Hyatt Regency Hotel
Bethesda, Maryland

The Osteoarthritis Initiative (OAI) is a multicenter, longitudinal, prospective observational study of knee osteoarthritis (OA). The overall aim of the OAI is to develop a public-domain research resource to facilitate the scientific evaluation of biomarkers for osteoarthritis as potential surrogate endpoints for disease onset and progression.

This 4-year longitudinal study is following 4,800 participants with either established osteoarthritis of the knee (˜1,300) or significant risk factors for the development of osteoarthritis of the knee (˜3,500) to produce a public resource database of imaging and clinical data to help identify and characterize the disease from onset to joint replacement.

The National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) and the National Institute on Aging (NIA) lead this initiative and are joined by the National Center for Complementary and Alternative Medicine, the Office of Research on Women's Health, the National Institute of Dental and Craniofacial Research, the National Center on Minority Health and Health Disparities, and the National Institute of Biomedical Imaging and Bioengineering to provide the Federal support for this initiative. The Foundation for the NIH (FNIH) organized and oversees the private-sector partners that include GlaxoSmithKline, Pfizer, Inc.; Novartis Pharmaceuticals; and Merck & Co., Inc. The Food and Drug Administration (FDA) participates in this initiative in an advisory manner.

On June 2, 2008, the Foundation for the NIH convened a meeting in Bethesda, Maryland for the purpose of presenting to current and future sponsors and users of the OAI an overview of the initiative and related ancillary studies, and to explore possibilities for the extension of followup of the cohort for additional years. The meeting consisted of a series of presentations from OAI investigators, experts in clinical epidemiology, and outside users of the OAI database. The presentations were followed by discussion periods to assess the overall interest in extending and enhancing this public-private initiative. The presentations and discussions are summarized below.

Welcome and Introductions

The meeting participants were welcomed by Stephen I. Katz, M.D., Ph.D., Director, NIAMS; Julie Wolf-Rodda, MA, Director of Partnership Development, Foundation for the NIH; and Gayle Lester, Ph.D., Project Officer for the OAI, NIAMS. Dr. Richard Hodes, M.D., Director, NIA, was unable to attend due to a commitment to meet with Senator Barbara Mikulski. The speakers noted that the public-private structure of the OAI was a paradigm for the type of endeavor needed to tackle this broad-based and long-term project, particularly for a project involving significant amounts of imaging research. The concept for this project dates back to 1998. A remarkable feature of the OAI is the precompetitive nature of the project that involves the participation of pharmaceutical company scientists in an effort to gather the type of information that is needed at the academic and commercial levels for developing treatment and prevention modalities. The impact of the OAI is already being seen at scientific meetings in which there are abstracts presented using the OAI cohort data. The PROMISE initiative, which is part of the NIH Road Map, is also providing information related to the OAI efforts.

Drs. Katz and Lester expressed the appreciation to the FNIH and FNIH staff for their efforts in organizing the meeting and supporting the OAI.

Session I: Where Is the OAI Now?

OAI: Where Did We Start and Where Are We Now?

Michael Nevitt, Ph.D., University of California San Francisco, PI, OAI Coordinating Center

Dr. Nevitt described the primary objectives of the OAI as: to serve as a shared clinical research resource to describe the structural and biochemical changes of early and progressive knee OA so as to understand the natural history of OA; to identify factors that influence knee OA onset and progression; and to characterize imaging, biochemical, and genetic biomarkers that predict and track the course and outcome of disease with the goal of biomarker qualification. To achieve the goals of the OAI, a longitudinal cohort study has been established. This cohort will allow for a well-defined and characterized community sample that can be assessed longitudinally for imaging, molecular, and genetic risk markers and for symptoms, function, disability, and surgery. The OAI will allow multiple stages in the spectrum of knee OA to be studied (at-risk, early/preclinical OA, established disease, end-stage disease) and will permit evaluation of biomarker levels (and changes) as predictors and correlates of disease and patient outcomes. The study will follow three subcohorts: an incidence group, a progression group, and a normal control group.

In terms of achieving the OAI objective of a public data resource, there will be open access to the data, images, and biospecimens so as to speed the generation of new knowledge about OA, and to enlist the community of OA investigators worldwide in understanding OA�s natural history and biomarker characterization. The data resource includes a downloadable clinical data archive on the Web, archived images on demand, and archived biospecimens by application.

Dr. Nevitt presented a brief history and overview of the OAI. The OAI began in 2003 with the installation of 3-Tesla MRI instruments at the clinical sites. Dr. Nevitt reviewed the study design and presented data on progress in a number of areas. This included the schedule of clinic visits, baseline and annual knee imaging, other joint imaging, and characteristics of the OAI participants (inclusion and exclusion criteria, demographics, baseline symptoms, baseline OA knee status). He also presented data on retention rates, followup status, and completion rates for biomarker ascertainment. He noted that between March 2004 and May 2006, 27 percent of individuals screened were enrolled in the study. There were many people who were not eligible for the OAI because of MRI contraindications or because the cohort for which they were eligible was already completely enrolled.

Dr. Nevitt discussed the schedule and process for public data release. Data and images are available on the OAI Web site, https://oai.epi-ucsf.org/datarelease/. He described the process for centralized image assessment, which is defined as standardized measurement/interpretation of well-defined samples of images using validated methods. There are two types of sponsorship for this: the assessment sponsored by OAI for public release and that sponsored by user groups with the intent of eventual public release. The centralized processing allows for comparability across analyses of structural outcomes (i.e., progression, incidence). Thus far, it is limited to the first-available set of baseline to 12-month images of 160 participants from the progression subset. He indicated that future enhancements to data access will include: release of data from ongoing centralized image assessments, release of biospecimens for analysis (through application to the Biospecimen Review and Approval Committee (BRAC)), customized sets of images based on user-defined samples/clinical characteristics, and images posted for browsing and the possibility of sample browsing and limited downloading on the National Cancer Imaging Archive (NCIA) Web site. Dr. Nevitt noted that abstracts and articles using the OAI resource are appearing at meetings and in journals, even at this early stage of data availability. He noted that the OAI resource is structured to allow for the development of nested case-control studies of biomarkers.

How Can Investigators Use the OAI Resource?

Charles McCulloch, Ph.D., University of California San Francisco, Biostatistics

Dr. McCulloch provided an overview of the various ways in which investigators can use the OAI resource. The Web site was developed by the data team at the OAI Coordinating Center, University of California, San Francisco. The OAI data and documentation are available to anyone simply by registering and agreeing to the terms of the data-use agreements. Thus far, there have been over 900 registrants.

There is a regular release schedule of data. Baseline, 12-month, 18-month, and 24-month clinical data and images have already been released for the first half of the cohort. Clinical data can be downloaded from the OAI Online Web site. Image datasets are large and are sent on portable hard drives.

Dr. McCulloch provided screen shots to illustrate the Web site and the multiple ways that data can be accessed. The online resources include a tutorial, a search-browse tool, descriptive statistics, and a data explorer tool. To increase awareness of the OAI, workshops have been held in conjunction with scientific meetings (Osteoarthritis Research Society International [OARSI] 2007). Future workshops have been planned (American College of Rheumatology, October 2008). There is also a Publications Committee whose members can provide critiques of analysis plans or who can put investigators in contact with experienced coauthors for possible collaboration.

