Accelerating the Development of Biomarkers for Drug Safety: Workshop Summary (2009)

In the 1990s, reports of potentially fatal cardiac arrhythmias in adverse event data focused attention on the potential of several drugs to cause cardiac toxicity. One effect of these drugs was to prolong the interval between the onset of the Q wave and the conclusion of the T wave in the heart’s electrical cycle—which is known as QTc when corrected for heart rate. This association with QTc prolongation and cardiac arrhythmias led to the removal of a series of drugs from the market, including terfenadine in 1998, astemazole and grepafloxacin in 1999, and cisapride in 2000. QTc is one of the oldest and best-known safety biomarkers used throughout drug development. The effect of a drug on QTc is an important input to regulatory decision making and has a major impact on how pharmaceutical companies design and prioritize drug development programs.

Compared with the newer safety biomarkers discussed later in this chapter, QTc has a number of strengths and weaknesses (Table 3-1). Among its strengths are that the technology needed to measure it is established and nearly universally available; a great deal is known about the molecular mechanisms of the ion channels that affect ventricular repolarization; a number of well-established in vitro and in vivo models exist; there is substantial clinical experience with patients who have a congenital prolonged

QT (LQT) syndrome; and a wide array of stakeholders are interested in advancing the understanding and use of this biomarker.

Despite these strengths, however, QTc also has several weaknesses as a biomarker for safety. First, there is no consensus on the optimal method of acquiring, measuring, and analyzing the QTc interval. This is due in part to the nature of the signal, which has low frequency and low amplitude, has a poor signal-to-noise ratio, is intrinsically variable, and is affected by a number of important confounding factors. Second, the link between the experimental models of QTc and the occurrence of rare and unpredictable clinical events is weak, in part because insufficient data have been collected to close this gap. Specifically, clinical epidemiology data have not been collected that would define the probability of an episode of the ventricular tachycardia known as torsade de pointes based on the QTc interval.

It should be noted that, while many biomarkers are used to understand a wide range of cardiovascular conditions—such as hyperlipidemia, inflammation, and ischemia—the scope of the discussion in this session of the workshop was limited to biomarkers of electrophysiologic toxicity, in particular, those related to QT interval prolongation.

This chapter begins by describing the regulatory response to the recognition that cardiac events were resulting from adverse reactions to drugs, the responses of drug developers, and effects on physician decision making. This is followed by a review of issues related to the development of potential cardiac safety biomarkers other than QTc, with a particular focus on troponin, and the possible contributions to this work of the Cardiac Safety Research Consortium (CSRC). Some lessons learned from experience to date with the development of cardiac safety biomarkers are then summarized. The chapter ends with highlights from the breakout discussion of key steps necessary for further progress.

The recognition that cardiac events were being caused by adverse reactions to drugs led to a variety of regulatory responses. In 1997, the FDA and the International Conference on Harmonisation (ICH) issued Guidance for Industry: S6 Preclinical Safety Evaluation of Biotechnology-Derived Pharmaceuticals (FDA, 1997). This was followed in 2001 by Guidance for Industry: S7A Safety Pharmacology Studies for Human Pharmaceuticals (FDA, 2001). Both of these documents stated that cardiovascular safety testing should be performed on new drugs, but provided no specific guidance on how this testing should be conducted. In 2001, the FDA announced that in fall 2002, it would begin collecting raw electrocardiogram (ECG) data from sponsors, and in 2002 a “points to consider” document was jointly authored by the FDA and Health Canada (FDA, 2002). This was followed by FDA/ICH guidance documents providing more specific recommendations regarding clinical (E14) (FDA, 2005a) and preclinical (S7B) (FDA, 2005b) testing approaches. The E14 guidance called for “thorough QT” (TQT) studies of new drugs to assess their potential for causing torsade de pointes. Even prolongation of QTc by just a few percent was considered to be clinically relevant. The FDA then established an interdisciplinary team to handle the review of QTc-related protocols and studies, to ensure a uniform response, and to accumulate experience in this area.

As the regulatory response was being crafted, the FDA made a public appeal for the development of standards for digital ECG data. This action was based on the idea that it will be critical to review the ECGs from TQT studies. Such a data standard was developed in 2002 and formally adopted by the Health Level 7 (HL7) standards organization in early 2003.²

As the data standard was being finalized, the FDA entered into a Cooperative Research and Development Agreement with Mortara Instruments to develop a web-accessible repository for conforming digital ECG data. This repository came online in 2005 and now hosts more than 2.5 million digital ECGs collected from more than 150 clinical studies.

As the ICH S7B and E14 guidance documents were being developed, responses from the pharmaceutical industry were mixed. In general, industry appreciated clarification of the standards for preclinical and clinical assessments of the effects of a drug on ventricular repolarization. In particular, industry was pleased that E14 created a clear definition of a compound with no QTc risk and made it clear that no further evaluation of QTc would be necessary for these compounds.

However, industry representatives raised two concerns related to the E14 guidance. First, E14 specified that every systemically available small molecule would require a clinical TQT study even if the results of the extensive preclinical studies related to ventricular repolarization outlined in S7B were completely normal. Second, E14 set an extremely high bar for declaring that a compound posed no QTc risk: at supratherapeutic exposures, a compound had to demonstrate an increase in QTc of less than 5 milliseconds (ms) (mean) or 10 ms (upper confidence limit) in a study that demonstrated assay sensitivity by detecting an increase in QTc of a similar magnitude with a positive control (usually moxifloxacin).

These two concerns were focused primarily on a fear that very small signals in QTc would be identified in compounds when there was no theoretical risk, when no preclinical evidence suggested future problems, and when early clinical evidence showed no signs of QTc prolongation. The initial lack of understanding of what it means when a compound demonstrates a 5–10 ms increase in QTc generated considerable uncertainty in drug development. In particular, drug developers asked questions such as the following:

• What was the clinical significance of such a small increase in QTc?

• What additional studies would be necessary in later phases of drug development to clarify the clinical significance of an increase in QTc of this magnitude?

• How would these additional studies affect the timelines and costs of drug development?

• What is the likelihood that these additional data would be able to offset the perceived risk associated with a small but clearly documented increase in QTc from a TQT study?

• How should a company weigh this potential increase in risk against the potential benefits of a drug?

• How would these issues be described on the drug label?

Because of the uncertainty surrounding these questions, some pharmaceutical and biotechnology companies avoided developing compounds with any potential for this liability. In the process of prioritizing compounds in a portfolio, companies began looking for ways to kill compounds with any potential QTc liability. Any increase in QTc in preclinical studies generated the perception that the compound would face enormous hurdles in drug development. Some companies began to discontinue compounds in development solely because of in vitro studies demonstrating an interaction with the hERG channel (a potassium ion channel involved in ventricular repolarization), even in the absence of evidence of prolonged QTc during in vivo animal studies. In addition, as compounds advanced through development, companies feared being penalized for evaluating supratherapeutic exposures and attempted to minimize this risk by limiting the maximum doses studied.

With regard to drug development, the ultimate success of the E14 and S7B guidance documents will be realized when there is a shared understanding between pharmaceutical companies and regulatory agencies of the clinical significance of a small increase in QTc interval in the context of the possible benefits of a new molecular entity. Excessive focus on this biomarker in the absence of true clinical risk could stifle innovation and lead to an unfortunate decision to discontinue the development of a drug that could offer patients benefits outweighing the actual risk.

One solution to this potential conundrum is to create an environment in which regulatory agencies, academics, and industry scientists can collaborate to better understand the link between the safety biomarker (in this case QTc) and the event it is intended to predict (in this case torsade de pointes). All parties involved would benefit from improved clinical epidemiology and greater understanding of how to measure and use this safety biomarker. If successful, this type of collaboration would likely result in better decision making that would place the risks of a drug in the context of its benefits. The potential of this approach is demonstrated by the CSRC, discussed later in this chapter.

The regulatory guidance discussed above has important effects on physician behavior and decision making. The provision of information to physicians on a product insert or label regarding how a drug might affect the QTc interval raises a number of important questions:

• How do physicians use the information on the label?

• How successful are physicians in measuring the QTc interval when instructed to do so by the label?

• How do physicians make risk/benefit decisions for an individual patient?

• Are physicians avoiding potentially beneficial medications because of the fear of a small increase in QTc?

• What is the impact of including new warnings on the labels of drugs that have been used for a long period of time (e.g., methadone)?

The recent developments related to QTc provide insight into the complexity facing the development of other cardiac safety biomarkers. Some examples of biomarkers that might merit further attention because of their link to cardiac morbidity and mortality include

• heart rate,

• blood pressure,

• lipids,

• troponin,

• C-reactive protein (CRP),

• brain or B-type natriuretic peptide (BNP),

• ex vivo platelet aggregation, and

• imaging biomarkers (cardiac magnetic resonance imaging).

It is beyond the scope of this chapter to discuss all of these potential cardiac safety biomarkers in any depth. However, examination of one example highlights both the challenges involved and the potential path forward.

Troponin is a protein complex involved in contraction in cardiac muscle. Subtypes of troponin can be sensitive indicators of damage to heart muscle caused by myocardial infarction or other cardiovascular conditions, and these uses are well established and supported by considerable research. Cardiac troponin also has been recognized as a potential biochemical marker of subclinical myocardial injury. Much less is known, however, about the use of troponin to identify drug-induced cardiotoxicity. For example, troponin has been studied as a potential biomarker of cardiotoxicity associated with two chemotherapeutic agents—the anthracycline doxorubicin and the humanized monoclonal antibody trastuzumab. Since the toxicity associated with anthracyclines varies considerably among individuals, the use of cardiac troponin has been suggested as potentially important in planning and monitoring treatment to allow maximum anthracycline dosages

without causing severe cardiac damage, and in developing preventative strategies to limit cardiomyopathy in later life. A complicating finding is that the early left ventricular dysfunction associated with doxorubicin may be reversible in the short term, even though clinical heart failure may not appear until much later.

Trastuzumab is an example of a drug whose use could be optimized by employing an appropriate biomarker. Trastuzumab has been used to prolong the lives of women with advanced breast carcinoma who have overexpression of the HER2 oncogene. Preclinical animal studies on mice and monkeys did not reveal cardiac toxicity for this drug; however, subsequent clinical trials demonstrated an unexpectedly high incidence of such toxicity. Despite the reversibility of trastuzumab-induced cardiac changes in most cases, this toxicity frequently leads to discontinuation of antibody therapy. If cardiac troponin were shown to be a reliable biomarker of patients at risk for this toxicity, it could help optimize the use of trastuzumab.

A number of important questions are raised by this approach:

• When should cardiac troponin be measured, and how should it be quantified?

• Which cardiac troponin assay should be used?

• What is the appropriate threshold to establish that an increase in cardiac troponin will be clinically significant?

• How will that threshold be determined in the context of the potential benefits of the drug?

• What should be done about events that are biochemically detectable but below that threshold and therefore may be clinically insignificant?

• How should investigators manage elevations in troponin in clinical studies?

• Which compounds need to undergo a cardiac troponin evaluation preclinically?

• Are the preclinical models sufficiently predictive? If not, which compounds warrant a cardiac troponin evaluation in clinical studies?

• How can a negative cardiac troponin evaluation be defined? Will a positive control be necessary to determine assay sensitivity? How would a positive control be used?

To examine the potential of QTc and other cardiac safety biomarkers, the Health and Environmental Sciences Institute (HESI), the FDA, and the CSRC hosted an open think tank forum on October 6–7, 2008, titled “Integrating Preclinical and Clinical Issues in Cardiac Safety: Translational Medicine Meets the Critical Path.” Experts from academia, industry, and the FDA gathered to discuss key topics in cardiac safety assessment, with

a particular focus on the translational gaps between the preclinical and clinical perspectives.

Plenary presentations titled “Collaboration, Critical Path, and Cardiac Safety: The FDA View” and “How Can Collaborations in Cardiac Safety Efforts Best Impact the Regulatory Landscape?” set the stage for examining the value of the collaborations promoted by the HESI and CSRC consortia. Organizational updates from HESI and CSRC summarized the challenges of and solutions to data-sharing processes, and presented the first proof-of-concept report illustrating the sharing of data from a number of companies in the ECG warehouse. The forum’s agenda encompassed the exploration of a number of potential biomarkers in addition to QTc, and included discussion of the following questions:

• Cardiotoxicity and troponin: Where do they fit in drug development?

• Preclinical and clinical testing for QTc proarrhythmia: How do they relate to one another and to the risk of life-threatening arrhythmic events?

• QTc evaluation of non-QTc proarrhythmia: What is appropriate preclinical and clinical testing?

• Biologics and large molecules: How should proarrhythmia and myotoxicity be evaluated?

• Risks and benefits of developing drugs with safety signals: What are the challenges?

• New horizons for cardiac safety programs: Do we need “thorough” blood pressure, heart rate, platelet, and lipid studies?

As the ECG warehouse was coming online, the FDA and the Duke Clinical Research Institute initiated the CSRC, a public–private partnership, to “advance scientific knowledge on cardiac safety for new and existing medical products by building a collaborative environment based upon the principles of the FDA’s Critical Path Initiative as well as other public health priorities.”³ This initiative brought together pharmaceutical companies, clinical research organizations, and academic partners in an effort to leverage the ECG warehouse and associated clinical data for mutual benefit.

The implementation of the CSRC has faced many challenges related to governance, infrastructure, resources (both funds and staff time), intellectual property, antitrust and other legal issues, and how to get companies to share data in a collaborative environment. Many of these challenges

have been or are being overcome. Companies have begun to share data, and CSRC research teams—including industry scientists, academics, and regulators—have begun to make progress on a number of projects.

An important accomplishment of the CSRC has been enhancing communication and education by promoting dialogue among scientists from the pharmaceutical industry, academia, and regulatory agencies. The CSRC has established common ground and an environment in which difficult issues can be discussed outside of formal regulatory channels. These discussions have included methods for evaluating the effects of chemotherapeutic agents and large molecules (antibodies and biologics) on QTc, as well as different statistical approaches to evaluating the effect of a drug on QTc, including concentration–response (PK-QTc) modeling. Recently, a number of pharmaceutical companies agreed to allow the FDA to share data from the ECG warehouse to create a meaningful data set that will enable companies and scientists to enhance the use of old measurements of QTc and develop new measurements of ventricular repolarization. This data set will also provide the opportunity to gain insight into the effect of moxifloxacin (the most commonly used positive control in TQT studies), including a better understanding of outliers and nonresponders. The potential will exist for informative studies in pharmacogenomics that might not be possible in a single company.

Combined with the technological and regulatory advances that have been achieved over the past few years, the CSRC has the potential to generate significant improvements in the utility of QTc as a safety biomarker. But it is not clear at this time whether the CSRC will be able to generate the clinical epidemiology studies and data necessary to provide a more refined link between drug-induced QTc prolongation and the risk of developing torsade de pointes. The next few years will determine whether the collaborations within the CSRC will generate the data sets necessary to provide meaningful and relevant answers to questions that limit the use of QTc as a safety biomarker.

The CSRC also hopes to foster collaborations among industry, academia, and regulatory agencies to further the development of new cardiac safety biomarkers. These advances in biomarker development will require investments in basic science to better elucidate the molecular mechanisms of cardiac toxicity and in preclinical models and clinical data to allow evaluation of the use of biomarkers. A coordinated approach to this effort is important to ensure that scientific issues are addressed appropriately, that regulatory strategies are crafted, that an infrastructure is developed to collect industrywide experience, and that the proper public–private partnerships are forged to profit from the aggregate experience.

A number of important lessons that may be applicable to the development of other safety biomarkers have been learned from the development of regulatory guidance on evaluating the potential of drugs to prolong QTc; the technological advances that enabled the formation of the ECG warehouse; and the healthy dialogue that has taken place among the pharmaceutical industry, academia, and regulatory agencies through the CSRC. This series of events has yielded a fairly complete (but still evolving) system for addressing a public health issue through regulatory and technical developments. The historical account makes the endeavor look like a coordinated response, but that is an inaccurate perception. Rather, individuals who recognized what needed to be done next made sure those steps were taken. The original “points to consider” document had its roots in a document authored by Health Canada’s Collette Strnad. The effort to develop a digital ECG data standard, which involved a team of people from pharmaceutical companies, clinical research organizations, device manufacturers, and academia, was initiated and managed by Scott Getzin of Eli Lilly. The CSRC came into being largely through the efforts of Christopher Cabell, then at the Duke Clinical Research Institute. Had any of these individuals failed to become involved when and to the extent that they did, the result would most likely have been significant delay and a suboptimal response. There is a pressing need to develop a quality-assured response to other perceived biomarker-based health risks.

The breakout group on cardiac toxicity identified several key steps necessary for progress on both the enhanced use of QTc as a biomarker and other biomarkers that can supplement the information provided by QTc. In the plenary session following the breakout, Alastair Wood described the group’s main conclusions.

The collection, annotation, curation, and submission of data need to be standardized across the entire research spectrum, including NIH, the FDA, and academia. Annotation and curation of data are especially important so that data will be usable, standardized, and accessible.

Without standardization, it is impossible to look across databases or even different studies and make comparisons or compare outcomes against biomarkers. In addition, patient data need unique identifiers, since frequently it is difficult to identify a patient who took part in more than one

study or developed toxicity after an event. Even drug names need to be better identified, since the trade names of drugs can change.

It is important to relate biomarkers to mechanisms of toxicity. Mechanistic understanding can be used to generate hypotheses that can then be tested experimentally. Identifying a biomarker can help clarify a mechanism and vice versa. And understanding mechanism can provide information on long-term clinical outcomes and on biomarkers that do and do not correlate with these outcomes.

Access to data held by the FDA and by private companies would be valuable for those involved in the development of biomarkers. For example, noncompetitive access to old drug data would benefit multiple stakeholders. Removing restrictions on access to FDA data would require legislation.

In general, broader access to compounds and past data associated with those compounds could improve productivity. For example, compounds that were abandoned because of toxicity concerns could yield data that relate to potential biomarkers currently being studied. Such data could reveal correlations or their lack and would allow for comparisons across studies.

A variety of organizations need to assume or be assigned responsibility for bringing stakeholders together and arranging for funding to advance the development of cardiac safety biomarkers. Among the issues that need to be resolved is who will support the needed research, what mechanisms will drive the research, and what is the proper balance of incentives and requirements to foster participation.

As part of this allocation of responsibilities, NIH’s role in biomarker development needs to be rethought and redefined. If NIH interprets its role too narrowly, it may not be willing to support clinical research that can have a major impact on patient outcomes. One option would be to convene a standing group including representatives of the National Heart, Lung, and Blood Institute, the FDA, industry, and academia to identify and prioritize high-impact opportunities in terms of public health and to recommend specific targets for research funding. Topics that NIH should consider include technology and animal model development aimed at translation to human studies; development of biomarkers through detailed studies of human

genomics, proteomics, and metabolomics; human studies to validate bio-markers in adequately sized longitudinal studies; and definition of appropriate institutional roles in the development of standards. Such initiatives are beyond the capability of either the FDA or most private companies unless they work together within a collaborative framework.

FDA (Food and Drug Administration). 1997. International conference harmonization guidance for industry: S6 preclinical safety evaluation of biotechnology-derived pharmaceuticals. http://www.fda.gov/cder/guidance/1859fnl.pdf (accessed October 17, 2008).

FDA. 2001. International conference harmonization guidance for industry: S7A safety pharmacology studies for human pharmaceuticals. http://www.fda.gov/Cber/gdlns/ichs7a071201.pdf (accessed October 17, 2008).

FDA. 2002. The clinical evaluation of QT/QTc interval prolongation and proarrhythmic potential for non-antiarrhythmic drugs. Preliminary concept paper. http://www.fda.gov/ohrms/dockets/ac/03/briefing/pubs%5Cprelim.pdf (accessed October 17, 2008).

FDA. 2005a. International conference harmonization guidance for industry: E14 clinical evaluation of QT/QTc interval prolongation and proarrhythmic potential for non-antiarrhythmic drugs. http://www.fda.gov/cber/gdlns/iche14qtc.pdf (accessed October 17,(accessed October 17, 2008).

FDA. 2005b. International conference harmonization guidance for industry: S7B nonclinical evaluation of the potential for delayed ventricular repolarization (QT interval prolongation) by human pharmaceuticals. http://www.fda.gov/cder/guidance/5533dft.htm (accessed October 17, 2008).