Drug development remains slow, unpredictable, and expensive. Estimates of the cost per approved medicine (1, 2) vary widely, from several hundred million to well over US$2 billion, with little sign that costs are falling.

The consequences are serious – for biotechs, investors, pharma companies, and patients. Biotechs fail, investors lose money, and pharma companies can struggle to recoup their investment, potentially making them more risk-averse and stymying further innovation. Patients, meanwhile, end up with costly medicines – or are denied drugs that might have helped them because those candidates were dropped during development.

There are many reasons why drug development remains so costly, but one of the biggest is the number of drugs that fail on the long road from discovery through clinical trials to regulatory approval and commercialization. Roughly nine out of 10 drugs that enter clinical trials fail to reach the market, and the cost of those failures is ultimately absorbed into the price of the drugs that succeed (3).

Drugs will always fail in preclinical or clinical development. Biology is too complex to imagine otherwise – and if all our drug candidates succeeded, we probably wouldn’t be pushing the boundaries hard enough. But the tools at our disposal to minimize failure are becoming increasingly sophisticated and could also help accelerate drug development and further reduce costs. It is imperative that we use them.

From single biomarkers to systems biology

Defining exactly what a patient is suffering from is a cornerstone of treatment. It is also essential when deciding who should be included in – and excluded from – a clinical trial. Historically, diagnostics have relied on a limited number of one-dimensional biomarkers, and clinical trial designs have often followed suit. As a result, trials can fail to accurately define a specific patient population, creating heterogeneous cohorts that dilute observed therapeutic effects. This increases the chance of a drug “failing” in a clinical trial when it might, in fact, have worked brilliantly in a more tightly defined subgroup.

Meanwhile, we have learned that conditions ranging from cancer and type 2 diabetes to metabolic dysfunction-associated steatohepatitis (MASH) and chronic kidney disease are not homogeneous diseases, but can encompass patient populations with very different underlying pathologies. Consequently, defining exactly who has what has become even more important.

In short, drug development in the era of precision medicine demands precision diagnostics. And precision diagnostics, in turn, depend on precision biomarkers.

Thankfully, biomarkers are improving. We are discovering more of them across multiple biological layers, including the genome, transcriptome, epigenome, proteome, and metabolome. Spatial biology is also helping us make sense of what is happening at molecular and cellular scales, while adding another layer of complexity.

Furthermore, we are learning how to analyze patterns of biomarkers – “constellations,” as we call them at numares – to uncover new insights.

Though the details of how this is done are complicated, the principle is straightforward.

Imagine trying to assess the height of a distant mountain using a theodolite and just one measurement. It would be little better than a guess. Now imagine taking measurements from a dozen different locations. The result would be far more accurate. Similarly, analyzing constellations of biomarkers allows us to triangulate biological information with much greater accuracy than using a single marker.

Our ability to identify biologically meaningful signatures among large numbers of biomarkers across biological systems is also improving rapidly, thanks to advances in data analysis, including AI and machine learning. It is like having satellite-guided GPS rather than a rusty old theodolite and a sextant.

One example is the use of biomarker constellations to identify liver disease earlier and more accurately using nuclear magnetic resonance (NMR) spectroscopy. NMR can detect biomarker entities in blood or urine, including lipid fractions, fatty acids, lipoproteins, and amino acids.

The biological information collected from a patient sample using NMR spectroscopy can be converted automatically into digital data, allowing researchers to interrogate it repeatedly in search of potential “digital biomarkers” without further physical processing or additional tests. This could accelerate the identification of biomarkers useful for patient stratification, clinical research, and, ultimately, patient care.

From episodic sampling to continuous monitoring

Another technological shift with the potential to improve clinical trials is the move toward more continuous patient monitoring.

At present, many trials rely on invasive sampling to generate data. Patients enrolled in liver disease trials, for example, may undergo multiple liver biopsies throughout a study to evaluate disease progression and treatment efficacy.

Such methods provide researchers with only periodic snapshots of disease progression or a patient's response to an experimental therapy. Invasive sampling that requires patients to travel to specialist centers can also increase the risk of trial dropout, which is costly and can compromise a study.

New biomarker and diagnostic approaches – including liquid biopsies using blood, urine, or saliva instead of tissue, alongside digital health monitoring – can allow patients to be assessed more frequently, or even continuously, and less invasively. This can increase the richness of trial datasets while helping retain participants.

These methods can also expand the geographical reach of trials, making participation more feasible for patients who live far from major research centers. In turn, this can broaden the pool of potential participants and help trial populations better reflect the populations ultimately receiving treatment.

From research assays to regulatory-grade diagnostics

Discovering new biomarkers is only the first step. To improve clinical trials, those biomarkers must be measured consistently, reproducibly, and to a standard that regulators can trust.

Historically, many biomarker assays have primarily been research tools, generating data that support scientific understanding but may not be robust enough to guide clinical or regulatory decisions.

That is beginning to change. Advances in analytical technologies, data science, and standardization are enabling diagnostics to deliver reproducible, quantitative measurements across different sites, studies, and populations. At the same time, the ability to analyze complex biomarker constellations rather than individual markers is improving how we identify disease subtypes, predict treatment response, and monitor disease progression.

This shift has important implications for clinical trials. More precise diagnostics can help researchers identify appropriate patients earlier, improve stratification strategies, and generate stronger evidence of therapeutic benefit. In some cases, they can help determine not only whether a drug works, but which patients are most likely to benefit.

As precision medicine evolves, diagnostics are moving from the periphery of drug development toward its center. Rather than simply measuring outcomes, they can increasingly help shape trial design, patient selection, and evidence generation from the outset.

Better biomarkers, better drug development?

Drug development will never be easy or cheap. But better biomarkers and diagnostic tools could help create clinical studies that are more tightly focused on relevant patient populations while making trials more efficient and informative.

By improving patient stratification, monitoring therapeutic response more effectively, and generating stronger evidence during development, these approaches could reduce avoidable clinical failures – and ultimately help bring effective medicines to patients more efficiently.

Newsletters

Receive the latest pharmaceutical news, personalities, education, and career development – weekly to your inbox.

I have read and understand the Privacy Notice*

Subscribe