He noted that some of the OAI approaches can impact design and analysis plans. This includes the facts that many variables are clustered by knee and that many variables are collected longitudinally over time on a person and/or knee. Prospective longitudinal studies such as the OAI are especially powerful for elucidating the trajectory of change over time, establishing causal relationships, and exploring intra-individual change. The consequences of the OAI study design are that failure to properly analyze clustered or longitudinal data can cause p-values and confidence intervals to be incorrect. Because expensive MRI measurements may not be available for all participants, case/control or case/cohort designs become more attractive. Dr. McCulloch provided examples of some of these issues and of approaches that could be used to analyze change in longitudinal data.

Dr. McCulloch concluded by noting that the size and prospective, longitudinal nature of the OAI study are particularly advantageous for development of research questions.

Discussion

Following the first two presentations, there was a discussion period in which several issues were addressed. It was clarified that there will not be any attempt by the OAI to filter analyses to avoid having multiple individuals address the same question. The OAI will provide a listing of ongoing projects and publications which may be helpful in this regard. Use of the Publication Committee by prospective investigators might be another way to avoid duplication. It was noted that some scientific duplication is appropriate in order to verify findings.

Since the analysis of MR data is not standardized, there was a query as to whether there is a way to compare how different investigators analyze the data. The OAI will provide centralized readings using established methodologies. Other investigators can cross-validate new approaches with these readings. It is hoped that such users will make their data available to the research community.

The availability of the full set of data over time will allow investigators to address the differences between the demographics of subjects in the first and second batch of released OAI data.

Role of Biomarker Initiatives in FDA Critical Path

Janet Woodcock, M.D., Director, CDER, FDA

Dr. Woodcock noted that one of the objectives of the OAI is to derive information that would spur treatment and prevention strategies. The availability of biomarkers is key to this goal as biomarkers: 1) will allow one to distinguish disease subgroups (e.g., rapid progression, anatomical differences, etc.); 2) will provide quantitative disease assessments for pharmacodynamic measures, including early response to therapy; and 3) will allow one to begin to evaluate surrogate endpoints for efficacy.

Dr. Woodcock defined biomarkers, validated biomarkers, clinical endpoints, and surrogate endpoints and how they relate to one another and to the drug development pathway. Biomarker qualification means developing the correlative information that allows one to understand the biomarker�s clinical meaning; this concept is often referred to as "fitness for use." Biomarkers may have broad or restricted use. Dr. Woodcock summarized the usefulness of biomarkers in early product development and decision making as well as their use in later aspects of drug development. She indicated the current problems/barriers in drug development, which include the enormous costs (which encompass the cost of failed products) and the long development time needed to bring a product to the clinic. Even after a product is on the market, there is still a need for safety and efficacy monitoring.

Biomarkers are key to bringing new products to clinical usage. The current practice of product development is empirical and population-based, as no other methods are available. Biomarkers would allow for: studies of drug metabolism, definition of disease subsets, estimation of response probability ("targeted therapy"), monitoring of responses, development of prevention products, assessment of medical product safety, and development of quantitative disease and intervention models that incorporate biomarkers. Once well-accepted biomarkers are available, clinical trials can be shortened.

There is a great interest in prevention of disease, but there has been little progress in this realm outside of vaccinations and cardiovascular disease. An important barrier to prevention research is related to the benefit/risk assessment at the levels of trials and clinical practice. There are ethical issues in asking people to take medications of unproven efficacy and unknown toxicity for long periods of time. However, risk stratification via biomarkers may allow one to undertake such studies in populations who can be identified as high risk for disease or disease progression.

Dr. Woodcock noted that practical business and conceptual models for biomarker development are lacking, and that there is little academic reward in this kind of research. In 2004, the FDA and European Medicine Agency (EMEA) issued the Critical Path Initiative, which calls for and encourages this type of translational science.

Dr. Woodcock closed by indicating that better understanding of disease natural history and development of qualified biomarkers will enable more rapid drug development in a given field. The FDA and EMEA are setting up procedures to formally evaluate biomarker status. Public-private partnerships like the OAI are an ideal venue for generating the needed scientific data.

Discussion

There was discussion about the use of imaging as a biomarker. The need for standardization to compare platforms was noted. Other considerations for using imaging as a biomarker are the availability of the method/technology and how the imaging measurement is related to the endpoint of interest. Dr. Woodcock noted that if an individual investigator has a specific proposed biomarker approach, the FDA and EMEA would be willing to review it and give feedback.

There was also mention of the potential for biochemical biomarkers. Developing such biochemical biomarkers would be a major scientific effort. The Europeans have launched a € 5 billion medical initiative for such a purpose.

Pharmaceutical Industry Perspective on Biomarker Initiatives

Monica Luchi, M.D., Novartis Pharmaceuticals

Dr. Luchi addressed biomarkers from the perspective of the pharmaceutical industry. She provided examples of biomarkers that may predict disease susceptibility. These include screening markers for early detection, e.g., PSA for prostate cancer; prognostic markers, e.g., estrogen receptor and breast cancer; and activity markers, e.g., C reactive protein and rheumatoid arthritis. Biomarkers for drug activity include markers of safety, e.g. liver functions for hepatotoxicity or proteinuria or changes in creatinine for renal toxicity; pharmacodynamic markers, e.g., GLP-1 and DPP4 inhibition; markers to measure efficacy, e.g., alkaline phosphatase for Paget�s disease of bone; and markers of long-term outcomes, e.g., x-rays in osteoarthritis.

One advantage of having biomarkers is that they accelerate drug development by facilitating evidence-based decision making. Biomarkers may shorten the length of trials; they may result in early ending of development for less effective products and thus, conserve resources for better candidates; they may identify patients more likely to respond; they may reduce trial size and number (by decreasing use of patients, thus decreasing costs); they may reduce safety risk (by predicting at-risk populations and by providing early flags, thus averting events); and they may improve sensitivity when monitoring for potential adverse events.

There is increasing interest in biomarkers outside of industry, including the FDA�s plans for development of process maps for biomarker evaluation and their issuance of a drug/diagnostic codevelopment concept paper. Public-private partnerships have been formed that focus on biomarkers. These include the Predictive Safety Testing Consortium, the FNIH/NIH/FDA/PhRMA Biomarkers Consortium, the Innovative Medicine Consortium (IMI) in Europe and the Serious Adverse Events (SAE) Consortium (which is an alliance of manufacturers with the goal of identifying genetic markers that put patients at more risk for adverse events). A specific example of a biomarker development effort is a 2-year Cooperative Research and Development Agreement (CRADA) with Novartis under the FDA�s Critical Path Initiative. The purpose of the CRADA is to define and test a process for qualifying preclinical safety biomarkers, particularly for detecting renal injury. The intent is to enhance industry-wide understanding of this issue.

Dr. Luchi indicated that biomarkers will be easier to define when there is clear understanding of disease pathology, of the population at risk, and the identification of objective means for both diagnosis and evaluation of response to treatment. The OAI can provide key information in this endeavor, as there are not animal models of OA that are entirely predictive of response to treatment in humans and no data exist that directly link structural outcomes with clinical and/or biochemical markers. One should also consider that the "full validation" of biomarkers can take extended periods of time. It is likely that rolling qualification of biomarkers will take place with data-driven "fit-for-use" claims in a small specified group and with the claims expanded by additional results over time. It is possible that composite outcome measures will be needed.

