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

One of the primary reasons for the attrition of promising therapeutic agents from the drug development pipeline is the observation of treatment-related histologic injury to the kidney in animal toxicology studies. At the same time, the ability to recognize when a therapeutic intervention is damaging the kidneys in humans and to predict when removal of that agent would arrest further deterioration is severely limited. Progress in identifying, evaluating, and qualifying biomarkers of kidney injury would therefore yield great benefits in drug development.

This chapter describes the current state of biomarker use and development in four specific areas: (1) the qualification of new kidney safety biomarkers to bridge the gap between preclinical animal toxicology testing and early human clinical investigations; (2) the flawed gold standard (serum creatinine and blood urea nitrogen [BUN]) that plagues the assessment of new kidney safety biomarkers; (3) collaborations among drug developers, regulatory authorities, funding agencies, and investigators; and (4) the development of in vitro and other screening model systems that could identify lead candidates for drug development. In each of these areas, the major questions that need to be answered and the hurdles that need to be

overcome are presented. The chapter then describes a vision for the future development and use of kidney safety biomarkers in drug development. The workshop’s breakout session on kidney safety biomarkers, summarized at the end of the chapter, focused on what needs to be done to achieve this vision.

The current state of kidney drug development is characterized by several notable deficiencies: a limited ability to screen compounds to predict kidney toxicities; problems in identifying agents that would result in human kidney toxicity; difficulties in confirming that in some instances, kidney toxicities are specific for the tested animal species and are not necessarily relevant to humans; and a limited ability to monitor kidney damage associated with drugs that have been approved for use in humans despite their potential for nephrotoxicity because they provide health benefits or address unmet medical needs. Current approaches to toxicity testing and decision making can waste time and resources and may not identify or support development of the best lead candidates for human use.

The kidney is a major site of drug metabolism and elimination, so it is not surprising that toxicologic findings are more common in the kidney than in most other organs. Merck has reported that renal injury is the second-leading cause of drug attrition, after liver toxicity (Merck, unpublished findings). Indeed, commercially marketed drugs with known nephrotoxic potential are contributing factors in at least 25 percent of acute kidney injury in critically ill humans, causing significant increases in mortality, morbidity, and health care costs (Pannu and Nadim, 2008).

As noted, the current biomarker gold standard for kidney toxicity is levels of serum creatinine and BUN. These measures of kidney function are not sensitive indicators of structural injury, however, in part because of the excess capacity—or “renal reserve”—of the kidneys (Ferguson et al., 2008; Parikh and Devarajan, 2008). In the absence of more sensitive biomarkers to detect acute changes in renal damage, early indications of kidney injury cannot be monitored without histologic examination.

Histology provides accurate detection of kidney injury in preclinical animal studies. However, the current standard of care for evaluation of potentially nephrotoxic agents in human studies does not extend to surveillance with renal biopsies. As a result of this inability to monitor for early indications of kidney injury in human subjects, the development of compounds found to cause kidney damage in preclinical animal studies may be suspended, even when the relevance of the findings to humans has not been established. Drug toxicities seen in animal studies account for more than 30 percent of the attrition of compounds from drug development (Kola and

Landis, 2004). It is important to note that adverse findings from one species in animal toxicology studies are not always seen in other species, and only 40 to 60 percent of such findings are predictive of toxicities in human trials (Olson et al., 2000; Knight, 2008). Thus, promising compounds are dropped from the development pipeline even though adverse kidney effects in humans might not ever occur or might be manageable through monitoring with sensitive biomarker tests.

As novel renal biomarkers for clinical use are developed to detect early indications of nephrotoxicity, more reliance will be placed on these biomarkers in making critical drug development decisions than is currently placed on the results of animal studies (which as noted often are not accurately predictive of human toxicity). It may be hoped that this will both improve the safety of drugs and decrease the need to conduct exhaustive preclinical studies (Sistare and DeGeorge, 2007).

Novel renal biomarkers may also ultimately play a role as surrogate markers of drug efficacy. Many companies are working to develop new drugs for treating diabetes, hypertension, obesity, heart failure, hyperlipidemia, and transplant rejection, conditions in which the possibility of kidney injury is an ever-present risk. Since current biomarkers of kidney injury and function lack sensitivity and specificity, it is difficult to properly assess kidney status at baseline or after treatment. New drugs might actually be ameliorating or slowing the progression of kidney diseases, but without sensitive and specific measures of clinically meaningful renal injury and/or function, improvements during treatment of these diseases is extremely difficult to assess. At this time, improvements in renal function can be assessed only by conducting large and lengthy clinical trials (using doubling of serum creatinine as the efficacy endpoint). Once it has been established that novel renal biomarkers are predictive of future renal morbidity, they will likely be accepted as surrogate markers of clinically meaningful renal disease and be relied upon to assess renal benefits during drug development.

Biomarkers have already been developed as surrogates in other areas. For example, large-scale trials have shown that statins reduce the risk of adverse cardiovascular outcomes, and the reduction they cause in serum low-density lipoprotein (LDL) levels is predictive of this effect. LDL cholesterol is now accepted as a surrogate endpoint for regulatory marketing approval. But examples of such biomarkers remain rare, and the burden of proof for marketing approval is high.

Table 4-1 lists several promising translational biomarkers of acute kidney injury that have been proposed in published studies (Dieterle et al., 2008; Ferguson et al., 2008; Parikh and Devarajan, 2008). A number of

groups are collaborating to advance understanding of these biomarkers for specific uses (Box 4-1), some of which involve early drug development. Many of these biomarkers could contribute unique and specific information to an overall assessment of the state of kidney function, structural perturbation of the kidneys, and the healing response. A prospectively defined and systematically collected data set could allow some of these biomarkers to gain broad acceptance and qualification for monitoring renal safety in early clinical human registration trials.

Key questions that need to be addressed include the following:

• Qualification of a biomarker as fit-for-purpose for safety monitoring is an antecedent to qualification of that biomarker as fit-for-purpose for an efficacy outcome. Some biomarkers may be ideally suited to safety monitoring and early detection of kidney injury but

may not be appropriate as surrogate endpoints of clinical efficacy. Which data are critical for establishing that a specific biomarker is fit-for-purpose (Wagner et al., 2007)?

• How can existing data sets from animal and human studies be used most effectively to support qualification of safety biomarkers?

• How can diverse stakeholders best collaborate and set reasonable expectations and evidentiary standards for fit-for-purpose qualification of candidate safety biomarkers and generate any needed new data?

• How can the decision-making processes be made transparent to all stakeholders?

There are a number of opportunities for answering these questions:

• Biomarkers that appear to outperform changes in serum creatinine measurements in animal toxicology studies need to be identified, and a harmonized lexicon needs to be established for histopathology as the gold standard for these preclinical studies.

• A critical review of published studies needs to be conducted to identify the most promising new renal biomarkers. If possible, data from these studies should be collected into a usable shared database.

• Standardized methodologies need to be used for preclinical and clinical data collection, sample collection, histopathology interpretation, biomarker measurements, and data interpretation. These methodologies should reflect the collaborative efforts of experts and vested stakeholders, including sponsors, investigators, and regulatory scientists.

• Biomarker studies are needed to better understand such biological processes as the anatomical region and cell types perturbed in kidney injuries, the functions perturbed, and the recovery response.

• Tissue sample and biomarker banks need to be created. The use of both fresh samples and archived, frozen, or formalin-fixed samples should be optimized to minimize the drain on resources and maximize the knowledge gained from such studies.

• Prospective outcome-based human clinical trials need to be conducted to assess the relative performance of biomarkers in real time; the temporal profiles of changes in biomarkers relative to changes in renal function should be assessed by more traditional methods.

• Collaboration needs to be improved across groups that have common interests in the development of kidney safety biomarkers to foster common goals and shared access to critical samples, assays, data, and funding.

• Funding needs to be optimized through partnering with NIH.

• Biomarker thresholds need to be established. Normal ranges based on variance and intervention thresholds need to be determined based on medical experiences with a variety of study populations.

• The output from diverse work streams needs to be directed to key regulatory decision and policy makers at the same time that drug developers are kept fully informed through written communications and regular meetings that include all interested stakeholders.

• Issues surrounding the commercial development of biomarker assay panels for use in human and animal studies need to be considered, including technical and fiscal issues surrounding the identification, validation, commercial development, and acceptance of multi-marker panels.

When comparing the sensitivity and specificity of two or more tests, it is important to have a standard against which to make the comparison. Otherwise, one can see only how two measures correlate with one another, and comparison of the sensitivity and specificity of new biomarkers to creatinine or other more traditional biomarkers of kidney function and injury in the clinical setting is difficult. For ethical reasons, kidney biopsy—which would be an ideal standard for assessment of the performance of new biomarkers when nephrotoxicity is suspected in the clinical setting—is not possible. Unfortunately, as discussed above, the current gold standard of serum creatinine and BUN is seriously flawed.

There are several major research opportunities for developing an improved standard:

• Conduct studies to explore the utility of adjudication committees or more complex algorithms of renal dysfunction as better gold standards. Consider the value of incorporating other measurements, such as casts or fractional excretion of sodium, into the gold standard.

• Conduct preclinical studies to explore the best fit between other laboratory variables (aside from novel renal biomarkers) and histopathology changes.

• Explore through further discussion whether measuring the time lag between early alterations in biomarkers and subsequent persistent serum creatinine elevations and/or other traditionally used biomarker changes (perhaps in a patient population with no renal reserve) might be an adequate way to qualify renal biomarkers for use in Phase 1 clinical trials.

• Establish the relationship among novel biomarkers, kidney damage, and serum creatinine by analyzing ongoing studies in which human biopsy samples are available. Examples include protocol transplant biopsies and biopsies of patients with hematuria or proteinuria in whom serum creatinine may still be normal despite kidney injury as reflected by histopathology.

While the initial focus must necessarily be on safety monitoring, the ultimate application of these endpoint considerations to efficacy trials should be kept in mind. To this end, it will be necessary to establish the relationships between biomarker elevation and appropriate short-, intermediate-, and long-term clinical outcome measures, such as dialysis, death, and hospitalization days; repeated hospitalization, infections, or resource utilization; or progression to chronic kidney disease, end-stage renal disease, or cardiovascular disease.

Many stakeholders share an interest in defining the appropriate use of new kidney safety biomarkers. An efficient approach to assessing the performance of emerging biomarkers, as well as gathering control patient data on new biomarkers, would be to add such measurements to ongoing animal and clinical trials. However, there currently are few incentives and many disincentives to adding measurements from unqualified exploratory safety biomarkers to such regulated studies (Sistare and DeGeorge, 2008). Key questions that must be addressed include the following:

• How can regulatory authorities establish and communicate an official regulatory policy to support the innovative development of safety biomarkers within the context of product development?

• Can regulatory authorities provide encouragement or support to advance the evaluation of a declared set of safety biomarkers using samples from ongoing registration studies?

• Can other, nonregulatory federal agencies provide access to samples from clinical trial sample sets that would allow a prioritized list of new safety biomarkers to be developed?

Major opportunities exist for answering these questions:

• Transparent and realistic communications need to occur between regulatory authorities and drug development scientists.

• Disincentives under the current drug development framework need to be removed.

• Issues of patient consent in clinical trials pertaining to testing of novel biomarkers need to be resolved.

• Exploratory studies and studies of negative controls are needed.

• A role for NIH-sponsored clinical natural history studies and clinical trials in this overall process needs to be identified.

• Roles for public–private partnerships and patient advocacy groups need to be identified.

• FDA–industry–NIH partnerships for monitoring drug safety and toxicity and providing broad access to study samples need to be explored.

In vitro and other screening model systems that could definitively rule out human kidney toxicity caused by test compounds and identify specific kidney toxicities in animal test species would be exceedingly helpful in drug development. Key questions in this area include the following:

• Are such systems viable and close to being established?

• How can their evolution be fostered, and how can they be efficiently validated?

There are several major opportunities for answering these questions:

• Studying the role and limitations of current primary cell cultures and of embryonic stem cells for assessing new biomarkers.

• Studying the utility of other in vitro model systems of nephrotoxicity (e.g., in zebrafish).

• Creating partnerships among academia, the pharmaceutical industry, NIH, industry technology providers, and advocacy groups to explore the advancement of in vitro and other model systems for early drug toxicity screening.

Table 4-2 summarizes a vision of the future for avoiding and addressing kidney safety issues in early drug development. The main features of such a vision are as follows:

• In vitro test systems would be available to predict the risk of kidney toxicity, including glomerular and tubule cell injury and functional

changes, and to assess whether the toxicity is species-specific or is relevant to all species, including humans.

• A limited number of rodent and nonrodent model systems would be available for short-term toxicology testing using a validated, easily accessible biomarker or multiple biomarkers that might include a combination of tissue and accessible biomarkers (e.g., genomic, metabolomic, protein, or imaging biomarkers). These biomarkers

would predict dose-dependent acute kidney injury in terms of dysfunction, anatomic alteration, or structural damage.

• Readily available qualified human biomarkers would be available to diagnose or rule out the involvement of specific anatomic regions, to assess the severity of an injury, to signal the need for early intervention, and to monitor the reversibility of injury-associated processes.

• Sponsors and investigators would be encouraged by regulatory authorities to demonstrate that a compound lacks the potential to injure the kidney or that kidney toxicity can be monitored and managed using qualified safety biomarkers that bridge the gap between preclinical animal studies and early human clinical trials.

• Qualified predictive biomarkers for kidney toxicity in humans would be available that would outperform functional measures currently employed to predict important clinical outcomes and the efficacy of intervention strategies. These biomarkers would be defined according to their applicability to patients with chronic kidney disease, as well as those with no known kidney disease.

• Where kidney involvement is a known comorbidity of disease, qualified biomarkers would be available to assess the effects of a drug on kidney processes in Phase II and III trials.

• Qualified biomarkers for kidney injuries would be used to diagnose and stage kidney diseases with pathologies related to kidney toxicities.

The breakout group that discussed biomarkers for kidney toxicity focused on prioritizing what needs to be done to achieve the future vision outlined above. Participants discussed the most critical obstacles and data gaps that need to be addressed in the four areas reviewed in the above section on the current state of renal biomarker development. In the plenary session following the breakout, Prasad Devarajan of the Cincinnati Children’s Hospital Medical Center, University of Cincinnati, presented the group’s main conclusions:

• Greater understanding of biomarkers—The group concluded that new biomarkers should outperform those already available. The current gold standard for preclinical biomarker discovery is animal toxicity studies based on histopathology, and the use of that standard should continue. Also needed are biomarkers that indicate the anatomic site of kidney injury in animal studies because different agents can affect different nephron segments, and biomarkers

should provide temporally ordered information regarding the stage of kidney injury. Needed as well is greater understanding of what biomarker combinations reveal about the stage of kidney injury or repair.

• Collection and analysis of data—Mechanisms for accumulating standardized preclinical data and for sharing and interpreting the data should be established through rules set by consortia. For example, assessments of histopathology need to be harmonized among the many groups working on kidney safety issues. Because data on long-term clinical outcomes can take many years to gather, it is important that short-term outcomes—including serum creatinine, dialysis requirements, and mortality—be correlated with biomarkers.

• Correlations with biopsies—Efforts should be made to correlate human biopsy histopathology with biomarkers. While not all patients with nephrotoxicity are biopsied, more can be learned from the biopsies that are being done, such as the protocol kidney transplant biopsies that are frequently performed in many institutions. Biopsies of patients with blood in the urine and glomerular disease should be simultaneously correlated with noninvasive biomarkers.

• National biorepositories—Standardized national biorepositories need to be established to enable comparison of biomarkers from ongoing clinical trials. Researchers should have access to de-identified samples, and it should be possible to share data generated from these samples. NIH has a public–private partnership office that could coordinate this cooperation. In addition, a system is needed for sharing clinical data, including data gathered by industry.

• Incentives, dialogue, and partnerships—Incentives could be structured to promote investigations of new safety biomarkers in regulated studies that support new product development. Incentives also need to be considered for exploratory studies and negative controls. An open dialogue between industry and regulatory agencies could identify and refine such incentives. In addition, an FDA–NIH partnership could consider funding research on biomarker development for monitoring drug safety and toxicity, making the FDA a funding as well as a regulatory agency.

• Preclinical model systems—In vitro and other screening models need to be developed for detecting nephrotoxicity. Clinical and animal studies are expensive. Cell culture systems have not been very reliable, but new systems, such as embryonic stem cells and zebrafish models, are promising, and additional models may emerge. A partnership between industry and NIH could accelerate

the development of these in vitro and other systems for detecting a compound’s nephrotoxicity potential.

Dieterle, F., E. Marrer, E. Suzuki, O. Grenet, A. Cordier, and J. Vonderscher. 2008. Monitor-Monitoring kidney safety in drug development: Emerging technologies and their implications. Current Opinion in Drug Discovery & Development 11(1):60–71.

Ferguson, M. A., V. S. Vaidya, and J. V. Bonventre. 2008. Biomarkers of nephrotoxic acute kidney injury. Toxicology 245(3):182–193.

Knight, A. 2008. Systematic reviews of animal experiments demonstrate poor contributions toward human healthcare. Reviews on Recent Clinical Trials 3(2):89–96.

Kola, I., and J. Landis. 2004. Can the pharmaceutical industry reduce attrition rates? Nature Reviews. Drug Discovery 3(8):711–715.

Olson, H., G. Betton, D. Robinson, K. Thomas, A. Monro, G. Kolaja, P. Lilly, J. Sanders, G. Sipes, W. Bracken, M. Dorato, K. Van Deun, P. Smith, B. Berger, and A. Heller. 2000. Concordance of the toxicity of pharmaceuticals in humans and in animals. Regulatory Toxicology and Pharmacology: RTP 32(1):56–67.

Pannu, N., and M. K. Nadim. 2008. An overview of drug-induced acute kidney injury. Critical Care Medicine 36(Suppl. 4):S216–S223.

Parikh, C. R., and P. Devarajan. 2008. New biomarkers of acute kidney injury. Critical Care Medicine 36(Suppl. 4):S159–S165.

Sistare, F. D., and J. J. DeGeorge. 2007. Preclinical predictors of clinical safety: Opportunities for improvement. Clinical Pharmacology and Therapeutics 82(2):210–214.

Sistare, F. D., and J. J. DeGeorge. 2008. Applications of toxicogenomics to nonclinical drug development: Regulatory science considerations. Methods in Molecular Biology 460:239–261.

Wagner, J. A., S. A. Williams, and C. J. Webster. 2007. Biomarkers and surrogate endpoints for fit-for-purpose development and regulatory evaluation of new drugs. Clinical Pharmacology and Therapeutics 81:104–107.