmation, or to hold the line until the controlled trial reports.
The stakes for the company are stark: Replimune has said that failure to win accelerated approval could force it to halt RP1 development in melanoma and consider restructuring. That raises the human temperature but should not change the scientific question, and the committee's task is to keep the two separate.
How to read it
The honest way to hold this is that the survival numbers and the "not interpretable" verdict are both accurate, because they answer different questions. The numbers describe what happened to the patients in the trial, which was, for many of them, encouraging. The verdict describes whether the trial can prove the drug caused it, which, because of the missing control group and the combination design, it cannot do cleanly. A drug can produce good-looking results in a study that is nonetheless unable to prove the drug produced them, and that is the situation here.
For patients following this, the tension is painful and worth naming plainly: the evidentiary standard that looks like an obstacle is the same standard that protects people from being given treatments that do not actually help. A drug that genuinely works would eventually clear it. The tragedy of a single-arm combination trial is that it leaves everyone, the company, the FDA, and the patients, unable to know from this data alone whether RP1 is a real advance or a good-looking result that a controlled trial would deflate. That uncertainty, not a villain on either side, is what the committee has to weigh this week, and it is weighing it for a drug whose confirmatory trial has not yet reported. The numbers earned RP1 its hearing. Whether they can carry an approval is the question that has now been asked three times, and the answer still depends on evidence the current trial was not built to provide.
Primary sources
- BioSpace for the FDA staff review calling the data package "not interpretable," the two prior rejections, and the Cellular, Tissue, and Gene Therapies Advisory Committee meeting this week.
- OncLive and Targeted Oncology for the IGNYTE trial design, the 33.6% blinded-central-review objective response rate, the 33.7-month median duration of response, and the accelerated-approval basis.
- StockTitan and Dermatology Advisor for the 47.8% and 83.5% three-year overall survival figures, the 32.9-month median overall survival, and the contribution-of-components question in the prior complete response letters.
- Dermatology Times for the regulatory history including FDA leadership disagreements over demonstrating RP1's contribution of effect, and IGNYTE-3 as the confirmatory trial.
- Replimune's 10-K for the company's warning that failure to gain accelerated approval could halt RP1 melanoma development and prompt restructuring, and its emphasis that the FDA did not raise safety as the central issue.