Evotec’s VP Translational Biomarkers, Michael Rohe, speaks with Hilary Wilkinson, Vice President Fluid Biomarkers at CHDI, about how biomarker strategies influence decision-making, patient selection, and neuroscience drug discovery & development.
MICHAEL: When do you implement biomarker strategies in preclinical projects and what proportion of early-phase clinical trials in neuroscience include biomarkers?
HILARY: Biomarker strategies can be implemented at any time in preclinical projects and should be discussed at the outset as part of the strategic evaluation for pursuing a new therapeutic intervention target. Between 75-90% of trials in neurodegenerative disease include biomarkers, because clinical outcomes usually take a relatively long time to change and biomarkers can help monitor drug response.
It is important to recognize that all biomarkers are useful, even if they are not always translatable. Well-characterized biomarkers in animal or cell-based models can be used for lead optimization and for building PK/PD (pharmacokinetic/ pharmacodynamic) models and predicting dose in people, even though one may need to develop a different biomarker for use in patients. For example, changes in regional mRNA transcription patterns in different parts of the brain can be used for drug discovery and lead optimization even if they cannot be imaged or detected in CSF in patients.
As we consider how we use biomarkers in drug discovery, however, prioritizing those biomarkers which are translatable is most common. There are several strategies for biomarker testing: First, if established biomarkers for the disease models already exist and are routinely used, then it is a matter of determining which biomarkers respond to the therapeutic approach. This can be done with tools like siRNAs directed at the new therapeutic target as well as testing the drug candidate. Second, one can develop novel biomarker assays, such as immunoassays or next-generation sequencing-based assays. It is particularly useful if assays can be directed to both human and preclinical species. This approach needs to begin as early as possible because it can take years to succeed. Once the assay is established, testing is needed in the appropriate cell-based or animal models. Third, adapting biomarker assays which are used in related disease areas. Testing these biomarkers in the relevant models can greatly accelerate discovery pipelines. Fourth, exploratory discovery using various ‘omics approaches in animal models, cell lines or human tissues to identify a novel biomarker or groups of biomarkers which have potential utility in drug development. This can be enhanced by comparing results in animal models and patient tissues to prioritize the utility and robustness of candidate biomarkers.
HILARY: Qalsody® (tofersen) is a good example where biomarkers significantly impacted the development of the drug. This is an antisense oligonucleotide used to treat amyotrophic lateral sclerosis in patients with a mutation in the superoxide dismutase gene. The FDA accepted the use of neurofilament light chain as a reasonably likely surrogate endpoint to predict clinical benefit. In addition, super oxide dismutase protein was measured as an indirect biomarker for target engagement. Robust and significant changes in both biomarkers were critical to the approval decision because a statistically significant treatment effect in the primary clinical endpoint was not demonstrated in the pivotal trial.
MICHAEL: How much confidence can we have in drug development decisions without biomarkers?
HILARY: Most drug development decisions depend on several pieces of information including pharmacokinetics, clinical safety and efficacy outcomes, and specific guidelines provided by regulatory authorities. Biomarkers are often used by sponsors to support early decision making in drug development because significant changes can be detected in a shorter time and with fewer participants versus the length and size of trials required to detect changes in clinical outcomes. They are not absolutely necessary but without these early biomarkers decision-making often carries a higher risk of failure.
An obvious example is cholesterol which changes in a relatively short time, versus the years it can take to observe an outcome such as myocardial infarction. Without cholesterol and its utility in dose selection, it can be very challenging to determine the best treatment paradigms and to detect whether the drug is having the desired effect. Most drug development paradigms in neurodegenerative disease depend on several biomarkers and several clinical assessments.
MICHAEL: What is your primary goal when incorporating biomarkers: dose selection, mechanism-of-action confirmation, or patient stratification?
HILARY: The primary goal when incorporating biomarkers depends on the nature of the drug development strategy. In neuroscience, the basic need for dose selection must be met and is generally achieved using pharmacodynamic biomarkers of drug response. Depending on the nature of the disease, patient stratification can be critical as having biomarker tools to identify a more homogeneous population can greatly improve the statistical power to detect a response. One of the most challenging biomarkers to identify is those that predict outcomes, and these can have the greatest impact on success.
MICHAEL: Would you recommend including target engagement or pharmacodynamic biomarkers to support dose selection in clinical trials?
HILARY: Pharmacokinetic modeling can be used very effectively in drug development for dose selection. However, having a target engagement or pharmacodynamic biomarker can greatly improve one’s confidence in the pharmacokinetic model and reveal unexpected differences if they exist. Target engagement is not always a pharmacodynamic biomarker. There is a need to correlate the pharmacokinetics to efficacy for dose selection. Prediction of efficacy can be difficult before actually running a trial, and often target engagement is combined with pharmacodynamic biomarkers to predict outcomes such as clinical signs or symptoms. The validity of these predictions is determined in the course of the trial.
MICHAEL: Do you prefer target engagement and pharmacodynamic biomarkers as secondary endpoints, or do you favor exploratory biomarkers for greater flexibility?
HILARY: Each drug development program is unique and there may be a rationale for using biomarkers as secondary endpoints versus exploratory endpoints. I don’t think this is something that can be decided independent of the disease or regulatory guidance. Target engagement biomarkers can be used as secondary endpoints, but in neurodegenerative disease testing multiple biomarkers as exploratory endpoints is the most conservative approach and can provide useful information for future clinical trial design.
MICHAEL: What biomarker technologies are currently used, and are there emerging platforms you are excited to adopt?
HILARY: Technologies that have persisted for several years are appealing because their strengths and weaknesses are well defined. Automated single molecule detection platforms can be robust and reproducible, but all platforms have technological challenges that require constant vigilance. Production of high-quality reagents for kits can often be one of the critical elements for selection of a bioassay. Having a dependable source for the kits, technical support for the machine, and robust methods for quality control are all critical needs to support biomarker utility. When companies go out of business or are acquired, whole platforms can disappear and this has a negative impact on drug development. Significant time is needed to test new platforms and build confidence in the robustness and reproducibility of the data. There are many new next-generation sequencing approaches, improved sensitivity in mass spectrometry, and seeding assays which are very exciting and will change the biomarker landscape as they become integrated into drug development programs. Many of these enable the detection of unique molecular signals that are present but difficult to distinguish with current ELISA-based methods.
MICHAEL: How do you see neuroscience biomarker strategies evolving over the next 5 years and what impact do you expect Al and machine learning to have?
HILARY: The use of machine learning to predict relationships between datasets is already in place and the sophistication and complexity of the analyses continues to evolve. In many omics approaches to biomarker discovery, these methods have been deployed to improve prioritization of biomarker candidates. Similarly, deep learning has been used for neural network analysis of MRI scans, for example. The use of artificial intelligence will allow the integration of results from these types of approaches to modeling with even larger and more diverse data and information. The hope is to be able to determine a patient’s biological disease state, understand the underlying processes driving progression, predict their trajectory, decide on a therapy and identify which biomarker measurements would be most useful for monitoring treatment.
Hilary Wilkinson
Vice President Fluid Biomarkers, CHDI
Biography
Hilary is a drug discovery and development expert with experience in leading in vitro and in vivo pharmacology teams. She has contributed to the advancement of clinical candidates for programs targeting oncology, bone, cardiovascular, muscle, and central nervous system disorders.
References
- Leisgang Osse et al (2026) Alzheimer’s Dement. 12 (1): e70217
- Vissers et al 2021 Int J Mol Sci.22(4):1615
- https://www.accessdata.fda.gov/drugsatfda_docs/nda/2023/215887Orig1s000IntegratedR.pdf