Recommendations for Soluble Biomarker Assessments in Osteoarthritis Clinical Trials

Abstract

Objective

To describe requirements for inclusion of soluble biomarkers in osteoarthritis (OA) clinical trials and progress toward OA-related biomarker qualification.

Methods

The Guidelines for Biomarkers Working Group, representing experts in the field of OA biomarker research from both academia and industry, convened to discuss issues related to soluble biomarkers and to make recommendations for their use in OA clinical trials based on current knowledge and anticipated benefits.

Results

This document summarizes current guidance on use of biomarkers in OA clinical trials and their utility at 5 stages, including preclinical development and phase I to phase IV trials.

Conclusions

Biomarkers can provide value at all stages of therapeutics development. When resources permit, we recommend collection of biospecimens in all OA clinical trials for a wide variety of reasons but in particular, to determine whether biomarkers are useful in identifying those individuals most likely to receive clinically important benefits from an intervention; and to determine whether biomarkers are useful for identifying individuals at earlier stages of OA in order to institute treatment at a time more amenable to disease modification.

Introduction

In 1996, a task force of the Osteoarthritis Research Society International (OARSI) published recommendations for the conduct of clinical trials in osteoarthritis (OA)[1]. At that time, they presciently stated that, “as part of the advancement of science, it is expected that OA protocols will contain both validated measures and investigational outcome measures still requiring validation”. Since then the field of OA-related biomarkers has undergone considerable advancement and there are many biomarkers currently in various states of “validation”, or to be more precise, qualification[2, 3]. Qualification is the evidentiary process of linking a biomarker with biological processes and clinical end points[4-6]. More recently, qualification has been described as a conclusion that within the stated context of use, the drug development tool (i.e. biomarker) can be relied on to have a specific interpretation and application in drug development and regulatory review[7]. Thus, the Steering Committee of the OARSI Task Force on Recommendations for Conducting Clinical Trials in Osteoarthritis deemed it appropriate and necessary to include a description of the process and issues related to inclusion of biomarkers in OA clinical trials.

A biomarker is defined as a characteristic that is objectively measured and evaluated as an indicator of normal biologic processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention[6]. Among other things, biomarkers in the OA field can be used in drug and other therapeutics development, treatment monitoring and the future basis of personalized evidence-based action plans, disease monitoring and prognosis. The long-term goal is to contribute to strategies that improve the lives of people with OA or who are at risk of OA. Biomarkers include not only soluble analytes measured in biospecimens such as blood and urine, but anatomic biomarkers such as findings detected by radiography and magnetic resonance imaging (MRI), physiological measurements such as gait analyses and even histological measurements produced as a result of a joint tissue biopsy such as a synovial biopsy. The goal of this report is to discuss issues related specifically to soluble biomarkers and to make recommendations for their use in OA clinical trials based on current knowledge and anticipated benefits.

Biomarker development generally stems from an understanding of the pathophysiology of a disease[4]. It is little wonder then that biomarker development in OA is burgeoning as we gain a more clear understanding of the disease, its stages, and various phenotypes. Many of the existing OA-related biomarker assays have grown directly out of an understanding of joint tissue metabolism[8, 9] and reflect catabolism or anabolism of joint tissues. To the extent that many of the products of joint tissue metabolism can stimulate an innate immune response (for instance fibronectin and hyaluronan fragments)[10-13], and be detected in biospecimens[3, 14, 15], there is a potential to have some OA-related biomarkers that are directly involved in the pathophysiology of an OA disease outcome. This is an enviable scenario that has rarely been achieved in other disease areas but represents the holy grail of biomarker development. Probably the paradigm of such a biomarker is cholesterol; high serum total and LDL-cholesterol are regarded as reflecting the pathophysiological events leading to atherosclerotic cardiovascular disease[4]. Work is currently underway in the OARSI/Foundation for NIH OA Biomarkers Consortium study[3] to qualify a panel of OA-related soluble markers (serum and urine) as predictors of a clinically relevant outcome consisting of the combination of knee OA radiographic worsening and pain worsening .

A great deal of guidance exists, and is growing rapidly, regarding the topic of biomarker development for use in clinical trials (described below). Reporting in 2010[16], and in 2012 applied to omics technologies[17], the Institute of Medicine (IOM) recommended a framework for the evaluation of biomarkers to lead to their clinical application. They recommended that the Food and Drug Administration (FDA) use the same degree of scientific rigor for evaluating biomarker use across all regulatory areas (drugs, medical devices, biologics, foods and dietary supplements) and for this purpose proposed a three-part framework for biomarker evaluation: (1) Analytical validation—evaluation of the analytical performance of the test to ensure biomarker tests are reliable, reproducible, and adequately sensitive and specific; (2) Qualification—to ensure the biomarker is associated with the clinical outcome of concern; (3) Utilization analysis—to determine that the biomarker is appropriate for the proposed use. They further recommended that the initial evaluation of analytical validation and qualification should be conducted separately from a particular context of use. They concluded that, “Modern medicine depends on biomarkers”[18].

We will cover each of these topics, first focusing on the issues and guidance related to the process of qualification of biomarkers for different contexts of use, second on their utilization pertaining to each phase of OA clinical trials, and third we discuss aspects of analytical validation of OA-related biomarkers. Although we refer primarily to the rich guidance available from the FDA and highlight, when available corresponding information provided by the European Medicines Agency (EMA), we anticipate that this information will be applicable and of use in all countries. We will summarize statistical considerations that are pertinent to biomarkers in OA trials. Finally, we propose a research agenda that emerges from this update in order to assist in advancing the field. We note that the research agenda from such endeavors can be very valuable as evinced by the previous OARSI/FDA white paper on the subject[19] that led to the current OARSI/Foundation for NIH OA Biomarkers Consortium study comparing a large panel of biochemical and imaging biomarkers for their predictive capabilities[3].