Dr. Luchi concluded by noting the likelihood that a large number of ancillary studies will be undertaken by private and industry researchers that will capitalize on the OAI and, along with other biomarker initiatives, will be able to better identify potential and safe disease-modifying OA drugs.

Discussion

Dr. Katz indicated the importance of the pharmaceutical industry collaborating with the OAI at the precompetitive intellectual level. The NIH Biomarkers Consortium as well as Alzheimer�s disease researchers have since followed this collaborative pathway.

Session II — View From Within the OAI
OAI Clinical Centers

Where Is OA Pain Located in the Knee? Methods of Assessment and Patterns

C. Kent Kwoh, M.D., PI, University of Pittsburgh Clinical Center

Dr. Kwoh noted that knee pain is the most common presenting symptom in knee OA. There are multiple methods of assessing knee pain, but the source of knee pain is not well-defined and there is not concordance between findings and report of knee pain. He summarized the baseline data on the percentage of subjects in the incidence and progression groups who self-assessed the frequency of the knee pain in three categories. The objectives of his study were to determine the frequency and patterns of knee pain location and to determine the association between knee pain frequency and severity with specific knee pain patterns: localized (eight sites), regional (three sites), diffuse. Additionally, WOMAC pain, stiffness, disability, and overall scores were measured. There were 1,177 participants from the Pittsburgh OAI site for which the 24-month data were analyzed.

The finding of the study was that distribution of knee pain patterns can be reliably assessed using the Knee Pain Map. Medial joint line, lateral joint line and patella were the most common sites of localized pain. Both knee-specific and host factors are important in understanding knee pain patterns. The diffuse pain pattern is associated with more frequent and severe pain, and with increased disability compared to localized or regional pain.

Discussion

There was a question about the meaning of diffuse pain as a psychosocial or long-term issue. It was noted that there will be a measure of coping skills added at the 48 month visit of the OAI, so that this issue could be addressed when those data are available. It is possible that diffuse pain is more severe but that it perhaps starts locally. Dr. Kwoh has not yet had a chance to examine the effect of previous injury, but may wait until there is a larger dataset to do that.

Effect of Weight Change on Knee Pain in OAI Participants

Marc Hochberg, M.D., MPH, PI, University of Maryland Clinical Center

Dr. Hochberg indicated that overweight is the strongest modifiable risk factor for the development of symptomatic radiographic knee OA. Both the European League Against Rheumatism and OARSI recommend weight reduction for patients with knee OA. However, a systematic review of four randomized controlled trials with data on 454 patients with knee OA was not strongly supportive of this recommendation.

Using the OAI data, Dr. Hochberg undertook a study that had as its objective examining the association between weight change and change in knee pain and function in adults with symptomatic radiographic knee OA. The sample included 711 persons enrolled in the progression subcohort with data from both baseline and 12-month followup visits for age, weight, height, WOMAC Index, KOOS, timed 20-meter walk, and use of medications for pain.

The results were that there were no significant correlations between weight change and change in pain and/or function as measured by WOMAC and KOOS in either crude or multiple variable-adjusted models. Weight change was significantly inversely correlated with change in walking speed in both crude and multiple variable-adjusted models (P < 0.01). There were no significant differences by tertile of weight change and change in pain and/or function as measured by WOMAC and KOOS in either crude or multiple variable-adjusted models. Change in walking speed increased across tertiles from weight gain to weight loss.

Thus, the 12-month followup data from this observational study did not demonstrate significant correlations between weight change and changes in pain and self-reported function in persons with symptomatic radiographic knee OA. Similarly, there were no significant differences in changes in pain and self-reported function in persons who lost weight compared to those who had stable weight. Future analysis will include examining 12-month followup data on the entire cohort and 24-month followup data on these 711 subjects and examining the correlation between weight change and change in knee structure.

Discussion

There was discussion as to whether weight at baseline was a correlate of pain severity and whether the presence or absence of malalignment may be a factor. There was a suggestion to look at strata among those who lose weight. It was also noted that weight gain could be a function of strengthening. There is also the potential to examine the level of physical activity after weight loss and whether that might be related to more or less pain.

Knee Pain, Physical Activity, and Function in OAI Participants

Rebecca Jackson, M.D., PI, Ohio State University Clinical Center

The goal of this project is to better understand the variation in knee pain and its impact on physical activity. Pain in knee OA is known to vary in frequency and intensity over time. Between any two assessments, data suggest that many subjects show worsening (or improvement). Variation in knee OA pain over time has been found to be related in a time-dependent way to variation in certain structural features of knee OA.

The hypothesis of this study is that fluctuations in knee pain intensity and persistence are valid indicators of clinical status. The study was undertaken using the data from the first half of the cohort as a self-matched case control. The predictor variable was Knee Pain Rating (0-10 scale, past 30 days) and the outcome variables were the WOMAC Functional Limitation Scale and two performance measurements (20-meter walk speed [m/sec] and chair stand [stands/sec]). One conclusion of the study was that variation in pain severity between visits within a knee is strongly associated with change between visits in self-reported functional limitation in the same knee. When pain improves, self-reported function improves and vice versa. Variation in pain severity between visits within a knee is also related--but not as strongly--to change in objective physical function (walk speed and chair stand). This contrast may be due to discrepancies between self-perceived and objective performance or limitations in the analysis. Factors that influence knee pain fluctuation over time will also affect knee-specific functional limitation. Factors other than pain fluctuation appear to play an important role in modifying changes in person-oriented measures of objective physical function.

Future work will include the ability to describe variability in pain frequency and intensity across 4 years of followup, to further categorize the interaction of pain variability and function, and to categorize the interaction of pain variability and subsequent disability.

Discussion

During the discussion, it was suggested that other factors affecting pain could be studied, including the use of analgesics and the use of alternative and complementary medicine interventions.

OAI Ancillary Studies

Thrust and Meniscal Lesions in Knee OA in the OAI

Leena Sharma, M.D., Northwestern University Ancillary Study, PI

Dr. Sharma indicated that varus malalignment (bow-leggedness, as measured by x-ray in the standing position) increases risk of medial OA progression and increases the load distributed to the medial compartment. Varus thrust is a movement assessed during simple observation of gait. The overall goals of her ancillary study were to examine: the relationship of baseline varus thrust to worsening of medial OA, separately in knees without and with OA at baseline; the relationship of medial meniscal tear and subluxation and subsequent medial OA worsening; whether thrust leads to meniscal tear and subluxation; and whether thrust and meniscal parameters contribute to function decline.

She traced how initial observations in the orthopedic clinic led to a thematic emphasis on mechanical factors and eventually to a longitudinal natural history study at Northwestern University (Mechanical Factors in Arthritis of the Knee, MAK). The type of study that she envisioned to investigate varus thrust in OA was not feasible within the MAK, but was feasible using the OAI cohort.

She noted that the OAI approach has many advantages over the traditional research approach. These include: lower cost and higher efficiency; a faster pace of knowledge advancement (less time on the redundant work of funding/study performance and more time on hypothesis testing, reporting, generation of new ideas, scientific discussion); a large community of thinkers who can participate in investigations and ideas; public data access; data of very high quality from state-of-the-art studies; and opportunities for investigators newer to OA to interact with a group of preeminent investigators with a variety of special expertise, so they can more rapidly advance their knowledge of OA investigation and methods.

