Biomarker-Driven Development: What Boards Need to Understand About Betting on a Signal
Biomarker-driven development is one of the most powerful ideas in modern drug development and one of the most quietly hazardous for a board to oversee. The promise is genuine: instead of testing a drug in a broad population and hoping enough patients respond to move the average, a company identifies a biomarker that predicts response and studies the drug in the patients most likely to benefit. The trial gets smaller, faster, and more likely to succeed. The eventual label points at a defined population. The commercial and clinical story becomes cleaner. When it works, it transforms a program's odds.
What boards frequently do not appreciate is that this strategy does not simplify the company's development challenge — it doubles it. A biomarker-driven program is not one development effort but two running in parallel: the drug, and the diagnostic test that identifies which patients should receive it. Each has its own science, its own risks, its own regulatory relationship, and its own ways of failing. And the two are coupled, so that a failure in the diagnostic can sink a perfectly good drug, and a shift in the biomarker strategy can invalidate work already done on both.
This article is about what a board needs to understand to govern that coupled bet well — not the assay chemistry, which is management's domain, but the strategic structure of the risk.
The Core Idea, and Why It Is Seductive
A predictive biomarker is a measurable characteristic — a genetic alteration, a protein expression level, a molecular signature — that identifies patients more likely to respond to a given therapy. Developing against one lets a company enrich its trial with likely responders, which can turn a drug that would fail in an unselected population into one that succeeds in a selected one.
The appeal to a clinical-stage company is obvious and real. Smaller trials cost less and read out faster. A concentrated effect in a defined population produces a cleaner result and a more defensible label. Investors and partners understand the story. In an environment where capital is scarce and speed matters, a credible biomarker strategy can be the difference between a fundable program and an unfundable one.
Because the appeal is so strong, the temptation is to adopt a biomarker strategy on a thinner foundation than the decision deserves — to treat a plausible biomarker hypothesis as if it were an established one, and to build the entire program on a signal that has not actually been shown to predict what the company needs it to predict. This is the central risk the board has to govern, and it has several distinct faces.
The Diagnostic Is a Second Development Program
The first thing a board should internalize is that committing to a biomarker strategy means committing to develop, validate, and — in most cases — secure regulatory approval for a companion diagnostic: the test that will be used in the real world to select patients for the drug.
In the United States, the FDA regulates a companion diagnostic as a medical device in its own right, distinct from the drug it accompanies, and the agency's stated best practice is for the drug and the diagnostic to be developed concurrently so that both are approved and available at the same time. This is not a formality. A drug can be approved on the strength of a biomarker-selected trial and then face a problem because the test used to select patients in the trial is not the same test that will be available to physicians at launch — and bridging between the two requires its own evidence, its own banked samples, and its own planning that has to begin years earlier.
For a board, the practical implications are concrete. The company needs diagnostic development capability, either internally or through a partner, and that partnership is itself a strategic relationship with its own risks. The timelines of the drug and the diagnostic have to be synchronized, and a diagnostic that falls behind can delay a drug that is otherwise ready. The samples needed to validate the eventual commercial test have to be collected and banked, with appropriate patient consent, during the clinical program — a requirement that is nearly impossible to satisfy retroactively if it was not planned for from the start. A board that approves a biomarker strategy without asking who is developing the diagnostic, on what timeline, and whether the sample and consent infrastructure is in place has approved half a plan.
The Ways a Biomarker Strategy Fails
The failure modes here are specific, and a board that understands them can ask questions that surface trouble before it becomes expensive.
The biomarker does not actually predict response. This is the fundamental risk. A biomarker adopted on the strength of preclinical data, a biological rationale, or a retrospective signal may simply not hold up when tested prospectively. The company has then built a trial, a diagnostic, and a commercial strategy on a foundation that does not exist. The board's question is how well established the predictive relationship actually is: is it grounded in prospective evidence, or in a biologically appealing story and a retrospective analysis that has never been confirmed? The confidence with which management presents the biomarker is not evidence of its validity; the strength of the underlying data is.
The cutoff is arbitrary. Many biomarkers are continuous — a protein is expressed at some level, not simply present or absent — and the company must choose a threshold that divides "positive" from "negative." Where that line is drawn determines who gets the drug, how large the eligible population is, and how strong the observed effect appears. A cutoff chosen after seeing the data, to maximize the apparent effect, is a form of the pre-specification problem that recurs throughout clinical development: it can produce an impressive result that does not replicate. The board should understand how the threshold was set, whether it was fixed in advance, and whether the commercial population it implies is large enough to matter.
The trial assay and the commercial test diverge. The test used to select patients during the trial is frequently not the test that will be sold when the drug launches — the trial may use a laboratory-developed test for speed, while the commercial product must be a validated, approved device. If the company has not planned the bridge between them, it can find at approval that it cannot demonstrate the commercial test selects the same patients the trial did. This is a planning failure, not a scientific one, which is what makes it so avoidable and so painful when it happens.
The biomarker-positive population is too small to build a business on. A biomarker strategy that produces a beautiful effect in three percent of patients may yield a drug that works and a company that cannot generate a return. The narrower the selected population, the stronger the effect often looks and the smaller the market becomes. The board has to hold both halves of that trade-off in view at once, and be wary of a strategy that has optimized for a clean trial result at the cost of a viable commercial population.
The strategy forecloses the broader opportunity. Committing to a narrow biomarker-defined population can be the right call, but it can also prematurely abandon patients outside the definition who might have benefited. Once a program is built around a biomarker, expanding back to a broader population is difficult and expensive. The board should understand what the biomarker strategy gives up, not only what it gains.
The Questions a Board Should Ask
None of the following requires the board to understand the assay. All of them go to the strategic soundness of the coupled bet.
How established is the predictive relationship — prospectively, or by hypothesis? This is the foundational question. The board should press until it is clear whether the biomarker has been shown to predict response in a way that will hold up, or whether the company is betting that it will. Both can be legitimate, but they carry entirely different risk, and the company should not be allowed to blur them.
Who is developing the companion diagnostic, and is it synchronized with the drug? A biomarker strategy without a concrete, resourced, appropriately timed diagnostic plan is incomplete. The board should know whether the diagnostic is being developed internally or with a partner, whether that partner is reliable, and whether the two timelines are actually aligned rather than merely assumed to be.
Was the biomarker cutoff pre-specified? As with any threshold in clinical development, a cutoff chosen after seeing the data is far weaker than one fixed in advance. This single question catches a large share of the results that look better than they will prove to be.
Have we banked the samples and consents we will need? Because the requirement to collect and consent samples for eventual diagnostic validation must be satisfied during the clinical program and cannot easily be fixed later, the board should confirm early that this infrastructure exists. It is the kind of unglamorous planning question that boards skip and companies regret.
Is the biomarker-positive population large enough to matter commercially? The board should insist on seeing the selected population's size alongside the effect it produces, and evaluate them together. A strong effect in a population too small to sustain a business is a scientific success and a commercial dead end.
What does this strategy give up? Understanding the foreclosed broader opportunity, and being satisfied that the narrowing is a deliberate strategic choice rather than an unexamined default, is part of governing the decision honestly.
The Partnership Dimension
Because most drug-developing companies do not have deep diagnostic development capability in-house, a biomarker strategy usually means a partnership with a diagnostics company. This introduces a governance dimension that boards focused on the drug can overlook entirely.
The diagnostic partner has its own priorities, its own timelines, and its own commercial interests, and the drug company's program depends on that partner delivering. A diagnostic partnership that is under-resourced, misaligned on timelines, or built on a partner for whom this program is not a priority can become the constraint that holds up an otherwise successful drug. The board should understand the health of this relationship as it would any other strategically critical partnership — not as a technical vendor arrangement but as a dependency on which the program's success rests. The same discipline the board would apply to a licensing or manufacturing partner applies here, and it is easy to forget because the diagnostic feels like a supporting detail rather than a coupled half of the program.
The Underlying Principle
Biomarker-driven development is a genuinely powerful strategy, and nothing in this article is an argument against it. It is an argument for governing it as what it actually is: a coupled bet on two development programs, a drug and a diagnostic, each with its own risks, joined so that a failure in either can sink both.
The board's contribution is not to evaluate the biomarker's biology — that is management's expertise. It is to ensure the company has been honest about how well established the predictive relationship really is, has resourced and synchronized the diagnostic as a real program rather than an afterthought, has fixed its thresholds before rather than after seeing the data, and has weighed the cleaner trial against the commercial population it implies. The strategy's power is real, and so is its capacity to fail quietly, on the half of the program the board was not watching.
Lawrence Fine is CEO of AGCP Farmacêuticos and has advised on licensing, regulatory, and partnership strategy across the pharmaceutical and advanced materials sectors.