The working group for this review included all individuals with extensive OA-related biomarkers expertise; the group consisted of two rheumatology physician scientists (VBK, FB), one orthopaedic surgeon physician scientist (SL), one clinical epidemiologist (ME), one basic scientist with expertise in physical therapy and rehabilitation (YH), one biostatistician with expertise in outcomes research (EL), one basic scientist with expertise in musculoskeletal tissue analysis (PO), and one research industry scientist with expertise in translational medicine and regulatory affairs (SP).

Guidance on use of biomarkers in OA clinical trials

Approach to clinical trial use of OA-related biomarkers

In 1992, the FDA posted rules and regulations in the Federal Register[20] regarding approval of drugs for “serious or life-threatening illnesses” based on “evidence from adequate and well-controlled studies of the drug's effect on a surrogate endpoint that reasonably suggests clinical benefit or on evidence of the drug's effect on a clinical endpoint other than survival or irreversible morbidity”. This guidance was updated in 2014 and defined a serious disease or condition as follows: “a disease or condition associated with morbidity that has substantial impact on day-to-day functioning. Short-lived and self-limiting morbidity will usually not be sufficient, but the morbidity need not be irreversible if it is persistent or recurrent. Whether a disease or condition is serious is a matter of clinical judgment, based on its impact on such factors as survival, day-to-day functioning, or the likelihood that the disease, if left untreated, will progress from a less severe condition to a more serious one”[21]. Importantly, this definition encompasses many aspects of the OA experience. This guidance describes a mechanism for expedited approval of drugs that may provide meaningful therapeutic benefit compared to existing treatment. After approval under these regulations, continued study of the drugs' clinical benefits and restrictions on use might be mandated, pending completion of studies to establish and define the degree of clinical benefits to study participants. Thus, this mechanism could be considered a conditional approval mechanism. This mechanism of approval is important for our consideration as soluble biomarkers might play a role in the conditional approval process and further, that this mechanism might be the stepping stone to a new drug development pathway in OA. This pathway would allow approval of “a drug that treats a serious condition AND generally provides a meaningful advantage over available therapies AND demonstrates an effect on a surrogate endpoint that is reasonably likely to predict clinical benefit or on a clinical endpoint that can be measured earlier than irreversible morbidity or mortality (IMM) that is reasonably likely to predict an effect on IMM or other clinical benefit (i.e., an intermediate clinical endpoint)”[21]. To support this mechanism of approval, it will be important to further our understanding and recognition of the serious nature of OA, promote development and widespread use of uniform definitions of OA, and gain a greater appreciation of the burden of OA at an individual and societal level.

A recent report underscores the major global disability burden of OA, including low back pain[22-24]. However, the systematic analysis for the Global Burden of Disease Study 2010 split OA-related conditions into separate categories such as low back pain, neck pain and OA (type not specified). Research is needed to account for all forms of OA so as to gain a holistic assessment of its true burden; for instance, it will be important to understand the proportion of neck and low back pain that is attributable to OA. Currently, our inability to definitively identify and quantify OA in all its forms, hampers our ability to form a holistic assessment of its true global impact. This problem in the field is potentially accessible by systemic biomarkers. Murray et al acknowledge that these musculoskeletal conditions are extremely common, that the burden is likely to grow, the symptoms are underestimated and research is urgently needed to develop effective and affordable strategies for dealing with these disorders[22]. These conclusions would seem to be consonant with the seriousness criterion needed for pursuing a conditional drug approval on the basis of surrogates.

Types of biomarkers

The BIPEDS classification system (standing for Burden of disease, Investigational, Prognostic, Efficacy of intervention, Diagnostic and Safety) describes one broad level of potential classification and contexts for qualification[19, 25]. The correspondence of these terms with those of the FDA is shown in Table 1. Note that the BIPEDS general category of prognostic biomarkers includes both prognostic and predictive markers per FDA definitions and the efficacy of intervention category includes pharmacodynamic (or activity) biomarkers per the FDA nomenclature in guidance documents. The three biomarker categories of prognostic, efficacy of intervention and safety are the highest development priorities for OA clinical trials; this is due to the fact that at present, the diagnosis of OA for meeting enrollment criteria is based on a combination of symptoms and established anatomic abnormalities identified by an imaging technique such as MRI or radiographic abnormalities. It is conceivable that in the future, OA trials will be conducted on individuals at earlier stages of disease development that might require diagnostic biomarkers to identify the appropriate patient population and verify a diagnosis of early OA that would not yet be possible by an existing imaging technique. A recent exciting event in the field has been the discovery of differences, at a molecular level, between the cartilages at different sites in the body[26]. Coupled with the recent demonstration that cartilages differ in their anabolic responses by joint site and disease state[27], joint type specific biomarkers would appear to be in reach for OA.

Table 1.

Summary of “BIPEDS” biomarker classification for OA and comparison with FDA terms.

Category
B	Burden of Disease	Biomarker associated with extent of severity of OA
I	Investigative	Biomarker not yet meeting criteria for another category
P	Prognostic	Predicts incidence or progression of disease (FDA prognostic biomarker) or likelihood of response to a treatment intervention (FDA predictive biomarker)
E	Efficacy of Intervention	Indicative of treatment efficacy (FDA pharmacodynamic or activity biomarker) and for which the magnitude of the change is considered pertinent to the response. Surrogates form a subset category of biomarkers intended to substitute for a clinical efficacy endpoint
D	Diagnostic	Differentiates diseased from non-diseased
S	Safety	Identify adverse effects and provide means of safety surveillance

Based on Bauer et al[25] and[7].

Contexts of use and importance of patient phenotyping

Patient phenotyping is critical to the success of biomarker qualification. OA is a tremendously heterogeneous group of different phenotypes of disease at different joint locations and different combinations thereof. It is possible that biomarkers will perform very differently in these different phenotypes; for instance, biomarkers for early onset post-traumatic knee OA might be very different from valid biomarkers for erosive hand OA.

