Biomarkers and Surrogate Endpoints in Clinical Drug Development

July 1, 2003
Bernd Rosenkranz

Bernd Rosenkranz, FFPM, MD, PhD, is vice president, Clinical Development, Jerini AG, Invalidenstr. 130, D-10115, Berlin, Germany, +49 30 97893 500, fax +49 30 97893 599, email:

Applied Clinical Trials

Applied Clinical Trials, Applied Clinical Trials-07-01-2003,

Both large and small pharmaceutical companies have learned that the value of their development candidates increases once clinical research has demonstrated their proof of concept. Sponsor companies face the challenge of moving new blockbuster drugs to market as rapidly as possible.

Both large and small pharmaceutical companies have learned that the value of their development candidates increases once clinical research has demonstrated their proof of concept. Sponsor companies face the challenge of moving new blockbuster drugs to market as rapidly as possible. During development, they try to recognize, and minimize their spending on, products that can never make it through the registration process because of toxicity or lack of efficacy. Focused drug development using the best scientific approach accepted by regulatory authorities has become (or at least should become) a major objective for clinical research departments. That approach can include the use of appropriate biomarkers during preclinical and clinical drug development.

As a general principle, a suitable clinical biomarker must be selected for which a reasonable relationship has been demonstrated to the clinically relevant endpoint. A few biomarkers are so closely linked to the clinical outcome that regulatory authorities have considered them valid primary variables in pivotal clinical trials and a substitute for a hard clinical endpoint during the marketing authorization process. It generally takes considerably less time and fewer subjects-and therefore less money-to demonstrate a significant effect of a drug on such a surrogate than on a "hard" endpoint. Based on these clinical biomarkers, preclinical models should be chosen that have some relevance to the effect of the drug in patients with the target disease.

Biomarker defined

In April 1999, the National Institutes of Health (NIH) and the Food and Drug Administration (FDA) cosponsored a conference on "Biomarkers and Surrogate Endpoints: Advancing Clinical Research and Applications."1 The concepts of biomarkers and surrogates have been summarized by the NIH Definitions Working Group as follows:

  • A biomarker is a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention.

  • A clinical endpoint is a characteristic or variable that reflects how a patient feels, functions, or survives.

  • A surrogate endpoint is a biomarker that is intended to substitute for a clinical endpoint. 2

Although more complex definitions have been suggested since then, the NIH definitions present a concise summary of the basic concepts. Preclinical project team members should choose appropriate (and, if possible, "humanized") biomarkers that are relevant to normal and/or pathological biology or physiologically associated with the target disease and the action of the new drug in humans.

Figure 1-3.

In early clinical studies, suitable biomarkers should be applied to help to demonstrate proof of concept and identify appropriate dose regimens for efficacy studies. Doing so-and learning, for example, which subpopulations are most likely to benefit from the new treatment-can help in planning later efficacy studies. Although the approach may seem straightforward, the use of biomarkers and surrogates carries with it a number of practical problems and pitfalls. Cardiovascular research will be used as an example here, because the concepts have often been applied in this clinical field.

It should be noted that any clinical observation in a given subject might serve different purposes, and each purpose can have implications for the way these characteristics are used, documented, validated, and interpreted. In clinical practice, for example, blood pressure (BP) is measured to assess the individual risk for a given patient and to decide upon that patient's further diagnostic and therapeutic management. The same characteristic may be determined in epidemiological research, for example. Finally, blood pressure may be used as an outcome measure for a company's decision making in a research project. It may help to answer such questions as "Does the antihypertensive drug work?" or "Does the new drug create any side effects related to blood pressure?" The same measure may become a valid surrogate and an essential element in the regulatory review process to be used for the decision about marketing approval for the new drug. Researchers in cardiovascular drug development have access to clearly defined and rigorously validated biomarkers, surrogate endpoints, and clinical endpoints.

Biomarkers include biochemical and functional characteristics or signals; ideally they should be noninvasive. Examples for biochemical biomarkers for cardiovascular conditions are high-density lipoprotein (HDL) cholesterol, lipoproteins, creatine kinase MB band (CK-MB), troponins (markers for myocardial cellular damage), high-sensitivity C-reactive protein (hs-CRP, a marker for inflammatory processes), fibrinogen, or plasminogen activator inhibitor-1 (PAI-1, an indicator of the activity of the blood coagulation system-and thus a risk for intravascular thrombus formation).3, 4 Functional or physiological characteristics include endothelial function, arterial or venous blood flow, arterial stiffness, and left ventricular systolic or diastolic volume or function. Finally, examples for signaling characteristics are ST depression recorded in the ECG (exercise or pharmacological stress testing, for example), and findings obtained by an imaging method such as quantitative coronary perfusion, coronary calcification (electron beam tomography), intravascular ultrasound, magnetic resonance imaging (MRI), or nuclear imaging (99mTc-SPECT) that include exercise or pharmacological stress testing.

