The previous chapter discussed why scientific developments and recent treatment successes have made many of us so excited about personalised medicine. This chapter describes how the search for predictive biomarkers is likely to change drug development research. It may lead to a reduction of costs and an increase in productivity of research & development projects. Predictive biomarkers will change the way clinical trials are performed.
We Need Predictive Biomarkers
Traditionally, biomarkers answer critical questions that arise during various stages of drug development—questions such as: Does the drug reach the target? Does it have the desired biological effect? Does it have an influence on other expected or unexpected targets? Does the drug affect characteristics that predict desired or undesired effects?
However, the concept of personalised medicine centres around predictive: biomarkers that can help select a patient population with a higher chance for a favourable response to a specific kind of medicine. In today’s drug development programmes, the identification and qualification of biomarkers are integrated. If we are able to determine the biologically relevant dose, range and selection of the optimal target population by means of biomarkers, we are convinced that biomarkers will increase R&D productivity by reducing development timelines and will prevent costly late stage attrition.
Unfortunately, predictive biomarkers are often not applied until rather late in the clinical development programme, when clinical data show that an optimal benefit-risk profile is only achieved in a subpopulation of patients. In such situations, the attention is suddenly turned to other available research data that might explain the underlying biological nature of the research results. Over the last years we gathered examples of biomarkers, which have proven to predict clinical response better because they are linked to the mode of actions of the compounds in question. Examples include Her-2/neu (Trastuzumab/Lapatinib), KRAS (Cetuximab/Panitumomab), BRAF (Vemurafenib), and CCR5 (Maraviroc).
We Lack Model Systems to Predict Drug Response
In our opinion, it is crucial to identify and qualify biomarkers that predict clinical response. But we lack good model systems that predict drug response.
Generally a prerequisite to enter clinical development is that efficacy is established in animal models. Despite convincing data from animal models, the translation from animals to humans is not always successful.
Alternative models or an alternative technology are needed. This had been attempted with ex-vivo systems in areas such as rheumatoid arthritis and Crohn’s disease. These assays have proven to be valuable for the evaluation of novel therapeutic targets.1 The application of such assays seems promising, but it remains to be fully investigated.
A caveat of ex-vivo systems may be that the generated biomarker signal derives from a local tissue environment, which might be difficult to capture in peripheral blood samples. The oncology field has the best access to tissue biopsies. That is one of the reasons why oncology is a front runner in personalised medicine with several tissue based companion diagnostics. On the other hand, still very little is known about the drug resistance as encountered in the oncology field. This asks for biomarkers that elucidate more clearly what is going on in individual tumours.
New Kind of Trials
Studying targeted medicines in traditionally defined patient populations is quite problematic. For example, if we want to test a medicine like gefitinib, we do not need patients who are clinically diagnosed with non-small cell lung cancer, but patients with a specific EGFR mutation. If we were to test the medicine on a population which is not preselected for the presence of the mechanism in question, we might conclude that the medicine does not have an appropriate benefit-risk ratio. However, if tested on patients with the underlying mechanism we might come to the opposite conclusion. Gefitinib is therefore indicated for patients with tumours showing specific EGFR mutations, while it is irrelevant whether the tumour is located in the lungs.
In order to develop targeted medicines, patients will have to be selected based on presence or absence of specific biomarkers. In some situations, that may be a relatively small subset of the traditionally defined patient population.
For instance: traditionally we would test a drug on patients with breast cancer. But now we would select patients with a similar disease pathway: e.g. Her-2 positive. Eventually, the patient population may be extended: it is to be expected that we can treat patients with malignant tumours in other organs with the same medicine, because the tumours are caused by the same molecular mechanisms (e.g. m-TOR mutations).
With this approach, patient populations participating in clinical trials can be smaller, or at least much more homogenous, and response to treatment will be more consistent and predictable as well.
Integrating Biomarker Research with Target Evaluation
We think continuous biomarker research, which ideally starts at least two to four years prior to first-in-man clinical trials, is needed. Before embarking on clinical trials, we should ideally have identified a broad panel of biomarker candidates that subsequently can be further qualified in appropriate in-vitro, ex-vivo, in-vivo, or in-silico model systems.
Bioanalytical assays for the selected candidates can then be established and validated prior to implementation in clinical trials. In order to be fully operational, this strategy requires competencies within three main areas: biomarker research, biomarker assay development and clinical biomarker implementation. These areas should collaborate to guarantee the quality of the biomarker data derived from clinical trials. This will strengthen the basis on which a personalised medicine programme can be evaluated.
Moreover, once embarked on clinical trials, it is advisable to continue the biomarker research activities in parallel with the clinical development programme. In this way it is possible to improve the biomarkers by applying data derived from early clinical trials to later stages of the programme.
This approach clinically validates pre-selected biomarker candidates but may also identify potential other and better biomarker candidates correlating with clinical outcomes.
Identifying Biomarkers is a Collaborative Effort
It is difficult, time consuming and costly to identify and qualify biomarkers for other uses, such as predicting clinical outcomes. A biomarker needs sufficient evidence of broader clinical utility and validation in multiple independent studies, across different cohorts, ethnic groups and clinical subgroups. This is in line with recent guidance for biomarker qualification from the Food Drug Administration (FDA) and the European Medicines Agency (EMA).
Fortunately, multiple pre-competitive collaborative initiatives have been established globally. In recent years, large consortia have succeeded in the qualification and validation of various types of biomarkers. Examples of collaborative biomarkerconsortia are the Biomarker Consortium (US) and Innovative Medicines Initiative (EU). They are supported by representatives from regulators, pharmaceutical industry, biotech industry and diagnostic industry, academia and governmental organisations.
Click here for the next chapter in this series, where you'll find out how developing diagnostic test kits has to be coordinated between two entities—pharmaceutical companies and diagnostic manufacturers.