The subject sample needs not only to be carefully detailed with respect to conventional demographic characteristics, i.e., age, sex, body mass index, comorbidities, etc., but also the targeted OA phenotype, joint location(s) and its disease stage depending on study purpose such as diagnostic, efficacy monitoring, or safety monitoring. As demonstrated by the experience with prostate cancer biomarkers, there is a need for a standardized procedure to validate new biomarkers to achieve comparability[28]. In addition to comprehensive phenotyping, very precise conditions are also needed for sample collection, handling and storage, as summarized below in the section on analytical validation.

One current major challenge is our lack of universal criteria to phenotype and characterize different stages of early OA. To facilitate interpretation and comparability of biomarker studies across trials, study participants need to be described in detail with respect to symptoms, structure, function, and other known risk factors and medications. The white papers on clinical trial recommendations of the other working groups of this task force are providing information to ultimately create a standardized environment in which to qualify OA biomarkers.

Here we provide only a few examples of OA phenotypes from the overall spectrum of phenotypes that may be suitable for biomarker trials:

• (a)Previously uninjured men and women, aged 25 to 35 with acute anterior cruciate ligament injury (< 4 weeks old) with or without primary arthroscopic ligament reconstruction or meniscectomy. The status of other joints is normal.
• (b)Postmenopausal women with symptomatic and disabling (>6 months) bilateral multi-joint erosive hand OA as detected by MRI; other joints asymptomatic.
• (c)Individuals aged 40 or older with new onset unilateral knee pain (<6 months) without knee trauma, but with normal knee x-ray. Knee MRI shows meniscal pathology (extrusion/degenerative tear), subchondral bone marrow lesion, effusion and/or minor tibiofemoral cartilage defect(s).
• (d)Symptomatic and radiographic knee OA.
• (e)Symptomatic and radiographic hip OA.

Structure-based outcomes, for instance from MRI, ultrasound or scintigraphy to name a few, need to be carefully defined, preferably using validated scoring systems[29]. The current fast development of compositional MRI, which has the potential for earlier detection of progression, creates a promising future in which to develop outcomes and endpoints that may require much shorter follow up times than the classic loss of joint space on conventional radiographs, or loss of cartilage thickness on MRI. Patient phenotyping is critical to the success of biomarker qualification. It is generally believed that OA trials have failed to succeed to date due to the insensitivity of the clinical outcome measure, namely radiographic joint space narrowing, mandated by regulatory agencies. Examples from other fields illustrate the impact of the clinical efficacy measure. For instance, the validity of blood pressure measures as surrogates depended quite strongly on the definition of the clinical efficacy measures because blood pressure measures were very predictive of effects on stroke, less predictive of effects on myocardial infarction, cardiovascular death and overall mortality, and poorly predictive of effects on heart failure (summarized by Fleming et al[30]). Thus, when using a biomarker as a substitute for a clinically meaningful endpoint, one must first be clear about the clinically meaningful endpoint for which the biomarker is a proposed surrogate[30].

The FDA guidance states (as described in Table 2) that their qualification process is intended to provide some degree of generalizability for use of the tools—such as their use across multiple clinical disorders, drugs, or drug classes. We would therefore anticipate that drug development programs could utilize biomarkers qualified for particular contexts of use. For instance, a biomarker qualified for OA progression, and modified by interventions that block progression, might be used as a tool for development of a chondroprotective agent.

Table 2.

Rationale for new approach to development of Drug Development Tools (DDTs).

Points	Description
1	Once a DDT is qualified for a specific context of use, industry can use the DDT for the qualified purpose during drug development, and CDER reviewers can be confident in applying the DDT for the qualified use without the need to reconfirm the suitability of the DDT.
2	Because of the substantial work needed to achieve qualification, CDER encourages the formation of collaborative groups to undertake these tool-development programs to increase the efficiency of joint efforts and to lessen the resource burden upon any individual person or company working to gain qualification for a tool.
3	DDT acceptance in the drug development and regulation process has previously been on a sponsor-by-sponsor, drug-by-drug basis.
4	If a DDT is qualified under this guidance, the qualified DDT will be made publicly available for use by sponsors of any drug or biologic investigational new drug (IND) or new drug application (NDA) or biologics license application (BLA).
5	The new guidance is intended to provide some degree of generalizability for use of the tool such as use across multiple clinical disorders, multiple drugs, or drug classes.

Based on two FDA guidance documents[7, 34]

Guidance on biomarkers as drug development tools

Several converging observations have led to the increasing emphasis on biomarkers in drug development and a process for their qualification. First, it is clear that not all individuals respond favorably to specific therapeutic interventions including drugs. The use of biomarkers can help identify individuals that are more likely to respond favorably to a given therapy. Increasingly, biomarkers are applied to the stratification of different patient groups in terms of clinical response, so as to develop personalized, preventive or therapeutic strategies[31]. The FDA has recognized the importance of biomarkers in personalized medicine[32]. Starting in 1998 with the approval of trastuzumab (Herceptin, Genentech) for the treatment of HER2-positive breast cancers, the FDA has approved more than 100 drugs that contain specific information about biomarkers in the labeling[33].

A second impetus to biomarker guidance originated with the Omnibus Appropriations Act of 2009 that included funds for an Institute of Medicine (IOM). Compelled by applications to the Center for Food Safety and Applied Nutrition of the FDA for approval of health claims for foods, most of which reflected claims of effects on a biomarker—the IOM was tasked to recommend a framework for the evaluation of biomarkers and to make ancillary recommendations for their application[18].