The OAI also provides important opportunities for hypothesis testing. There are an immense number of potential questions, dealing with, for example, outcomes, disease natural history, risk factors for pain/function/structure outcomes, and identifying targets for prevention and intervention, and for imaging markers and other biomarkers. Although certain questions require ancillary support, a vast number do not. The OAI also provides for investigator career development, e.g., through scientific presentations and papers, methodologic knowledge advancement, enrichment of one�s own "local" work, and preliminary studies to support new/novel questions. The OAI facilitates methodologic decision-making, e.g., the decision to use a given instrument for one�s research.

There are a number of important reasons to extend the OAI, including that it is unprecedented to have such a rigorously studied and well-characterized cohort followed for the duration that the 4-year extension will provide. OA in a human life spans decades, yet current knowledge is based upon a relatively short duration study consisting of 2-4 year "snapshots." Not continuing the OAI, after the outstanding achievement of the current 4-year cycle, would be a great loss to the research community and to persons with this prevalent and disabling condition.

The Role of Bone in OA Progression

Timothy McAlindon, M.D., M.P.H., Tufts University, Ancillary Study PI

Dr. McAlindon indicated that changes in bone were noted in the earliest description of OA. However, over the last few decades, the paradigm of OA has been as a cartilaginous disorder, and that has lead to a focus on measures that reflect cartilage loss, such as joint-space width. Although there is much evidence against this focus, perhaps the strongest is biomechanical studies that investigate which structures bear load in a joint. The findings are that articular cartilage bears only 1-3 percent of the load, whereas peri-articular bone bears 30-50 percent of the load. Another surprising finding is that the structure of the trabeculae in the periarticular bone suggests that force is dispersed in bone through a hydrodynamic mechanism rather than a mechanical mechanism. In OA, subchondral plate thickening occurs early, or even antedates, the development of cartilage loss to the extent that some have suggested that changes in bone are one of the initiating events in OA. Similarly, the increased bone turnover rate leads to reduced mineralization and material density factors that further compromise bone resilience.

Dr. McAlindon referred to findings he had made a number of years ago in one of the early studies of OA using MRI in a convenience sample. Among these findings were that bone marrow lesions are of clinical importance in terms of pain and their prediction of structural progression. There are several inferences that can be drawn from the changes observed: disruption of periarticular bone is a prominent characteristic of OA; the bone response in OA may determine whether the disease stabilizes or progresses; and bone processes need to be studied longitudinally to better understand the mechanisms of progression from the perspective of their being predictors, biomarkers, and targets.

In terms of synergy between his ancillary project and the OAI, Dr. McAlindon noted that the OAI has provided an unequalled opportunity to study the role of bone in a large sample whose comprehensive evaluations include annual 3-Tesla MRI. The OAI did not have a measure of bone processes; thus, these ancillary studies will bring such measures to the cohort. Within the OAI, there are huge inherent efficiencies, opportunities for collaboration, and the availability of intellectual resources. The results of the ancillary project will contribute to the public data set.

He noted that developing the ancillary project was a highly collaborative endeavor, including the development of the MR bone sequence (Dr. Erika Schneider), the testing of measurement of approaches to bone morphometry (Dr. Sharmila Majumdar), and choosing other bone measures as well as developing logistical plans.

The aims of the ancillary project are to determine if periarticular BMD (measured by DXA) and apparent bone volume fraction (measured by MRI) are predictors of the rate of knee OA progression (cartilage loss) and to test associations of systemic indicators of bone health on cartilage loss. These aims integrate well with the parent study and use the progression cohort. Dr. McAlindon provided an update on the status of the project thus far, and indicated that 456 of the 600 participants have already been recruited and enrolled.

Relation of Knee Alignment and Other Risk Factors to Knee OA

David Felson, M.D., M.P.H., Boston University, Ancillary Study PI

Dr. Felson noted that much of the work to establish the importance of frontal-plane alignment in knee OA was done by Dr. Sharma in the MAK study. With change in frontal-plane alignment, valgus, varus, or knock-kneed or bowlegged, there is marked change in the focal stress on cartilage. Dr. Felson noted that studies in rabbits using springs to create medial (varus) malalignment result in knee OA in the animals. Surgery in humans to correct varus malalignment (HTO) relieves knee pain/medial progression for a number of years.

Dr. Felson described the way to measure frontal-plane alignment using long-limb x-rays. He noted that the knee is the fulcrum of the largest lever arm of the body (extending from the hip to the ankle). The angle circumscribed by the tibial mechanical and femoral mechanical axis constitutes the frontal alignment which normally is 0° or 1° varus.

He noted that collaborative work with Dr. Sharma in the MAK study had defined the rate for knee OA progression (defined by x-ray) in subjects with valgus or varus malalignment. Two-hundred-forty subjects with knee OA were followed for 18 months. In those with varus malalignment, there was a 3.5-fold increased risk (p<.001) for medial progression; in those with valgus malalignment, there was a 3.8-fold increased risk (p<.001) for lateral progression. Severe varus and valgus knees had worsening disability. In the MOST study, Drs. Felson and Sharma found that varus and valgus knees have a much higher rate of incident knee pain than neutral knees. It appears that all of the adverse effects of knee OA may be induced by malalignment.

Dr. Felson indicated that one of the most important aspects of malalignment is that in any analysis of risk factors for OA, all other risk-factor effects on knee OA appear to be determined in part by the state of the malalignment. An example of this is that the effect of obesity on knee OA progression is dependent on mechanical alignment. This example relates to Dr. Hochberg�s talk, since in neutral knees, obesity has no effect on OA progression. In severely varus knees, obesity has no significant effect on OA progression. However, in moderately misaligned knees, obesity has a statistically significant effect on OA progression. Thus, the effect of obesity on progression is limited to moderately misaligned knees; obesity is not a factor in severely malaligned knees that inevitably progress or in neutral knees that rarely progress.

The major goal of the ancillary R01 is to determine if these findings could be corroborated and to determine if other OA risk factors differ by different alignment states. Dr. Felson noted that the importance of understanding malalignment relates to the choice of study populations for drug trials. Thus, if in severe malalignment a risk factor does not cause disease progression, then drugs targeting that risk factor will not prevent disease progression in that group of individuals; the malalignment is too severe and drives out all of the other factors such that the drugs won�t show an effect. Dr. Felson suggested that there might be a Goldilocks phenomenon for choosing groups of patients for OA drug testing. Thus, severely malaligned knees may be "too hot" (progression inevitable) for study; neutral knees may be "too cool" (progression too slow) and moderate malaligned knees may be "just right" for study.

Dr. Felson addressed the need for long-limbed films in the OAI. These films provide the most accurate and precise measurement of alignment. He also described the reasons for mechanical and anatomical alignment not being well-coordinated with one another. Dr. Felson noted that initially, the long-limbed films were only to be obtained in the progression cohort. With the funding of the R01, they will also be obtained in the incidence cohort.

The primary outcomes of the R01 are cartilage volume loss over 1 year in one knee in all of the OAI subjects in the incidence and progression cohorts and WOMAC pain change in these subjects.

Dr. Felson noted that a major strength of the OAI is the very large cohort and the potential to study subsets of individuals. This is very important, given that OA is not a homogenous disease, in part because different individuals have different knee mechanics. Subsets of individuals are likely to have different courses of disease, and this factor is very relevant to choice of subjects for OA drug trials.

Dr. Felson concluded that malalignment is a potential risk factor for OA progression and is likely to determine how drugs and other risk factors affect OA.

