RELATIVITY-047’s Five-Year Data Just Made LAG-3 Biomarker Selection a Clinical Trial Design Imperative

Pull up the five-year data from the RELATIVITY-047 phase 3 trial, published in Nature Medicine, and one number stops you cold: the reported survival benefit in patients with LAG-3-high CD8+ T cell expression. The trial randomized patients with unresectable or metastatic melanoma to relatlimab plus nivolumab versus nivolumab alone, and across the full intention-to-treat population, the combination already outperformed monotherapy on progression-free survival. But the biomarker subgroup data tells a sharper story, and the industry hasn’t fully reckoned with what it demands operationally.

This five-year readout lands at a moment when three converging signals are pushing immuno-oncology trial design toward a breaking point. Sponsors are designing LAG-3 programs the same way they designed PD-1 programs fifteen years ago: enroll broadly, hope the signal clears, then do the biomarker archaeology post-hoc. RELATIVITY-047’s mature data makes that approach look not just inefficient, but strategically reckless. Call this pattern the Biomarker Selection Debt, and the bill is coming due.

The Signal RELATIVITY-047 Is Actually Sending

Bristol-Myers Squibb received FDA approval for Opdualag in March 2022, based on the initial RELATIVITY-047 progression-free survival data. That approval, for adult and pediatric patients 12 years and older with unresectable or metastatic melanoma, was a landmark: the first LAG-3 inhibitor combination approved in the United States. The label was broad. The biomarker strategy was not prospectively enriched. And for a first approval, that was acceptable regulatory logic.

Five years later, the data architecture looks different. The RELATIVITY-047 biomarker analyses, reported in this Nature Medicine publication, identify LAG-3-high CD8+ T cell tumor infiltration as a signal that concentrates the survival benefit. This matters because LAG-3 expression measured by immunohistochemistry has already shown clinical validation potential: research published in Oncotarget using the D2G4O clone found LAG-3 expression significantly higher in responders to anti-PD-1 plus anti-LAG-3 combination therapy versus non-responders (p = 0.0210), with an 84% response rate in the LAG-3-positive cohort. The RELATIVITY-047 five-year data now gives that signal a long-horizon survival context that earlier studies could not.

What makes this a trend signal rather than a single data point is the competitive context. In CheckMate-067, published five-year overall survival rates for nivolumab plus ipilimumab versus nivolumab alone in advanced melanoma. That benchmark established dual checkpoint blockade as a viable long-term strategy. RELATIVITY-047’s five-year readout now asks whether LAG-3 inhibition, particularly in biomarker-selected populations, can approach or exceed that bar with a more favorable tolerability profile. The answer from the data is that the unselected population benefit is real, but the biomarker-defined population benefit is where the durable argument lives.

The Competitive Landscape Forces the Question

Opdualag has reported substantial worldwide revenue in 2024 and holds a significant share of the U.S. metastatic melanoma market. Those are not numbers that suggest a drug in trouble. But the next LAG-3 entrant, fianlimab, is already in clinical development, and every competing sponsor has watched RELATIVITY-047’s trial architecture and learned from it.

Here is the counterintuitive framing the industry keeps avoiding: broad enrollment in checkpoint combination trials does not protect market position. It erodes it. When a competitor designs a prospectively biomarker-selected LAG-3 trial and shows a cleaner hazard ratio in a defined population, the FDA’s own enrichment guidance framework becomes the mechanism by which that sponsor stakes out a differentiated label claim. The FDA’s guidance on enrichment strategies defines predictive enrichment as the prospective use of patient characteristics to select populations where detection of a drug effect is more likely. Read that definition against RELATIVITY-047’s post-hoc biomarker analysis and the gap is visible: the evidence exists to support prospective selection, but the trial was not designed that way from the start.

That gap will not remain academic for long. Fianlimab’s clinical program enters a landscape where RELATIVITY-047 has already generated five years of mature survival and biomarker data. Any sponsor designing a LAG-3 trial today without prospective CD8-high or LAG-3-high patient selection is making a deliberate choice to compete on the same unselected ground that BMS occupied in 2019, while the field has moved.

What This Demands From Trial Operations Right Now

For sponsors, the RELATIVITY-047 five-year data reframes the protocol design conversation from “do we need a biomarker strategy?” to “how do we operationalize one without destroying enrollment timelines?” Prospective LAG-3 immunohistochemistry screening using validated assays adds site-level complexity, sample handling requirements, and central laboratory turnaround pressure. Sites that have not run prospective IHC-stratified oncology trials will need training and infrastructure support that standard site qualification visits do not assess. The FDA’s enrichment framework is clear on what predictive enrichment requires methodologically; the operational burden of executing it across a global Phase 3 network is where sponsor teams typically underestimate scope.

For CROs, this trend demands a reckoning with companion diagnostic partnership timelines. Retrospective biomarker analysis, which RELATIVITY-047 used to generate the CD8-high LAG-3 signal now driving the field’s attention, is a discovery exercise. Prospective stratification requires a validated, analytically locked assay available at screening, not at database lock. The lag between biomarker discovery and assay validation can extend well beyond a year in immuno-oncology programs. A CRO that cannot integrate companion diagnostic development into its Phase 2 to Phase 3 transition planning is not competitive in this segment of the market anymore.

Technology vendors serving clinical data management and RTSM platforms face a related pressure. Prospective biomarker-stratified randomization requires real-time central laboratory data flowing into the IRT system before a patient can be randomized. In decentralized or hybrid trial architectures, where sample logistics are already a friction point, that data pipeline needs to be designed into the protocol from the IND stage. Bolt-on solutions added mid-study after the biomarker data becomes compelling are the operational equivalent of retrofitting a bridge while traffic is running.

The broader pattern, the Biomarker Selection Debt, accumulates when sponsors treat post-hoc biomarker analyses as trial design inputs for future programs rather than as protocol requirements for the program currently in design. RELATIVITY-047 paid that debt forward: its five-year data now funds the scientific rationale that the next LAG-3 sponsor needs to justify prospective enrichment to an FDA reviewer. The question facing every immuno-oncology program director today is whether they are spending that currency or accumulating more debt.

If and when a LAG-3 trial designed with prospective CD8-high IHC stratification as a primary randomization criterion reads out, it will either show a sharper hazard ratio than RELATIVITY-047’s full-population result, or it won’t. If it does, every broad-enrollment LAG-3 program in Phase 2 will face a protocol amendment conversation that should have happened at the IND. If it doesn’t, the field will debate whether LAG-3 expression is truly predictive or merely prognostic, and that debate will take place in FDA advisory committee rooms, not in oncology conference halls. Either outcome resets the evidentiary bar, and every sponsor currently running on the assumption that five-year overall survival data alone justifies broad enrollment is about to find out which side of that bar they landed on.

References

1. Nature Medicine, “Relatlimab plus nivolumab versus nivolumab for advanced melanoma: 5-year survival and biomarker analyses of the phase 3 RELATIVITY-047 trial”
2. ASCO, “FDA Approves Opdualag for Unresectable or Metastatic Melanoma” (March 18, 2022)
3. PubMed/NEJM, CheckMate-067: Five-year overall survival rates for nivolumab plus ipilimumab in advanced melanoma
4. PMC/Oncotarget, LAG-3 IHC (D2G4O clone) expression and response rates in anti-PD-1 plus anti-LAG-3 combination therapy
5. Bullfincher, Bristol-Myers Squibb Opdualag worldwide revenue 2024 ($928 million) and U.S. market share data
6. ImmunoMark Summit, FDA enrichment strategy guidance framework: predictive enrichment definition and categorization