Biomarker strategy is one of the most consequential decisions made in early-stage drug development, and one of the most commonly deferred. The reasoning is understandable. Early programmes have limited data, limited budgets, and a long list of competing priorities. Biomarker work can feel like a refinement for later stages, something to address properly once the core mechanism is better understood. This is a mistake, and it is one that consistently costs companies time, money, and in some cases the programme itself.
A well-designed biomarker strategy does several things simultaneously. It sharpens the scientific rationale for the asset. It defines the patient population most likely to respond. It generates evidence that satisfies regulators and informs clinical trial design. And it builds the kind of credibility with investors and partners that generic efficacy data rarely achieves on its own. Getting biomarker strategy right early is not a scientific luxury. It is a commercial necessity.
What a biomarker strategy actually needs to answer
The starting point is clarity on what the biomarker is intended to do. The regulatory and scientific communities distinguish between several biomarker types, and the distinction matters for how a strategy is designed and resourced.
A prognostic biomarker stratifies patients by likely disease outcome independent of treatment. A predictive biomarker identifies patients most likely to respond to a specific treatment. A pharmacodynamic biomarker demonstrates that the treatment is engaging its intended target. A surrogate endpoint biomarker serves as a proxy for a clinical outcome, with the implication that changes in the biomarker are reasonably likely to predict clinical benefit.
Each of these requires a different evidence base, a different validation approach, and a different regulatory conversation. Companies that conflate them, or that develop biomarker data without being explicit about which category they are building evidence for, typically find themselves in difficulty when they reach late-stage development or regulatory submission. The FDA and EMA are precise about these distinctions. Early-stage programmes should be equally precise.
The qualification question
Biomarker qualification is the process by which a biomarker is formally accepted by regulators as fit for a specific purpose in drug development. A qualified biomarker can be used across programmes without requiring de novo validation each time. The FDA’s biomarker qualification programme and the EMA’s parallel process exist specifically to address this, and engagement with these processes, even at early stages, can significantly accelerate later development.
Most early-stage companies are not in a position to pursue formal qualification for a novel biomarker. The evidentiary bar is high and the process is long. But understanding where a biomarker sits on the qualification spectrum matters for how it is communicated to investors and regulators. A biomarker supported by mechanistic rationale and early clinical data is a different proposition from a qualified biomarker, and presenting the two equivalently is a credibility risk that sophisticated reviewers will identify quickly.
The more practical question for most early-stage programmes is whether an existing qualified or accepted biomarker can serve the programme’s needs, either as a primary endpoint, a patient selection tool, or a safety signal. The landscape of established biomarkers in oncology, immunology, and neurology is extensive, and building on existing regulatory acceptance is almost always faster and more cost-effective than generating de novo validation data for a novel biomarker.
Analytical validation comes before clinical validation
A common sequencing error in early biomarker development is moving to clinical validation before analytical validation is complete. Analytical validation establishes that the assay measuring the biomarker is accurate, precise, reproducible, and fit for purpose in the specific sample type and clinical context in which it will be used. Clinical validation establishes that the biomarker is meaningfully associated with a biological or clinical state. The two are distinct, and clinical validation data generated with a poorly validated assay is not interpretable.
The implications for early-stage programmes are practical. Before investing in clinical studies designed to generate biomarker evidence, the assay needs to be locked and analytically validated. This includes defining the reference range, the lower limit of quantification, the inter- and intra-assay variability, and the stability of the analyte under relevant pre-analytical conditions. For tissue-based biomarkers, this means establishing consistent scoring criteria and demonstrating inter-reader reproducibility. For digital and AI-driven biomarkers, it means establishing algorithm performance characteristics across the relevant sample distribution.
GoldWhite’s experience developing and validating AI-driven biomarkers in oncology, including eight CE-marked products launched at Paige.AI, has consistently reinforced this principle. The programmes that moved fastest through regulatory review were those that had invested most thoroughly in analytical validation before clinical data collection began.
Patient selection and the enrichment question
One of the most valuable things a biomarker strategy can do for an early-stage drug development programme is define a patient population in which the probability of observing a treatment effect is meaningfully higher than in the unselected population. This is the logic of biomarker-driven patient enrichment, and it is a strategy that has transformed the economics of drug development in oncology and increasingly in immunology, neurology, and metabolic disease.
The commercial and regulatory arguments for enrichment are well established. A smaller, better-defined patient population reduces the sample size required to demonstrate efficacy. It reduces trial duration. It reduces cost. And it increases the probability of a positive result, which affects both the programme’s probability of success and its attractiveness to investors and partners at interim analysis.
The challenge is that enrichment requires confidence in the predictive biomarker before the trial begins. Companies that pursue enrichment strategies on the basis of insufficiently validated biomarkers run the risk of selecting the wrong population, missing a treatment effect in the broader population, or generating results that regulators cannot interpret cleanly. The validation work needs to precede the enrichment decision, not follow it.
Building a biomarker strategy investors believe
Biomarker data is increasingly central to investor evaluation of early-stage drug development programmes. A mechanistically coherent biomarker strategy, supported by analytically validated assays and early clinical data, signals that a programme is scientifically rigorous and clinically realistic. It also provides a framework for communicating interim results in a way that sophisticated investors can evaluate.
The most credible biomarker narratives for investors share several characteristics. They are explicit about what type of biomarker is being developed and what regulatory role it is intended to play. They distinguish clearly between analytical validation, clinical validation, and qualification, and they are honest about where the programme currently sits on that spectrum. They connect the biomarker strategy to the clinical development plan in a way that makes the path to a pivotal trial legible. And they address the assay commercialisation question: whether the biomarker will be developed as a companion diagnostic, a complementary diagnostic, or an internally used trial tool, and what the regulatory and commercial implications of each option are.
Investors who have evaluated multiple drug development programmes will probe all of these dimensions. Programmes that have thought through the answers in advance are significantly better positioned than those that treat biomarker strategy as a scientific detail rather than a commercial argument.
Where to start
For most early-stage programmes, the practical starting point is a systematic review of existing biomarker data in the relevant therapeutic area. What biomarkers have been validated in similar indications? What assay platforms have regulatory precedent? Where does existing qualified biomarker data support the biological rationale for the asset?
From that foundation, the programme can define which biomarker types are strategically relevant, prioritise analytical validation for the assay or assays that will be used in early clinical studies, and develop a plan for generating the clinical validation data required to support the regulatory strategy.
Biomarker strategy is not a downstream activity. The decisions made in early development about which biomarkers to pursue, how to validate them, and how to integrate them into clinical trial design have consequences that compound throughout the life of the programme. Companies that treat these decisions with the same rigour they apply to the therapeutic hypothesis itself are consistently better positioned, scientifically, regulatorily, and commercially, than those that do not.
Marc Goldfinger is Director of Life Sciences and Health-Tech Strategy at GoldWhite. He holds a PhD in Clinical Neuroscience from Imperial College London, has led biomarker strategy and regulatory programmes across oncology, immunology, and fibrosis, and is a named inventor on a US patent for AI-driven biomarker detection.