Discussion

Dr. Felson noted that the findings related to malalignment as a risk factor are stronger with long-limb films versus short-limb films, which support the need for their use. There was discussion about the frequency of use of long-limb films in some settings and whether two short-limb films could be combined to provide relevant information.

There was discussion about whether correcting alignment decreases the risk of progression. There is not a clear answer to this question.

Session III- View From Outside the OAI

Change in Cartilage Morphometry Over One Year: Studies in the Progression Cohort of the OAI

David Hunter, M.B.B.S., Ph.D., M.Sc., New England Baptist Hospital

The goal of Dr. Hunter�s studies was to assess the rate of natural disease progression of cartilage morphometry measures from baseline to 1 year in knees with OA from a subset of participants from the OAI Progression subcohort and to compare different methods to measure the changes. The study compared the cartilage morphology changes using the Virtual Scopics� method versus Chondrometrics� method and the responsiveness of x-ray versus MRI morphometry. Within the dataset, there was an attempt to analyze for predictors of progression. Dr. Hunter noted that much of this work has been done in collaboration with the pharmaceutical partners. Much of the image analysis work has been done in collaboration with Virtual Scopics.

The current gold standard for regulatory approval in drug studies is joint space narrowing on an x-ray, which provides a proxy of cartilage health by measuring joint space width and its change. A lot of effort has been expended on developing MR as a better way to measure cartilage morphometry and to provide a direct method for assessing cartilage health. However, MR hasn�t been exploited in datasets to assess its responsiveness over short periods of time in clinical trial applications. Although data from Australia and Canada suggested that MR could detect changes in the 7 percent range, recent studies have not been able to achieve that level of change.

The study sample (150) was derived from 160 subjects from the OAI Progression subcohort, all of whom had both frequent symptoms and, in the same knee, radiographic OA (ROA) based on a screening reading done at the OAI clinics. MRI postprocessing was accomplished using sagittal 3D DESSwe MR images from the baseline and 12 followup month visits. A segmentation algorithm was applied to the cartilage plates of the index knee to compute the cartilage volume, normalized cartilage volume (volume normalized to bone surface interface area), and percent denuded area (total cartilage bone interface area denuded of cartilage).

The analysis focused on the statistics of changes, both absolute and percentage from baseline to 1 year. The study conclusion was that annualized rates of change are small with the central medial femur showing the greatest consistent change. The rate of annualized change is about -2.5 percent. Some subjects seem to be losing cartilage volume and others seem to be "gaining" cartilage volume during the course of the 1-year interval. These types of findings have been replicated and published by other investigators using other image analysis tools and software algorithms. These results should stimulate discussion as to which MRI parameters should be measured in longitudinal studies of OA progression and how best to perform these measurements.

In order to compare two measurement methods, Dr. Hunter studied the rate of disease progression over 1 year using two pulse sequence and image analysis approaches in knees with OA from a subset of participants from the OAI. He measured the change in knee cartilage thickness (VCtAB / ThCtAB) over 1 year for sagittal DESSwe (Virtual Scopics) and coronal FLASHwe (Chondrometrics). He found that the magnitude of change is small and is not a function of the analytical methods.

Dr. Hunter also undertook collaborative studies to compare the results of MRI and x-ray findings in assessing minimal joint space, and found that both methods were about the same in terms of responsiveness.

In other studies with Dr. Grace Lo, he assessed several factors that might be predictive of progression, and found that the strongest predictors in the central medial femur in univariate analysis were the maximal bone marrow lesion, a large bone marrow lesion, the maximal median meniscus grade measured as a change in denuded area between baseline and 1 year. The pain and alignment categories did not appear to be important predictors.

Dr. Hunter noted that another way to look at the data to increase responsiveness is to focus on regions where there is a pre-existing focal loss of cartilage to see if denuded areas beget denuded areas. He stratified the subjects by levels of denuded surface area. This somewhat increased the standardized response means. These findings suggest that if one focuses on those people who have more cartilage thickness loss, it may be possible to differentiate a group that is at greater risk of progression.

Dr. Hunter noted that this work was done with only a small subset of the OAI. In order to get a better understanding of the predictors of progression, it is very important that the dataset be extended through the proposed additional 4 years.

Discussion

It was noted that the amount of denuded area seemed to only have an effect when the change in normalized volume, but not change in volume, was measured.

It was clarified that the changes in thickness in volume were examined within the denuded area.

There was discussion about what the increase in cartilage volume might mean. It may be related to swelling.

Strong Association of Bone Marrow Lesions and Effusion with Pain in Osteoarthritis, Data from the OAI

Grace Lo, M.D., Tufts University

Dr. Lo indicated that she is presenting from the point of view of a rheumatologist who 2 years ago started her current junior faculty position at an academic institution. She has an interest in epidemiologic research in OA. To become an independent investigator, she needs to obtain a mentored award that will support her career development. Consequently, she must propose a project that is adequately powered. However, the majority of mentored funding only provides salary support with little allocation of funds for data acquisition. From this perspective, the OAI is very valuable as it provides a publicly available dataset with three large cohorts of participants, including data from questionnaires, physical exams, functional assessments, x-rays, and 3-Tesla MRIs.

For Dr. Lo, the OAI has provided a setting in which to test her research questions and has provided an opportunity to meet, network, and collaborate with other researchers with interests in OA. The OAI was instrumental in allowing her to successfully obtain and execute a career development award (Arthritis Foundation/American College of Rheumatology) and to participate through abstracts and a presentation at national and international osteoarthritis meetings, thus assisting with her goal of becoming an independent investigator.

Dr. Lo described her research goal: to identify the specific tissue level of pain in OA, the source of which has been elusive. She took advantage of the fact that MRIs provide a wealth of anatomic information and could thereby allow for the identification of features associated with pain. Her hypothesis was that in people with established knee OA, BMLs, synovitis, and effusion are associated with pain independently. She designed a cross-sectional study of participants with symptomatic OA in the progression cohort of the OAI. The MRI features were scored using the BLOKS (Boston Leeds Osteoarthritis Knee Score) scoring method for BML size, synovitis of the infrapatellar fat pad, and effusion size. For BMLs and synovitis, she relied most heavily on IW TSE FS images with some use of DESS sequences for localization of findings. For synovial effusion, she relied most heavily on axial MPR DESS sequences. Knee pain was defined as moderate-to-extreme pain (score 2-4) on any of the three weight-bearing WOMAC pain questions (1, 2, 5). Both univariate and multivariate analyses were carried out. The data were also analyzed using questions 3 and 4 of the WOMAC, which relate to non-weight-bearing pain. The overall trend was still the same. The conclusion of the study was that cross-sectionally, in patients with symptomatic knee OA, maximal BML and effusion scores are highly associated with pain. The strength of the association persists when mutually adjusting for other features. Each feature is independently associated with pain, supporting the idea that pain in OA is multifactorial. If treatments for OA target features that are associated with pain, including BMLs and effusion, and possibly synovitis, those treatments may also modify pain. The ability to find an association between maximal score of the weight-bearing WOMAC pain questions and features as measured by BLOKS provides validation for this pain-scoring method and for BLOKS.

Instrument Development for the OAI

Sharmila Majumdar, Ph.D., University of California San Francisco

The overall goal of Dr. Majumdar�s project is to develop an image registration scheme to register knee MR images in subjects with OA. The first specific aim involves: 1) utilizing fat-suppressed driven equilibrium steady state (DESS) images obtained at 3T, 2) developing a shape-based atlas of the knee; 3) developing a knee registration algorithm that will enable one to register all baseline and followup scans to the knee atlas, so that all scans are analyzed in a common coordinate system with common shape-based anatomical regions; and 4) in close collaboration with musculoskeletal radiologists, defining anatomical compartments, such as the medial and lateral femur, patello-femoral compartment, and medial and lateral tibia, and determining cartilage thickness and volume in those compartments. An additional aspect will be the development of a graphic user-interface and platform for the software developed in the first group of subaims for use by other researchers. The second specific aim is to register the multiecho spin echo OAI images to the knee atlas and generate T2 maps, and to use texture measures such as entropy to obtain measures of T2 spatial heterogeneity in the compartments. The third specific aim is to use the tools developed in the first two specific aims to assess longitudinal changes in cartilage morphology, T2 and to determine whether baseline T2 and its heterogeneity predicts progression, as determined by changes in cartilage morphology, using a subset of images from the progression cohort.

Dr. Majumdar presented examples of how registration based on shape-mapping is achieved, of how cartilage thickness comparisons can be made, and of how cartilage relaxation time can be assessed, and showed examples of OAI data results (DESS, T2 map, 3D thickness map). She noted that this approach could be used to examine changes in tissue heterogeneity and biochemistry as related to progression.

Dr. Majumdar indicated that the OAI is very suitable for this goal, as it is a well-characterized cohort and minimizes costs for image acquisition. 3T imaging, quality control, followup studies, other correlative measures, and different grades of disease are available through the OAI.

A Consortium for Analysis of MR Images from the OAI

Felix Eckstein, M.D., Paracelsus University, Salzburg, Austria and Chondrometrics GbmH, Ainring, Germany

Dr. Eckstein indicated that we need to study structure and imaging in OA because there is not good overlap between symptoms (e.g., pain, function) and structural changes (e.g., in cartilage and other tissues). Because of this lack of overlap, if one wants to ascertain whether a patient is improving structurally as the result of an intervention, there is a need to measure structure independently from the symptoms.

A disease-modifying OA drug, in contrast to a symptom-modifying OA drug, needs to create structural benefit. A biomarker that would be used for such studies needs to demonstrate structural change, which relates to the characteristics of good imaging. However, it is not clear whether a person would take—or an agency approve—a drug that only demonstrated a structural change, but did not have a symptomatic effect. Thus, one needs to demonstrate that structural change is associated with a long-term clinical benefit, and if it is, what the time frame is for that change. One would have to determine which structural endpoint is most closely associated with the clinical endpoint. However, the best structural endpoint/clinical outcome is not known; i.e., is the most reproducible biomarker the best one or is it the biomarker that is most sensitive to change? Clinical outcome data are needed to address this issue. The OAI presents a great opportunity to address these questions, as it will provide long-term data on the cohort and allow for the assessment of the impact of short-term changes on clinical outcomes.

Dr. Eckstein noted that there is a need to have good measures of short-term cartilage changes and a need to determine if they are related to long-term changes and to outcome. He presented examples of data on quantitative cartilage analysis, and noted that strong quality control is the key to obtaining solid information. He demonstrated the range of information that can be assessed, and noted that referring to cartilage morphology rather than volume is a better way to describe this variable. He noted that recently several groups have started to use a subregional approach so that one is not just looking at the total cartilage plate; this is important because cartilage loss is not uniform across the plate. He noted that reports in the literature on cartilage loss per year provide varying levels of loss, which may be due to the cohorts or to the way the data were analyzed. There is a need for more firm data in this area.

Dr. Eckstein noted that when the first OAI imaging data were becoming available, there was not a lot of funding available for quantitative analysis of the data. Therefore, he tried to assemble a Consortium (Consortium for the Analysis of OAI MR Image Data) of industry sponsors who would fund the analysis of subsets of the data. Dr. Eckstein enumerated the general aims of the OAI Consortium proposal. These are to analyze (regional) cartilage morphology changes in the OAI cohort over (relatively) short time periods (1 or 2 years [Y1, Y2]) and to use these data to inform clinical trial design. The Consortium would try to find sponsors who are willing to each contribute funds to analyze 320 data sets (160 pairs [baseline to 1 year], 107 triplets [baseline to 1 year and 2 years]) and share results amongst all contributors. To achieve coordinated analysis of a large data set and to obtain high statistical power, more than 1,000 cases and more than 8 sponsors will be needed. An additional goal is to try to collaborate with other researchers/proposals/networks to maximize use of data.

The specific aims of the Consortium proposal are to determine: 1) the rate of change and sensitivity to change of cartilage thickness in different subregions of the knee, 2) the relationship with clinical parameters at baseline, 3) the relationship with x-ray changes at baseline, 4) the relationship with other MRI structural changes at baseline (semiquantitative scores, read by WORMS, BLOKS), 5) the relationship with long-term change (4 years or longer), and 6) the relationship with clinical outcome (WOMAC or total knee arthroplasty [TKA] years later). The OAI extension would be particularly important for the last two goals.

He noted that the specific value gained and the specific questions that need to be answered include: How do disease-modifying OA drug studies need to be powered? Which imaging protocol should be selected? Which patients should be selected? Which anatomical (sub) region and parameters should be measured? Does short-term change (Year 1, Year 2) track with long-term structural change? Are these imaging biomarkers valid surrogates of clinical outcome? The extension of the OAI is particularly relevant to these issues.

Dr. Eckstein provided an overview of the types of analysis that each of the four sponsors already engaged in the Consortium would provide, as well as an indication of the analysis that the OAI would support and how these analyses related to one another. In terms of data sharing, he noted that each new sponsor joining the Consortium receives immediate access to data delivered within Consortium for internal use. Each sponsor retains all rights to its particular set of funded data (publication, release to others, etc.) Publications that contain data from several sponsors need agreement from all sponsors involved. Sponsors have full control of their own data, but cannot prevent publication of other sponsor�s data. The aim of this approach is to maximize the collaborative use of data, but still keep it attractive for new sponsors to join, with the overall goal of increasing the number of cases analyzed. Dr. Eckstein provided examples of how the data were provided to Consortium participants.

Dr. Eckstein concluded by indicating that the OAI public-use database provides an ideal basis for collaborative efforts to answer questions that can only be addressed in really large cohorts. The current Consortium Proposal is one way of structuring data analysis in large parts of the cohort in order to stimulate expanded analyses, and this effort may be paralleled by other consortium proposals or initiatives for other parameters (clinical, x-ray, sq MRI, serum biomarkers) so as to maximize usage of the data.

Discussion

Following this presentation, there was a question about how researchers outside of the Consortium might get access to the data on cartilage morphology change. Drs. Lester and Nevitt have tried approaches to facilitate the release of some of these data. For example, sponsors who have supported baseline-to-year 1 analyses will be provided with the 24-month analyses in return for the sponsor making the baseline data public. Dr. Lester is also working with Dr. Eckstein to see what other kinds of approaches can be used.

The status of the Virtual Scopics analysis and the relationship of that data to the Consortium were briefly described. Abstracts have been presented on some of these data, and a paper is in preparation.

Dr. Eckstein noted that the intent of the Consortium is to have data made public after the sponsor gets some exclusivity and output from it. Additionally, even before the data become public, individual investigators with ideas can approach the Consortium, and if there is mutual interest, work with them on the investigator�s idea.

State of OA Clinical Research: Value of Longitudinal Cohorts

Stefan Lohmander, M.D., Ph.D., Lund University, Lund, Sweden

Dr. Lohmander addressed the value of longitudinal cohorts in understanding OAI and the range of factors and variables that one needs to consider in trying to develop an understanding of the etiology and progression of OA.

Dr. Lohmander noted that in examining OA cohorts, one cannot readily visualize the participants moving along the symptom-structure change axis. He presented findings from the EUROHIP study, which may provide insight about the etiology of OA. EUROHIP is a longitudinal study involving 12 countries and 1,300 hip replacements. In that study, significantly worse WOMAC scores were found at surgery in older people, women, those with obesity, those with worse general health, and those with lower education. Radiographic severity showed no correlation with WOMAC scores. There were no major differences between countries. While patients reporting mild symptoms may be operated on, surgery is unlikely in the absence of severe radiographic changes.

Obesity, lack of exercise, and weak muscles are associated with high OA risk. The Framingham Study found that weight loss can reduce risk for knee OA; however, the effect of weight loss on symptoms is less well-studied.

Insight into the etiology of knee OA may be derived from studies of young people who have suffered sports knee injuries. Studies in Sweden have found that 10-20 years after ligament or meniscus injury, about 50 percent of the study participants have radiographic evidence of OA. Thus, these are young patients with old knees.

Knee ligament lesions are, in fact, common. MRI examination of unselected populations of persons older than 50, and often without any history of knee injury, determined that the prevalence of ACL rupture was approximately 5 percent and of significant meniscus pathology was 30 to 50 percent.

Thus, it seems that two modifiable risk factors for knee OA are obesity and injury.

Dr. Lohmander then described a Swedish study, LUMEN (Lund meniscus cohort), which has followed persons postsurgery. In this study, there were more than 300 individuals plus controls; they had been followed for more than 25 years. The subjects had isolated unicompartmental meniscus lesions and no radiographic OA at the index arthroscopy. On followup, the prevalence of radiographic OA lesions in the index knee was 50 percent in the 45-54 age group and 70 percent in the 55-64 age-group. In the uninjured knee, there were lesions in 50 percent of the subjects in the 55-64 age-group. Using Odds Ratios, some of the predictors of progression in a person with meniscal tear were: BMI greater than 30, hand OA, being female, or having a degenerative meniscal tear as compared to a traumatic tear. There was little change in pain as measured by the KOOS over 2 years, but there were changes in the KOOS over 8 years.

Dr. Lohmander discussed data from the 1991-96 Malm� Diet and Cancer Cohort that could provide insight into OA. Incidence of severe knee and hip OA was studied in relation to: body mass index, weight, waist, waist/hip ratio (WHR), height, and percent body fat. It was found that BMI, weight and waist were major risk factors for incident severe knee OA (TKR, THR). WHR and percent body fat showed substantially weaker associations with incident severe knee OA. All measures of overweight showed substantially weaker associations with incidence of hip OA than with knee OA. Dr. Lohmander indicated that these findings support the "mechanical" OA hypothesis. The incidence of KOA in the MOA was studied in relationship to the presence of the Metabolic Syndrome. There was increased incidence of severe knee OA in women with Metabolic Syndrome, which was mainly explained by increased BMI.

Dr. Lohmander closed by noting that there is much focus on measuring cartilage. However, muscles, menisci, ligaments and bone, obesity, and injuries are factors that need study. Perhaps there should be a consideration of "street OA" (nonsymptomatic persons with radiological evidence of knee pathology). There also needs to be an assessment of the perceptions, attitudes, concerns, and strategies adopted by people with OA, and the confusion regarding OA and other musculoskeletal diseases or age-related "aches and pains." It is also not clear how pain is a gauge of OA severity. To approach these issues, large multidimensional, carefully designed, population-based, prospective long-term studies such as the OAI are needed. He added that it is not clear what a "true" normal control cohort would be. There is the potential to use studies that focus just on aging as a type of control cohort.

Session IV — View of the Future

Proposal for and Discussion of Extension of Followup of the OAI Cohort

Michael Nevitt, Ph.D., and Charles McCulloch, Ph.D.

Dr. Nevitt noted that the participants had been given several documents describing the goals and details of the proposed extension (e.g., key exams and questionnaire measures). He therefore provided a brief overview of the extension. The primary goals of the extended followup are to follow established knee OA to consequential clinical outcomes (pain that limits function and activity, disability, end-stage disease, and total joint replacement) and to accrue additional endpoints (x-ray knee OA, clinical/symptomatic knee OA, TKR; adequate statistical power for key subgroups; enhanced statistical power for main analyses).

The extended followup will enhance the overall OAI goals by allowing for the investigation of the natural history of knee OA across the spectrum of disease and for following the evolution of early OA to clinically significant disease. At each stage of disease, there will be the opportunity to evaluate imaging, biochemical, and genetic biomarkers with respect to long-term course and outcome. There will be the ability to study the association of changes in biomarkers and risk factors identified during early followup with structural and clinical outcomes captured during extended followup. Additionally, the extension will provide an overall enrichment of the public data resource.

Dr. Nevitt detailed the value of the extended followup of established knee OA across a number of domains: pain and function, disability, end-stage structure/clinical outcomes, and the value of extended followup for incident knee OA. With respect to pain and function, he stressed that in general, knee OA is a slowly developing disease with a highly variable course. Systematic review of longitudinal pain and function outcome studies have not provided clear or consistent evidence relating to the worsening of pain and function. In terms of disability, studies have shown that the transition from "no disability" to "moderate-extreme" disability (as measured with the Late Life Disability Index) is slow, about, 3 to 5 percent /year. Even if the rate increases with the aging of the cohort, the extension of followup will be needed to better study this issue. With respect to end-stage structure/clinical outcome, although current knee imaging techniques are able to detect small amounts of cartilage loss, the rate of cartilage loss detected in the OAI over first 12 months of the study is low. Additionally, the risk of a knee reaching end stage is only 3.2 percent /yr, and the time for 25 percent of participants to reach end-stage is estimated as 8 years. In terms of the value of extended followup for incident knee OA, it is anticipated that the extension would more than double the number of incident radiographic knee OA endpoints and provide enhanced power for subgroups in the main analyses. It would also allow for analysis of the relationships between observations in the first 4 years of followup and incident OA in the second 4 years, including changes in biochemical markers and risk factors and evolution of pathology on MRI.

Dr. Nevitt noted that understanding and recognizing the markers for "early OA" represents an opportunity for development of treatment and secondary prevention strategies before normal joint biomechanics are substantially disrupted and the downward slope toward progression becomes too steep.

(Due to time constraints, Dr. McCulloch did not make his presentation entitled: Rationale for OAI extension: Analytic issues.)

Q & A Session With OAI Advisory Committee to Dr. Nevitt and Speakers,

Maximizing Use/Analysis of OAI Data:

It was commented that a tremendous amount of data will be collected, but that there are not sufficient funds available within the OAI for their analysis. Although set-aside funding is unlikely to be available, NIH plans to issue Program Announcements soliciting applications to use the clinical data, biospecimens, images, and risk factor data. This would provide a mechanism for scientists with hypotheses to test to use the data and datasets to improve our understanding of OA.