The most important criteria for valid biomarkers to be used in drug development are their clinical relevance; their sensitivity, specificity, and reliability (do they measure what they are supposed to); and their practicality and simplicity.5

The list of cardiovascular biomarkers that have been accepted as surrogate endpoints is much shorter, of course, and includes such characteristics as blood pressure, cholesterol, and low-density lipoprotein (LDL).

Validity of biomarkers and surrogates

The most important criteria for valid surrogates are summarized in the International Conference on Harmonisation (ICH) Guideline E9, Statistical Principles for Clinical Trials.6 These mainly define the relationship between the surrogate endpoint (high blood pressure, for example) and the "hard" clinical endpoint (such as stroke) actually relevant when treating the condition. To show this relationship, the following must be demonstrated:

  • biological plausibility

  • statistical relationship in epidemiological studies

  • evidence from clinical studies that treatment effects on the surrogate correspond to the clinical outcome.

For the same disease, the relationship between surrogate and clinical endpoints may depend on the mode of action. This means that for the same condition, regulatory authorities may accept that the surrogate is valid for one class of drugs, but not for other classes.

"Hard" clinical endpoints for cardiovascular disease include such events as death, myocardial infarction, and stroke. These require no further validation, but are clinically relevant per se. However, use of these measures is often impossible in early proof-of-concept studies, because those studies require a large number of subjects and a long time period-often years from the beginning treatment until the event.

Biomarkers in preclinical development

Drug development was traditionally described as a sequential occurrence of formal, well-defined phases-usually described as preclinical development followed by clinical Phases 1 through 4. Today, drug development is seen as an integral data-driven process in which the phases overlap and-more importantly-use all information gained to create an optimal design for the next steps.7 Furthermore, unexpected findings at a later stage of development may necessitate planning new preclinical studies or Phase 1 and Phase 2 trials.

In preclinical development, biomarkers are an essential element of the processes used to optimize the science of drug development. These include a number of modern technologies, such as functional genomics, proteomics, imaging, modeling and simulations, computational methods/informatics, modern analytical technologies, and specific humanized preclinical models.

A well-documented example for the selection of appropriate and validated preclinical biomarkers is cardiac safety models to assess the potential of a drug causing prolongation of QTc, an ECG biomarker for life-threatening ventricular arrhythmia. Several regulatory guidance documents have recently been published on appropriate biomarkers for the proarrhythmic potential of drugs, and a harmonized guideline for the United States and Europe is expected for this year.8–11 Relevant preclinical models include in vitro tests such as ionic channel assays and hERG (human ether-a-gogo) channel binding, guinea pig myocytes, rabbit or dog Purkinje fibers, anesthetized guinea pig or dog, and conscious dog studies. These models are considered reliable and practical. They fulfill the requirements for suitable biomarkers and are related to the corresponding clinical biomarker, QTc prolongation.

Preclinical or clinical studies frequently are performed without assessing the concentration dependency of the observed effects. This information would be very valuable, however, to predict appropriate dosage regimens in initial studies that use pharmacokinetic/pharmacodynamic (PK/PD) modeling techniques. Thus, in vitro concentrations or in vivo exposure, by sparse sampling methods (example: toxicokinetics), should regularly be determined in preclinical models.

A recent "Position Paper on the Non-Clinical Safety Studies to Support Clinical Trials with a Single Low Dose of a Compound" drafted by the Committee for Proprietary Medicinal Products (CPMP) proposed minimum requirements for early pilot studies on the pharmacokinetics of a new chemical entity in human subjects using subpharmacological doses.12 This procedure allows for selecting appropriate animal species for preclinical studies-especially safety studies-and also for better use of preclinical information to predict dose dependency of efficacy or toxicity data in humans.

Interpreting outcomes

Biomarkers play an especially critical role during Phase 2a of drug development (that is, during early proof-of-concept studies in a well-defined target population). This phase has been defined as "human trials providing sound evidence supporting the postulated effects of a new therapeutic drug product." The effects may be a relevant pharmacological action or a change in disease biomarkers, established surrogate endpoints, or clinical outcomes. Particularly, it should be noted that the relevant drug effects might be beneficial (that is, related to efficacy) and/or toxic (related to potential side effects).

