The use of biomarkers and surrogate endpoints in drug development

By Tomasz Burzykowski, Sr. Statistician MSOURCE and Associate Prof at the University of Hasselt

Introduction

Development of a new medicine is a long and complex process. The candidate drug needs to pass through a series of phases of clinical trials, in which the efficacy and safety of the drug have to be rigorously established.

The complexity and the long duration of (especially phase III) clinical trials mostly result from the use of a long-term (e.g., clinical progression, survival) clinical endpoint to assess the clinical benefit of a new treatment. In diseases with a long natural history, the final result of comparative trials with survival endpoints is often not known before five to ten years after the study start. Also, the use of a low-frequency (e.g., hip fracture rate, post-MI mortality) clinical endpoint may result in the need to accrue a large number of patients, with obvious consequences for the duration and the complexity of a trial.

Therefore, to replace such a clinical endpoint (the “true” endpoint) by another, that could be measured earlier, more conveniently or more frequently, and that would adequately reflect the benefit of new treatments on the true endpoint, seems an attractive solution. Such replacement endpoints are termed “surrogate” endpoints.

The selection and use of surrogate endpoints is not a straightforward problem. In recent years, however, important developments have been made that allow for a practical implementation of the concept.

Selection of candidate surrogate endpoints

Biomarkers (like prostate specific antigen, PSA, for prostate cancer, or bone mineral density, BMD, in osteoporosis) are generally regarded as best candidate surrogate endpoints. Biomarker is a physical sign or laboratory measurement that occurs in association with a pathological process or that have putative diagnostic and/or prognostic utility. It is an intermediate outcome that is correlated with the true clinical outcome for an individual patient.

It is a common misconception, however, that established biomarkers – those for which the correlation with the true endpoint for an individualpatient was proven – necessarily make valid surrogate endpoints. This misconception led in the past, at least in some applications, to erroneous, or even harmful, conclusions. Probably the best known case is the approval by the Food and Drug Administration (FDA) in the United States of the use of three drugs: encainide, flecainide, and moricizine. The drugs were approved based on the fact that they were shown to effectively suppress arrhythmias. It was believed that, because arrhythmia is associated with an almost fourfold increase in the rate of cardiac-complication-related death, the drugs would reduce the death rate. Thus, arrhythmia occurrence was effectively treated as a surrogate for overall survival. However, a clinical trial conducted after the drugs had been approved by FDA and introduced into clinical practice showed that in fact the death rate among patients treated with encainide and flecainide was more than twice the one among patients treated with placebo (The Cardiac Arrhythmia Suppression Trial (CAST) Investigators, NEJM 1989). An increase of the risk was also detected for moricizine.

This and other examples of unsuccessful replacement of true endpoints led to the scepticism about usefulness of surrogate endpoints. This is despite the fact that not all early applications were failures. For example, the impressive therapeutic results for treatment of AIDS obtained early on with zidovudine, and the pressure for an accelerated evaluation of new therapies, have all led to first the use of CD4 blood count and, with the advent of highly active antiretroviral therapy (HAART), viral load, as endpoints that replaced time to clinical events and overall survival (DeGruttola et al., Journal of AIDS 1993), in spite of some concerns about their limitations as surrogates for clinically relevant endpoints (Lagakos and Hoth, Annals of Internal Medicine 1992).

Required characteristics of surrogate endpoints

The failed past attempts to use surrogate endpoints made it clear that there is a need for a formal definition and framework that would allow to check the validity of the use of a biomarker as a surrogate endpoint. During last two decades several definitions of a surrogate endpoint have been proposed. The one that allows for a practical implementation of the concept states that it is required that “the effect of treatment on a surrogate endpoint must be reasonably likely to predict clinical benefit” (Biomarkers Definitions Working Group, Clinical Pharmacology & Therapy 2001). Surrogacy is thus a concept that relates to groups of patients. To demonstrate surrogacy, a high association between the treatment effects on the surrogate and on the true endpoint needs to be established across groups of patients treated with the new and standard interventions.

Methods for practical validation of surrogate endpoints

Along with the development of different definitions of a surrogate endpoint, methods for verifying whether the definitions hold for a biomarker have been proposed. Recently, a new methodology, known as the “meta-analytic” validation approach, has been developed (Buyse et al., Biostatistics 2000). This method uses data from multiple randomized clinical trials and aims at assessing directly the precision of prediction of treatment effect on the true endpoint from the effect on the surrogate. Thus, it allows to assess whether “the effect of treatment on a surrogate endpoint is reasonably likely to predict clinical benefit”, as required by the Biomarkers Definitions Working Group definition. As such, it is a powerful tool that makes it possible to assess the suitability of candidate surrogate endpoints and to prevent erroneous use of them.

A big advantage for such a structured approach is that it allows quantifying the evidence in favor or against the use of a surrogate. The methodology has been already used, e.g., to evaluate the validity of using response rate and progression-free survival as surrogates for overall survival in colorectal cancer (Buyse et al., Lancet 2000; Sargent et al., Journal of Clinical Oncology 2005; Buyse et al., Journal of Clinical Oncology 2007); the validity of using PSA as a surrogate for overall survival in metastatic prostate cancer (Collette et al., Journal of Clinical Oncology 2005); or the validity of using response rate, disease control rate, time to progression, and progression-free survival as surrogates for overall survival in metastatic breast cancer (Burzykowski et al., Journal of Clinical Oncology 2007). Some of these results have been taken into account in the “FDA Project on Cancer Drug Approval Endpoints”, launched by FDA “to evaluate potential endpoints for cancer drug approval” (http://www.fda.gov/cder/drug/cancer_endpoints). Within the project, FDA holds public workshops to identify important issues that are discussed in meetings of the Oncologic Drugs Advisory Committee (ODAC). Subsequently, guidance documents are published describing FDA's view on endpoints for cancer drug approval. Thus far, workshops for multiple myeloma, ovarian cancer, primary brain tumors, lung cancer, colorectal cancer, prostate cancer, and acute leukemia, have been organized.

Conclusions

One needs to be aware of the fact that, even if a surrogate has been validated for a particular class of treatments in a particular disease, this does not automatically mean that the use of the surrogate is valid in another disease, or for another class of treatments. This is because the mechanisms of action of different treatments may differ, and their relevance for different diseases may also vary. Thus, the use of a surrogate endpoint will most likely always require a careful consideration whether the result of a validation exercise can apply to a particular situation.

Nevertheless, there are numerous reasons why using a surrogate endpoint does constitute an attractive option:

1. The number of candidate biomarkers and ultimately the number of potential surrogate endpoints based upon them is increasing dramatically.
2. New discoveries in medicine and biology create a possibility for development of many potentially effective treatments for a particular disease. In such a situation a need to cope with a large number of new promising treatments that should be quickly evaluated with respect to their efficacy might appear. This can already be observed happening in, e.g., oncology.
3. There is also increasing public pressure for new, promising drugs to be approved for marketing as rapidly as possible. Such an approval will have to be based on biomarkers rather than on some long-term clinical endpoint (Lesko and Atkinson, Annual Review of Pharmacology & Toxicology 2001). Regulatory agencies are, in fact, already responding to this pressure by allowing like, for instance, FDA, “fast tracks” for the approval of drugs (www.accesdata.fda.gov/scripts/cder/onctools).
4. If the approval process is shortened, there will be a corresponding need for earlier detection of safety signals that could point to toxic problems with new drugs. The use of surrogate endpoints (for toxicity-related clinical endpoints) might allow one to obtain information about toxic effects even during the clinical testing phase.
5. Finally, from a practical point of view, shortening the duration of a clinical trial also limits possible problems with non-compliance and missing data, which are more likely in longer studies, and therefore increases effectiveness and reliability of the research.

For all these reasons, the concept of a surrogate endpoint gives a promise of a quicker development of new medicines. Now, with the availability of the meta analytic validation approach, we are much closer to a practical implementation of the concept and a realization of the promise.