Status of Nonimaging Markers:

There was discussion about the status of the nonimaging biomarkers, e.g., serum and urine. Dr. Lester clarified that the specimens are in a repository and that NIAMS has developed an independent process for the expert review of applications to use these specimens. The requests for access and the review of applications will take place three times a year. The process should be finalized in the next 2 months. The specimens requested by successful applicants will be put on hold until the applicant provides documentation that funding has been secured for the analyses. Both the NIAMS and OAI Web sites will have information about the application process.

Status of Imaging Data:

There was discussion about why more MRI analysis data are not available. Part of this is the lack of resources (funds). Importantly, the study protocol requires pairs of data; thus, there is a need for a sufficient number of subjects to undergo assessments beyond the baseline. Efficiency and sensitivity are gained by using paired data. Blinding of the data also requires paired data. Presently, the OAI is working on preparing MRI analysis data for release. Additionally, other groups, such as those led by Dr. Eckstein, are doing analyses of the data on their own.

Dr. Nevitt noted that a lot of investigators are not able to do structural outcomes on their own, and for this reason the OAI is analyzing the baseline-to-24-month data using a variety of analytical approaches. Dr. Eckstein�s group is working with the OAI on this. NIAMS has also tried to leverage efforts.

Dr. Lester indicated that when the OAI was first envisioned, it was hoped that all of the analyses would be able to be done. However, this is very expensive to do, given the size of the study cohort that was needed to achieve the statistical goals. Dr. Nevitt also noted that in 2002, there were very limited methods available for the analysis of the MRI data so the plan to do analyses was deferred until such time when the data would be available.

It was also noted that one of the strengths of the OAI is that it has accrued so many images. There are major costs associated with assembling the cohort and getting the images; not having to request funds or invest time in image acquisition has allowed investigators to seek funds for and invest time only on the analysis of the images. Having the images available will also drive research for new techniques to study them.

It was suggested that imaging data from other organ systems, such as brain, have been released in formats that make them quickly available for analysis. An example of this is the release of atlases that allow groups of investigators to take the data and develop algorithms.

It was suggested that when the OAI makes imaging data available, it would be helpful to include a set of training data. Such training data would help to stimulate the field.

There was discussion about making segmentation data available. Such data should be available when analyses are purchased. Making high-quality segmentations available to investigators would be helpful to test new algorithms and useful for training purposes. Alternatively, segmentation can be generated in a blinded fashion or by multiple sets of individuals. One way to stimulate this type of work might be through a program announcement.

Dr. Eliot Siegel (Chief of Nuclear Medicine, University of Maryland), who represented the National Cancer Imaging Archive (NCIA), indicated that their group had been able to post a number of organ-specific imaging datasets (e.g., colonoscopy, lung nodes) in such a way that investigators can browse the Web site, sign on to it, and mark up the data in various ways. Dr. Siegel indicated a willingness to work with the OAI and allow the OAI to use the NCIA server.

Treatment of Imaging Data From Incident and Progression Studies:

There was a question about whether the MRI data of the incidence and progression cohorts will be treated in the same way in the project extension. Dr. Nevitt indicated that there will be comparable information on subjects in both groups. Early disease will be present only in the incidence cohort. Outside investigators have the ability to "design their own study" by selecting participants from each group according to their own hypothesis.

Full Limb Radiography:

There was discussion about the feasibility of full limb x-rays at the 96-month assessment as a way to assess gait and changes over time. There were concerns about the cost of such x-rays and whether the IRB would approve them because of the level of radiation. It was noted that there are data on alignment from other studies that show very little change in alignment over time.

It was noted that full limb radiography might be a good long-term outcome that could help in validating biomarkers.

Ancillary Studies for the Extension Period:

There was discussion about the timing for ancillary studies; i.e., whether they should coincide with the extension planned clinical assessments at 72 and 96 months or whether they might be better scheduled at 60 and 84 months. The nature of the proposed ancillary study might be the best criterion for that. Ancillary studies in which participant involvement was brief might be easily fitted into the scheduled clinical assessment visits at 72 and 96 months. Ancillary studies that involved more time per subject might be scheduled at the 60- and 84-month time points. Additionally, the timing of the ancillary studies should take into account the scientific need for measurement of related study variables. There also may be cost-efficiency and time-efficiency assessments and subject-burden/retention assessments for the timing of the ancillary studies in relation to the main study measures.

Q & A and Comments from Others in Audience to Dr. Nevitt and Speakers,

Consideration should be given to the future involvement of the proteomic and genomics communities.

In response to a question about when the LLDI questionnaire was to be added during the followup, Dr. Nevitt noted that it is not displacing other studies; it is inexpensive to collect and seemed relevant as the study population aged. Dr. Lester added that an application studying the LLDI and physical activity was going to NIAMS Council and, if approved, would add this measure and others to the OAI.

It was commented on that the ancillary study involving the role of bone in OA was a good addition to the OAI.

It was noted that information on the use of a walking aid is collected as is information related to walking 20 meters and 40 meters. The collection information on positive rehabilitation was suggested.

It was noted that for orthopaedic surgeons, total knee replacement, rather than being the endpoint (as it is for the OAI), is the starting point for failure. Any markers/factors that predicted early failure of total knew replacement would be valuable if these could be identified. Dr. Nevitt noted that subjects with TKR would continue to be followed, but not with imaging. However, the subjects� surgeons would be following them with imaging.

Extension Process: It was noted that the OAI is a contract and not a grant, making the renewal process different. For extension of a contract, the scope of the work is continued rather than changed as it would be in a grant.

Dr. Rebecca Jackson noted that for other large longitudinal studies, it was really in the long-term followup that significant progress was made.

Dr. Leena Sharma noted that both the heterogeneity and duration of OA justify extending the OAI to 8 years.

Dr. Joshua Jacobs noted that the data from the OAI will be extremely valuable to orthopaedic surgeons.

Dr. Rowley Moskowitz, Chairperson of the Observational DSMB, commented that the DSMB had written to Dr. Katz detailing their strong support for the renewal of the OAI and the rationale for that support. The program has been very successful to date, and the gain of information is more than what was anticipated. This type of program and data cannot be readily replicated, and represents a once-in-a-lifetime opportunity. The organizational structure of the OAI itself is a very important achievement and should be sustained.

Comments and Recommendation of the Advisory Committee

The Advisory Committee noted that the accomplishments of the OAI have been superb. The OAI has established an impressive resource that can be readily accessed from a user-friendly Web site. The OAI clearly defines the paradigm of a public database. It is a remarkable resource that cannot be easily reproduced. One of the strongest aspects of the OAI is the MRI data.

The Advisory Committee indicated their complete endorsement of an extension period.

Dr. Nevitt indicated that the extension is primarily to continue to do the ongoing work. The OAI welcomes specific suggestions, but the ability to implement new areas depends on availability of resources.

Wrap up and Summary

Stephen I. Katz, MD, PhD Director, NIAMS

Dr. Katz thanked all of the participants for their efforts. He expressed his appreciation to the Advisory Committee for their constructive comments. He noted that the meeting had achieved its goals of addressing the current status of the OAI and the opportunities for the future. He reiterated that the OAI is a public-private partnership in which all of the partners have made significant contributions and commitments, and which should serve as a resource for both public and commercial endeavors as well as a way for new investigators to enter the field.

Last Reviewed: