The p-Value That Wasn't: What Regeneron's Fianlimab Miss Reveals About the Limits of Dual Checkpoint Blockade in Melanoma

There is a particular kind of clinical trial result that is harder to interpret than a clean failure. Regeneron's Phase 3 HARMONY trial of fianlimab, its LAG-3 inhibitor, in combination with cemiplimab for first-line metastatic melanoma is exactly that kind of result. The trial did not fail. It also did not succeed. And the distance between those two outcomes, measured in a single p-value of 0.0627, is now one of the more consequential data points in immuno-oncology.

On May 15, 2026, Regeneron announced that the Phase 3 trial evaluating two dose levels of fianlimab plus cemiplimab against pembrolizumab monotherapy in patients with unresectable locally advanced or metastatic melanoma did not reach statistical significance for its primary endpoint of progression-free survival improvement. The high-dose combination arm, fianlimab at 1600 mg plus cemiplimab at 350 mg every three weeks, produced a median PFS of 11.5 months compared to 6.4 months for pembrolizumab alone. That is a 5.1-month numeric improvement. The hazard ratio was 0.845. The confidence interval was 0.709 to 1.008. The p-value was 0.0627. The prespecified threshold for statistical significance was not crossed.

What the Numbers Actually Mean

A 5.1-month improvement in median PFS is not a trivial signal. In a disease where pembrolizumab monotherapy has been the dominant first-line standard for years, a combination that extends median progression-free survival by more than five months would, in most contexts, be considered a meaningful clinical advance. The problem is that the trial was not designed to detect a trend. It was designed to detect a statistically significant difference, and the data did not clear that bar. The confidence interval crossing 1.0 means the result is consistent with no benefit at all, even if the point estimate suggests otherwise.

The low-dose arm, fianlimab at 400 mg plus cemiplimab, performed worse. Its hazard ratio was 0.931 with a p-value of 0.4661, a result that offers no encouragement for that dose level. The dose-response relationship, or lack of it, adds a layer of complexity to interpreting the high-dose signal. If the mechanism were working cleanly, one would expect a cleaner separation between dose levels. The fact that the high-dose arm showed a numerically meaningful but statistically insufficient benefit while the low-dose arm showed essentially nothing raises questions about whether the trial was adequately powered, whether the patient population was optimally selected, or whether the combination's benefit is real but more modest than the trial was designed to detect.

The LAG-3 Landscape and What This Changes

To understand the significance of this result, it helps to understand where LAG-3 inhibition stands in the broader checkpoint inhibitor landscape. The only approved LAG-3 combination therapy is Opdualag, Bristol Myers Squibb's fixed-dose combination of nivolumab and relatlimab, which received FDA approval in March 2022 for unresectable or metastatic melanoma. Opdualag's approval was based on the RELATIVITY-047 trial, which demonstrated a statistically significant improvement in PFS versus nivolumab monotherapy, with a median PFS of 10.1 months versus 4.6 months and a hazard ratio of 0.75. That result established the proof of concept for dual LAG-3 and PD-1 blockade in melanoma and set the competitive benchmark that fianlimab was attempting to surpass.

Regeneron's strategy was to pair its own LAG-3 inhibitor with cemiplimab, its approved PD-1 inhibitor, and test the combination not against PD-1 monotherapy in a general population, but specifically against pembrolizumab, the market-leading PD-1 inhibitor. That head-to-head design was ambitious. It required the combination to demonstrate superiority over an already-effective standard of care rather than simply outperforming a single-agent control. The trial enrolled 1,546 patients, a substantial investment in a single hypothesis. The hypothesis was not confirmed.

The Trial That Is Still Running

The fianlimab story is not over. A separate Phase 3 trial, also in first-line unresectable or metastatic melanoma, is evaluating the high-dose fianlimab plus cemiplimab combination directly against Opdualag. That head-to-head design is strategically interesting for a different reason: it tests whether fianlimab's LAG-3 inhibition, paired with cemiplimab's PD-1 blockade, can outperform the only approved LAG-3 combination. If that trial succeeds, Regeneron would have a result that is arguably more commercially meaningful than a pembrolizumab comparison, because it would establish fianlimab as the superior LAG-3 combination rather than simply a competitive one.

The ongoing trial also provides a path for Regeneron to salvage the program's commercial potential even after the HARMONY miss. Oncologists and payers do not make decisions based on a single trial in isolation. A positive result against Opdualag, combined with the numerically encouraging but statistically insufficient HARMONY data, could still support a regulatory submission and a compelling clinical narrative. The question is whether the ongoing trial is powered and designed to deliver the kind of unambiguous result that the HARMONY trial did not.

What This Means for the Broader Field

The HARMONY result is a useful reminder that the checkpoint inhibitor field is not a simple story of incremental progress. Adding a second checkpoint target to a PD-1 inhibitor does not automatically produce a statistically significant benefit, even when the numeric signal is encouraging. The biology of LAG-3 inhibition is real, as Opdualag's approval demonstrates. But the magnitude of benefit from LAG-3 blockade appears to be more modest than the field initially hoped, and translating that modest benefit into a statistically significant result in a large, well-powered trial requires either a larger effect size or a more precisely defined patient population.

That population question is one the field has not yet answered satisfactorily. LAG-3 expression on tumor-infiltrating lymphocytes is a plausible biomarker for predicting response to LAG-3 inhibition, but it has not been validated as a patient selection tool in the way that PD-L1 expression has been used to enrich populations for PD-1 inhibitor trials. If fianlimab's benefit is concentrated in a LAG-3-high subpopulation, a trial that enrolls an unselected population will dilute that signal. The HARMONY data, when presented in full at an upcoming medical meeting, may shed light on whether any subgroup showed a more pronounced benefit that could inform future trial design.

For Regeneron, the immediate financial impact is limited. Cemiplimab has established commercial value across multiple indications, and the company's pipeline is broad enough to absorb a single Phase 3 miss without existential consequences. But the HARMONY result does narrow the near-term regulatory pathway for fianlimab in melanoma and shifts the program's fate to the ongoing Opdualag comparison. That trial's outcome, expected in the coming years, will determine whether fianlimab becomes a meaningful addition to the melanoma treatment landscape or a cautionary example of how close a drug can come to statistical significance without crossing the line that matters.

In oncology, 0.0627 is not a failure. It is an invitation to ask harder questions about trial design, patient selection, and the biology of the target being inhibited. The field would do well to accept that invitation before the next LAG-3 trial reads out.