Fianlimab

Fianlimab is Regeneron's anti-LAG-3 monoclonal antibody, developed in combination with cemiplimab (the company's anti-PD-1). On May 15, 2026, Regeneron disclosed that the key Phase 3 HARMONY trial (NCT05352672) in first-line advanced melanoma MISSED its primary endpoint of progression-free survival versus pembrolizumab monotherapy: the high-dose fianlimab + cemiplimab arm posted HR 0.845 (p=0.0627), narrowly missing the prespecified 0.05 threshold, while the low-dose arm posted HR 0.931 (p=0.4661) [8]. Median PFS was 11.5 months on high-dose combo versus 6.4 months on pembrolizumab, but the trial does not meet the statistical bar for approval on this endpoint. The remaining program value sits in a second Phase 3 versus Opdualag in melanoma and the parallel NSCLC Phase 3 program [9]. The original strategic thesis - a proprietary IO doublet to compete with Bristol's Opdualag and Keytruda - is materially damaged in melanoma.

Novel compound. Fianlimab (REGN3767) has never been approved for any indication. The lead Phase 3 (HARMONY, NCT05352672) READ OUT NEGATIVE on May 15, 2026: 1,546 patients randomized across four arms (high-dose fianlimab + cemiplimab, low-dose fianlimab + cemiplimab, pembrolizumab monotherapy, cemiplimab monotherapy), with neither fianlimab arm clearing statistical significance for PFS versus pembrolizumab [4][8]. Regeneron has not committed to an FDA filing on this dataset; overall survival data remain immature and will be a watch item for any accelerated-approval pathway. A separate Phase 3 head-to-head against Opdualag (nivolumab + relatlimab) in 1L melanoma is ongoing [8]. The parallel NSCLC Phase 3 (NCT05785767) testing fianlimab + cemiplimab versus cemiplimab monotherapy in 1L PD-L1≥50% NSCLC remains active and is now the most consequential remaining catalyst for the program [9]. A Phase 2 neoadjuvant NSCLC study (NCT06161441, n=195) - neoadjuvant meaning treatment given before surgery to shrink the tumor - is active but not recruiting, with pathological complete response (pCR; no viable cancer cells found in the surgical specimen) as primary endpoint [5]. Investigator-initiated Phase 2 studies continue in MSI-H colorectal, ovarian, and ipilimumab combinations at MSKCC. No public FDA breakthrough, fast-track, or orphan designations have been disclosed. The strategic posture - fianlimab tested almost exclusively with cemiplimab as a proprietary doublet - now means the asset's fortunes are tightly coupled to the NSCLC readout and the head-to-head vs. Opdualag, not a stand-alone melanoma path.

LAG-3 (lymphocyte activation gene 3) is a brake on T cells. When a T cell encounters certain signals on tumor cells or antigen-presenting cells - mainly MHC class II proteins and a hepatocyte-derived protein called FGL1 - LAG-3 fires and tells the T cell to stand down. Tumors exploit this by displaying these ligands to shut off the immune attack. Blocking LAG-3 with an antibody like fianlimab releases the brake. The catch: LAG-3 alone is a weaker brake than PD-1 (the dominant checkpoint), so single-agent LAG-3 inhibition does very little. The clinical story has always been about combinations with anti-PD-1. The mechanism is validated by approval: Bristol-Myers Squibb's relatlimab + nivolumab combination (Opdualag) won FDA approval in March 2022 for first-line metastatic melanoma based on RELATIVITY-047, where the doublet roughly doubled progression-free survival over nivolumab alone [7]. That approval remains the entire commercial proof point for LAG-3 as a target. Fianlimab was positioned as the next-most-advanced LAG-3 antibody behind relatlimab - but the HARMONY miss now reframes the class question: relatlimab + nivolumab cleared PFS against nivolumab alone, while fianlimab + cemiplimab failed to clear PFS against pembrolizumab. The biology is real, but the magnitude of LAG-3-added benefit on top of an already-strong PD-1 backbone is now in active doubt. Open Targets scores LAG3 at 0.567 in melanoma and 0.304 in NSCLC - non-trivial but not overwhelming, and the HARMONY result is consistent with the lower end of that evidence.