To properly interpret the outcome of clinical studies in which the primary outcome variable is a biomarker that substitutes for the relevant clinical outcome, researchers must consider the complex relationship between the medical condition, the biomarker, and the clinical outcome. Simplified relationships between disease, biomarker (surrogate endpoint), and clinical outcome (clinical endpoint) are shown in Figures 1 and 2.

Hypertension is a straightforward example; it is associated with the biomarker elevated blood pressure-one of the few generally accepted surrogate endpoints in cardiovascular drug development-is associated with a clinically relevant outcome such as stroke, myocardial infarction, or renal failure. But the apparently straightforward association between disease, biomarker, and outcome oversimplifies the underlying (patho-) physiology. Thus, a disease generally is associated with several clinical manifestations, which may or may not serve as biomarkers, and which may be associated with the relevant clinical outcome. An example of a complex clinical condition is the metabolic syndrome, which is defined as insulin resistance (with or without manifest type 2 diabetes) and at least two of the following criteria-hypertension, dyslipidemia, obesity, and microalbuminuria. Clinically, this disease is associated with an increased cardiovascular risk.

The possibility that the clinical outcome is time dependent must also be taken into account when interpreting the outcome of studies using biomarkers (Figure 3). An example is the pharmacological group of phosphodiesterase (PDE) 3 inhibitors (amrinone and milrinone, for example) for which a significant increase in cardiac contractility could be demonstrated in subjects with congestive heart failure. Following chronic administration of these drugs, mortality increased in this subject population.

It can be concluded that biomarkers and clinical endpoints are not completely dependent on one another, but that there is a more or less well-defined relationship between the two. The use of a clinical biomarker as the primary outcome of a pharmacological intervention must therefore be interpreted very cautiously, taking into account the underlying (patho-)physiology, the possibility of false positive or false negative results, and time dependency of the clinical outcome.

In cardiovascular medicine, blood pressure is a biomarker with a clearly demonstrated relationship to clinically relevant endpoints. BP has been accepted as surrogate endpoint for cardiovascular morbidity and mortality, by both the medical community (as reflected in current World Health Organization guidelines) and regulatory authorities. Major prospective, randomized clinical studies (including meta-analyses) and epidemiological evidence (the Framingham Heart Study conducted over more than 40 years13) have clearly demonstrated that high blood pressure is associated with increased morbidity and mortality, and that lowering blood pressure is beneficial. Some of the main findings:

  • Treated hypertensive subjects have a 10-year mortality rate about 25% (for men) to 33% (for women) lower than that of untreated hypertensive subjects.

  • Reducing diastolic blood pressure by 5 mm Hg results in a decreased incidence of stroke (down 42%) and coronary artery disease (down 14%).

  • The maximum benefit seems to be achieved with a target blood pressure of 130–140/80–85 mm Hg.14

  • Antihypertensive treatment is indicated in both men and women and both young and elderly people.

To use a biomarker as a surrogate endpoint in a regulatory submission, researchers must validate the clinical study method for reproducibility, sensitivity, specificity, interrater reliability, variability, and bias. When using blood pressure as biomarker, for example, a considerable diurnal fluctuation exists. Because the maximum BP values occur in the late morning and early evening, ambulatory blood pressure monitoring (ABPM) is more appropriate than occasional office blood pressure recording.

If a company wants to use additional claims in its marketing application, it needs to use auxiliary biomarkers in the clinical program, such as:

  • Cardiac hypertrophy

  • Ventricular arrhythmias

  • Left ventricular function

  • Left ventricular hypertrophy

  • Endothelial dysfunction

  • Atherosclerosis

  • Cerebral lacunar lesions

  • Retinal vascular changes

  • Renal function

  • (Micro-) Albuminuria.15

Other biomarkers have been accepted as surrogates for more specific endpoints. For the marketing authorization of a new low-dose modified-release aspirin (acetylsalicylic acid) formulation, for example, the European Medical Evaluation Agency (EMEA) accepted that pharmacodynamic equivalence is shown based on the reduction of serum thromboxane B2 levels as surrogate for platelet aggregation and efficacy in secondary prevention of cardiovascular events.16 The decision is based on the following facts:

  • Thromboxane B2 is a metabolite of arachidonic acid, a powerful vasoconstrictor and inducer of platelet aggregation. Thromboxane B2 suppression (via inhibition of the enzyme cyclooxygenase) is the main mechanism of action of aspirin.

  • There is no evidence for other mechanisms of action that play a role in the clinical effects of the drug.

  • The lack of a dose effect in randomized clinical trials is consistent with the observation that thromboxane B2 suppression is saturable at low aspirin concentrations in vivo and in vitro. This finding supports the conclusion that thromboxane B2 suppression is a valid surrogate for the clinical effect of the drug.

Testing false

Below, examples of false positive and false negative biomarkers (used as surrogates for clinical endpoints) are discussed. The examples illustrate the need for a fact-based assessment of the relationship between a biomarker and the associated clinical outcome, even if such a relationship appears to be clearly evident.

False positive biomarker: Using anti-arrhythmics in patients with ventricular arrhythmias. For a long time, cardiologists have generally accepted the empirical clinical hypothesis that prevention of sudden death can be achieved by pharmacological suppression of cardiac arrhythmias. This clinical dogma was based on the observation that about 75% of sudden deaths are due to ventricular tachycardia or fibrillation and that complex ventricular arrhythmias are a risk factor for sudden death in patients with myocardial infarction (MI), especially when associated with left ventricular dysfunction. The CAST (Cardiac Arrhythmia Suppression Trial) study published in 1991 proved the cardiologists wrong.17

Even though arrhythmia as a risk factor was diminished by administering antiarrhythmic drugs such as encainide or flecainide (so-called class IC-antiarrhythmics) in 1498 post-MI subjects with ventricular extrasystoles and a low left ventricular function (ejection fraction <55%), mortality was significantly higher in subjects who received the antiarrhythmic drug than in the control group. Drug-induced alterations in intracardiac conduction properties most likely explain this result, but other effects of the antiarrhythmics, such as an effect on the adrenergic system, also may be responsible for the negative outcome associated with treatment in this population. This example highlights the need for extensive preclinical work in order to understand drug action and to increase the usefulness of a clinical biomarker such as arrhythmia, especially after unexpected results have been obtained in clinical studies.18

In conclusion, a clinically reasonable association between biomarker and clinical outcome does not necessarily mean that a treatment-related effect on this biomarker will improve the clinical outcome. This has been noted by the regulatory authorities, which require a sound database on the association between drug-induced effects on the biomarker and on relevant clinical endpoints before they accept a biomarker as definitive surrogate endpoint.

False negative biomarker: Using beta-blockers in patients with congestive heart failure. About 20 years ago, administering beta-blockers to patients with congestive heart failure was considered medical malpractice. This conclusion was based on pharmacological reasoning-mainly the negative inotropic effects of this drug class. Furthermore, cardiac output was considerably reduced by beta-blockers such as practolol (by as much as 12%), atenolol (by as much as 25%), and propranolol (by as much as 28%), when given intravenously to patients with ischemic heart disease-even those with no overt signs of heart failure.19 That study concluded that "it will probably not be acceptable to give beta-blockers, even those with a considerable degree of intrinsic sympathomimetic activity, to subjects with congestive heart failure." Today, beta-blockers belong to the standard-of-care of many subjects with congestive heart failure, based on a number of well-controlled mortality studies and scientific evidence that excess activity of the adrenergic system contributes to the pathophysiology of this widespread disease.

This example shows that the assumed lack of a clinically beneficial effect of a drug on a biomarker does not necessarily mean that the drug will not improve the clinical outcome if used properly. It may be hard for scientists in a pharmaceutical company, though, to convince management to continue with development of a drug once the negative data are available. Sound scientific reasoning is necessary when choosing appropriate biomarkers during early drug development.

Be wary

Biomarkers have the potential to be useful in early clinical drug development; they can assist in compound selection and in demonstrating early clinical proof of concept. Sound preclinical and clinical investigations are necessary, however, to show a linkage between disease, pharmacological mechanism, and clinical endpoints (efficacy and/or toxicity)-and to select an adequate method for conducting definitive clinical studies. The limitations of biomarkers as discussed earlier must always be considered in this context, and biomarkers should be considered suspect as surrogate endpoints in pivotal registration studies until they have been fully validated.


1. NIH-FDA Conference: Biomarkers and Surrogate Endpoints: Advancing Clinical Research and Applications. Abstracts. Disease Markers, 14 187–334 (1998).

2. Wayne A. Colburn, "Optimizing the Use of Biomarkers, Surrogate Endpoints, and Clinical Endpoints for More Efficient Drug Development," Journal of Clinical Pharmacology, 40 1419–1427 (2000).

3. William W. Chu, Robert S. Dieter, and Charles K. Stone, "Evolving Clinical Applications of Cardiac Markers: A Review of the Literature," Wisconsin Medical Journal, 101 49–55 (2002).