The FDA recognized, in the Critical Path Initiative launched in 2004, that the advancement in knowledge about the biology of diseases had not resulted in the commensurate increase in effective new drugs (http://www.fda.gov/scienceresearch/specialtopics/criticalpathinitiative/default.htm). To improve the drug development process, the FDA recognized that the application of those scientific advances to drug development should make this process more efficient. Such tools should ultimately produce products with greater safety and efficacy and most importantly, reduce the attrition rate, i.e. drugs not reaching a marketing authorization during advanced clinical development. This engendered a third type of guidance, a rationale for a new approach to development of Drug Development Tools (DDTs) that was provided by the FDA in 2010 (summarized in Table 2)[34]. In 2014 the FDA released a second guidance further describing the qualification process for DDTs[7]. The guidance provides a framework for interaction between the Center for Drug Evaluation and Research (CDER) and the sponsor (pharmaceutical firms, academic institutions, consortia, etc.) proposing the DDT for qualification.

A DDT could be a biomarker, a clinical outcome assessment, or an animal model. Biomarker qualification has been recognized by the FDA as a significant area of interest, either as a single biomarker or as a composite biomarker, the latter consisting of several individual biomarkers combined in a stated algorithm to obtain an easily interpretable readout. The guidance also contains an indication on how a sponsor should formulate the so-called “Context Of Use” (COU). These include a “use statement” (the name of the biomarker and the specific purpose for use in drug development) and a description of conditions for the biomarker to be used in the qualified setting that are termed “condition for qualified use” (the conditions for the use of the biomarker in the qualified setting). In summary, the process of biomarker qualification is made up of several steps:

• (1)Identification of the biomarker as soluble/'wet' (biochemical analyte, genomic, etc.) or non-soluble/'dry' (imaging physical examination finding, scales, etc.);
• (2)Aspect of the biomarker measured and indication for correct interpretation (i.e., single time point vs. area under the curve (AUC) of a receiver operator characteristic curve, post-treatment vs. pre-treatment);
• (3)Animal species or subjects studied (strain, age, sex, ethnicity, disease phenotype, etc.);
• (4)Purpose in drug development (patient phenotyping, assessment of efficacy and/or safety, etc.);
• (5)Drug development circumstances for applying the biomarker (nonclinical vs. clinical);
• (6)Interpretation and decision/action based on the biomarker (threshold that indicates beneficial physiological response, threshold that indicates organ toxicity, individuals that are more likely to respond to treatment, individuals that are more likely to experience adverse effects).

Finally, the steps described above should be integrated in a decision tree that clarifies, in the context of use, the actions that would be taken based on the biomarker results. The steps involved in qualification are listed in Figure 1 and based on the most recent FDA guidance[7]. With the advent of a clear pathway to biomarker qualification, work is ongoing to establish DDTs for OA-related trials and ultimately clinical use.

Figure 1. The qualification process for Drug Development Tools (DDTs).

This schematic is based on a 2014 FDA guidance document related to the biomarker qualification process[7]. Since 2008, 4 submitters have been granted qualified biomarker status from the FDA for biomarkers related to nephrotoxicity, cardiotoxicity and lung infection with aspergillosis. Draft guidance and supporting information on these qualified biomarkers can be found listed at the FDA website on the Biomarker Qualification Program (http://www.fda.gov/Drugs/DevelopmentApprovalProcess/DrugDevelopmentToolsQualificationProgram/ucm284076.htm). CDER=the Center for Drug Evaluation and Research; DDT=Drug Development Tool; QRT=Qualification Review Team; FDA=United States Food and Drug Administration

Although this chapter refers primarily to the FDA, the EMA also offers scientific advice related to novel methodologies intended for use as tools in clinical drug development (http://www.ema.europa.eu/docs/en_GB/document_library/Regulatory_and_procedural_guideline/2009/10/WC500004201.pdf). This guidance was approved by the Committee for Medicinal Products for Human Use (CHMP) in 2009 and has been revised twice. It now includes dates and deadlines for submission of letters of intent for qualification of novel methodologies and addresses the following procedures, which are quite similar to the process adopted currently also by the FDA (from http://www.gmp-compliance.org/enews_04089_New-EMA-Guidance-Qualification-of-novel-Methodologies-for-Drug-Development.html):

• Intention to submit a request

• Appointment of the Coordinator and the Qualification team on behalf of the Committee for Medicinal Products for Human Use CHMP)

• Preparatory meeting

• Evaluation of data and discussion with the applicants

• Scientific Advice Working Party (SAWP) review

• CHMP adoption of Qualification Advice and discussion of Qualification Opinion

• Public consultation (for Qualification Opinion only)

• Adoption of the final CHMP Qualification Opinion

An overview of the comments received on the EMA draft guidance document on qualification of biomarkers can be found at the following url: www.ema.europa.eu/docs/en_GB/document_library/Regulatory_and_procedural_guideline/2009/10/WC500004202.pdf.

Utilization of Biomarkers as DDTs for OA Clinical Trials

Biomarkers could provide value and be involved in OA-related drug development at preclinical stages and throughout clinical development in phase I to IV trials. Below we summarize considerations for each of these phases of drug development and trial work.

• Preclinical Work

• – Assess drug safety

• – Assist with selection of animal models and lead compounds

• – Assess drug mechanism of action

The majority of biomarker use in drug development is in the early phases involving the internal decision making process related to determining which drugs to progress[35]. At the preclinical level, three major applications of biomarkers are their use in toxicology studies [36], their use to help select appropriate animal models and lead compounds[4, 37] and their use to determine or refine the mechanism of action of a compound[4]. In toxicology studies, the goal is to use safety biomarkers in assisting in the selection of drug candidates that are more likely to be tolerated in humans, thereby reducing cost and time required for preclinical safety evaluation[36]. Potential clinical biomarkers of drug efficacy are often identified through pre-clinical studies or basic research[37]. Biomarkers verified for use in clinical trials can confirm a drug's pharmacological or biological mechanism of action, guide protocol design, aid patient and dose selection, and help to minimize safety risks[37]. Floyd and McShane separate biomarkers into three tiered sets that predict compound efficacy with increasing confidence: level 1 to confirm pharmacological mechanism of action; level 2 to demonstrate biological mechanism of action; and level 3 to predict clinical outcome[37]. All three of these levels represent different types of efficacy of intervention biomarkers per the BIPEDS classification. They can be employed both in preclinical studies and clinical trials in humans. As they note for the oncology field, “Most companies now have requirements for the incorporation of biomarker strategies into clinical development plans, and success of this approach to drug development will require the consistent and timely delivery of new biomarkers having a level of robustness and validation that matches or exceeds the significance of the decisions they will be designed to support”[37].

