How to Validate an IO Biomarker in a Clinical Trial: Expert Guidance

How to Validate an IO Biomarker in a Clinical Trial: Expert Guidance

You have a promising IO biomarker. You have data behind it. Now comes the harder part: getting it through a clinical trial and into a form that regulators will actually accept.

This is where most programs struggle. Not because the science is weak, but because the strategy around it is built too late, defined too loosely, or structured in a way that cannot survive the move from discovery into a regulated trial context.

We wrote this guide to help you avoid that. Whether you are planning your first biomarker substudy or refining a strategy ahead of a pivotal trial, here is what the validation process actually requires and how to do it well.

Key takeaways

• Your COU statement is the foundation. Write it before your IND and let it drive every analytical and clinical validation decision that follows.
• Analytical and clinical validation are separate requirements. A validated assay does not equal a validated biomarker.
• Biomarker strategy has to be embedded in trial design from the start. A program retrofitted after the protocol is locked will not produce regulatory-grade evidence.
• Early, specific FDA engagement changes what is possible. Arriving at pre-IND or Type C meetings with a clear plan—not just data—is what makes those conversations productive.

Why most IO Biomarker programs fall short

Only three IO biomarkers have received clear FDA validation: PD-L1, TMB, and MSI. Even these come with significant limitations. PD-L1 is predictive in fewer than 30% of the FDA approvals where it was used. Fewer than one in three patients on immune checkpoint inhibitors achieves a durable response.

The shortfall is rarely a drug problem. More often, it reflects a biomarker strategy problem.

Assay characterization happens without a defined regulatory purpose. Biomarker hypotheses get generated after outcome data is already in hand. Sample collection fails across sites and there is no recovery plan. The biomarker-treatment interaction that was supposed to prove predictive value turns out to be a prognostic signal that was present all along.

These are fixable problems. All of them stem from decisions made too late or not made at all.

Step 1: Write your context of use statement first

The Context of Use (COU) statement is the foundation of your entire validation program. It defines what your biomarker is supposed to do, in which patient population, at what stage of drug development, and for what specific regulatory purpose.

Every downstream decision flows from it: which assay to run, how to power your substudy, what pre-analytical controls you need, and what FDA will require from you at each stage.

Step 2: Separate analytical from clinical validation

Analytical validation proves your assay works. Clinical validation proves your biomarker works for its intended clinical purpose. These are two separate things, and having one does not give you the other.

Pre-analytical variables deserve particular attention. How samples are collected, frozen, shipped, and processed directly affects assay performance.

Step 3: Know whether your biomarker is predictive or prognostic

A prognostic biomarker tells you about disease outcome regardless of treatment. A predictive biomarker tells you which patients will benefit from your specific treatment.

To validate a predictive claim, you need a control arm, pre-specified biomarker hypotheses, and statistical evidence of a biomarker-by-treatment interaction.

Any analysis not pre-specified in your protocol will be classified by FDA as exploratory. Exploratory data can support a hypothesis. It cannot validate a biomarker for labeling.

Step 4: Build the biomarker substudy into trial design from the start

The biomarker program cannot be attached to a trial that is already designed. The major decisions need to be in place before your first patient is enrolled.

• Lock your assay before enrollment begins
• Power the substudy for the biomarker endpoint
• Account for sample attrition
• Establish your central lab infrastructure early

Step 5: Address tumor heterogeneity as a design problem

Tumor biology does not cooperate with the assumption of a uniform, stable biomarker signal. PD-L1 expression, TMB, and most tumor microenvironment markers vary across regions of the same tumor, between primary and metastatic sites, and over time as patients receive treatment.

Circulating tumor DNA (ctDNA) offers a partial answer. Early ctDNA clearance correlates with improved survival in IO trials across multiple tumor types.

For your trial, the practical choices are serial sampling to capture longitudinal change, liquid biopsy as a complement to tissue, and clear pre-specification of which sample type is the primary biomarker source for your endpoint.

Step 6: Engage FDA early and with a specific plan

These conversations are far more productive when you come prepared with a written COU, an analytical validation plan, and a clear articulation of what regulatory claim you are trying to support.

Step 7: Harmonize across sites before the trial starts

Running a biomarker substudy across global trial sites is a fundamentally different challenge from validating an assay in a single central lab.

For liquid biopsy, pre-analytical standardization is critical. Tube type, centrifugation protocol, time from collection to processing, freeze conditions, and shipping temperature all affect ctDNA yield and quality.

Harmonization is not a lab operations problem. It is a trial design problem. Build it into your protocol from day one.

Get the IO Biomarker validation checklist

We built a step-by-step checklist that covers every stage of IO biomarker validation, from writing your COU to assay lock, substudy powering, sample logistics, site harmonization, and FDA engagement planning.

Download the checklist