4. William W. Chu, Robert S. Dieter, and Charles K. Stone, "A Review of Clinically Relevant Cardiac Biochemical Markers," Wisconsin Medical Journal 101 40–48 (2002).

5. Lawrence J. Lesko and A. Atkinson, "Use of Biomarkers and Surrogate Endpoints in Drug Development and Regulatory Decision Making: Criteria, Validation, Strategies," Annual Review of Pharmacology and Toxicology, 41 347–366 (2001).

6. Committee for Proprietary Medicinal Products (CPMP), Note for Guidance on Statistical Principles for Clinical Trials (ICH Topic E9) CPMP/ICH/363/96 (CPMP, London, UK, 1998).

7. Lawrence J. Lesko, Malcolm Rowland, Carl C. Peck, and Terrence Blaschke, "Optimizing the Science of Drug Development: Opportunities for Better Candidate Selection and Accelerated Evaluation in Humans," Journal of Clinical Pharmacology, 40 803–814 (2000).

8. Committee for Proprietary Medicinal Products (CPMP), Points to Consider: The Assessment of the Potential for QT Interval Prolongation by Non-Cardiovascular Medicinal Products CPMP/986/96 (CPMP, London, UK, 1997).

9. Health Canada Sante Canada, Draft Therapeutic Products Directorate Guidance Document: Assessment of the QT Prolongation Potential of Non-Antiarrhythmic Drugs, (Health Canada Sante Canada, Ottawa, Ontario, Canada, 2001).

10. Committee for Proprietary Medicinal Products (CPMP), Note for Guidance on Safety Pharmacological Studies for Assessing the Potential for Delayed Ventricular Repolarization (QT Interval Prolongation) by Human Pharmaceuticals (ICH Topic S7B) CPMP/ICH/423/02 (CPMP, London, UK, 2002).

11. FDA (CDER) and Health Canada Therapeutics Product Division (TPD), Preliminary Concept Paper: The Clinical Evaluation of QT/QTc Interval Prolongation and Proarrhythmic Potential for Non-Antiarrhythmic Drugs, (FDA, Rockville, MD, 2002).

12. Committee for Proprietary Medicinal Products (CPMP), Position Paper on the Non-Clinical Safety Studies to Support Clinical Trials with a Single Low Dose of a Compound CPMP/SWP/2599/02 (draft) (CPMP, London, UK, 2002).

13. Pamela A. Sytkowski, Ralph B. D'Agostino, Albert J. Belanger, and William B. Kannel, "Secular Trends in Long-Term Sustained Hypertension, Long-Term Treatment and Cardiovascular Mortality," Circulation, 93 697–703 (1996).

14. Gastone Leoneti and Alberto Zanchetti, "Principal Results of Hypertension Optimal Treatment (HOT) Study and their Clinical Impact," Clinical Hemorheology and Microcirculation, 21 217–224 (1999).

15. Giuseppe Mancia, Antonio Lanfranchi, Carlo Turri, and Guido Grassi, "Can Good Surrogate End-Points Predict the Prognosis of Hypertensive Patients?" Journal of Hypertension, 16 (5) S3–S7 (1998).

16. Committee for Proprietary Medicinal Products (CPMP), Position Paper on the Regulatory Requirements for the Authorization of Low-Dose Modified Release ASA Formulations in the Secondary Prevention of Cardiovascular Events EMEA/CPMP/EWP/282/02 (CPMP, London, UK, 2002).

17. D.S. Echt, P.R. Liebson, L.B. Mitchell, R.W. Peters, D. Obias-Manno, A.H. Barker, D. Arensberg, A. Baker, L. Friedman, H.L. Greene, et al., "Mortality and Morbidity in Patients Receiving Encainide, Flecainide or Placebo: The Cardiac Arrhythmia Suppression Trial," New England Journal of Medicine, 324 781–787 (1991).

18. Craig M. Pratt, "Predicting Antiarrhythmic Performance," Journal of Cardiovascular Electrophysiology, 10 302–306 (1999).

19. T.L. Svendsen, O.J. Hartling, J. Trap-Jensen, A. McNair, and J. Biddal, "Adrenergic Beta Receptor Blockade: Hemodynamic Importance of Intrinsic Sympathomimetic Activity at Rest," Clinical Pharmacology and Therapeutics, 29 711–718 (1981).

Bernd Rosenkranz, FFPM, MD, PhD, is vice president, Clinical Development, Jerini AG, Invalidenstr. 130, D-10115, Berlin, Germany, +49 30 97893 500, fax +49 30 97893 599, email:

Related Content: