Key Takeaways

• Biomarker failure is pervasive - of 883 EMA assessment reports, only 37 predictive biomarkers for 41 drugs were identified, and most candidates fail early in discovery or analytical validation
• The most common failure causes are confirmational bias in discovery, premature promotion before comprehensive validation, and poorly designed clinical studies with inadequate blinding
• When biomarkers are used to select clinical trial populations, drug approval probability increases from 10% to 25% - making a rigorous biomarker strategy one of the highest-value investments in drug development

Jump to section

The Potential of Biomarkers

In Drug Development

The drug development process takes time and money and is fraught with difficulties; many drugs do not make it to the approval stage. However, when biomarkers are used to select clinical trial populations, the probability of successful drug approval increases from 10% to 25%. Biomarkers have the potential to decrease the time, costs, and failure rate of drug development.

In Disease Detection

Significant efforts are ongoing to generate reliable and sensitive early-detection biomarkers that allow diseases to be caught early. These efforts have the potential to benefit patients by enabling more effective treatment and screening for predictors of future disease. Effective biomarkers could provide us with greater knowledge of the mechanisms of disease and pave the way for preventative measures such as informed lifestyle changes.

Perhaps the most notable recent success story is HER2 in breast cancer. HER2-targeted therapy has been available for over two decades for patients expressing high levels. Results from the DESTINY-Breast04 study showed patients with even low levels of HER2 treated with T-DXd had a two-fold increase in progression-free survival - dramatically expanding the patient population who could benefit and shifting how HER2 status is categorised.

In this article, we discuss why biomarkers fail at different stages of development. Let's start with an overview of the biomarker life cycle.

The Biomarker Life Cycle

All InVitro Diagnostics (IVDs) exist because of a diagnostic, predictive or prognostic clinical need. The biomarker life cycle begins with identifying the clinical need, which forms the basis of the IVD's intended purpose and defines the intended use population.

The clinical need drives the search for datasets to discover candidate biomarkers and confirm representation across multiple studies in the intended use population. A test method is developed once a candidate has been established as representative and specific. Clinical validation demonstrates the clinical sensitivity and specificity of the IVD. Planning for clinical adoption begins in parallel with the preceding stages. We discuss the validation process in depth in our Guide to Biomarker Validation.

The Difficulties of Biomarker Development

Several studies highlight the benefits of including biomarkers in drug development. This has generated enthusiasm and massive investment. However, a recent analysis showed that whilst the EMA published 883 European public assessment reports, only 37 predictive biomarkers for 41 drugs were mentioned.

A few overarching difficulties face researchers investing in biomarkers:

• Skills gaps and lack of expertise underlie the difficulty in validating clinically that a marker has utility. This is a process failure that could be avoided by effective planning. Sonrai's governed platform removes many of these process barriers, and our guide to free bioinformatics tools helps researchers build capability independently.
• Lack of regulatory oversight for the masses of data arising from parallel assays such as microarrays, mass spectroscopy, or DNA sequencing. Most publications on biomarkers have not been validated clinically by outside investigators.
• Investigator-specific biases persist because of the omission of the necessary "blinding". Some biases are especially common to initial reports - the "beginner's luck bias" or "declining effect phenomenon".

Reasons for Failure

Failure can occur at any phase of the lifecycle. However, most candidates fail early in the development process. There is potentially a pressure to fail early rather than late, as less time and resources will have been invested. The FDA and NIH recently developed the BEST resource to aid researchers in understanding biomarker definitions and applications - which could help mitigate future failures.

Failures during Discovery

A combination of poor methods, selective publication and selective or incomplete reporting contributes to considerable waste in biomarker discovery. Two common pitfalls are applying a hypothesis-driven or supervised selection of biomarkers based upon existing knowledge of the disease, and feature selection using exhaustive machine learning and statistical modelling searches.

Hypothesis-driven methods, when driven by confirmational bias, lead to cherry-picking biomarkers. Many machine learning techniques can result in overfitting when applied blindly to the data. Both produce biomarkers that fail to generalise across independent datasets. Unbiased analytical approaches - as described in our bioinformatics tools guide - are essential at this stage.

Failures during Analytical Validation

Premature promotion of biomarker potential in advance of comprehensive performance evaluation is a critical failure mode. One notable example was the discredited Lancet publication on proteomic patterns in ovarian cancer - reporting near-perfect discriminating ability before comprehensive independent validation had been completed.

Failures of Clinical Validation

Some biomarkers fail clinical validation and are shown to have little additional predictive ability in a comprehensive clinical setting. The regulatory landscape surrounding biomarker approval is less robust than those for drugs and biologics, and proof of effectiveness from trials is not typically required.

The example of haemoglobin A1C reduction with rosiglitazone provides a cautionary tale. Rosiglitazone was approved for Type 2 Diabetes based on evidence demonstrating its ability to lower haemoglobin A1C. However, a 2007 New England Journal of Medicine study reported it resulted in a significant increase in myocardial infarction risk - a systemic risk-benefit failure that the approval studies were not designed to detect.

Non-Optimised Clinical Translation

Many gene expression profiles for cancer outcomes fail to move into clinical development because studies have small sample sizes, and selection for translation is often determined by commercial viability rather than comparative evidence of performance and merit. Discrepancies in access to biomarkers also impact implementation and depend on geographical distribution of expertise across hospitals and medical centres.

Clinical Relevance

After implementation, continued surveillance sometimes proves the biomarker is not as useful as proposed, or may even be harmful. The PSA test - considered accurate and standardisable with simple assays - had poor predictive discrimination, overdiagnosis rates of 17-50%, and treatment came with the risk of serious harm.

Examples of fraud - such as the papers retracted by Potti and the Theranos scandal - remind us that ethics and the fundamental standards of reproducible research must always exist in biomedical research.

Success Stories

Recent notable successes shed light on how this field is moving forward. The uptick in adopting gene expression profiles in oncology for prognosis and treatment guidance is a great example. Commercially available diagnostic biomarker tests include MammaPrint and Prosigna in breast cancer, the Oncotype Dx tests (breast, prostate and colorectal) and the Prolaris prostate cancer test.

The key aspect is the extensive and intensive development programmes undertaken for gene expression assays - thorough analytical validation followed by careful clinical validation involving hundreds of patient samples from multiple collections. These are the standards that underpin successful biomarker discovery.

In cardiovascular medicine, brain natriuretic protein (BNP) - first identified in 1988 - took almost thirty years to become part of clinical practice, established through the study of thousands of patients in many controlled trials. This timeline illustrates both the rigour required and the value delivered when validation is done properly.

In inflammatory diseases, the Vectra test for rheumatoid arthritis has shown that changes in the test score over as little as two weeks of therapy can predict future clinical responses measured at six weeks - potentially avoiding costly therapy trials in patients who won't benefit.

Strategies for Success

Creating a biomarker strategy at the earliest point possible in drug development increases the opportunity to select optimal targets and models for the preclinical stage. An early biomarker strategy also offers an increased chance to develop a companion diagnostic.

Critical factors to an in-depth understanding of a candidate biomarker include:

• Access to well-annotated biospecimens
• Early attention to the characterisation of the biomarker and standardisation of assay methods
• Clinical trial designs that allow early evaluation of the effectiveness of the biomarker in predicting responsiveness to the drug - often referred to as co-development
• Ability to incorporate advancing technology and emerging data suggesting new or additional biomarkers that better characterise the target populations

In conjunction with a sound discovery strategy, a rigorous validation strategy is needed that considers:

• Unbiased data analytic strategies for reproducible biomarker discovery
• Use stratification and dimension reduction to address disease heterogeneity
• Study design as a means of minimising confounding factors
• Avoid bias in data acquisition and analysis
• Avoid overfitting during model building and optimisation

For an in-depth look at what lies ahead in biomarker research and applications, read our piece on The Future of Biomarkers.

Common Questions about Biomarker Failure

References

• Global Engage | Why do biomarkers fail?
• Waste, Leaks, and Failures in the Biomarker Pipeline | Clinical Chemistry
• Why Your New Cancer Biomarker May Never Work | Cancer Research
• Biomarker definitions and their applications | PMC
• Stratified medicine in EMA licensing | PubMed
• Does biomarker use in oncology improve clinical trial failure risk?
• Why 90% of clinical drug development fails and how to improve it
• Unbiased data analytic strategies to improve biomarker discovery in precision medicine
• Discovery and validation of biomarkers for safe and effective pain therapeutics
• Recent progress in biomarker development | APM

Talk to our team

Sonrai acts as an extension of your team, combining scientific expertise, transparent and reproducible analysis, and a governed research environment to generate decision-ready evidence that can be revisited, reproduced, and built upon throughout the life of a programme.