• Phase I Trials

• – Assess mode of action

• – Assist with dose finding and selection

• – Assess safety via surveillance of effects on joint metabolism

Phase I studies during drug development are the first studies involving humans and provide answers to important questions about a potential new drug[38]. The main objective of Phase I studies is to assess the safety of the product after single and repeated doses. Secondary objectives are to evaluate the pharmacokinetics (what the body does to the drug) and, if possible, the pharmacodynamics (what the drug does to the body). The latter two issues are relevant as they provide the first preliminary indication in humans of whether a new molecule has the potential to be effective based on ease of use (dose and frequency of dosing) and potency (based on preclinical in vitro and in vivo studies). The pharmacokinetic (PK) profile determined in Phase I studies provides an understanding of whether the plasma concentrations are adequate to engage the pharmacological target with at most two or three administrations per day of the drug while simultaneously maintaining a suitable safety margin compared to the exposure observed in toxicological studies. Therefore, in order to be a pharmacodynamic (PD) biomarker in Phase I, the biomarker must be linked to the mechanism of action of the new chemical entity, demonstrate whether the administered compound reaches the target receptor and does so in sufficiently high concentrations, and that this occurs at systemic exposure unlikely to generate adverse events based on the product toxicological profile. The biomarker is subsequently used in phase I studies to build a dose response curve by linking the plasma concentration observed at different doses with the concentrations/response curves obtained in vitro on the target receptor[39, 40]. When the above-mentioned link is made, this is referred to as Mechanism-based PK-PD modeling.

These processes differ from simple PK-PD models in that they include the quantitative evaluation of additional steps from drug administration to effect. This includes target site distribution, binding and activation, pharmacodynamic interactions, transduction and homeostatic feedback mechanisms. Important progress has been made in the field of mechanism-based PK-PD modeling through the incorporation of concepts from receptor theory[39, 40]. Receptor theory is the application of receptor models to explain drug behavior. For instance, the inclusion of receptor theory concepts for characterization of target binding and target activation processes results in improved prediction of the pharmacological active dose and/or doses. Specifically, receptor theory constitutes a scientific basis for the prediction of in vivo drug concentration-effect relationships. To this end, biomarkers need to be used early during the drug development process and information needs to be gathered and continuously updated to ensure the reliability of the results.

Phase I studies are usually conducted in healthy volunteers and sometimes in individuals with the disease. Regardless of the population enrolled, since OA is a chronic disease, it is obvious that the availability of biomarkers in Phase I for OA drugs will accelerate the understanding of the potential efficacy of a new chemical entity as modulation of the biomarker, linked to the drug mechanism of action, would be expected to occur within a significantly shorter period than that needed to reach the clinical endpoint[41]. OA drug development is one of the therapeutic indications that could potentially stand to benefit most from the use of biomarkers during early clinical development. An early indication of efficacy would reduce the risk of failure during long and expensive Phase III clinical trials and therefore, make the drug development process more efficient, shortening time to market, and allowing more accurate decisions regarding investments.

A biomarker in Phase I can be used not only to assess the expected therapeutic effects but equally important, to monitor for unwanted side effects. Ideally, drug development should benefit from biomarkers capable of monitoring the efficacy (linked to the mechanism of action of the drug) and the safety (allowing for instance, the monitoring of the function of an organ known to be a target for toxicity from toxicological investigations) of an experimental drug. These would allow an understanding of whether a favorable therapeutic index can be expected in treated individuals.

• Phase II Trials

• – Serve as an early objective indicator of drug effect

• – Assist in identifying the minimal effective dose and dose response profile

• – Facilitate Multiple Comparison Procedure Modeling (MCP-Mod)—use of modeling and simulation during Phase II to aid dose selection for Phase III clinical trials in OA

Typically, biomarkers play a less-exploratory role but are rather used as key decision points in phase IIb and phase III stage development [35]. Phase II studies are the most critical trials in drug development as they generate the first proof of concept regarding potential clinical efficacy. Careful study design and conduct is of paramount importance to reduce the risk of failure during later Phase III trials. One of the most important aspects that must be taken into consideration is the construction of an accurate dose response relationship for both safety and efficacy to ensure the most appropriate dose range and/or ranges are taken further and tested in subsequent Phase III trials. Recent advances in the design of Phase II studies include application of computer tools to model and simulate clinical trials[42]. Modeling and simulation applied in Phase II are based on the availability of a biomarker linked either to the mechanism of action of the drug, or to a potential unwanted effect, or both. Modeling and simulation allow an appropriate dose selection for Phase III studies to ensure optimal efficacy and safety in real clinical practice, thus moving from “efficacy” (the drug is capable of affecting the disease phenotype in a controlled trial) demonstrated in Phase II and III to “effectiveness” (the drug is capable of producing clinically relevant benefits in the “real” practice of medicine). To streamline the drug development process and reduce the risk of failure, major pharmaceutical companies are developing modeling and simulation tools alone or in conjunction with other companies. One example is the Multiple Comparison Process-Modeling (MCP-Mod) developed by Novartis[43] that was also recently qualified by requesting the EMA opinion on the topic[44].

US and European Regulatory Agencies encourage pharmaceutical companies to increase the use of modeling and simulation tools and to this end have even created dedicated divisions specialized in the relatively new science called pharmacometrics (http://www.fda.gov/AboutFDA/CentersOffices/OfficeofMedicalProductsandTobacco/CDER/ucm 167032.htm). Pharmacometrics is becoming a fundamental part of drug development that quantifies drug, disease and trial information to aid not only efficient drug development, but also regulatory decisions[45]. Pharmacometrics adds the individual patient characteristics to better describe the relationship between exposure (pharmacokinetics) and response (pharmacodynamics) for both desired and undesired effects. Disease models describe the relationship between biomarkers and clinical outcomes, time course of disease and placebo effects. Modeling will thus facilitate the definition of the inclusion/exclusion criteria, study participant discontinuation and adherence based not only on the safety and efficacy of the drug, but also on the interaction of these with the individual's characteristics. This will ultimately allow identification of populations that are most likely to respond to treatment and/or those that are more likely to experience adverse events enabling enrichment study designs[46]. Therefore, pharmacometrics aids in the selection of the right drug dosing for all individuals thus greatly contributing to the determination of the value of a biomarker for both efficacy and safety across clinical trials for a given disease and/or for a drug class.

• Phase III Trials

• – Increase study power and reduce sample size through enrichment of an appropriate target population

• – Shorten duration of trial

• – Gain approval for different dose range not in original trial based on clinical trial simulations with biomarkers

As described in a recent FDA guidance document[47], enrichment study designs are increasingly viewed as providing potential benefit to the drug development process. We can therefore envision biomarkers playing a role in identifying specific types of individuals (progressors, drug responders, etc.) for trial inclusion. In our prior OARSI/FDA white paper on biomarkers[19], we considered biomarkers as likely serving as secondary rather than primary endpoints. We can now conceive however, that with the formal qualification of a panel of OA-related biomarkers[3], and the development of a conditional approval mechanism for OA-related therapeutics (as described above), that biomarkers could potentially form a component of a primary endpoint. As described above, were OA to be formally accepted as a serious disease for purposes of drug approval by the FDA[21], it might be possible for biomarkers that predict clinical benefit to be used for marketing approval of the drug, with subsequent postmarketing confirmatory trials to verify the clinical benefit.

• Phase IV Trials

• – Define cost-effective means of safety surveillance during the post-marketing phase of drug development

• – Identify subgroups of responder or non-responder individuals

• – Monitor drug effectiveness and safety in real life conditions

The need for post-market surveillance becomes increasingly important in this new era of personalized medicine as more and more products are approved on the basis of very small clinical trials[32]--smaller pre-market exposure requires greater post-market monitoring in the general population. In daily medical practice it will be important to determine long-term risks and long-term benefits of drugs; to this end, biomarker surveillance could provide a minimally invasive and relatively cost-effective method to monitor for both risks and benefits of a drug on joint metabolism and health.

Analytical Validation

Guidance on bioanalytical validation

Validation refers to the measurement performance characteristics of a biomarker[4]. Validation of a bioanalytical method is needed to demonstrate that it is reliable and reproducible for the intended quantitative measurement of the biomarker(s) in a given biological matrix (e.g., blood, plasma, serum, or urine). When changes are made to a previously validated method, additional validation may be needed[4]. The FDA has provided comprehensive guidance on analytical validation processes appropriate to biomarkers; this document covers accuracy, precision, selectivity, sensitivity, reproducibility and stability[48].

• Strict standard operating procedures (SOPs) must be implemented to control preanalytical variability and ensure a complete system of quality control and assurance. SOPs should cover all aspects of analysis, from the time the sample is collected and reaches the laboratory until the results of the analysis are reported. For details on methods of sample collection, type and timing of samples, see Appendices from prior publications[1, 19], and Rai et al[49].

• Key reagents, such as reference standards, antibodies, tracers, and matrices should be characterized appropriately and stored under defined conditions. When there are changes in any of these, additional optimization or validation may be needed (e.g., checking cross-reactivities of an antibody).

• Matrix effects should be addressed such as ion suppression, ion enhancement, or alterations of extraction efficiency. The FDA guidelines advise measuring matrix effects by comparing spiked minus unspiked donor samples in a number of donors (e.g., 10) and looking at the appropriateness of the increase relative to the original level of the spiked in sample.

• A calibration (standard) curve should be generated for each biomarker in the sample, using standards that can contain more than one analyte. The curve should be prepared in the same biological matrix as the samples in the intended study by spiking the matrix (e.g., serum, urine or synovial fluid) with known concentrations of the standards. Concentrations of standards should be chosen on the basis of the concentration range expected in the particular study. A calibration curve should consist of a blank sample (matrix sample processed without analyte or internal standard), a zero sample (matrix sample processed without analyte but with internal standard), and at least six non-zero samples (matrix samples processed with analyte and internal standard) covering the expected range, including the lower limit of quantification (LLOQ).

• Most calibration curves of Ligand-Binding Assays (LBAs) are inherently nonlinear and, in general, more concentration points may be recommended to define the fit over the standard curve range than for chromatographic assays.

• The analytical method should be shown to be selective for the biomarker, ensuring this selectivity at the LLOQ. Evidence should be provided that the substance quantified is the intended biomarker, thus blank samples should be tested for interferences from substances similar to the biomarker tested; cross-reactivity can also be evaluated for metabolites, concomitant medications and their significant metabolites, or endogenous compounds.

• Accuracy is assessed by measuring recovery of the spiked analyte tested in triplicate, in accordance with FDA guidance for industry on bioanalytical method validation[48]. When possible, LBAs (such as ELISA) should be compared with a validated reference method (such as LC-MS) using incurred samples (selected real study samples) and predetermined criteria to assess the accuracy of the LBA method.

• Precision is measured using a minimum of five determinations per concentration. The precision determined at each concentration level should not exceed 15% of the coefficient of variation (CV) except for the LLOQ, where it should not exceed 20% of the CV.

• Reproducibility of the method is assessed by replicate measurements using the assay, including quality controls and possibly incurred (selected real study samples) samples.

• The chemical stability of the biomarker in the matrix, under specific conditions for given time intervals, should be assessed in several ways. Pre-study stability evaluations should cover the expected sample handling and storage conditions during the conduct of the study, including, among others, conditions at the clinical site and any secondary sites, conditions during shipment and effects of freeze-thaw. Stability samples should be compared to freshly made calibrators and/or freshly made quality controls (QCs). Testing is recommended to include a minimum of three concentrations in the range of expected study sample concentrations. The mean value should be within 15% of the nominal value except at LLOQ, where it should not deviate by more than 20%[48].

• The extent of recovery of a biomarker and its reference standard from the sample should be consistent, precise, and reproducible.

• A full validation of the bioanalytical method is important for the analysis of a new biomarker entity. In cases where modifications are made on already validated methods, partial validations can be performed. These can range from an intra-assay accuracy and precision determination, to a nearly full validation.

Bioanalytical platforms

Bioanalytical procedures for biomarker evaluation generally belong to three different platforms: chromatographic methods, such as liquid chromatography (LC) or gas chromatography (GC), combined chromatography and mass spectrometry (MS) procedures, such as liquid chromatography–mass spectrometry (LC-MS or LC-MS-MS), or ligand-binding assays (LBAs), such as ELISA or radioimmunoassays (RIA). Most recently, approaches combining LBAs with MS have been also developed[50].

Among these different analytical platforms, ELISA technology provides an unsurpassed high sensitivity combined with high throughput, which makes this platform desirable for the final clinical assay. However, reliance on this technology has created a bottleneck during biomarker verification, since antibodies are often poorly validated and the development of novel ELISAs is costly and lengthy. This important limitation prevents ELISAs from being used to interrogate large numbers of candidate biomarkers (peptides, proteins or specific isoforms), which have been identified in MS-based discovery strategies. The advent of proteomics has raised the challenge of detecting and quantifying a larger number of biomarkers from samples in a systematic and efficient fashion. It is clear that for complex diseases like OA, a multi-analyte biomarker approach should be valuable not only for prognostic and diagnostic procedures, but also for understanding disease mechanisms. Following this reasoning, analysis of biomarker panels is highly desirable in clinical trials, and therefore, multiplexing capabilities are becoming of key importance for biomarker verification and validation. In this sense, alternative multiplexed antibody-based protein quantification methods, such as protein arrays[51] and bead-based immunoassays[52], offer improved potential although they still require costly and lengthy antibody development.

Biomarker identification and qualification in clinical proteomics need specific recommendations[53, 54]; technical specifications related to the procedure should be reported e.g., deviations of mass and other parameters (retention time, migration on gel, etc.). It is also necessary to report the observed deviation in identifying parameters and (relative) abundance when the same sample is analyzed repeatedly.

In comparison with antibody-based strategies, (LC-MS)-based procedures exhibit a supreme high specificity towards precise epitopes, peptides or protein isoforms; it is well-suited for multiplexing, but with the drawbacks of limited sensitivity (especially in complex matrices such as plasma or serum) and throughput. Therefore, novel approaches combining immunoenrichment (which reduces sample complexity and allows higher sensitivity) followed by high-throughput, highly-selective, targeted MS analysis, appear to be promising strategies for routinely screening multiple biomarkers in hundreds of samples as required for biomarker verification[55]. In our field of OA biomarkers, this strategy has been applied to great advantage in the measurement of the aggrecan-ARGS protein neo-epitope using specific antibodies for the affinity purification of protein fragments combined with detection using mass spectrometry [56].

Other methodological considerations

Some fundamental concerns can be raised during the development of diagnostic tests, which command additional validation steps to suitably assess the value of the bioanalytical method before clinical application. These include the following:

• (1)Bias from inappropriate subject selection. To discern “discriminatory” characteristics (e.g., OA vs controls) in a rigorous and reproducible fashion, it is imperative to collect well-annotated biosamples prospectively or retrospectively, and in compliance with a well-designed protocol that is powered to obtain statistically important outcomes. Since they can influence biomarker levels, data should be collected on other skeletal alterations, other conditions, and concomitant medications.
• (2)Inappropriate sample size. Clinical biomarker studies routinely suffer from a lack of proper experimental design, resulting in analysis of too few biospecimens to achieve adequate statistical power at each stage of a biomarker pipeline. In proteomics research, a consensus is being built to set statistical criteria for clinical relevance and an approach to calculate sample size[57]. Rigorous evaluation of sample size and statistical power should be conducted prior to study execution to ensure conclusive results and to minimize misattribution of lack of effect due to low statistical power.
• (3)Type and timing of samples. Biomarker levels may vary with gender, age, menopausal status, ethnicity and OA risk factors such as body mass index. Furthermore, serum and urinary levels of most proposed OA markers, such as Cartilage Oligomeric Matrix Protein (COMP), the C-propeptide of collagen type II (CPII) or the carboxy-terminal cross-linked telopeptide of type II collagen (CTX-II) can be modified by diurnal change, food intake or physical activity, as previously reviewed[19]. Regarding the choice of plasma or serum for analysis it is important to highlight that, while both types of samples have similar protein compositions at a global level, the expression, stability or recovery of individual proteins may vary greatly. While the HUman Proteome Organization (HUPO) recommends EDTA-plasma for proteomics studies[58], biomarker data generated in OA clinical trials to date have been most commonly reported in serum.
• (4)Bias from sample collection, processing and storage. As mentioned above, strict standard operating procedures (SOPs) for sample harvest, management and distribution must be implemented to control preanalytical variability. Reliable biosample banks are required to assess the ability of the markers to discriminate among groups of subjects in an unbiased manner.
• (5)Data over-fitting. This refers to the tendency of models trained on large numbers of variables measured on small numbers of samples to produce extraordinarily high sensitivity and specificity and then fail on independent validation sets. This problem has been exacerbated by “omics” technologies, which produce thousands of variables for each sample or subject. To solve this concern, cross-validation of the results across several multiparametric statistical models and independent validation sets are needed [59].

Statistical Considerations

Details of statistical analysis procedures and reporting of results can be obtained in the article concerning statistical considerations that accompanies this special issue. Generally, the level of sophistication in planning or reporting biomarker studies has been very low, and inconsistent. Rarely does reporting extend beyond a simple association of an outcome with the concentration of a biomarker whose source is ill defined. Here we provide some basic information that should be included in describing data analysis and reporting of results of biomarker analyses, and discuss some methodological issues pertinent to biomarker statistical analyses.

• Proper analytic techniques to address non-independence of data points

If study data are gathered across multiple data points, the analysis plan should be consistent with methods used for sample size and power analysis that is centered around primary outcomes. Appropriate methods should be used accounting for data non-independence (two knees of the same person), repeated measures over time. For example, regressions with variance adjustments using generalized estimating equations (GEE) should be considered if multiple joints of the same persons are considered in the analysis. Alternatively, analysis could use data from only one joint per person, selected either at random or by severity (e.g. worst joint, based on standardized assessment).

• Methods for imputing out of range values

The data should be checked for completeness to ensure discrimination between missing data and out of range values. Out of range values should be imputed in a standardized fashion that is consistent across studies. For example, consensus should be reached if the actual imputed values should be the lowest level of detection, the mid-value between 0 and the lowest level of detection or some other method such as interpolation from the standard curve extended below the lowest standard (method most appropriate for cases where the standard curve can be shown to be linear in this range and presumably superior to random imputation of these low values).

• Prognostic biomarkers can be used to enrich a trial for progressors or for predicting treatment response

There are two analytical approaches to this type of biomarker. First, it is possible to use the entire distribution of responses to the treatment and correlate with the baseline biomarker concentrations. Alternatively, it is possible to compare biomarkers by groups based on a responder criterion, such as the OMERACT-OARSI responder criteria for OA clinical trials[60], and evaluate the biomarkers in those that respond compared to non-responders. The type of analysis should correspond to the pre-defined analysis plan and be consistent with analysis performed for sample size estimation. A variety of responder definitions can be identified along the cumulative distribution of the response curve.

• Standardization of biomarkers

To ensure comparability across studies and across biomarkers it is useful to utilize standardized scores that could express clinically meaningful changes in units of standard deviation. Such an approach could standardize reporting and aggregation of data across studies.

Summary of Recommendations and Research Agenda

As stated by Fleming et al[30], a prognostic factor does not an effect modifier make and a correlate does not a surrogate make. A long-term advantage to biomarker development and use in the drug development process is the advancement of tools that could contribute to personalized medicine. Personalized medicine may be defined as “a medical model using molecular profiling technologies for tailoring the right therapeutic strategy for the right person at the right time, and determin[ing] the predisposition to disease at the population level and to deliver timely and stratified prevention” (Definition from European Commission Health Research Directorate[31]). These principles can be seen as forming the foundation for the research agenda to advance biomarkers for clinical trial applications. The suggestions put forward here should be regarded as strong recommendations rather than requirements. Only by including biomarkers in trials, both at the hypothesis generating and verification stages, will their full potential be realized in the drug development process.

• 1)
  We recommend collection of biospecimens in all OA clinical trials because,

  • As noted in the clinical trial guidelines of 1996[1] and prior OARSI FDA white paper on biomarkers[19], here we encourage ongoing collection and analysis of samples from clinical studies and trials. Evidence from randomized controlled trials are needed to determine whether biomarkers are useful in identifying those individuals most likely to receive clinically important benefits from an intervention; this will establish a rich resource of biospecimens in which to qualify biomarkers as efficacy of intervention markers in different treatment contexts.
  • We also encourage collection of biospecimens from case control and longitudinal studies. Although diagnostic markers are not of primary importance in the current structure of OA clinical trials, we anticipate that they will be increasingly important, as we evolve to treating earlier disease, at the metabolic derangement stage, prior to structural abnormalities by MRI or radiograph. The Standards for the Reporting of Diagnostic Accuracy (STARD)[61], are generally accepted guidelines in diagnostic studies.
• 2)An extensive overview of biomarker results from trials is needed to give reliable estimates of the net effects of the intervention on the clinically meaningful endpoint and on the biomarker.
• 3)Collect data on concomitant medications—little is known of their effects on biomarker clearance.
• 4)Gain greater recognition of OA as a serious disease to establish a basis for OA drug development pathways permitting the use of surrogate endpoints.
• 5)Reporting guidelines for OA biomarker publications need to be established, and be openly available at e.g., OARSI website. These guidelines should include both “generic” aspects for any biomarker reporting (which may be adopted from other biomarker areas), and “OA specific” aspects. The current state of OA biomarker literature is very heterogeneous with regard to quality and details of reporting, making it almost impossible to perform a systematic review / meta-analysis, in spite of hundreds of publications. Examples of generic information are antibody and assay specifics, sampling and sample storage specifics, subject demographics, and more. Examples of OA-specific information are index joint structure information by radiography or MRI or other, information on other joints, joint symptoms (by what patient reported outcome), comorbidities, medication, and more. Statistical analysis methodology should be carefully described. Manuscripts not adhering to these guidelines will be less likely to be accepted for publication. By analogy to the CONSORT checklist for randomized clinical trial manuscripts, a checklist for authors should be developed to specify where in the submitted manuscript the different information is available, and this should accompany manuscript submission to facilitate review. Ultimately, higher quality and transparency of biomarker reporting will support a better understanding of the role of biomarkers in OA. Developing reporting guidelines will require an effort separate from the present.