The lead Phase 3 HARMONY (NCT05352672) was a four-arm, 1,546-patient randomized trial in previously untreated unresectable or metastatic melanoma: (1) high-dose fianlimab + cemiplimab, (2) low-dose fianlimab + cemiplimab, (3) pembrolizumab monotherapy (comparator), and (4) cemiplimab monotherapy [4][8]. Primary endpoint was PFS by blinded independent central review, with OS as a key secondary [4]. Pembrolizumab as comparator was the right choice - global standard of care for first-line melanoma without targetable mutations - but it raised the bar substantially. The hazard ratio (HR) is the headline efficacy metric: HR=1.0 means identical disease progression rates between arms, HR=0.65 means 35% slower progression on the experimental arm, and lower is better. Because two fianlimab dose arms were tested simultaneously against the same pembrolizumab comparator, the prespecified p-value threshold had to account for multiplicity - the actual hurdle for the high-dose arm was tighter than the headline 0.05, and the high-dose result (p=0.0627) fell short even before multiplicity adjustment. The cemiplimab monotherapy arm provides a critical internal calibration: it isolates the LAG-3 contribution by comparing the doublet against the PD-1 backbone alone. The supporting Phase 1 data (Hamid et al. JCO 2024) reported ORR of 57% in the combined treatment-naive advanced melanoma cohort (95% CI 47-67%, n=98), with individual cohort ORRs around 61-63% from n=40 each [1] - well above the 40-45% typically seen with pembrolizumab monotherapy, but Phase 1 melanoma cohorts of 40 patients routinely run hot relative to confirmatory trials, and the HARMONY result confirms that pattern. The design contained no LAG-3 expression cutoff or other biomarker enrichment, which RELATIVITY-047 also lacked [7]. The Phase 2 neoadjuvant NSCLC study (NCT06161441, n=195) using pCR as endpoint remains a cleaner, faster signal-finding design [5].

Our model puts this drug's chance of eventual approval at 37%. That starts from the historical approval rate for Phase 3 drugs in this area - about 48% - then shifts up or down based on ten facts about the trial and sponsor. On the positive side, the trial has more secondary endpoints and larger enrollment than typical for this phase; on the negative side, the blinding is heavier than usual and earlier-phase results were weak or limited. The remaining facts were close to average, so they left the estimate near where it began.

The primary efficacy risk has now resolved unfavorably: HARMONY missed PFS in 1L melanoma with HR 0.845 (high dose, p=0.0627) and HR 0.931 (low dose, p=0.4661) [8]. The forward risks are now different. Regulatory risk: Regeneron has not committed to a filing path. A submission attempting to use near-miss PFS plus immature OS is possible but historically unfavorable at FDA without unmet medical need, which a crowded 1L melanoma market does not provide. Class read-through risk: a confirmed PD-1 + LAG-3 doublet (Opdualag, RELATIVITY-047) cleared PFS over PD-1 monotherapy, while fianlimab + cemiplimab failed to do so. This raises the possibility that fianlimab itself, cemiplimab as a backbone, or both, is the weaker component versus the BMS asset - important for the NSCLC Phase 3 (NCT05785767), which uses the same drugs against the same biology in a different tumor [9]. Safety risk is manageable and unchanged: Regeneron stated no new safety signals emerged in HARMONY [8]. Phase 1 had shown a notable rate of adrenal insufficiency with fianlimab + cemiplimab - higher than seen with anti-PD-1 alone, an on-target immune-related toxicity that can be permanent and requires lifelong hormone replacement [1][2]. Commercial risk: Opdualag's modest commercial trajectory in 1L melanoma already showed oncologist hesitance to adopt LAG-3 combinations over single-agent pembrolizumab. A fianlimab + cemiplimab approval - even if achievable - would enter the same crowded space with a numerically weaker Phase 3 dataset than Opdualag had. Execution risk: low. Regeneron runs clean trials, and the HARMONY conduct is not in question - the molecule under-delivered, not the operations.

The Phase 3 miss has resolved the buy/sell question I had previously framed: under the prior thesis, an HR at or below 0.65 with separating OS curves would have been a buy signal - the actual HR of 0.845 is in the 'technically meaningful PFS delta, statistically insufficient, commercially difficult' zone, and in fact did not even clear statistical significance. Three things matter now. (1) Regulatory path: watch for Regeneron's stated filing strategy. The narrow miss (p=0.0627) means OS maturation or a post-hoc biomarker subset could provide a path, but the base-case expectation should be that this dataset does not yield a 1L melanoma label. (2) NSCLC pivot: the parallel Phase 3 NCT05785767 (fianlimab + cemiplimab vs. cemiplimab monotherapy in 1L PD-L1≥50% NSCLC) is now the most consequential remaining catalyst for the entire fianlimab program [9]. The HARMONY result is a negative read-through - the LAG-3 increment over PD-1 monotherapy is empirically smaller than RELATIVITY-047 had suggested, and that increment is exactly what the NSCLC trial needs to demonstrate. The neoadjuvant NSCLC Phase 2 (NCT06161441) reading pCR (no viable tumor cells in resected tissue at surgery) provides a faster, cleaner secondary signal [5]. (3) LAG-3 class implications: Opdualag remains the only approved LAG-3 asset, and HARMONY does not improve confidence in the class beyond the relatlimab + nivolumab specific combination. Stock impact for Regeneron: limited but real. Regeneron's 2025 revenue base (~$14.3B, anchored by Eylea, Dupixent collaboration profits, and cemiplimab [6]) makes fianlimab strategically important but not financially critical - the HARMONY miss removed a peak-sales scenario but does not threaten core franchise economics. Net assessment: the 'track but don't trade' framing has resolved; the fianlimab melanoma story is materially over, and the watch list shifts entirely to NSCLC readouts and any Regeneron commentary on regulatory intent for HARMONY. Do not extrapolate from a recovery narrative on the secondary OS data without seeing it.