The Miller Lab The Miller Lab The Miller Lab
  • Our Approach
  • Meet The Team
  • Publications
  • Presentations
  • Posts
  • Software
  • Funding
  • Slides
  • Donate

Persuaded, Not Certain: Why I Voted Yes

David M. Miller

2026-08-07

Editorial · Journal of Cutaneous Oncology

Persuaded, Not Certain: Why I Voted Yes

Reflections on the FDA advisory committee review of RP1 plus nivolumab after progression on PD-1–based therapy—and on how to make a consequential regulatory decision when the evidence is meaningful but imperfect.

Melanoma RP1 + Nivolumab FDA Advisory Committee Accelerated Approval Regulatory Science
Journal Journal of Cutaneous Oncology
Published August 7, 2026
Article type Editorial · Volume 4, Issue 2

Citation

Miller DM. Persuaded, Not Certain: Why I Voted Yes. Journal of Cutaneous Oncology. 2026;4(2). doi:10.59449/joco.2026.08.07.

View DOI Read Full Article FDA Advisory Committee Materials
Post-Publication Context

On August 6, 2026, after the substantive analysis in this editorial was completed, the FDA granted accelerated approval to RP1 plus nivolumab for the indication discussed here. The reflections below are based on the evidentiary record available to the July 30 advisory committee.

Why this editorial matters

Regulatory decisions rarely occur in a world of perfect evidence. The July 30, 2026 FDA advisory committee meeting on RP1 plus nivolumab forced a particularly difficult question: whether a single-arm dataset with substantial methodological limitations still contained an efficacy signal that was sufficiently evaluable and clinically meaningful to support accelerated approval.

The editorial explains why the FDA’s concerns were persuasive, why those concerns did not disappear, and why—after discounting the evidence most vulnerable to bias—a meaningful treatment signal still remained.

Central Perspective

A yes vote did not require certainty, nor did it require accepting the sponsor’s most favorable interpretation of the data. It required deciding whether clinically meaningful activity remained after the weaknesses of the trial were taken seriously.

One trial, three response rates

Applicant analysis 33.6%

47 responders

The published analysis counted overall RECIST responses while patients received RP1 plus nivolumab.

FDA primary analysis 15.7%

22 of 140 patients

FDA focused on patients in whom systemic activity could be assessed through noninjected target lesions while retaining the full denominator.

FDA sensitivity analysis 24.7%

22 of 89 evaluable patients

Patients without evaluable noninjected target lesions were excluded from the denominator.

The Estimand Problem

The question was not simply, “What was the response rate?” Each estimate answered a different question about what counted as evidence of systemic activity. The disagreement was fundamentally about the estimand, the denominator, and the claim the data could support.

Why injected lesions complicated interpretation

Local control is clinically meaningful

Durable regression of an injected lesion can prevent bleeding, ulceration, pain, infection, functional compromise, surgery, or radiation. Local benefit should not be dismissed simply because it is local.

But local response is not the same as systemic activity

A directly injected lesion receives an exposure that distant disease does not. A broad melanoma indication reasonably requires evidence that benefit extends beyond the lesions selected for injection.

Why FDA’s skepticism was persuasive

  • IGNYTE was not originally designed as a pivotal trial and its protocol evolved substantially.
  • There was no concurrent control capable of isolating the contribution of RP1 from nivolumab continuation.
  • Treatment beyond progression, reinjection of lesions, missing assessments, surgery, and biopsy complicated response classification.
  • The published 33.6% response rate therefore could not be treated as a simple, fixed estimate of systemic efficacy.
  • The available historical controls differed in patient population, prior therapy, resistance definitions, and response-assessment methods.

Why I still voted yes

Even after applying FDA’s more conservative analytical framework, the remaining signal included deep and durable responses, regression in noninjected disease, activity in visceral metastases, and responses among patients with clinically difficult forms of PD-1 resistance.

The timing of prior therapy also mattered. Patients had recently progressed on anti–PD-1 treatment, making this less like a distant rechallenge and more like progression during essentially continuous PD-1 blockade followed by the addition of RP1. In that context, nivolumab continuation alone seemed unlikely to explain the entire response pattern.

Contribution of Components

IGNYTE did not prove synergy. But the combination was biologically coherent: RP1 alone had not demonstrated activity approaching that observed in IGNYTE, while patients had already progressed on PD-1 blockade. The most plausible interpretation was that both components contributed to the regimen.

The historical-control problem

Replimune emphasized an expected response rate of roughly 6% to 7% with continued PD-1 therapy after progression. FDA argued that no historical control was sufficiently comparable to IGNYTE to make that benchmark reliable.

That criticism was valid. Yet uncertainty about the exact counterfactual does not make every possible counterfactual equally plausible. Data from MASTERKEY-115, which prospectively distinguished metastatic PD-1 resistance from recurrence after adjuvant therapy, provided useful context and supported the view that conventional responses after established metastatic PD-1 resistance are uncommon.

The accelerated-approval context

A provisional decision

Accelerated approval permits action on an endpoint reasonably likely to predict clinical benefit while requiring subsequent confirmation. It does not eliminate the requirement for substantial evidence.

A randomized answer is still needed

IGNYTE-3 is intended to answer the question the single-arm study could not. A negative confirmatory result would directly challenge continued approval.

What testimony added—and what it could not

Patient testimony did not establish efficacy, but it made treatment burden, limited alternatives, and the practical meaning of durable disease control impossible to reduce to abstract endpoints alone.

Expert testimony similarly provided information that was clinically meaningful but incompletely measured: the pace and coherence of response, how distant lesions behaved, and how unusual certain outcomes appeared relative to ordinary experience after PD-1 failure.

Such evidence is vulnerable to recall, selection, attribution, and social influence. It cannot repair a flawed design. But acknowledging that it influenced judgment is more transparent than pretending consequential decisions occur in a social vacuum.

The Decision Test

Would I recommend RP1 plus nivolumab to an appropriate patient progressing on PD-1–based therapy? Yes. If I were that patient, fully informed of the uncertainty and alternatives, would I consent to receive it? Yes.

What remains uncertain

  • The precise magnitude of the treatment effect remains uncertain.
  • Overall survival from the single-arm study is not readily interpretable as a treatment effect.
  • The individual contribution of RP1 and nivolumab was not definitively established.
  • Response frameworks for intratumoral therapy need to prospectively distinguish local control from systemic activity.
  • The randomized confirmatory trial remains essential to determine whether the advisory committee’s judgment was correct.
The Practical View

The FDA review succeeded in making the uncertainty visible. But uncertainty is not the same as absence of evidence. After discounting the responses most vulnerable to methodological bias, I concluded that a clinically meaningful signal remained—persuasive enough to vote yes, but not strong enough to justify certainty.

Author

David M. Miller.

This page highlights selected elements of the published editorial rather than reproducing the complete article. See the original publication for the full analysis, regulatory context, references, disclosures, and discussion. This site represents our opinions only. See our full Disclaimer and Terms of Use Agreement.